This chapter serves to describe and analyse the experimental set-up used in the study of the hot filament diamond CVD environment. Chapter 3 describes the analysis of results from a gas-phase study of atomic hydrogen and CH3 radicals, while chapter 4 discusses studies of the influence of nitrogen addition to the HF-CVD environment.
The study of atomic hydrogen (also termed H atoms) and CH3 radicals have for sometime been paramount in the understanding of the diamond CVD process, as both species are widely considered to be of great importance in the deposition of high quality diamond. Due to its relative simplicity, versatility, and low cost, the hot filament CVD system is the most commonly studied reactor type. This is also true of gas-phase and surface modelling studies, many of which substantiate experimental findings.
The technique of choice for the detection of both species was resonance-enhanced multiphoton ionisation (REMPI), which has been mentioned previously in chapter 1. This species and quantum state selective optical technique, has previously been used in the detection of species within a diamond-depositing environment,. REMPI has the additional advantages of being highly sensitive while allowing a degree of spatial resolution. Despite the presence of a platinum electrode for electron/ion collection, positioned close to the laser focus, REMPI is only minimally intrusive. The use of a biased electrode immersed in the gas-phase environment is justified as the production of charged gas-phase species is found to be minimal. REMPI also has the advantage of being so versatile as to allow detection of H atoms or CH3 radicals simply by altering the wavelength of the incident laser light.
This chapter will define the characteristics of the hot filament reactor and the laser detection system. The use of REMPI for the detection of both H atoms and methyl radicals together with measured temperature profiles will also be discussed and justified.
2.1 Reactor Considerations
In order to study the gas-phase chemistry prevalent in diamond deposition the hot filament CVD system was considered to be the most favourable. The reactor body is designed around a stainless steel six-way cross with an internal diameter of 100 mm. The construction of the reactor is such that spatially resolved measurements may be obtained of species number densities and gas temperature as a function of distance from the hot filament.
This has been achieved by utilising a translation mechanism that allows movement of the filament and substrate assembly, relative to that of the fixed laser focus. This approach allows the laser focus to remain unchanged throughout the series of studies, thereby negating any problems associated with manipulating the laser light, and maintaining a constant laser focus. The major components of the reactor are outlined in figure 2.1.
As the diagram shows, the filament assembly is mounted onto a translation stage allowing vertical separation of the filament and laser focus up to
~25 mm. The translation assembly also allows the addition of a substrate holder and heater, however, for the purpose of the studies in this thesis the substrate was removed so that the gas-phase chemistry may be probed with minimal gas-surface interactions.
The excitation laser light passes through a 20 cm focal length quartz lens, prior to entering the chamber, thus allowing the definition of a well-defined focal volume. This volume will be mentioned in some detail later in this chapter with respect to signal collection. The exiting light passes through another quartz window positioned in the back of the reactor. A fraction of the exiting beam is reflected from a quartz beamsplitter, into a cuvette containing a dilute solution of the dye Rhodamine 6G. The resulting fluorescence is monitored by a photodiode, the signal from which is passed into a digital oscilloscope and used to normalise laser output pulse energies.
The Tantalum filament, produced from 250 mm diameter wire (7 windings from a length of 110 mm wire, ~3 mm coil diameter), is clamped into position on the translation assembly. One end of the filament wire is secured onto a tungsten electrical feedthrough and the other fixed onto a stainless steel post that forms part of the chamber and therefore shares a common ground. The filament is resistively heated using a DC power supply (Isotech IP1810H) operating in constant current mode. This power supply unit allows constant monitoring of the current and voltage supplied to the filament and typically operates at approximately 12.5 V and 6.5 A to produce a filament temperature of 2100 ºC.
The translation mechanism is mounted onto the top flange allowing the filament assembly to be vertically aligned with respect to the laser focus. Sub-millimetre precision in translation is allowed by having a set of digital callipers mounted onto the assembly.
The temperature of the filament is of primary importance in the gas-phase studies mentioned. The filament temperature is monitored using a two-colour optical pyrometer (Land Infrared FRP12) which views the filament through a quartz window positioned in the same horizontal plane as the laser light (the pyrometer head and the viewing quartz window are not shown in figure 2.1). Filament temperature is constantly measured throughout experimental studies. The use of optical pyrometry to measure the filament temperature in this system has been discussed previously in much detail. Previous studies concentrating on the measured filament temperature as a function of the power-induced resistivity, established that a unity non-greyness factor should be applied. The emissivity of the radiating material was also set to unity.
Optical pyrometers are designed for use primarily in the measurement of hot metal plates which fill the entire field of view of the pyrometer head. The determination of temperatures from the blackbody radiation emitted by a heated coiled wire, which does not fill the entire viewing window, is more problematic. Our normal procedure is to direct the pyrometer so it returns the highest filament temperature value; this offers a reproducible strategy but may introduce an error into the temperature measurement.
Since REMPI is a technique that relies on the detection of either produced ions or electrons, a biased probe is positioned close to the laser focus as shown in figure 2.2.
Figure 2.2 Representation of detection area showing filament, laser focus and REMPI probe. Note that diagram is not to scale.
The tip of a platinum wire probe is positioned ~1 mm from the laser focus and lies on the same horizontal plane. The 750 mm diameter wire is held in place and insulated from the chamber by a ceramic tube that is fixed to the chamber through a stainless steel post. The probe bias voltage and REMPI signal is passed through electrical feedthroughs, and via BNC cable, to a filter circuit which allows the signal from the probe to be collected and transmitted to a digital oscilloscope, while allowing a bias voltage to be applied. The filter circuit diagram is shown in figure 2.3. The bias applied to the probe throughout the studies was set at -49.5 V relative to the grounded chamber.
The reaction chamber is connected via vacuum tubing to an Edwards 8 two-stage rotary pump, prior to any experiment the chamber was evacuated to a base pressure of ~10-2 Torr. Gases are flowed through Mass Flow Controllers (MFC) (Tylan), and mixed in a manifold, prior to injection into the reaction chamber. In order to maintain a suitable chamber pressure a needle valve is used to throttle the gas flow to the vacuum pump. The chamber pressure is constantly monitored with a Baratron gauge (Tylan CDL-21), with evacuation base pressure measured via a Pirani gauge (Edwards 11). Throughout the series of experiments reported in this thesis, the pressure was typically maintained at 20 Torr with a total gas flow of 100 sccm.
With a filament temperature reaching, at times ~2400 ºC, the walls of the reactor need to be cooled to prevent damage to the chamber flanges and connections. Cooling is achieved via the use of four fans positioned outside the chamber.
Although no diamond films were deposited during the studies discussed, previous studies have demonstrated the conditions necessary for film growth. These conditions have been used to ensure that the probed gas-phase chemistry is representative of that existing during typical deposition.
2.2 Laser System
The laser radiation used throughout the studies documented in this thesis was produced through a Nd:YAG pumped dye laser. The radiation from a Quanta-Ray DCR-2A Nd:YAG laser, equipped with necessary frequency doubling and mixing crystals, and dichroic mirrors to allow first, second and third harmonic isolation, was used to pump the dye laser. Triggering of the laser and the digital oscilloscope was via an in-house built mains frequency divider, producing a repetition rate of 10 Hz. Light with pulse durations of 7-8 ns and energy of ~100 mJ was passed into the dye laser via the appropriate pair of dichroic mirrors.
The dye laser used throughout these studies was a Spectra Physics PDL-3, with an oscillator stage arranged in the Littrow configuration. This type of dye laser has a quoted bandwidth of 0.05 cm-1 in the visible. The beam waist of the exiting light was approximately 4 mm and adjusted to be non-convergent or divergent. Frequent measurements of the laser power were taken to maintain dye laser output pulse energies of approximately 10 mJ. The coherency of the laser light was also frequently checked via a monitor etalon.
The dye laser radiation was then frequency doubled in a b-barium borate (BBO) B-type crystal, producing light with pulse energies of ~1 mJ. In order to ensure optimum power output from the frequency doubling crystal during frequency scans, the angle of the BBO crystal relative to the laser propagation axis has to be continually changed. This was achieved using an in-house built tracking device, which constantly changes the angle of the BBO crystal to ensure maximum laser pulse energy output.
Prior to entering the CVD reactor a number of optics facilitate the manipulation of the beam position and allow continual monitoring of the frequency of the fundamental laser output via an etalon. This was achieved by inserting a quartz beam splitter and directing a portion of the fundamental output of the laser light through a train of two irises to an etalon. The intensity transmitted through the etalon was monitored by a photodiode, with signal collection and integration carried out by a Boxcar averager (SRS SR250) operating in linear-gate mode. Collected, gated and integrated processed signals were then transmitted to a PC for collection. Interference fringes produced via the etalon were collected simultaneously with the REMPI signal so as to ascertain the frequency dispersion as a function of the frequency scan. The fringe separation with this particular etalon had previously been measured4, via calibration with the transitions of atomic neon, to be
1.250(4) cm-1 in the visible and has been used extensively in the study of Doppler profiles of H atoms in the CVD reactor.
The excitation laser beam is passed through a Pellin-Broca prism, which separates the dye laser fundamental output, from the doubled laser light. While the doubled laser output passes through a lens and an iris prior to entering the chamber, the separated fundamental output is terminated at a beam block. The use of a 20 cm focal point lens and an iris forms a clearly defined probe laser beam that focuses into the centre of the chamber directly beneath the filament. Figure 2.2 schematically outlines the relative positions of the filament, laser focal point and probe wire.
Clearly within the focal point of the laser beam ionisation of species will occur more readily. This allows a degree of spatial resolution and represents an advantage of REMPI over line-of-sight absorption techniques. Absorption techniques, such as Cavity ring down spectroscopy, rely on absorption throughout the length of a defined cavity. REMPI, on the other hand, operates by formation of an ionised species centralised in the laser focus, and then detection of the ion or electron by a biased probe positioned approximately 1 mm away. REMPI is therefore the technique of choice in documented studies to assess the relative species number densities with a high degree of spatial resolution. However, the volume of gas encompassed by the focal volume is poorly defined allowing only relative species number densities to be measured.
The signal from the biased probe is passed, via the filter circuit, into a digital oscilloscope (LeCroy 9361). Collection and analysis of the REMPI signal, the laser power reading, and where applicable the etalon fringes, is carried out in a PC with direct connections to the oscilloscope and Boxcar averager. Experimental data were collected using the Drive program, which allows the simultaneous collection of a number of channels. The REMPI probe signal and the laser power reading were collected simultaneously, as any ‘shot-to-shot’ variations in the laser power will inevitably have an effect on the measured REMPI signal. Both the H atom and CH3 radical detection schemes used here involve an initial coherent two photon excitation scheme, followed by a further one photon ionisation process as discussed below. The initial two photon absorption is rate limiting, and in such cases, excitation probability, p, is related to the intensity of the exciting radiation by p = Ix, where in this REMPI scheme, x=2. The experimentally measured REMPI signals are corrected to allow for this dependence.
2.3 REMPI Spectroscopy
The use of multiphoton ionisation (MPI) for the detection of species within a diamond CVD environment has been previously explored and its use justified4. The basis behind the multiphoton ionisation technique is that the absorption of a number of photons (with equivalent or different energies) by a species, forces ionisation to occur by excitation into the ionisation continuum.
MPI techniques then rely on the detection of either the ion or electron produced. The probability that the ionisation process is accomplished is greatly enhanced by promotion, via a real excited electronic state, and is termed resonance enhanced MPI (REMPI). As the absorption of multiple photons will allow the one-photon selection rules to be broken, states that may not otherwise be reached may be accessed. Figure 2.4 shows two different multiphoton ionisation schemes.
Figure 2.4 Ionisation by a) 2+1 and b) 3+1 REMPI
In the reactor described, the detection of the ions or electrons is achieved by the use of a biased platinum probe wire. While biased metal plate detectors are more commonly used in general MPI techniques, a probe wire was used in the present work to minimise the intrusion on the gas-phase chemistry. Similar wire probes have been used previously3 in REMPI studies of a diamond depositing gas-phase environment.
REMPI allows the detection of either the ions or electrons produced both with equal propensity. Experimentally it was found that the collection of ions gave on-resonance signals with less noise. Thermionic emission of electrons from the hot filament is presumed to be responsible for the larger background signals when the probe is biased for electron detection. All REMPI studies reported here were therefore conducted with a negatively biased probe for the detection of ions.
2.3.1 REMPI study of Atomic Hydrogen
H atoms were studied using 2+1 REMPI via the 2s¬1s two-photon transition with excitation by ~243.1 nm radiation. Light was produced by 355 nm radiation pumping the dye Coumarin 480, with the output being frequency doubled to form light with a wavelength of ~240 nm. The absorption of another photon at this wavelength ionises the atom. Figure 2.5 shows a simple energy level diagram for atomic hydrogen.
Figure 2.5 Energy level diagram for atomic hydrogen showing allowed transitions
Careful studies of atomic hydrogen have shown that the n = 2; 2S½ ( l = 0 ) and n = 2; 2P½ ( l = 1 ) energy levels are essentially degenerate. For a two-photon transition the selection rules allow Dl = 0, ± 2, thereby allowing sufficient population of the 2S½ level from the ground state. Fluorescent decay from the 2S½ excited state back to the ground state is not permitted by the one-photon selection rules. It is worth noting that while the 2S½ level is metastable with respect to the ground state, collisional transfer allows the 2P½ level to become populated, which via one-photon fluorescence (dotted line in fig. 2.5), provides a route to the ground state.
The collisional transfer route signifies a competing route to that of ionisation. Previous calculations4 have shown that for a chamber containing 20 Torr of molecular hydrogen at 1000 K, the collisional transfer rate is likely to be at least two orders of magnitude lower than the rate of loss due to ionisation. It seems fair to conclude that loss due to collisional transfer is insignificant at the typical process gas temperatures and pressures employed here.
The collection of REMPI signal, and its consequent conversion into a frequency dependent spectrum, is carried out by sampling transient time dependent signals. Each laser pulse will excite a number of the detectable species, some fraction of which, will be ionised and detected via the biased probe. The transport time taken for an ionised species to hit the detector will obviously be dependent on the velocity of the species, which in turn will be a function of the local gas temperature. Figure 2.6 shows a number of transient H+ ion signals collected at different filament temperatures (Tfil).
Figure 2.6 Transient H atom REMPI signals collected at d = 4 mm from the filament surface as a function of filament temperature. The REMPI laser fires at t = 0 ms. Each transient is the summation of 10 laser shots.
Clearly time-gated signal collection will only give a reliable indication of the relative species number density if the gate is set sufficiently wide to encompass the whole signal; a narrow time-gate (e.g. ~20 ms width) set near the transient peak will not sample a constant fraction of the total ion current. Therefore in the REMPI studies reported here a large time-gate, ~800 ms width, was used to incorporate the majority of the ion current.
2.3.2 H atom Doppler Profiles
For a study focusing on the detection of gaseous atomic species the most important type of spectral broadening is due to the Doppler effect. Doppler broadening arises as a result of the moving gaseous species ‘visualising’ the monochromatic radiation as that of a different frequency. This frequency shift is a function of the species velocity relative to the direction of radiation. Since the velocity of the gaseous species fundamentally influences the Doppler broadened profiles, local gas temperatures may also be obtained from their analysis. Figure 2.7 shows representative Doppler broadened H atom lineshapes recorded at two different distances from the hot filament.
Figure 2.7 Doppler broadened H atom lineshape from the 2s ¬ 1s two-photon transition showing a Gaussian fit. Both were taken with a gas flow of 100 sccm H2, chamber pressure of 20 Torr, Filament temperature (Tfil) of 2375 K and at a filament distance (d) of (i) 0.5 and (ii) 10 mm. The quoted Doppler widths, Dn, are measured Full Width at Half Maximum height (FWHM).
The broadened lineshape can be approximated to a Gaussian. If we consider each particle as having an independent velocity, and also assume that thermal equilibrium exists, so that a Maxwell-Boltzmann distribution may be used to define the distribution of kinetic energies, then equation 2.1 may be used to describe the velocity distribution, f(nx).
where m is the mass of the particle, k is the Boltzmann constant, T is the local gas temperature and nx defines the particle velocity along the viewing axis, x. The high incidence of gas phase collisions in the hot filament reactor ensures that atomic hydrogen is rapidly thermalised so ensuring local thermodynamic equilibrium (in the scale of the probe volume) and thus allowing the Maxwell-Boltzmann distribution to be assumed.
If we consider that an atom or molecule is travelling towards a detector with a velocity nx, then the frequency, u at which the transition is observed to occur is related to the actual transition frequency, u0 in a stationary atom or molecule by,
where c is the speed of light. Due to the usual Maxwell velocity distribution there is a spread of values of nx which lead to characteristic line broadening given by equation 2.3,
where Du is the FWHM of the Gaussian lineshape in frequency space and T is the local gas temperature.
Equation 2.3 therefore defines a Gaussian to which the experimental Doppler broadened lineshapes may be fitted. By fitting the lineshapes the intensity and the FWHM of the Gaussian fit provides information of the H atom number density and local gas temperature. The relative H atom number density was obtained by integrating the lineshape, therefore detecting H atom signal regardless of the Doppler broadening effect. From the FWHM of the lineshapes the local gas temperature may be obtained.
As previously mentioned the probe laser has a finite linewidth, which has to be taken into account when analysing the broadened lineshapes. The probe laser has an assumed Gaussian linewidth of 0.05 cm-1 in the visible.
The result of combination of the laser and H atom broadened Gaussian lineshapes is a convolution forming a Pythagorean combination and can be deconvoluted by using equation 2.4,
(DuMeasured)2 = (DuDoppler)2 + (DuLaser)2 Equation 2.4
Given that the etalon fringes are recorded in the visible, the Doppler lineshape may be written as,
As the total REMPI scheme involves the absorption of two photons at a doubled frequency, the gas temperature is determined from DuDoppler multiplied by four.
Other typical sources of spectral broadening will now be discussed with reference to a hot filament CVD reactor. In any gas-phase study the presence of pressure broadening must be investigated. The total increase in linewidth due to pressure broadening is typically ~10 MHz Torr-1. Since 1 cm-1 is approximately 3 ´ 104 MHz and the chamber is maintained at 20 Torr, pressure broadening will be negligible.
Due to the nature of species excitation by high intensity incident radiation, caution had to be taken to ensure that power broadening did not occur. The occurrence of power broadening in terms of REMPI would occur via saturation of the ionisation step via excitation with too high intensity light. In the case of atomic hydrogen, this would lead to a reduction in the lifetime of the 2S½ excited state. Any power broadening would manifest itself as an invariance of the linewidth to process conditions. All experimental data collection during these studies was carried out with radiation of intensity below a threshold so as to prevent any noticeable power broadening.
The studies of H atoms in the presence of CH4 and C2H2 together with studies of the radial dependence of the atomic hydrogen number density are included in chapter 3.
2.4 REMPI Studies of Methyl Radicals
Methyl radicals (CH3) were probed using 2+1 REMPI, studying the origin band of the two-photon transition. Radiation at ~333 nm was produced by using the second harmonic (532 nm) of the Nd:YAG laser to pump a dye mixture of DCM and LDS 698. The tuneable dye laser output was frequency doubled through a potassium di-hydrogen phosphate (KDP) doubling crystal. The UV output was attenuated to pulse energies of ~1 mJ. Simultaneous recording of atomic neon transitions in an opto-galvanic lamp and the CH3 REMPI spectra provided an accurate frequency calibration.
The REMPI set-up is the same as that used in the detection of H atoms. Studies were carried out with the addition of quantities of CH4 and C2H2 in the region of 0 – 5 sccm, together with a constant flow of 100 sccm H2.
2.5 Filament Carburisation
The addition of trace quantities of methane (or, we presume, any other gaseous hydrocarbon) to a chamber containing pure H2, and a glowing pure Tantalum filament, causes the filament surface to change. This change is due to carburisation of the filament surface with the formation of tantalum carbides. Since the filament surface is the main initiator of the gas-phase chemistry a change in the number and the nature of available sites on the surface will affect the production of H atoms. Previous studies4 showed that under standard chamber conditions (i.e. 20 Torr of 100 sccm H2, initial Tfil of 2360 K ), upon first introducing 0.5% CH4, the H atom number density rose as a function of time as the filament became more carburised and settled after ~6-7 hours. It was also stated that with a constant filament power input the filament temperature, monitored by an optical pyrometer, rose indicating a change to the whole filament material. This trend was also reflected in the local gas temperatures obtained via the analysis of H atom Doppler lineshapes.
The carburisation process has been shown to occur radially through the diameter of the metal wire by the production of a series of filaments carburised as a function of time. Figure 2.8 shows cross-sectional SEM images of the cleaved filaments after a period of carburisation.
Figure 2.8 SEM images showing the extent of carburisation to filaments carburised for a) 1 hr, b) 4 hr and c) 8 hr. The final panel summarises the results of the SEM analysis and shows the growth in carbide layer thickness as a function of time.
Clearly over time carbon dissolves radially further into the filament material. This is possibly encouraged by the damage caused to the surface by the formation of surface cracks which enhance the surface area. The increase in filament surface area may also explain the enhanced H atom production after carburisation. However, as the carbon content in the filament increases the resistivity of the filament will also rise, resulting in an increased filament temperature at constant input power, which in turn increases the H atom production at the surface.
Consequently all gas-phase studies discussed in this thesis have been carried out using a filament that has undergone a carburisation process of at least 8 hours.
Chapter 3 focuses on monitoring atomic hydrogen and the band origin of CH3, as a function of added CH4 and C2H2, distance from filament, and as a function of other process conditions. This chapter also contains discussion of the factors that need to be appreciated when turning measured REMPI signal recorded at a single excitation wavelength into total CH3 number densities.
 D.G. Goodwin and J.E. Butler, in ‘Handbook of Industrial Diamonds and Diamond Films’,
ed. M.A. Prelas, G. Popovici and L.K. Bigelow, p. 1-68, and references therein.
 J. Heinze, N. Heberle and K. Kohsehoinghaus, Chem. Phys. Lett., 1994, 223, 305.
 F.G. Celli and J.E. Butler, J. Appl. Phys., 1992, 71, 2877.
 S.A. Redman, Ph.D. Thesis, University of Bristol, 1999.