ANÁLISIS PARTICULAR DEL MERCADO Y LA EMPRESA

For the purposes o f this project, work described as having been carried out in a Clean Room, refers to a particle, temperature and humidity controlled environment specified to Class 1000 with laminar flow workspaces locally Class 100 and yellow light throughout. Unless stated otherwise, all other work was carried out in open laboratory space under normal ambient room temperature and white working light.

4.2.1 Surface Degrease

Every sample used in the course o f this project has been subjected to an initial degrease as summarised below:

warm T richloroethy lene —> Acteone IP A —> DI —> N2

The trichloroethylene and the acetone serve as general purpose solvents to loosen or remove grease, moisture and particulate contamination, IPA follow s the acetone as it is the purest o f the electronic grade organic solvents [4.1] and is readily m iscible in DI water, which completes the sequence as the purest solvent available. A rapid blow-dry in N2 is followed by a 20 minute bake at 85“C to yield a uniformly clean, dry surface.

Som e samples were received with visible evidence o f substantial particulate surface contamination, such as powdered graphite residue from laser dicing operations. In these cases the degrease was preceded by a brief rinse in DI water under ultrasonic agitation.

4.2.2 Oxidising Etches

It has been established by [4.2] that the as-grown surface o f a CVD diamond film can be expected to carry a substantial degree o f contamination in the form o f amorphous, or other non-diamond, carbon. The polished surface of natural diamond is also known to be similarly contaminated [4.3], and a number o f strongly oxidising processes have been proposed to remove this material [4.2, 4.3, 4.4, 4.5, 4.6]. For the purposes o f this project, the follow ing hybridised regime was derived from the published literature and found to be effective; the process was subsequently evaluated in more detail elsewhere

Chapter 4; Experimental Methods

(a)

(b)

:V.<5 X ) aà T . , - . - T ';< ^

m #

Figure 4.1: Optical micrographs indicating the effect of acid treatment on a typical freestanding CVD diamond tile (a) untreated - extensive non-diamond carbon deposits can be seen along laser-cut edges, sample lacks lustre (b) after acid treatment - the cutting debris has been removed and the tile appears

generally brighter and clearer.

The sample is placed in a saturated solution of ammonium persulphate ((NH4)2S2 0g) in sulphuric acid (H2SO4) which is then heated to around 230°C, being maintained above 170°C for 20 minutes. The sample is then transferred to a solution o f hydrogen peroxide (H2O2) (70%) and ammonium hydroxide (NH4OH) (30%) combined in equal quantities by volume. The first stage specifically attacks non-diamond carbon [4.2] whilst the second stage is identical to the industry standard RCA clean, SC-1 which is widely employed to effect wet oxidation of organic surface films and expose the surface for desorption of trace metals including gold, silver, copper, nickel, zinc and chrome [4.1]. The effectiveness of this cleaning regime is illustrated in figure (4.1) which shows the contaminated edge and surface of a laser cut tile of freestanding CVD diamond before and after cleaning, in which the extent of material removal is clearly visible. Baral et. al. [4.7] have used auger electron spectroscopy to confirm that the process effectively removes graphitic and non-diamond carbon material and surface hydrogen, but commented that it leaves some residual surface contamination in the form of sulphur and oxide com plexes. This contamination has not been found to be a problem for the majority of devices of interest. Given the wide range o f sample types investigated and the variety of conditions in which they were received, this cleaning

Chapter 4; Experimental Methods

regime has been essential in asserting a baseline condition against which differences can be measured for different samples and different forms o f processing.

4.2.3 Photolithography on Rough Surfaces

Standard photolithographic methods for defining device geometries on a smooth, flat semiconductor surface are a well developed and characterised technology [4.8]. Films o f photoresist are deposited on the sample surface either by spraying or by dripping a pool o f resist onto the centre o f the wafer and then spinning it with a speed and duration specified by the manufacturer to achieve the desired thickness, typically in the order of «1.3 to 1.8p.m. The coverage and uniformity o f such films are generally excellent and the device image is then created by exposing the resist to a patterned light source of the appropriate U V wavelength, typically « 3 0 0 to 400nm. D eveloping the resist will remove either the illuminated or the unilluminated regions depending upon the chosen regime, so that the sample surface is then left exposed though a pattern o f 'windows' in a protective film o f resist. The exposed surface can be selectively etched away, or an overlayer can be deposited through the 'windows' so that when the resist is subsequently removed, only the exposed regions o f the sample surface have been coated, (this latter process is known as 'lift off).

The main challenge in processing the randomly oriented 4.2 x 4 .2 m m ^ polycrystalline CVD diamond tiles, as used predominantly in this project, is the difficulty in achieving total and reasonably uniform coverage o f the very rough surface. This problem has resonances in other fields o f small-dimension manufacturing such as micromechanics and optoelectronic interconnects, so thick photoresist technologies are commercially available. Even with specialised resists however, the nature o f the work undertaken required that they be used at and beyond the limits o f their data sheet specifications. To achieve this, extensive iterative optimisation was carried out to establish a feasible regime o f spin speed, spin time, exposure intensity and exposure time that was capable o f yielding successful device definition on a range o f substrates.

Hoechst A Z 4620A is a viscous positive photoresist specifically designed for thick coating of semiconductor surfaces [4.9]: manufacturer's spin speed/time parameters to achieve a given film thickness on a smooth surface are reproduced in figure (4.2). Figure (4.3) shows a cross-sectional electron micrograph of a typical freestanding CVD diamond film used in this project with an inset indicating the surface roughness to have a peak-to-trough height of around 15)j.m.

Chapter 4; Experimental Methods

Film Wckn### [pm]

00

1 0 2 0 4 0 2 0 0 0 4 0 0 0 SOOO 6 0 0 0 7 0 0 0

Figure 4.2: Manufacturer's recommended spin parameters to achieve a given thickness of AZ4620P photoresist on a smooth surface. [Reproduced from 4.9].

Figure 4.3: Cross sectional electron micrograph of typical CVD diamond tile with inset indicating a peak-to-trough surface roughness of = 15|im.

Chapter 4; Experimental Methods

The surface roughness gives rise to two primary problems: com plete and effective coating o f the surface requires that a substantial thickness o f resist be employed, placing the parameter selection in the 2000rpm region o f figure (4.2), whilst the flow of the viscous resist across the non-smooth surface serves to undermine the applicability of these data sheet parameters. Under these circumstances the use o f data sheet parameters to achieve adequate coverage at the centre o f the tile results in an extensive occurrence o f the 'edge-head effect' [4.10] whereby excess resist fails to be expelled by centripetal force during spinning and collects instead at the perimeter o f the sample, leading to both a substantial inhom ogeneity in resist thickness across the sample and extensive spreading o f this usually minor effect across a significant proportion o f the surface of the tile.

It was found that a spin o f 2000rpm for 20 seconds follow ed immediately by a second spin at SOOOrpm for 20 seconds yielded a tile-centre resist thickness sufficient to achieve unpenetrated coverage with an edge bead width <lm m . Exposing this resist for 720 seconds to 400nm , 7.3mW /cm^ follow ed by a develop o f « 5 0 seconds in a formulation o f dilute commercial 351 developer (351:DI, 1:2) achieved satisfactory reproduction o f device features in the order o f 25|im . It was noted that this regime operates substantially in the tolerance region o f the photoresist, beyond the conditions recommended by the manufacturer, with the result that very little latitude remains for variations in the process parameters or the nature o f the sample surface before critical failure o f the process w ill occur. Some low-grade 10p,m device features were also achieved on a one-off basis, however this work was highly irreproducible and could not be regarded as a reliably characterised system.