Chapter 6 -Results for C/H/N Systems

“I don’t know what it was, but it had either prussic acid or white lead in it. I should fancy it was prussic acid, as she seems to have died instantaneously.”

Oscar Wilde ‘The Picture of Dorian Gray’

6.1 Introduction

Nitrogen is the most common impurity found in natural diamond [[1]]. Natural diamond is a good electrical insulator, however, when the doping level (the amount of impurity atoms, e.g. nitrogen, incorporated into the diamond lattice) increases, the electrical properties of the diamond change to those of a semiconductor. The potential of diamond to be both insulator and semiconductor has led many people to conceive of novel diamond-based micro-electronic devices. Much interest has also been expressed in making so-called ‘cold cathode’ electron emitters [[2]-[3][4]], and field emission displays from doped diamond [[5]].

The modern era of CVD diamond growth allows diamond to be deposited as a thin film, on Si wafers (as used in the micro-electronics industry). In order to make diamond semiconducting, it is necessary to introduce a second element (the dopant), with a different electronic structure to carbon, into the diamond lattice. Attention has mainly focused on group 13 or group 15 elements (such as boron [[6][7]-[8]], nitrogen [[9][10]-[11]], or phosphorus [[12]-[13][14]]) as possible dopants for CVD diamond. This may be done by post-deposition ion implantation [13,[15],[16]], but a more attractive approach would be to carry out the doping during growth by the addition of a dopant gas to the CVD process gas mixture (typically CH₄/H₂).

The presence of nitrogen in the CVD reactant gases is known to affect the film morphology, quality, and growth rate when producing diamond films by MWCVD [9,[17],[18]]. Nitrogen (as N₂) added to a typical CH₄/H₂ CVD mixture in the 10-100 ppm (parts per million) range was found to promote the formation of (100) surface texture, and produce a ~ fivefold increase in growth rate over that obtained from CH₄/H₂ alone [9,[19],[20]]. At higher levels of nitrogen input the growth rate [10,19,[21]] and film quality [19,21] (assessed by Raman spectroscopy), was found to decrease. In addition to examining surface morphology, many previous studies of nitrogen addition to diamond CVD reactants have focused on the incorporation of nitrogen into the growing film [[22],[23]]. This was found to be quite inefficient, in that only a small amount of the input nitrogen (relative to the input carbon) was incorporated into the film. There is, to date, no firm consensus as to whether gas-surface, or gas-phase chemical reactions (or a combination of both) are necessary to account for these observations. Most studies have concentrated on the properties of films deposited in the presence of nitrogen, and hence neglected the fate of nitrogen in the gas-phase, prior to incorporation into the growing film. In order to supplement the data concerning properties of diamond films deposited in the presence of nitrogen, this chapter presents a study of the gas-phase composition during MWCVD with nitrogen present in the CVD feed gas. The technique of in situ molecular beam mass spectrometry (MBMS) [[24],[25]] has been used to determine concentrations of stable and radical species in the plasma, and the results have been published in Diamond & Related Materials [[26]]. This study parallels previous work carried out in the Bristol laboratories for hot filament CVD by Tsang et al [[27]]. Some growth experiments have also been carried out, and the amount of nitrogen incorporated into the film has been examined using SIMS (Chapter 2), and the results published in Thin Solid Films [[28]].

6.2 Effect of applied Microwave Power

A carbon to nitrogen ratio, C:N=1:1 was used in this study. Nitrogen was added to a typical 1% C:H₂ CVD gas mixture in one of the following forms: N₂, NH₃, CH₃NH₂, or HCN. For the experiments using nitrogen as N₂ and NH₃, CH₄ was used as the carbon source. CH₃NH₂ and HCN required no extra carbon to be added to the system. These four nitrogen-containing gases were chosen as the nitrogen occurs in a variety of chemically bonded forms (i.e. NºN, N-H, C-N, and CºN). The effect of microwave power on the gas phase composition was investigated initially, as this parameter seemed to have a marked effect on the gas-phase composition measured by MBMS, when hydrocarbon/H₂ source gas mixtures were used (Chapter 4).

The same techniques were used for the MBMS determination of gas-phase composition as those of Tsang et al [27,[29]], who performed an analogous study of the effect of nitrogen additions during HFCVD diamond growth. There were, however, some slight differences in methodology between that work and the present study. Chiefly, we used the modified data correction procedure given in Section 3.8.5. As noted in that section, this was to ensure that the background corrected data did not become negative. Additionally, electron ioniser energies had to be chosen for the detection of the nitrogen containing species. These are given in Table 6.1.

*m/q*	*Species detected*	*Electron energy used for detection (eV)*
2	H₂	16.0
15	CH₃	13.6
16	CH₄	16.0
17	NH₃	13.6
26	C₂H₂	13.2
27	HCN	16.0
28	C₂H₄	11.9
28	N₂	30.0
31	CH₃NH₂	13.6

Table 6.1 Electron energies used to detect species for all experiments in this chapter.

One complication arises from the fact that C₂H₄ and N₂ have equal masses (28 a.m.u.), hence the mass spectrometer measures a combined C₂H₄+N₂ signal, I(C₂H₄+N₂), at 30.0 eV. In order to determine I(N₂) separately, an additional correction to those described in Chapter 3 must be carried out to subtract the C₂H₄ signal from the combined signal, I(C₂H₄+N₂). At 11.9 eV C₂H₄ can be detected (as IP(C₂H₄) = 10.51 eV), but N₂ cannot (as IP(N₂) = 15.55 eV). Therefore the 11.9 eV m/q=28 signal arises solely from the species C₂H₄. During stable species calibration, we measure signals for a known concentration (e.g. 1%) of C₂H₄ at 11.9 eV (as in, for example, Chapter 4) but also at 30.0 eV (the energy at which the combined signal, I(C₂H₄+N₂), is measured). Therefore, the C₂H₄ signal at 11.9 eV can be related to the C₂H₄ signal at 30.0 eV, and a subtraction made to correct the 30.0 eV I(C₂H₄+N₂) data for the C₂H₄ contribution, yielding I(N₂).

This procedure was found to be satisfactory, however the data produced by this process sometimes produced highly scattered data, as noticed during the previous HFCVD studies [29,[30]]. N₂ presents several experimental detection problems, as it is present in all stages of the vacuum system as a consequence of the non-zero base pressure attainable by the vacuum pumps. Combined with the interference of C₂H₄ at the same m/q, and the extra corrective procedure this implies, it is not surprising that the N₂ signal should be harder to determine accurately. For most of the experiments this was not too much of a problem. In one case, however (Fig. 6.1), where there was a significant quantity of N₂ present, it was necessary to use another approach to determine the N₂ mole fraction. A C:N feed ratio of 1:1 was used in most experiments (Section 6.2.1-6.2.4). It was therefore possible to require the total measured N mole fraction to equal the total measured C mole fraction, in order to be consistent with the input values. As all the other nitrogen containing species of interest had their mole fractions determined in the MBMS experiment, an upper bound to the N₂ mole fraction could be estimated by subtracting the mole fraction of all measured N-containing species from the total N mole fraction ( = total C mole fraction), and halving the result (N₂ is diatomic). It was only found to be necessary to do this for one particular experiment (Fig. 6.1).

6.2.1 1% CH₄ / 0.5% N₂ / H₂ Gas Mixture

The gas-phase composition determined by MBMS for a 1% CH₄/0.5% N₂/H₂ gas mixture at 20 Torr is shown as a function of microwave power in Fig. 6.1 (overleaf). These data should be compared with the analogous graph in Chapter 4 for a CH₄/H₂ gas mixture (Fig. 4.18). As in that case, the measured carbon total decreased as the microwave power was increased, due to mass-dependent thermal diffusion in the plasma. The CH₄ mole fraction fell, and C₂H₂ was produced, as in the case of hydrocarbon/H₂ feed gases (Chapter 4). However, some HCN (containing ~5% maximum of the input N), and NH₃, are also produced, indicating that a small proportion of the input N₂, with its strong NºN bond, is dissociated in the plasma.

Table 6.1 Key to all figures in this Chapter:

Figure 6.1. Graph showing species mole fractions determined by MBMS for a 1% CH₄/0.5% N₂/H₂ microwave plasma. Conditions 20 Torr, 700°C substrate temperature. See Table 6.1 for key to symbols.

The comparative unreactivity of N₂ under these conditions allows the ‘normal’ gas-phase reactions encountered in diamond CVD using CH₄/H₂ gas mixtures to proceed almost unimpeded. This is evidenced by the production of C₂H₂ at high microwave powers, as in the results of Chapter 4. These results are also similar to those seen during HFCVD at low filament temperatures (~1600°C), where very little N₂ dissociation was found to occur [27,29]. At high filament temperatures more N₂ was found to dissociate, which was not seen under the conditions of this study. This indicates that the hot filament dissociates N₂ more readily than the plasma environment (under the conditions of this study). In the hot filament system gas activation is believed to occur on the filament surface (~2600 K), whereas in MWCVD, activation occurs throughout the plasma whose temperature is typically 1400 K [[31]], also reported in Chapter 5. Thus the results for MWCVD can be compared meaningfully with those for HFCVD when the temperature of the activation zone (filament/plasma) is similar. It may be possible to decompose a larger fraction of the input N₂ in the microwave plasma by using higher pressures, or higher microwave powers, due to the increased gas temperature under these conditions.

6.2.2 1% CH₄ / 1% NH₃ / H₂ Gas Mixture

Fig. 6.2 shows the effect of using a 1% CH₄/1% NH₃ mixture in H₂. As the microwave power was increased, NH₃ and CH₄ were consumed in the plasma, presumably by reaction with H to produce NH₂ and (the observed) CH₃. NH₂ radicals, and/or CH₃ radicals are necessary for the production of species containing a C-N bond (Reactions 6.1 and 6.2). H abstraction reactions of such a species containing a C-N bond will result in the formation of the highly stable moiety HCN.

Figure 6.2. Graph showing species mole fractions determined by MBMS for a 1% CH₄/1% NH₃/H₂ microwave plasma. Conditions 20 Torr, 700°C substrate temperature. See Table 6.1 for key to symbols.

Fig. 6.2 shows that the NH₃ mole fraction falls more dramatically with microwave power than the CH₄ mole fraction. This observation was also noted in the MBMS study of Tsang [29], where the NH₃ mole fraction fell more rapidly than the CH₄ mole fraction as a function of filament temperature. It was concluded that this indicated a higher NH₂ mole fraction than CH₃ mole fraction. Further, Tsang concluded that this made the reaction

CH₄ + ^·NH₂ ® CH₃NH₂ + H^·, Reaction 6.1

more likely than the reaction

^·CH₃ + NH₃ ® CH₃NH₂ + H^·, Reaction 6.2

for the formation of C-N bonded species, due to the greater abundance of NH₂ and CH₄, compared to the abundances of NH₃ and CH₃. The ^· symbol is used here to emphasise the radical species in Reactions 6.1 and 6.2. The present results indicate that the mechanism for C-N bond formation (Reaction 6.1) may also be important in the microwave plasma environment.

Between 200 W and 600 W a sizeable signal at m/q=28 was detected in this experiment (and in those reported in sections 6.2.3 and 6.2.4). Common ions with m/q=28 are CO⁺, N₂⁺, and C₂H₄⁺. Out of these species, only C₂H₄ could ionise at the electron kinetic energy used, but a high C₂H₄ concentration would be very surprising given the previous work on hydrocarbon/H₂ mixtures, which showed C₂H₄ to be present at a concentration close to the ultimate resolution of our instrument under all conditions studied (Chapter 4, and [31]). Hence the m/q=28 signal is not believed to be due to C₂H₄. Instead, Fig. 6.2 shows that by increasing the applied power, the conversion of CH₄ and NH₃ to HCN is being effected. This requires an intermediate species containing a C-N bond, which will undergo H abstraction reactions to form HCN. Species such as CH₃NH₂, CH₂NH₂, or CH₃NH are possible intermediates. The latter two are radical species, and will have low ionisation potentials. Electron bombardment of CH₂NH₂ would produce CH₂=NH₂⁺ and, conceivably, HCºNH⁺ by dissociative ionisation. HCºNH⁺ has m/q=28, and a stable electronic structure analogous to C₂H₂. Consistent with this view, the m/q=28 signal is found at power levels where the NH₃ and CH₄ signals are falling, yet where the HCN mole fraction is rising, as would be expected for an intermediate in the conversion of CH₄ and NH₃ to HCN. We thus assign the m/q=28 signal to a radical species, which we cannot calibrate to produce an absolute mole fraction. Therefore measured carbon totals on all graphs in this chapter are those determined excluding the m/q=28 signal. The m/q=28 signal is thus in arbitrary units (a.u.) in Figs. 6.1-6.4.

6.2.3 1% CH₃NH₂ / H₂ Gas Mixture

The gas-phase composition determined using CH₃NH₂ as the source of C and N is shown in Fig. 6.3. Even at the lowest powers used, the CH₃NH₂ concentration falls close to the minimum detectable, indicating rapid destruction of CH₃NH₂ in the plasma environment. CH₄, NH₃ and HCN are observed instead. Two reaction pathways arise for a CH₃NH₂ molecule in the plasma. Firstly, H abstraction reactions could occur, producing a strong, stable CºN bond (as in the observed product HCN). This is analogous to the chemistry observed in hydrocarbon/H₂ gas mixtures, where the CH₃ combination reaction (Reaction 4.1), followed by H abstraction reactions produced C₂H₂. Secondly, fission of the C-N bond in the CH₃NH₂ molecule is possible, followed by reaction with H or H₂ to produce the observed products CH₄ and NH₃. The measured CH₄ and NH₃ concentrations track each other closely, as would be expected if they were produced by fission of the C-N bond in CH₃NH₂. The amounts of CH₄ and NH₃ are reduced with increasing microwave power, with a concomitant rise in the HCN mole fraction, relative to the total measured carbon mole fraction. Again, a sizeable quantity of species with m/q=28 was measured in the low power region (200-600 W), where the conversion of CH₄ and NH₃ to HCN occurs.

Figure 6.3 Graph showing species mole fractions determined by MBMS for a 1% CH₃NH₂/H₂ microwave plasma. Conditions 20 Torr, 700°C substrate temperature. See Table 6.1 for key to symbols.

6.2.4 0.5% HCN / H₂ Gas Mixture

As mentioned earlier, only a small quantity of HCN was synthesised for each experiment, and therefore less MBMS data were taken for this gas. A smaller C:N:H₂=1:1:200 was also used for this experiment to make the HCN supply last longer. Approximately 0.5 dm³ HCN was produced by reaction of 1 g NaCN (0.02 mol) with excess 0.2 M H₃PO₄, in vacuo. A diagram of the apparatus used for this synthesis is shown in Figure 6.4. The gas was produced in flask #1, and collected in flask #2, using liquid N₂ to freeze the HCN in the ‘cold finger’.

Figure 6.4 Vacuum line apparatus used in the synthesis of HCN for the experiment described in Section 6.2.4.

The gas-phase composition for a 0.5% HCN/H₂ plasma, determined by MBMS is shown in Fig. 6.5. Most of the gas remains as HCN, but at low powers small quantities of CH₄ and NH₃ are seen. The fact that little HCN is decomposed emphasises the stability of the HCN molecule under conditions found in the type of plasma discharge studied here. The situation for HCN is very similar to that of N₂ noted previously - both are highly stable in the plasma conditions used here. Once again, below ~600 W, a significant m/q=28 signal was observed.

Figure 6.5 Graph showing species mole fractions determined by MBMS for a 0.5% HCN/H₂ microwave plasma. Conditions 20 Torr, 700°C substrate temperature. See Table 6.1 for key to symbols.

6.3.1 Effect of Nitrogen in the gas-phase using N₂ as Nitrogen source

Diamond films were deposited from a 1% CH₄/H₂ gas mixture, with N₂ added in the range 0-4.7%. This produced smooth nanocrystalline films ~1 mm thick after 6 hours deposition for all experiments where N₂ was added (minimum N₂ added was 0.5%), and a faceted diamond film when no N₂ was added. The variation of growth rate with input N₂ mole fraction is presented in Fig. 6.6.

Figure 6.6 Variation of diamond film growth rate as a function of N₂ mole fraction added to a 1% CH₄/H₂ feed. Conditions: 1000 W microwave power, 700°C substrate temperature, 20 Torr pressure. Also shown are the results of Prawer et al [19], which have been divided by a factor of 10 for comparison on the same graph.

Also shown are the results of Prawer et al [19], who were able to examine the region where very little N₂ was added (<1000 ppm), which could not be investigated in the present study due to the comparatively large 0-10 sccm MFC employed to deliver the N₂. Also, the present growth rate ‘without’ nitrogen in the feed is very high compared with the results of Prawer et al. This is consistent with the view that a small quantity of N₂ was in fact present in the CVD chamber as a consequence of the non-zero base pressure attainable by the vacuum pumps, small leaks in the vacuum system, or as an impurity in the gas supplies.

The diamond films examined in Fig. 6.6, grown using a 1% CH₄ gas mixture with N₂ mole fractions of 0-4.7%, were examined for incorporated nitrogen by the technique of secondary ion mass spectrometry (SIMS) [28]. In common with previous authors it was found that nitrogen has a very low incorporation efficiency (<10^-3) [22,23], even when several percent of the CVD feed gas is N₂.

A 1% CH₄/H₂ plasma was examined by MBMS as a function of the mole fraction of N₂ added to the feed, and the results are shown in Fig. 6.7. The measured carbon total remains constant at a lower value than the input amount due to thermal diffusion effects. As the mole fraction of N₂ was increased, the mole fractions of the hydrocarbon species were reduced slightly, due to the formation of HCN. This indicates that under these conditions N₂ only participates in the gas phase chemistry to a limited extent, even when several percent of N₂ is added. This has implications for growth studies using N₂ additions to form p-type diamond [9,10,17-23], since almost all added nitrogen remains as gas phase N₂ (under the conditions of this study).

Figure 6.7 Gas-phase composition of a 1% CH₄/H₂ microwave plasma as a function of added N₂, determined by MBMS. See Table 6.1 for key to symbols.

6.3.2 Effect of Nitrogen in the gas phase using NH₃ as Nitrogen source

The previous MBMS experiment was repeated with the addition of NH₃ to the feed gas instead of N₂. The data are shown in Fig. 6.8 and it is apparent that the NH₃ reacts more readily than N₂ with the hydrocarbon species to produce HCN. Most of the measured carbon has been ‘locked up’ as HCN by the time the NH₃ concentration becomes stoichiometric with the input CH₄ (1%). Once the NH₃ concentration increases beyond the 1% input carbon limit additional HCN can no longer be formed, and excess NH₃ is largely converted to N₂. Presumably this occurs in an analogous fashion to the case of CH₃ combination followed by H abstraction reactions forming C₂H₂, which occurs in the absence of nitrogen. For nitrogen, however, NH_x species (x=0-3) can combine to produce N₂H_y (y=0-4). Such species were not observed using the molecular beam mass spectrometer, indicating rapid destruction in the plasma environment, presumably by H atom abstraction reactions to produce the observed N₂. These results suggest a method to selectively tailor the gas phase composition between species containing one nitrogen atom and N₂ by varying the concentration of ammonia relative to CH₄.

Figure 6.8 Gas-phase composition of a 1% CH₄/H₂ microwave plasma as a function of added NH₃, determined by MBMS. See Table 6.1 for key to symbols.

6.4 References