Workshop on diamond thin films 771

Stephen 0. Hay, Ward C. Roman, Meredith B. Colket III

United Technolo ‘es Research Center East Hart ord, CT 06108 i”

The driving force behind the strong interest in diamond deposition processes is the outstanding combination of unique natural properties of this material. A wide variety of techniques has been

including hot filament, thermal plasma, CVD, PACVD rf, dc, beam, laser beam, oxyacet lene torch, and numerous 6 ybrid

dual-beam con Thus, there are many routes available or producing diamond coatin T s in individual crystals, amorphous coatings, polycrystalline films or single crystal fi lms

under con~tio~ far removed from the th~~~~~y stable region nominally associated with diamond growth. CVD of diamond coatings from hydrocarbon containing gases can have an almost infinite number of compositions and structures; each with differing amounts of sps (diamond) and sps

6, graphite) bonding. This variation has contributed to confusion both in the working definition of amond coatings and in understanding the controlling processes of form@ these films. In fact, the

mechanisms involved in the gas phase processing, the nucleation and growth structures, aud especially their correlation are poorly understood. Interestingly, this limitation has not hindered the headstrong rush for commercial development.

Commercial deployment of this nascent technology hinges on the ability to form uniform coatings at high deposition rates over large and irregular surface areas. Intrinsic to the ability to scale existing techniques for required deposition rates over large areas is detailed mechanistic knowledge for both the gas phase and gas surface reactions which result in high purity deposition of the proper morphology. In the case of diamond deposition, the focus of most studies has been to document the effect of varying macroscopic physical observable8 input energy, flow rates, total pressures, etc.) upon coating

6 ~~acte~sti~ such as purity, morp ology, hardness, and adhesion. Chemical kinetrc modeling has concentrated on boundary conditions. Modelers generally possess detailed knowledge of both initial conditions and the final state but have an incomplete picture of the transitional process.

To address this dearth of experimental data, a variety of remote, nonintrusive optical techniques has been applied to probe the species, concentratrons and ener etics of these transitional re ‘ens. We have applied Coherent Anti-Stokes Raman Spectroscopy ( e ARS), using a narrowban , scanned P collinear configuration to measure temperatures and relative concentrations and to detect species in low pressure CVD of poly~yst~ne diamond. CARS rne~~~ts were obtained for methane, hydrogen and acetylene in either or both a rf plasma reactor and a hot filament reactor.

In the rf PACVD experiments, a mixture of 1% CH4 in Hr was used at a total pressure of 5 Torr. The rf power input to the plasma was 300 Watts and the Hs and CH4 flow rates were 99 and 1 seem, respectively. As acetylene (CsHr) has been proposed as an intermediate in diamond growth, it was selected for the initial series of measurements. In the absence of rf power, a sensitivity of 5 mtorr was observed, while in the plasma downstream of the rf coils, no observable signal attributable to CsHr was evident. This places an upper limit to conversion of methane to acetylene at 2046, a figure reprinting the observed sensitivity to CrHr.

In the hot filament reactor, gas flow was 200 seem of 1% CH4 in Hs at a total pressure of 150 Torr. Under these conditions! CsHp was detectable. Absolute concentrations were not calculated but the observed spectra are withm an order of magnitude of our sensitivity limit. This allows estimation of the CrHs partial pressure near the substrate as 5-50 mTorr or from 0.33 to 3.3% conversion from methane. In view of this low conversion percentage, the absence of a signal in the rf experiments must be taken as inconclusive.

CARS spectra of methane were also obtained in both reactors. In the rf reactor, under similar conditions to those described previously, the methane relative concentration decreased to 25% as the rf Power was increased from zero to 400 Watts. In the hot filament reactor, CH4 CARS signal profiles were obtained as a function of axial distance from the hot filament, and parametrically as a function of filament temperature. Comparison of these pro&s, in which the observed signal decayed monoto~~ly as the filament was approached and increased monoto~~y downstream of the filament, was made with theoretical calculations. This comparison showed that the variations were attributable to temperature/pressure effects and not to chemistry.

In an effort to determine if the observed depletion in the rf plasma was similarly attributable, the CARS signal of hydrogen was observed as a function of tial distance downstream of the rf coil

712 Workshop on diamond thin films

centerline and parametrically a8 a function of rf power. In contrast to expected behavior in the thermal hot filament reactor, little rotational excitation WBB observed in the plasma. Rotational temperature8 were assigned to hydrogen based upon comparison with theoretically derived spectra. At 450 Watt8 rf power, rotational temperature8 of 340 K were observed 4 to 6 cm downstream of the coil, the re ‘on where the 25% decrease in CH4 wa8 observed. Little or no density fluctuation8 will accrue due to t P e8e temperatures, indicating that the observed depletion in methane signal is attributable to decomposition or chemical reaction in the plasma.

In summary, CARS is applicable to reactant specie8 (CH4) axial profiling in both reactors, but cau be limited by sensitivity in the detection of intermediate or product specie8 (C8H8). In addition, CARS thermometry can be utilized to profile the rotational temperatures of selected species.

Research supported in part by DOE Basic Energy Science8 Research-Grant BES-ER-13560

Effect8 of Filament Conditioning and Temperature on the Compozition ofGase8intheFilamentA88i&dCVDdDiamond

L. Heatherly, Jr. and R.E. Clausing

Metals and Ceramics Division Oak Ridge National Laboratory

P. 0. Boz 9008 Oak Ridge, TN 37831-6093

Introduction

Filament as&ted chemical vapor deposition (FACVD) is a very useful and easily controlled technique used to study diamond film deposition. Although quite dimple compared to many other deposition processes, it still contains a number of important proce88 parameter8 that deserve attention. The monitoring of these parameter8 can be very helpful in determining the condition of the process and relating the growth processes to the morphology and quality of the diamond films.

This paper describe8 change8 in the filament characteristics and the gas composition during the conditioning and operation of a tungsten-tun

! sten

diamond film8 using the hot filament assiste carbide filament. In addition, the deposition of

CVD technique are analyzed. The variation of the electrical resistance of the filament a8 the tungsten carburized wa8 determined by monitoring the applied voltage and current over a period of time. The composition of the gas from the chamber was observed during the filament carburization and film deposition periods. The gas composition wa8 also studied a8 a function of filament temperature. This paper describes the result8 of these measurements and discusses how these results may be used to better understand and control the diamond deposition prOCe88.

Exuerimental

A schematic drawin of the FACVD apparatus is shown in Fig. 1. A premixed gas containing CH4 and hydrogen flowe j through a 6 mm diameter quartz tube and acro88 a hot filament. The substrate wa8 positioned near the filament and wa8 heated by radiant heat from the filament. The exhaust gas wa8 pumped through a 25 mm diameter quartz tube using a mechanical roughing pump. A differentially pumped quadrupole residual gas analyzer wa8 used to sample the chamber gas. The voltage drop acro88 the filament, the filament current and the output of the residual simultaneously monitored a8 a function of time, and the data wa8 stored on magnetic

a8 analyzer were

11/34 computer and CAMAC interfacing. isks using a PDP

Result@

The electrical resistance of a typical new filament a8 a function of conditioning time is shown in Fig. 2. During the first phase of the conditioning period most of the carbon in the gas mixture reacted with and wa8 absorbed by the tungsten filament. This resulted in an increase in the electrical resistance of the filament. Analysis of the chamber gasee of acetylene in the gas durin

3) indicated depletion of the methane and absence

wa8 removing carbon rather conditioning period, confirming that the filament

om the gas mixture. The amount of carbon absorbed from the gas phase wae estimated and compared to the amount required to carburize the filament. After a period of approximately thirty minutes, the second phase of the conditioning process began. The carbon uptake by the filament decreased and eventually became negligible compared to the uptake of