EiAMOND RELATED MATERIALS ELSEVIER
Diamond
Substantial
and Related
Materials
effect of linear velocity of combustion growth of diamond S. Marinkovic,
Depurtment
4 (1995) 186-190
gf Materials
Science,
Received
Institute
of Nuclear
Sciences
1 August 1994; accepted
flame on oriented
S. Zec “Vinca”,
POB 522, 1 JO01 Belgrade,
in final form 19 October
Minor Yugoslutk
1994
Abstract Diamond coatings were deposited on diamond-polished molybdenum substrates using combustion flame CVD. The total (acetylene plus oxygen) flow rate was between 75 and 310 1 h-‘, the acetylene-to-oxygen flow rate ratio was between 1.03 and 1.22 and the substrate temperature was from 620 to 900 “C. A strong dependence of preferred orientation and crystalline order of the coatings, as determined by X-ray diffraction, on the total flow rate is obvious in spite of the wide range of other experimental conditions used for preparation of the samples. A strong (110) preferred orientation observed at low flow rates rapidly decreases and changes to (111) preferred orientation at the highest flow rates. The preferred orientation is apparently restricted to the direction of the coating growth. The change in preferred orientation, which reflects a change in the relative growth rates in the
(110) and (111) crystal directions, is related to a change in surface morphology: instead of the { 111) faces appearing at low fow rates, cube-octahedral or cubic morphology was observed at high flow rates. The crystalline size values L,,, and L,zO. taken as a measure of crystalline order, behave differently: LII1 steadily increases with increasing flow rate, while I&, remains virtually unchanged. Kr~ords:
Combustion
flame CVD; Polycrystalline
diamond
films; Oriented
1. Introduction A preferential alignment of diamond { 110) (and sometimes { 1111) planes parallel to the substrate in coatings obtained by CVD has often been reported [l-4]. Yarbrough et al. [l] found at low substrate temperatures (down to 450 “C) a strong (110) orientation of diamond coatings prepared from a CH,-H, gas mixture using remote thermal CVD. Celii et al. [2], working with microwave CVD from a CH,-H, mixture, found that the preferred orientation of diamond films depended on the substrate nature: it was (111) on diamond-seeded Si, but on microcrystalline diamond (obtained by CVD on Si) it was (110). Wild et al. [ 31, who used a combination of hot filament and microwave plasma CVD, found that the diamond crystals formed at the beginning of the coating growth were randomly oriented and that a preferential orientation of { 110) planes parallel to the substrate surface gradually developed with increasing coating thickness. The present authors found a (110) preferred orientation, ranging from high to low, in diamond coatings prepared by acetylene-oxygen combustion flame CVD [4]. OY25-9635/95/$09.50 0 1995 Elsevier Science S.A. All rights reserved &SD1 0925-9635(94)00242-8
growth;
Orientation
This paper contains results on the effect of the linear velocity of the combustion flame on the preferred orientation of the deposited diamond. The dependence of the crystalline order on the linear flame velocity has also been studied. Regarding the latter point, Snail and Craigie [S] have found a positive effect of high flame velocity (turbulent flame) on diamond crystalline order.
2. Experimental
details
The simple apparatus used to deposit diamond has been described previously [ 11. A conventional welding torch fitted with a 1 mm welding tip has been used. A molybdenum substrate in the form of a plate, polished with diamond paste, has been fixed on a water-cooled copper block. The substrate temperature was regulated by adjusting the thermal contact between the substrate and the copper holder; it was monitored by means of a chromelLalume1 thermocouple placed in a crater drilled on the back of the substrate to a depth of approximately 0.5 mm below the substrate surface. Acetylene and oxygen flow rates and their ratios were controlled by
S. Marinkovic. S. Zec / Diamond and Related Materials 4 (1995) 186-190
means of flowmeters. The ranges covered by the whole group of deposition experiments performed were: substrate temperature from 620 to 900 “C; total (acetylene plus oxygen) flow rate between 75 and 310 1 h-‘; acetylene-to-oxygen flow rate ratio between 1.03 and 1.22. Diamond growth occurred within an approximately circular area on the substrate centred along the flame axis. The region of growth had a diameter of about 5 mm. The diamond coating often peeled off the substrate and was examined in such a case as a freestanding sample. The diamond coatings were characterized by X-ray diffraction, scanning electron microscopy and Raman spectroscopy. X-Ray diffraction was used to determine: the diamond lattice constant from the (311) Ka, peak position; the apparent crystalline sizes in the (110) and (111) directions, L,,, and L,,, respectively, as derived from the full width at half-maximum (FWHM) of the respective Ka, profiles corrected for instrumental broadening; the intensity ratio of the (220) and ( 111) KU, reflections, 1220/1111. Raman spectra were obtained using an argon ion laser with the 514.5 nm line as the excitation wavelength.
3. Results and discussion The effect of the linear velocity of the combustion flame on the intensity ratio of the X-ray reflections, taken to represent preferred orientation, is I,,,lI,,,, shown in Fig. 1. In spite of scattering of the experimental points, a strong dependence of preferred orientation on total flow rate is obvious. The scattering of the points
is presumably due to the fact that results for all the samples were used, irrespective of the considerable variation in experimental conditions other than flow rate (substrate temperature, acetylene-oxygen ratio; see Section 2) used for their preparation. A marked preferred orientation of the crystals with their { 110) planes parallel to the deposition surface was observed at low flow rates. According to Wild et al. [ 31, the preferred orientation is created during growth by a competition of differently oriented grains, whereby the crystals with their { 1 lo} planes parallel to the deposition surface grow faster. It follows from Fig. 1 that the (110) orientation decreases with increasing flow rate and tends to change to (111) orientation at the highest flow rate values. Supposing that the preferred orientation is a consequence of different rates of growth of differently oriented grains, this change should result in a change in surface morphology, because the planes with the slowest growth rate should determine the morphology. Such a change was found to take place: only octahedral { 11 l} faces were observed at low flow rates, while cubooctahedral or cubic crystals were sometimes present at high flow rates. In order to get information on the dependence of the crystalline order on the linear velocity of the flame, the apparent crystalline sizes in the (110) and (111) directions as well as their ratio were plotted against total flow rate. The L values show the extent of coherent crystalline domains (in the respective crystal directions), so that they can be taken to represent the crystalline order. In fact, it has been suggested [4] that Lll, (or FWHM,,,) should be used as a measure of diamond quality. Fig. 2 shows that the behaviour of the crystalline
200
150
187
250
Total flowrate, l/h Fig. 1. Dependence
of preferred
orientation,
expressed
as X-ray reflection
intensity
ratio I,,,/I,,,,
on total flow rate.
188
.
Total flowrate, l/h Fig. 2. Dependence
of crystallite
size ratio L,,,/L,,, is qualitatively similar to that of the intensity ratio. As in the case of Fig. 1, all available results were used for Fig. 2, neglecting differences in substrate temperature and gas mixture composition in the sample preparation. At low flow rates the crystalline size ratio is high, but it decreases as the flow rate increases, becoming less than unity at the highest flow rates studied. Figs. 3(a) and 3(b) show the behaviour of crystalline sizes L,,, and I,,,, respectively with increasing flow rate. shows a distinct increase (Fig. 3(a)), while L,,, 2;;. 3(b)) remains virtually unchanged, showing that the increase in L,,, values is responsible for the behaviour of the L,,,/L,,, ratio with increasing flow rate (Fig. 2). It follows from the data presented in Figs. 1 and 3(a) that a relative increase in the rate of growth in the (111) direction (i.e. decreasing I,,,/I,,, ratio), provoked by the increasing flow rate, occurs together with an increased order along this direction. However, these results do not permit us to determine whether the (absolute) rate of growth in the (111) direction increases with increasing flow rate. The increased crystalline order of the coatings obtained at high flow rates is evident also from Raman results. The Raman diamond line at 1332 cm ’ becomes considerably narrower for these coatings, in agreement with a previously observed correlation between the FWHM of the X-ray (111) reflection and the FWHM of the Raman diamond line 141. In addition, the band at about 1350 cm-l due to disordered or microcrystalline graphitic carbon [6] is usually present in the Raman spectra of samples prepared at low flow rates, but it is mostly absent at high flow rates. This shows that the
size ratio LZ2D;L,,
I on total flow rate.
diamond purity also improves with increasing flow rate, in agreement with the correlation between the nondiamond phase content and the diamond quality [7]. The results discussed so far indicate what happens in the direction of coating growth. The question of how the crystals grow in other directions could not be answered on the basis of these results. In order to try to answer this question, some of the samples were powdered and examined by X-ray diffraction. The effect of powdering is that some of the crystals which during the deposition had been oriented so that their crystal planes were not parallel to the substrate surface and consequently not in the “visible” position were moved to such a position. The opposite, i.e. that some of the crystals which gave their contribution to the diffraction peaks were moved by powdering into ineffective positions, should also occur. Thus the effect of powdering is to “randomize” the oriented samples. The observed effect of powdering (Table 1) is that the preferred orientation (i.e. the intensity ratio I,,,,jl,,,) is shifted towards the value for a random sample (0.25). Thus, for the samples deposited at high flow rate (with ( 111) preferred orientation), the intensity ratio became Table 1 Effect of powdering Total
on preferred
orientation
I 220;1,, ,
and crystalline
size
L,, A
flow rate (I h-l)
As deposited
Powdered
122 267 274 280
2.1 0.57 0.14 0.18
1.1 0.48 0.30 0.21
As deposited
Powdered
260 400 680
200 255 X20
1200
830
189
S. Marinkovic, S. Zec / Diamond and Related Materials 4 (1995) 186-I 90
n
200
150
(a)
250
Total flowrate, l/h
lool--____ 50
100
(b) Fig. 3. Dependence
150
200
300
250
.
0
Total flowrate, l/h of crystallite
size in (a) (Ill)
higher after powdering. Conversely, for the samples with (110) preferred orientation, as deposited at low flow rates, the intensity ratio after powdering became less. These results suggest that preferential alignment exists only in the direction of growth, i.e. that only in this direction is there an oriented crystal growth. Results of crystalline size measurements (Table 1) point to the same conclusion: the Li,, values for the powdered samples differ considerably (being either higher or lower) from those for the as-deposited ones. The results presented here indicate that an increased flame velocity leads to the following changes in the direction of coating growth: (a) a rapid decrease in (110) preferred orientation which finally becomes (11 l), with
and (b) (110)
crystal
directions
on total flow rate.
a related change in morphology; (b) an increased crystalline order; (c) a decrease in non-diamond carbon content. It is interesting to compare the known effects of supersaturation and substrate temperature on oriented diamond growth and morphology [8,9] with the effects of flame velocity as found in this study. According to Angus and Hayman [S], attachment of single-carbon species on { lOO} planes that under “slow” growth conditions (low supersaturation, low temperature) leads to octahedral { 111) faceted crystals is energetically favoured, because these species can make two bonds with the surface. Therefore it proceeds at low supersaturation or low substrate temperature. Growth on { 11 l} planes, leading to cubic { lOO} faceted crystals,
190
S. Murinkovic,
S. Zrc i Diamondund
becomes rapid only at high supersaturation and/or temperature, because multicarbon nuclei have to be formed in that case. Applying this reasoning to the observed flow rate dependence of preferential growth, it follows that an increased flame velocity has an effect analogous to that of increased supersaturation and/or temperature. In this respect it is worth considering the work of Ravi and Joshi [lo], who found the morphology of diamond grown by an oxyacetylene flame to be a strong function of temperature, with temperatures below 1000 “C resulting in an octahedral habit and temperatures above 1000 “C resulting in a cubic habit. The authors postulate that the propagation of ( 100) faces at high temperatures is enhanced by the presence of oxidizing species (0, CO, OH) in the flame that preferentially remove nondiamond-bonded carbon from the deposit as well as etch the diamond surface to create growth ledges, which are most readily formed on { 100) surfaces. A considerable role of oxidizing species in flame deposition has also been pointed out by Snail and Craigie [S]. According to them, the low temperature boundary layer near the substrate becomes thinner as the flame velocity increases, thus shortening the transport time for diffusion across the layer. At very high flow rates, when the flame becomes turbulent, the diffusion rate is one to two orders of magnitude higher than that for the laminar flame [S]. The quality of the deposited diamond is improved at high flow rates owing to an increased diffusion of etchant species, while the rate of diamond growth is lower. The difference in composition between laminar and turbulent flames has recently been confirmed [ 111: unlike the laminar flame, the turbulent flame was found to contain the etchant species 02, CO, and H,O, although the amount of atmospheric air entrainment apparently did not increase. Thus it seems that the effect of increasing flame velocity on both the oriented growth and crystalline perfection as observed in the present work may have a common cause in an increasing flow of oxidizing species.
4. Conclusions A strong (110) preferred orientation observed in coatings prepared with a low linear flame velocity rapidly
Related Materi&
4 ( 1995) 186-190
decreases with increasing flame velocity and tends to change to (111) preferred orientation at the highest velocities used. The preferred orientation is apparently restricted to the direction of coating growth and is presumably a consequence of different growth rates in this direction of differently oriented crystals. As expected, the change in preferred orientation, which reflects a change in the relative growth rates, leads to a change in surface morphology: instead of the octahedral faces appearing at low flame velocity, cube-octahedral or cubic morphology is observed at the highest flow rates. The effect of increased flame velocity on preferential growth and morphology is analogous to that of increased supersaturation and/or temperature. The observed increase in crystalline size L,,, with flame velocity, indicating higher crystalline perfection, is presumably a consequence of an increased diffusion of etchant species with increasing gas velocity. Thus the relative increase in growth rate in the (111) direction occurs along with an increased crystal order in this direction. The observed effects of increased flame velocity on both oriented growth and crystalline perfection may have a common cause in an increased flow of oxidizing species.
References [I]
W.A. Yarbrough, A.R. Badzian, D. Ptckrell, Y. Liou and A. Inspektor, J. Cr?.st. Grorvth, 99 ( 1990) 1177. [?I F.G. Celii, D. White Jr. and A.J. Purdes, J. Appl. Phys., 70 (1991) 5636. [3] Ch. Wild, N. Hexes and P. Koidl, J. Appl. Phys., 68 (1990) 973. [4] S. Marinkovic and S. Zec, J. Serb. Ckrm. Sot., 58 (1993) 679. [S] K.A. Snail and C.J. Craigie, Appl. Phys. L&t., 5X (1991) 1875. [6] R.J. Nemanich and S.A. Solin, Phys. Rue. B. 20 (1979) 392. [7] S. Prawer, A. Hoffman, S.-A. Stuart, R. Manory, P. Weiser, C.S. Lim, J.M. Long and F. Ninio, J. Appl. Phys., 69 ( 199 1 ) 6625. [X] J.C. Angus and C.C. Hayman, Science, 241 (1988) 913. [9] M.N.R. Ashfold, P.W. May and C.A. Rego, Chrm. Sot. Rev.. 23 (1994) 21. [lo] K.V. Ravi and A. Joshi, Appl. Phys. L&t., 58 (1991) 246. [ 111 C.M. Marks, H.R. Burris, J. Grun and K.A. Snail. J. Appl. Ph.v.s.. 73 (1993) 755.