3.3.2 Spectral Characterisation 3.3.2 Speed of Response 3.4 References
3.1 Introduction
This chapter describes the experimental methods used during the fabrication and characterisation of the diamond photoconductive devices and the optically-modulated field effect transistor, which form the basis of this thesis.
Chapter 3: Experimental M ethods
3.2 Device Fabrication
Device fabrication is performed in one of the two departmental cleanrooms. By controlled use of air conditioning and filtering, the rooms are designed to conform to Class 100(F) (BS 5295(1989) class E) under the laminar flow hoods and class 10,000(F) (BS 5295(1989) class J) or better in the main body of the cleanrooms; these class levels refer to the maximum number of particles greater than 0.5 p,m in diameter per cubic foot of air. To maintain a low level of particle contamination special clean room clothing is worn over normal wear and there is a ban on naked paper.
The rooms are maintained at a stable temperature o f 21°C and relative humidity of 45% to provide suitable conditions for photolithographic processing. The clean rooms are illuminated by safe yellow light to allow processing with UV sensitive photoresists.
3.2.1 Photoconductor Device Fabrication
Photoconductor device structures were fabricated on polycrystalline diamond grown by microwave plasma assisted chemical vapour deposition (CVD). The diamond film has a thickness of about 1 0 0 }im and is free-standing; the silicon substrate has been
chemically etched away. The growth side of the diamond film has a random morphology, with average grain size o f 20-40 pm, and is also very rough, with surface peak-to-trough roughness of the order o f 10-15 pm. The diamond wafer has been laser cut into square tiles, 4.2 x 4.2 mm.
The diamond surface in the "as-grown" state is conductive due to the hydrogen plasma used during the growth of the film; this conductivity is discussed in Chapter 8. Before
fabrication of a photoconductor the film needs to be cleaned in a strongly oxidising solution to remove the hydrogen and also any non-diamond carbon which may be present on the diamond surface, particularly at the edges, where the laser cutting may have caused localised heating and possible graphitisation. The clean surface is then covered with a thin layer of gold before device patterning using photolithography. Finally, the device may be subjected to a two-stage heat treatment in methane and air environments [McKeag et al., 1995] prior to wire bonding and device packaging.
3.2.1.1 Preparation of Diamond Surface
This surface cleaning process was developed in at UCL during the early part of 1994 as part o f the UCL/Centronic Ltd Photodetector Project. The intended purpose of this treatment is to remove any graphite and other non-diamond carbon from the surface of the diamond sample as a prelude to device fabrication. This treatment consists of two distinct processes: a degrease and an acid bath. The material is subjected to a degrease, then an acid bath and another degrease. Both processes are to be carried out in a fume cupboard, with particular care being taken during the acid bath.
De-grease reagents : 1,1,1 -trichloroethane Acetone
2-propanol (isopropanol, IP A) De-ionised (DI) water
De-grease procedure:
i) Heat sample in 1,1,1-trichloroethane at 60°C for 5 minute - do not allow to boil.
ii) Transfer sample to acetone at room temperature. iii) Transfer sample to 2-propanol at room temperature. iv) Transfer sample to de-ionised water at room temperature. v) Blow dry with nitrogen.
When transferring samples between solvents it is important to prevent the solvent on the sample from drying out as a residue which is insoluble in the following solvent may be left on the surface.
Acid bath reagents: Concentrated sulphuric acid (C-H2SO4)
Ammonium persulphate ((NH4)2S2 0 g)
Hydrogen peroxide (H2O2)
Ammonium hydroxide (NH4OH)
De-ionised (DI) water
Chapter 3: Experimental Methods
Acid bath procedure:
i) Place 6 teaspoons of ammonium persulphate in a beaker. Carefully
add 30 ml of concentrated sulphuric acid. This will give the ETCH solution.
ii) Prepare the RINSE solution which consists of a 1:1 mix of hydrogen peroxide ( 1 0 ml) and ammonium hydroxide ( 1 0 ml).
iii) Heat the ETCH solution and add the diamond samples when the temperature is about 85°C. It may fizz slightly.
iv) Continue heating. Above about 140°C fumes of SO3 will be released.
v) Maintain the temperature of the solution at about 200°C for 20 minutes.
vi) Allow the solution to cool to about 50°C before carefully transferring the sample from the ETCH solution to the RINSE solution. Take care, as there will be some heavy fizzing as the acid is neutralised.
vii) Heat the RINSE solution and sample to about 50°C for 10 minutes. viii) Dispose of the ETCH solution by infinite dilution in running water. ix) Transfer sample from the RINSE solution to de-ionised water. x) Dispose of the RINSE solution by infinite dilution in running water. xi) Remove sample from DI water and blow dry with nitrogen.
During the development of this procedure a selection of literature was considered: Das
et al. [1992], Fang et al. [1989], Gildenblat et al. [1991], Grot et al. [1990], Mori et al.
[1991]. The development work at UCL included an Auger electron spectroscopy study [Baral et al., 1996] of various reported treatments: a simple degrease (isopropyl alcohol- de-ionised water), hydrogen plasma (-700 °C, 15 minutes) at the termination of the growth run followed by a degrease, a saturated solution of H2S0 4:(NH4)2S2 0g (30
minutes, 200 °C) followed by a rinse in 1:1 mixture of H2 0 2:NH4 0 H (15 minutes,
boiling), a saturated solution of H2S0 4 :Cr0 s (10 minutes, 170 °C) followed by a rinse in
1 : 1 mixture of H2 0 2:NH4 0H (15 minutes, boiling), and also excimer laser irradiation in
air (ArF, 193 nm, 100 shots) at 1 Jcm'^. They concluded that exposing CVD polycrystalline diamond to a hydrogen plasma at the termination of the growth run or the use of strongly oxidising solutions are both capable o f removing non-diamond carbon from the surface, but high fluence excimer laser irradiation has the opposite
sulphur within the surface region and produced some form of oxide phase, but the sulphuric acid-ammonium persulphate etchant solution gave a carbon KVV Auger electron spectrum which more closely resembled single crystal diamond. Hence the use of this solution here.
3.2.1.2 Metallisation
All photoconductors described in this thesis required metallisation of the diamond surface. The adhesion of gold to a diamond surface is generally poor and so a carbide- forming metal, such as titanium, is usually used. Various capping layers, such as Au, Pt/Au, Ag/Au, have been used to prevent out-diffusion of the titanium. However, it was found that after thorough and careful cleaning of the diamond surface good adhesion of gold to the diamond surface, with good ohmic contact behaviour, could be obtained. The samples are given another de-grease should they require it. The gold (purity 99.999%) pellets are also de-greased; approximately 6 cm length of 1 mm diameter gold
wire should be sufficient for a 300 nm thick gold layer.
Once clean, the gold is placed in a molybdenum boat inside the evaporator, an Edwards E306 evaporator. The diamond is attached to a glass slide, by double-sided sticky tape, so that the growth side of the diamond film is exposed. This is then loaded into the evaporator, with the diamond sample(s) suspended beneath the glass slide and directly above the gold.
The evaporator uses a combination of rotary pump and oil-diffusion pump to obtain a pressure o f about 10'^ mbar or less within the evaporation chamber. A cold trap is filled with liquid nitrogen to assist in obtaining the required pressure. The pressure within the chamber can be read using a Pirani gauge and a Penning gauge; from atmospheric pressure down to about 1 0'^-1 0'^ mbar, and from about 1 0’^ mbar to high vacuum,
respectively.
Once at the correct pressure the gold is evaporated using resistance heating o f the molybdenum boat. A high tension voltage is applied across the boat and the current is turned up to gradually heat the boat to the correct temperature without causing the boat to fail. The temperature needed to evaporate gold requires a current of about 90 A to be
Chapter 3: Experimental Methods
passed through the molybdenum boat. The gold is then deposited on any surface with a line-of-sight of the gold in the boat, see Figure 3.1.
S K j(Jov»
Figure 3.1 Schematic o f evaporation chamber. (Reproduced from Edwards literature.)
The thickness of the metal film which has been deposited is estimated using a crystal film thickness monitor. The crystal is suspended near the substrate, with one side exposed to the evaporating metal. As a film deposits onto the crystal its oscillating frequency falls in proportion to the loading on the crystal. Thus given the density and acoustic impedance of the évaporant the unit can accurately monitor the rate of deposition and thickness.
3.2.1.3 Device Patterning
Patterning of the device structure was done by defining the structure in photoresist using photolithography, then etching away the exposed gold. This is relatively routine on the smooth surface of a polished silicon wafer, but these devices are defined on the growth surface of the CVD diamond which, as shown in Figure 4.8, can have a peak-to-peak roughness of about 10-20 pm for a film about 100 pm thick. This can make the lithography rather difficult. The process is shown in Figure 3.2 and is described below:
After deposition of the metal, in this case gold (see Figure 3.2a), a layer of photoresist is spun on top of the metal (see Figure 3.2b). The photoresist is an ultra-violet sensitive polymeric compound. Upon exposure to UV light the structure of the polymer changes, making it either more or less soluble (depending on whether it is a positive or negative photoresist - as in a photographic image) in a particular chemical solution, called a developer. The photoresist is selectively patterned by exposure through lithographic
mask (Figure 3.2c). This mask consists of a quartz glass plate on which a chromium layer has been deposited; it is this layer which has the device structure in it. The sample is then immersed in a developer solution (Figure 3.2d) to remove either the exposed or unexposed photoresist. This leaves a pattern in the photoresist; the resist pattern is the same as the pattern on the mask if a positive photoresist is used, and the inverse if a negative resist is used. The unexposed photoresist does dissolve in the developer but at a slower rate, so to prevent development of the unexposed photoresist the sample is immersed in de-ionised water and then dried. The metal which is now exposed is then removed by use of an etch solution (Figure 3.2e). Finally, the photoresist can be removed with acetone to leave the desired metallisation pattern on the substrate (Figure 3.2f).
Photoresist
< G o ld >
Diamond
Etch solution ambient
UV Mask Photoresist - >
Developer solution ambient
<--- G o ld --->
(c)
< — Diamond —>■(d)
Photoresist < Gold DiamondFigure 3.2 Schematic o f the photolithographic process.
The photolithography was performed using a Karl Süss MJB3 UV300 mask aligner, using a mercury lamp at 300 nm wavelength. The photoresists used were AZ4620 and AZ4562 (a replacement for AZ4620 used when AZ4620 was no longer commercially available). The resist was spun onto the surface of the sample for 20 seconds at 2000 rpm followed by 20 seconds at 5000 rpm. It was then left to dry at room temperature for 15 minutes before heating on a hotplate at 100°C for 1 minute. The photoresist
Chapter 3: Experimental Methods
required 600-720 seconds exposure to the UV light from the mask aligner, depending on the age of the photoresist.
The exposed photoresist was removed using a solution of 351 developer - 1:2 de-ionised water. Exposed gold was etched using an iodine etching solution; 150g potassium iodide and 50g iodine dissolved in 800ml de-ionised water.
Optical microscopy was used to assess the quality of the developing and etching during the fabrication process. As the diamond samples used are free-standing, it was possible to backlight the sample during optical microscopy. This yielded a high contrast image, with the diamond transmitting the light and the gold creating a shadow. Such a contrast could not be seen if the sample was lit from above.
The structure contained on the lithographic mask is shown in Figure 3.3. The external dimension of the photoconductor is 2 x 2 mm. There are 15 pairs of fingers, with the width of each finger being 25 pm and the spacing between fingers also being 25 pm. The two bus bars are 50 pm wide and the contact pads are 300 x 300 pm.
Figure 3.3 Photoconductor interdigitated transducer structure.
This patterning process is a bit of a "black art" due to the rough nature of the diamond surface, which causes a variation in photoresist thickness. It is likely that the troughs between the diamond grains will be covered by a thicker photoresist layer than the peaks. This means that it is not possible to specify exact developing and etching times. Instead, the developing and etching stages (Figure 3.2d and e, respectively) are repeated at the user's discretion, making use of the optical microscope to assess the device structure. Photoresist remaining on the surface can create a short circuit between device
fingers, whereas over-developing and etching (sometimes required to eliminate short circuits) can result in very narrow or broken fingers. Initially the yield of working devices was very low and rather disappointing, but after much practise the yield improved considerably, with over 80 % being obtained.
3.2.1.4 Gaseous Treatment
It was found that by subjecting a photoconductor to certain gaseous treatments, using methane and air, the diamond surface could be modified slightly to improve its characteristics, i.e. reducing the dark current and improving the UV-visible discrimination in its spectral response. McKeag et al. at UCL reported this in 1995. The process involves a 1 hour treatment in methane at about 700°C followed by a 1 hour treatment in air at about 400°C.
In 1997 McKeag et al. reported a series of experiments involving treatments with a range of other hydrocarbons and inert gases, none o f which led to the positive effects seen with the methane-air combination; some gaseous treatments quenched the photoconductivity. Auger electron spectroscopy showed a feature associated with oxygen (-514 eV) under some conditions, persisting for methane-air treated samples. Variations in the fine structure of the carbon KVV range (-270 eV) were also seen. Room temperature photoluminescence (PL) was also reported, with a sharp peak at 1. 6 8
eV and a broad peak between - 1. 8 and 2.3 eV; the 1. 6 8 eV peak has been identified in a
number of studies of CVD diamond and is attributed to a Si-vacancy defect, the broad peak is also commonly seen in PL studies o f polycrystalline films but has not yet been clearly assigned. McKeag et al. reported that the methane-air treatment led to a reduction in the 1. 6 8 eV peak height and a reduction and shift in the broad peak. It is
suggested that the methane step o f this treatment is donating carbon or a carbon containing species to the surface, thereby making the surface more graphitic and conductive. The subsequent air step then oxidises the surface removing the surface conductivity which would result from an graphitised surface. As diffusion through diamond grains is not probable at these temperatures a model which involves transport along grain boundaries within the film, followed by passivation o f the photoconductive defects that are local to the grain boundaries, would appear reasonable. Discussion of
Chapter 3: Experimental Methods
the experiments carried out which led to the development of this process are contained in McKeag [1999].
Deep level transient spectroscopy (DLTS) measurements have also been carried out on treated and untreated surfaces, showing a range of trap levels of which some are reduced by this methane-air treatment. A detailed discussion of these measurements is contained in Gaudin [2002].
3.2.1.5 Device Packaging
Once the device has been fabricated and preliminary electrical testing carried out to check for short circuits the device is bonded to an industry standard type T05 header. The device contact pads are connected to the terminals on the header via aluminium bond wires. The standard window supplied with these headers is not transparent in the deep ultra-violet, where these devices are to operate, so the window is removed and the devices operated with a windowless lid; window materials which are transparent over the wavelength range required are being investigated by Centronic, who commercially produce these devices. A packaged device is shown in Figure 3.4.
3.2.2 OPFET Device Fabrication
For fabrication of the OPFET (Optically-modulated Field Effect Transistor, see Chapter
8) devices the surface was not treated with an acid bath because this design is to utilise
the near-surface "hydrogen-doped" conductive layer which would be removed by such a treatment.
3.2.2.1 Metallisation
The design of an OPFET requires two Ohmic contacts with a Schottky contact between them. A simple stripe configuration was chosen, using gold for the Ohmic contacts and aluminium for the Schottky contact; Figure 3.5. The choice of metals used for this device structure is discussed in Chapter 8, where the design, characteristics and
operation of a diamond OPFET is reported. A set of shadow masks with displaced apertures (for Ohmic and Schottky contacts) was designed and fabricated, along with a shadow mask jig for alignment of contacts.
diamond crv'st
“H-doped’ p-type
Figure 3.5 Schematic o f OPFET structure.
The evaporation was carried out in an Edwards E306, as described above for the gold evaporation and using a tungsten helix and approximately 30-40 A for the aluminium evaporation. Gold was evaporated to a thickness of about 300 nm and the aluminium to a thickness of about 50 nm. Aluminium contacts of thickness -300 nm were made along one side of the aluminium strips.
Chapter 3 : Experimental Methods
3.3 Device Characterisation
The device characterisation techniques used throughout the work reported in this thesis are basic current-voltage measurements, including at different gate bias levels when fleld-effect transistor (FET) structures are considered, calculation of the spectral response of photodetectors and the temporal response o f these detectors when illuminated by a laser pulse.
3.3.1 I-V Characterisation
For the electrical current-voltage characterisation the devices were placed in a probe station and were carefully probed with needle-point probes. The probe station was electrically shielded (as well as being shielded from light), the probe mounts were grounded and triax cables used for connection from the probes and shielding to the parameter analysers. Two different Hewlett Packard semiconductor parameter analysers (HP4061A & HP4145B) were used for the electrical characterisation.
3.3.1.1 Hewlett Packard 4061A
The Hewlett Packard 4061A was used for basic I-V measurements. It was connected to a PC using an IEEE488 interface and was controlled by a programme written in Testpoint (a piece o f laboratory equipment software). This software enabled user definition of the start and stop voltages, the voltage step size and the "settling time" (the