- Email: [email protected]
TEM study of the initial oxide scales of Ti2AlC
Available online at www.sciencedirect.com
Acta Materialia 59 (2011) 5216–5223 www.elsevier.com/locate/actamat
TEM study of the initial oxide scales of Ti2AlC J.C. Rao a,b,⇑, Y.T. Pei a, H.J. Yang a, G.M. Song c, S.B. Li d,c, J.Th.M. De Hosson a,⇑ a
Department of Applied Physics, Zernike Institute for Advanced Materials and Materials Innovation Institute M2i, University of Groningen, Nijenborgh 4, 9747 AG Groningen, The Netherlands b The Institute for Advanced Ceramics, Department of Materials Science, Harbin Institute of Technology, Harbin 150001, People’s Republic of China c Department of Materials Science and Engineering, Delft University of Technology, Mekelweg 2, 2628 CD Delft, The Netherlands d Institute of Materials Science and Engineering, School of Mechanical and Electronic Control Engineering, Beijing Jiaotong University, Beijing 100044, People’s Republic of China Received 17 January 2011; received in revised form 21 April 2011; accepted 25 April 2011 Available online 31 May 2011
Abstract This paper presents a detailed microstructural analysis of the crystallographic relationships between a-Al2O3 oxide scale and Ti2AlC parent material, and examines the atomic diffusion and formation of oxide scale on Ti2AlC during the initial oxidation stage at 1200 °C. It is shown that the a-Al2O3 oxide scale can be either a continuous or a discontinuous capping layer on Ti2AlC. A Ti-rich intermediate layer that consists mostly of TiC interrupts the continuity of the a-Al2O3 layer. The channels for inward diffusion of O and outward diffusion of Ti and Al run not only along grain boundaries of a-Al2O3, but also through the Ti-rich intermediate layer. The outward diffusion of Al atoms is either parallel to the (0 0 1) basal plane or parallel to prism planes of Ti2AlC. Crystallographic orientation relationships between a-Al2O3 oxide scale and Ti2AlC were observed: ½ 1 1 2a-Al2 O3 // ½0 1 0Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð0 0 1ÞTi2 AlC and ½0 0 1a-Al2 O3 // ½3 3 1Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð1 1 0ÞTi2 AlC . Ó 2011 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Ceramics; TEM; Ti2AlC; Microstructure; Oxidation
1. Introduction Ti2AlC is a fascinating member of the family of layered ternary compounds, also referred to as Ha¨gg phases in 1930 or later as MAX phases, with the general formula of Mn+1AXn (n = 1, 2 or 3), where M is a transition metal, A is a group IIIA–VIA element and X is C or N [1,2]. Ti2AlC has attracted considerable attention due to its unusual combination of properties, i.e. low density, high modulus and strength, high electrical conductivity, good thermal conductivity, low coefficient of friction, good ther⇑ Corresponding authors. Address: Department of Applied Physics, Zernike Institute for Advanced Materials and Materials Innovation Institute, University of Groningen, Nijenborgh 4, 9747 AG Groningen, The Netherlands (J.C. Rao). E-mail addresses: [email protected] (J.C. Rao), [email protected] (J.Th.M. De Hosson).
mal stability and oxidation resistance [1–6]. These unique properties make the material attractive in structural components for high-temperature applications, oxidation-resistant coatings on alloy surfaces, and as a conducting ceramic in harsh environments. For all these applications, the oxidation resistance of Ti2AlC is a critical property. The oxidation behavior of Ti2AlC has been studied by a number of research groups including Barsoum and Zhou [7–10] as well as in our previous work [11,12]. The results will be discussed in detail below. In general, the results demonstrate that oxidation occurs by the inward diffusion of oxygen and the outward diffusion of Al and Ti atoms. Ti2AlC shows a preferential oxidation mechanism, i.e. first a protective a-Al2O3 layer is formed, and then a discontinuous TiO2 layer follows and covers the top of the previously formed a-Al2O3 layer. The high diffusivity of Al contributes to this selective oxidation phenomenon. The excellent oxidation resistance of Ti2AlC is due to the for-
1359-6454/$36.00 Ó 2011 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actamat.2011.04.058
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223
mation of a protective a-Al2O3 scale on the surface of Ti2AlC. Ganguly et al. [13] studied the interdiffusion process in Ti3SiC2–Ti3GeC2 and Ti2AlC–Nb2AlC diffusion couples and concluded that at comparable temperatures, the absolute values of the diffusivities of the A group elements are about seven times higher than those of the M group elements. The activation energies for the diffusion of the M elements are also greater by a factor of 1.7. The M–X bonds are stronger than the M–A bonds. Ganguly et al. also reported that Ti2AlC would, like other MAX phases, decompose to form TiCx and the diffusion had been halted after a short time, apparently due to the formation of a diffusion barrier of TiC at 1600 °C. Although research on MAX phases has focused on their preparation and structural properties, little microstructural analysis, especially concerning atomic-scale aspects of the oxidation process, is available. Published research on high-temperature oxidation has focused on the macroscale. Questions about the initial stage of the oxidation and possible crystallographic relationships between the oxide phase and the Ti2AlC substrate have not been addressed previously. This paper presents a detailed and systematic transmission electron microscopy (TEM) analysis of the initial stage of the high-temperature oxidation of Ti2AlC at 1200 °C. The atomic diffusion and formation mechanism of oxides will be discussed within the framework of the crystallographic analysis. 2. Experimental Ti2AlC ceramics were fabricated using an in situ solid– liquid reaction of Ti, Al and graphite powders under hotpressing conditions. A detailed description of the preparation can be found elsewhere [14,15]. The starting bulk material was confirmed by X-ray diffraction (XRD) to be Ti2AlC [12]. Several rectangular bars with dimensions of 2 1 8 mm were cut using a slow-speed saw with a diamond blade from the center of the sintered bulk disk in order to obtain homogenous samples. Both surfaces of each bar were abraded to 4000 grit with SiC paper and polished down to 1 lm diamond paste followed by degreasing in acetone and cleaning in ethanol with 10 min ultrasonic vibration. The bars were placed as quickly as possible on a SiC plate in a furnace preheated to 1200 °C. There was no obvious temperature drop as the samples were small, as measured by a thermal couple that had already been positioned next to the sample. The oxide scales were obtained by annealing these bars at 1200 °C for 3 min in flowing air. The samples were then removed from the furnace and transferred to another cold SiC plate for cooling down to ambient temperature in air. The cooling time was also very short since the separated bars were very small in size. This means the oxidation time was as far as possible controlled to be 3 min. The upper sides of the samples were chosen for subsequent analysis of the morphology and phases. The oxidation time is short since we are interested
5217
in studying the initial stage of the oxidation by means of microstructural analysis. The surface morphology of the oxidized specimens was studied by scanning electron microscopy (SEM) using a Philips XL-30 field emission gun microscope, operating at 10 kV and equipped with an energy-dispersive X-ray spectroscopy (EDS) system for chemical composition analysis. In order to investigate the oxidation mechanism, cross-sectional laminae for transmission electron microscopy (TEM) observation were sliced with a focused ion beam (FIB) in a dual-beam FIB/SEM (Lyra, Tescan, Czech Republic) with a Ga ion source operating at 30 kV. Two directions were chosen for preparation of the TEM laminae: one is parallel to the longitude of the lathy grain; the other one is perpendicular to it. A protective tungsten layer (20 1 lm, 2 lm thick) was deposited first on the surface area of interest of the oxide scale for protection before making each lamina. The laminae were finally thinned to 50–100 nm with a very fine ion beam current. A JEM 2010F transmission electron microscope, operating at 200 kV, equipped with an EDS system, was used for selected-area electron diffraction (SAED) analysis and high-resolution TEM observations. The high-resolution images were recorded around Scherzer defocus, i.e. 43 nm for the microscope used here. Fast Fourier transformation (FFT) was carried out using a digital micrograph software package (Gatan, USA). Computer simulations were also performed in order to interpret the high-resolution TEM images. The microscope was described in the calculations with the experimental parameters according to specifications of the microscope. The main parameters used for simulation are: operating voltage 200 kV, spherical aberration (Cs) 0.5 mm, defocus around the Scherzer value of 43 nm. 3. Results and discussion 3.1. Morphology Fig. 1 shows a typical SEM micrograph revealing the surface morphology of Ti2AlC oxidized at 1200 °C for 3 min. The oxide scale is obviously inhomogeneous in thickness and grain size. The scales formed on the surface of Ti2AlC were composed of a-Al2O3 and TiO2 (rutile) [2,8,11,16]. The smooth and dark scale covering the lathy Ti2AlC grains is a-Al2O3. The light gray piled-up oxide, as shown in Fig. 1a, was identified as TiO2 (rutile), which covered the a-Al2O3 oxide on top. This work aimed to reveal the formation mechanism of the initial oxides, and therefore the smooth area was chosen for further detailed TEM analysis. Fig. 1b is a magnification of the central area of Fig. 1a where the TEM lamina was sliced. It is proposed that it was easier for Al to diffuse outward along basal planes than along any other planes of the Ti2AlC structure. Usually the basal planes are parallel to the longitudinal direction of Ti2AlC grains. The rectangular frame in Fig. 1b indicates the position where a
5218
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223
Fig. 2. Bright-field TEM micrograph showing a cross-section of Ti2AlC oxidized at 1200 °C for 3 min. The oxide layers are TiO2 (rutile) and aAl2O3 from the outside to the inside, respectively. A discontinuous Ti-rich layer is observed between the a-Al2O3 layer and the Ti2AlC matrix. The darkest outer layer is tungsten deposited to protect the oxidation surface before FIB milling.
Fig. 1. SEM micrograph of the Ti2AlC oxidized at 1200 °C for 3 min: (a) overview and (b) magnified image of the middle area with the rectangular frame indicating the position where a TEM lamina was sliced out by FIB.
small lamina along the longitudinal direction was picked up with FIB for TEM observations. Fig. 2 is a typical TEM bright-field micrograph of the cross-section of Ti2AlC with oxide scale. The darkest outer layer is tungsten (W) that was deposited to protect the oxidation surface before milling with FIB. Underneath this protective coating, there is a layer about 250 nm thick consisting of polyhedral grains. The selected-area diffraction patterns (SADPs) from different zone axes in TEM confirmed that these are TiO2 (rutile). The next layer, about 350–500 nm thick, is confirmed by SADPs to be a-Al2O3. The shape of the grains is irregular, not like the TiO2 (rutile) layer which clearly consists of polyhedral grains. The two dashed lines indicate the third thin layer, about 100 nm thick, marked as a Ti-rich layer. This layer is just between the a-Al2O3 oxide layer and Ti2AlC parent material. SADPs from different zone axes confirmed that it consists of cubic TiC phase rather than Ti3AlC2 phase. Furthermore, the diffraction pattern can also be indexed as the Ti2C phase in the [0 0 1] direction. TiC and Ti2C are both cubic phases, while the space groups are Fm 3 m (2 2 5) and Fd 3m (2 2 7) for TiC and Ti2C, respectively. Ti2C has the superstructure of TiC but a lattice parameter
twice as large [17]. This phase is formed by an ordered arrangement of carbon atoms in the octahedral voids of the metal lattice. Ti2C phase has rarely been reported, so here the phase was indexed as TiC. The expression “Ti-rich layer” is in any case used for this layer hereafter. This layer is not continuous but interrupted by a-Al2O3 oxide grains. In some other areas, the a-Al2O3 oxide layer was formed directly on the outer surface of the Ti2AlC substrate without this intermediate layer. In general, it is concluded that the a-Al2O3 oxide layer on the surface of Ti2AlC is either continuous or discontinuous. It should be remembered that Fig. 2 only shows a discontinuous a-Al2O3 oxide layer interrupted by a very thin Ti-rich layer. There are, however, several ways for Al to diffuse out from Ti2AlC substrate. First, as mentioned above, the Ti-rich layer is discontinuous. Secondly, the Ti-rich layer is not very big, e.g. 2 lm long in Fig. 2. Thirdly, the Al atom can diffuse across this layer (if needed) along grain boundaries, in a similar fashion to the diffusion of Ti out along the grain boundaries of a-Al2O3. Wang and Zhou [8] reported a continuous a-Al2O3 oxide layer on Ti2AlC with good adhesion to the substrate. Their oxidation experiment was performed over the temperature range 1000–1300 °C for 20 h. It seemed that there was enough time at these temperatures both for Al atoms to form a continuous inner a-Al2O3 oxide scale and for Ti atoms to further diffuse out, forming TiO2 (rutile) oxide scale on the outside. In our work, the oxidation time was only 3 min. This is also the reason why a discontinuous a-Al2O3 oxide layer was observed to form on the Ti2AlC matrix. Fig. 3 shows typical SADPs from Ti2AlC matrix, TiC, oxide layers of a-Al2O3 and TiO2 (rutile). These phases were confirmed by the SADPs from different, but adjacent, areas of each layer. The outward growth of large isolated
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223
Fig. 3. Electron diffraction patterns from Ti2AlC substrate, TiC, oxide layers of a-Al2O3 and TiO2 (rutile), respectively.
TiO2 (rutile) grains suggests that the growth of TiO2 (rutile) is governed by the outward diffusion of Ti. 3.2. Crystallographic orientation relationships As mentioned above, little research has been reported on the microstructural analysis of Ti2AlC after oxidation; in particular, the crystallographic relationships between the newly grown oxide phases and the Ti2AlC matrix are still unknown. The present work focused on possible crystallographic orientation relationships both between the a-Al2O3 layer and the substrate and between these two outer oxide layers of TiO2 (rutile) and a-Al2O3. Several cross-sectional laminae were cut out from different Ti2AlC grains after oxidation along two directions either parallel or perpendicular to the longitudinal directions of the lathy grains. No regular crystallographic relationship was found between these two outer oxide layers of TiO2 (rutile) and a-Al2O3. In contrast, several crystallographic orientations were recorded between the a-Al2O3 oxide and Ti2AlC matrix by electron diffraction. Fig. 4 shows the predominant orientation relationships: ½ 1 1 2a-Al2 O3 // ½0 1 0Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð0 0 1ÞTi2 AlC as well as schematic drawings of the unit cells of a-Al2O3 oxide and Ti2AlC matrix. It is very clear that the basal plane (0 0 1) of Ti2AlC is parallel to the secondary prism plane (1 1 0) of a-Al2O3 oxide. The lattice mismatch between these a-Al2O3 and Ti2AlC orientated phases was also calculated: there is only 3% mismatch between every two unit cells of Ti2AlC along [0 0 1] and three unit cells of aAl2O3 along [1 1 0], implying that the observed orientation relationship is reasonably stable. Ti2AlC has a layered structure composed of three different types of bonds: the relatively weak Ti 3d–Al 3p bond
5219
Fig. 4. (a) SADP showing the crystallographic orientation relationship between the a-Al2O3 oxide and Ti2AlC matrix; (b and c) schematic drawings of their unit cells according to the orientation relationship: ½1 1 2a-Al2 O3 // ½0 1 0Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð0 0 1ÞTi2 AlC .
1 eV below the Fermi level and the stronger Ti 3d–C 2p and Ti 3d–C 2s bonds with energies of around 2.5 and 10 eV below the Fermi level, respectively [18]. Insertion of Al monolayers into face-centered cubic (fcc) TiC matrix in such a way that every second monolayer of C atoms is replaced by an Al atom layer leads to the formation of hexagonal close-packed (hcp) Ti2AlC. As a result, the strong Ti–C bonds are broken up and replaced by weaker Ti–Al bonds at the expense of internal energy. The TiC layers surrounding the Al monolayers are then twinned with the Al layer as the mirror plane. As shown in Fig. 4b, the thermodynamically stable Ti–C–Ti slabs are separated by softer Ti–Al–Ti slabs with weak metallic bonds. Thus the Al atoms may easily decompose at high temperatures and diffuse outward along the direction parallel to the basal plane according to the weaker bonding between Ti and Al. The Al atoms are located in the planes parallel to the basal plane in Ti2AlC. Fig. 4b shows the hexagonal structure of Ti2AlC, and the unit cell of a-Al2O3 is shown in Fig. 4c. As mentioned above, it is easier for Al atoms to diffuse outward parallel to the basal planes of Ti2AlC. When this Al monolayer is diffused out from Ti2AlC, an aluminum oxide layer forms at high temperatures. Although several aluminum oxide phases, e.g. c-Al2O3, d-Al2O3, hAl2O3 as well as TiO2 (anatase), have been reported by other researchers [9,19], in this work only the a-Al2O3 phase has been confirmed by SADPs. The most compacted Ti, Al or C monolayers are perpendicular to the c axis of the unit cell, though Al atoms also lie in other planes, e.g. parallel to the ð1 1 0Þ plane of the Ti2AlC unit cell. Therefore there are various ways for Al
5220
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223
Fig. 6. High-resolution TEM micrograph of the interface between aAl2O3 oxide and Ti2AlC matrix. The two inserts are the simulated images under the same conditions used for TEM. Some crystalline planes of both Ti2AlC matrix phase and a-Al2O3 oxide are indicated. The crystallographic orientation can be deduced from the figure to be: ½1 1 1a-Al2 O3 // ½4 5 1Ti2 AlC , ð1 3 4Þa-Al2 O3 // ð1 1 1ÞTi2 AlC .
Fig. 5. (a) SADP showing another crystallographic orientation relationship between the a-Al2O3 oxide and Ti2AlC matrix; (b and c) schematic drawings of their unit cells according to the orientation relationship: ½0 0 1a-Al2 O3 // ½3 3 1Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð1 1 0ÞTi2 AlC .
atoms to diffuse out of Ti2AlC structure. Fig. 5 shows another observed crystallographic orientation: ½0 0 1a-Al2 O3 // ½3 3 1Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð1 1 0ÞTi2 AlC . This means that besides diffusion parallel to the (0 0 1) basal plane, the Al atoms can also diffuse out in the atom monolayers parallel to the ð1 1 0Þ plane, and perpendicular to the basal (0 0 1) plane. From these crystallographic considerations, we can reason that there may be other orientations between the oxide grains and the Ti2AlC parent grains. Fig. 6 shows a highresolution TEM micrograph of the interface between aAl2O3 oxide and Ti2AlC. Fast Fourier transformation was applied to each high-resolution image of both phases. The results show the following crystallographic orientation:
½1 1 1a-Al2O3 // ½4 5 1Ti2AlC , ð1 3 bar4Þa-Al2O3 // ð1 1 1ÞTi2 AlC . The two inserts are simulated images according to the above direction of each phase, respectively, under the same condition (voltage = 200 kV, spherical aberration (Cs) = 0.5 mm, sample thickness = 37 nm, defocus = 46 nm, which is near to the Scherzer focus of 43 nm). Some crystalline planes of Ti2AlC matrix and a-Al2O3 oxide phases are indicated directly on the picture. It is concluded that besides diffusion of Al parallel to the basal plane of Ti2AlC, other mechanism may operate as observed for the first time in this work. It should be noted that the crystallographic orientation relationships observed in this work are different from that reported by Magnuson et al. [18]. In the latter study, TiCx and Ti2AlC films were epitaxially grown on single-crystal Al2O3. In the present study, the a-Al2O3, TiC and TiO2 layers were oxidation products of Ti2AlC and formed via a different growth mechanism. Another point which should also be emphasized concerns the relationship between the basal plane direction and oxide scale formation. During our experiments, some TEM laminae were prepared parallel to the longitudinal directions of the Ti2AlC lathy grains. However, this longitudinal direction is perpendicular to the c axis of the Ti2AlC unit cell. In other words, it is parallel to the basal planes of Ti2AlC. However, the prepared laminae are not, in general, parallel to these basal planes. It seems that there are more columnar oxide grains on the Ti2AlC surface after prolonged oxidation if the surface is perpendicular to the basal planes since it is easier for Al and Ti outward diffusion and O inward diffusion along these par-
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223
allel basal planes. There are definitely fewer columnar oxide grains if the surface is parallel to the basal planes since it is more difficult for the atoms to diffuse perpendicular to the basal planes. 3.3. Formation mechanism of oxide scales Oxidation of alloys or composites consists in general of two processes: the dissolution of oxygen, and the formation of an oxide scale. During the initial stage of oxidation, molecular oxygen from the gas phase is adsorbed on the substrate surface and dissociates. A thin layer of oxide film is formed by nucleation and lateral growth of oxide crystallites. In general, the oxidation process includes the thermodynamically controlled initial surface reaction and the subsequent growth of an oxide film. The growth is determined predominantly by the rate of diffusion of oxygen ions through the oxide scale and substrate surface. To explain why a-Al2O3 is present in the inner part and TiO2 in the outer part of the scale, Wang et al. assumed that a-Al2O3 and TiO2 (rutile) nucleate simultaneously on the surface of Ti2AlC during the initial oxidation stage due to the oxygen partial pressure being sufficiently high at the ambient air–Ti2AlC interface [8]. Upon the intake of oxygen, the oxygen partial pressure at the oxide scale– Ti2AlC interface decreases with increasing oxidation time. Once the oxygen pressure is lower than the TiO2–Ti2AlC equilibrium pressure, TiO2 ceases further growth because a-Al2O3 is more thermodynamically stable than TiO2. aAl2O3 can grow continuously due to the fact that the formation of a-Al2O3 needs a much lower oxygen pressure [8]. However, this is unlikely to be the physical picture of the oxidation process of Ti2AlC. The electronic structure investigation by Zhou and Sun [20] has shown that the bonding between Ti and C is strongly covalent and directional, whereas the bonding between Ti and Al is weak in the layered ternary compound of Ti2AlC. The strong interaction of Ti and C would decrease the activity of Ti and result in an activity of Al high enough for it to be preferentially oxidized. Based on the TEM observations, therefore, only a-Al2O3 nuclei nucleate on the Ti2AlC matrix and grow in the initial stage of oxidation, forming the inner oxide scale. Once the depletion of Al atoms occurs in the surface layer of Ti2AlC matrix, Ti atoms start to diffuse out and form the outer TiO2 layer. Due to the weak bonding between Ti and Al in layered MAX phases, the cleavage of laminar grains were reported mainly parallel to the hexagonal basal plane (0 0 1) in a MAX-like system (Ti3AlC2). Al will first diffuse out from Ti2AlC matrix and form a-Al2O3 with oxygen. TiC is decomposed from the matrix after Al diffuses out, and hence Ti will diffuse out thereafter through the grain boundaries of preformed a-Al2O3 and react with oxygen to form TiO2 (rutile). This means that a-Al2O3 starts from pure Al while TiO2 starts from TiC. In order to know which oxide will form readily, it is important to compare the free energies for the formation of a-Al2O3 and TiO2
5221
(rutile). The Gibbs free energy for Al2O3 formation from the reaction of Al atom with O at 1473 K is 408 kJ mol–1 and for TiO2 formation from TiC it is 295 kJ mol–1 [11,21]. It was concluded in our previous work [11] that the cleavage surface with Al terminal layer is more easily oxidized than that with a Ti3C2 layer on top in Ti3AlC2 system. Therefore, more Al2O3 nuclei than TiO2 nuclei may be expected to form. In Ref. [11] the formation mechanisms of inner a-Al2O3 and outer TiO2 oxide scales in the Ti3AlC2 system were also analyzed from the viewpoint of diffusion. a-Al2O3 is a fairly stable protective oxide, the growth rate of which is about five orders of magnitude lower than that of TiO2 during the oxidation of Ti– Al alloy [22]. As a result, TiO2 outgrows the inner Al2O3 to form a TiO2 outer layer. 3.4. Atom diffusion and formation of a Ti-rich intermediate layer Wagner studied internal oxidation and selective scale formation as long ago as 1959 [23]. According to his theory, a high content as well as high diffusivity of Al in the substrate, and the low solubility and diffusivity of oxygen in the substrate, favors the preferential oxidation of Al. In the case of Ti2AlC, it was reported that the diffusivity of Al is high and the solubility of oxygen in Ti2AlC is low [8]. Both the low solubility of oxygen and high diffusivity of Al in the substrate are favorable factors for the preferential oxidation of Al. Moreover, the densities of Ti2AlC and a-Al2O3 are 4.21 and 3.98 g cm–3, respectively. Therefore, there is little volume change and a small difference in the thermal expansion coefficient between the a-Al2O3 oxide scale and Ti2AlC substrate. In addition, the oxide scale exhibits good adhesion to the substrate. The formation of a continuous a-Al2O3 layer on Ti2AlC and the good adhesion to the substrate are the main contributions to the excellent oxidation resistance of Ti2AlC. This is also crucial for the recovery of the strength of self-healed materials based on the formation of oxides [16]. Nevertheless, it is a drawback in self-healing process since further oxidation (healing) will be slowed down by the continuous scale. Because the diffusivity of Ti in a-Al2O3 is very low [24] and no Ti concentration gradient was detected across the oxide scale–Ti2AlC interface for samples oxidized at high temperature, Wang et al. concluded that the outward diffusion of Ti via short-circuit diffusion along grain boundaries of the as-formed a-Al2O3 is reasonably fast [8]. In the present study, a discontinuous a-Al2O3 oxide inner layer was observed with high-resolution TEM, which was interrupted by Ti-rich intermediate layer (mostly TiC). This implies that there should be another route for Ti diffusion apart from the short-circuit diffusion through grain boundaries of the as-formed a-Al2O3. The oxide scale of Ti–Al alloys consists of TiO2 and Al2O3 instead of only Al2O3; this was proven both experimentally by Meier et al. [25] and theoretically supported by Rahmel et al. with thermodynamic calculations [26]. According to their results for the
5222
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223
tion mechanism in Ti2AlC during the initial oxidation stage at 1200 °C for 3 min, the following conclusions can be drawn.
Fig. 7. High-resolution TEM micrograph of the interface between Ti2AlC matrix and the Ti-rich intermediate layer containing some randomly orientated a-Al2O3 clusters. Some atomic voids or channels are visible in both the Ti2AlC matrix and the Ti-rich layer for Al and Ti outward diffusion and also for O inward diffusion during the oxidation process.
1. The a-Al2O3 oxide scale can be either a continuous or discontinuous layer on the surface of Ti2AlC matrix. The TiO2 outer scale is formed due to the outward diffusion of Ti after depletion of Al atoms. 2. An intermediate layer is observed between the a-Al2O3 inner scale and Ti2AlC matrix, which is Ti rich and consists mostly of TiC. It transforms from Ti2AlC matrix due to depletion of Al atoms. 3. There are some regular crystallographic orientations between a-Al2O3 oxide scale and Ti2AlC matrix: ½1 1 2a-Al2 O3 // ½0 1 0Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð0 0 1ÞTi2 AlC and ½0 0 1a-Al2 O3 // ½3 3 1Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð1 1 0ÞTi2 AlC . 4. The outward diffusion of Al atoms occurs in certain atomic planes of the Ti2AlC structure. These planes are either (0 0 1) basal planes or prism planes of Ti2AlC. Both will form the secondary prism planes {1 1 0} of Al in hexagonal a-Al2O3 during oxidation.
Acknowledgements Ti–Al–O system, the stable oxide changed from TiO to Al2O3 in coexistence with the TiAl phase because the Al– Al2O3 equilibrium pressure is lower than that for Ti–TiO. The Al content is about 50% in Ti–Al alloys. It is obvious that the Al content in Ti2AlC is much less than this critical amount and the oxidation of Ti2AlC should not yield a continuous Al2O3 layer if the influence of carbon on the activity of Ti and Al was not taken into account [8]. Fig. 7 is a high-resolution TEM micrograph of the interface between the intermediate layer (Ti-rich layer) and Ti2AlC. Small a-Al2O3 clusters are randomly distributed in this layer. Some atomic voids or channels are clearly visible in both the intermediate layer and Ti2AlC. The Ti-rich intermediate layer forms as a consequence of atom diffusion due to the depletion of Al atoms in Ti2AlC. It is therefore a barrier to the outward diffusion of Al atoms but a source of Ti atoms that successively diffuse outwards. Some crystalline planes of both Ti2AlC matrix and a-Al2O3 oxide phases are indicated in the micrograph. In conclusion, we may state that the diffusion channels for Ti are not only the grain boundaries of as-formed a-Al2O3, but also the intermediate Ti-rich layer with randomly distributed aAl2O3 crystalline clusters. Since the intermediate layer subsequently transforms from Ti2AlC matrix, its formation does not destroy the orientation relationships observed between the a-Al2O3 scale and Ti2AlC matrix. 4. Conclusions From the detailed microstructure analysis of the crystallographic relationships between the a-Al2O3 oxide scale and Ti2AlC, and the atom diffusion and oxide scale forma-
J.C.R. and S.B.L. are on leave from China. All the research work was done in Groningen and Delft, the Netherlands. This work was supported by the Netherlands Innovation Oriented Program on Self Healing Materials under Grant No. IOP-SHM 0871. The authors would thank Professor Michel Barsoum (Drexel University, Philadelphia, USA) for providing the Ti2AlC bulk sample. References [1] Barsoum MW, El-Raghy T. J Am Ceram Soc 1996;79:1953. [2] Barsoum MW. Prog Solid State Chem 2000;28:201. [3] Gupta S, Amini S, Filimonov D, Palanisamy T, El-Raghy T, Barsoum MW. J Am Ceram Soc 2007;90:3566. [4] Farber L, Barsoum MW, Zavaliangos A, El-Raghy T, Levin I. J Am Ceram Soc 1998;81:1677. [5] Amini S, Barsoum MW, El-Raghyy T. J Am Ceram Soc 2007;90:3953. [6] El-Raghy T, Zavaliangos A, Barsoum MW, Kalidindi SR. J Am Ceram Soc 1997;80:513. [7] Barsoum MW. J Electrochem Soc 2001;148:C544. [8] Wang XH, Zhou YC. Oxid Met 2003;59:303. [9] Wang XH, Zhou YC. J Mater Res 2002;17:2974. [10] Barsoum MW, Tzenov N, Procopio A, El-Raghy T, Ali M. J Electrochem Soc 2001;148:C551. [11] Song GM, Pei YT, Sloof WG, De Hosson JTM, Van der Zwaag S. Mater Chem Phys 2008;112:762. [12] Song GM, Li SB, Zhao CX, Sloof WG, vd Zwaag S, Pei YT, et al. J Eur Ceram Soc 2011;31:855. [13] Ganguly A, Barsoum MW, Doherty RD. J Am Ceram Soc 2007;90:22002. [14] Wang XH, Zhou YC. Acta Mater 2002;50:3141. [15] Barsoum MW, Brodkin D, El-Raghy T. Scripta Mater 1997;36:535. [16] Song GM, Pei YT, Sloof WG, Li SB, De Hosson JTM, vd Zwaag S. Scripta Mater 2008;58:13. [17] Goretzki H. Phys Status Solidi (b) 1967;20:k141.
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223 [18] Magnuson M, Wilhelmsson O, Palmquist J-P, Jansson U, Mattesini M, Li S, et al. Phys Rev B 2006;74:195108. [19] Babu N, Balasubramaniam R, Ghosh A. Corros Sci 2001;43: 2239. [20] Zhou YC, Sun ZM. Phys Rev B 2000;61:12570. [21] Lide DR. Handbook of chemistry and physics. 76th ed. New York: CRC Press; 2000. p. 5–72. [22] Yang MR, Wu SK. Bull College Eng NTU 2003;89:3.
5223
[23] Wagner C. Z Elektrochem 1959;63:772. [24] Langensiepen RA, Tressler RE, Howell PR. J Mater Sci 1983;18:2771. [25] Meier GH, Appalonia D, Perkins RA, Chiang KT. In: Grobstein T, Doychak J, editors. Oxidation of high-temperature intermetallics. Cleveland (OH): TMS; 1988. p. 379. [26] Rahmel A, Spencer PJ. Oxid Met 1991;35:53.
Acta Materialia 59 (2011) 5216–5223 www.elsevier.com/locate/actamat
TEM study of the initial oxide scales of Ti2AlC J.C. Rao a,b,⇑, Y.T. Pei a, H.J. Yang a, G.M. Song c, S.B. Li d,c, J.Th.M. De Hosson a,⇑ a
Department of Applied Physics, Zernike Institute for Advanced Materials and Materials Innovation Institute M2i, University of Groningen, Nijenborgh 4, 9747 AG Groningen, The Netherlands b The Institute for Advanced Ceramics, Department of Materials Science, Harbin Institute of Technology, Harbin 150001, People’s Republic of China c Department of Materials Science and Engineering, Delft University of Technology, Mekelweg 2, 2628 CD Delft, The Netherlands d Institute of Materials Science and Engineering, School of Mechanical and Electronic Control Engineering, Beijing Jiaotong University, Beijing 100044, People’s Republic of China Received 17 January 2011; received in revised form 21 April 2011; accepted 25 April 2011 Available online 31 May 2011
Abstract This paper presents a detailed microstructural analysis of the crystallographic relationships between a-Al2O3 oxide scale and Ti2AlC parent material, and examines the atomic diffusion and formation of oxide scale on Ti2AlC during the initial oxidation stage at 1200 °C. It is shown that the a-Al2O3 oxide scale can be either a continuous or a discontinuous capping layer on Ti2AlC. A Ti-rich intermediate layer that consists mostly of TiC interrupts the continuity of the a-Al2O3 layer. The channels for inward diffusion of O and outward diffusion of Ti and Al run not only along grain boundaries of a-Al2O3, but also through the Ti-rich intermediate layer. The outward diffusion of Al atoms is either parallel to the (0 0 1) basal plane or parallel to prism planes of Ti2AlC. Crystallographic orientation relationships between a-Al2O3 oxide scale and Ti2AlC were observed: ½ 1 1 2a-Al2 O3 // ½0 1 0Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð0 0 1ÞTi2 AlC and ½0 0 1a-Al2 O3 // ½3 3 1Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð1 1 0ÞTi2 AlC . Ó 2011 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Ceramics; TEM; Ti2AlC; Microstructure; Oxidation
1. Introduction Ti2AlC is a fascinating member of the family of layered ternary compounds, also referred to as Ha¨gg phases in 1930 or later as MAX phases, with the general formula of Mn+1AXn (n = 1, 2 or 3), where M is a transition metal, A is a group IIIA–VIA element and X is C or N [1,2]. Ti2AlC has attracted considerable attention due to its unusual combination of properties, i.e. low density, high modulus and strength, high electrical conductivity, good thermal conductivity, low coefficient of friction, good ther⇑ Corresponding authors. Address: Department of Applied Physics, Zernike Institute for Advanced Materials and Materials Innovation Institute, University of Groningen, Nijenborgh 4, 9747 AG Groningen, The Netherlands (J.C. Rao). E-mail addresses: [email protected] (J.C. Rao), [email protected] (J.Th.M. De Hosson).
mal stability and oxidation resistance [1–6]. These unique properties make the material attractive in structural components for high-temperature applications, oxidation-resistant coatings on alloy surfaces, and as a conducting ceramic in harsh environments. For all these applications, the oxidation resistance of Ti2AlC is a critical property. The oxidation behavior of Ti2AlC has been studied by a number of research groups including Barsoum and Zhou [7–10] as well as in our previous work [11,12]. The results will be discussed in detail below. In general, the results demonstrate that oxidation occurs by the inward diffusion of oxygen and the outward diffusion of Al and Ti atoms. Ti2AlC shows a preferential oxidation mechanism, i.e. first a protective a-Al2O3 layer is formed, and then a discontinuous TiO2 layer follows and covers the top of the previously formed a-Al2O3 layer. The high diffusivity of Al contributes to this selective oxidation phenomenon. The excellent oxidation resistance of Ti2AlC is due to the for-
1359-6454/$36.00 Ó 2011 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actamat.2011.04.058
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223
mation of a protective a-Al2O3 scale on the surface of Ti2AlC. Ganguly et al. [13] studied the interdiffusion process in Ti3SiC2–Ti3GeC2 and Ti2AlC–Nb2AlC diffusion couples and concluded that at comparable temperatures, the absolute values of the diffusivities of the A group elements are about seven times higher than those of the M group elements. The activation energies for the diffusion of the M elements are also greater by a factor of 1.7. The M–X bonds are stronger than the M–A bonds. Ganguly et al. also reported that Ti2AlC would, like other MAX phases, decompose to form TiCx and the diffusion had been halted after a short time, apparently due to the formation of a diffusion barrier of TiC at 1600 °C. Although research on MAX phases has focused on their preparation and structural properties, little microstructural analysis, especially concerning atomic-scale aspects of the oxidation process, is available. Published research on high-temperature oxidation has focused on the macroscale. Questions about the initial stage of the oxidation and possible crystallographic relationships between the oxide phase and the Ti2AlC substrate have not been addressed previously. This paper presents a detailed and systematic transmission electron microscopy (TEM) analysis of the initial stage of the high-temperature oxidation of Ti2AlC at 1200 °C. The atomic diffusion and formation mechanism of oxides will be discussed within the framework of the crystallographic analysis. 2. Experimental Ti2AlC ceramics were fabricated using an in situ solid– liquid reaction of Ti, Al and graphite powders under hotpressing conditions. A detailed description of the preparation can be found elsewhere [14,15]. The starting bulk material was confirmed by X-ray diffraction (XRD) to be Ti2AlC [12]. Several rectangular bars with dimensions of 2 1 8 mm were cut using a slow-speed saw with a diamond blade from the center of the sintered bulk disk in order to obtain homogenous samples. Both surfaces of each bar were abraded to 4000 grit with SiC paper and polished down to 1 lm diamond paste followed by degreasing in acetone and cleaning in ethanol with 10 min ultrasonic vibration. The bars were placed as quickly as possible on a SiC plate in a furnace preheated to 1200 °C. There was no obvious temperature drop as the samples were small, as measured by a thermal couple that had already been positioned next to the sample. The oxide scales were obtained by annealing these bars at 1200 °C for 3 min in flowing air. The samples were then removed from the furnace and transferred to another cold SiC plate for cooling down to ambient temperature in air. The cooling time was also very short since the separated bars were very small in size. This means the oxidation time was as far as possible controlled to be 3 min. The upper sides of the samples were chosen for subsequent analysis of the morphology and phases. The oxidation time is short since we are interested
5217
in studying the initial stage of the oxidation by means of microstructural analysis. The surface morphology of the oxidized specimens was studied by scanning electron microscopy (SEM) using a Philips XL-30 field emission gun microscope, operating at 10 kV and equipped with an energy-dispersive X-ray spectroscopy (EDS) system for chemical composition analysis. In order to investigate the oxidation mechanism, cross-sectional laminae for transmission electron microscopy (TEM) observation were sliced with a focused ion beam (FIB) in a dual-beam FIB/SEM (Lyra, Tescan, Czech Republic) with a Ga ion source operating at 30 kV. Two directions were chosen for preparation of the TEM laminae: one is parallel to the longitude of the lathy grain; the other one is perpendicular to it. A protective tungsten layer (20 1 lm, 2 lm thick) was deposited first on the surface area of interest of the oxide scale for protection before making each lamina. The laminae were finally thinned to 50–100 nm with a very fine ion beam current. A JEM 2010F transmission electron microscope, operating at 200 kV, equipped with an EDS system, was used for selected-area electron diffraction (SAED) analysis and high-resolution TEM observations. The high-resolution images were recorded around Scherzer defocus, i.e. 43 nm for the microscope used here. Fast Fourier transformation (FFT) was carried out using a digital micrograph software package (Gatan, USA). Computer simulations were also performed in order to interpret the high-resolution TEM images. The microscope was described in the calculations with the experimental parameters according to specifications of the microscope. The main parameters used for simulation are: operating voltage 200 kV, spherical aberration (Cs) 0.5 mm, defocus around the Scherzer value of 43 nm. 3. Results and discussion 3.1. Morphology Fig. 1 shows a typical SEM micrograph revealing the surface morphology of Ti2AlC oxidized at 1200 °C for 3 min. The oxide scale is obviously inhomogeneous in thickness and grain size. The scales formed on the surface of Ti2AlC were composed of a-Al2O3 and TiO2 (rutile) [2,8,11,16]. The smooth and dark scale covering the lathy Ti2AlC grains is a-Al2O3. The light gray piled-up oxide, as shown in Fig. 1a, was identified as TiO2 (rutile), which covered the a-Al2O3 oxide on top. This work aimed to reveal the formation mechanism of the initial oxides, and therefore the smooth area was chosen for further detailed TEM analysis. Fig. 1b is a magnification of the central area of Fig. 1a where the TEM lamina was sliced. It is proposed that it was easier for Al to diffuse outward along basal planes than along any other planes of the Ti2AlC structure. Usually the basal planes are parallel to the longitudinal direction of Ti2AlC grains. The rectangular frame in Fig. 1b indicates the position where a
5218
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223
Fig. 2. Bright-field TEM micrograph showing a cross-section of Ti2AlC oxidized at 1200 °C for 3 min. The oxide layers are TiO2 (rutile) and aAl2O3 from the outside to the inside, respectively. A discontinuous Ti-rich layer is observed between the a-Al2O3 layer and the Ti2AlC matrix. The darkest outer layer is tungsten deposited to protect the oxidation surface before FIB milling.
Fig. 1. SEM micrograph of the Ti2AlC oxidized at 1200 °C for 3 min: (a) overview and (b) magnified image of the middle area with the rectangular frame indicating the position where a TEM lamina was sliced out by FIB.
small lamina along the longitudinal direction was picked up with FIB for TEM observations. Fig. 2 is a typical TEM bright-field micrograph of the cross-section of Ti2AlC with oxide scale. The darkest outer layer is tungsten (W) that was deposited to protect the oxidation surface before milling with FIB. Underneath this protective coating, there is a layer about 250 nm thick consisting of polyhedral grains. The selected-area diffraction patterns (SADPs) from different zone axes in TEM confirmed that these are TiO2 (rutile). The next layer, about 350–500 nm thick, is confirmed by SADPs to be a-Al2O3. The shape of the grains is irregular, not like the TiO2 (rutile) layer which clearly consists of polyhedral grains. The two dashed lines indicate the third thin layer, about 100 nm thick, marked as a Ti-rich layer. This layer is just between the a-Al2O3 oxide layer and Ti2AlC parent material. SADPs from different zone axes confirmed that it consists of cubic TiC phase rather than Ti3AlC2 phase. Furthermore, the diffraction pattern can also be indexed as the Ti2C phase in the [0 0 1] direction. TiC and Ti2C are both cubic phases, while the space groups are Fm 3 m (2 2 5) and Fd 3m (2 2 7) for TiC and Ti2C, respectively. Ti2C has the superstructure of TiC but a lattice parameter
twice as large [17]. This phase is formed by an ordered arrangement of carbon atoms in the octahedral voids of the metal lattice. Ti2C phase has rarely been reported, so here the phase was indexed as TiC. The expression “Ti-rich layer” is in any case used for this layer hereafter. This layer is not continuous but interrupted by a-Al2O3 oxide grains. In some other areas, the a-Al2O3 oxide layer was formed directly on the outer surface of the Ti2AlC substrate without this intermediate layer. In general, it is concluded that the a-Al2O3 oxide layer on the surface of Ti2AlC is either continuous or discontinuous. It should be remembered that Fig. 2 only shows a discontinuous a-Al2O3 oxide layer interrupted by a very thin Ti-rich layer. There are, however, several ways for Al to diffuse out from Ti2AlC substrate. First, as mentioned above, the Ti-rich layer is discontinuous. Secondly, the Ti-rich layer is not very big, e.g. 2 lm long in Fig. 2. Thirdly, the Al atom can diffuse across this layer (if needed) along grain boundaries, in a similar fashion to the diffusion of Ti out along the grain boundaries of a-Al2O3. Wang and Zhou [8] reported a continuous a-Al2O3 oxide layer on Ti2AlC with good adhesion to the substrate. Their oxidation experiment was performed over the temperature range 1000–1300 °C for 20 h. It seemed that there was enough time at these temperatures both for Al atoms to form a continuous inner a-Al2O3 oxide scale and for Ti atoms to further diffuse out, forming TiO2 (rutile) oxide scale on the outside. In our work, the oxidation time was only 3 min. This is also the reason why a discontinuous a-Al2O3 oxide layer was observed to form on the Ti2AlC matrix. Fig. 3 shows typical SADPs from Ti2AlC matrix, TiC, oxide layers of a-Al2O3 and TiO2 (rutile). These phases were confirmed by the SADPs from different, but adjacent, areas of each layer. The outward growth of large isolated
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223
Fig. 3. Electron diffraction patterns from Ti2AlC substrate, TiC, oxide layers of a-Al2O3 and TiO2 (rutile), respectively.
TiO2 (rutile) grains suggests that the growth of TiO2 (rutile) is governed by the outward diffusion of Ti. 3.2. Crystallographic orientation relationships As mentioned above, little research has been reported on the microstructural analysis of Ti2AlC after oxidation; in particular, the crystallographic relationships between the newly grown oxide phases and the Ti2AlC matrix are still unknown. The present work focused on possible crystallographic orientation relationships both between the a-Al2O3 layer and the substrate and between these two outer oxide layers of TiO2 (rutile) and a-Al2O3. Several cross-sectional laminae were cut out from different Ti2AlC grains after oxidation along two directions either parallel or perpendicular to the longitudinal directions of the lathy grains. No regular crystallographic relationship was found between these two outer oxide layers of TiO2 (rutile) and a-Al2O3. In contrast, several crystallographic orientations were recorded between the a-Al2O3 oxide and Ti2AlC matrix by electron diffraction. Fig. 4 shows the predominant orientation relationships: ½ 1 1 2a-Al2 O3 // ½0 1 0Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð0 0 1ÞTi2 AlC as well as schematic drawings of the unit cells of a-Al2O3 oxide and Ti2AlC matrix. It is very clear that the basal plane (0 0 1) of Ti2AlC is parallel to the secondary prism plane (1 1 0) of a-Al2O3 oxide. The lattice mismatch between these a-Al2O3 and Ti2AlC orientated phases was also calculated: there is only 3% mismatch between every two unit cells of Ti2AlC along [0 0 1] and three unit cells of aAl2O3 along [1 1 0], implying that the observed orientation relationship is reasonably stable. Ti2AlC has a layered structure composed of three different types of bonds: the relatively weak Ti 3d–Al 3p bond
5219
Fig. 4. (a) SADP showing the crystallographic orientation relationship between the a-Al2O3 oxide and Ti2AlC matrix; (b and c) schematic drawings of their unit cells according to the orientation relationship: ½1 1 2a-Al2 O3 // ½0 1 0Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð0 0 1ÞTi2 AlC .
1 eV below the Fermi level and the stronger Ti 3d–C 2p and Ti 3d–C 2s bonds with energies of around 2.5 and 10 eV below the Fermi level, respectively [18]. Insertion of Al monolayers into face-centered cubic (fcc) TiC matrix in such a way that every second monolayer of C atoms is replaced by an Al atom layer leads to the formation of hexagonal close-packed (hcp) Ti2AlC. As a result, the strong Ti–C bonds are broken up and replaced by weaker Ti–Al bonds at the expense of internal energy. The TiC layers surrounding the Al monolayers are then twinned with the Al layer as the mirror plane. As shown in Fig. 4b, the thermodynamically stable Ti–C–Ti slabs are separated by softer Ti–Al–Ti slabs with weak metallic bonds. Thus the Al atoms may easily decompose at high temperatures and diffuse outward along the direction parallel to the basal plane according to the weaker bonding between Ti and Al. The Al atoms are located in the planes parallel to the basal plane in Ti2AlC. Fig. 4b shows the hexagonal structure of Ti2AlC, and the unit cell of a-Al2O3 is shown in Fig. 4c. As mentioned above, it is easier for Al atoms to diffuse outward parallel to the basal planes of Ti2AlC. When this Al monolayer is diffused out from Ti2AlC, an aluminum oxide layer forms at high temperatures. Although several aluminum oxide phases, e.g. c-Al2O3, d-Al2O3, hAl2O3 as well as TiO2 (anatase), have been reported by other researchers [9,19], in this work only the a-Al2O3 phase has been confirmed by SADPs. The most compacted Ti, Al or C monolayers are perpendicular to the c axis of the unit cell, though Al atoms also lie in other planes, e.g. parallel to the ð1 1 0Þ plane of the Ti2AlC unit cell. Therefore there are various ways for Al
5220
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223
Fig. 6. High-resolution TEM micrograph of the interface between aAl2O3 oxide and Ti2AlC matrix. The two inserts are the simulated images under the same conditions used for TEM. Some crystalline planes of both Ti2AlC matrix phase and a-Al2O3 oxide are indicated. The crystallographic orientation can be deduced from the figure to be: ½1 1 1a-Al2 O3 // ½4 5 1Ti2 AlC , ð1 3 4Þa-Al2 O3 // ð1 1 1ÞTi2 AlC .
Fig. 5. (a) SADP showing another crystallographic orientation relationship between the a-Al2O3 oxide and Ti2AlC matrix; (b and c) schematic drawings of their unit cells according to the orientation relationship: ½0 0 1a-Al2 O3 // ½3 3 1Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð1 1 0ÞTi2 AlC .
atoms to diffuse out of Ti2AlC structure. Fig. 5 shows another observed crystallographic orientation: ½0 0 1a-Al2 O3 // ½3 3 1Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð1 1 0ÞTi2 AlC . This means that besides diffusion parallel to the (0 0 1) basal plane, the Al atoms can also diffuse out in the atom monolayers parallel to the ð1 1 0Þ plane, and perpendicular to the basal (0 0 1) plane. From these crystallographic considerations, we can reason that there may be other orientations between the oxide grains and the Ti2AlC parent grains. Fig. 6 shows a highresolution TEM micrograph of the interface between aAl2O3 oxide and Ti2AlC. Fast Fourier transformation was applied to each high-resolution image of both phases. The results show the following crystallographic orientation:
½1 1 1a-Al2O3 // ½4 5 1Ti2AlC , ð1 3 bar4Þa-Al2O3 // ð1 1 1ÞTi2 AlC . The two inserts are simulated images according to the above direction of each phase, respectively, under the same condition (voltage = 200 kV, spherical aberration (Cs) = 0.5 mm, sample thickness = 37 nm, defocus = 46 nm, which is near to the Scherzer focus of 43 nm). Some crystalline planes of Ti2AlC matrix and a-Al2O3 oxide phases are indicated directly on the picture. It is concluded that besides diffusion of Al parallel to the basal plane of Ti2AlC, other mechanism may operate as observed for the first time in this work. It should be noted that the crystallographic orientation relationships observed in this work are different from that reported by Magnuson et al. [18]. In the latter study, TiCx and Ti2AlC films were epitaxially grown on single-crystal Al2O3. In the present study, the a-Al2O3, TiC and TiO2 layers were oxidation products of Ti2AlC and formed via a different growth mechanism. Another point which should also be emphasized concerns the relationship between the basal plane direction and oxide scale formation. During our experiments, some TEM laminae were prepared parallel to the longitudinal directions of the Ti2AlC lathy grains. However, this longitudinal direction is perpendicular to the c axis of the Ti2AlC unit cell. In other words, it is parallel to the basal planes of Ti2AlC. However, the prepared laminae are not, in general, parallel to these basal planes. It seems that there are more columnar oxide grains on the Ti2AlC surface after prolonged oxidation if the surface is perpendicular to the basal planes since it is easier for Al and Ti outward diffusion and O inward diffusion along these par-
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223
allel basal planes. There are definitely fewer columnar oxide grains if the surface is parallel to the basal planes since it is more difficult for the atoms to diffuse perpendicular to the basal planes. 3.3. Formation mechanism of oxide scales Oxidation of alloys or composites consists in general of two processes: the dissolution of oxygen, and the formation of an oxide scale. During the initial stage of oxidation, molecular oxygen from the gas phase is adsorbed on the substrate surface and dissociates. A thin layer of oxide film is formed by nucleation and lateral growth of oxide crystallites. In general, the oxidation process includes the thermodynamically controlled initial surface reaction and the subsequent growth of an oxide film. The growth is determined predominantly by the rate of diffusion of oxygen ions through the oxide scale and substrate surface. To explain why a-Al2O3 is present in the inner part and TiO2 in the outer part of the scale, Wang et al. assumed that a-Al2O3 and TiO2 (rutile) nucleate simultaneously on the surface of Ti2AlC during the initial oxidation stage due to the oxygen partial pressure being sufficiently high at the ambient air–Ti2AlC interface [8]. Upon the intake of oxygen, the oxygen partial pressure at the oxide scale– Ti2AlC interface decreases with increasing oxidation time. Once the oxygen pressure is lower than the TiO2–Ti2AlC equilibrium pressure, TiO2 ceases further growth because a-Al2O3 is more thermodynamically stable than TiO2. aAl2O3 can grow continuously due to the fact that the formation of a-Al2O3 needs a much lower oxygen pressure [8]. However, this is unlikely to be the physical picture of the oxidation process of Ti2AlC. The electronic structure investigation by Zhou and Sun [20] has shown that the bonding between Ti and C is strongly covalent and directional, whereas the bonding between Ti and Al is weak in the layered ternary compound of Ti2AlC. The strong interaction of Ti and C would decrease the activity of Ti and result in an activity of Al high enough for it to be preferentially oxidized. Based on the TEM observations, therefore, only a-Al2O3 nuclei nucleate on the Ti2AlC matrix and grow in the initial stage of oxidation, forming the inner oxide scale. Once the depletion of Al atoms occurs in the surface layer of Ti2AlC matrix, Ti atoms start to diffuse out and form the outer TiO2 layer. Due to the weak bonding between Ti and Al in layered MAX phases, the cleavage of laminar grains were reported mainly parallel to the hexagonal basal plane (0 0 1) in a MAX-like system (Ti3AlC2). Al will first diffuse out from Ti2AlC matrix and form a-Al2O3 with oxygen. TiC is decomposed from the matrix after Al diffuses out, and hence Ti will diffuse out thereafter through the grain boundaries of preformed a-Al2O3 and react with oxygen to form TiO2 (rutile). This means that a-Al2O3 starts from pure Al while TiO2 starts from TiC. In order to know which oxide will form readily, it is important to compare the free energies for the formation of a-Al2O3 and TiO2
5221
(rutile). The Gibbs free energy for Al2O3 formation from the reaction of Al atom with O at 1473 K is 408 kJ mol–1 and for TiO2 formation from TiC it is 295 kJ mol–1 [11,21]. It was concluded in our previous work [11] that the cleavage surface with Al terminal layer is more easily oxidized than that with a Ti3C2 layer on top in Ti3AlC2 system. Therefore, more Al2O3 nuclei than TiO2 nuclei may be expected to form. In Ref. [11] the formation mechanisms of inner a-Al2O3 and outer TiO2 oxide scales in the Ti3AlC2 system were also analyzed from the viewpoint of diffusion. a-Al2O3 is a fairly stable protective oxide, the growth rate of which is about five orders of magnitude lower than that of TiO2 during the oxidation of Ti– Al alloy [22]. As a result, TiO2 outgrows the inner Al2O3 to form a TiO2 outer layer. 3.4. Atom diffusion and formation of a Ti-rich intermediate layer Wagner studied internal oxidation and selective scale formation as long ago as 1959 [23]. According to his theory, a high content as well as high diffusivity of Al in the substrate, and the low solubility and diffusivity of oxygen in the substrate, favors the preferential oxidation of Al. In the case of Ti2AlC, it was reported that the diffusivity of Al is high and the solubility of oxygen in Ti2AlC is low [8]. Both the low solubility of oxygen and high diffusivity of Al in the substrate are favorable factors for the preferential oxidation of Al. Moreover, the densities of Ti2AlC and a-Al2O3 are 4.21 and 3.98 g cm–3, respectively. Therefore, there is little volume change and a small difference in the thermal expansion coefficient between the a-Al2O3 oxide scale and Ti2AlC substrate. In addition, the oxide scale exhibits good adhesion to the substrate. The formation of a continuous a-Al2O3 layer on Ti2AlC and the good adhesion to the substrate are the main contributions to the excellent oxidation resistance of Ti2AlC. This is also crucial for the recovery of the strength of self-healed materials based on the formation of oxides [16]. Nevertheless, it is a drawback in self-healing process since further oxidation (healing) will be slowed down by the continuous scale. Because the diffusivity of Ti in a-Al2O3 is very low [24] and no Ti concentration gradient was detected across the oxide scale–Ti2AlC interface for samples oxidized at high temperature, Wang et al. concluded that the outward diffusion of Ti via short-circuit diffusion along grain boundaries of the as-formed a-Al2O3 is reasonably fast [8]. In the present study, a discontinuous a-Al2O3 oxide inner layer was observed with high-resolution TEM, which was interrupted by Ti-rich intermediate layer (mostly TiC). This implies that there should be another route for Ti diffusion apart from the short-circuit diffusion through grain boundaries of the as-formed a-Al2O3. The oxide scale of Ti–Al alloys consists of TiO2 and Al2O3 instead of only Al2O3; this was proven both experimentally by Meier et al. [25] and theoretically supported by Rahmel et al. with thermodynamic calculations [26]. According to their results for the
5222
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223
tion mechanism in Ti2AlC during the initial oxidation stage at 1200 °C for 3 min, the following conclusions can be drawn.
Fig. 7. High-resolution TEM micrograph of the interface between Ti2AlC matrix and the Ti-rich intermediate layer containing some randomly orientated a-Al2O3 clusters. Some atomic voids or channels are visible in both the Ti2AlC matrix and the Ti-rich layer for Al and Ti outward diffusion and also for O inward diffusion during the oxidation process.
1. The a-Al2O3 oxide scale can be either a continuous or discontinuous layer on the surface of Ti2AlC matrix. The TiO2 outer scale is formed due to the outward diffusion of Ti after depletion of Al atoms. 2. An intermediate layer is observed between the a-Al2O3 inner scale and Ti2AlC matrix, which is Ti rich and consists mostly of TiC. It transforms from Ti2AlC matrix due to depletion of Al atoms. 3. There are some regular crystallographic orientations between a-Al2O3 oxide scale and Ti2AlC matrix: ½1 1 2a-Al2 O3 // ½0 1 0Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð0 0 1ÞTi2 AlC and ½0 0 1a-Al2 O3 // ½3 3 1Ti2 AlC , ð1 1 0Þa-Al2 O3 // ð1 1 0ÞTi2 AlC . 4. The outward diffusion of Al atoms occurs in certain atomic planes of the Ti2AlC structure. These planes are either (0 0 1) basal planes or prism planes of Ti2AlC. Both will form the secondary prism planes {1 1 0} of Al in hexagonal a-Al2O3 during oxidation.
Acknowledgements Ti–Al–O system, the stable oxide changed from TiO to Al2O3 in coexistence with the TiAl phase because the Al– Al2O3 equilibrium pressure is lower than that for Ti–TiO. The Al content is about 50% in Ti–Al alloys. It is obvious that the Al content in Ti2AlC is much less than this critical amount and the oxidation of Ti2AlC should not yield a continuous Al2O3 layer if the influence of carbon on the activity of Ti and Al was not taken into account [8]. Fig. 7 is a high-resolution TEM micrograph of the interface between the intermediate layer (Ti-rich layer) and Ti2AlC. Small a-Al2O3 clusters are randomly distributed in this layer. Some atomic voids or channels are clearly visible in both the intermediate layer and Ti2AlC. The Ti-rich intermediate layer forms as a consequence of atom diffusion due to the depletion of Al atoms in Ti2AlC. It is therefore a barrier to the outward diffusion of Al atoms but a source of Ti atoms that successively diffuse outwards. Some crystalline planes of both Ti2AlC matrix and a-Al2O3 oxide phases are indicated in the micrograph. In conclusion, we may state that the diffusion channels for Ti are not only the grain boundaries of as-formed a-Al2O3, but also the intermediate Ti-rich layer with randomly distributed aAl2O3 crystalline clusters. Since the intermediate layer subsequently transforms from Ti2AlC matrix, its formation does not destroy the orientation relationships observed between the a-Al2O3 scale and Ti2AlC matrix. 4. Conclusions From the detailed microstructure analysis of the crystallographic relationships between the a-Al2O3 oxide scale and Ti2AlC, and the atom diffusion and oxide scale forma-
J.C.R. and S.B.L. are on leave from China. All the research work was done in Groningen and Delft, the Netherlands. This work was supported by the Netherlands Innovation Oriented Program on Self Healing Materials under Grant No. IOP-SHM 0871. The authors would thank Professor Michel Barsoum (Drexel University, Philadelphia, USA) for providing the Ti2AlC bulk sample. References [1] Barsoum MW, El-Raghy T. J Am Ceram Soc 1996;79:1953. [2] Barsoum MW. Prog Solid State Chem 2000;28:201. [3] Gupta S, Amini S, Filimonov D, Palanisamy T, El-Raghy T, Barsoum MW. J Am Ceram Soc 2007;90:3566. [4] Farber L, Barsoum MW, Zavaliangos A, El-Raghy T, Levin I. J Am Ceram Soc 1998;81:1677. [5] Amini S, Barsoum MW, El-Raghyy T. J Am Ceram Soc 2007;90:3953. [6] El-Raghy T, Zavaliangos A, Barsoum MW, Kalidindi SR. J Am Ceram Soc 1997;80:513. [7] Barsoum MW. J Electrochem Soc 2001;148:C544. [8] Wang XH, Zhou YC. Oxid Met 2003;59:303. [9] Wang XH, Zhou YC. J Mater Res 2002;17:2974. [10] Barsoum MW, Tzenov N, Procopio A, El-Raghy T, Ali M. J Electrochem Soc 2001;148:C551. [11] Song GM, Pei YT, Sloof WG, De Hosson JTM, Van der Zwaag S. Mater Chem Phys 2008;112:762. [12] Song GM, Li SB, Zhao CX, Sloof WG, vd Zwaag S, Pei YT, et al. J Eur Ceram Soc 2011;31:855. [13] Ganguly A, Barsoum MW, Doherty RD. J Am Ceram Soc 2007;90:22002. [14] Wang XH, Zhou YC. Acta Mater 2002;50:3141. [15] Barsoum MW, Brodkin D, El-Raghy T. Scripta Mater 1997;36:535. [16] Song GM, Pei YT, Sloof WG, Li SB, De Hosson JTM, vd Zwaag S. Scripta Mater 2008;58:13. [17] Goretzki H. Phys Status Solidi (b) 1967;20:k141.
J.C. Rao et al. / Acta Materialia 59 (2011) 5216–5223 [18] Magnuson M, Wilhelmsson O, Palmquist J-P, Jansson U, Mattesini M, Li S, et al. Phys Rev B 2006;74:195108. [19] Babu N, Balasubramaniam R, Ghosh A. Corros Sci 2001;43: 2239. [20] Zhou YC, Sun ZM. Phys Rev B 2000;61:12570. [21] Lide DR. Handbook of chemistry and physics. 76th ed. New York: CRC Press; 2000. p. 5–72. [22] Yang MR, Wu SK. Bull College Eng NTU 2003;89:3.
5223
[23] Wagner C. Z Elektrochem 1959;63:772. [24] Langensiepen RA, Tressler RE, Howell PR. J Mater Sci 1983;18:2771. [25] Meier GH, Appalonia D, Perkins RA, Chiang KT. In: Grobstein T, Doychak J, editors. Oxidation of high-temperature intermetallics. Cleveland (OH): TMS; 1988. p. 379. [26] Rahmel A, Spencer PJ. Oxid Met 1991;35:53.