A simple, space constrained NIRIM type reactor for chemical vapour deposition of diamond

In this paper the design of a simple, space constrained chemical vapour deposition reactor for diamond growth is detailed. Based on the design by NIRIM, the reactor is composed of a quartz discharge tube placed within a 2.45 GHz waveguide to create the conditions required for metastable growth of diamond. Utilising largely off-the-shelf components and a modular design, the reactor allows for easy modification, repair, and cleaning between growth runs. The elements of the reactor design are laid out with the CAD files, parts list, and control files made easily available to enable replication. Finally, the quality of nanocrystalline diamond films produced are studied with SEM and Raman spectroscopy, with the observation of clear faceting and a large diamond fraction suggesting the design offers deposition of diamond with minimal complexity.


I. INTRODUCTION
The outstanding and unique properties of diamond 1 have led to the pursuit of new methods to synthesise the material for both industrial and academic purposes 2 . Almost concurrently in the 1950s the two vastly different techniques of high-pressure high-temperature (HPHT) 3 and chemical vapour deposition (CVD) were pioneered. Doing away with the extreme conditions required during HPHT, the low pressure, kinetically governed growth process of CVD was successfully developed by Eversole and independently by Angus et al. and Deryagin et al. 4 Unreported till the early 1960s, the first work published on the technique by Eversole demonstrated the successful deposition of diamond on natural diamond powder under less than atmospheric pressure using thermally decomposed hydrocarbons or carbon monoxide 5,6,7 . The subsequent introduction of hydrogen in the gas phase in 1966 8 as an etchant of non-diamond sp 2 carbon, first as molecular hydrogen 9 and then disassociated through either the use of a hot filament 10 or plasma activation 11 , then made CVD a viable method for the growth of high quality diamond.
After two decades of the technique remaining largely unchanged, work performed at National Institute for Research in Inorganic Materials (NIRIM) in Japan in the 1980s brought about significant advances in the CVD process and reactor design 12 . Studies published by the group demonstrated growth on non-diamond substrates using a mixture of hydrogen and methane, initially ionised with a hot-filament 10 and shortly after within a plasma 11 , with sufficient detail such that the methods used could be replicated by others. In particular the microwave plasma CVD reactor proposed, now termed the NIRIM reactor, offered simple operation and the prospect of high quality films free from filamentary impurities. With the design subsequently made commercially available by New Japan Radio, the reactor brought microwave plasma CVD within the reach of those interested in academia and industry.
The reactor overview proposed by the group consists of a quartz vacuum tube fed with hydrogen and methane gaseous precursors, placed through a 2.45 GHz waveguide. Through careful placement of the substrate within the waveguide and the use of a sliding short, the sinusoidal variation in electric field of the TE 10 mode along the width of the cavity results in the formation of a plasma within the centre of the cavity suitable for diamond growth 13 . While still routinely used for high quality diamond growth 14 , the constraints placed on the plasma by the dimensions of the waveguide resulted in the end of dominance of the design. Commercially available reactors targeted towards research and development applications then moved towards designs coupling the TE 10 waveguide to the TE 01 mode of a cylindrical cavity. However, the increased deposition areas attainable came with both an increase in complexity of the design and the use of non-standard parts, resulting in a significant barrier to entry into the field. As such, researchers were dissuaded from investing in the microwave plasma CVD technique and growth capability limited to a select few. This manuscript therefore details the design of a simple NIRIM type reactor, largely constructed from off-the-shelf microwave and vacuum components, as a simple and economic alternative to commercial systems, with sufficient detail to allow replication. The design lends itself well to both doped and intrinsic growth, with any deposition on the reactor walls largely present on an easily interchangeable quartz tube comprising the majority of the discharge tube.
The pertinent aspects of the design and peripheral items used are detailed, while the CAD files, parts list, and control program of the reactor described below are all freely available to download and modify, all with the aim of expanding diamond growth capability 15 .

A. Reactor Design
Based upon the design by NIRIM 11 , the reactor consists of a quartz vacuum tube placed within the anti-node of a 2.45 GHz rectangular waveguide to create a plasma discharge suitable for CVD of diamond. To minimise complexity all microwave components are within the WR340 series by Sairem. The vacuum tube has been designed around an easily replaceable quartz tube and predominantly relies on readily available KF fittings. Figure 1 gives an overview of the complete reactor within panel (c), while panels (b) and (a) detail the pertinent microwave components and show a cross section of the discharge tube within the waveguide respectively.  4 The pressure within the chamber is then monitored through the use of a Vacuubrand capacitive VSK3000 pressure gauge, and regulated with a variable speed Vacuubrand MV 2 NT VARIO diaphragm pump and CVC 3000 controller connected to one of the lower KF flanges to ensure gas movement through the chamber, 'E'. To prevent over pressurisation of the quartz tube a MKS 51A baratron pressure switch with a setpoint of 70 Torr and an above atmosphere Pfeiffer AVA 016 X pressure relief valve are connected to a KF25 cross, 'F', along with a manual diaphragm valve to allow venting to atmospheric pressure. Finally a Williamson DWF-24-36C dual wavelength pyrometer is attached to a custom made tilt stage with vacuum tight quartz window, 'G', situated atop the reducing tee to monitor the temperature of the sample during growth.  Table 1 Table 1. Should the flow rate of a MFC deviate from its set point for an extended period of time the VI will shut off all MFCs and evacuate the discharge tube. Once the pressure within the tube is no longer able to sustain a plasma, the large increase in power reflected towards to the magnetron will be promptly detected by the Sairem components, and the magnetron switched off. Similarly, should the pressure within the reactor inadvertently exceed 70 Torr the hardwired Baratron pressure switch will shut off the gas supply through the pneumatic solenoid valves placed in between the MFCs and reactor inlet. Finally, should the water flow to the reactor and microwave components fall below 6 L/min a flow switch will shut off the magnetron to prevent overheating of the reactor components. As such the reactor is designed to operate safely and without supervision during long growth runs, with processes in place to gracefully shut down should problems arise.

A. Film Growth
To test the capability of the design and the quality of resulting growth a series of nanocrystalline diamond (NCD) films were grown with the constructed reactor. Silicon (100) one-inch wafers were used as substrates and cleaned by RCA standard clean 1 (SC-1) 16 before use. To obtain the high nucleation densities required for the formation of coalesced thin films 17 the substrates were first seeded in a mono-dispersed colloid containing 5 nm hydrogen-terminated nanodiamond seeds/DI water, as explained by Williams et al 18 . The substrates were then placed in the ceramic holder sitting on top of the quartz rod previously described ('T' within Figure   2(b)), and introduced through the bottom of the reactor, as shown in Figure 1. The amount that the rod protrudes into the vacuum chamber is then adjusted such that the sample sits just below the entry to the waveguide ensuring the plasma is atop the holder and, as heating of the sample arises solely from the plasma, that the sample temperature is conducive to diamond growth.
Previous work conducted on the effect of feed gas composition has indicated a reduction in grain size and a corresponding increase in non-diamond content present within films upon increasing the methane admixture 19 , with these changes expected to result in a degradation of the mechanical 19 , optical 20   To confirm the quality of the films grown, Raman spectroscopy measurements were carried out with an inVia Renishaw confocal Raman microscope equipped with a 532 nm laser. Figure   4 plots the normalized Raman intensity from 1000 cm -1 to 1650 cm -1 of the sample after background subtraction. The spectra shows a clear first-order diamond peak at 1332 cm -1 24,25 , broadened due to the small size of the crystallites as characteristic of NCD films 26 . Less pronounced is the G-band peak, located around 1550 cm -1 and attributable to the in-plane stretching of sp 2 bonds within graphitic-like material 27 . Finally, the peaks at 1130 cm -1 and at 1450 cm -1 arise from the presence of C=C and C-H bonds within transpolyacetylene (TPA) chains at grain boundaries between neighbouring crystallites 28 . With the scattering efficiency of graphite previously reported to be ~50 times larger than that of diamond at an excitation of 514 nm, in combination with the morphology observed with SEM, it can therefore be concluded that the diamond fraction within the NCD film is high 29 .

B. Reactor cleaning
After approximately 20 hours of operation at the conditions mentioned above a thin band of amorphous carbon built up around the interior of the quartz tube immediately adjacent to the plasma. To prevent the conductive lacquer from affecting the propagation of microwaves through the waveguide the quartz tube was then removed, placed within a furnace, and heated in air at 600-700 °C until clean. The tube was then reused for subsequent intrinsic growth.
Alternatively, the deposits can also be removed through abrading with tissue paper and solvent.
The quartz rod meanwhile remained relatively clean as a result of being enshrouded within the sample holder within the plasma, while the holder itself was cleaned with isopropyl alcohol and DI after each run.

IV. CONCLUSIONS
The design of a simple and space constrained NIRIM type CVD reactor has been laid out, enabling replication as an alternative to commercial diamond plasma deposition systems. The reactor consists of predominantly off-the-shelf fittings and utilises a modular construction to allow for simple modification, repair and cleaning, while also ensuring ease of use during growth runs. SEM images of NCD films grown with the reactor show high crystallinity while Raman spectra indicate a significant diamond fraction with minimal non-diamond content, demonstrating the efficacy of the reactor design.