- Email: [email protected]
alumina multilayer coatings grown on sapphire single crystals
Micron 42 (2011) 808–818
Contents lists available at ScienceDirect
Micron journal homepage: www.elsevier.com/locate/micron
CVD TiC/alumina multilayer coatings grown on sapphire single crystals S. Canovic a,∗ , B. Ljungberg b , M. Halvarsson a a b
Dept. of Applied Physics, Chalmers University of Technology, Fysikgränd 3, SE-412 96 Göteborg, Sweden AB Sandvik Tooling, SE-126 80 Stockholm, Sweden
a r t i c l e
i n f o
Article history: Received 2 November 2009 Received in revised form 9 May 2011 Accepted 9 May 2011 Keywords: CVD Alumina Sapphire Single crystal TiC Multilayers
a b s t r a c t Multilayers of TiC/␣-Al2 O3 consisting of three (1 m thick) alumina layers separated by thin (∼10 nm) oxidized TiC layers have been deposited onto c-, a- and r-surfaces of single crystals of ␣-Al2 O3 by chemical vapour deposition (CVD). The aim of this paper is to describe and compare the detailed microstructure of the different multilayer coatings by using transmission electron microscopy (TEM). The general microstructure of the alumina layers is very different when deposited onto different surfaces of ␣-Al2 O3 single crystal substrates. On the c- and a-surfaces the alumina layers grow evenly resulting in growth of single crystal layers of TiC and alumina throughout the coating. However, when deposited on the r-surface the alumina layers generally grow unevenly. No pores are observed within the alumina layers, while a small number of pores are found at the interfaces below the TiC layers. The TiC and alumina layers grow epitaxially on the c- and a-surface substrates. On the r-surface, epitaxy is present only at some rare locations. The TiC layers were oxidized in situ for 2 min in CO2 /H2 prior to the alumina layer deposition. For all three samples chemical analyses show that the whole TiC layer is oxidized. On the c- and a-surfaces the TiC layer was oxidized to an fcc TiCO phase. On the r-surface the oxidation stage resulted in a transformation of the initially deposited fcc TiC to a monoclinic TiCO phase, which appears to be a modified TiO structure with a high carbon content. © 2011 Elsevier Ltd. All rights reserved.
1. Introduction Chemical vapour deposition (CVD) is still today the economically most favourable technique for producing high-quality alumina coatings for cutting tools (Ruppi, 2005). Due to chemical stability and favourable thermal properties, TiX [X = C; N; or (C, N)] and Al2 O3 coatings are often deposited on cemented carbide (WC/Co) cutting tools by CVD in order to increase their wear resistance (Chatfield et al., 1989; Colombier and Lux, 1989; Funk et al., 1976; Halvarsson et al., 1993a,b; Halvarsson and Vuorinen, 1997; Lux et al., 1986; Ruppi, 2005; Sjöstrand, 1979; Vuorinen and Skogsmo, 1988, 1990). Today alumina can be deposited by CVD as three different phases: ␣-Al2 O3 , -Al2 O3 and ␥-Al2 O3 (Vuorinen and Skogsmo, 1990; Larsson and Ruppi, 2001; Ruppi, 1992; Ruppi and Larsson, 2000). In many applications the stable ␣-Al2 O3 is the preferred phase since the other two are metastable and can transform to ␣-Al2 O3 at the high temperatures generated during a metal cutting operation (Dragoo and Diamond, 1967; Levin and Brandon, 1998; Okumiya et al., 1971; Stumpf et al., 1950). In most CVD applications alumina is deposited onto an intermediate layer of TiX, either as a single layer or as a multilayer, where the TiX layer inhibits diffusion of elements from the WC/Co substrate that may
∗ Corresponding author. E-mail address: [email protected] (S. Canovic). 0968-4328/$ – see front matter © 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.micron.2011.05.003
cause uneven nucleation and growth of the alumina layer. Nevertheless, small amounts of W and Co can still diffuse upwards along the TiX grain boundaries (Chatfield et al., 1989; Lux et al., 1986; Vuorinen and Skogsmo, 1988). Earlier studies have shown that the quality of the CVD layers is highly affected by the nucleation and initial growth on the underlying layer (Canovic et al., 2007; Halvarsson et al., 1993a,b; Halvarsson and Vuorinen, 1996). In particular, the structure and chemistry of the nucleation surface is of crucial importance for the phase content of the deposited alumina layer. For instance, non-oxidized TiX surfaces have been shown to favour nucleation of -Al2 O3 (Halvarsson and Vuorinen, 1995; Vuorinen and Skogsmo, 1990). Once nucleated -Al2 O3 is stable and can grow to a thickness of about 10 m (Halvarsson et al., 1993a,b; Halvarsson and Vuorinen, 1995, 1996). However, oxidized TiX surfaces promote nucleation of ␣-Al2 O3 (Chatfield et al., 1989). Halvarsson and Vuorinen (1995) have shown that ␣Al2 O3 can grow epitaxially on chemically vapour deposited Ti2 O3 . Thus, by modifying the nucleation surface, the phase content of the alumina layer can be modified. In addition, both the nucleation step (Ruppi, 2005) and process parameters (Ruppi, 2008) can influence the texture of the ␣-Al2 O3 layer. Many of the published articles in the area of CVD of wear resistant coatings focuses on TiX/alumina multilayer coatings deposited on a few microns thick polycrystalline TiX layer that in turn is deposited on cemented carbide substrates. This means that the nucleation surface for the alumina layer is not well defined and
S. Canovic et al. / Micron 42 (2011) 808–818
known beforehand and diffusion of W and Co from the (cemented carbide) substrate to the nucleation surface can occur. In order to avoid diffusion of elements such as W and Co and complications connected to nucleation surfaces that are not well defined, it is desirable to study model systems with given surfaces and substrates without diffusion of these elements. In a previous study, multilayer coatings of TiC/alumina and TiN/alumina were deposited on ␣-Al2 O3 (sapphire) single crystal substrates with two different surfaces, c-surface (0 0 0 1) and r-surface (0 1 1¯ 2) (Canovic et al., 2010). It was found that the general microstructure of the alumina layers was very different when deposited onto different surfaces of the ␣-Al2 O3 single crystal substrates. On the c-surface the alumina layers grew evenly resulting in growth of single crystal layers of TiX and alumina throughout the coating. However, when deposited on the r-surface the alumina layers generally grew unevenly. In addition, it was found that the TiC/alumina layers grew epitaxially on the c-surface substrates with close-packed planes growing on close-packed planes. On the r-surface, epitaxy was present only at some rare locations in TiC/alumina, while no epitaxy was found in TiN/alumina multilayer coatings in this case. In this work sapphire single crystals with three well-specified surfaces are used as substrates: c-surface (0 0 0 1), a-surface (1 1 2¯ 0) and r-surface (0 1 1¯ 2). Onto these surfaces a TiC/␣-Al2 O3 multilayer coating is deposited. The aim of this paper is to (i) describe the orientation relationships, if present, between ␣-Al2 O3 and TiC, (ii) link interfacial structure, grain shape and grain size to the underlying surface, (iii) describe differences and similarities between the different samples and the different interfaces of each sample, (iv) compare these results with our previous results for TiC/␣-Al2 O3 multilayer coatings deposited on c- and r-surfaces of sapphire single crystals. The investigation of the TiX/alumina multilayer coatings was carried out by using transmission electron microscopy (TEM).
809
ture. The spherical aberration coefficient (Cs ) of the microscope is 1.2 mm. In all experiments presented in this paper the samples were mounted in an FEI low-background double-tilt holder. In order to investigate the chemical composition of the samples two different methods were used: (i) energy dispersive X-ray (EDX) analysis with an Oxford Inca EDX system including evaluation software; and (ii) electron energy loss spectroscopy (EELS) using a Gatan Tridiem 866 GIF system. All EEL spectra were acquired in TEM diffraction mode with a convergence angle ˛ of 1.2 mrad and collection angle ˇ of 4 mrad. The collected EEL spectra were processed using the Gatan Digital Micrograph software. An extraction voltage of 4.0 kV and an intermediate gun lens strength with a spot size of typically < 1 nm was used during the EDX and EELS analysis. During the measurements the sample was tilted less than 5◦ . Cross-section TEM thin foils were prepared using an FEI Strata 235 DB FIB/SEM (focused ion beam/scanning electron microscopy) instrument with an in situ lift-out technique. First an area of interest was chosen and a protective 2 m thick and approximately 20 m × 2 m wide Pt-strip was deposited on top of the area of interest. The next step was to mill a trench (40 m wide) on each side of the Pt-strip by using high ion current. These trenches were subsequently made larger by milling with a lower ion current, making the lift-out specimen below the Pt-strip thinner. The trenches had to be large because of the low milling rate and high re-deposition rate of alumina. The next step was to mill a Ushaped pattern from the side, around the lift-out specimen. A thin Omniprobe needle was inserted, positioned and attached to the lift-out specimen before it was cut free from the bulk specimen. The approximately 4–5 m thick specimen was lifted out and then attached to a Cu-grid. Finally ion milling was carried out until the lift-out specimen was less than 100 nm thin, i.e. electron transparent. 3. Results and discussion
2. Experimental 3.1. Coating deposited on the c-surface The coatings were deposited in a computer-controlled hot-wall CVD reactor. The TiC layers were deposited in a two-step process. First TiC was deposited for 7 min by using TiCl4 –CH4 –H2 . Then the TiC layer was oxidized by flushing CO2 /H2 for 2 min prior to the Al2 O3 layer deposition. The Al2 O3 layers were deposited using H2 , AlCl3 , CO2 and H2 S. The AlCl3 was generated through the chlorination of Al with HCl. The temperature in the reactor during all deposition steps was 1000 ◦ C and the pressure was 55 mbar. Multilayer TiC/alumina coatings are used as they provide several TiC/alumina interfaces in one specimen, which gives better statistics when performing the TEM analysis. In order to make it possible to maintain epitaxy throughout the coating, it is desirable to have an intermediate TiC layer that is thin and not more than one grain thick. Therefore the time used for deposition of TiC was short corresponding to a nominal thickness of 10 nm. ␣Al2 O3 single crystal substrates with three different top surfaces (c-surface, a-surface and r-surface) were coated with three approximately 1 m thick ␣-Al2 O3 layers, separated by thin (10 nm) layers of TiC. The three samples are in this article therefore referred to as “␣c-sub -TiC/Al2 O3 ”, “␣a-sub -TiC/Al2 O3 ”, “␣r-sub -TiC/Al2 O3 ”. ¯ with ␣-Al2 O3 has a trigonal crystal structure (space group R3c) lattice parameters a = 4.75 A˚ and c = 12.97 A˚ (Kronberg, 1957), while ¯ TiC has sodium-chloride structure (Fm3m) with lattice parameter a = 4.33 A˚ (Pierson, 1996). The main part of this work was carried out by transmission electron microscopy (TEM) using an FEI Titan 80-300 TEM/STEM working at 300 kV and a Phillips CM 200 FEG TEM working at 200 kV. The former was used for high-resolution electron microscopy (HREM) imaging, which was performed with slightly underfocused conditions and without using an objective aper-
3.1.1. General microstructure Fig. 1 shows a bright field TEM and a high angle annular dark field (HAADF) scanning transmission electron microscopy (STEM) crosssection image of ␣c-sub -TiC/Al2 O3 . Three ␣-Al2 O3 layers, with a thickness of approximately 1 m each, separated by thin (∼10 nm) TiC layers can be seen in the figure. The thin TiC layers are more clearly visible in Fig. 1(b) that provides atomic number (Z) contrast. The ␣-Al2 O3 layers have grown as single crystals. The dark fringes appearing in Fig. 1(a) are mainly due to strain contrast. This contrast, which extends through several layers, is probably mainly associated with the growth of the multilayer coating but is to some extent also due to bending of the thin TEM sample. On top of the TiC/␣-Al2 O3 coating in Fig. 1(a) there is a dark layer of Pt, deposited during the FIB/SEM sample preparation. The alumina (and TiC) layers grew with an even thickness resulting in straight and parallel interfaces, which resulted in epitaxial growth. Epitaxy and orientation relationships are discussed in more detail below. Some porosity can also be observed in the TiC/␣-Al2 O3 coating. Within the ␣-Al2 O3 layers no pores were observed, while interfacial pores were observed below the TiC layers. The presence of pores has been observed in earlier studies of TiX/alumina multilayer coatings (Canovic et al., 2010; Halvarsson et al., 2006; Lindulf et al., 1994; Skogsmo et al., 1992; Vuorinen and Karlsson, 1992). They may have a negative impact on the coating quality since they act as stress centers and as phase transformation nucleation sites in the coating (Lindulf et al., 1994; Skogsmo et al., 1992; Vuorinen and Karlsson, 1992). The pores are faceted, which indicates the absence of a high-pressure gas trapped inside of them. Further, they are not believed to have formed during specimen preparation
810
S. Canovic et al. / Micron 42 (2011) 808–818
Fig. 1. (a) Bright field TEM and (b) high angle annular dark field STEM micrographs of ␣c-sub -TiC/Al2 O3 . The individual ␣-Al2 O3 layers are separated by thin TiC layers. There is a dark layer of Pt on top of the coating, deposited during sample preparation. Also several pores are present below the TiC layers (some of them indicated with black arrows).
as the ion beam was almost at grazing incidence in the FIB/SEM. In addition, there are also pores present in the substrate below the innermost TiC layer. This indicates that the pores observed below the TiC layers are not caused by the ␣-Al2 O3 deposition process in TiC/␣-Al2 O3 multilayer coatings but that they are formed during the initial deposition of the TiC layers, for instance due to chemical etching. 3.2. Orientation relationships In order to determine the orientation relationships between the layers, selected area electron diffraction (SAED) was used. SAED patterns from the substrate and the three ␣-Al2 O3 layers are shown in Fig. 2(a–d). The patterns shown in Fig. 2(a–c) are almost identical with only small crystal misorientations, meaning that the two innermost ␣-Al2 O3 layers have basically the same orientation as the substrate. However, the pattern from the outermost ␣-Al2 O3 layer in Fig. 2(d) has a different orientation. The micrograph in Fig. 2(f) shows the correct orientation of the sample relative to the diffraction patterns. The diffraction indices are given in Fig. 3, which shows that the ␣-Al2 O3 (0 0 0 6) substrate is parallel to ␣-Al2 O3 (0 0 0 6) layers (the two innermost layers). Thus, the two innermost ␣-Al2 O3 layers grow along the c-axis. The growth direction of the 3rd ␣Al2 O3 layer is also parallel to (0 0 0 6) but the grain (Fig. 3(b)) is rotated 180◦ around the c-axis relative to the other ␣-Al2 O3 layers. A diffraction pattern from the substrate, TiC and 1st ␣-Al2 O3 layer is shown in Fig. 2(e) and the indexing is shown in Fig. 3(c). It can be ¯ seen that ␣-Al2 O3 (0 0 0 6) is parallel to TiC(1¯ 1 1). In ␣c-sub -TiC/Al2 O3 epitaxial growth of TiC was followed by epitaxial growth of ␣-Al2 O3 etc. where every alumina layer is a single crystal with the same orientation as the substrate (except the outermost ␣-Al2 O3 layer that was rotated 180◦ around the c-axis). It should be mentioned here that this rotation could have occurred in any of the other alumina layers. The TiC layers were also single crystals that were related to the alumina layers according to the following orientation relationship: ¯ TiC //(0 0 0 1)␣ //(1¯ 1 1) ¯ TiC //(0 0 0 1)␣ //(1¯ 1 1) ¯ TiC //(0 0 0 1)␣ (0 0 0 1)c-sub //(1¯ 1 1) [2 1¯ 1¯ 0]c-sub //[1¯ 1 2]TiC //[2 1¯ 1¯ 0]␣ //[1¯ 1 2]TiC //[2 1¯ 1¯ 0]␣ //[1¯ 1 2]TiC //[2¯ 1 1 0]␣
¯ and vice versa, i.e. Thus, ␣-Al2 O3 (0 0 0 1) grows on TiC(1¯ 1 1) close-packed planes grow on close-packed planes, which is fre-
quently observed in CVD multilayer coatings (Canovic et al., 2007, 2010; Halvarsson et al., 2006; Halvarsson and Vuorinen, 1997). In conclusion, all the layers in ␣c-sub -TiC/Al2 O3 are single crystals that grow epitaxially. 3.3. High-resolution electron microscopy and chemical analysis High-resolution electron microscopy (HREM) was used in order to study the microstructure at the interfaces in more detail. Fig. 4 shows an HREM micrograph of the substrate/TiC/␣-Al2 O3 region. Both layers consist of well-crystallized grains and both the substrate/TiC and the TiC/␣-Al2 O3 interfaces are relatively even on the atomic level. The coating layers in Fig. 4 have the same orientation as shown above, which was confirmed by using Fast Fourier Transform (FFT). As mentioned above, the TiC layers were oxidized prior to the Al2 O3 layer deposition. But the question is how the oxidation affects the TiC layers. Does only the outermost monolayer(s) get oxidized or does oxygen get incorporated into the whole TiC layer? In order to investigate the chemical composition of the TiC layer, EDX and EELS were used. EDX analysis was performed within the innermost ␣-Al2 O3 layer (∼100 nm above TiC). The resulting EDX spectrum is shown in Fig. 5(a). It can be seen that the intensity of the Al peak from Al2 O3 is twice as high as the intensity of the O peak. The chemical composition of the TiC layer was investigated by performing three analyses at different heights within the TiC layer. Three rectangles in Fig. 4 indicate the location of the measurements, which were performed within a rectangular area (instead of point analysis) to minimize contamination. All three analyses resulted in similar spectra, one of which is shown in Fig. 5(b). In this spectrum it can be seen that the TiC layer contains Ti, C, O and very small amounts of Al (less than 1 at.%). The Al level is much lower than the O level, meaning that only small amounts of the measured O could come from the surrounding Al2 O3 area, and most of the oxygen comes from the TiC layer. Small amounts (just above the detection limit) of Cu and Ga were present in both TiC and Al2 O3 layers. For simplicity they are only marked out in Fig. 5(b). These peaks come from the Cu grid, where the specimen is mounted, and the ion implantation (Ga) during the specimen preparation. Since the EDX analyses from all the three heights resulted in similar spectra this means that the whole TiC layer is oxidized and not only the outermost nanometers. However, the quantification of C and
S. Canovic et al. / Micron 42 (2011) 808–818
811
Fig. 2. SAED patterns from (a) the substrate, (b) 1st, (c) 2nd and (d) 3rd ␣-Al2 O3 layer. (e) SAED pattern consisting of a superposition of patterns from the substrate, TiC and the 1st ␣-Al2 O3 layer. The diffraction indices of the patterns are shown in Fig. 3. (f) A bright field image of the substrate and the alumina layers. The image is rotated correctly relative to the diffraction patterns.
O is difficult with EDX due to the low yield and high absorption of low energy X-rays. Therefore, in order to confirm the oxidation of the TiC layer, a complementary method (EELS) was used. An EELS spectrum from the TiC layer is shown in Fig. 6. The results show presence of Ti, C and O edges, which confirms the results obtained from the EDX analyses. The two chemical analyses (EDX and EELS) show that C and O account for around 25 at.% each (Ti being 50 at.%). The exact levels of C and O are not possible to give, due to the possible carbon contamination and surface oxidation. However, from a comparison with the result from a pure TiC sample (not shown) it is clear that the TiC layer is oxidized and in fact is a Ti(Cx ,Oy ) layer (x and y ≈ 0.5). The oxidized TiC layers will generally be referred to as “TiC” in this article, except in the discussion where the oxidation is treated explicitly. Then the oxidized TiC layers will be referred to as TiCO.
3.4. Coating deposited on the a-surface 3.4.1. General microstructure Fig. 7 shows a cross-section overview of ␣a-sub -TiC/Al2 O3 showing the three ␣-Al2 O3 layers separated by the thin TiC layers. As in the case of ␣c-sub -TiC/Al2 O3 the ␣-Al2 O3 layers are single crystals growing on single crystals of TiC. The dark strain contrast extending through several layers that was present in Fig. 1(a) also appears in Fig. 7. Also in this case no pores were observed within the ␣-Al2 O3 layers, while a small number of interfacial pores were observed below the TiC layers, also in the substrate below the innermost TiC layer. As in the case of ␣c-sub -TiC/Al2 O3 the alumina (and TiC) layers grew with an even thickness resulting in straight and parallel interfaces, which resulted in epitaxial growth. Epitaxy and orientation relationships are discussed in more detail below.
-
131 0006
0330
0112
-0222
N,GD
0114
- -
--
-
0222
0000
0114
0330
0112
-
0330
-
0222
0114
- -
0112
-
a
b
0006
GD
0000
0006
0006
0006
0330 220
-
0112
-0114
0112 --
--
0222
-
311 0222
-- -
- -111 0114 311 GD 0222 0000 -220
-
0330
-0114 111 -Al2O3 TiC
-
0006
- -
-
0112
--
0330
131
c
Fig. 3. The diffraction indices of SAED patterns from Fig. 2. (a) Indices for the substrate, 1st and 2nd ␣-Al2 O3 layers with the zone axis [2 1¯ 1¯ 0]. GD = growth direction, N = substrate normal. (b) Indices for the 3rd ␣-Al2 O3 layer with the zone axis [2¯ 1 1 0]. (c) Indices for the substrate, TiC and the 1st ␣-Al2 O3 layer. The zone axis for TiC is [1¯ 1 2].
812
S. Canovic et al. / Micron 42 (2011) 808–818
Al
a Counts
1000
O 500
Ti 0
0
1500
1
2
3
E (keV )
4
5
6
Ti
b
500 Ti C 0 Fig. 4. HREM image of the inner part of ␣c-sub -TiC/Al2 O3 showing the substrate, TiC and 1st ␣-Al2 O3 layer. Both the substrate/TiC and TiC/␣-Al2 O3 interfaces are relatively even on the atomic level. EDX analysis was performed at three different heights in the TiC layer (marked with rectangles). The results are shown in Fig. 5.
3.4.2. Orientation relationships In order to determine the orientation relationships between the layers, electron diffraction was used. In Fig. 8(a–d) SAED patterns from the substrate and the three ␣-Al2 O3 layers are shown. The patterns in Fig. 8(b–d) are almost identical with only small crystal misorientations, which means that the three ␣-Al2 O3 layers have basically the same orientation. However, the pattern from the substrate shown in Fig. 8(a) is different. The indexing of the diffraction patterns (Fig. 9(a) and (b)) shows that the ␣-Al2 O3 layers grow along the m-surface normal (0 3¯ 3 0). Fig. 8(e) shows a diffraction pattern from the substrate, TiC and the 1st ␣-Al2 O3 layer and Fig. 9(c) shows its indices. It can be seen that (1¯ 2 1¯ 0) from the substrate is parallel to TiC(2 2 0) and ␣-Al2 O3 (0 3¯ 3 0) from the first alumina layer. This means that TiC(1 1 0) grows on the a-surface substrate. Onto this TiC surface an ␣-Al2 O3 layer, which grows
0
O
Cu/Ga
Counts
1000
Ti
Ga Al
1
2
3
E (keV )
4
5
6
Fig. 5. EDX spectra from (a) the 1st ␣-Al2 O3 layer and (b) the TiC layer. The three analyses from the TiC layer marked in Fig. 4 resulted in similar spectra.
along the m-axis, is deposited. Fig. 8(f) is a pattern from the 1st ␣-Al2 O3 , TiC and the 2nd ␣-Al2 O3 layer and the indices are shown in Fig. 9(d). It can be seen that (0 3¯ 3 0) from the first alumina layer is parallel to TiC(2 2 0) and (0 3¯ 3 0) from the second alumina layer. This means that TiC(1 1 0) grows on the ␣-Al2 O3 layer with morientation and onto this TiC surface another ␣-Al2 O3 layer with m-orientation is grown. In ␣a-sub -TiC/Al2 O3 epitaxial growth of TiC was followed by epitaxial growth of ␣-Al2 O3 etc. where every alumina and TiC layer was a single crystal with the same orientation with the following orientation relationship: (1¯ 2 1¯ 0)a-sub //(1 1 0)TiC //(0 1¯ 1 0)␣ //(1 1 0)TiC //(0 1¯ 1 0)␣ //(1 1 0)TiC //(0 1¯ 1 0)␣ [1 1¯ 0 1]a-sub //[1 1¯ 2]TiC //[2 1¯ 1¯ 0]␣ //[1 1¯ 2]TiC //[2 1¯ 1¯ 0]␣ //[1 1¯ 2]TiC //[2 1¯ 1¯ 0]␣
Fig. 6. EELS spectrum from the innermost TiC layer in ␣c-sub -TiC/Al2 O.
S. Canovic et al. / Micron 42 (2011) 808–818
813
same orientation relationship was observed in ␣c-sub -TiC/Al2 O3 . The difference is that the grains in this case are rotated 90◦ , i.e. another (macroscopic) coating growth direction is present. The steps present on the surface are probable nucleation sites for the ␣Al2 O3 layer. At these locations there are “vertical” TiC(1 1 1) faces present, see Fig. 10(b). It is likely that the ␣-Al2 O3 layer initially grows on the vertical TiC(1 1 1) face, to the right in Fig. 10, locally forming the same orientation relationship as in ␣c-sub -TiC/Al2 O3 . As the ␣-Al2 O3 layer continues to grow it will also cover TiC(2 2 0) faces forming the orientation relationship described above. However, it is possible that ␣-Al2 O3 can also nucleate and grow directly on TiC(2 2 0), especially in areas without TiC surface steps. According to this and the orientation relationship discussed above, epitaxial growth of ␣-Al2 O3 on TiC can take place in two ways: upwards in Fig. 10(a) where ␣-Al2 O3 (0 3¯ 3 0) grows epitaxi¯ ally on TiC(2 2 0) and to the right in Fig. 10(a) where ␣-Al2 O3 (0 0 0 6) ¯ Thus, at horizontal TiC(2 2 0) faces grows epitaxially on TiC(1 1¯ 1). Fig. 7. Bright field TEM micrograph of ␣a-sub -TiC/Al2 O3 . The individual ␣-Al2 O3 layers are separated by thin TiC layers.
Thus, TiC(1 1 0) grows on the single crystal substrate with aorientation, followed by growth of ␣-Al2 O3 single crystal with morientation. This is followed by growth of TiC(1 1 0) on the sapphire single crystal with m-orientation. 3.4.3. High-resolution electron microscopy and chemical analysis Fig. 10(a) shows an HREM micrograph of the substrate/TiC/␣Al2 O3 region. Both layers consist of well-crystallized grains. In this case the TiC/␣-Al2 O3 interface is not as even as in ␣c-sub -TiC/Al2 O3 . There are steps of ∼2 nm along the interface, such as that arrowed ¯ is paralin Fig. 10(a). In Fig. 9(c) it can be seen that TiC(1 1¯ 1) ¯ perpendicular to the growth direction. The lel to ␣-Al2 O3 (0 0 0 6)
only the former orientation relationship is present while at ver¯ faces as that shown in Fig. 10(a) both orientation tical TiC(1 1¯ 1) relationships are possible. This is also schematically illustrated in Fig. 10(b). As in the case of ␣c-sub -TiC/Al2 O3 chemical analysis was used in order to investigate the chemical composition of the TiC layer in the same way as was described earlier. The results were similar to the ones shown in Figs. 5 and 6, meaning that the whole TiC layer is oxidized. Thus, the TiC layer is in fact a TiCO layer also in this case. 3.5. Coating deposited on the r-surface 3.5.1. General microstructure Fig. 11(a) shows a cross-section overview of ␣r-sub -TiC/Al2 O3 with the three individual ␣-Al2 O3 layers separated by the thin TiC
Fig. 8. SAED patterns from (a) the substrate, (b) 1st, (c) 2nd and (d) 3rd ␣-Al2 O3 layer. (e) A diffraction pattern from the substrate, TiC and the 1st ␣-Al2 O3 layer. (f) A pattern from the 1st ␣-Al2 O3 , TiC and the 2nd ␣-Al2 O3 layer. The diffraction indices of the patterns are shown in Fig. 9. The growth direction (GD) is indicated in (d).
814
S. Canovic et al. / Micron 42 (2011) 808–818
Fig. 9. The diffraction indices of the SAED patterns from Fig. 8. Indices for (a) the substrate, (b) 1st, 2nd and 3rd ␣-Al2 O3 layers, (c) substrate, TiC and 1st ␣-Al2 O3 layer and (d) 1st ␣-Al2 O3 , TiC and 2nd ␣-Al2 O3 layer. The zone axes are [1¯ 0 1 1] and [2 1¯ 1¯ 0] for the substrate and the alumina layers, respectively, and [1 1¯ 2] for TiC.
layers. The first part of the multilayer coating is shown in Fig. 11(b) in higher magnification. As can be seen in Fig. 11 the first alumina layer grows with an uneven thickness on the first TiC layer and accordingly the following TiC layer is not parallel to the innermost TiC layer. At some rare locations the second TiC layer was parallel to the first one, see, e.g. the area within the circle in Fig. 11(b). At these locations epitaxial columns of grains, extending throughout the coating thickness, were observed. Epitaxy and orientation relationships are discussed in more detail below. Generally, more dislocations and defects were present within the ␣-Al2 O3 layers. For example, the dark fringes (marked with white arrows in Fig. 11(b)) have steps due to local errors present in the crystal during growth, which indicates a more disturbed growth of the coating on the r-surface. 3.5.2. Orientation relationships, high-resolution electron microscopy and chemical analysis Generally no epitaxy is present in the coating; only some few epitaxial columns were found. Fig. 12(a–d) shows SAED patterns from the substrate and the three ␣-Al2 O3 layers from such an epitaxial column. The corresponding diffraction indices are shown in Fig. 12(e). It can be seen that the substrate and all the ␣-Al2 O3 layers have basically the same orientation; they are only slightly
misoriented. Thus, the alumina grains grow with the r-surface ¯ parallel to the growth direction. In order to deternormal (0 1¯ 1 2) mine the orientation relationship between ␣-Al2 O3 and TiC FFTs from HREM images were used. Fig. 13 shows an HREM micrograph of the substrate/TiC/␣-Al2 O3 region. Both layers consist of wellcrystallized grains and both the substrate/TiC and the TiC/␣-Al2 O3 interfaces are quite even on the atomic level. FFT patterns from the TiC and the ␣-Al2 O3 layer from Fig. 13 are shown in Fig. 14. ␣-Al2 O3 could be indexed as in Fig. 12, see Fig. 14(a). The indexing of the TiC layer is shown in Fig. 14(b). Consider especially the spots labeled a, b and c in the figure. The interplanar distances corresponding to the spots a and b fit well with (2 0 0)a and (1 1 1)b for a face centered cubic structure (TiC). However, the angle between these two spots is ∼75◦ instead of ∼55◦ , which it should be for an fcc structure. The interplanar distance corresponding to spot c does therefore not fit with (3 1 1), i.e. the sum of a and b, for the fcc structure. Nevertheless, chemical analyses from the TiC layer showed similar results as for the TiC layers in ␣c-sub -TiC/Al2 O3 and ␣a-sub -TiC/Al2 O3 , meaning that the whole TiC layer is oxidized to TiCO also in this case with around 25 at.% C and 25 at.% O. It has previously been shown that several different titanium oxides (e.g. Ti3 O5 , Ti2 O3 , TiO2 ) can be formed by oxidizing a TiC surface in a CVD reactor (Chatfield et al., 1989; Lhermitte-Sebire
S. Canovic et al. / Micron 42 (2011) 808–818
815
Fig. 10. (a) HREM image of the inner part of ␣a-sub -TiC/Al2 O3 showing the substrate, TiC and 1st ␣-Al2 O3 layer. There is an approximately 2 nm large step (arrowed) at the TiC/␣-Al2 O3 interface. (b) A schematic drawing illustrating how ␣-Al2 O3 grows on TiC. The arrows indicate the plane (index) normals.
et al., 1986). In the present case a short oxidation stage was applied, which should favour a phase with low oxygen content. Therefore, the titanium oxides with high oxidation number should be less probable. Hence, Ti(II)O is the most probable titanium oxide to form. Most of the titanium oxide candidates could also be excluded based on matching of their interplanar distances to the FFT results. However, one specific phase of TiO matches the FFT results, namely the low temperature form of the phase TiO. The unit cell is monoclinic (space group A2/m) with a = 5.855, b = 9.340, c = 4.142 A˚ and = 107.32◦ (Watanabe et al., 1967). The structure is similar to that of NaCl, but has an ordered array of vacant lattice sites. Half of the titanium and half of the oxygen atoms are missing alternately in every third (1 1 0) plane (Watanabe et al., 1967). This phase is stable
in the composition range TiO0.9 –TiO1.1 and at temperatures up to 990 ◦ C (Watanabe et al., 1967). Thus, the TiCO phase formed in the present case seems to be a modified TiO structure, i.e. monoclinic TiO with a relatively large carbon content. This should be possible due to the high density of vacancies present in the TiO structure. Since the observed phase also contains C the temperature stability and the composition range may be influenced compared to pure TiO. The monoclinic TiCO phase matches the FFT results even better if the lattice parameter b is slightly increased from the value for pure TiO. The presence of C in the structure could explain why the unit cell is somewhat distorted compared to pure TiO. From the indexing in Fig. 14 it can be seen that TiCO(3¯ 4¯ 0)m is ¯ In the schematic parallel to the coating growth direction, (0 1¯ 1 2).
Fig. 11. (a) Bright field TEM micrograph of ␣r-sub -TiC/Al2 O3 . (b) An area of the inner part of the coating. The TiC layers are marked with arrows. A region with epitaxy is circled in (b). Several pores are present below the TiC layers (some of them are indicated with black block arrows in (b)). Also several stacking faults are present (some of them indicated with white block arrows).
816
S. Canovic et al. / Micron 42 (2011) 808–818
Fig. 12. SAED patterns from (a) the substrate, (b) 1st, (c) 2nd and (d) 3rd ␣-Al2 O3 layer. (e) The diffraction indices of the patterns. The zone axis is [2 2¯ 0 1].
drawing in Fig. 15 it can be seen that (3 4 0)m , which intersects the monoclinic unit cell at a/3 and b/4, deviates by only a few degrees from (1 1 0)m . This leads to the following orientation relationship:
Generally the coating grows unevenly with random orientation when deposited onto the ␣-Al2 O3 single crystal substrate with
¯ r-sub ∼//(1¯ 1¯ 0)TiCO ∼//(0 1¯ 1 2) ¯ ␣ ∼//(1¯ 1¯ 0)TiCO ∼//(0 1¯ 1 2) ¯ ␣ ∼//(1¯ 1¯ 0)TiCO ∼//(0 1¯ 1 2) ¯ ␣ (0 1¯ 1 2) ¯ TiCO ∼/[2 2¯ 0 1]␣ ∼//[1 1¯ 2] ¯ TiCO ∼//[2 2¯ 0 1]␣ ∼//[1 1¯ 2] ¯ TiCO ∼//[2 2¯ 0 1]␣ [2 2¯ 0 1]r-sub ∼//[1 1¯ 2] Fig. 15 also illustrates the relation between the cubic parent cell and the monoclinic unit cell (Watanabe et al., 1967). For instance it can be seen that (1 1 0)m //(1 0 0)c , i.e. on the r-surface TiCO grows as the monoclinic equivalent to (1 0 0)c . The orientation relationship described here is not similar to the one observed in a previous paper (Canovic et al., 2010).
r-orientation. However, at some rare locations epitaxial columns ¯ forms on the of grains are observed where monoclinic TiCO(1¯ 10) ␣-Al2 O3 single crystal substrate. The originally deposited TiC layer had a cubic structure. After the oxidation stage the cubic TiC incorporated oxygen and transformed to a monoclinic structure. It is well known that TiC has an fcc structure (Pierson, 1996), while TiO has a monoclinic structure (Watanabe et al., 1967). It is likely that the structure of the TiCO formed during oxidation is strongly dependent on its C/O ratio. A high C/O ratio increases the probability for formation of a TiC-like structure (fcc), while a low C/O ratio probably favours formation of a TiO-like structure (monoclinic). It is possible that the oxidation process progresses faster in some TiC layer regions containing, e.g. defects and/or specific grain orientations (i.e. the r-surface sample), which would lead to a higher amount of oxygen in some regions. Thus, in some cases it should be possible to have a local transformation to a monoclinic phase in some regions, while the structure remains fcc in others.
4. Concluding discussion
Fig. 13. HREM image of the substrate, TiC and 1st ␣-Al2 O3 layer. Both the substrate/TiC and TiC/␣-Al2 O3 interfaces are relatively even on the atomic level.
Multilayers of TiC/␣-Al2 O3 consisting of three alumina layers separated by thin oxidized TiC layers, have been deposited onto c-, a- and r-surfaces of single crystals of ␣-Al2 O3 . In all three cases chemical analyses showed that the whole TiC layer is oxidized and is in fact a TiCO layer. When deposited on the c- and asurface substrates the alumina layers could grow evenly resulting in growth of single crystal layers of TiCO and alumina. However, when deposited on the r-surface the alumina layers exhibited uneven, faceted growth. A small number of pores were observed at the interfaces below the TiCO layers, probably due to the initial TiC
S. Canovic et al. / Micron 42 (2011) 808–818
817
Fig. 14. FFTs from (a) the ␣-Al2 O3 layer and (b) the TiC layer in Fig. 13. ␣-Al2 O3 could be indexed as in Fig. 12, TiC is indexed as a monoclinic structure with the zone axis ¯ The spots labeled a, b and c are discussed in the text. [4 3¯ 5].
deposition process step. The presence of pores has been observed in earlier studies of TiX/alumina multilayer coatings (Canovic et al., 2010; Halvarsson et al., 2006; Lindulf et al., 1994; Skogsmo et al., 1992; Vuorinen and Karlsson, 1992). In ␣c-sub -TiC/Al2 O3 and ␣a-sub -TiC/Al2 O3 epitaxy was frequently found. However, on the r-surface the alumina layers did not grow with an even thickness, which resulted in non-parallel TiC layers. At some rare locations straight and parallel layers were found also in ␣r-sub -TiC/Al2 O3 . At these locations epitaxial columns of grains were observed. Schematics summarizing the orientation relationships, when present, obtained in the different coatings are shown in Fig. 16. In ␣c-sub -TiC/Al2 O3 epitaxial growth of TiC was followed by epitaxial growth of ␣-Al2 O3 etc. where every alumina layer is a single crystal with the same orientation as the substrate (except
Fig. 15. Schematic drawing of the TiO structure viewed in the [0 0 1] direction at z = 0. The monoclinic unit cell parameters a, b and are indicated. The pseudo-cubic unit cell (ac and bc ) is also indicated. One plane for (3 4 0)m and one for (1 1 0)m (corresponding to (1 0 0)c ) are given.
the outermost ␣-Al2 O3 layer that was rotated 180◦ around the c-axis, which could have occurred in any of the other alumina layers). The TiC layers were also single crystals that were related to the alumina layers according to the orientation relationship shown in Fig. 16 (left). This is the same orientation relationship as was observed in a previous paper (Canovic et al., 2010). Thus, first TiC(1 1 1) grows on ␣-Al2 O3 (0 0 0 1). Then, the TiC layer is oxidized during the oxidation step and forms fcc TiCO followed by growth of ␣-Al2 O3 (0 0 0 1) on TiCO(1 1 1). In both cases close-packed planes grow on close-packed planes, which is frequently observed in CVD multilayer coatings (Canovic et al., 2007, 2010; Halvarsson et al., 2006; Halvarsson and Vuorinen, 1997). In conclusion, all the layers in ␣c-sub -TiC/Al2 O3 are single crystals that grow epitaxially. Similarly, in ␣a-sub -TiC/Al2 O3 epitaxy was frequently present and the orientation relationship shown in Fig. 16 (center) was found. Thus, TiC(1 1 0) grows on the single crystal substrate with a-orientation. It is oxidized to form fcc TiCO and followed by growth of ␣-Al2 O3 single crystal with m-orientation. This is followed by growth of TiC(1 1 0) on the ␣-Al2 O3 m-surface. Then the TiC layer is oxidized and the sequence is repeated for the other ␣-Al2 O3 and TiC layers. On the r-surface, which is not close-packed, more irregular coating growth takes place. In this case, the alumina layers do not grow evenly, i.e. the top surfaces of the first alumina layer are generally not the same as the top surface of the substrate. At some rare locations, where they are the same epitaxy is observed. In these cases the orientation relationship shown in Fig. 16 (right) was observed. In ␣r-sub -TiC/Al2 O3 the oxidation step resulted in a transformation of the fcc TiC to a monoclinic TiCO phase. From FFT and chemical analyses it was found that the originally deposited TiC structure (fcc) incorporates oxygen and transforms to monoclinic TiCO. It is possible that the oxidation process progresses faster in some TiC layer regions containing, e.g. defects and/or specific grain orientations, which would lead to a higher amount of oxygen in some regions. Thus, in some cases it should be possible to have a local transformation to a monoclinic phase in some regions, while the structure remains fcc in others. In the present work, it is shown that the TiC layers can be affected differently when oxidized. This is important since the TiC surface affects the subsequent deposition of Al2 O3 . However, in this work ␣-Al2 O3 formed in all cases, even when the TiC layer was oxidized to a TiCO phase with a monoclinic structure. Thus, during a short oxidation step the originally deposited fcc TiC layer can incorporate oxygen and form an fcc TiCO phase (on the c- and a-surfaces), or monoclinic TiCO (on the r-surface). An extended oxidation step
818
S. Canovic et al. / Micron 42 (2011) 808–818
c (180° rot)
m
111
110
TiCO (fcc)
c 111
c
TiCO (fcc)
m TiCO (fcc)
c 111
r
110
c-sub
110
TiCO (m.c.)
r TiCO (fcc)
110
TiCO (m.c.)
r
m TiCO (fcc)
110
TiCO (fcc)
a
a-sub
110 r
TiCO (m.c.) r-sub
Fig. 16. Schematics showing the coating sequences for the three samples. (m.c.) = (monoclinic).
may lead to formation of different titanium oxides (e.g. TiO, TiO2 , Ti2 O3 , Ti3 O5 ). Excessive oxidation of the TiC layer prior to or during initial alumina layer deposition may result in considerable expansion of the TiC layer. Therefore, it is probably important to have low oxidizing conditions for the TiC layer during the oxidation step and the initial alumina layer deposition. In general, the TiC oxidation step is a sensitive and important part of the CVD process of these types of multilayer coatings. Acknowledgement Financial support was provided by the Swedish National Graduate School in Materials Science (NFSM). References Canovic, S., et al., 2007. TEM and DFT investigation of CVD TiN/-Al2 O3 multilayer coatings. Surf. Coat. Technol. 202, 522–531. Canovic, S., et al., 2010. CVD TiC/alumina and TiN/alumina coatings grown on sapphire single crystals. Int. J. Refract. Met. Hard Mater. 28, 163–173. Chatfield, C., Lindström, J.N., Sjöstrand, M.E., 1989. Microstructure of CVD Al2 O3 . J. Phys. C5, 377–387. Colombier, C., Lux, B., 1989. Formation of mixed TiC/Al2 O3 layers and ␣- and -Al2 O3 on cemented carbides by chemical vapour deposition. J. Mater. Sci. 24, 462–470. Dragoo, A.L., Diamond, J.J., 1967. Transitions in vapor-deposited alumina from 300◦ to 1200 ◦ C. J. Am. Chem. Soc. 50 (11), 568–574. Funk, R., et al., 1976. Coating of cemented carbide cutting tools with alumina by chemical vapor deposition. J. Electrochem. Soc. 123 (2), 285–289. Halvarsson, M., Nordén, H., Vuorinen, S., 1993a. Microstructural investigation of CVD ␣-Al2 O3 /-Al2 O3 multilayer coatings. Surf. Coat. Technol. 61, 177–181. Halvarsson, M., Trancik, J.E., Ruppi, S., 2006. The microstructure of CVD -Al2 O3 multilayers separated by thin intermediate TiN or TiC layers. Int. J. Refract. Met. Hard Mater. 24, 32–38. Halvarsson, M., Vuorinen, S., Nordén, H., 1993b. Microstructural investigation of -modification layers in CVD ␣-Al2 O3 /-Al2 O3 multilayer coatings. Mater. Res. Soc. Symp. Proc. 314, 83–88. Halvarsson, M., Vuorinen, S., 1995. The influence of the nucleation surface on the growth of CVD ␣-Al2 O3 and -Al2 O3 . Surf. Coat. Technol. 76/77, 287–296. Halvarsson, M., Vuorinen, S., 1996. Epitaxy in multilayer coatings of -Al2 O3 . Surf. Coat. Technol. 80, 80–88.
Halvarsson, M., Vuorinen, S., 1997. Microstructural investigation of CVD TiN/-Al2 O3 multilayer coatings on cemented carbides. Int. J. Refract. Met. Hard Mater. 15, 169–178. Kronberg, M.L., 1957. Plastic deformation of single crystals of sapphire: basal slip and twinning. Acta Metall. 5, 507–524. Larsson, A., Ruppi, S., 2001. Microstructure and properties of CVD ␥-Al2 O3 coatings. Int. J. Refract. Met. Hard Mater. 19 (4–6), 515–522. Levin, I., Brandon, D., 1998. Metastable alumina polymorphs: crystal structures and transition sequences. J. Am. Ceram. Soc. 81 (8), 1995–2012. Lhermitte-Sebire, I., et al., 1986. Microanalysis of the Al2 O3 –TiC interface in aluminacoated cemented carbides by Auger electron and glow discharge emission spectroscopies. Thin Solid Films 138, 209–220. Lindulf, N., et al., 1994. Microstructural investigation of the -Al2 O3 → ␣-Al2 O3 transformation in multilayer coatings of chemically vapour deposited -Al2 O3 . Thin Solid Films 253, 311–317. Lux, B., et al., 1986. Preparation of alumina coatings by chemical vapour deposition. Thin Solid Films 138 (1), 49–64. Okumiya, M., et al., 1971. The formation of - and -Al2 O3 from the dehydration of thodite 5Al2 O3 × H2 O. Bull. Chem. Soc. Jpn. 44 (2), 418–423. Pierson, H.O., 1996. Handbook of Refractory Carbides and Nitrides. Noyes Publications, New Jersey, USA. Ruppi, S., 1992. Multi-oxide coated carbide body and the method of producing the same. U.S. Patent 5,137,774. Ruppi, S., Larsson, A., 2000. Chemical vapour deposition of -Al2 O3 . Thin Solid Films 388 (1/2), 50–61. Ruppi, S., 2005. Deposition, microstructure and properties of texturecontrolled CVD ␣-Al2 O3 coatings. Int. J. Refract. Met. Hard Mater. 23, 306– 316. Ruppi, S., 2008. Enhanced performance of ␣-Al2 O3 coatings by control of crystal orientation. Surf. Coat. Technol. 202 (17), 4257–4269. Sjöstrand, M.E., 1979. Deposition of wear resistant TiN on cemented carbides using mixtures of NH3 /N2 and TiCl4 /H2 . Proc. CVD, 452–453. Skogsmo, J., Halvarsson, M., Vuorinen, S., 1992. Phase transformation of CVD Al2 O3 . Micron Microsc. Acta 23 (1/2), 219–220. Stumpf, H.C., et al., 1950. Thermal transformation of aluminas and alumina hydrates. Ind. Eng. Chem. 42 (7), 1398–1403. Vuorinen, S., Karlsson, L., 1992. Phase transformation in chemical vapour-deposited -alumina. Thin Solid Films 214, 132–143. Vuorinen, S., Skogsmo, J., 1988. In: Sudarshan, T., Bhat, D. (Eds.), Microstructural Characterization of Wear Resistant Coatings. Surface Modification Technologies. TMS, Phoenix, AZ, pp. 143–168. Vuorinen, S., Skogsmo, J., 1990. Characterization of ␣-Al2 O3 , -Al2 O3 and ␣ multioxide coatings on cemented carbide. Thin Solid Films 193–194, 536–546. Watanabe, D., et al., 1967. The ordered structure of TiO. Acta Cryst., 23.
Contents lists available at ScienceDirect
Micron journal homepage: www.elsevier.com/locate/micron
CVD TiC/alumina multilayer coatings grown on sapphire single crystals S. Canovic a,∗ , B. Ljungberg b , M. Halvarsson a a b
Dept. of Applied Physics, Chalmers University of Technology, Fysikgränd 3, SE-412 96 Göteborg, Sweden AB Sandvik Tooling, SE-126 80 Stockholm, Sweden
a r t i c l e
i n f o
Article history: Received 2 November 2009 Received in revised form 9 May 2011 Accepted 9 May 2011 Keywords: CVD Alumina Sapphire Single crystal TiC Multilayers
a b s t r a c t Multilayers of TiC/␣-Al2 O3 consisting of three (1 m thick) alumina layers separated by thin (∼10 nm) oxidized TiC layers have been deposited onto c-, a- and r-surfaces of single crystals of ␣-Al2 O3 by chemical vapour deposition (CVD). The aim of this paper is to describe and compare the detailed microstructure of the different multilayer coatings by using transmission electron microscopy (TEM). The general microstructure of the alumina layers is very different when deposited onto different surfaces of ␣-Al2 O3 single crystal substrates. On the c- and a-surfaces the alumina layers grow evenly resulting in growth of single crystal layers of TiC and alumina throughout the coating. However, when deposited on the r-surface the alumina layers generally grow unevenly. No pores are observed within the alumina layers, while a small number of pores are found at the interfaces below the TiC layers. The TiC and alumina layers grow epitaxially on the c- and a-surface substrates. On the r-surface, epitaxy is present only at some rare locations. The TiC layers were oxidized in situ for 2 min in CO2 /H2 prior to the alumina layer deposition. For all three samples chemical analyses show that the whole TiC layer is oxidized. On the c- and a-surfaces the TiC layer was oxidized to an fcc TiCO phase. On the r-surface the oxidation stage resulted in a transformation of the initially deposited fcc TiC to a monoclinic TiCO phase, which appears to be a modified TiO structure with a high carbon content. © 2011 Elsevier Ltd. All rights reserved.
1. Introduction Chemical vapour deposition (CVD) is still today the economically most favourable technique for producing high-quality alumina coatings for cutting tools (Ruppi, 2005). Due to chemical stability and favourable thermal properties, TiX [X = C; N; or (C, N)] and Al2 O3 coatings are often deposited on cemented carbide (WC/Co) cutting tools by CVD in order to increase their wear resistance (Chatfield et al., 1989; Colombier and Lux, 1989; Funk et al., 1976; Halvarsson et al., 1993a,b; Halvarsson and Vuorinen, 1997; Lux et al., 1986; Ruppi, 2005; Sjöstrand, 1979; Vuorinen and Skogsmo, 1988, 1990). Today alumina can be deposited by CVD as three different phases: ␣-Al2 O3 , -Al2 O3 and ␥-Al2 O3 (Vuorinen and Skogsmo, 1990; Larsson and Ruppi, 2001; Ruppi, 1992; Ruppi and Larsson, 2000). In many applications the stable ␣-Al2 O3 is the preferred phase since the other two are metastable and can transform to ␣-Al2 O3 at the high temperatures generated during a metal cutting operation (Dragoo and Diamond, 1967; Levin and Brandon, 1998; Okumiya et al., 1971; Stumpf et al., 1950). In most CVD applications alumina is deposited onto an intermediate layer of TiX, either as a single layer or as a multilayer, where the TiX layer inhibits diffusion of elements from the WC/Co substrate that may
∗ Corresponding author. E-mail address: [email protected] (S. Canovic). 0968-4328/$ – see front matter © 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.micron.2011.05.003
cause uneven nucleation and growth of the alumina layer. Nevertheless, small amounts of W and Co can still diffuse upwards along the TiX grain boundaries (Chatfield et al., 1989; Lux et al., 1986; Vuorinen and Skogsmo, 1988). Earlier studies have shown that the quality of the CVD layers is highly affected by the nucleation and initial growth on the underlying layer (Canovic et al., 2007; Halvarsson et al., 1993a,b; Halvarsson and Vuorinen, 1996). In particular, the structure and chemistry of the nucleation surface is of crucial importance for the phase content of the deposited alumina layer. For instance, non-oxidized TiX surfaces have been shown to favour nucleation of -Al2 O3 (Halvarsson and Vuorinen, 1995; Vuorinen and Skogsmo, 1990). Once nucleated -Al2 O3 is stable and can grow to a thickness of about 10 m (Halvarsson et al., 1993a,b; Halvarsson and Vuorinen, 1995, 1996). However, oxidized TiX surfaces promote nucleation of ␣-Al2 O3 (Chatfield et al., 1989). Halvarsson and Vuorinen (1995) have shown that ␣Al2 O3 can grow epitaxially on chemically vapour deposited Ti2 O3 . Thus, by modifying the nucleation surface, the phase content of the alumina layer can be modified. In addition, both the nucleation step (Ruppi, 2005) and process parameters (Ruppi, 2008) can influence the texture of the ␣-Al2 O3 layer. Many of the published articles in the area of CVD of wear resistant coatings focuses on TiX/alumina multilayer coatings deposited on a few microns thick polycrystalline TiX layer that in turn is deposited on cemented carbide substrates. This means that the nucleation surface for the alumina layer is not well defined and
S. Canovic et al. / Micron 42 (2011) 808–818
known beforehand and diffusion of W and Co from the (cemented carbide) substrate to the nucleation surface can occur. In order to avoid diffusion of elements such as W and Co and complications connected to nucleation surfaces that are not well defined, it is desirable to study model systems with given surfaces and substrates without diffusion of these elements. In a previous study, multilayer coatings of TiC/alumina and TiN/alumina were deposited on ␣-Al2 O3 (sapphire) single crystal substrates with two different surfaces, c-surface (0 0 0 1) and r-surface (0 1 1¯ 2) (Canovic et al., 2010). It was found that the general microstructure of the alumina layers was very different when deposited onto different surfaces of the ␣-Al2 O3 single crystal substrates. On the c-surface the alumina layers grew evenly resulting in growth of single crystal layers of TiX and alumina throughout the coating. However, when deposited on the r-surface the alumina layers generally grew unevenly. In addition, it was found that the TiC/alumina layers grew epitaxially on the c-surface substrates with close-packed planes growing on close-packed planes. On the r-surface, epitaxy was present only at some rare locations in TiC/alumina, while no epitaxy was found in TiN/alumina multilayer coatings in this case. In this work sapphire single crystals with three well-specified surfaces are used as substrates: c-surface (0 0 0 1), a-surface (1 1 2¯ 0) and r-surface (0 1 1¯ 2). Onto these surfaces a TiC/␣-Al2 O3 multilayer coating is deposited. The aim of this paper is to (i) describe the orientation relationships, if present, between ␣-Al2 O3 and TiC, (ii) link interfacial structure, grain shape and grain size to the underlying surface, (iii) describe differences and similarities between the different samples and the different interfaces of each sample, (iv) compare these results with our previous results for TiC/␣-Al2 O3 multilayer coatings deposited on c- and r-surfaces of sapphire single crystals. The investigation of the TiX/alumina multilayer coatings was carried out by using transmission electron microscopy (TEM).
809
ture. The spherical aberration coefficient (Cs ) of the microscope is 1.2 mm. In all experiments presented in this paper the samples were mounted in an FEI low-background double-tilt holder. In order to investigate the chemical composition of the samples two different methods were used: (i) energy dispersive X-ray (EDX) analysis with an Oxford Inca EDX system including evaluation software; and (ii) electron energy loss spectroscopy (EELS) using a Gatan Tridiem 866 GIF system. All EEL spectra were acquired in TEM diffraction mode with a convergence angle ˛ of 1.2 mrad and collection angle ˇ of 4 mrad. The collected EEL spectra were processed using the Gatan Digital Micrograph software. An extraction voltage of 4.0 kV and an intermediate gun lens strength with a spot size of typically < 1 nm was used during the EDX and EELS analysis. During the measurements the sample was tilted less than 5◦ . Cross-section TEM thin foils were prepared using an FEI Strata 235 DB FIB/SEM (focused ion beam/scanning electron microscopy) instrument with an in situ lift-out technique. First an area of interest was chosen and a protective 2 m thick and approximately 20 m × 2 m wide Pt-strip was deposited on top of the area of interest. The next step was to mill a trench (40 m wide) on each side of the Pt-strip by using high ion current. These trenches were subsequently made larger by milling with a lower ion current, making the lift-out specimen below the Pt-strip thinner. The trenches had to be large because of the low milling rate and high re-deposition rate of alumina. The next step was to mill a Ushaped pattern from the side, around the lift-out specimen. A thin Omniprobe needle was inserted, positioned and attached to the lift-out specimen before it was cut free from the bulk specimen. The approximately 4–5 m thick specimen was lifted out and then attached to a Cu-grid. Finally ion milling was carried out until the lift-out specimen was less than 100 nm thin, i.e. electron transparent. 3. Results and discussion
2. Experimental 3.1. Coating deposited on the c-surface The coatings were deposited in a computer-controlled hot-wall CVD reactor. The TiC layers were deposited in a two-step process. First TiC was deposited for 7 min by using TiCl4 –CH4 –H2 . Then the TiC layer was oxidized by flushing CO2 /H2 for 2 min prior to the Al2 O3 layer deposition. The Al2 O3 layers were deposited using H2 , AlCl3 , CO2 and H2 S. The AlCl3 was generated through the chlorination of Al with HCl. The temperature in the reactor during all deposition steps was 1000 ◦ C and the pressure was 55 mbar. Multilayer TiC/alumina coatings are used as they provide several TiC/alumina interfaces in one specimen, which gives better statistics when performing the TEM analysis. In order to make it possible to maintain epitaxy throughout the coating, it is desirable to have an intermediate TiC layer that is thin and not more than one grain thick. Therefore the time used for deposition of TiC was short corresponding to a nominal thickness of 10 nm. ␣Al2 O3 single crystal substrates with three different top surfaces (c-surface, a-surface and r-surface) were coated with three approximately 1 m thick ␣-Al2 O3 layers, separated by thin (10 nm) layers of TiC. The three samples are in this article therefore referred to as “␣c-sub -TiC/Al2 O3 ”, “␣a-sub -TiC/Al2 O3 ”, “␣r-sub -TiC/Al2 O3 ”. ¯ with ␣-Al2 O3 has a trigonal crystal structure (space group R3c) lattice parameters a = 4.75 A˚ and c = 12.97 A˚ (Kronberg, 1957), while ¯ TiC has sodium-chloride structure (Fm3m) with lattice parameter a = 4.33 A˚ (Pierson, 1996). The main part of this work was carried out by transmission electron microscopy (TEM) using an FEI Titan 80-300 TEM/STEM working at 300 kV and a Phillips CM 200 FEG TEM working at 200 kV. The former was used for high-resolution electron microscopy (HREM) imaging, which was performed with slightly underfocused conditions and without using an objective aper-
3.1.1. General microstructure Fig. 1 shows a bright field TEM and a high angle annular dark field (HAADF) scanning transmission electron microscopy (STEM) crosssection image of ␣c-sub -TiC/Al2 O3 . Three ␣-Al2 O3 layers, with a thickness of approximately 1 m each, separated by thin (∼10 nm) TiC layers can be seen in the figure. The thin TiC layers are more clearly visible in Fig. 1(b) that provides atomic number (Z) contrast. The ␣-Al2 O3 layers have grown as single crystals. The dark fringes appearing in Fig. 1(a) are mainly due to strain contrast. This contrast, which extends through several layers, is probably mainly associated with the growth of the multilayer coating but is to some extent also due to bending of the thin TEM sample. On top of the TiC/␣-Al2 O3 coating in Fig. 1(a) there is a dark layer of Pt, deposited during the FIB/SEM sample preparation. The alumina (and TiC) layers grew with an even thickness resulting in straight and parallel interfaces, which resulted in epitaxial growth. Epitaxy and orientation relationships are discussed in more detail below. Some porosity can also be observed in the TiC/␣-Al2 O3 coating. Within the ␣-Al2 O3 layers no pores were observed, while interfacial pores were observed below the TiC layers. The presence of pores has been observed in earlier studies of TiX/alumina multilayer coatings (Canovic et al., 2010; Halvarsson et al., 2006; Lindulf et al., 1994; Skogsmo et al., 1992; Vuorinen and Karlsson, 1992). They may have a negative impact on the coating quality since they act as stress centers and as phase transformation nucleation sites in the coating (Lindulf et al., 1994; Skogsmo et al., 1992; Vuorinen and Karlsson, 1992). The pores are faceted, which indicates the absence of a high-pressure gas trapped inside of them. Further, they are not believed to have formed during specimen preparation
810
S. Canovic et al. / Micron 42 (2011) 808–818
Fig. 1. (a) Bright field TEM and (b) high angle annular dark field STEM micrographs of ␣c-sub -TiC/Al2 O3 . The individual ␣-Al2 O3 layers are separated by thin TiC layers. There is a dark layer of Pt on top of the coating, deposited during sample preparation. Also several pores are present below the TiC layers (some of them indicated with black arrows).
as the ion beam was almost at grazing incidence in the FIB/SEM. In addition, there are also pores present in the substrate below the innermost TiC layer. This indicates that the pores observed below the TiC layers are not caused by the ␣-Al2 O3 deposition process in TiC/␣-Al2 O3 multilayer coatings but that they are formed during the initial deposition of the TiC layers, for instance due to chemical etching. 3.2. Orientation relationships In order to determine the orientation relationships between the layers, selected area electron diffraction (SAED) was used. SAED patterns from the substrate and the three ␣-Al2 O3 layers are shown in Fig. 2(a–d). The patterns shown in Fig. 2(a–c) are almost identical with only small crystal misorientations, meaning that the two innermost ␣-Al2 O3 layers have basically the same orientation as the substrate. However, the pattern from the outermost ␣-Al2 O3 layer in Fig. 2(d) has a different orientation. The micrograph in Fig. 2(f) shows the correct orientation of the sample relative to the diffraction patterns. The diffraction indices are given in Fig. 3, which shows that the ␣-Al2 O3 (0 0 0 6) substrate is parallel to ␣-Al2 O3 (0 0 0 6) layers (the two innermost layers). Thus, the two innermost ␣-Al2 O3 layers grow along the c-axis. The growth direction of the 3rd ␣Al2 O3 layer is also parallel to (0 0 0 6) but the grain (Fig. 3(b)) is rotated 180◦ around the c-axis relative to the other ␣-Al2 O3 layers. A diffraction pattern from the substrate, TiC and 1st ␣-Al2 O3 layer is shown in Fig. 2(e) and the indexing is shown in Fig. 3(c). It can be ¯ seen that ␣-Al2 O3 (0 0 0 6) is parallel to TiC(1¯ 1 1). In ␣c-sub -TiC/Al2 O3 epitaxial growth of TiC was followed by epitaxial growth of ␣-Al2 O3 etc. where every alumina layer is a single crystal with the same orientation as the substrate (except the outermost ␣-Al2 O3 layer that was rotated 180◦ around the c-axis). It should be mentioned here that this rotation could have occurred in any of the other alumina layers. The TiC layers were also single crystals that were related to the alumina layers according to the following orientation relationship: ¯ TiC //(0 0 0 1)␣ //(1¯ 1 1) ¯ TiC //(0 0 0 1)␣ //(1¯ 1 1) ¯ TiC //(0 0 0 1)␣ (0 0 0 1)c-sub //(1¯ 1 1) [2 1¯ 1¯ 0]c-sub //[1¯ 1 2]TiC //[2 1¯ 1¯ 0]␣ //[1¯ 1 2]TiC //[2 1¯ 1¯ 0]␣ //[1¯ 1 2]TiC //[2¯ 1 1 0]␣
¯ and vice versa, i.e. Thus, ␣-Al2 O3 (0 0 0 1) grows on TiC(1¯ 1 1) close-packed planes grow on close-packed planes, which is fre-
quently observed in CVD multilayer coatings (Canovic et al., 2007, 2010; Halvarsson et al., 2006; Halvarsson and Vuorinen, 1997). In conclusion, all the layers in ␣c-sub -TiC/Al2 O3 are single crystals that grow epitaxially. 3.3. High-resolution electron microscopy and chemical analysis High-resolution electron microscopy (HREM) was used in order to study the microstructure at the interfaces in more detail. Fig. 4 shows an HREM micrograph of the substrate/TiC/␣-Al2 O3 region. Both layers consist of well-crystallized grains and both the substrate/TiC and the TiC/␣-Al2 O3 interfaces are relatively even on the atomic level. The coating layers in Fig. 4 have the same orientation as shown above, which was confirmed by using Fast Fourier Transform (FFT). As mentioned above, the TiC layers were oxidized prior to the Al2 O3 layer deposition. But the question is how the oxidation affects the TiC layers. Does only the outermost monolayer(s) get oxidized or does oxygen get incorporated into the whole TiC layer? In order to investigate the chemical composition of the TiC layer, EDX and EELS were used. EDX analysis was performed within the innermost ␣-Al2 O3 layer (∼100 nm above TiC). The resulting EDX spectrum is shown in Fig. 5(a). It can be seen that the intensity of the Al peak from Al2 O3 is twice as high as the intensity of the O peak. The chemical composition of the TiC layer was investigated by performing three analyses at different heights within the TiC layer. Three rectangles in Fig. 4 indicate the location of the measurements, which were performed within a rectangular area (instead of point analysis) to minimize contamination. All three analyses resulted in similar spectra, one of which is shown in Fig. 5(b). In this spectrum it can be seen that the TiC layer contains Ti, C, O and very small amounts of Al (less than 1 at.%). The Al level is much lower than the O level, meaning that only small amounts of the measured O could come from the surrounding Al2 O3 area, and most of the oxygen comes from the TiC layer. Small amounts (just above the detection limit) of Cu and Ga were present in both TiC and Al2 O3 layers. For simplicity they are only marked out in Fig. 5(b). These peaks come from the Cu grid, where the specimen is mounted, and the ion implantation (Ga) during the specimen preparation. Since the EDX analyses from all the three heights resulted in similar spectra this means that the whole TiC layer is oxidized and not only the outermost nanometers. However, the quantification of C and
S. Canovic et al. / Micron 42 (2011) 808–818
811
Fig. 2. SAED patterns from (a) the substrate, (b) 1st, (c) 2nd and (d) 3rd ␣-Al2 O3 layer. (e) SAED pattern consisting of a superposition of patterns from the substrate, TiC and the 1st ␣-Al2 O3 layer. The diffraction indices of the patterns are shown in Fig. 3. (f) A bright field image of the substrate and the alumina layers. The image is rotated correctly relative to the diffraction patterns.
O is difficult with EDX due to the low yield and high absorption of low energy X-rays. Therefore, in order to confirm the oxidation of the TiC layer, a complementary method (EELS) was used. An EELS spectrum from the TiC layer is shown in Fig. 6. The results show presence of Ti, C and O edges, which confirms the results obtained from the EDX analyses. The two chemical analyses (EDX and EELS) show that C and O account for around 25 at.% each (Ti being 50 at.%). The exact levels of C and O are not possible to give, due to the possible carbon contamination and surface oxidation. However, from a comparison with the result from a pure TiC sample (not shown) it is clear that the TiC layer is oxidized and in fact is a Ti(Cx ,Oy ) layer (x and y ≈ 0.5). The oxidized TiC layers will generally be referred to as “TiC” in this article, except in the discussion where the oxidation is treated explicitly. Then the oxidized TiC layers will be referred to as TiCO.
3.4. Coating deposited on the a-surface 3.4.1. General microstructure Fig. 7 shows a cross-section overview of ␣a-sub -TiC/Al2 O3 showing the three ␣-Al2 O3 layers separated by the thin TiC layers. As in the case of ␣c-sub -TiC/Al2 O3 the ␣-Al2 O3 layers are single crystals growing on single crystals of TiC. The dark strain contrast extending through several layers that was present in Fig. 1(a) also appears in Fig. 7. Also in this case no pores were observed within the ␣-Al2 O3 layers, while a small number of interfacial pores were observed below the TiC layers, also in the substrate below the innermost TiC layer. As in the case of ␣c-sub -TiC/Al2 O3 the alumina (and TiC) layers grew with an even thickness resulting in straight and parallel interfaces, which resulted in epitaxial growth. Epitaxy and orientation relationships are discussed in more detail below.
-
131 0006
0330
0112
-0222
N,GD
0114
- -
--
-
0222
0000
0114
0330
0112
-
0330
-
0222
0114
- -
0112
-
a
b
0006
GD
0000
0006
0006
0006
0330 220
-
0112
-0114
0112 --
--
0222
-
311 0222
-- -
- -111 0114 311 GD 0222 0000 -220
-
0330
-0114 111 -Al2O3 TiC
-
0006
- -
-
0112
--
0330
131
c
Fig. 3. The diffraction indices of SAED patterns from Fig. 2. (a) Indices for the substrate, 1st and 2nd ␣-Al2 O3 layers with the zone axis [2 1¯ 1¯ 0]. GD = growth direction, N = substrate normal. (b) Indices for the 3rd ␣-Al2 O3 layer with the zone axis [2¯ 1 1 0]. (c) Indices for the substrate, TiC and the 1st ␣-Al2 O3 layer. The zone axis for TiC is [1¯ 1 2].
812
S. Canovic et al. / Micron 42 (2011) 808–818
Al
a Counts
1000
O 500
Ti 0
0
1500
1
2
3
E (keV )
4
5
6
Ti
b
500 Ti C 0 Fig. 4. HREM image of the inner part of ␣c-sub -TiC/Al2 O3 showing the substrate, TiC and 1st ␣-Al2 O3 layer. Both the substrate/TiC and TiC/␣-Al2 O3 interfaces are relatively even on the atomic level. EDX analysis was performed at three different heights in the TiC layer (marked with rectangles). The results are shown in Fig. 5.
3.4.2. Orientation relationships In order to determine the orientation relationships between the layers, electron diffraction was used. In Fig. 8(a–d) SAED patterns from the substrate and the three ␣-Al2 O3 layers are shown. The patterns in Fig. 8(b–d) are almost identical with only small crystal misorientations, which means that the three ␣-Al2 O3 layers have basically the same orientation. However, the pattern from the substrate shown in Fig. 8(a) is different. The indexing of the diffraction patterns (Fig. 9(a) and (b)) shows that the ␣-Al2 O3 layers grow along the m-surface normal (0 3¯ 3 0). Fig. 8(e) shows a diffraction pattern from the substrate, TiC and the 1st ␣-Al2 O3 layer and Fig. 9(c) shows its indices. It can be seen that (1¯ 2 1¯ 0) from the substrate is parallel to TiC(2 2 0) and ␣-Al2 O3 (0 3¯ 3 0) from the first alumina layer. This means that TiC(1 1 0) grows on the a-surface substrate. Onto this TiC surface an ␣-Al2 O3 layer, which grows
0
O
Cu/Ga
Counts
1000
Ti
Ga Al
1
2
3
E (keV )
4
5
6
Fig. 5. EDX spectra from (a) the 1st ␣-Al2 O3 layer and (b) the TiC layer. The three analyses from the TiC layer marked in Fig. 4 resulted in similar spectra.
along the m-axis, is deposited. Fig. 8(f) is a pattern from the 1st ␣-Al2 O3 , TiC and the 2nd ␣-Al2 O3 layer and the indices are shown in Fig. 9(d). It can be seen that (0 3¯ 3 0) from the first alumina layer is parallel to TiC(2 2 0) and (0 3¯ 3 0) from the second alumina layer. This means that TiC(1 1 0) grows on the ␣-Al2 O3 layer with morientation and onto this TiC surface another ␣-Al2 O3 layer with m-orientation is grown. In ␣a-sub -TiC/Al2 O3 epitaxial growth of TiC was followed by epitaxial growth of ␣-Al2 O3 etc. where every alumina and TiC layer was a single crystal with the same orientation with the following orientation relationship: (1¯ 2 1¯ 0)a-sub //(1 1 0)TiC //(0 1¯ 1 0)␣ //(1 1 0)TiC //(0 1¯ 1 0)␣ //(1 1 0)TiC //(0 1¯ 1 0)␣ [1 1¯ 0 1]a-sub //[1 1¯ 2]TiC //[2 1¯ 1¯ 0]␣ //[1 1¯ 2]TiC //[2 1¯ 1¯ 0]␣ //[1 1¯ 2]TiC //[2 1¯ 1¯ 0]␣
Fig. 6. EELS spectrum from the innermost TiC layer in ␣c-sub -TiC/Al2 O.
S. Canovic et al. / Micron 42 (2011) 808–818
813
same orientation relationship was observed in ␣c-sub -TiC/Al2 O3 . The difference is that the grains in this case are rotated 90◦ , i.e. another (macroscopic) coating growth direction is present. The steps present on the surface are probable nucleation sites for the ␣Al2 O3 layer. At these locations there are “vertical” TiC(1 1 1) faces present, see Fig. 10(b). It is likely that the ␣-Al2 O3 layer initially grows on the vertical TiC(1 1 1) face, to the right in Fig. 10, locally forming the same orientation relationship as in ␣c-sub -TiC/Al2 O3 . As the ␣-Al2 O3 layer continues to grow it will also cover TiC(2 2 0) faces forming the orientation relationship described above. However, it is possible that ␣-Al2 O3 can also nucleate and grow directly on TiC(2 2 0), especially in areas without TiC surface steps. According to this and the orientation relationship discussed above, epitaxial growth of ␣-Al2 O3 on TiC can take place in two ways: upwards in Fig. 10(a) where ␣-Al2 O3 (0 3¯ 3 0) grows epitaxi¯ ally on TiC(2 2 0) and to the right in Fig. 10(a) where ␣-Al2 O3 (0 0 0 6) ¯ Thus, at horizontal TiC(2 2 0) faces grows epitaxially on TiC(1 1¯ 1). Fig. 7. Bright field TEM micrograph of ␣a-sub -TiC/Al2 O3 . The individual ␣-Al2 O3 layers are separated by thin TiC layers.
Thus, TiC(1 1 0) grows on the single crystal substrate with aorientation, followed by growth of ␣-Al2 O3 single crystal with morientation. This is followed by growth of TiC(1 1 0) on the sapphire single crystal with m-orientation. 3.4.3. High-resolution electron microscopy and chemical analysis Fig. 10(a) shows an HREM micrograph of the substrate/TiC/␣Al2 O3 region. Both layers consist of well-crystallized grains. In this case the TiC/␣-Al2 O3 interface is not as even as in ␣c-sub -TiC/Al2 O3 . There are steps of ∼2 nm along the interface, such as that arrowed ¯ is paralin Fig. 10(a). In Fig. 9(c) it can be seen that TiC(1 1¯ 1) ¯ perpendicular to the growth direction. The lel to ␣-Al2 O3 (0 0 0 6)
only the former orientation relationship is present while at ver¯ faces as that shown in Fig. 10(a) both orientation tical TiC(1 1¯ 1) relationships are possible. This is also schematically illustrated in Fig. 10(b). As in the case of ␣c-sub -TiC/Al2 O3 chemical analysis was used in order to investigate the chemical composition of the TiC layer in the same way as was described earlier. The results were similar to the ones shown in Figs. 5 and 6, meaning that the whole TiC layer is oxidized. Thus, the TiC layer is in fact a TiCO layer also in this case. 3.5. Coating deposited on the r-surface 3.5.1. General microstructure Fig. 11(a) shows a cross-section overview of ␣r-sub -TiC/Al2 O3 with the three individual ␣-Al2 O3 layers separated by the thin TiC
Fig. 8. SAED patterns from (a) the substrate, (b) 1st, (c) 2nd and (d) 3rd ␣-Al2 O3 layer. (e) A diffraction pattern from the substrate, TiC and the 1st ␣-Al2 O3 layer. (f) A pattern from the 1st ␣-Al2 O3 , TiC and the 2nd ␣-Al2 O3 layer. The diffraction indices of the patterns are shown in Fig. 9. The growth direction (GD) is indicated in (d).
814
S. Canovic et al. / Micron 42 (2011) 808–818
Fig. 9. The diffraction indices of the SAED patterns from Fig. 8. Indices for (a) the substrate, (b) 1st, 2nd and 3rd ␣-Al2 O3 layers, (c) substrate, TiC and 1st ␣-Al2 O3 layer and (d) 1st ␣-Al2 O3 , TiC and 2nd ␣-Al2 O3 layer. The zone axes are [1¯ 0 1 1] and [2 1¯ 1¯ 0] for the substrate and the alumina layers, respectively, and [1 1¯ 2] for TiC.
layers. The first part of the multilayer coating is shown in Fig. 11(b) in higher magnification. As can be seen in Fig. 11 the first alumina layer grows with an uneven thickness on the first TiC layer and accordingly the following TiC layer is not parallel to the innermost TiC layer. At some rare locations the second TiC layer was parallel to the first one, see, e.g. the area within the circle in Fig. 11(b). At these locations epitaxial columns of grains, extending throughout the coating thickness, were observed. Epitaxy and orientation relationships are discussed in more detail below. Generally, more dislocations and defects were present within the ␣-Al2 O3 layers. For example, the dark fringes (marked with white arrows in Fig. 11(b)) have steps due to local errors present in the crystal during growth, which indicates a more disturbed growth of the coating on the r-surface. 3.5.2. Orientation relationships, high-resolution electron microscopy and chemical analysis Generally no epitaxy is present in the coating; only some few epitaxial columns were found. Fig. 12(a–d) shows SAED patterns from the substrate and the three ␣-Al2 O3 layers from such an epitaxial column. The corresponding diffraction indices are shown in Fig. 12(e). It can be seen that the substrate and all the ␣-Al2 O3 layers have basically the same orientation; they are only slightly
misoriented. Thus, the alumina grains grow with the r-surface ¯ parallel to the growth direction. In order to deternormal (0 1¯ 1 2) mine the orientation relationship between ␣-Al2 O3 and TiC FFTs from HREM images were used. Fig. 13 shows an HREM micrograph of the substrate/TiC/␣-Al2 O3 region. Both layers consist of wellcrystallized grains and both the substrate/TiC and the TiC/␣-Al2 O3 interfaces are quite even on the atomic level. FFT patterns from the TiC and the ␣-Al2 O3 layer from Fig. 13 are shown in Fig. 14. ␣-Al2 O3 could be indexed as in Fig. 12, see Fig. 14(a). The indexing of the TiC layer is shown in Fig. 14(b). Consider especially the spots labeled a, b and c in the figure. The interplanar distances corresponding to the spots a and b fit well with (2 0 0)a and (1 1 1)b for a face centered cubic structure (TiC). However, the angle between these two spots is ∼75◦ instead of ∼55◦ , which it should be for an fcc structure. The interplanar distance corresponding to spot c does therefore not fit with (3 1 1), i.e. the sum of a and b, for the fcc structure. Nevertheless, chemical analyses from the TiC layer showed similar results as for the TiC layers in ␣c-sub -TiC/Al2 O3 and ␣a-sub -TiC/Al2 O3 , meaning that the whole TiC layer is oxidized to TiCO also in this case with around 25 at.% C and 25 at.% O. It has previously been shown that several different titanium oxides (e.g. Ti3 O5 , Ti2 O3 , TiO2 ) can be formed by oxidizing a TiC surface in a CVD reactor (Chatfield et al., 1989; Lhermitte-Sebire
S. Canovic et al. / Micron 42 (2011) 808–818
815
Fig. 10. (a) HREM image of the inner part of ␣a-sub -TiC/Al2 O3 showing the substrate, TiC and 1st ␣-Al2 O3 layer. There is an approximately 2 nm large step (arrowed) at the TiC/␣-Al2 O3 interface. (b) A schematic drawing illustrating how ␣-Al2 O3 grows on TiC. The arrows indicate the plane (index) normals.
et al., 1986). In the present case a short oxidation stage was applied, which should favour a phase with low oxygen content. Therefore, the titanium oxides with high oxidation number should be less probable. Hence, Ti(II)O is the most probable titanium oxide to form. Most of the titanium oxide candidates could also be excluded based on matching of their interplanar distances to the FFT results. However, one specific phase of TiO matches the FFT results, namely the low temperature form of the phase TiO. The unit cell is monoclinic (space group A2/m) with a = 5.855, b = 9.340, c = 4.142 A˚ and = 107.32◦ (Watanabe et al., 1967). The structure is similar to that of NaCl, but has an ordered array of vacant lattice sites. Half of the titanium and half of the oxygen atoms are missing alternately in every third (1 1 0) plane (Watanabe et al., 1967). This phase is stable
in the composition range TiO0.9 –TiO1.1 and at temperatures up to 990 ◦ C (Watanabe et al., 1967). Thus, the TiCO phase formed in the present case seems to be a modified TiO structure, i.e. monoclinic TiO with a relatively large carbon content. This should be possible due to the high density of vacancies present in the TiO structure. Since the observed phase also contains C the temperature stability and the composition range may be influenced compared to pure TiO. The monoclinic TiCO phase matches the FFT results even better if the lattice parameter b is slightly increased from the value for pure TiO. The presence of C in the structure could explain why the unit cell is somewhat distorted compared to pure TiO. From the indexing in Fig. 14 it can be seen that TiCO(3¯ 4¯ 0)m is ¯ In the schematic parallel to the coating growth direction, (0 1¯ 1 2).
Fig. 11. (a) Bright field TEM micrograph of ␣r-sub -TiC/Al2 O3 . (b) An area of the inner part of the coating. The TiC layers are marked with arrows. A region with epitaxy is circled in (b). Several pores are present below the TiC layers (some of them are indicated with black block arrows in (b)). Also several stacking faults are present (some of them indicated with white block arrows).
816
S. Canovic et al. / Micron 42 (2011) 808–818
Fig. 12. SAED patterns from (a) the substrate, (b) 1st, (c) 2nd and (d) 3rd ␣-Al2 O3 layer. (e) The diffraction indices of the patterns. The zone axis is [2 2¯ 0 1].
drawing in Fig. 15 it can be seen that (3 4 0)m , which intersects the monoclinic unit cell at a/3 and b/4, deviates by only a few degrees from (1 1 0)m . This leads to the following orientation relationship:
Generally the coating grows unevenly with random orientation when deposited onto the ␣-Al2 O3 single crystal substrate with
¯ r-sub ∼//(1¯ 1¯ 0)TiCO ∼//(0 1¯ 1 2) ¯ ␣ ∼//(1¯ 1¯ 0)TiCO ∼//(0 1¯ 1 2) ¯ ␣ ∼//(1¯ 1¯ 0)TiCO ∼//(0 1¯ 1 2) ¯ ␣ (0 1¯ 1 2) ¯ TiCO ∼/[2 2¯ 0 1]␣ ∼//[1 1¯ 2] ¯ TiCO ∼//[2 2¯ 0 1]␣ ∼//[1 1¯ 2] ¯ TiCO ∼//[2 2¯ 0 1]␣ [2 2¯ 0 1]r-sub ∼//[1 1¯ 2] Fig. 15 also illustrates the relation between the cubic parent cell and the monoclinic unit cell (Watanabe et al., 1967). For instance it can be seen that (1 1 0)m //(1 0 0)c , i.e. on the r-surface TiCO grows as the monoclinic equivalent to (1 0 0)c . The orientation relationship described here is not similar to the one observed in a previous paper (Canovic et al., 2010).
r-orientation. However, at some rare locations epitaxial columns ¯ forms on the of grains are observed where monoclinic TiCO(1¯ 10) ␣-Al2 O3 single crystal substrate. The originally deposited TiC layer had a cubic structure. After the oxidation stage the cubic TiC incorporated oxygen and transformed to a monoclinic structure. It is well known that TiC has an fcc structure (Pierson, 1996), while TiO has a monoclinic structure (Watanabe et al., 1967). It is likely that the structure of the TiCO formed during oxidation is strongly dependent on its C/O ratio. A high C/O ratio increases the probability for formation of a TiC-like structure (fcc), while a low C/O ratio probably favours formation of a TiO-like structure (monoclinic). It is possible that the oxidation process progresses faster in some TiC layer regions containing, e.g. defects and/or specific grain orientations (i.e. the r-surface sample), which would lead to a higher amount of oxygen in some regions. Thus, in some cases it should be possible to have a local transformation to a monoclinic phase in some regions, while the structure remains fcc in others.
4. Concluding discussion
Fig. 13. HREM image of the substrate, TiC and 1st ␣-Al2 O3 layer. Both the substrate/TiC and TiC/␣-Al2 O3 interfaces are relatively even on the atomic level.
Multilayers of TiC/␣-Al2 O3 consisting of three alumina layers separated by thin oxidized TiC layers, have been deposited onto c-, a- and r-surfaces of single crystals of ␣-Al2 O3 . In all three cases chemical analyses showed that the whole TiC layer is oxidized and is in fact a TiCO layer. When deposited on the c- and asurface substrates the alumina layers could grow evenly resulting in growth of single crystal layers of TiCO and alumina. However, when deposited on the r-surface the alumina layers exhibited uneven, faceted growth. A small number of pores were observed at the interfaces below the TiCO layers, probably due to the initial TiC
S. Canovic et al. / Micron 42 (2011) 808–818
817
Fig. 14. FFTs from (a) the ␣-Al2 O3 layer and (b) the TiC layer in Fig. 13. ␣-Al2 O3 could be indexed as in Fig. 12, TiC is indexed as a monoclinic structure with the zone axis ¯ The spots labeled a, b and c are discussed in the text. [4 3¯ 5].
deposition process step. The presence of pores has been observed in earlier studies of TiX/alumina multilayer coatings (Canovic et al., 2010; Halvarsson et al., 2006; Lindulf et al., 1994; Skogsmo et al., 1992; Vuorinen and Karlsson, 1992). In ␣c-sub -TiC/Al2 O3 and ␣a-sub -TiC/Al2 O3 epitaxy was frequently found. However, on the r-surface the alumina layers did not grow with an even thickness, which resulted in non-parallel TiC layers. At some rare locations straight and parallel layers were found also in ␣r-sub -TiC/Al2 O3 . At these locations epitaxial columns of grains were observed. Schematics summarizing the orientation relationships, when present, obtained in the different coatings are shown in Fig. 16. In ␣c-sub -TiC/Al2 O3 epitaxial growth of TiC was followed by epitaxial growth of ␣-Al2 O3 etc. where every alumina layer is a single crystal with the same orientation as the substrate (except
Fig. 15. Schematic drawing of the TiO structure viewed in the [0 0 1] direction at z = 0. The monoclinic unit cell parameters a, b and are indicated. The pseudo-cubic unit cell (ac and bc ) is also indicated. One plane for (3 4 0)m and one for (1 1 0)m (corresponding to (1 0 0)c ) are given.
the outermost ␣-Al2 O3 layer that was rotated 180◦ around the c-axis, which could have occurred in any of the other alumina layers). The TiC layers were also single crystals that were related to the alumina layers according to the orientation relationship shown in Fig. 16 (left). This is the same orientation relationship as was observed in a previous paper (Canovic et al., 2010). Thus, first TiC(1 1 1) grows on ␣-Al2 O3 (0 0 0 1). Then, the TiC layer is oxidized during the oxidation step and forms fcc TiCO followed by growth of ␣-Al2 O3 (0 0 0 1) on TiCO(1 1 1). In both cases close-packed planes grow on close-packed planes, which is frequently observed in CVD multilayer coatings (Canovic et al., 2007, 2010; Halvarsson et al., 2006; Halvarsson and Vuorinen, 1997). In conclusion, all the layers in ␣c-sub -TiC/Al2 O3 are single crystals that grow epitaxially. Similarly, in ␣a-sub -TiC/Al2 O3 epitaxy was frequently present and the orientation relationship shown in Fig. 16 (center) was found. Thus, TiC(1 1 0) grows on the single crystal substrate with a-orientation. It is oxidized to form fcc TiCO and followed by growth of ␣-Al2 O3 single crystal with m-orientation. This is followed by growth of TiC(1 1 0) on the ␣-Al2 O3 m-surface. Then the TiC layer is oxidized and the sequence is repeated for the other ␣-Al2 O3 and TiC layers. On the r-surface, which is not close-packed, more irregular coating growth takes place. In this case, the alumina layers do not grow evenly, i.e. the top surfaces of the first alumina layer are generally not the same as the top surface of the substrate. At some rare locations, where they are the same epitaxy is observed. In these cases the orientation relationship shown in Fig. 16 (right) was observed. In ␣r-sub -TiC/Al2 O3 the oxidation step resulted in a transformation of the fcc TiC to a monoclinic TiCO phase. From FFT and chemical analyses it was found that the originally deposited TiC structure (fcc) incorporates oxygen and transforms to monoclinic TiCO. It is possible that the oxidation process progresses faster in some TiC layer regions containing, e.g. defects and/or specific grain orientations, which would lead to a higher amount of oxygen in some regions. Thus, in some cases it should be possible to have a local transformation to a monoclinic phase in some regions, while the structure remains fcc in others. In the present work, it is shown that the TiC layers can be affected differently when oxidized. This is important since the TiC surface affects the subsequent deposition of Al2 O3 . However, in this work ␣-Al2 O3 formed in all cases, even when the TiC layer was oxidized to a TiCO phase with a monoclinic structure. Thus, during a short oxidation step the originally deposited fcc TiC layer can incorporate oxygen and form an fcc TiCO phase (on the c- and a-surfaces), or monoclinic TiCO (on the r-surface). An extended oxidation step
818
S. Canovic et al. / Micron 42 (2011) 808–818
c (180° rot)
m
111
110
TiCO (fcc)
c 111
c
TiCO (fcc)
m TiCO (fcc)
c 111
r
110
c-sub
110
TiCO (m.c.)
r TiCO (fcc)
110
TiCO (m.c.)
r
m TiCO (fcc)
110
TiCO (fcc)
a
a-sub
110 r
TiCO (m.c.) r-sub
Fig. 16. Schematics showing the coating sequences for the three samples. (m.c.) = (monoclinic).
may lead to formation of different titanium oxides (e.g. TiO, TiO2 , Ti2 O3 , Ti3 O5 ). Excessive oxidation of the TiC layer prior to or during initial alumina layer deposition may result in considerable expansion of the TiC layer. Therefore, it is probably important to have low oxidizing conditions for the TiC layer during the oxidation step and the initial alumina layer deposition. In general, the TiC oxidation step is a sensitive and important part of the CVD process of these types of multilayer coatings. Acknowledgement Financial support was provided by the Swedish National Graduate School in Materials Science (NFSM). References Canovic, S., et al., 2007. TEM and DFT investigation of CVD TiN/-Al2 O3 multilayer coatings. Surf. Coat. Technol. 202, 522–531. Canovic, S., et al., 2010. CVD TiC/alumina and TiN/alumina coatings grown on sapphire single crystals. Int. J. Refract. Met. Hard Mater. 28, 163–173. Chatfield, C., Lindström, J.N., Sjöstrand, M.E., 1989. Microstructure of CVD Al2 O3 . J. Phys. C5, 377–387. Colombier, C., Lux, B., 1989. Formation of mixed TiC/Al2 O3 layers and ␣- and -Al2 O3 on cemented carbides by chemical vapour deposition. J. Mater. Sci. 24, 462–470. Dragoo, A.L., Diamond, J.J., 1967. Transitions in vapor-deposited alumina from 300◦ to 1200 ◦ C. J. Am. Chem. Soc. 50 (11), 568–574. Funk, R., et al., 1976. Coating of cemented carbide cutting tools with alumina by chemical vapor deposition. J. Electrochem. Soc. 123 (2), 285–289. Halvarsson, M., Nordén, H., Vuorinen, S., 1993a. Microstructural investigation of CVD ␣-Al2 O3 /-Al2 O3 multilayer coatings. Surf. Coat. Technol. 61, 177–181. Halvarsson, M., Trancik, J.E., Ruppi, S., 2006. The microstructure of CVD -Al2 O3 multilayers separated by thin intermediate TiN or TiC layers. Int. J. Refract. Met. Hard Mater. 24, 32–38. Halvarsson, M., Vuorinen, S., Nordén, H., 1993b. Microstructural investigation of -modification layers in CVD ␣-Al2 O3 /-Al2 O3 multilayer coatings. Mater. Res. Soc. Symp. Proc. 314, 83–88. Halvarsson, M., Vuorinen, S., 1995. The influence of the nucleation surface on the growth of CVD ␣-Al2 O3 and -Al2 O3 . Surf. Coat. Technol. 76/77, 287–296. Halvarsson, M., Vuorinen, S., 1996. Epitaxy in multilayer coatings of -Al2 O3 . Surf. Coat. Technol. 80, 80–88.
Halvarsson, M., Vuorinen, S., 1997. Microstructural investigation of CVD TiN/-Al2 O3 multilayer coatings on cemented carbides. Int. J. Refract. Met. Hard Mater. 15, 169–178. Kronberg, M.L., 1957. Plastic deformation of single crystals of sapphire: basal slip and twinning. Acta Metall. 5, 507–524. Larsson, A., Ruppi, S., 2001. Microstructure and properties of CVD ␥-Al2 O3 coatings. Int. J. Refract. Met. Hard Mater. 19 (4–6), 515–522. Levin, I., Brandon, D., 1998. Metastable alumina polymorphs: crystal structures and transition sequences. J. Am. Ceram. Soc. 81 (8), 1995–2012. Lhermitte-Sebire, I., et al., 1986. Microanalysis of the Al2 O3 –TiC interface in aluminacoated cemented carbides by Auger electron and glow discharge emission spectroscopies. Thin Solid Films 138, 209–220. Lindulf, N., et al., 1994. Microstructural investigation of the -Al2 O3 → ␣-Al2 O3 transformation in multilayer coatings of chemically vapour deposited -Al2 O3 . Thin Solid Films 253, 311–317. Lux, B., et al., 1986. Preparation of alumina coatings by chemical vapour deposition. Thin Solid Films 138 (1), 49–64. Okumiya, M., et al., 1971. The formation of - and -Al2 O3 from the dehydration of thodite 5Al2 O3 × H2 O. Bull. Chem. Soc. Jpn. 44 (2), 418–423. Pierson, H.O., 1996. Handbook of Refractory Carbides and Nitrides. Noyes Publications, New Jersey, USA. Ruppi, S., 1992. Multi-oxide coated carbide body and the method of producing the same. U.S. Patent 5,137,774. Ruppi, S., Larsson, A., 2000. Chemical vapour deposition of -Al2 O3 . Thin Solid Films 388 (1/2), 50–61. Ruppi, S., 2005. Deposition, microstructure and properties of texturecontrolled CVD ␣-Al2 O3 coatings. Int. J. Refract. Met. Hard Mater. 23, 306– 316. Ruppi, S., 2008. Enhanced performance of ␣-Al2 O3 coatings by control of crystal orientation. Surf. Coat. Technol. 202 (17), 4257–4269. Sjöstrand, M.E., 1979. Deposition of wear resistant TiN on cemented carbides using mixtures of NH3 /N2 and TiCl4 /H2 . Proc. CVD, 452–453. Skogsmo, J., Halvarsson, M., Vuorinen, S., 1992. Phase transformation of CVD Al2 O3 . Micron Microsc. Acta 23 (1/2), 219–220. Stumpf, H.C., et al., 1950. Thermal transformation of aluminas and alumina hydrates. Ind. Eng. Chem. 42 (7), 1398–1403. Vuorinen, S., Karlsson, L., 1992. Phase transformation in chemical vapour-deposited -alumina. Thin Solid Films 214, 132–143. Vuorinen, S., Skogsmo, J., 1988. In: Sudarshan, T., Bhat, D. (Eds.), Microstructural Characterization of Wear Resistant Coatings. Surface Modification Technologies. TMS, Phoenix, AZ, pp. 143–168. Vuorinen, S., Skogsmo, J., 1990. Characterization of ␣-Al2 O3 , -Al2 O3 and ␣ multioxide coatings on cemented carbide. Thin Solid Films 193–194, 536–546. Watanabe, D., et al., 1967. The ordered structure of TiO. Acta Cryst., 23.