Introduction:

Diamonds are one of the more extraordinary materials on our planet. Its tightly packed carbon structure is responsible for making it the hardest material on Earth. This makes it an excellent cutting material, yet diamond has many other intrinsic properties. Diamond has a very large bandgap of 5.47 eV, making it a wide bandgap semiconductor (WBGS) along with materials such as SiC and GaN. Note that this report is a discussion of single-crystal CVD diamond, as it is much more predictable and easier to work with than naturally occurring diamond. The material possesses the highest thermal conductivity of any known material, above 24 𝑊𝑐𝑚−1𝐾 −1 as well as the highest charge carrier mobility among WBGS’s at room temperature [1] . Along with an extremely high breakdown field of 10 𝑀𝑉𝑐𝑚−1 , diamond would be ideal for electronic devices subject to high temperatures and high-power inputs[1] . Where typical field-effect transistors (FETs) made of Si or GaAs might fail, a diamond FET would not.

However, as attractive as diamond may seem, the material presents many challenges in the face of its commercial viability. Diamond is not naturally conductive and functions traditionally as a dielectric. Doping, a classic technique in semiconductor materials, is also very difficult due to the tightly packed face-centered cubic (FCC) structure of diamond crystals. This is typically done through substitutional doping, when foreign elements are introduced into a metal to alter its properties. This is often done through ion implantation or during crystal growth [2] . Doping may either occur with a donor or acceptor material to form a ptype or n-type semiconductor respectively. Normally, this leads to an increase of the density of charge carriers within a material, which naturally leads to higher conductivity. For diamond, the best-known dopants are boron, nitrogen, and phosphorus; although these dopants have high activation energies compared to common dopants in other semiconductors[2] . An alternative to traditional doping and the challenges it presents is surface transfer doping by controlling the surface termination.

Transfer doping is an easier and promising alternative to substitutional doping. Instead of trying to force electron acceptors or donors into diamond’s tight carbon lattice, transfer doping can be done by terminating the diamond surface, specifically with hydrogen, and exposing it to an appropriate absorbate. The abundance of hydrogen on the diamond surface after hydrogenation will pull electrons to the surface because of a negative electron affinity. Newly “summoned” electrons will then jump from diamond’s conduction band into the atmospheric layer resting above the surface (if the surface is exposed to air). This is due to the potential difference between the chemical potential of the atmospheric layer and fermi level of diamond [2] . The process is pictured in figure 1. With so many electrons pulled out of diamond, an abundance of holes will remain on the surface, which leads to p-type semiconduction. Another term for the behavior of the diamond surface in this state is 2-dimensional hole gas. Transfer doping can also occur when hydrogenated diamond is in contact with an appropriate material. If the conduction band of the acceptor material sits below the “bent” valence band of diamond and has an electron affinity greater than 4.2 eV, then electron transfer to the absorbate will occur. Examples of appropriate materials are MoO3, V2O5, and WO3 [2].

Two primary methods of hydrogenating a diamond surface have been developed: plasma and thermal. Hydrogen plasma treatment involves exposing a diamond surface to atomic hydrogen for a short time inside a plasma reactor and produces notably different results compared to thermal treatment. Plasma treatment may lead to a rougher diamond surface, lower sheet resistance, and higher charge carrier density (fig. 5) [6]. Thermal hydrogenation utiltizes molecular hydrogen gas and high-temprature annealing. Results with thermal may not produce as low resistances as plasma treatment, but this process is much cheaper and has industrial potential. Seshan et al provides relevant proceudres for carrying out both methods in detail [7].

The current field of thermal hydrogenation of diamond has developed substantially over the past 20 years. A 1993 study took an extensive look at the temperature dependence of this hydrogenation process when applied to diamond powder. The study examined the annealing process with pure hydrogen and reported peaks in molecular surface bond vibrations[3]. Fig. 2 displays their results of examining C-H surface concentration with respect to temperature and shows a peak in intensity around 900° Celsius. A study by Maier [4] et al continued to look at the cause of increased surface conductivity in diamond. They concluded that not only is surface termination necessary to achieve higher conductivity, but so is air exposure (or more specifically an atmospheric absorbate) post-hydrogenation.

In 2007, an Italian team [5] have wrote a critical paper on this subject. They tested the annealing process with pure molecular hydrogen gas for 1 hour at several temperatures to examine optimal temperature for lowest sheet resistance [5]; the results of which are shown in figure 3. Exponential decay is shown with significant results beginning at 700° C and ideal results at about 850° Celsius, which falls in line with the 1993 study on C-H concentration [3]. Figure 4 shows sample resistance measured in time intervals after the sample was exposed to air. Along with confirming previous research that diamond surface conductivity requires air exposure [4] , they showed the ideal thermal hydrogenation process requires temperatures of at least 800° C and several hours of air exposure.

Furthermore, hydrogenation studies on diamond can be examined at the surface orientation level[9,10] . An article by Manfredotti [9] et al on hydrogenation revealed best surface conductivty and C-H bond stretching on surfaces (111) and (110) at temperatures even below 800°C, as well as at 800°C for (100). Temperatures above 800°C were concluded to be ideal for thermal hydrogenation which “represents a cleaner, simpler and cheaper way to hydrogenate diamond surfaces [9].” Another recent study simulated thermal treatment to determine optimal orientation during hydrogenation [10]. Figure 6 is a table of the number of unfulfilled carbon bonds per square nanometer on each surface post-hydrogenation. The (001) surface appears most Figure 4 (Fizzotti, 2007) – Sample resistance as a function of time after air exposure Figure 3 (Fizzotti, 2007) - Sheet resistance in hydrogenated diamond as a function of temperature Figure 5 (Liu, 2017) – Comparison of thermally hydrogenated diamond and plasma hydrogenated diamond susceptible to annealing, but the author notes that more research needs to be done to understand what exactly is happening on the molecular level.

Methods:

Four total experiments of thermal treatment were completed. The first three trials tested only one sample; the final trial tested four samples at once. Synthetic diamond samples measuring 2×2 and 3×3 mm were used. Prior to thermal treatment, all samples were cleaned at ~250°C in a solution of nitric and sulfuric acid for at least 30 minutes. Deionized water was used to further decontaminate each sample. Following proper cleaning, samples were placed inside a glass quartz tube which sat inside a ceramic furnace. Following preliminary pumping and venting to cleanse the system, a pressure range of 0.05 – 0.25 torr was achieved, and a forming gas composed of 93% argon and 7% hydrogen was delivered into the system at half of atmospheric pressure (5 - 7 psi). The gas delivery line remained connected to the system (including the pumping machine and exhaust line) for safety and to ensure a constant flux of gas at 4-5 sccm. Note that this procedure may be done with pure molecular hydrogen gas with appropriate vacuum conditions[8] .

With the gas flow in place, the furnace temperature was ramped to 800° Celsius over a period of two hours. The temperature change required adjustment of the pumping intake valve to ensure a gas flux ideally between 4-5 sccm. Samples were typically annealed for ~48 hours at 800°C, before letting cool to room temperature for an hour. For the 3rd experiment, the annealing process was carried out for 64 hours. Analysis of the H-terminated diamond surface was done using both two-probe and four-probe methods. For the two-probe setup electrical contacts were placed on the corners of the sample and a potential difference created between them. Current measurements were taken and plotted. The four-probe setup was done using a specially ordered piece of equipment for testing electrical characteristics of small square samples. Resistance measurements were taken.

Data:

[see pdf for figures]

Discussion:

Looking at graph 1, the linear I-V relationship at low (V<1 V) bias and especially in the third quadrant manifests ohmic behavior demonstrating the newly conductive diamond surface. This is a success and appeared after 10 minutes of air exposure. These graphs have regions where they are not linear (non-ohmic), but this is more indicative of imperfect electrical contacts by probes. Graph 4 is also not perfectly linear which could be due to the probes or incomplete hydrogenation on the opposite side. This could be evidence of the importance of the surface orientation’s effect during the process. Also worth investigation may be the use of forming gas as opposed to hydrogen gas. Hydrogenation is often performed with 99.999% pure hydrogen gas for optimal results; however, this implies additional safety risks. Lower purity hydrogen gas may require higher temperatures or more than 2 days to properly hydrogen terminate a diamond surface.

In any case, hydrogenation with the forming gas was successful. Resistance of the diamond sample also decreased with time upon its exposure to air. Specifically, from 167 𝑘Ω to 143 𝑘Ω. This is observed by applying Ohm’s law to the slopes of graphs 2 and 3. A further decrease in resistance is likely after several hours of air exposure [5] . Graph 3 shows current output after rotating the sample 90 degrees (changing one of the probes), and it shows a slightly more linear trend than previously, further indicating a faulty probe. Graph 4 is analysis of the opposite side of the sample and is not quite as convincing. Examining the third quadrant of graph 4, the opposite side conducts up to about 7 volts where the trend ceases to be linear and is more unpredictable. Not equivalent exposure to the hydrogen flow may be the cause, resulting in some surfaces of the sample being more susceptible to hydrogen than others.

The table is a concise list of sheet resistances measured with the four-probe method. The best results came from samples 1 and 5, with sheet resistances in the range of 700-1,000 kΩ. These results are at least an order of magnitude off from some of the typical published data [4] . The remaining results fell in the range of 1,000-40,000 kΩ, as well as some trials that were unsuccessful. There is a great deal of variability here, which could be due to several factors. The initial surface roughness/polishing could be affecting charge mobility or hydrogenation effectiveness [6] . One of the benefits of thermal hydrogenation is that it does not damage the material’s surface, but the samples need to have relatively smooth surfaces to begin with. Gas impurity, either from the low concentration of hydrogen in the forming gas or possible leaks in the experimental setup could also be responsible. In addition, the long trial times paired with slightly varying pressures inside the system made it difficult to ensure a constant flux of forming gas. This aspect of the experiment would ideally be automated in an industrial setting.

Comparisons with contemporary semiconductors could be made with more concrete values for resistivity, charge carrier mobility and carrier density. Larger samples could also be tested to see how well thermal hydrogenation affects larger surface areas. More tests with forming gas should be performed to test its efficiency against high-purity hydrogen gas. Both sample orientation, as well as hydrogenation’s dependence on forming gas flux, should be investigated further. This experiment was an overall success, and the takeaway is how simple this thermal process is. Compared to the cost of plasma or electron beam treatment, thermal hydrogenation is a simple and inexpensive soft treatment that produces comparable results.