Influence of Process Conditions on the Growth and Texture of CVD Alpha-Alumina

Abstract

In the present study, the influence of experimental variables on the growth characteristics and texture development in chemically vapour deposited α-Al₂O₃ coatings was investigated. The emphasis was on the effects of H₂S. The α-Al₂O₃ layers were deposited from the AlCl₃-H₂-CO₂-HCl-H₂S precursor system onto intermediate Ti(C,N) layers. The substrate was cemented carbide. The coatings were characterized with respect to microstructure and texture using X-ray diffraction, scanning electron microscopy and electron back-scattering diffraction. The observations were unpredictable, suggesting that in addition to the growth rate, H₂S strongly affected the texture of the α-Al₂O₃ coatings. The uncatalyzed α-Al₂O₃ coatings (H₂S = 0) developed (11$\overline{2}$0) texture. With the increment of H₂S concentration the texture changed from (11$\overline{2}$0) texture through (10$\overline{1}$0), (10$\overline{1}$2) and (10$\overline{1}$4) textures to (0001) texture. Growth regimes for (11$\overline{2}$0), (10$\overline{1}$0), (10$\overline{1}$2), (10$\overline{1}$4) and (0001) textures were depicted as a function of the H₂S/CO₂ ratio. Examples of grain-boundary and microstructural manipulation of CVD α-Al₂O₃ layers by applying texture control were presented.

1. Introduction

Alumina is uniquely suited for metal cutting tools due to its chemical inertness and high hot-hardness at the temperatures typically reached in these applications [1,2,3,4]. Ti(C,N)-Al₂O₃ coated cemented carbide tools are used, for example, in turning and milling of steels and cast irons. The Al₂O₃ coatings are typically obtained using chemical vapour deposition (CVD). CVD has been used for about 50 years for the industrial deposition of wear resistant coatings and is still today the only technique that can economically produce thick and uniform Al₂O₃ coatings on cemented carbide tools.

CVD of Al₂O₃ is usually performed from the AlCl₃–H₂–CO₂ system based on hydrolysis of AlCl₃ according the following generally accepted reactions:

2AlCl₃(g)+3H₂O(g) → Al₂O₃(s)+6HCl(g)

(1)

The amount of water needed for the reaction is formed in situ in the deposition chamber as a result of the water gas shift reaction:

CO₂ (g) + H₂(g) → CO(g) + H₂O(g)

(2)

The overall reaction is thus:

2AlCl₃(g) + 3CO₂(g) + 3H₂(g) → Al₂O₃ (s) + 3CO(g)+6HCl(g)

(3)

The reaction is complex and involves both homogenous gas-phase reactions and heterogeneous surface reactions [5,6,7,8,9,10,11,12,13,14,15]. The rate determining reaction is the slow homogeneous water gas shift reaction (2).

The heterogeneous surface reactions that may contribute to the growth of Al₂O₃ are probably direct surface reactions between aluminum halides and an oxygen donor. The relative contributions of homogenous and heterogeneous reactions to the growth of Al₂O₃ depend strongly on the free gas volume to surface ratio adjacent to the surface to be coated [5,6]. Long chains of reactions starting with the decomposition of AlCl₃ (or Al₂Cl₆) in the presence of H₂ (H₂S = 0) have been suggested, together with different intermediate transition states such as AlCl₂(OH) and AlClO [13,14,15]. According to the simulations by Tan et al. [15], the AlClO molecule could be a major intermediate Al-species catalyzing the slow water gas shift reaction at short residence times. A recent thermodynamic study [16] supports this hypothesis even though the authors emphasized that AlCl₃ and AlCl₂OH are thermodynamically more stable under CVD conditions.

The early Al₂O₃ coatings deposited from the system AlCl₃-H₂-CO₂ exhibited a very low deposition rate and often consisted of a mixture of α-Al₂O₃ and κ-Al₂O₃. Typically, the first-generation alumina coatings were composed of coarse grained α-Al₂O₃ embedded in a matrix of fine-grained κ-Al₂O₃. Besides the fact that it was not possible to control the phase content, the overall growth rate was very low. Further, the early Al₂O₃ coatings also suffered from a pronounced “dog-bone” effect, i.e., the deposition rate was higher at the cutting edges than on the other surfaces of the insert. This phenomenon is commonly accepted to be a result from a higher gas volume relative to the surface area around the edges of the inserts. The introduction of H₂S doping [17] enhanced not only the over-all deposition rate but also suppressed the dog-bone effect. The use of H₂S as a catalyst facilitated thicker alumina layers with a good coating uniformity. It is emphasized that the α-Al₂O₃ found in these kinds of coatings as well as in early commercial α-Al₂O₃ coatings was mainly a result of the κ → α phase transformation [18].

The introduction of specific nucleation and oxidation steps enabled H₂S-catalyzed deposition of Al₂O₃ coatings being composed of either as-grown κ-Al₂O₃ or α-Al₂O₃ phases [19,20]. Following this development, it was established that as-grown α-Al₂O₃ exhibited a superior wear resistance to κ-Al₂O₃ and earlier α-Al₂O₃ layers. Empowered by the nucleation control, it was soon discovered that not only the phase, but also the texture of the α-Al₂O₃ coatings could be determined by careful control of the process parameters, together with a proper nucleation step [20,21,22,23,24]. Consequently, it became possible to deposit α-Al₂O₃ coatings with, e.g., (10$\overline{1}$2), (10$\overline{1}$4), (0001) growth textures. Recently, the (0001) textured α-Al₂O₃ coatings were found to show a substantially enhanced cutting performance, in some applications by several hundreds of per cent over the other α-Al₂O₃ textures [20,21,23,24]. Introduction of the texture-control on an industrial scale [23,24], facilitating the commercial production of (0001) textured α-Al₂O₃ coatings on medium temperature MTCVD Ti(C,N) intermediate layers has thus been one of the most important recent advances in CVD tool coatings. These α-Al₂O₃ coatings have been considered state-of-the-art for more than ten years [25].

Although H₂S has been used for about 30 years in industrial CVD processes as a catalyst, the underlying mechanisms have not yet been fully understood. The earlier experimental as well as thermo-dynamic studies have mainly dealt with deposition of α-Al₂O₃ from the AlCl₃-H₂-CO₂ and AlCl₃–H₂–CO₂–HCl systems [5,6,7,8,9,10,11,12,13,14,15] and there are only few studies which have tried to understand the role of H₂S in the CVD process of α-Al₂O₃ [26,27]. According to these studies, H₂S shifts the rate determining step from the gas phase to the surface. It has also been assumed that H₂S acts as a true catalyst, dissociating and reforming via adsorption desorption reactions on the surface, these reactions being much faster than H₂O adsorption explaining the much higher growth rates obtained in the presence of H₂S. Slager and Amberg [28] suggested steps consisting of H₂S adsorption on Al⁺- ions forming an Al–S surface bond H-bonding to neighboring O and OH-species followed by its decomposition with water formation. According to this mechanism, both S and O adsorption takes place and the presence of OH-groups is needed. Consequently, sulfur adatoms must be removed from the surface and this step could be rate limiting for the suggested mechanism. Recent work has suggested that H₂S acts as true catalyst for the surface reaction [16] and that H₂S will not solve in α-Al₂O₃ during deposition [29].

Even though advances in the nucleation and deposition technology have enabled the controlled deposition of strongly textured α-Al₂O₃ coatings, there are only a limited number of reports on deposition of such textured CVD α-Al₂O₃ layers [12,20,21]. Most importantly, there are no studies trying to elucidate the effects of H₂S on the texture development in CVD α-Al₂O₃ coatings. This work addresses the above-described situation and elucidates the influence of process conditions, including H₂S, on CVD of α-Al₂O₃.

2. Experimental

2.1. CVD Process

The α-Al₂O₃ coatings were deposited from the AlCl₃–H₂–CO₂–H₂S–HCl system at 1000 °C. The coatings were deposited in a computer-controlled, hot-wall CVD reactor. AlCl₃ was generated in a separate generator by letting HCl pass over and react with aluminum chips (purity 99.999 %) at a temperature of about 380 °C. Approximately 22% of the total flow (H₂, HCl) passed through the AlCl₃ generator. The coatings were deposited as multilayers and single layers. The multilayer coatings were mainly used to investigate the growth rate as a function of H₂S concentration at several levels of CO₂, while the single layers were deposited to investigate the texture development in the α-Al₂O₃ layers.

The α-Al₂O₃ coatings were deposited onto (MTCVD) Ti(C,N) layers. The MTCVD Ti(C,N) layers were deposited onto cemented carbide substrates composed of 93.3 wt.% WC, 6.2 wt.% Co and 0.5 wt.% (Ta,Nb)C at a temperature of about 860 °C from the CH₃CN-TiCl₄-N₂-H₂ system. Before depositing the MTCVD Ti(C,N) layer, a thin TiN layer was first applied on the cemented carbide substrate from the TiCl₄-N₂-H₂ precursor mixture. The MTCVD Ti(C,N) layer was deposited to a thickness of ~3 μm onto the TiN layer. After this, the process temperature was raised to 1000 °C and a high temperature Ti(C,N) layer was deposited on top of the MTCVD layer from the TiCl₄-CH₄-N₂-H₂ system. Onto this layer, an oxidizing deposition step was applied from the CO₂-CO-H₂ system to ensure the nucleation of α-Al₂O₃.

2.2. α-Al₂O₃ Coatings

2.2.1. Multilayer Coatings

The multilayer coatings were deposited to examine the effects of CO₂ and H₂S on the growth of α-Al₂O₃. The α-Al₂O₃ multilayers were composed of 8 individual layers of α-Al₂O₃, each layer being deposited at different process conditions. The individual α-Al₂O₃ layers of the of the multilayer coating were separated by thin Ti(C,N) layers combined with an oxidizing nucleation step to stabilize α-Al₂O₃. The deposition time for each single layer in a multilayer coating was 65 min. H₂S was added into the process after a 5 min delay and the procedures were identical for all layers. The effect of H₂S doping (0.1–1.6 vol %) was studied at three different CO₂ levels of 3, 4 and 6 vol % (

Table 1. Deposition parameters for multilayer experiments.

H2S (vol %)	CO2 (vol %)	AlCl3 (vol %)	HCl (vol %)	H2 (vol %)	P (mbar)
0.1–1.6	3	2.5	1	Balance	75
0.1–1.6	4	2.5	1	Balance	75
0.1–1.6	6	2.5	1	Balance	75

). By using the multi-layer approach, several α-Al₂O₃ coatings with different process parameters could be deposited in the same coating run and then inspected with respect to growth rate and, to a limited degree, with respect to the texture using EBSD. This approach has been previously used to investigate the deposition characteristics of κ-Al₂O₃ [10].

2.2.2. Coatings for Texture Analysis

Single α-Al₂O₃ layers were deposited to elucidate the texture development. The process parameters are summarized in

Table 2. Deposition parameters for texture experiments.

	Texture Experiments	Reference Coating
H2S (vol %)	0.02–1.6	0
CO2 (vol %)	3.0–6.5	3.25
H2S/CO2 ratio	0.004-0.40	0
AlCl3 (vol %)	2.5	2.5
HCl (vol %)	1.0	1.0
H2 (vol %)	Balance	Balance
P (mbar)	75	75
Time (min)	480	1200

. H₂S concentration was varied from about 0.02 vol % to about 1.6 vol %. CO₂ was varied from about 3 vol % to about 6.5 vol %. Consequently, H₂S/CO₂ ratios from ~0.004 to ~0.40 were applied. The α-Al₂O₃ coatings presented in Table 2 were deposited on MTCVD Ti(C,N) layers using identical bonding and oxidation steps.

The first 30 min of all the α-Al₂O₃ coatings was performed uncatalyzed (H₂S = 0). After this period, H₂S was introduced into the reactor at controlled H₂S/CO₂ ratios. The AlCl₃/CO₂ ratio was kept constant (as well as other process parameters) during the depositions and the deposition time was 480 min, resulting in the formation of 6–10 μm thick α-Al₂O₃ layers. The reference layer of α-Al₂O₃ was deposited without H₂S (H₂S = 0). In this case, due to the low deposition rate of non-catalyzed alumina [10], the deposition time of 1200 min was applied to achieve a sufficient thickness, Table 2. Coatings deposited at H₂S = 0.03 vol % (CO₂ = 6.25 vol %), H₂S = 0.08 vol % (CO₂ = 6.25 vol %), H₂S = 0.2 vol % (CO₂ = 3.25 vol %), H₂S = 0.7 vol % (CO₂ = 6.25 vol %) and at H₂S = 0.7 vol % (CO₂ = 3.25 vol %) referred to as Coatings a, b, c, d and e were selected to be analysed in more detail by XRD and SEM.

2.2.3. Manipulation of α-Al₂O₃ Microstructures

To reduce the process time a thin MTCVD layer (~ 3 µm) was used in the experiments described in Section 2.2.2. However, the MTCVD Ti(C,N) layer has an important function as a diffusion barrier for W and Co from the cemented carbide substrate to the α-Al₂O₃-Ti(C,N) interface. These elements may disturb the initial growth of α-Al₂O₃ [1,30]. The problem is well-known in cutting tool industry and confirmed in a recent TEM study [30]. Further, the texture of the MTCVD Ti(C,N) layer, which is more profound in thicker layers, may also affect nucleation and growth of the alumina layer. Consequently, additional α-Al₂O₃ layers were deposited on thicker and textured MTCVD Ti(C,N) layers using state-of-the-art nucleation and bonding layers. The aim of these experiments was to validate the obtained results and explore further possibilities of texture control.

The similar CVD process and the procedure as described in Section 2.1 was applied. The MTCVD Ti(C,N) layers were, in this case, deposited to a thickness of ~8 μm. Additionally, between the MTCVD and the high-temperature CVD Ti(C,N) layers, a thin N-rich Ti(C,N) layer was applied at 1000 °C. The MTCVD Ti(C,N) coating exhibited (211) texture. Three different α-Al₂O₃ layers referred to as Coatings f, g and h were deposited on these Ti(C,N) layers: Coating f (

Table 3. Deposition conditions for Coating f.

	Coating A
H2S/CO2 ratio	0.40
AlCl3 (vol %)	2.65
HCl (vol %)	1.1
H2 (vol %)	Balance
P (mbar)	75
Time (min)	480

) was composed of an α-Al₂O₃ layer deposited at a constant H₂S/CO₂ ratio of 0.4 (after a period of 30 min at H₂S = 0) to a thickness of about 8 µm (deposition time 480 min). Coating g comprised an α-Al₂O₃ layer deposited using the H₂S/CO₂ ratios 0.40 and <0.01 applied in sequences favoring the growth along the c-axis and perpendicular to it, i.e., along the <0001> and <11$\overline{2}$0> directions, respectively. Deposition was started by applying (0001) process conditions (H₂S/CO₂ = 0.40) first (after a period of 5 min at H₂S = 0), followed by (11$\overline{2}$0) conditions (H₂S/CO₂ < 0.01). The combination of (0001) and (11$\overline{2}$0) sequences was repeated 7 times, the uppermost sequence being thus deposited using (11$\overline{2}$0) conditions. The resulting coating was consequently composed of 14 sub-layers: 7 layers with (0001) conditions and 7 layers with (11$\overline{2}$0) conditions. The deposition time for the first (0001) sequence was 45 min and 30 min for the following sequences. H₂S was added after 5 min delay in each step. The process parameters for α-Al₂O₃ were otherwise the same as those used for Coating f. The α-Al₂O₃ layer of Coating h (

Table 4. Deposition conditions for Coating h.

	First Part	Second Part
H2S/CO2 ratio	0.40	<0.01
AlCl3 (vol %)	2.65	2.65
HCl (vol %)	1.5	1.5
H2 (vol %)	Balance	Balance
P (mbar)	85	55
Time (min)	420	60

) was deposited using H₂S/CO₂ = 0.4 (after a period of 30 min at H₂S = 0) to a thickness of about 7 µm (deposition time 420 min) and after this period, the process conditions were switched over to H₂S/CO₂ < 0.01 and applied for 60 min.

2.3. Analysis

The thicknesses of the individual α-Al₂O₃ layers in multilayers were determined with a scanning electron microscope (SEM) run preferably in backscattered electron imaging mode. The Ti(C,N) interlayers appeared as white lines in the darker alumina matrix (Figure 1a). The coating thickness was measured in the middle of the upper side of the insert (average of five measurements). The general morphology of the multilayers as well as the single layers was studied using secondary electrons. The Scanning Electron Microscopy (SEM) and Electron Back-Scattering Diffraction (EBSD) investigations were carried out using LEO Ultra 55 FEGSEM (equipped with Chanel 5 EBSD software) operating at 15–20 kV with a working distance of 15 mm.

The texture of the single layers was studied by means of X-ray diffraction (XRD) using a Bruker D8 Advance X-ray diffractometer with θ–2θ setup and CuK_α radiation. The 2θ range used was from 20° to 70°. Texture coefficients (TC) for the α-Al₂O₃ coatings were calculated from the Harris equation [31]:

$$TC(hkil)=\frac{I(hkil)}{I_0(hkil)}\{\frac{1}{n}{{\displaystyle \sum \frac{I(hkil)}{I_0(hkil)}}\}}^{-1}$$

(4)

where I(hkil) are the measured intensities of the (hkil) reflection, I₀(hkil) powder diffraction intensities according to the ICCD’s PDF card no 46-1212 and n is the number of reflections used in the calculation. It is noted that Miller-Bravais (hkil) indices for a hexagonal unit cell were applied. The following (hkil) reflections were used: (10$\overline{1}$2), (11$\overline{2}$0), (10$\overline{1}$4), (0006), (11$\overline{2}$3), (11$\overline{2}$6) and (30$\overline{3}$0). In this case, 7 reflections were used and therefore, the maximum value of $TC(hkil)$ was 7 for perfectly oriented material (referred later as TC_max). Additional analyses using a 2θ range of 20°–130° were performed for selected coatings and in this case, ICCD’s PDF card no 14-4268 was used to identify the reflexes. It is emphasised that we are here dealing with a fiber texture, meaning that in the (hkil) textured coating, the crystallographic planes {hkil} are preferentially oriented parallel to the substrate, while there is a rotational degree of freedom around the fiber axis which is perpendicular to this plane.

The EBSD studies were carried out on coating surfaces mechanically polished using colloidal silica as well as ion polished cross-sections. The specimens were ultrasonically cleaned before EBSD examination. The sample surface was tilted 70° to the incident beam and analysis was carried out at 20 kV. Data collection was performed at 80 Hz on a scan area of 30 µm × 15 µm. A step size of 0.05 um was used during the EBSD scans. EBSD inverse pole figure (IPF) maps, grain boundary (GB) statistics as well as Σ values were acquired and processed by using commercial integrated software. The constrained coincidence site lattice (CLS) model adapted for hexagonal symmetry by Grimmer was used [32]. Brandon criteria [33] was applied to account for the allowed deviation from the theoretical values. Special GBs corresponding to given Σ values were counted and expressed as a fraction of the total GBs. A confidence index (CI) ≥0.2 was applied [34]. In this context, only the Σ3 values were dealt with.

3. Results and Discussion

3.1. Influence of H₂S on Growth Rate of α-Al₂O₃

The results presented below were obtained by depositing multilayers of α-Al₂O₃ as described in Section 2.2.1. The aim of this experiment was to obtain a general understanding of the effects of H₂S and CO₂ on the growth rate of α-Al₂O₃ as well as to elucidate the limits for the gas-phase nucleation. The detailed process data for the coatings are presented in Table 1.

SEM cross-section images (Figure 1a–c) show alumina multilayer coatings composed of 8 layers of α-Al₂O₃ deposited on the MTCVD Ti(C,N) layers described above. The layers were deposited at H₂S contents of about 0.1 vol % (first layer), 0.2 vol % (second layer), 0.4 vol % (third layer), 0.6 vol % (fourth layer), 0.8 vol % (fifth layer),1.0 vol % (sixth layer), 1.4 vol % (seventh layer) and 1.6 vol % (eight layer). The CO₂ content was 4 vol % for the coatings shown in Figure 1a and 6 vol % for the coating shown in Figure 1b. From the SEM image (Figure 1b), it is evident that the uppermost layer deposited using a high H₂S (1.6 vol %) concentration showed gas phase nucleation. As is clear from Figure 1a, this was not the case when the same coating was deposited at CO₂ = 4 vol %. Figure 1c shows an inverse pole figure map of ion polished cross-section of the coating shown in Figure 1b. It is interesting to note that the inverse pole figure map suggests that the preferred growth direction of α-Al₂O₃ changes from (11$\overline{2}$0), as seen in the first and second layers, to (0001) with increasing H₂S doping. The uppermost layer, however, which was affected by gas-phase nucleation was composed of more randomly oriented grains. It is emphasized that these experiments where the catalyzed growth period of each layer was only 50 min cannot be used to draw any conclusions of the effects of H₂S on the texture development. Figure 1c gives only a clue that H₂S concentration may affect the texture of α-Al₂O₃ and was a triggering factor to investigate the role of H₂S concentration in more detail as described in Section 3.2.1.

The growth rates measured using cross-section SEM are presented in Figure 2 as a function of H₂S content at CO₂ levels of 3, 4 and 6 vol %. As shown in Figure 2, the growth rate of α-Al₂O₃ was strongly dependent on H₂S concentration and the effect of H₂S was more pronounced at higher CO₂ levels. Consequently, at higher CO₂ levels, very high growth rates could be reached at increased levels of H₂S doping. This is in accordance with earlier studies for κ-Al₂O₃ [10]. The maximum in the growth rate occurred at about 0.3, 0.6 and 0.9 vol % H₂S at CO₂ levels of 3, 4 and 6 vol %, respectively. The maximum growth rate was about 2.2 μm/h at CO₂ = 6 vol % and H₂S ~0.9 vol %. Compared with deposition rates of less than 0.1 µm/h obtained without H₂S doping (see Section 3.1), the growth rates could thus be increased by more than 20 fold. After having reached the maximum, the growth rates decreased with the H₂S increment, as has also been reported previously for κ-Al₂O₃ [10]. This behavior was observed at all the studied CO₂ levels and was more pronounced at the highest CO₂ level, possibly due to the tendency for gas-phase nucleation, as also speculated in connection with κ-Al₂O₃ [10]. Even though these results give a general idea of the possibilities to manipulate the growth of α-Al₂O₃ by H₂S and CO₂, it is not relevant to compare the growth rates presented in Figure 2 with the growth rates of textured α-Al₂O₃ layers discussed later in connection with texture development.

3.2. Texture and Microstructure of α-Al₂O₃

The results presented below were obtained by depositing thick single layers of α-Al₂O₃ as described in Section 2.2.2. The detailed process data for the coatings is presented in Table 2. The aim of this part was to elucidate the influence of H₂S on texture development.

3.2.1. Texture Development

The reference layer of α-Al₂O₃ (H₂S = 0) showed a thickness of about 1.2 μm, despite a long deposition time of 1200 min exhibiting thus the average growth rate of only 0.06 µm/h. This coating showed a clear (11$\overline{2}$0) texture with TC(11$\overline{2}$0) = 4.5, which can be considered high as far as the coating thickness is considered. The measured TCs for this layer are presented in

Table 5. Texture coefficients for α-Al₂O₃ coating deposited at H₂S = 0.

(hkil)	TC
(10$\overline{1}$2)	0.67
(10$\overline{1}$4)	0.65
(11$\overline{2}$0)	4.50
(0006)	0
(11$\overline{2}$3)	0.50
(11$\overline{2}$6)	0.59
(30$\overline{3}$0)	0.09

and these values are used in Figure 3a,b at H₂S = 0.

Figure 3a,b shows the developments of the TCs according to equation (4) for the reflections (10$\stackrel{\_}{1}$2), (10$\stackrel{\_}{1}$4), (0006), (11$\stackrel{\_}{2}$0), (11$\stackrel{\_}{2}$6) and (30$\stackrel{\_}{3}$0) as a function of increasing the H₂S concentration from about 0.03 vol % to about 1.25 vol %. The depositions were carried out at CO₂ concentrations of about 3.25 vol % (Figure 3a) and CO₂ 6.25 vol % (Figure 3b), respectively. The coatings selected for this X-ray study were deposited at the following H₂S concentrations: ~0.03, ~0.08, ~0.16, ~0.32, ~0.5, ~0.7, ~0.95 and ~1.25 vol %. Consequently, 8 layers were deposited at each CO₂ level. The corresponding H₂S/CO₂ ratios were 0.004 < H₂S/CO₂ < 0.4. The deposition time for all the α-Al₂O₃ coatings was 480 min and the obtained coating thicknesses varied from about 6.5 µm to about 10 µm.

As shown in Figure 3a (CO₂ = 3.25), the maximum values for TC(30$\overline{3}$0), TC(10$\overline{1}$2) and TC(10$\overline{1}$4) appeared at about 0.08, 0.16 and 0.32 vol % H₂S, respectively. At H₂S levels exceeding ~0.6 vol %, a strong TC(0006) started to develop reaching TC(0006) > 6 at H₂S levels from ~0.6 to ~1.1 vol %. The total absence of the (0006) reflection at H₂S < 0.3 vol % should be noted. Further, it should be noted that very low TCs for (11$\overline{2}$0) were detected in all experiments at CO₂ = 3.25 as compared with the reference coating deposited at H₂S = 0 exhibiting TC (11$\overline{2}$0) ~4,5. The results also suggest that at the H₂S concentrations exceeding ~0.9 vol %, the intensity of (0006) reflection started to decline with the re-appearance of TC(11$\overline{2}$0). Figure 3b shows the development of TCs as a function of H₂S concentrations at CO₂ = 6.25 vol %. As compared with the results presented in Figure 3a, the differences are clear and surprisingly, a very high TC(11$\overline{2}$0), was now detected at very low concentrations of H₂S (H₂S ~ 0.03 vol %). Furthermore, very strong TCs (> 5.5) for (10$\overline{1}$0), (11$\overline{2}$0) and (10$\overline{1}$4) at H₂S levels of about 0.08, 0.32 and 0.7 vol %, respectively, could be distinguished. At H₂S concentrations exceeding ~1 vol % the TC(0006) started to dominate. The highest TC (0006) = 6.3 was obtained at H₂S = 1.2 vol %. As compared with the results obtained at CO₂ = 3.25 vol %, all TCs reached ~80% of TC_max and were clearly separated with respect to H₂S concentration which was not the case at CO₂ = 3.25 vol %.

Figure 4 is a summary of texture development (including the data presented in Figure 3a,b) where identified and verified textures (based on 2 repetitions) are plotted as a function of H₂S/CO₂ (0.004 < H₂S/CO₂ < 0.4). The CO₂ concentrations used in the experiments were from about 3.25 vol % to about 6.50 vol %. As depicted on Figure 4, (11$\overline{2}$0) texture could only be obtained when H₂S/CO₂ ≤ 0.01 (field a in Figure 4) and (10$\overline{1}$0) texture existed when 0.01 < H₂S/CO₂ < 0.02 (field b in Figure 4). Thus, to facilitate deposition of (11$\overline{2}$0) and (10$\overline{1}$0) textures, very low H₂S/CO₂ ratios must be applied. Sufficiently low H₂S/CO₂ to deposit (11$\overline{2}$0) texture could only be reached at higher CO₂ levels used in these experiments.

The (11$\overline{2}$0) and (10$\overline{1}$0) reflections were absent at higher H₂S/CO₂ values and instead, TC(10$\overline{1}$2) was found in the experiments when 0.02 ≤ H₂S/CO₂ < 0.1 (field c, Figure 4). When H₂S/CO₂ exceeded ~0.2 the (0001) texture dominated (field e, Figure 4). In between the (10$\overline{1}$2) and (0001) windows, the results suggest the presence of (10$\overline{1}$4) texture (field d, Figure 4). The process windows for (11$\overline{2}$0), (10$\overline{1}$0) and (10$\overline{1}$4) textures appeared to be narrow as compared with the process windows for (10$\overline{1}$2) and (0001) textures. As clear from Figure 4, the H₂S/CO₂ ratio is an important parameter, which can be used to control the texture of α-Al₂O₃.

It is emphasized that there are naturally no abrupt boarders separating the growth regimes (fields a, b, c, d and e) depicted in Figure 4. Further, the hypothetical process windows presented here can be affected by many other process parameters (which were not in scope of this study) e.g., total pressure, AlCl₃/CO₂ ratio, texture of the MTCVD Ti(C,N) layer as well as nucleation and bonding procedures, to name a few. In the experiments discussed above a thin (~3µm) layer of MTCVD Ti(C,N) without optimized texture was applied, which could have affected the alumina growth. In addition, the thin Ti(C,N) layer may have allowed diffusion of the substrate elements (Co, W) to the Ti(C,N)-Al₂O₃ interface disturbing the process [1,30]. However, the suggested process windows presented in Figure 4 are assumed to hold reasonably well for CO₂ values higher than ~3 vol %. In the first commercial α-Al₂O₃ coatings, the most commonly found texture was (10$\overline{1}$2) [22].

Figure 5a–e shows X-ray diffractograms of the coatings a, b, c, d, and e deposited at H₂S = 0.03 vol % (CO₂ = 6.25 vol %), H₂S = 0.08 vol % (CO₂ = 6.25 vol %), H₂S = 0.2 vol % (CO₂ = 3.25 vol %), H₂S = 0.7 vol % (CO₂ = 6.25 vol %) and at H₂S = 0.7 vol % (CO₂ = 3.25 vol %), respectively. Consequently, coatings a, b, c, d and e were deposited at H₂S/CO₂ ratios of the order of 0,005, 0.013, 0.06, 0.1 and 0.2, respectively. As clear from the X-ray diffractograms in Figure 5a–e, very profound textures could be obtained in all studied coatings. It should further be noted that the coatings which exhibited strong (10$\overline{1}$4) and (0006) textures (Figure 5d,e, respectively) were both deposited using the same H₂S concentration of about 0.7, vol %, but at different CO₂ concentrations of 6.25 vol % (Figure 5d) and 3.25 vol % (Figure 5e) respectively. The H₂S/CO₂ ratios used to deposit Coatings d and e were consequently of the order of 0.1 and 0.2, respectively.

3.2.2. Microstructure

Figure 6A–E shows SEM cross-sections and surface morphologies of Coatings a, b, c, d and e deposited, as mentioned above, at H₂S/CO₂ ratios of the order of 0,005, 0.013, 0.06, 0.1 and 0.2, respectively (see Figure 5a–e). The growth rates for (11$\overline{2}$0) textured α-Al₂O₃ (Coating a), (10$\overline{1}$0) textured α-Al₂O₃ (Coating b), (10$\overline{1}$2) textured α-Al₂O₃ (Coating c), (10$\overline{1}$4) textured α-Al₂O₃ (Coating d) and (0001) textured α-Al₂O₃(Coating e) were 1.3, 1.1, 0.8, 0.8. and 1.1 µm/h, respectively (see Figure 6A–E). It is plausible that the identified textures may exhibit different growth rates related to this specific orientation. However, since the same texture can be deposited with very different process data within the specific H₂S/CO₂ regime, according to this work, it is not possible to define the typical growth rates for different preferred growth directions (textures) based on the present experiments. It should further be noticed that (11$\overline{2}$0) and (10$\overline{1}$0) textured layers showed higher or equal growth rates as compared with the (10$\overline{1}$2), (10$\overline{1}$4), and (0006) textured α-Al₂O₃ layers deposited at higher H₂S concentrations but lower CO₂ levels. It must be emphasized that the growth rates obtained in the first part of this work (presented in Figure 2) are not comparable with those obtained for strongly textured coatings.

All the α-Al₂O₃ layers were composed of columnar grains. The (10$\overline{1}$2) textured coating was composed of slightly smaller and more uniform columnar grains terminated by sharp pyramidal-shaped grains (Coating c, Figure 6) as compared with the other textured layers. A clear tendency for widening of the columnar grains perpendicular to the growth direction and the formation of grains with flat surfaces instead of sharp, faceted grains was observed when deposition was carried at H₂S/CO₂ < 0.02. Consequently, (11$\overline{2}$0) and (10$\overline{1}$0) textured α-Al₂O₃ layers exhibited larger grain sizes as compared with the other textured α-Al₂O₃ layers.

The Σ3 CLS GB fractions and TCs for the (11$\overline{2}$0), (10$\overline{1}$0), (10$\overline{1}$2) and (0006) α-Al₂O₃ coatings deposited at CO₂ = 6.25 vol % are given in

Table 6. Fraction of Σ3 CSL GBs in textured α-Al₂O₃ coatings.

Texture (hkil)	TC (hkil)	Fraction of Σ3 CSL GBs (%)
(11$\overline{2}$0)	5.9	~5
(10$\overline{1}$0)	6.5	~5
(10$\overline{1}$2)	5.9	~12
(0001)	6.2	~50

. The (0001) textured α-Al₂O₃ exhibited high Σ3 CLS GB fractions of about 50% compared with low values (5–12%) for the other textured coatings. Similar results have been published previously [35,36,37]. Volanthen and Grobety [38] measured Σ3 CLS GB fractions of about 15–16% for slightly textured sintered alumina and about 3% for non-textured sintered alumina. As concluded in [38], the high fraction of Σ3 CLS GBs in the (0001) textured α-Al₂O₃ is a result of crystallographic control and generally depends on the orientation of the CSL misorientation axis with respect to the [0001] axis. This is assumed to explain the high Σ3 CLS GB fractions also in the studied CVD (0001) textured α-Al₂O₃ layers.

3.2.3. Texture-Manipulated Structures

The results presented below were obtained by depositing α-Al₂O₃ as described in Section 2.2.3 and Table 3 and Table 4. The aim of the experiments was to deposit (0001) textured α-Al₂O₃ coatings first on thicker and textured MTCVD Ti(C,N) layers using state-of-the-art nucleation and bonding layers (Coating f) and then investigate the effects of changing the fibre axis (growth direction) during deposition by 90° degrees (Coatings g and h). It should be noted that directions <0001> and <11$\overline{2}$0> as well as <0001> and <10$\overline{1}$0> are perpendicular to each other. In this experiment the (0001)–(11$\overline{2}$0) system was used. These types of process modifications could open interesting possibilities to manipulate the grain size and grain boundary structures of α-Al₂O₃ coatings.

A cross-section SEM micrograph and an EBSD IPF surface map of Coating f are shown in Figure 7a,b. The corresponding X-ray diffractogram is shown in Figure 7c. The interfacial part of the α-Al₂O₃ layer was composed of a fine-grained region, as can clearly be seen in Figure 7a. This region was deposited at H₂S = 0 for 30 min and obviously grew having <11$\overline{2}$0> as preferred growth direction (see Section 3.2.1). The addition of H₂S to the process after 30 min changed abruptly the growth direction from <11$\overline{2}$0> to <0001>. Hardly any region of evolutionary selection, i.e., competing growth of differently oriented crystals can be seen and the fiber axis had adopted the [0001] direction almost immediately. As shown in Figure 7a, the α-Al₂O₃ layer above the fine-grained region was composed of columnar crystals growing along [0001].

As clear from the EBSD IPF surface map (Figure 7b) and from the X-ray diffractogram (Figure 7c), the α-Al₂O₃ layer exhibits a pronounced (0001) texture (TC(0006) = 5.8). In the diffractogram shown in Figure 7c, the (0006) reflection and its second order reflection (000.12) at (2θ) angles of 41.71° and 90.74°, respectively, can be seen (JCPDF Card 14-4268). In addition, a TiN(422) reflection, which originates from the textured Ti(C,N) is present. Consequently, in this case, also the MTCVD Ti(C,N) layer is textured. This experiment demonstrates the importance of the MTCVD Ti(C,N) layer, together with optimized bonding and nucleation processes to fully utilize the texture control. A high density of Σ3 CLS GBs of the order 52% was detected in this α-Al₂O₃ layer.

Figure 8 shows a cross-section SEM micrograph and EBSD IPF surface map of Coating g deposited using the H₂S/CO₂ ratios 0.40 and <0.01 applied in sequences (see Section 2.2.3). The resulting coating was composed of 14 sub-layers: 7 layers deposited with (0001) conditions and 7 layers deposited with (11$\overline{2}$0) conditions. The deposition time for the first (0001) sequence was 45 min and 30 min for the following sequences. H₂S was added in the process after 5 min delay and consequently, the fine-grained α-Al₂O₃ at the α-Al₂O₃-Ti(C,N) interface, which can clearly be distinguished in Coatings f and h (Figure 7a and Figure 9a, respectively) is almost totally absent here.

This coating exhibited a—slightly unpredicted—strong (0001) texture TC(0006) = 5.78 having been determined by the first step, which, in this case, was (0001). The grains started to grow along the [0001] direction, which was also maintained during the subsequent steps deposited with (1$\overline{1}$20) process conditions. During the sequences using (1$\overline{1}$20) process conditions, the growth rate perpendicular to the growth direction [0001] was increased. During the next step using (0001) process conditions, the opposite effect occurred. Consequently, the variations of the growth rate perpendicular to the main grain growth direction of [0001] resulted in the formation of a “zigzag” grain boundaries, as clearly seen in Figure 8a. The wavelength of the “zigzag“ grain boundary remained constant, suggesting that the growth rate along the [0001] direction did not change markedly, even though a different process parameter were applied. The “zigzag” feature can also be obtained by applying the process conditions for (10$\overline{1}$2) and (10$\overline{1}$0) in combination with (0001) process conditions in sequence [39]. It remains to be seen if one can enhance the wear properties of α-Al₂O₃ layers by applying these types of processes.

In Coating h (Table 4), the α-Al₂O₃ coating was deposited at H₂S/CO₂ = 0.4 (after a period of 30 min at H₂S = 0) to a thickness of about 7 µm and only after this period, the process conditions were switched over to H₂S/CO₂ < 0.01. These conditions were applied for one hour. From Figure 9a an abrupt change in the growth direction can be seen. Obviously, this effect can be preferably obtained in α-Al₂O₃ layers with the established preferred growth along the [0001] direction. This is due to availability of several sets of {11$\overline{2}$0} growth facets parallel to the [0001] direction. Figure 9b shows the surface morphology of Coating h. The α-Al₂O₃ surface was dominated by (0001) facets, as confirmed by EBSD. As a result, the surface roughness was reduced considerably. This type of process could be used to increase surface quality of the coated cutting tool inserts and has already been applied in some commercial products [40].

3.2.4. Discussion

Only a few studies can be found in the literature on the influence of doping on texture development in CVD α-Al₂O₃ coatings. A strong (10$\overline{1}$0) texture was observed to develop in α-Al₂O₃ as a result of ZrCl₄ doping [41] and a strong (11$\overline{2}$0) texture as a result of TiCl₄ doping [42]. In the case of ZrCl₄ doping, it was speculated that Zr-species were adsorbed on the growing α-Al₂O₃ surfaces blocking the growth the sites, which could lead in the formation of other growth textures, such as (10$\overline{1}$2) and (10$\overline{1}$4). The texture development observed in the present experiments will also be explained below by selective CVD growth on certain alumina facets or sets of alumina facets [43,44].

There has been much interest in understanding the behavior of alumina surfaces [45,46,47,48,49]. Using first-principles density functional theory, Hinneman et al. [45] found that the sulphur atom preferentially adsorbs to an oxygen. The effects of the environment on the Gibbs free energy of α-Al₂O₃ (0001) surfaces were studied in [46] and surface terminations as a function of oxygen partial pressure were discussed. The (0001), (10$\overline{1}$2), (11$\overline{2}$3), (11$\overline{2}$0), (10$\overline{1}$0) α-Al₂O₃ surfaces have been studied by Marmier and Parker [47], applying the ab initio method combined with thermodynamics. The authors found that at ambient conditions, only two types of surfaces—stoichiometric or fully hydroxylated—are thermodynamically stable, except at extremes of oxygen or hydrogen partial pressures. A recent ab-initio study of the α-Al₂O₃ (10$\overline{1}$4) surface [48] discusses the possible surface terminations of and emphasizes the effect of temperature and O₂-partial pressure. It has also been reported that H₂O adsorbs much more rapidly on the α-Al₂O₃ (10$\overline{1}$2) surface than on the α-Al₂O₃ (0001) surface [49]. However, in CVD, several Cl-containing species are present in the gas phase and this type of conditions have not been considered in the above-mentioned studies. A recent study [16] using thermodynamic modelling and density functional theory (DFT) discusses the behavior of H₂S on (0001) alumina surfaces under more realistic CVD conditions. The oxygen terminated surface was found to be hydrogenated, in agreement with several previous studies and Al-terminated surfaces were found to be Cl-terminated. In accordance with the model of Slager and Amber [28], the authors suggest the following scenario to take place on the Cl-terminated (0001) surface:

H₂S + Al₂O₃:Cl → Al₂O₃:SH + HCl

H₂O + Al₂O₃:SH → Al₂O₃:OH + H₂S

The proposed construction nicely describes the recycling of H₂S on a Cl-terminated α-Al₂O₃ (0001) surface and explains why very small amounts of H₂S (usually much less than 0.5 vol % of the total flow), are needed to achieve and maintain the high growth rate.

As clear from the above-cited literature and references therein, different α-Al₂O₃ crystal facets (surfaces) display very different reactivities for molecular (or dissociative) adsorption, adsorbate stability and desorption, depending on crystallography and termination of an individual facet. Obviously, there may exist several types of competing adsorption/desorption modes that are favored by certain α-Al₂O₃ surfaces and/or terminations. It is speculated that certain crystal facets together with related catalyzed reactions could be stabilized by H₂S and especially by H₂S/CO₂. H₂S acts on these facets as a mediator for the hydrolysis of Al-containing species. The understanding of the surface reactions is further complicated by the fact that the CVD environment most probably also affects the homogenous decomposition of AlCl_x and the resulting intermediate Al-species and their surface reactions on α-Al₂O₃ [13,14,15].

It is hypothesized that texture development in α-Al₂O₃ coatings could be a result of catalyzed reactions and growth occurring preferentially on specific growth facets of Al₂O₃ being determined by H₂S partial pressure and the H₂S/CO₂ ratio. This hypothesis, however, requires that there exist specific equilibrium conditions for each set of growth facets, which is not contradicted by the present results. The fact that the growth direction could be rotated by 90° by altering the process condition supports the above discussed mechanisms. Deposition without H₂S and with very low amount of H₂S (especially together with relatively high amount of CO₂) resulted in the (11$\overline{2}$0) texture. This texture could be a result of same surface reactions on the same set of facets that occurred also in the absence H₂S. With the increment of H₂S, the surface reactions on these planes were retarded or blocked and growth shifted to next set of growth facets. Since we can see the effect of increased H₂S on all textures at all levels of CO₂, the limiting step could be desorption reactions related to the removal of sulfur adatoms or sulfur-containing species. This assumption is supported by fact that S is not found in CVD Al₂O₃ [10] and H₂S does not dissolve in α-Al₂O₃ [29]. Furthermore, it is noted that the limiting steps as well as the surface reactions do not necessarily need to be same on all sets of facets. It should also be noted that it is not clearly established on which crystallographic surfaces (facets) the growth of different α-Al₂O₃ textures occurs.

This work shows that as a result of the H₂S increment at constant CO₂ concentrations, the texture of α-Al₂O₃ changed from (11$\overline{2}$0) texture, which was obtained at very low H₂S concentrations, through (10$\overline{1}$0), (10$\overline{1}$2) and (10$\overline{1}$4) textures to (0001) texture, which was obtained at ~ 0.6 vol % H₂S (at CO₂ = 3.25 vol %) and ~ 1 vol % H₂S (at CO₂ = 6.25 vol %). Consequently, the process windows for the observed textures should preferably be expressed as a function of the H₂S/CO₂ ratio. The use of texture control to modify coating microstructures was demonstrated and validated. This type of processes could be used to further improve α-Al₂O₃ coatings. As demonstrated in Figure 2, the growth rate of α-Al₂O₃ can be increased considerably by applying high CO₂ and H₂S partial pressures. Application of high CO₂ and H₂S partial pressures, together with texture control (H₂S/CO₂ ratio) introduced in this work opens possibilities to deposit e.g., (0001) textured α-Al₂O₃ coatings at high growth rates. The importance of intermediate layers including bonding and oxidation steps for the texture development was emphasized.

4. Conclusions

This paper studied the influence H₂S and CO₂ on texture and microstructure of CVD α-Al₂O₃ deposited from the AlCl₃-H₂-CO₂-HCl-H₂S system. The main emphasis was on the texture development in CVD α-Al₂O₃. The following conclusions could be drawn:

• The experiments confirmed that the growth rate of α-Al₂O₃ could be increased considerably by applying high CO₂ and H₂S partial pressures.
• When deposited uncatalyzed (H₂S = 0) and using very low H₂S concentrations the resulting α-Al₂O₃ layer exhibited (11$\overline{2}$0) texture.
• As a result of the increment of H₂S at constant CO₂ concentration the texture of α-Al₂O₃ changed from (11$\overline{2}$0), which was obtained at very low H₂S concentrations, through (10$\overline{1}$0), (10$\overline{1}$2) and (10$\overline{1}$4) textures to (0001) texture. Consequently, the fiber axis of α-Al₂O₃ was rotated by 90° as a result of H₂S increment.
• The H₂S/CO₂ ratio was identified as an important parameter to control the texture of α-Al₂O₃. Growth regimes for (11$\overline{2}$0), (10$\overline{1}$0), (10$\overline{1}$2), (10$\overline{1}$4) and (0001) textures were depicted as a function of the H₂S/CO₂ ratio. With the increment of H₂S/CO₂ ratio the texture of α-Al₂O₃ changed from (11$\overline{2}$0), which was obtained at H₂S/CO₂ < 0.01 through the (10$\overline{1}$0), (10$\overline{1}$2) and (10$\overline{1}$4) textures to the (0001) texture, which was obtained at the H₂S/CO₂ ratios exceeding ~0.2.
• Even though this work has focused on the effect of H₂S concentration at constant CO₂ levels, it should be noted that based on the present results is straightforward to conclude that the H₂S/CO₂ ratio can also be varied by varying the CO₂ concentration at a constant H₂S level.
• This work also unveiled new possibilities to manipulate the surface and grain boundary structures of the CVD α-Al₂O₃ coating by utilizing the texture-control.
• The work demonstrated that the properties of the MTCVD Ti(C,N) layer and especially the bonding, nucleation and oxidation procedures are of crucial importance for full utilization of the process windows according to this work.

5. Suggestions for the Further Work

The results of this work give rise to several fundamental questions concerning texture development and the basic mechanisms of H₂S catalysis. A TEM study together with 3D-EBSD of the textured coatings as well as of the faceted grain boundaries shown in Figure 8 could give valuable information. Since α-Al₂O₃ single crystals are commercially available, it would also be advisable to study adsorption-desorption mechanisms of H₂S on, e.g., [10$\overline{1}$2], [11$\overline{2}$0] and [0001] oriented single crystals and combine these studies with practical CVD deposition experiments. Deposition of α-Al₂O₃ using different H₂S/CO₂ ratios on the crystallographically controlled α-Al₂O₃ single crystal surfaces would contribute to enhanced understanding of this complex phenomena. Further, it would be of interest to study the influence of H₂S on the growth and texture development in CVD κ-Al₂O₃.