Influence of atomic hydrogen gradients on the growth rate and nucleation of diamond produced by microwave plasma assisted deposition

Influence of atomic hydrogen gradients on the growth rate and nucleation of diamond produced by microwave plasma assisted deposition

298 Diamond and Related Materials, 2 (1993) 298-303 Influence of atomic hydrogen gradients on the growth rate and nucleation of diamond produced by ...

485KB Sizes 0 Downloads 10 Views

298

Diamond and Related Materials, 2 (1993) 298-303

Influence of atomic hydrogen gradients on the growth rate and nucleation of diamond produced by microwave plasma assisted deposition A. O h l , J. R 6 p c k e a n d W . S c h l e i n i t z Institutefor Low-Temperature Plasma Physics, R.-Blum-Strasse 8-10, 0-2200 Greifswald (Germany)

Abstract Experiments to show the influence of the relative atomic hydrogen content on diamond deposition in hydrogen plasmas with small admixtures of alcohols are reported. By using a planar microwave plasma source and placing the substrates outside the active plasma excitation region, we could ensure that the substrate heating was dominated by the surface recombination of atomic hydrogen. Therefore,the concentration of atomic hydrogen could be assumed to be proportional to the substrate temperature. By varying the position of the substrate, deposition experiments with a variation in concentration of 60% were possible. Growth conditions yieldingwell isolated crystals were used. A dependenceof the medium crystal diameter on position, similar to the spatial dependence of the atomic hydrogen concentration was observed. This indicated a power law dependenceof the linear growth rate on the concentration of atomic hydrogen.An analysis of the crystallitesize revealed an asymmetricsize distribution. In contrast to the crystal diameters, the nucleation densities were independent of the atomic hydrogen concentration.

1. Introduction Low-pressure microwave plasmas are well adapted both to the processing conditions of low-pressure diamond growth and to the typical requirements of high-tech processes. They are therefore expected to be a useful tool for growing diamond films for future advanced applications such as optical coatings and passive or active constituents of electronic devices. However, for these applications, the requirements of crystalline quality and growth conditions are very high, as is known from the growth of other high-quality crystalline materials. There is no reason to assume different, less rigorous requirements for diamond growth. On the contrary, the specific process conditions cause additional difficulties. One specific problem of diamond film growth is the relatively low growth rates obtained under conditions useful for high-quality film growth. It is well known that lower admixtures of carbon-containing gas to the already large proportion of carrier gas improve the crystalline quality. Often admixtures of less than 1% hydrocarbon to hydrogen are used, but this results in very low growth rates. Therefore, it is not surprising that the tendency is to use as high a working pressure as possible. In this way growth rates can be increased, but at the same time spatial inhomogeneities in the growth medium and in growth increase. So it is of considerable interest to study not only the influence of external macroscopic process parameters such as the reaction gas composition, sub-

0925-9635/93/$6.00

strate temperature, gas flow rates and gas pressure on the deposition process, but also the correlation between the spatial variation of internal plasma parameters and diamond growth. In our previous investigations we studied in detail the heat transport to and from substrates during diamond growth in low-pressure microwave plasmas [1, 2]. The specific properties of the planar microwave plasma source used allowed us to distinguish between different heat transport processes. As a rule, at pressures of ~ few kilopascals, substrate heating is dominated by surface recombination of atomic hydrogen and substrate cooling is dominated by thermal conductivity of the neutral gas [2]. Detailed neutral gas temperature measurements demonstrated the non-isothermal character of the plasma. Gas temperatures down to ambient temperature were determined. The role of atomic hydrogen in the microscopic growth process is to suppress the much faster formation of non-diamond carbon structures and to stabilize the diamond surface lattice, which is thermodynamically unstable under these growth conditions. Taking this into account, complete or at least nearly complete coverage of the growing surfaces by atomic hydrogen can be assumed. It is known that this complete coverage allows us to assume direct proportionality between the substrate temperature and the concentration of atomic hydrogen in the gas phase [3]. This provides a very direct means of studying the influence of variations in the relative atomic hydrogen content on the deposition process. Temperature differences between different growth condi-

© 1993 -- ElsevierSequoia. All rights reserved

A. Ohl et al. / Influence of atomic hydrogen gradients

tions then can be linked directly to variations in gas phase atomic hydrogen concentration. Our present work uses this correlation. We grew diamond in plasmas with different concentrations of atomic hydrogen. The results of experiments concerning the correlation between the atomic hydrogen concentration, growth rates and nucleation densities are given in this paper.

2. Experimental details All the experiments were carried out using our planar microwave plasma source. This source uses the principle of distributed coupling of microwave power which allows large-area lateral homogeneous planar microwave plasmas to be produced. Microwaves penetrate into a metallic vacuum vessel through a long rectangular quartz microwave window. This quartz window is also part of the wall of a rectangular waveguide. In the vacuum vessel near the microwave window, a lateral homogeneous plasma layer is generated. This acts as the waveguide wall at this place. The thin physically active plasma layer near the microwave window is followed by a preferentially chemically active region. Therefore the plasmachemical properties of the source can be described by a onedimensional two-layer model which uses the distance to the microwave window as coordinate. Since most chemically active species are generated inside the active plasma region, the chemically active region exhibits decaying activity with increasing distance from the microwave window. The adequate match of this approximate model with the real situation was confirmed by various experiments: probe measurements of the electron density, actinometric optical measurements of active species distribution, optical measurement of the neutral gas temperature and calorimetric measurements with substrates at different distances from the microwave window. Details of these experiments, the plasma source itself, the experimental set-up for plasma diagnostics and the set-up for diamond deposition have been described elsewhere [1, 2, 4]. As in the previous investigations, the experiments described here were carried out in a test plasma source with 4 cm x 14 cm window dimensions. The substrates were placed in a flat, thin molybdenum boat of area 2 cm x 13 cm, which could be placed at different distances parallel to the microwave window. To ensure precise control of the substrate temperature during deposition, the boat could be heated directly by an alternating electrical current. Reaction gas flowed in the direction normal to the microwave window with flow rates of 370 standard cm 3 min 1 hydrogen and admixtures of a few per cent of hydrocarbon gas. All substrates

299

were cut from (l 1l) silicon wafers and scratched with fine grain diamond. Figure 1 shows stationary substrate temperatures, which were achieved by locating the substrates at different distances from the microwave window. Furthermore, the hydrocarbon admixture was varied. Hydrogen plasmas with no admixture are compared with plasmas with methane, methyl alcohol and ethyl alcohol admixtures respectively. A minimum distance from the microwave window of 20 mm was maintained to ensure that the substrates were always outside the electrically active plasma layer and therefore preferably heated by surface recombination of atomic hydrogen. The hydrocarbon admixtures were relatively high compared with typical diamond growth conditions, to make their effects on heating clearly visible. A variation of 40 mm in the distance from the microwave window results in a variation in substrate temperature of about 450 K. This temperature variation corresponds to a variation in atomic hydrogen concentration of 60%, since the temperature of substrates located in the chemically active region can be assumed to be proportional to the gas phase atomic hydrogen concentration [2]. Under the same assumption, maximum surface recombination rates can be estimated to be about 102°cm 2 s-1 (at x = 20 mm). Since the corresponding wall collision rate has to be of the same order of magnitude or higher, the

1000

900 3

800 0..

E

700 ..Io

6130

500

,

0

i

10

,

I

20

,

I

30

,

I

,

40

I

50

60

,

70

position [mm] Fig. I. Dependence of stationary substrate temperature on the distance between the substrate and the microwave window at a typical deposition pressure of 1800 Pa. Hydrogen plasmas with no admixture (©), admixture of 5% methane ([21), 5% methyl alcohol (O) and 5% ethyl alcohol ( I ) are compared. The difference between the substrate temperature and ambient temperature can be assumed to be proportional to the atomie hydrogen concentration.

300

A. Ohl et al. / Influence of atomic hydrogen gradients

minimum partial pressure of atomic hydrogen must be of the order of 10 Pa. That means that at least about 1% of the molecular hydrogen is dissociated. Obviously, admixtures of larger molecules result in lower concentrations. The alcohols in particular drastically reduce the atomic hydrogen concentration. This can be explained by their larger collision cross-sections for inelastic collisions, resulting in losses of atomic hydrogen by the formation of methyl radicals or ethyl radicals and molecular hydrogen. However, this is not the only possible explanation. Methane, methyl alcohol and ethyl alcohol exhibit growing total electron ionization collision crosssections in the order cited [5]. This can also reduce atomic hydrogen generation in the active plasma layer. In the diamond deposition experiments, the substrates were first located within the chemical active region. Then plasmas were ignited. From the beginning the reaction gas mixture flowed through the vessel. Admixtures of 2% hydrocarbon gas were always used. After the substrates had reached their recombination-induced stationary temperature, their temperature was raised by additional heating to a standard temperature of 750 °C. This constant temperature regime was used to ensure maximum compatibility between different deposition processes. It should be noted that this kind of experiment gives very direct information about the influence of varying atomic hydrogen contents and the corresponding concentration gradients. In particular, the influence of simultaneous variations in plasma parameters can be excluded. This is in contrast to many other deposition devices yielding spatial variations in deposition. Within the accuracy of the present results, remaining additional variations in the heavy particle components and neutral gas temperature could be neglected. To illustrate this in more detail, results of actinometric measurements concerning the atomic hydrogen distribution inside the active plasma layer are reported in Fig. 2. Measurements with different substrate locations are compared. At a distance of 20 mm between the substrate and the microwave window, the maximum atomic hydrogen concentration inside the active layer has already changed remarkably. However, the influence on the near exponential decay of the substrate temperature given in Fig. 1 with the opposite spatial dependence remains low since the substrates were always located in the tail of the spatial distributions. Therefore, by first approximation the dependence on position of the substrate temperature is really an effect of an atomic hydrogen concentration gradient. Plasma properties inside the active layer remain unchanged. This observation is in agreement with previous measurements [4]. The applicability of the actinometric method was tested by taking the intensity ratio between the Halph a

40 Q

d ix. 30 ._o

._~, ffl

2o .c_ ®

.c_ tO -~ 10 .~_ E

0

0

J

I

k

5

10

15

distance [mm] Fig. 2. Actinometricmeasurementsof the dependenceof the intensity ratio Ha~phaline to Ar-750, 4 nm line on the distance to the microwave window. Distance of the substrate to the microwave window were 20ram (• • •) and 60ram (O [] A): O • 5% ethyl alcohol admixture; [] • 5% methyl alcohol admixture; A • 5% methane admixture; p= 1800 Pa. The atomic hydrogen suppression indicated here has little effecton the exponentialdecay of the substrate temperature given in Fig. 1 with its opposite spatial dependence.

and the Hbeta lines, which gives a measure of relative changes in electron temperature [6, 7]. Correspondingly, the electron temperature seemed to be nearly constant inside the region of interest. Only a slight increase of a few per cent was observed.

3. Results and discussion

Several deposition experiments in which the atomic hydrogen content of the reaction gas was varied by varying the distance between substrate and microwave window were carried out. As already mentioned, the substrate temperature was brought to the same value of 750°C by additional heating in each case. However, different hydrocarbon admixtures and substrates with different preliminary scratching procedures were used. Process starting conditions were also changed in some experiments. Previous scratching procedures and deposition times were chosen so as to grow well isolated crystals. Scanning electron micrographs of the substrates were used to determine the medium diameters and the surface number density of the isolated crystals. Figure 3 summarizes results of the dependence of the medium crystal diameter on substrate location. In all cases the deposition time was the same. It can be seen

A. Ohl et al. / Influence of atomic hydrogen gradients

301

6 / / / /

"~

j~

6

E

©

"3

i

4

,.,._a

i

\\\\\

E "0

1/0

j~

E

,m

t~ "0

4

i

~2 E"

"O

o

"l

2

0

i

i

L

0

L

I

20

40

60

position [rnm]

i

i

i

200

400

600

temperature difference [K]

Fig. 3. Dependence of the medium diamond crystallite diameter on substrate location relative to the microwave window. The position dependence is similar to that of the spatial variation of the atomic hydrogen concentration given in Fig. 1, p = 1800 Pa: © ~ results of two experiments using 2% methyl alcohol admixture; • results of an experiment using 2% ethyl alcohol admixture.

Fig. 4. Plot of the crystal diameters vs. difference between the substrate temperature and ambient temperature. Since this temperature difference can be assumed to be proportional to the atomic hydrogen recombination, this plot indicates a square law correlation between the growth rate and atomic hydrogen concentration; depositions using methyl alcohol, same as in Fig. 3.

that this position dependence is similar to the spatial variation of the atomic hydrogen concentration. There is no difference in position dependence between methyl and ethyl alcohol, but growth rates were lower for ethyl alcohol. This can be explained by the lower oxygen atomic ratio in the deposition gas mixture. In some experiments the stationary substrate temperature achieved under the influence only of atomic hydrogen recombination was determined. A plot of the crystal diameters v s . this temperature (see Fig. 4) exhibits a square law correlation. This indicates a power law dependence of the linear growth rate normal to the surface on atomic hydrogen concentration. Assuming the crystallite surfaces to be similar to spheres a maximum power exponent of four can be expected. It should be noted that the deposited crystals were well faceted. They exhibited the well known pentagonal twin shape and some secondary nucleation. Crystallites deposited from ethyl alcohol exhibited broader Raman diamond peaks than deposits from methyl alcohol (about 20cm-1 v s . 10cm-1). This is in good agreement with the experience of many researchers, that the best crystal quality can be achieved with carbon and oxygen atomic ratios in equal quantities in the deposition gas mixture. Since the crystal morphology of the deposits did not change remarkably with distance, it can be assumed that there is a range of atomic hydrogen concentration in which the generation of hydrocarbon diamond growth

precursors can be controlled by atomic hydrogen concentration while the growth of non-diamond carbon phases is suppressed. The growth rate is then limited by the concentration gradients of the hydrocarbon growth precursors. The crystal diameters given in Fig. 3 are arithmetic averages of large numbers of crystals measured at different places on the substrates. A detailed analysis of the crystallite size distribution revealed that taking the arithmetic average is not completely correct since the real particle-size distributions are asymmetric. In Fig. 5 we show histograms of the particle-size distribution. Their shape is similar to the shape of particle-size distributions reported in ref. 8, which were also for (111) silicon substrates. A slight narrowing of the peak width with reduced atomic hydrogen content might be noted, but this effect needs investigation in more detail. Following ref. 8, this asymmetric distribution shape indicates the occurrence of intermediate species at the substrate surface during growth. Figure 6 is a plot of the densities of crystals on the substrate surfaces obtained in different deposition series v s . the distance from the microwave window. In contrast to the crystal diameters, these densities are independent of the atomic hydrogen concentration. This result can be understood assuming complete control of nucleation by pretreatment-induced nucleation sites. If substrates with different pretreatment procedures were used, entire films as well as isolated

302

A. Ohl et al. / Influence of atomic hydrogen gradients

4. Summary

40 C

b

30

8

1

20

i;

ii iI

10

~-iiiii !ii ~ii

17 ......

all f ] I"t r

0

2O

10 crystallite diameter [a.u.]

Fig. 5, Typicalhistogramsof crystallite-sizedistribution,with varying distance from the microwave window, corresponding to decreasing atomic hydrogen concentration: a 20mm, b 27.5mm, c 57.5mm; depositions usingmethylalcohol,sameas in Fig.3.

10

8

f~ eQ) "10

•"~

s

4

0

0

O

2

0

0

D

D

0

[]

[]

[]

I

0

20

~

I

40

J

The present paper reports results of experiments carried out to improve our understanding of the role of atomic hydrogen in diamond deposition. The planar microwave plasma source used for these experiments allows well defined activated particle concentration gradients to be produced. By placing substrates at different positions within this gradient but outside the physically active plasma excitation region, the atomic hydrogen concentrations could be varied without simultaneous variations in other plasma parameters. Considering that in this case surface heating is dominated by the surface recombination of atomic hydrogen [2], and assuming complete surface coverage, the atomic hydrogen concentration is then proportional to the substrate temperature. Variations in concentration of 60% were possible. Diamond was grown from hydrogen plasmas with small admixtures of methyl alcohol or ethyl alcohol. In all cases, during deposition the substrate temperature was held at the same value of 750 °C by additional heating. Previous scratching procedures of the (111) silicon substrates and deposition time were chosen so as to grow well isolated crystals. Scanning electron micrographs of the substrates were used to determine the medium diameters and the surface number density of isolated crystals. A position dependence of the medium crystal diameter similar to that of the spatial dependence of the atomic hydrogen concentration could be observed. The plot of crystal diameters vs. recombination induced substrate temperature indicates a power law dependence of the linear growth rate on atomic hydrogen concentration. Detailed analysis of the crystallite size revealed an asymmetric size distribution. A slight narrowing of the peak with reduced atomic hydrogen content might be noted, but further investigations are necessary. In contrast to crystal diameters, the local densities of crystallites are independent of the atomic hydrogen concentration. This can be related to complete control of the nucleation density by pretreatment-induced nucleation sites.

I

60

position [ram] Fig. 6. Densities of crystallites on the substrate surfaces, obtained in four differentdepositionseries at differentsubstrate positions with respect to the microwave window (different atomic hydrogen concentrations).

crystals could be grown using the same deposition process. The differences which can be seen in Fig. 6 have to be assigned to random parameter variations in substrate preparation and processing conditions.

Acknowledgment The authors wish to thank U. Kellner and D. Grtt for technical assistance. We are also indepted to the Daimler-Benz Research Laboratories Ulm for support in substrate preparation and Raman spectroscopy. This research was supported by the Federal Ministry of Research of the FRG under grant No. 03M2727F5. Responsibility for the content of the present paper is taken by the authors.

A. Ohl et al. / Influence of atomic hydrogen gradients

References 1 A. Ohl and M. Schmidt, Surf. Coat. Technol. 47 (1991) 29-38. 2 A. Ohl and J. R6pcke, Diamond Relat. Mater., 1 (1992) 243. 3 H. Wise and B. J. Wood, in D. R. Bates and I. Estermann (eds.), Advances in Atomic and Molecular Physics, Vol. 3, Academic Press, New York, 1967, p. 291.

303

4 J. R6pcke and A. Ohl, Contr. Plasma Phys., 31 (1991) 669. 5 R. Basher and M. Schmidt, Proc. I2th Int. Mass Spectrometry Conf., Amsterdam, August 26-30, 1991, in Adv. Mass Spectrom, 12, in press. 6 E. Lopata and J. Contrywood, J. Vac. Sci. Technol., A,6 (1988) 2949. 7 J. W. Coburn and M. Chen, J. Vac. Sci. Technol,, 18 (1981) 353. 8 M. Tomellini, R. Polini and V. Sessa, J. Appl. Phys., 70 (199l) 537.