Diamond and Related Materials, 2 (1993) 417 424 417

Diamond deposition in a bell-jar reactor: influence of the plasma and substrate parameters on the microstructure and growth rate

A. Gicquel and E. Anger Laboratoire d'lngbni~rie des Matbriaux et des Hautes Pressions, CN RS, Avenue Jean Baptiste CIbment, F-93430 Villetaneuse (France

M. F. Ravet Laboratoire de Microstructures et de Microdlectronique, CN RS, 196 Avenue H. Ravera, BP 107, F-92225 Bagneux (France)

D. Fabre and G. Scatena Laboratoire d'lngbnibrie des MatOriaux el des Hautes Pressions, CNRS, Avenue Jean Baptiste ClOment, F-93430 Villetaneuse (France

Z. Z. Wang Laboratoire de Microstructures et de Micro~lectronique, CN RS, 196 Avenue H. Ravera, BP 107, F-92225 Bagneux (France)-

Abstract

This paper describes diamond films that have been grown in a bell-jar (10 cm diameter) microwave plasma reactor. The microstruc- ture, purity of the films and growth rate are controlled by the initial nucleation density, substrate temperature and plasma parameters. The diamond membranes produced have qualities and microstructures which are compatible with X-ray lithographic mask applications.

I. Introduction

The microstructure, diamond purity level and growth rate of polycrystalline diamond films are strong functions of the nucleation density, deposition temperature and plasma local conditions (nature and concentrations of deposition and etching agents). The initial density of nucleation sites can be modified by varying the nature and crystallographic orientation of the substrate or by submitting the substrate to different physicochemical pretreatments (scratching, diamond powder ultrasonic impacts, ion bombardment). We report here results obtained in a low pressure microwave plasma reactor which allows deposition of diamond on substrates of diameter 2in. The design of the reactor allows the substrate temperature and plasma characteristics to be varied independently.

The investigation consisted of two parts. In the first part we studied the chemical contribution induced by ultrasonic impacts of powders to the diamond nucleation process by comparing the effects of three different powders: diamond, SiC and A1203. The roughness generated by the different powders was measured and the resulting diamond deposit was analysed. In the second part we focused our attention on the production of diamond films with controlled qualities and micro- structures. For a given composition of the plasma the

role of the substrate temperature in the diamond film purity level, microstructure and nucleation density has been shown, and for a given temperature the role of some global plasma parameters in the characteristics of the films tested. Using Raman spectroscopy, the films have been grouped in isoquality regions on a graph where the deposition temperature (local parameter) and percentage of methane in the gas (global parameter) were plotted. In addition, for a given set of deposition parameters we compared the microstructure and quality of individual diamond particles and continuous films. Finally, we report here measurements of the internal stress of continuous films and the visible optical transmission and thickness of diamond membranes made with films possessing a tensile stress. The intrinsic stress is discussed in terms of the sp 2 and structural defect concentrations. The quality of these membranes is compatible with X-ray lithographic and micromechan- ical mask applications.

2. Experimental details

The reactor was composed of a quartz bell-jar low pressure chamber [1]. The plasma was created in the cavity between an antenna and the substrate holder (5 cm diameter). A 1200 W Sairem 12 KE/T microwave

0925-9635/93/$6.00 i~;) 1993 - - Elsevier Sequoia. All rights reserved

418 A. Gicquel et al. / Diamond deposition in a bell-jar reactor

= " /~" / "d '< ' , v t j_..~,d._..__.,//. ' - , /

I I

(a)

<t , - \ , I X, ,1

0 1 , 0 0 2 , 0 0 VN

I

I I 0 1 . 0 0 2 . 0 0

IJN

(b)

~~ (c)

0 1 . 0 0 2 . 0 0 ul, l

Fig. l. Roughness deduced from a statistical treatment of AFM photo- graphs of Si(100) substrates treated by ultrasonic impacts with (a) A1203, (b) SiC and (c) diamond powders for 1 h.

TABLE 1. Roughness measurements of Si(100) surface pretreated by ultrasonic impacts with different 45 gm powders

As received A1203 SiC Diamond

R o (,~) 3-5 50 68 60 Rz (A) 350 480 760 Nucleation density 2.5 x 104 3.1 x 105 107 3 X 109 (particles cm 2) Ratio I 12 4 × 103 1.2 x 105

genera to r was used to p rov ide power to the p lasma. The subs t ra te was heated bo th by the p l a sma and by an add i t i ona l heat ing source. The t empera tu re of the subs t ra te was measured by an I rcon b i ch roma t i c p y r o m e t e r and var ied from 700 to 1000°C. The gases were in t roduced th rough a quar tz tube at the top of

Fig. 2. (a) SEM image of an Si(100) substrate treated by ultrasonic impacts with 45 gm diamond powder for I h. Note the diamond crystals implanted in the substrate. (b) SEM image of an Si(100) substrate treated by ultrasonic impacts with 45 gm diamond powder for 1 h and coated with a polycrystalline diamond layer which has been partially peeled off.

the bell-jar. Mi r ro r -po l i shed sil icon subst ra tes 275 lam thick were submi t t ed to 45 l~m ul t rasonic impacts of powders for 1 h p r io r to p lac ing them in the reactor. [2, 3] Different powders were tested: d i amond , SiC and A1203. M a c r o - R a m a n analysis of the deposi ts was per formed using a m i c r o - R a m a n spec t romete r equ ipped with a Di lor opt ica l mul t ichanne l ana lyser work ing with two exci ta t ion lines (514.5 and 647.1 nm). Scanning e lect ron mic roscopy (SEM) was per formed systemat i - cally and the growth rate was de te rmined by measure- ment of the weight as a funct ion of time. Atomic force

A. Gicquel et al. Diamond deposition in a bell-jar reactor 419

Fig. 3. (al, bl) Continuous films (diamond powder ultrasonic impacts) and (a2, b2) individual particles (SiC powder ultrasonic impacts) grown on Si(100) substrates under the same operating conditions except for the CH4 content of the feed gas (deposition temperature 850 ~C, 25 mbar, 600 W, flow rate 300 sccm (standard cm 3 min- 1), 4 h: (a) 2% CH4, (b) 0.5% CH 4.

microscopy (AFM) pictures were taken with a commer- cial instrument (Nanoscope III AFM system, Digital Instrument Inc.) operating in air. The roughness and surface morphology of the diamond films were visualized in the constant-force mode ( f= 10 8 N) with microfabri- cared Si3N 4 tips. Typical scan rates were 1 ~tm s-1. The AFM pictures were representative of a number of images taken for each sample with different tips, The thickness of the films was measured by interferometry in the reflection mode and has been found to be homogeneous within _+5% on a central circle of diameter 20 ram. The residual stress was determined for different diamond films grown on substrates of diameter 2 in by comparing the curvature measured on a Fizeau interferometer before and after deposition and using the Stoney

formulation. Because of the thickness gradient of the layers, the deflection was determined over a central circle 23 mm in diameter. The precision of the residual stress was estimated to be _+15%. Sch~ifer et al. [2] mentioned that plastic deformation of the silicon substrate may occur during deposition above T-- 850 °C. Therefore we have deduced the stress by removing the diamond film from the substrate by reactive ion etching in O 2 and measuring the curvature of the substrate afterwards. Only a substrate submitted to a 900°C deposition temperature suggested any residual deforma- tion. We obtained edge-supported membranes on 2 in samples from which a window of diameter 15 mm was chemically back etched in an H F - H N O 3 - C H 3 C O O H solution.

420 A. Gicquel et al. / Diamond deposition in a bell-jar reactor

(a) % CH 4 = 2 - . , '~

Isolated Particles

Macro-Raman : )~= 514,5 nm

/.

llm <_--- 1900 1700 1500 1300 I 100 900 700

Wavenumber (cv5 !)

(b) % C H 4 = 0 . 5 ~ M a c r o - R a m a n : ~ = 5 1 4 , 5 n m I [

, Isolated Particles II

, rl 1

C o n t i n u o u ~

500

1900 1700 1500 1300 1100 900 700 500 Wavenumber (cnil)

Fig. 4. Macro-Raman spectra corresponding to the deposits presented in Figs. 3(a) and 3(b).

Fig. 5. Continuous polycrystalline diamond films produced under the same experimental conditions except for the deposition temperature (25 mbar, 600 W, flow rate 300sccm, 0.5% CH4, 4h: (a) 900°C, (b) 750 °C.

3. Results

The quality and microstructure of the films are strong functions of the nucleation density, local characteristics of the reactive medium at the plasma-surface interface and deposition temperature.

3.1. Influence o f the Si( lO0) surface state Under the same deposition conditions, a variation in

the nucleation density induces very different microstruc- tures. The nature of the 45/am powder used when

submitting the Si(100) substrate to ultrasonic impacts was varied. Under the same deposition parameters, ultrasonic impacts by A1203 produced a very low nucleation density (3 × 105 particles cm-2), SiC pro- duced a low nucleation density (107 particles cm 2) and diamond produced a very high nucleation density (3 × 10 9 particles cm 2). The roughness, deduced from a statistical treatment of the AFM pictures, indicated a higher value of Rz (maximum peak-to-valley ratio) and showed steeper peaks when using the diamond powder than when using SiC or Al20 3 (Fig. 1, Table 1). The

A. Gicquel et al. /' Diamond deposition in a bell-jar reactor 421

;I

(a) T = 9 0 0 ° C /

(b) T = 750°C

Macro-Raman : )x= 514,5 nm

x

\

1900 1700 1500 1300 1100 900 700 500

Wavenumber(cni 1)

Fig. 6. Macro -Raman spectra corresponding to the films presented in Figs. 5(a) and 5(b}.

1100

(J o._.

Y.

1000

g00

800

700

600

500

" "~, I t 4 J

J II \ . 1

Pol~er Like

Diamond Like Carbon (+Ions)

! u

0 1 2

Methane Percentage

Fig. 7. Raman isoquality regions as a function of the CH 4 content of the feed gas and the deposit ion temperature (25 mbar, 600 W, 300 sccm, 4h).

difference in Rz between alumina and SiC was less than 25%. The fact that diamond nucleated on the substrate when SiC was used for pretreatment of the silicon substrate, while very few diamond crystals were formed when using alumina, indicates that the diamond nuclei formation cannot be attributed only to the physical defects generated by the ultrasonic impacts.

200

[] %CH4 = 0,5

* %CH4 = 1

[ ] %CH4 = 1,5

o %CH4 = 2

100 o []

D u

o i

600 700

[ ]

[ ]

[ ]

o •

o

[]

[ ] [ ]

[ ] [ ]

( 32/4g/h.cm = 0.1 ,um/h )

u i u

800 900 1ooo ~oo

Temperature ( °C )

Fig. 8. Polycrystalline d iamond film growth rates as a function of the deposit ion temperature for different CH 4 contents of the feed gas.

Diamond formation must be attributed to a chemical effect. As a matter of fact, SiC can superficially graphitize or dissociate (if ultrasonic impacts provoke local temperatures as high as 1400 K) and generate disordered carbon seeds in the induced defects on the surface [3, 4]. These disordered seeds later transform into sp 3 carbon seeds on the surface during the first stages of deposition (dangling bonds reacting with H atoms) [5, 6]. The few crystals observed on the "as- received" Si substrates and those treated with alumina powder can be attributed to superficial disordered seeds provided by dust or uncontrolled carbon content impurities. The difference between the SiC and diamond powders could probably be attributed to a number of parameters. During the ultrasonic treatment with diamond powder, both amorphous and crystallized carbon seeds are able to be incorporated into the induced superficial defects of the surface. In addition,

422 A. Gicquel et al. / Diamond deposition in a bell-jar reactor

diamond crystals are implanted in the silicon substrate with a density of 3.3 x 109 particles cm -2 as seen by SEM (Fig. 2(a)). Furthermore, SEM of a silicon surface where a deposited diamond layer has been partially peeled off (Fig. 2(b)) showed evidence of holes, which probably correspond to the reciprocal of the grains observed in Fig. 2(a). Homoepitaxial growth is then likely on these implanted diamond grains. On the surface the net result is a mixture of physical defects filled with amorphous carbon compounds and diamond grains. Both seeds and (mainly) diamond grains can contribute to the surface roughness and diamond nucleation process. However, we cannot exclude the possibility that the holes observed on the silicon surface could be due to heterogeneous nucleation of diamond embedded in a silicon carbide layer grown during the first stages of deposition.

3.2. Influence of the local conditions at the plasma- surface interface

The nucleation density depends on the initial nucle- ation site density and local deposition conditions. The local conditions include the surface temperature and local plasma parameters. They strongly influence the quality of the deposit, its microstructure and its growth rate. The local plasma parameters depend on one hand on the history of the molecules and species in the plasma bulk (composition of the feed gas, dissociation efficiency, etc.) and on the other hand on the substrate boundary conditions (recombination processes, mass transfer, energy accommodation) [7]. Determining the key parameters then requires measurements of a number of local parameters which are not easily accessible experimentally. We report here the influence of the percentage of CH4 in the feed gas and of the deposition temperature on the microstructure and quality of the films and on the growth rate. These parameters were varied independently.

3.2.1. Quality of the films As reported in the literature, the quality of the films

was enhanced by lowering the CH 4 concentration in the feed gas [8, 9]. Spectroscopic measurements of the plasma revealed that the atomic hydrogen concentration increased as the CH4 percentage in the feed gas decreased [ 10]. Obviously, since the CxHr concentration was simultaneously decreased, the ratio of the concen- tration of the etching agent (H) to that of the deposition agent (CxHy) had increased as the hydrocarbon content in the feed gas was decreased. The enhancement in quality of the films can be attributed to a decrease in the graphite deposition at the grain boundaries (inducing secondary nucleation there) or a reduction in the secondary nucleation on each entity on which there is a codeposit of graphite. A comparison of the role of

the methane concentration in the microstructure of the deposits has been made for two nucleation densities which were obtained by preparing the substrate with ultrasonic impacts with either SiC or diamond powders. For both nucleation densities a decrease in the CH4 in the feed gas from 2% to 0.5% obviously lowered the secondary nucleation on the individual entities (Fig. 3). A decrease in the microcrystalline diamond component at 1140 cm-1 in the Raman spectrum is simultaneously observed (Fig. 4). Thus for the individual entities (aggre- gate or pseudocrystal) the enhancement of the quality is attributed to a lowering of the number of diamond nuclei necessarily formed on the non-diamond phase, since the codeposition of graphite has decreased. The deposition of graphite at the boundaries between aggregates (or crystals) is also simultaneously reduced. In a similar way a decrease in the substrate temperature from 900 to 750°C reduced the secondary nucleation (Figs. 5 and 6) on high nucleation density diamond films. Since the temperature and plasma parameters are coupled, Raman isoquality regions have been traced in graphs where the CH 4 percentage in the feed gas and the temperature are plotted (Fig. 7). If the data were available, this diagram should be constructed with the local concentration of the active species instead of the CH4 concentration.

3.2.2. Growth rate The growth rate is a strong function of the deposition

temperature. In the range of operating conditions presented here, a maximum in the growth rate has always been observed at a deposition temperature of 900°C (Fig. 8). This behaviour has been attributed to the competition between the deposition and etching reactions, which are temperature-activated processes. Note that the growth rate measurements are not able to separate the contribution of the diamond phase from that of graphite. Since at the highest temperatures the graphite phase is preferentially formed [8], the net decrease may probably be attributed to an enhancement of the etching reaction rates of the graphite phase.

3.3. Stress- and edge-supported membranes The residual stress in the very high nucleation density

films was found to be either compressive (negative) or tensile (positive) according to the deposition parameters. The residual stress is the result of the intrinsic stress related to the microstructure and of the compressive thermal stress due to the mismatch of the expansion coefficients of silicon and diamond. The thermal stress can be considered as relatively constant and equal to -240 _+ 35 MPa over the deposition temperature range, in agreement with Windischmann and Epps [-11]. The intrinsic stress was then deduced from the residual stress by substracting this value. As reported in Table 2, for

A. Gicquel et al. / Diamond deposition in a bell-jar reactor

TABLE 2. Stresses and thicknesses of d iamond films and apertures of corresponding membranes

423

Sample T (°C) %C H 4 % 0 2 e (nm) a,o, (MPa) aio~ (MPa) Membrane aperture

A 750 0.5 0 1200 - 220 + 20 Silicon still not etched B 750 1.5 0.8 1200 - 285 45 Broken C 750 3 0 1100 - 306 66 Broken D 850 0.5 0 820 + 68 + 308 Taut E 900 0.5 0 940 - 115 + 125 Taut F 800 0.75 0 933 + I + 241 Taut

samples A, B and C at 750 °C the compressive intrinsic stress increased with increasing CH4 ratio when all other conditions were held constant. Consequently, the compressive intrinsic stress increased with increasing concentration of sp 2 bonding. The corresponding mem- branes were crumpled, brittle and immediately broke. This suggests that the sp 2 bonds are at the origin of the compressive stress feature in CVD polycrystalline diamond films. For samples A, D and E, deposited under the same 0.5% CH4 ratio and plasma parameters, the tensile intrinsic stress presented a maximum at 850°C as the temperature was lowered from 900 to 750 °C. The resulting decrease in the s p 2 bond concen- tration as the temperature is decreased, as mentioned above, should provoke an increase in the tensile intrinsic stress. Therefore the variation in the intrinsic stress must be attributed to additional factors. Observation of both the morphology (Fig. 5) and fluorescence observed by Raman spectroscopy (Fig. 6) of the films produced at 900 and 750°C suggests that a higher concentration of structural defects (such as pseudopen- tagonal symmetry induced by twins [12]) is produced at higher temperature. These structural defects are likely responsible for the tensile stress. Thus competition between compression due to an increase in the s p 2 bond concentration and tension due to an increase in the structural defect concentration is occurring as the temperature is increased. Finally, sample F, which was obtained at lower temperature than samples D and E (lower s p 2 concentration) but at a higher CH4 concen- tration (higher s p 2 concentration), had a tensile intrinsic stress value very similar to that of sample D. The membranes made from samples D, E and F, having a tensile intrinsic stress, were quite taut. Their measured optical transmission was 60%-65% at 630 nm and their roughness was 30 nm (RQ).

4. Conclusions

Polycrystalline CVD diamond films have been pro- duced in a bell-jar microwave plasma reactor. The quality and microstructure were controlled by varying

the surface state preparation, deposition temperature and plasma parameters. By preparing high quality diamond films presenting a low concentration of sp 2 bonds and a somewhat high concentration of structural defects, we were able to produce taut membranes. Their mechanical and physical properties (roughness, stress, transparency) were compatible with X-ray lithography for applications in microelectronics and micromechanics. Finally, by controlling the concentration of structural defects, a film with almost no intrinsic stress has also been produced.

Acknowledgments

The DRET is gratefully acknowledged for financial support of this work. Dr. R. Chiron from the LPMTM is thanked for his help in the SEM characterizations, Dr. J. Chen and Dr. J. Bournei× from the L2M for the optical transmission and membranes aperture measure- ments respectively, J. J. Dubray for helpful discussions and Dr. C. Scott for the English revision and helpful discussions.

References

1 A. Gicquel, Nato Advanced Research Workshop on Microwave Discharges. Fundamentals and Applications, Plenum, New York, in press.

2 L. Schfifer, X. Jiang and C. P. Klages, in Y. Tzeng, M. Yoshikawa, M. Murakawa and A. Feldman (eds.), Applications of Diamond Films and Related Materials, Elsevier, Amsterdam, 1991, p. 121.

3 B. Goss Levi, Phys. Today, November (1991) 18. 4 G. J. Exardos, M. S. Donley and R. H. Geiss (eds.), Microheam

Analysis, 1987, p. 125. 5 P. Koidl, Brite/Euram Workshop on European Diamond Technology

]or the '90s, Brussels, December 1991, Diamond Relat. Mater., 1 (1992) 1065.

6 J. J. Dubray, C. G. Pantano, M. Meloncelli and E. Bertran, in R. d 'Agostino (ed.), J. Vac. Sci. Technol. A, 9 (1991) 3012.

7 A. Gicquel, M. Cappelli, A. Y. Chang and R. K. Hanson, in R. d 'Agostino (ed.), Proc. Int. Symp. on Plasma Chemistry, ISPC 9, Pugnochiuso, 1989, Vol. III, pp. 1613 1618.

8 A. Gicquel, C. H6au, D. Fabre and J. Perri6re, Diamond Relat. Mater., 1 (1992) 776.

9 K. Kobashi, K. Nishimura, Y. Kawate and T. Horiuchi, in

424 A. Gicquel et al. / Diamond deposition in a bell-jar reactor

K. Akashi and A. Kinbara (eds.), Proc. Int. Syrup. on Plasma Chemistry, ISPC 8, Tokyo, 1987, Vol. 4, pp. 2475-2480.

10 A. Gicquel, R. Lamendola, Y. Breton and L. Saint Onge, in preparation.

I l H. Windischmann and G. F. Epps, Diamond Relat. Mater., 1 (1992) 656.

12 B. E. Williams, H. S. Kong and J. T. Glass, J. Mater. Res., 5 (1990) 801.