Exploring SiC CVD growth parameters compatible with remote epitaxy Free

Silicon carbide is a technologically important wide-bandgap semiconductor that is commonly used in the fabrication of power electronic and opto-electronic devices due to its temperature stability, high breakdown voltage, and availability of large diameter (>6 in.) wafers. In addition, it has become a critical substrate in the field of 2D materials, as it can be annealed in an inert environment to form epitaxially oriented graphene layers.^1,2 Graphene formed this way is commonly referred to as epitaxial graphene (EG). These EG layers are highly desirable for 2D electronics due to their single crystallinity, uniformity, controllable synthesis, and high electrical mobility relative to graphene grown on other substrates.

Wafer-scale continuous graphene layers have become increasingly relevant to the larger field of semiconductor thin-film production due to a process called remote epitaxy (RE), where the graphene layers are utilized as a barrier between a single-crystal substrate and an epilayer.^3–8 If the graphene is sufficiently thin and the substrate atomic bonds are sufficiently ionic, then the potential field of the base substrate can penetrate through the thin barrier layer and direct epitaxial growth instead of the graphene lattice.⁴ Due to the chemical stability of graphene, the graphene layer can remain intact and be used as a release layer after deposition, allowing the epilayers to be mechanically exfoliated similarly to 2D material layers. This allows for expensive substrates to be reused for further epitaxial growth after the epilayers are removed. The ability to transfer thin epitaxial layers, potentially containing electronic devices such as LED heterostructures,³ to arbitrary substrates opens a wide range of possibilities for next-generation electronics. These transferred layers can have features such as flexibility, atomically abrupt interfaces, and close surface proximity to device layers inaccessible to traditional semiconductor devices tied to a bulk substrate. These properties make RE an excellent technique to fabricate advanced devices such as quantum sensors (using optically active defects very close to the surface), photonic crystal cavities, or power electronic devices with higher cooling capacity through the removal of substrate material which may have poor thermal conductivity (e.g., GaN or β-Ga₂O₃). While the potential applications and initial demonstrations of RE are encouraging, fundamental aspects of this technique are still under investigation and must be better understood to broaden the applications of this growth process.

EG provides a promising material platform for RE, complete with single-crystal graphene layers, atomically smooth surfaces, and well-defined layer numbers.⁹ Several studies have utilized a layer-resolved EG transfer step to select a single monolayer (or buffer layer) to use as the barrier;^3,10,11 however, the mechanical graphene transfer complicates device formation and can introduce physical defects or chemical impurities to the system. While EG/SiC substrates have been used for remote epitaxy of GaN (Refs. 7 and 12) and β-Ga₂O₃ (Ref. 13) thin films, little work has focused on the homoepitaxial SiC/graphene/SiC(0001) system or the effect of native EG/SiC morphology on the RE process.

This study on SiC remote homoepitaxy via CVD explores the significant modifications to the SiC growth process required for successful RE. Relative to other RE experiments, such as those involving III-V compound semiconductors, the SiC CVD growth conditions are expected to be even more severely damaging to graphene layers due to high growth temperatures (>1400 °C) combined with the use of H₂ as a carrier gas.^14–16 For this reason, this study explored modifications to the growth conditions, namely, carrier gas type, growth temperature, and C/Si ratio, to attempt to prevent EG degradation. In addition, preparation of the EG layers for RE was a key variable—using SiC substrates with different offcut angles allowed for various SiC macrostep profiles and EG layer numbers to be investigated simultaneously in a single growth run. As the EG was grown in situ before the SiC remote epitaxy, these experiments also allowed for improved process efficiency and reduced externalities affecting the graphene layer (contamination or physical damage) by avoiding ex situ transfer. The effect of the changes to the CVD growth parameters and the variations in EG morphology on the subsequent SiC RE were studied and optimal growth parameters are presented in this paper for future use in other RE experiments.

The epitaxial graphene and SiC were synthesized in a dual cell Aixtron/Epigress VP508 horizontal hot-wall reactor. Hydrogen gas (H₂), purified via a point-of-use Pd membrane purifier, was utilized as an initial etchant to remove substrate surface damage and contamination, while Ar was used as a carrier gas during growth. Silane (SiH₄, diluted 2% in H₂) and propane (C₃H₈) gas were used as Si and C sources, respectively, for SiC deposition. At our default carrier gas flow rate of 25 standard liters per minute (slm), the additional H₂ from the silane source (0.49 slm) accounted for ∼2% of the total flow. Substrates were heated by RF induction of a SiC- and TaC-coated susceptor and a TaC-coated rotating satellite. Temperature was monitored via infrared pyrometry and the reading was calibrated to the Si melting temperature. These Si melt tests yielded a temperature uniformity of ±2 °C over 100 mm at 1414 °C. Additional information on growth in this system can be found in Ref. 17.

Each growth utilized two separate 8 × 8 mm SiC coupons of different surface offcut and polytype. By changing the SiC offcut, growth temperature, and growth time, the number of EG layers can be controlled.¹⁸ This experiment used “on-axis” (<0.1° offcut), semi-insulating 6H-SiC(0001) for nominally single-monolayer (ML) EG substrates and “off-axis” (4° offcut), N-type 4H-SiC(0001) for >2 ML EG substrates. Two different offcut substrates were used for each growth to examine the effect of graphene layer thickness on SiC remote epitaxy. All SiC substrates were sourced from II-VI Advanced Materials.

A Thermo DXRxi Raman microscope using a 532 nm laser at 9.6 mW with a spot size of ∼0.7 μm (NA 0.9/100× objective) was used to acquire 20 × 20 μm² Raman full width at half maximum (FWHM) maps of the graphene 2D peak at room temperature. The FWHM of graphene’s 2D peak provides a relatively accurate measure of the number of monolayers, up to a limit of ∼3 ML.¹⁹ A Bruker Dimension FastScan AFM operating in tapping mode was used to investigate the surface morphology. A FEI Helios NanoLab G3 scanning electron microscope (SEM) was utilized to check morphology and EG layer coverage over a large area using in-lens detected secondary electron contrast to distinguish layer numbers.²⁰ Nomarski interference contrast (NIC) microscopy provided information on macrostep morphology and other surface features.

RE SiC morphology was also examined via NIC microscopy, SEM, AFM, and high-angle annular dark field cross-sectional scanning transmission electron microscopy (HAADF-STEM). Cross-sectional STEM specimens were prepared using Ga focused ion beam (FIB) milling and lift-out in the dual-beam FEI Helios NanoLab G3. The Ga ion accelerating voltage was stepped down from 30 to 2 kV during the polishing process in order to minimize damage to the material during thinning of the <100 nm thick lamellae. STEM imaging was conducted using an aberration-corrected Nion UltraSTEM system operating at 200 keV.

The process steps utilized for the SiC remote epitaxial growths with in situ EG barriers and minimal H₂ carrier gas were as follows:

1. Temperature ramp and etch: SiC(0001) substrates were etched in 5 slm of high-purity H₂ at 200 mbar while the sample temperature was ramped to growth temperature (1620 °C).

2. EG growth: After reaching growth temperature, the carrier gas was switched to Ar and in situ EG was grown in 30 slm Ar at 100 mbar for 20–60 min.

3. Ar carrier gas and temperature ramp: After EG growth, the Ar carrier gas was immediately ramped from 30 to 25 slm.

   • 3.1. If the RE SiC growth temperature was lower than the graphene growth temperature of 1620 °C, the temperature setpoint was reduced and the cell was allowed to cool. Cooling with PID temperature control to 1540 or 1450 °C generally took around 15 min. Upon reaching the setpoint temperature, the cell was allowed to stabilize for 2 min.

4. Precursor introduction: Upon stabilization, precursors were introduced to the chamber. A flow of 500 sccm dilute silane was held constant while the propane flow was varied to modify the C/Si ratio, unless otherwise noted. This process was varied to see the effect on the stability of the epitaxial graphene layer. The steps, shown in Fig. 1(b), were conducted as follows:

   • 4.1. Propane dose: A constant, small flow of propane was introduced to the reactor without any silane flow.

   • 4.2. Precursor ramp: Both propane and silane were ramped from a starting C/Si ratio to the C/Si ratio used for the remainder of the RE growth.

5. SiC RE growth: SiC was grown for 10–30 min. The range of C/Si ratios was 0.9–1.20.

6. Cooldown: After growth, the precursor gas flows were stopped and the heater was turned off to let the reactor cool to ∼1000 °C. The samples were cooled down in H₂ or Ar carrier gas. Ar flow was reduced to 5 slm for the entirety of the cooldown step. Cooling in Ar allowed for more accurate determination of film thickness, especially for thin (<1 μm) films. The gas flow of 5 slm was chosen since previous experiments found ex situ graphene was removed from 4H- and 6H-SiC substrates upon heating to 1620 °C and immediately cooling in this growth reactor at this flow. Higher flows allowed for graphene to survive.

   • 6.1. After initial cooling, the chamber pressure was increased to 950 mbar to improve cooling.

   • 6.2. After cooling for 30–75 min (to below ∼700 °C as estimated by the lack of visible blackbody radiation), gas flows were stopped and the chamber was evacuated, completing the growth.

These process steps are labeled and indicated by dashed vertical black lines in Fig. 1, along with substrate temperature (black solid lines with filled square markers) and gas flow (blue solid lines with open circles for Ar, open triangles for H₂, and open squares for precursors) as a function of time. A control growth that omitted the in situ EG growth step (step 2 in Fig. 1) was conducted to examine any differences between standard and remote homoepitaxy. A growth temperature of 1540 °C and a C/Si ratio of 1.0 were used for this control SiC growth.

Raman 2D peak FWHM maps and AFM height maps of representative EG substrates can be seen in Fig. 2. The duration of the graphene growth (step 2) for these samples was 60 min. The measured EG thickness on nominally on-axis (−0.06° offcut) 6H-SiC(0001) is a mix of monolayer (FWHM ∼36 cm⁻¹) and bilayer (∼55 cm⁻¹), with regions of bilayer thickness aligned to step edges²¹ [Fig. 2(a)]. These step edges on on-axis EG are typically on the order of 1–10 nm high, separating atomically flat terraces that are ∼5 μm wide, as evidenced by the AFM height map in Fig. 2(b). The flat terraces largely correspond with 1 ML EG regions, as is regularly seen in the literature.²² Comparatively, EG on 4° off-axis 4H-SiC(0001) shows 2D peaks with uniform FWHM of 68 ± 5 cm⁻¹ corresponding to 2–3 ML (∼trilayer FWHM ∼75 cm⁻¹) of graphene [Fig. 2(c)]. The 4° offcut from the (0001) axis produces a more densely macrostepped surface [Fig. 2(d)] with step heights of ∼50 nm and terrace widths of ∼500 nm. RMS roughness of the 4° off-axis EG is an order of magnitude higher than the on-axis EG, with the region in Fig. 2(d) having an RMS roughness of 11.4 nm compared to 1.51 nm for the region in Fig. 2(b).

When the SiC growth chamber was used to grow in situ graphene, an Ar flow of 30 slm was necessary to achieve full coverage of EG at the SiC growth temperature of 1620 °C. Test growths with parameters comparable to EG depositions that use an uncoated graphite susceptor (10 slm Ar for 20 min at 1580 °C) resulted in incomplete coverage when attempted using the TaC/SiC coated graphite susceptor in the SiC growth chamber used for this RE study. For comparison, example Raman spectral maps and AFM height micrographs of EG grown using an uncoated graphite susceptor can be seen in Fig. S1 in the supplementary material. The fact that the in situ EG required higher growth temperature and higher Ar flow (reducing the boundary layer during growth, thus promoting Si sublimation along with increasing mass transport from the substrate) is indicative of a difference in the growth environment between the two cells. A lower carbon concentration in the in situ EG growth environment due to the use of coated graphite instead of uncoated graphite may have inhibited EG formation. The lower C concentration may also be caused by residual SiC from previous RE and SiC growths in this cell, which would act as additional, potentially variable, Si sources during EG growth.

A summary of the SiC remote epitaxy growth parameters is shown in Table I.

Initial high-temperature (1620 °C) SiC growths using Ar carrier gas (HT 1 and HT 2 in Table I) produced very uniform, high-quality epilayers on both SiC substrate offcuts. By utilizing a C/Si ratio of 0.8 with only Ar carrier gas at a flow of 25 slm, films with smooth morphology without observable grain boundaries were achieved. This deposition (HT 1) on on-axis EG/6H-SiC(0001) had relatively featureless surfaces in NIC microscopy [Fig. 3(a)], while depositions on 4° off-axis EG/4H-SiC(0001) had evidence of macrosteps visible in NIC micrographs [Fig. 3(b)]. AFM height scans showed both substrates resulted in macrosteps, but on-axis substrates had wider plateau widths of several μm and macrostep heights of ∼5 nm [Fig. 3(c)], while off-axis substrates had plateau widths on the order of 100 nm wide [Fig. 3(d)]. This is similar to the starting EG surface [Fig. 2(d)]. HT 2, with the same growth conditions as HT 1 except for a reduced SiC growth duration (15 min), was used for cross-sectional TEM and attempted epilayer transfer. The cross-sectional HAADF-STEM, however, showed no evidence of a graphene interlayer surviving this deposition (Fig. S2 in the supplementary material).

With the negative results of the high-temperature remote epitaxy attempts, further growth studies focused on lower temperature (1540–1450 °C) depositions. The resulting film morphology can be seen in the NIC micrographs in Fig. 4. The SiC control growth on the on-axis substrate (without an in situ graphene barrier layer) shows high-quality surface morphology with minimal features present [Fig. 4(a)]. Figure 4(b) shows mostly the same morphology on the off-axis substrate, but with a circular region containing macrostepped morphology in the center of the micrograph. This defect likely stemmed from a Si droplet that was etched or evaporated after the silane flow ended. As the typical C/Si ratio for SiC growth in this reactor is 1.55, operating in this Si-rich regime and lower temperature may have provided excess Si resulting in the droplet. The remote epitaxial growth most comparable to the SiC control (growth A) shows a dramatically different surface morphology. Growth A on the on-axis substrate [Fig. 4(c)] was polycrystalline, with relatively large, smooth grains. The orientation of the grain boundaries appears to be somewhat anisotropic, with many edges aligning to θ = ∼45° from the bottom of the image (parallel to [11 $2 ¯$0]). The film grown on the off-axis substrate [Fig. 4(d)] is rougher than the growth on the on-axis substrate, with a polycrystalline surface that exhibits macrostep bunching. This macrostep bunching appears to be larger in scale than a standard epitaxial graphene surface. The macrosteps are oriented such that they “step down” and follow the intentional offcut direction of [11 $2 ¯$0]. The step edges are approximately parallel to [1 $1 ¯$00], the direction perpendicular to the offcut.

The modifications to the growth for growth B (longer EG growth time, additional propane dose, C/Si ramp, and lower C/Si ratio) had relatively small effects on the film morphology. The on-axis surface [Fig. 4(e)] has similar flat, anisotropic polycrystalline grains as growth A [Fig. 4(c)]. In this case, the angle of the elongated grain edges, θ, is roughly 60°–70°. This may be tied to the crystallographic direction of [ $2 ¯$110], but sample-to-sample variations suggest that the angle is more likely due to the local surface offcut of the substrate. The off-axis surface [Fig. 4(f)] is also polycrystalline, but both the lateral grain size and the magnitude of step bunching appear reduced compared to growth A.

Increasing the starting C/Si ratio for the C/Si ramp from 1.8 in growth B to 6.0 in growth C produced the most significant changes in film morphology for the growth on the nominally on-axis substrate. The resulting film [Fig. 4(g)] has elongated narrow ridges of polycrystalline material with small grains, separated by wider, flatter plateaus with larger grain sizes. These ridges tend to run along ∼45° from the bottom of the image, much like the grain edges from growth A. This morphology is analogous to the starting epitaxial graphene macrostep morphology, where atomically flat, wide terraces are separated by ∼1–50 nm high macrostep edges.¹⁷ Growth C on the off-axis substrate [Fig. 4(h)] has a higher degree of step bunching than growth B but with a similar degree of polycrystallinity, as determined by the density of horizontal grain boundaries. Increasing the pre-SiC growth propane dose by 50% from 3 sccm for 60 s (growth C) to 90 s (growth D) did not substantially change the film morphology. The higher dose growth still exhibits a similar “ridged” growth on the on-axis substrate [Fig. 4(i)] with perhaps a slightly more uniform microstructure. The difference in the ridge angle (∼80° compared to the previous ∼45°) may be more due to underlying substrate surface morphology/local substrate offcut or curvature than any differences in the epitaxial growth. There is no apparent difference in the film grown on the off-axis substrate [Fig. 4(j)]. Growth at a reduced temperature of 1450 °C and a higher final C/Si ratio of 1.2 (growth E) produced the most disordered film morphology, although linear features could still be resolved in both the on-axis [Fig. 4(k)] and the off-axis film [Fig. 4(l)]. The on-axis film had linear features oriented ∼80° from [11 $2 ¯$0].

The fact that the remote epilayer in growth A is polycrystalline, while the SiC control growth is single-crystalline provides evidence that the epitaxial graphene barrier layer is negatively affecting the epitaxial alignment of the remote epilayer. This implies that the graphene layer may be too thick at the point of nucleation and is preventing the underlying substrate penetrative electrostatic field from directing the alignment of the film. To corroborate this theory, cross-sectional STEM was required in order to inspect the growth interface for the presence of graphene. Figure 5 shows HAADF-STEM of the approximate growth interface of growth A, which was determined by the starting point of 3C-SiC inclusions as these inclusions are not present in the starting substrate and only occur due to the stability of the polytype during these CVD conditions.

To accurately determine where the 3C regions begin, an atomic-resolution HAADF-STEM image was Fourier filtered to produce contrast between 3C and 4H polytypes. The original HAADF-STEM image can be seen in Fig. 5(a), with the 2D fast Fourier transform (FFT) pattern shown in the inset. The red circles on the FFT pattern represent the openings in the masked FFT pattern—only these reflections were used for the inverse FFT (IFFT) seen in Fig. 5(b). The approximate growth interface was determined by the start of the 3C region in Fig. 5(b), labeled as an orange dashed line. To determine if graphene was present at this interface, an integrated line profile of the HAADF-STEM contrast was taken across the interface [along the light blue arrow in Fig. 5(c)]. The region was averaged across the width of the rectangle (along the Si atomic columns) to increase the signal-to-noise ratio. As HAADF-STEM contrast is mass-thickness dependent (approximately proportional to Z²), the difference in atomic mass between SiC and pure C atomic columns will produce a darker contrast for graphene in a brighter SiC matrix. The profile in the inset of Fig. 5(c) only shows uniform contrast associated with SiC, with no dip in contrast as expected from layers of graphene. Other regions (not pictured) were also probed for the presence of graphitic carbon, but only uniform SiC was found.

Our hypothesis for the lack of graphene at the interface in Growth A is that the initial SiC growth conditions were etching the epitaxial graphene before SiC could be deposited on top of the layer. As graphene can be etched in high-temperature (>1500 °C) hydrogen environments,²³ it is likely that hydrogen in the dilute silane source, the majority contribution to the 2% H₂ overall concentration in the reactor, was responsible. Unfortunately, no other silane source was available to reduce the H₂ concentration for these experiments. In an effort to protect the epitaxial graphene from the effects of H₂, Growth B incorporated a pre-SiC propane dose and C/Si ramp as outlined in Fig. 1(b). The rationale for this decision was the protective carbon layer may shield the underlying EG from H₂ in the silane as the precursor was slowly introduced. The slow introduction and ramp from C-rich to Si-rich was attempting to nucleate SiC in a C-rich (minimal H) environment, which might bury the EG and protect it during growth with a more nominal (H-rich) C/Si ratio. This strategy also prevents any conversion of graphene to SiC via excess silane/Si. This conversion has been shown to occur if graphitized carbon-face SiC is exposed to a silane flux at 1100 °C.²⁴ A longer EG growth duration was also incorporated to provide a slightly thicker initial graphene layer.

While the attempted methods were far from ideal conditions for SiC growth, the goal was graphene survivability over high SiC crystalline quality. Despite that, NIC micrographs in Figs. 4(e) and 4(f) show the resulting films had comparable morphology to Growth A for depositions on both on-axis and off-axis substrates, respectively. Increasing the initial C/Si ratio from 1.8 to 6.0 resulted in a dramatic degradation in film microstructure, as evidenced by the ridge-like structure of Growth C in Fig. 4(g). Growth D also utilized a 50% longer propane dose than Growth C and resulted in smoother terraces than Growth C. The Growth D on-axis sample was cross-sectioned for HAADF-STEM imaging (Fig. 6). In this growth, islands of multilayer graphitic carbon could be resolved at the approximate growth interface through the low contrast associated with a carbon particle in a SiC matrix. The graphitic carbon region is spanned by an orange dotted box in the line profile in Fig. 6(a). HAADF-STEM contrast line profiles in an atomic-resolution image [Fig. 6(b)] show the growth interface to the left of the carbon particle has a uniform contrast associated with pure SiC and a lack of a graphene interfacial layer [blue line #1 in Fig. 6(c)]. In comparison, a line profile through the graphitic carbon particle [orange line #2 in Fig. 6(c)] shows a measurable dip in contrast indicative of the lower atomic mass of the carbon. The carbon was determined to be largely graphitic in nature as it exhibited a laminar structure and the measured interplanar spacing of 3.23 Å is close to that of graphite (3.35 Å) instead of SiC (2.52 Å). Ripples, overlap with other SiC planes in the FIB lamella, and other distortions in the graphitic carbon lattice can be responsible for the slight mismatch in measured interplanar spacing compared to pure graphite. While other regions outside of the displayed area in Fig. 6 were inspected, no signs of a uniform carbon layer at the growth interface were detected in this sample. However, the fact that some patches of graphitic carbon survived the remote epitaxial growth procedure is a positive indicator of improvement brought about by the changes in precursor introduction.

With these initial results, additional HAADF-STEM was conducted on the lower-temperature Growth E. During the FIB lift-out process, voids in the film [black regions indicated by the (yellow) arrows in Fig. 7(a)] were detected using plan-view SEM. A HAADF-STEM cross section of one such void is shown in Fig. 7(b). These voids may be due to the low temperature (1450 °C) and reduced growth time (10 min) resulting in incomplete coalescence of the film, but they may also be indicative of thick patches of graphitic carbon. These regions of deposited graphitic carbon may have locally prevented SiC nucleation, resulting in film voids. This hypothesis was corroborated by Raman spectral mapping of this sample [Fig. 7(c)] that showed similar elongated regions with much higher intensity graphitic carbon 2D peaks than the rest of the film. While hard to distinguish in the optical micrographs from the Raman mapping system, the density of these linear spectral features implies that most of the graphitic carbon regions appeared to be buried under the SiC remote epilayer, with only some regions resulting in the voids seen in SEM. Away from these streaks of high 2D peak area, a relatively weak, but uniform Raman 2D peak was detected. A representative full Raman spectrum [Fig. 7(d)] from a region away from the streaks of high 2D peak area [white rectangle in Fig. 7(c)] shows a substantial D peak at 1365 cm⁻¹, indicating that the 2D peak may correspond to graphitic carbon with a high defect density.

The HAADF-STEM cross section of a void in the film shows that graphitic carbon has survived underneath the 3C-SiC epilayer to either side of the void [Fig. 7(b)]. Despite a longer lateral continuous graphitic carbon film (90 nm long region found under the SiC epilayer) than the past depositions, no evidence of graphitic carbon was found on either side of this region. The exact structure of the graphitic carbon in the center of the void in the film is unfortunately impossible to accurately determine due to potential FIB damage. The changes in SiC film coverage resulted in nonuniform milling of material, with more material milled where there was no protective SiC epilayer over the voided region. Likewise, the region in the center of the micrograph that appears to be a depression into the SiC substrate may be related to carbon embedding into the substrate during CVD growth or may also be an artifact from FIB preparation. Additional TEM studies using advanced sample preparation are necessary to rule out FIB-related artifacts.

The cross-sectional HAADF-STEM studies show various degrees of graphitic carbon survived the SiC epitaxial growth process, but none contain the expected uniform ∼2–3 ML epitaxial graphene barrier. The continuous Raman map signal may again be due to the convolution of the Raman laser spot width/step size and the smaller sections of thick, isolated graphitic carbon resulting in an apparently continuous map despite gaps in the graphene coverage. While it may be reasonable to assume the few-monolayer epitaxial graphene layers were etched away by the 2% H₂ in the reactor environment, it does not explain the formation of thick, isolated regions of graphitic carbon.

To attempt to gain a better understanding of the stable species in this growth environment, NASA’s Chemical Equilibrium for Applications (CEA) program was utilized to calculate the equilibrium species for the growth conditions.²⁵ It is important to note that this does not take into account the kinetics and fluid dynamics of CVD growth, nor does it take into account the effects of solid surfaces (such as heterogeneous nucleation). It is only a simple approximation of the stable species in the reactor environment and not a full model of SiC growth. The results of these calculations can be seen in Fig. 8. These plots show the majority of stable species as a function of temperature for the C/Si ratios used in the initial stages of the remote epitaxy C/Si ratio ramp (1.8 and 6.0). The specific temperatures for each growth at the given C/Si ratio are indicated by dotted vertical lines.

The conditions for Growth B [dashed line in Fig. 8(a)] show approximately equivalent equilibrium concentrations of solid SiC (orange), atomic H (dotted light blue), and graphitic carbon (dashed green) at a mole fraction of ∼2 × 10⁻⁴. Conversely, the growths that utilize a starting C/Si ratio of 6.0 and a growth temperature of 1540 °C [Growths C and D, as seen in Fig. 8(b)] have around an order of magnitude greater equilibrium concentration of graphitic carbon than atomic H and a factor of 5 greater than SiC. Reducing the temperature to 1450 °C reduces the concentration of atomic H even further, but the ratio of graphitic carbon to SiC is largely unchanged.

Qualitatively, this result provides insight that (1) equilibrium graphitic carbon concentration is mainly determined by the C/Si ratio and is largely independent of temperature and (2) atomic H concentration increases approximately exponentially with temperature, while SiC and graphitic carbon concentrations remain predominantly constant. The first insight is at odds with our experimental findings, as Growths C, D, and E all have the same initial C/Si ratio of 6.0, but Growth E had substantially more graphitic carbon present at the growth interface and had higher Raman signatures of sp² carbon. Importantly, Growth E had a lower Growth temperature (1450 °C) than C and D (1540 °C). This suggests that temperature may be a larger factor for the stability of graphitic carbon on a SiC surface as compared to the stability in a homogeneous gas phase, since the gas-phase/precipitate graphitic carbon concentration in the CEA result is relatively insensitive to temperature.

Increased etching via atomic H at higher temperatures and the effects of SiC surface sublimation can explain the discrepancy between the CEA results and the experimental data. At lower temperatures, there is a lower concentration of atomic H (Fig. 8, long-dashed green line), which more readily etches graphitic carbon than molecular H₂.¹⁶ The competing effects of temperature and C/Si ratio on graphitic carbon and atomic H concentration are visually represented in Fig. 8(c), which shows the mole fraction ratio of C/H versus temperature and C/Si ratio. It clearly shows that lower temperatures and higher C/Si ratios increase the relative concentration of graphitic carbon over atomic H. This agrees with our hypothesis that stable graphitic carbon will more likely survive the SiC RE step in these conditions, at least partially due to reduced etching by atomic H.

In addition, higher temperatures may result in a competition between SiC etching and graphene growth that may limit the graphene thickness and reduce the efficacy of the pre-Si propane dose. It has been found in propane-based CVD graphene growths that the graphene thickness actually decreased with increasing propane dose for higher temperature (1550 °C) depositions at 100 mbar.²³ Since the conditions in that study are similar (albeit with different H₂ carrier gas concentrations) to our pre-RE propane dose and initial RE growth conditions (with high C/Si ratios), a similar deposition/etching competition may be preventing our protective carbon layers from forming with our pre-RE propane dose. It may also influence the initial stages of SiC growth at high C/Si ratios. This competing effect was found to be temperature dependent in Ref. 23, where the lower temperature (1350 °C) instead showed an expected monotonic increase in graphene thickness with increasing propane dose. This limitation of graphene thickness would be due to the SiC surface decomposing and would not be captured in the CEA calculations.

Both of these temperature-dependent phenomena may explain the lower experimentally observed graphitic carbon concentration on the SiC surface at increased temperatures compared to the CEA gas-phase simulations. Overall, these simulations provide trends as a function of temperature and relative chemical composition and show that it may be possible for graphitic carbon to exist at these growth conditions, but they are relatively simple models that only account for a uniform gas phase reaction.

While increasing the C/Si ratio to 6.0 does have an improved effect on the stable concentration of graphitic carbon, these results still do not necessarily explain the change from thin, uniform epitaxial graphene to clusters of thick graphitic carbon seen in the cross-sectional STEM images. Further simulations that account for growth kinetics and adatom mobility may be required to understand any clustering effects from surface migration that could explain the thicker regions of graphitic carbon. One possible explanation for the thicker graphitic carbon regions is that the stable concentration of carbon seen in the CEA simulations for high C/Si ratios indicates that carbon deposits/gas-phase reactions are likely to occur. The anisotropy of the patches on the surface [highly aligned streaks of sp² carbon as seen in Fig. 7(c)], however, make depositions from the gas-phase reaction to be an unlikely cause unless significant migration on the SiC surface occurs, as any particulates forming in the gas phase are expected to be spherical particles.

Instead, it may be that graphitic carbon preferentially nucleates on step edges of the substrate, explaining the anisotropy of the graphitic carbon clusters. Preferential 3D growth of these nuclei over homogeneous nucleation in the middle of an EG-covered terrace would explain their large size and lack of uniform graphene deposition across the sample. These larger carbon clusters could then locally block SiC nucleation, eventually leading to their overgrowth by surrounding SiC. If carbon is kinetically driven to cluster, the surface C/Si ratio away from these clusters may be more Si-rich, potentially making EG/graphitic carbon thermodynamically unstable away from the carbon clusters. This would explain the lack of EG outside of the larger carbon clusters. Confirming these hypotheses, however, requires more advanced simulations that consider surface adatom kinetics and local chemical equilibria that are beyond the scope of this work.

It may be that reduction of atomic H, and therefore etching, is more important than an increased stable concentration of graphitic carbon as it may help maintain the high-quality, uniform epitaxial graphene layer. These CEA results show that it is possible to independently reduce atomic H without dramatically affecting the graphitic carbon stability concentration by lowering the temperature, indicating that lower temperature growths may be a path forward for exploration into graphene stability during remote epitaxy. Of course, simply lowering the growth temperature may result in poorer film quality and these low temperature growth attempts would require optimization of the C/Si ratio to prevent Si droplet formation (Si-rich growth) and undesired thick graphitic carbon deposition (C-rich growth). Further growth studies and characterization of the propane source in an Ar ambient may also be required to fully explain the carbon clusters.

A uniform, thin carbon deposition is critical to the success of remote epitaxy—nonuniform deposition may not allow for a protective carbon layer to fully cover the epitaxial graphene barrier, exposing it to the damaging effects of hydrogen during the initial stages of remote epitaxy. The multilayer graphitic carbon islands will also be too thick to allow for remote epitaxy, introducing crystallographic defects and preventing the continuation of the underlying substrate crystal lattice. If the use of propane in Ar carrier gas is shown to fundamentally result in nonuniform deposition, moving toward remote epitaxy with CVD graphene barriers, which can utilize H₂ carrier gas, may be necessary.

Growth parameters for SiC/EG/SiC(0001) remote epitaxy, specifically C/Si ratio, growth temperature, and precursor dosing procedure, were examined for their effect on the graphene interface and the epilayer crystalline quality. These parameters were tested on both 1 ML on-axis EG and ∼3 ML 4° off-axis EG/SiC substrates. The EG layers were grown in situ directly before the RE growth for improved cleanliness and to eliminate the need for any transfer process. It was found that smooth surface morphology was achieved with 1620 °C growths using a C/Si ratio of 0.8 using 2% H₂ in Ar as a carrier gas, showing high-quality epilayers could be grown with minimal H₂ (2% from the SiH₄ dilutant). Despite the reduction in H₂, graphene could not be detected at the epilayer/substrate interface using these conditions. The growth was then modified to minimize exposure of the EG barrier to hydrogen present in the silane source through staggering the introduction of the SiC precursor gases and ramping from a carbon-rich environment to the nominal growth C/Si ratio. Lower growth temperatures were also used to minimize hydrogen etching by reducing the concentration of atomic H. These modified growths resulted in detectable levels of graphitic carbon at the growth interface of polycrystalline films, but cross-sectional TEM imaging showed the carbon layer to be discontinuous. Without a continuous barrier, any epitaxial growth was the result of epitaxial lateral overgrowth instead of remote epitaxy. That said, the findings from our experiments are promising for further remote epitaxy experiments, as they show the feasibility for graphitic carbon to remain stable at the growth interface throughout SiC CVD growth. Future work will focus on producing uniform, monolayer graphene barriers instead of discontinuous, multilayer graphitic carbon islands. Complete elimination of H₂ through Ar-diluted precursor sources will also be explored.

See the supplementary material for additional characterization data. Figure S1 contains Raman spectral maps and AFM height micrographs of the graphene 2D peak FWHM for on-axis and 4° off-axis EG grown using an uncoated graphite susceptor. These data can be directly compared to Fig. 2, which shows the same characterization of EG grown using a coated graphite susceptor. Figure S2 contains a representative cross-sectional HAADF-STEM micrograph of growth HT 2. The STEM analysis of this sample did not provide any evidence that graphene survived the subsequent SiC deposition.

Research at the NRL was supported by the ONR. This research was performed while D.J.P. held an NRC Research Associateship award and while J.R.H. held an ASEE postdoctoral fellowship at NRL. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of the U.S. Naval Research Laboratory, NRC, ASEE, the U.S. Navy, or the U.S. government.

The authors have no conflicts to disclose.

Daniel J. Pennachio: Conceptualization (equal); Data curation (lead); Formal analysis (equal); Investigation (equal); Methodology (equal); Visualization (lead); Writing – original draft (lead); Writing – review & editing (equal). Jenifer R. Hajzus: Conceptualization (equal); Data curation (supporting); Formal analysis (equal); Investigation (equal); Methodology (equal); Visualization (supporting); Writing – review & editing (equal). Rachael L. Myers-Ward: Conceptualization (equal); Funding acquisition (lead); Investigation (equal); Methodology (equal); Project administration (lead); Resources (lead); Supervision (lead); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.