Awadesh K. Mallik ^1,2* Sankaran K. J. ^2,3 Derese Desta ² Paulius Pobedinskas ² Rani M. Joy ² Rozita Rouzbahani ² Fernando Lloret ^2,4 Hans-Gerd Boyen ² Ken Haenen ²

• ¹ Nanyang Technological University, Singapore, Singapore
• ² University of Hasselt, Hasselt, Limburg, Belgium
• ³ Institute of Minerals and Materials Technology (CSIR), Bhubaneswar, Orissa, India
• ⁴ University of Cádiz, Cádiz, Spain

There are non-diamond substrates [1] incompatible with diamond growth due to the harsh CVD processing conditions. The temperature window of 700 o -1200 o C, which is suitable for the metastable growth of the CVD diamond phase, is detrimental for many low melting point substrates, for example, polymers. Many inorganic substrates are also reactive at such high temperatures of CVD diamond processing. GaN [2,3] is a reactive substrate prone to etching beyond 600 o C [4]. It has been proposed by May et al. [5] that the following three possible reactions occur under a CVD growth environment:1. 2N(surface) + 3H2 (g) → 2NH3 (g) ----------at 800 o C 2. 2Ga(surface) + H2 (g) → 2GaH (g) ----------at 800 o C 3. 2GaN(s) → Ga(l) + N2(g) -----------above 800 o C is necessary to remove the intermediate buffer layer andThere are reports of directly growing the diamond on the GaN substrate [25]. GaN-on-diamond is a complex multi-step technology [2219].Direct deposition of diamond for capping the GaN devices is a simpler approach. Sun et al. [13] demonstrated the variation of TBR with the SiN interlayer thickness and by extrapolating the SiNx layer thickness to zero, they showed a TBR value below 7 m 2 K/GW, taken into account the uncertainty of the linear fit. Yates et al. showed very high TBR (> 40 m 2 K/GW) without SiN interlayer. Without an interlayer diamond only forms a weak van der Waals bond to GaN. This results in a TBR which is seven times higher than for GaN-on-diamond which includes a silicon nitride interlayer [26]. Earlier simulation work [27] showed that the small crystal sizes during the initial growth period of diamonds on GaN significantly impacted the device's thermal resistance as a function of CVD diamond film thickness. Phonons are primarily responsible for thermal conduction inside diamonds. The typical length of such phonons is 1µm in the bulk diamond. The nanocrystalline nature of the GaN-diamond interface decreases the effective thermal conductivity [28]. Moreover, critics may argue that GaN is not a carbide-forming substrate [1,1417]; therefore, it will not form any chemical bond with diamond. The weak Van der Waals bonding between the GaN substrate and the grown diamond coating will eventually fail as the thickness of the diamond film is Formatted: Font: Times New Roman Formatted: Font: Times New Roman Formatted: Not Superscript/ Subscript increased with deposition time. Fortunately, thick coatings are not typically grown by the linear antenna CVD system used in this research [29].One approach to avoiding GaN substrate etching during the CVD growth of diamond is to seed the GaN substrate surface, either by novel nucleation procedure of the substrate surface with amorphous carbon [30] or by hydrogen-terminated detonation nanodiamond (DND) solution[2125] before CVD processing. Researchers have also tried two-step growth recipes: first using high methane recipes and, subsequently, normal diamond CVD growth recipes [5,2530]. The next methodological approach is to keep the GaN substrate temperature as low as possible. Goyal et al. [31] used argon as the main precursor gas to grow ultra-nanocrystalline diamond (UNCD) films at low 450º-500ºC of CVD process temperatures. However, UNCD has more grain boundaries, which promote phonon scattering, which is not conducive to effective thermal management [32,33]. Variation of filament to substrate distance [5] has also been tried to optimiseoptimize GaN surface etching during hot filament CVD growth of diamond [34]. Authors have shown that at a 12-14 cm filament to substrate distance, optimal formation of diamond film takes place, although the main drawback of their paper is that it does not mention the corresponding substrate temperatures. Malakoutian et al. [35] deposited near-isotropic microcrystalline diamonds on top of SiO2 and Si3N4 at relatively lower temperature range of 300-500 o C, by using oxygen gas in the recipe. They have reported a very low TBR and high thermal conductivity of the deposited diamonds. However, their work was done with conventional 2.45 GHz microwave plasma resonant cavity reactor. Recently Barba et al. investigated the effect of substrate to linear antenna (pulse mode power) distance in growing NCD at 450 o C and found the thermal conductivity (10 2 W/m.K) is not altered with distance variation. [36]. Nitrogen gas was sometimes added to the precursor recipe to suppress the GaN surface decomposition reaction. However, all their growth of diamond at as low as 300 o C substrate temperature [37]. In the present paper, substrate temperature between 400°C to 900°C, possible with a linear antenna CVD system, has been used to investigate the GaN etching issues in diamond CVD. It is also to be noted that the working pressure is much lower inside linear antenna CVD reactor compared to resonant-cavity or hotfilament CVD reactors. Different combinations of precursor gases have been used for the successful CVD growth of polycrystalline diamonds on non-diamond substrates [38,39,40]. The most widely used precursor [41] is a mixture of hydrogen gas with some hydrocarbon source like methane, acetylene, etc.Argon has been used recently as the major precursor gas for its effectiveness in depositing ultra nanocrystalline diamond (UNCD) films [42]. There are also dopant gases like nitrogen, di-borane, phosphine, etc. which are added to the precursor gas mixtures in order to dope the diamond films and making them n-type or p-type semiconductor [43,44,45,46,47]. The reason for using a gas like germane was to create optical colour centres [48]. The addition of such gases, which creates the electron states within the bandgap of diamond, has helped to change the intrinsic properties of CVD-grown diamonds. Researchers are always in search of alternative gas recipes (for example, H2S) which can introduce a high concentration of donor impurities (for example, S) inside the diamond lattice [49,50] without losing the diamond film quality. Nitrogen gas is routinely used for giving n-type conductivity [3844] in diamonds, but it has its own limitation for any practical use, due to the high activation energy of donor states. Phosphorous is another potential n-type donor [3945,4046,4147]. Here is an attempt to add phosphine (0.1% in hydrogen) in the precursor gas recipe [51,3743] to study the effect of such a new PH3-based precursor recipe on the diamond films grown over the GaN/sapphire substrate heterostructure. This work will report the direct CVD growth of diamond at lower temperatures to address the problem of GaN substrate etchingimportantly, growing diamond on the GaN/sapphire heterostructures without a silicon nitride layer on top of the substrates. In theory [13], it is possible to reach low TBR (around 6 m 2 K/GW) even without any interlayer. 2.1. Substrate material and CF4 plasma treatment:Commercially available GaN on sapphire substrates (7×7 mm 2 sizes) were used in the present study (silicon substrates of 10×10 mm 2 sizes were kept alongside for CVD growth reference). The as-received GaN surface was characterised by X-ray diffraction (XRD) which was performed with a Bruker D8 theta-theta diffractometer equipped with a Göbel mirror (line focus, mostly Cu Ka radiation). The X-rays are detected with a 1D lynxeye detector. X-ray scan was from 10º to 80º 2 theta range, with 0.04º stepsize and 10 s/step counting time. As nitride surfaces are not compatible with direct diamond growth [9], the heterostructure substrates were pre-treated with CF4 plasma (it cleaned the substrate as well) before seeding the substrate with detonation nanodiamond (DND) [17,52]. 300W power with negative bias on the stage was used in a DCsputtering chamber, etching with 42 sccm CF4 gas flowing in for 20 s at 10 -2 Torr working pressure [1417]. The chemical composition and chemical state of the substrates were analysedanalyzed using X-ray photoelectron spectroscopy (XPS) (PHI 5500 XPS, PerkinElmer) equipped with a monochromatic Al Kα X-ray source (hν=1486.6 eV) and a hemispherical of 150 mm in diameter. The spot size of the X-ray source on the sample was 1 mm in diameter. Afterwards, the etched and cleaned GaN substrates were seeded with DND as described earlier [53] before being loaded into the CVD reactor. DND powders were mixed in de-ionised water by 0.05 wt% and sonicated to produce a water based DND suspension. A single drop of the suspension was added onto the substrate surface, which was vacuum mounted onto a spin coater. A monolayer coating of the DND seeds was obtained [54], after spinning for 40 sec at a speed of 4000 rpm, with short rinsing with de-ionised water, and successive drying by a nitrogen gun. Whereas the corresponding reference silicon substrates (cleaned in oxygen plasma) were not pretreated with CF4 plasma but were only DND seeded before CVD diamond growth. The linear antenna microwave plasma enhanced CVD system [55,56,57] was used to grow diamond at substrate temperature as low as 400 o C, thereby avoiding the possibility of GaN surface reactions described already in the introduction section 1.1. [5]. Moreover, such low temperature also solves the problem of cooling-induced (thermal stress) delamination caused by putting two dissimilar materials together. Table 1 describes the CVD processing conditions used in the study.In the first series of experiments (I), PH3 gas (1000 ppm in H2) was added to the plasma such that the [P]/[C] ratio in the plasma was varied from 0 ppm to 8090 ppm, while keeping the CH4 and CO2 gases at fixed concentrations of 5% and 6%, respectively. In the second series of experiments (II), the substrate temperature was increased gradually from 400 o to 900 o C by keeping the maximum [P]/[C] ratio of 8090 ppm in the plasma. The depositions were carried out with GaN/sapphire heterostructures kept along with reference silicon substrates. The vacuum level was Formatted: Not Superscript/ Subscript in the range of 10 -4 Torr before introducing the precursor gases inside the reactor, whereas during the CVD growth, a shallow pressure of 0.23 Torr was maintained. There were two 2.45 GHz microwave power sources, each delivering 2800 W average power (total 5600 W) inside the CVD chamber [4956]. The thickness of the growing film was monitored by laser in-situ interferometer (wavelength 405 nm), and each time the experiment was stopped after achieving 250 nm thickness of the growing diamond film on the reference silicon substrates. It was 6 hours of deposition approximately. Typical LACVD growth rate was around 40 nm/hr [51]. The distance between the quartz tube and the substrate stage was kept fixed at 5 cm, and the substrate was heated only by the plasma to achieve a fixed substrate temperature of (400±20) º C. The temperature reported here is the stage temperature measured by thermocouple placed underneath the substrate stage -thus measurement was not affected by the plasma environment. We did not observe any fluctuation in substrate temperature when different mixing percentage combinations of the H2 and diluted PH3 gases were used while keeping their total volume flow rate (150 sccm) constant. In the second set of experiments (IIa-IIe), when only diluted phosphine (no pure hydrogen) was used, the heater underneath the stage was turned on to achieve substrate temperatures beyond 400 o C. The CVD plasma chemistry was monitored by an optical emission spectrometer (OES). An optical fiber (Avantes FC UV600-2) collected the light via one of the lateral viewing ports. The spectra were recorded by an AvaSpec -2048 (Avantes) spectrometer, which comprises of a 2048 pixel CCD detector array and covers the wavelength range from 200 nm -1100 nm. The effect of more reactive CVD plasma chemistry produced from diluted phosphine gas will be discussed in section 3.2. AFM was performed in tapping mode to observe the effect of CF4 plasma treatment on substrates and was also used to observe the nanocrystalline diamond (NCD) grains after CVD growth.Scanning electron microscopy (SEM -FEI Quanta 200 FEG) also revealed the as-grown morphology of the diamond films along with the elemental analysis from energy dispersive X-ray spectroscopy (EDS). Raman spectra were recorded with a HORIBA Jobin Yvon T64000 spectrometer using a laser (power 50 mW) light with a wavelength of 488 nm. Table 1. Linear antenna microwave plasma CVD diamond deposition conditions (film thickness was kept constant at 250 nm with in-situ laser interferometer monitoring). as can be found from the atomic force microscopy (AFM) images of the fluorinated and untreated GaN substrate surfaces, shown in Figure 2. As-received GaN had a very smooth surface (figure 2a) which is detrimental for mechanical anchoring of the coating, but it is found that CF4 plasma not only incorporates fluorine atoms on the GaN surface but also etches the surface to create planar defects (figure 2b), helpful in nucleation enhancement of CVD grown diamond. The initial difference in height between asperities and valleys for the untreated GaN surface was 15 nm, which increased to 312.4 nm after 20 s of CF4 plasma pretreatment. The CF4 plasma pre-treatment is not detrimental to the GaN two-dimensional electron gas (2DEG)properties [17]. Some additional effects of CF4 plasma treatment will also be discussed in the later sections with Raman spectroscopy results. CVD diamond growth. Table 1, samples Ia and If, common processing conditions were MW input power 2×2800W, 9 sccm CO2, 7.5 sccm CH4. Figure 3a shows the optical emission spectra (OES) spectra from the diluted phosphine-rich plasma, and figure 3b shows the OES spectra from pure hydrogen plasma chemistries used during the CVD growth of diamond films on GaN/sapphire substrates. Hα and Hβ are the two most prominent lines at about 655 nm and 485 nm due to the electronic transition from hydrogen atom inner shell 3 to 2 and 4 to 2 respectively [58,59]. All other spectral lines corresponding to different plasma species (atomic hydrogen=602nm, CH=433nm, C2=580nm) identical for both pure H2 and diluted PH3 plasma chemistries, are also present in Figure 3. The critical point to be noted is that there is no additional spectral line for diluted phosphine (the amount of P is negligible in these experiments and OES spectrometer is not sensitive enough to detect very low levels of P) plasma than from the pure hydrogen plasma, but the intensities of such spectral lines are higher (IHα 11928 vs 10810, or IHβ 5272 vs 4800) for the diluted phosphine rich gas chemistry than pure hydrogen plasma chemistry during CVD growth of diamond. The absolute intensity variation observed here is more than 10%. which is typical variation of plasma intensity during the same CVD run [60]. Figure 4 shows the diamond film morphologies that appeared on the reference Si substrates as the phosphine gas (diluted with H2) was increased gradually in the CVD precursor gas recipe (table 1, Ia-If). In the beginning, when there was no diluted phosphine in the precursor gas, only pure hydrogen was used as the main precursor ingredient; the average size of the nanocrystalline grains is about 50-150 nm (figure 4a). With the addition of diluted phosphine gas to the precursor recipe with a 5 sccm flow rate (3.3%), the average NCD grain size becomes bigger, many grains are more than 250 nm (Figure 4b). However, the effect of successive addition of diluted phosphine gas on the average grain sizes of NCDs is not significant and does not follow any trend, as evident from the corresponding morphologies in Figure 4. Wide variation of NCD grain size distribution (as small as 30-40 nm) appeared when 7.7% or 11.6 sccm diluted PH3 (figure 4c) was present in the precursor gas, whereas the bigger and more uniform NCD grain morphology (150-200 nm) was observed, when the diluted PH3 flow rate was 59.4 sccm (figure 4e, 39.6%). The respective grains (figure 4a-4e) are irregular in shape and sizes, as typical for nanocrystalline diamond films. But the morphology of the diamond film, when diluted phosphine gas (0.1% PH3 in H2) completely replaced all the pure hydrogen (100% H2) in the precursor recipe with a flow rate of 133.5 sccm (figure 4f, 89%), is observed to be somewhat different, with elongated (about 190 nm in length and 50 nm in breadth) worm-like grainy appearance. Such morphology is uniformly distributed, unlike the previous ones with lower diluted phosphine gas levels. The possible explanation for getting preferential growth along one lateral direction may be because a linear antenna CVD reactor promotes lateral growth [4956] when it is allowed to grow for longer diamond growth periods.Figure 4. Effect of phosphine gas (0.1% in H2) addition on the diamond film morphologies grown over reference silicon substrates, as evident from their SEM micrographs, from (a) to (f) with increasing phosphine gas flow rates. Figure 5 shows the effect of phosphine gas (0.1% or 1000 ppm in H2) addition on the diamond film morphologies and their qualities grown over the GaN/sapphire substrates. It has been reportedearlier [5] that under diamond CVD growth conditions the GaN substrate reacts with the hydrogen gas plasma due to harsh environment. But as shown in Figure 5a, the diamond film grown on the GaN surface with 100% pure hydrogen as the main precursor gas does not show any sign of substrate etching. The corresponding Raman signals (figure 5c) also clearly indicate the presence of microcrystalline diamond (MCDsp 3 peak position 1332 cm -1 ) in contrast to NCD morphologies simultaneously observed on the reference Si substrate, Figure 4a. However, it was also observed that substrate etching is prevalent in the case of the samples prepared with the addition of diluted phosphine gas precursor (0.1% PH3 in H2), as shown in Figures 5b 6. It might be that the phosphine-doping hydrogen plasma conditions were harsher (more reactive species with more energy) than pure 100% hydrogen CVD plasma environment. However, due to natural variation in OES spectra, harsher plasma chemistry cannot be conclusively claimed. Such harsh CVD processing conditions have clearly etched the GaN substrate, as shown in Figures 5b and6.The corresponding Raman signal (figure 5d) indicates the presence of both the microcrystalline (1331.1 cm -1 ) and nanocrystalline (trans-polyacetylene, 1140 cm -1 and 1484 cm -1 ) diamonds. The magnified peak contours shown for the position 1332 cm -1 in the insets of Figures 5c and5d are significantly different. The sp 3 peak for the pure hydrogen precursor sample is sharply present at the theoretical 1332 cm -1 position (stress free), whereas, for the sample prepared with diluted phosphine gas, the diamond peak is broadened and split into three positions (1327 cm -1 , -2.83GPa tensile, 1330 cm -1 -1.13 GPa tensile and 1335 cm -1 ) -1.7 GPa compressive), indicating smaller grain sizes and anisotropic stress present inside [61]. Diamond on GaN films reportedly [62] is associated with thermal stress (due to cooling down after CVD growth) [63] processing which is unlikely to induce thermal stress with slow cooling. In fact, thermal stress was absent in the film prepared with H2 gas, however, PH3 gas in the recipe etched the substrate surface and produced defects in the diamond-on-GaN films which is evident from the blue shift and red shift of the respective Raman signals from the theoretical sp3 Raman peak position. More evidence of GaN surface etching is also present in their corresponding Raman signatures. The film grown with pure hydrogen plasma shows a crystalline Raman peak of GaN at 571 cm -1 from the base substrate (upshifted and stressed GaN), whereas the Raman signal from the film grown under diluted phosphine plasma chemistry does not have any GaN crystalline peak from the base substrate (range not shown here for better clarity of the graph, but details will be discussed in section 3.4.). The GaN theoretical peak position should have been at 569 cm -1 (E2 mode), and there is also could have been a peak at 578 cm -1 (Eg mode) for the underlying sapphire substrate, as reported before [64]. GaN (and the top diamond film) was thick enough to prevent the Raman laser probe from reaching the underlying sapphire. In addition, there is also the appearance of a Raman peak at 670 cm -1 for the sample prepared with pure hydrogen plasma, which is typical for ion-implanted GaN on sapphire substrates, as previously reported [65]. Researchers have suggested that the appearance of a peak at 670 cm -1 is not due to the scattering from the implanted atom but due to the vibrational mode of the host GaN lattice [66]. The absence of a GaN Raman peak (full range is not shown in Figure 5d) indicates that CVD plasma conditions under a diluted phosphine environment destroyed the GaN crystal structure, as previously reported by May et al. To investigate the effect of diluted phosphine gas addition on the level of phosphorous doping inside the diamond lattice, firsthand conductivity measurements by laboratory electrometer (touching the diamond surface with two probing pins) reveals that the diamond films (table 1, samples Ia-If) deposited at the lower substrate temperature of 400 o C has very high resistance (in the order of 5-20 MΩ). eV to 140 eV binding energy range shows signature peaks corresponding to phosphorous atom 2p3/2 and 2p1/2 electron energy levels for the sample series IIe. Phosphorous has a larger diameter than carbon atoms, so lattice incorporation needs a higher diamond growth temperature to accommodate P atoms [7][8][9]. It is to be noted that, the XPS signals were received from the surface of the CVD-grown diamond films. However, XPS does not reveal the composition from the bulk.Secondary ion mass spectroscopy (SIMS) will be a more appropriate method of determining P doping levels inside the diamond films [67]. Nevertheless, considering the thin (250 nm) nature of the NCD films grown by LA MW PE CVD, XPS data proves the incorporation of P in the diamond crystal structure, at least confirmatively on the NCD film top layer surface. In the following section, such efforts with diluted phosphine precursor gas and with increasing substrate temperature on the diamond film morphology and quality will be presented and discussed. AFM image (figure 8) of the NCD grains deposited at a higher substrate temperature of 800 o C (sample IId) shows a slightly smaller and smoother film compared to the NCD grown at 500 o C (sample IIa) when diluted phosphine was the main precursor gas (table 1). The triangular-shaped grains indicate prevalent (111) faceted morphologies. The AFM image also reveals a porous NCD structure, which is further confirmed by the SEM images in Figure 9 which shows the effect of the successive rise of substrate temperatures at 600 o C (figure 9a), 700 o C (figure 9b), 800 o C (figure 9c) and 900 o C (figure 9d) on diamond morphologies. The microstructure shown in Figure 9c is from the same sample shown in Figure 8b (sample IId), which also validates the fact that the NCDs grown at higher temperatures have smaller grain sizes (250-300 nm) than NCDs grown at lower temperatures (600 o Cfigure 9a, sample IIb, average sizes 400-500 nm). Although the diamond sample (IIe) deposited at 900 o C substrate temperature, shown in figure 9d, does not (300-400 nm)show the smallest grain size of them all. Therefore, it can be inferred that NCD grain size does not follow a definitive trend with increasing substrate temperature, but they are all fully covered substrates without the evidence of any GaN substrate etching, in contrast to the earlier findings in their SEM microstructure grown at a low temperature of 400 o C (figure 5b and figure 6). The reason for not showing substrate surface etching at higher temperatures is counterintuitive. The logic behind substrate etching by the earlier report [5] was that at high temperature GaN reacts with the hydrogen gas. By using a lower substrate temperature inside the linear antenna system, substrate etching could be avoided, while using hydrogen as the precursor gas. However, when diluted phosphine was added to the gas mixture recipe, it was found to etch the GaN substrate, even at low substrate temperature. Surprisingly, the GaN substrate etching disappears as we increase the 10). So, it is concluded that the increase in substrate temperature favors diamond growth and suppresses GaN substrate decomposition. The favorable high CVD substrate temperature helped the diamond film to form quickly over the GaN substrate, thus not letting the corrosive plasma species like atomic hydrogen (Hα & Hβ) come in direct contact with the GaN substrate surfaces to cause chemical reactions or etching. Diamond crystals grow both in the vertical columns and horizontal planar direction. Quick substrate surface coverage with growing diamond layer is essential to prevent etching of GaN.Since LACVD reactor favors lateral growth over columnar growth, GaN etching is further prevented inside this type of reactor. Microcrystalline and nanocrystalline diamond films were deposited on GaN/sapphire heterostructure substrates with hydrogen as precursor gas without causing any apparent substrate surface etching, as previously reported in the literature. It was also possible to avoid having any SiN barrier layer between the diamond and GaN/sapphire substrate, which is different from the routine practice by other researchers. Three essential research problems were addressed using a linear antenna microwave plasma-enhanced CVD system to grow diamonds at a very low substrate temperature of 400 o C. However, such low-temperature CVD growth conditions were found to etch the GaN substrate when diluted phosphine gas was used as a doping precursor. Phosphine precursor gas could not render room-temperature conductivity to the growing diamond films. The highly resistive diamond film deposition was due to the low substrate temperature of 400 o C, which was not high enough to incorporate phosphorous atoms inside the diamond lattice. Therefore, low CVD growth temperature not only etched the GaN surface but was also futile in doping the diamond film. However, when higher substrate temperatures were adopted, then it was found that a substrate temperature of 900 o C could achieve phosphorous doping. Unfortunately, higher substrate temperatures (600 o -900 o C) made the diamond film porous. Contrary to the earlier literature reports, this work finds that higher substrate temperatures do not damage the GaN/sapphire substrates due to favorable conditions of CVD diamond lateral growth in competition over the underneath GaN substrate decomposition in the linear antenna CVD reactor.On the other hand, lower substrate temperature accelerated GaN decomposition, by not letting diamond film form faster enough to cover the substrate, when diluted phosphine was used as the precursor gas.The objectives of the paper were: Formatted: Line spacing: Double