Evaluating the Influence of Growth Parameters on Cvd Diamond Deposition Using Factorial Analysis

The deposition of diamond films by low pressure CVD methods has been demonstrated using a number of thermal and plasma assisted CVD techniques. Addition of oxygen to the methane-hydrogen feedstock has been reported to improve the quality of diamond deposited and decrease the temperature at which diamond deposition is feasible. Oxygen also effects the growth rate and possibly the uniformity of the film. In this paper an experimental design methodology, factorial analysis, is presented which is, in principle, applicable to any CVD process. The factorial analysis identifies the most influential growth parameters and any interactions between independently variable parameters. f i e analysis has been applied to diamond growth using CH4 0 2 -Hz mixtures in a microwave assisted CVD process.The effect of oxygen addition on growth rate and film quality is discusst;d. The chemical vapour deposition of diamond films has been demonstrated by a wide range of methods including thermal filament, DC discharge, rf and microwave plasma assisted CVD, oxyacetylene and plasma torch 111. Growth conditions for the deposition of diamond requires substrate temperatures to be monitored between 700 and 1000T using a gas feedstock of 0.35% methane diluted in hydrogen. Other parameters such as the excitation source (plasma or filament temperature), its distance from the substrate, flow rates and pressure also influence the growth rate and the quality of the deposited film. Some general conclusions may be drawn from the extensive range of studies that have been made in an attempt to 'optimise' the diamond growth process. Firstly, the amount of amorphous-graphitic carbon that is co-deposited with the diamond increases with methane concentration. The amorphous-graphitic carbon content can be reduced and hence the quality of the diamond film increased at higher substrate temperatures up to some threshold value of approximately l000T. A compromise between growth rate and film quality occurs as the growth rate decreases with decreasing methane partial-pressure. It has been proposed that the addition of oxygen to the methane-hydrogen feedstock has the effect of reducing the temperature at which good quality diamond films can be grown ie.below 70095 /2/. This is advantageous in the coating of temperature sensitive substrates, with the possibility of opening up new application areas such as hard coatings on polymeric materials. It has also been postulated that oxygen additions permit the deposition of high quality films using extended methane concentrations which would overcome the sacrifice in deposition rate. Hirose and Terasawa I31 have observed qualitative improvements in film quality and growth Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jp4:19912110 JOURNAL DE PHYSIQUE IV rate using a range of oxygen containing precursors such as alcohols, aldehydes and ketones. Subsequently, Kawato and Kondo 141 reported experiments using a similar thermal filament CVD reactor and feed stocks of CH4 Hz Oz mixtures.These authors observed that the addition of 0.4% oxygen to 1.6-4% methane /hydrogen increased both growth rate and diamond purity, compared to experiments performed without oxygen. A more quantitative evaluation of the effect of oxygen additions was performed by Saito et all51, who used CH4 Hz Hz Omixtures in a microwave plasma to deposit diamond thin films. Enhanced growth rates of 1-5pmIhr were reported and maximum growth rate as achieved when the II2 OlCH4 ratio was between one-third and a half. These deposition parameters produced better crystallinity and reduced non-diamond impurities. These authors also investigated the etching of diamond and graphite using microwave discharges of &-I320 mixtures 151. Addition of water vapour to the discharge was found to double the etch rate of graphite. This observation was used to infer that the primary role of the oxygen containing addition is the preferential etching on graphite carbon which is co-deposited with the diamond hence improving the quality of the film. Liou et a1 121 have also used microwave assisted CVD to investigate the influence of oxygen additions to diamond forming plasmas of methanehydrogen mixtures over the range of substrate temperatures from 300°-100093. In the absence of oxygen the amorphous-graphitic content of the deposited films was found to increase above 900T. Addition of a few percent of oxygen or less, was found to improve film quality, growth rate, and to possibly extend the deposition region. At temperatures below 50093 the successful growth of diamond was attributed to the etching of graphite deposits by oxygen. Below 50095 a white soot was deposited in the absence of oxygen. In this paper we present the results of a statistical experimental design method, factorial analvsis 16.71. which has been amlied to the microwave assisted CVD de~osition of diamond film; f r o m ' ~ 4 Hz 02 mixtu;is. The design method provides a quantiiative ranking of the most important parameters and interactions between them. 2.Experirnental Details 2.1 The MACVD reactor and substrate pretreatment The schematic desrgn of an MACVD reactor is shown in figure 1 and consists of a 1kW 2.45GHz microwave source, heated platen capable of holding substrates up to 3" in diameter. Pressure is monitored in a side-arm using an absolute pressure capacitance manometer. In this experiment no external heating was used and the substrate temperature was measured via a chromel-alum01 thermocouple attached to the backface of the molybdenum substrate holder and also by an optical pyrometer. Gas flows were controlled by rf suppressed mass flow controllers which were premixed before admission into the reactor. Films were deposited onto 30mm silicon wafers which had been polished mechanically for one hour using a lpm grade diamond polishing paste. The substrates were cleaned in methanol, acetone and a 1.1.1 trichloroethylene vapour degreasing bath prior to weighing. The substrates were mounted into a fixed substrate holder to ensure constant positioning of the wafers. Deposition times were 22 hours and the substrate platen was positioned just below the plasma ball during deposition. 2.2 Experimental design using factor analysis The dependence of diamond film quality and deposition rate on parameters (factors) such as pressure, substrate temperature, methane concentration, oxygen concentration, flow rates (residence time), microwave input power etc, makes CVD diamond growth an ideal experiment for factor analysis. The traditional approach which varies one factor at a time in isolation to monitor the response does not allow the interaction between parameters to be fully assessed. Factorial experiments consist of a series of trials including combinations of parameters at set levels. In the simplest case, as used here, the factor-levels are either high or low value. Therefore a full factorial experiment involving four factors at each of two-levels would require Dummy load n Axial applicator 2.45GHz microvave Forwardlreverse Mass Flow Controllers Figure 1. Schematic diagram of the microwave CVD reactor. 2 4 or 16 trials to be performed. The outcome of each combination produces a 'response' which is used to estimate the effect of each factor. In this paper we have chosen to study four factors, pressure P (at 20 and 50 mbar), microwave power M (at 500 and 800 W), methane flow rate C (at 1 and 3 sccm), and oxygen flow rate 0 (at 0.5 and 1.5 sccm). The combination of factors for the full factorial experiment are shown in Table l(a). Under each factor are 8 runs with the factor at a high (+) level and 8 runs at a low (-) level. The 'factor-effect' is defined as a change in the average response (weight gain) arising from the change in factor-level combinations, ie effect = (mean response at high level)-(mean response at low level) The full factorial would provide information concerning 4 main (individual) factors, 6 two factor interactions, 4 three factor interactions and one four factor term. The level of multi factor effect is found by multiplying the level of its component factors eg. the term PC0 would be high (+) if P(+l)xC(l)xO(-1) and low (-) if P(l)xC(l)xO(1) for instance. The number of experiments that have to be performed can be halved if certain information can be sacrificed. If this is done the same factors and their effects can be investigated, however it is no longer possible to identify the effect of each individual term and 'confounding' occurs. Depending upon which 8 runs are selected from the full factorial several factors will have the same high and low levels in each run and so it is impossible to attribute the response to either uniquely. The experiment is referred to as a fractional or half factorial experiment (112 (2)4). Table l(b) shows the interactions that can be estimated by this method. The systematic choice of combinations used here is designed to ensure that the main effects are confounded with multi-factor effects having negligible values. In general, effects including two, three or four factors are likely to have increasingly smaller values if the factors are independently variable which is a prerequisite of factorial analysis experimental design. In these experiments the response was determined by: the weight gain of each substrate due to C2-918 JOURNAL DE PHYSIQUE IV br (a) full ((2)4) factorial and (b) the half (la (2)4) factorial used . the introduction of systematic errors. (+)I (-) denote high 1 low Table 1 Experimental designs in "randomised" order to avoi factor levels.