Effects of aluminum on epitaxial graphene grown on C-face SiC

The effects of Al layers deposited on graphene grown on C-face SiC substrates are investigated before and after subsequent annealing using low energy electron diffraction (LEED), photoelectron spectroscopy, and angle resolved photoemission. As-deposited layers appear inert. Annealing at a temperature of about 400 °C initiates migration of Al through the graphene into the graphene/SiC interface. Further annealing at temperatures from 500 °C to 700 °C induces formation of an ordered compound, producing a two domain √7× √7R19° LEED pattern and significant changes in the core level spectra that suggest formation of an Al-Si-C compound. Decomposition of this compound starts after annealing at 800 °C, and at 1000 °C, Al is no longer possible to detect at the surface. On Si-face graphene, deposited Al layers did not form such an Al-Si-C compound, and Al was still detectable after annealing above 1000 °C.


I. INTRODUCTION
Since the experiment realization of graphene as the first two-dimensional atomic crystal by Novoselov et al., 1 graphene has become one of the hottest research fields.3][4] Thermal decomposition of SiC is regarded as a powerful method for large area epitaxial graphene fabrication directly on a semi-insulating substrate. 5Moreover, this technique is compatible with the existing CMOS fabrication technology. 6arge and homogeneous monolayer graphene on Si-face SiC substrates was achieved using high temperature sublimation under 1 atm Ar pressure. 7,8On C-face SiC substrates, however, much smaller domain sizes and multilayer graphene were typically obtained. 9,10Growth on C-face SiC is faster, requires lower temperature 9,10 and gives graphene with a higher mobility.An order of magnitude higher electron mobility has been reported for C-face graphene, 6 compared to Si-face graphene.Suitable metal/graphene contacts are required for device fabrication.To investigate if and how a deposited metal layer affects C-face graphene is therefore of interest and also to verify if the effects are similar or not as on Si-face graphene.
To date, only a few such investigations [11][12][13] have been reported concerning graphene grown on SiC(000 1).Our previous study 14 of the effects of deposited Al layers on graphene grown on Si-face SiC revealed an annealing temperature of 400 C to be required for decoupling the interaction between the carbon buffer layer and the substrate, and consequently transform from single layer to bi-layer graphene electronic properties.Different ordered phases were also found to form on the surface after annealing at temperatures of about 750 C and above.To determine if this is also the case for graphene grown on C-face SiC, experiments for a similar set of sample preparations and measurements are reported here.We find that an annealing temperature of around 400 C is required also for C-face graphene samples, for Al to be able to penetrate through the graphene layers and into the graphene/substrate interface, which is similar to the Si-face case.Aluminum silicide is observed after annealing at temperatures from 500 C to 700 C, and the formation of an ordered compound is revealed by the appearance of a new two domain ͱ7Â ͱ7R19 low energy electron diffraction (LEED) pattern as well as the significant changes induced in the core level PES spectra.Surprisingly, the graphene signal then becomes extremely weak and the C 1s spectrum is dominated by a broad component, overlapping the carbon signal from the substrate.From the simultaneous changes observed in the Si 2p and Al 2p spectra formation of an ordered Al-Si-C compound is suggested to occur in this temperature range.This pronounced decrease in graphene signal was not observed on Si-face graphene, 14 where formation of mainly aluminum silicide was suggested.After annealing at 800 C, the present ordered Al-Si-C compound starts to decompose, the Al content in the surface selvedge region starts to decrease, and the graphene C 1s component starts to increase, gradually with increasing temperature.No Al was possible to detect after annealing at 1000 C or higher temperatures, which also is different when compared to Si-face grapheme, where an Al signal was still possible to detect after annealing at 1200 C.

A. Sample preparation
The graphene samples on C-face SiC were prepared using the procedure described in Ref. 15: first heating at 800 C under a Si flux to create a Si rich 3Â3 reconstructed a) surface, followed by in-situ heating at around 1050 C for a few minutes to obtain a C-rich 3Â3 reconstructed surface.The substrates were n-doped 6H-SiC (000 1) from Si crystal.Three samples, each with an average graphene coverage between 1 and 2 ML, were typically investigated in each case.Al was then thermally evaporated in-situ onto the graphene sample kept at room temperature.From measured core level intensity ratios, the amount of deposited Al was estimated to be $10 A ˚and $4 A ˚for the samples studied with photoelectron spectroscopy (PES), LEED, and angle resolved photoemission (ARPES), respectively, as presented below.Annealing was performed in situ during 1 min at each specified temperature.

B. Characterization
The experiments were carried out at beamline I311 and I4 at the MAX-lab synchrotron radiation laboratory.Beamline I311 is equipped with a modified SX-700 monochromator, which provides light for two end-stations.The first station is equipped with a conventional LEED optics in the preparation chamber and a large hemispherical Scienta electron analyzer in the analysis chamber.A total energy resolution of 100 meV at photon energies from 33 to 450 eV and 300 meV at photon energies up to 750 eV was selected in the high resolution PES studies the C 1s, Si 2p, and Al 2p core levels reported.Only relative binding energies are given, which are referenced to the G component in the C 1s spectra, the substrate SiC component in the Si 2p spectra and the surface A1 component in the Al 2p spectra.The second end-station is equipped with a spectroscopic photoemission and low-energy electron microscope (SPELEEM) instrument.This microscope has a spatial resolution better than 10 nm in the LEEM mode and also selected area LEED (micro-LEED) patterns were collected on this instrument.ARPES experiments were carried out at beamline I4, which is equipped with a SGM monochromator and a PHOIBOS 100 2D CCD Specs energy analyzer.The Wide Angle lens mode (WAM) was utilized that provided an acceptant angle of 614 .The base pressure was about 1 Â 10 À10 mbar in all three end-stations used.

III. RESULTS AND DISCUSSION
The morphology and order of graphene prepared using this in situ procedure on C-face SiC were checked using LEEM, and selected area LEED and typical results for one sample are displayed in Fig. 1.The LEEM image in Fig. 1(a) shows the presence of small domains with different contrast over the whole surface, which is typical for graphene samples prepared this way.The micro-LEED pattern in Fig. 1(b) shows a mixture of a "ring-like" graphene diffraction pattern and a 3Â3 reconstructed substrate.This indicates a nonhomogenous surface containing small domains of ordered single-and multi-layer graphene of different azimuthal orientations as well as areas on the substrate having a smaller concentration of carbon producing the 3Â3 reconstruction.
In the conventional LEED setup in the PES end-station, the "ring-like" graphene pattern of the as prepared graphene samples was barely visible, as illustrated in Fig. 2(a).
However, the (3Â3) reconstruction appeared quite pronounced.Therefore, we first checked if Al deposition and annealing at different temperatures induced any structural changes.After Al deposition, only an overall lower diffracted intensity resulted but after annealing at 500 C, additional diffraction spots appeared, as shown in Fig. 2(b).Most pronounced is the extra set of hexagonal spots, indicating the presence of a phase having a somewhat different lattice constant than the SiC substrate.Upon closer inspection, however, twelve weaker diffraction spots oriented in a circle around each of the zero and first order hexagonal spots (enclosed in the rings of Fig. 2(b), see supplemental material for details 16 ) appear at the same time, which indicates the presence of an ordered two-domain ͱ7Â ͱ7R19 phase on the surface.Such a ͱ7Â ͱ7R19 LEED pattern was earlier observed in a study of Au/SiC(000 1) and suggested to originate from Au-silicide formation. 17This reconstruction remains after annealing at 600 C, see Fig. 2(c), but after annealing at higher temperatures, the diffraction pattern gradually transforms back to the original 3Â3 reconstruction (see Fig. A set of C 1s spectra collected before and after Al deposition and after annealing at different temperatures is shown in Fig. 3 when recorded using (a) 450 eV and (b) 330 eV photons.For the as prepared initial sample, the spectrum is composed of two components, the graphene (G) and bulk SiC substrate (SiC) peaks.The G/SiC intensity ratios provide, using a simple layer attenuation model, an estimate of 0.9 ML for the average surface carbon layer thickness.This carbon layer represents most probably a mixture of graphene and the surface carbon enrichment that give rise to the 3Â3 reconstruction.Earlier studies of initial graphitization of the SiC(000 1) surface 18,19 have shown that the binding energy of graphene, and these additional carbon atoms on the surface are very similar.After Al deposition, the separation between the G and SiC peaks increases slightly, ca.0.1 eV, and increases further, ca.0.4 eV, after annealing at 400 C when the bulk component is labeled SiC'.The graphene component is surprisingly/dramatically reduced after annealing at 500 C, i.e., at the temperature when the LEED pattern showed presence of an ordered two-domain ͱ7Â ͱ7 R19 reconstructed phase.This has as far as we know not been observed earlier, that graphene prepared on SiC is almost totally destroyed/eliminated, after annealing at such a low temperature.The two new and dominating components, C1 and SiC1, are interpreted to originate from an Al-C-Si ternary compound formed on top of the SiC substrate, but underneath the graphene remaining on the surface, and a SiC substrate component that is shifted due to the ternary compound formed at the interface.This interpretation is supported by the C 1s data shown in Fig. 3(b), collected using 330 eV photons (and also by the Si 2p data discussed).This photon energy provides a higher surface sensitivity and therefore, a considerably larger G/SiC intensity ratio, as illustrated by the bottom curve in Fig. 3(b) in which the substrate SiC signal shows a considerably lower relative strength than in Fig. 3(a).After annealing at 500 C, the new and dominating SiC1 and C1 components appear at fairly similar binding energies as the shifted bulk SiC' peak at the lower temperature, while the strongly reduced G component shows a somewhat higher relative strength in more surface sensitive spectrum in Fig. 3(b).This suggests that most of the surface carbon layer has been transformed into an Al-C-Si compound on top of the SiC substrate.The considerably larger width of the SiC1 peak, compared to the SiC and SiC' peaks present at lower temperatures, suggests that this peak probably contains two or more unresolved components.The LEED data above showed that the 3Â3 reconstruction disappeared and instead, the two-domain ͱ7Â ͱ7 R19 reconstructed pattern appeared, indicating the formation of an ordered phase with a different periodicity on top of the SiC substrate.This very efficient reduction/elimination of graphene induced by a metal adsorbate upon heating at around 500 C has not been reported earlier, only in the case of adsorbed Si layers has some transformation of graphene into SiC been reported. 20The newly formed Al-C-Si compound is not stable after annealing at temperatures at and above 800 C, it then starts to decompose and contribute to regrowth of graphene.At the same time, the SiC1 component becomes much sharper, which suggest that it again originates mainly from SiC in the substrate.At 900 C, the C 1s spectrum contains a prominent G component, see Fig. 3, and above $1000 C, it closely resembled the spectrum recorded from the initial surface with no detectable SiC1 or C1 components.After annealing at 1100 C, the G/SiC intensity ratio is considerably higher when compared to the initial surface, which indicates that this annealing created a thicker surface carbon layer.
The set of Si 2p spectra shown in Fig. 4(a) recorded at a photon energy of 240 eV, before and after Al deposition and after annealing exhibits spectral changes with annealing temperature that correspond quite well with those in the C 1s spectra.For the initial sample, the spectrum is composed of one spin-split Si 2p doublet, originating from the bulk SiC substrate.No noticeable change in the spectrum is observed.After deposition of Al, only a slight shift, 0.1 eV, of the doublet is discernable.However, after annealing at 400 C, the spectrum clearly contains two doublets, a bulk peak now labeled SiC', shifted 0.55 eV, and an additional component labeled S1 shifted ca.0.9 eV to lower binding energy in both cases.The appearance of these shifted components verifies 14,20,21 that the deposited Al has penetrated the surface carbon layer, reacted with Si and induced changes in the dipole layer at the interface.The additional component labeled S1 is assigned to silicon in the interface region, i.e., in the uppermost Si-C bilayer, that has reacted with Al during annealing.It is evident that it originates from atoms located on top of the SiC substrate from the clear decrease in the S1/SiC' peak intensity ratio with increasing photoelectron kinetic energy, illustrated in Fig. 4(b).That Al actually has started to react is clearly demonstrated by the appearance of additional broad and shifted components in the Al 2p spectra, as presented below.Upon increasing the annealing temperature to 500, 600, and 700 C, the S1 and SiC 0 components are observed to gradually broaden and merge into one broad peak at 700 C. When checking the energy separation between the centroid of this broad Si 2p feature and of the SiC1 component in the C 1s spectrum, using a peak fit procedure, one finds an average value of 182.27 6 0.07 eV for these three cases.The average value extracted between the SiC and SiC' components in the C 1s and Si 2p spectra for the other six cases, i.e., for the clean surface after Al deposition and heating at the other temperatures, is 182.23 6 0.08 eV.Thus, the energy separation between these components remains the same, within error bars, during formation and decomposition of the ordered surface compound, which can be taken to suggest growth of SiC at these low temperatures.However, a temperature of at least 850 C has typically been reported, which was required earlier 22 for growth of SiC and therefore, we suggest formation of an Al-Si-C ternary compound after annealing at temperatures between 500 and 700 C and of an Al-Si only at 400 C.After annealing at 800, 900, and 1100 C, the Si 2p spectrum again only contains a single doublet peak, which also clearly indicate that the newly formed Al-C-Si compound decomposes after annealing at temperatures above 700 C, and that the peak now originates mainly from SiC.Thus, a mixture of reactions at the interface region occurs; first, Al-Si formation and then consumption of graphene followed by formation of an ordered ternary Al-Si-C compound, resulting in the ͱ7Â ͱ7 R19 reconstruction observed by LEED.
The formation of Al containing compounds after annealing is clearly revealed by the Al 2p spectra, shown in Fig. 5, which were collected using a photon energy of 140 eV.One spin-split Al 2p doublet only (labeled A1), at a binding energy of 72.7 eV, is observed after deposition.It originates from metallic Al on the surface and has a very similar peak width as determined earlier for metallic Al on graphene/ SiC(0001). 14After annealing at 400 C, the intensity of the A1 component is reduced by a factor of two, and a wide broad feature has developed on the high binding energy side, indicating that part of the Al has reacted chemically.Since the Si 2p spectrum at this stage shows the additional S1 component, while the C 1s spectrum show only an increase in the separation between the G and SiC' components, formation of an aluminum silicide (Al-Si) at the interface is quite obvious.However, after annealing at 500 C, the broad feature becomes the dominant (see Fig. 5), modeled by two broad components A2 and A3, while the A1 has decreased even more, although still present.This is at the temperature when the G component has almost disappeared, and the SiC1 is the dominant in the C 1s spectrum, and the S1 and SiC 0 components are broadened and almost merged in the Si 2p spectrum.Formation of a ternary Al-Si-C compound in the surface region is therefore inferred to occur from this temperature and up to 700 C. At 800 C, the A1 component is no longer visible and these compounds have started to decompose, but some Al is still detectable in the surface region.Only after annealing at !1000 C, there is no Al found to remain within the probing depth, so no metal poisoning effect as reported for other cases earlier 14,21 occurs for Al deposited on graphene/SiC(000 1) samples.
The results are significantly different to our earlier findings for Al layers on graphene/SiC(0001) 14 for what concerns annealing temperatures higher than 400 C. In the present case, the graphene is found to be almost completely consumed and a ternary ordered Al-Si-C compound to form in the temperature range of 500-700 C. At higher temperatures, this compound is found to decompose and after heating at 1100 C, no Al is detectable on the surface and a graphene/SiC(000 1) structure is recreated although with a thicker layer of surface carbon/graphene than for the initial surface condition.A possible reason why so different reactions are observed for Al on C-face graphene compared to Si-face graphene, we suggest it is because of the very different morphology of these graphene films.It is well known that C-face graphene grows faster, in smaller domains (grains), with different azimuthal orientations and with a large spread in the number of layers in different grain.On the Si-face, the first carbon layer formed is tightly bound to the substrate and does not show the characteristic p-band of graphene.Graphene layers grow on top of this carbon buffer layer and form considerably larger domains with a much smaller spread in the number of graphene layers compared to on the C-face.Formation of small islands with different number of graphene layers and of different azimuthal orientations produces an abundance of grain boundaries and edges on C-face graphene.The corresponding terminating C of the C-face graphene has domains and effectively more reactive sites for reaction with Al when compared to Si-face graphene.
Other Al on graphene/SiC(000 1) samples were prepared for the purpose to study the p-band structure.ARPES spectra recorded at around the K-point of the graphene Brillouin zone using a photon energy 33 eV are shown in Fig. 6.Two intersecting p-bands separated by 8 are observed on the initially prepared sample that showed a 3Â3 reconstructed surface.This is due to that graphene grown on the SiC(000 1) surface forms smaller domains having different azimuthal orientations, and the Dirac point for graphene on C-face SiC is normally located fairly close (within 0.1 eV) to the Fermi level. 10,13,19,23However, the preparation method used in this case, Si deposition and annealing, resulted in slightly ndoped graphene with the Dirac point $0.18 eV below the Fermi level, see Fig. 6(a), which corresponds to an estimated 24 doping concentration 2 Â 10 12 cm À2 , if a linear dispersion of p-bands is assumed around the K point.No distinct changes in the p-bands were possible to identify after Al deposition and annealing up to $500 C.After annealing, the sample at $700 C, the p-bands showed a considerably lower intensity, as illustrated in Fig. 6(b).This unambiguously reveals the almost total consumption of graphene, as suggested from the core level data presented above.The Dirac point is in this case shifted downwards slightly further, by $0.08 eV, indicating a slight increase in the n-type doping upon formation of additional SiC and the ternary Al-Si-C compound.After annealing at 800 C and higher temperatures, the p-bands again appear sharper and more intense, i.e., showing a higher signal to background noise ratio.At 1100 C, re-growth of graphene can be concluded, see Fig. 6(c), and a lower n-type doping, since the Dirac point is now located at $0.13 eV below the Fermi level.These observations are well in line with both LEED and PES results presented above.

IV. CONCLUSIONS
Detailed LEED, PES, and ARPES investigations of how deposited Al layers affects graphene samples prepared in-situ on SiC(000 1) substrates are reported.Annealing is required before any effects are observed.A temperature of about 400 C is found to be required to trigger migration of Al through the graphene and into the graphene/SiC interface.Further annealing at temperatures from 500 C to 700 C results in the formation of an ordered Al-Si-C compound, as revealed by the appearance of a two domain ͱ7Â ͱ7R19 LEED pattern and simultaneous pronounced changes in the C 1s, Si 2p, and Al 2p core level spectra.In this temperature range, the core level spectra reveal that most of graphene is consumed upon formation of the Al-Si-C ordered compound.This is quite different when compared to the effects observed on Si-face graphene samples 14 in this temperature range and may be explained by the more reactive C-sites on smallerdomain size graphene on C-face compared to Si-face basalplane SiC.This Al-Si-C compound decomposes gradually during annealing at 800 C. No Al can be detected on the sample after annealing at 1000 C, when the original C-face graphene sample is essentially regenerated.Annealing at even higher temperature just creates more carbon on the surface, as revealed by the increase in the G/SiC peak intensity ratio in the C 1s spectrum.The electron p band structure recorded around the K-point does not show any dramatic changes after Al deposition and annealing.There is a pronounced reduction in the amount of graphene on the surface, as judged from a significant broadening of the p bands, only small shifts in the location of the Dirac point, and a considerably lower intensity of the core levels over the wide temperature range of 600-700 C.
2(d) at 900 C).Further increase of the annealing temperature to 1000 C makes the pattern sharper, see Fig. 2(e), but at 1100 C, see Fig. 2(e), a considerably more diffuse and weaker LEED pattern results.These changes in the LEED pattern after Al deposition and annealing indicate a sequence of reactions: chemical reaction, compound formation, and decomposition in certain temperature ranges.Detailed high-resolution photoemission studies of the C 1s, Si 2p, and Al 2p core levels were therefore performed in an effort to reveal changes in chemical composition after Al deposition and annealing.

FIG. 1 .
FIG. 1.(a) LEEM image acquired at 4.6 eV with a field of view (FOV) of 15 lm from the initial graphene sample grown on SiC(000 1).(b) Micro-LEED pattern recorded at E kin ¼ 45 eV and a probing diameter of 0.5 lm from the graphene sample, where the arrows indicate graphene (G) and substrate (SiC) spots.

FIG. 2 .
FIG. 2. LEED patterns recorded from (a) the initial graphene sample grown on SiC(000 1), and after Al deposition and annealing at (b) 500 C, (c) 600 C, (d) 900 C, (e) 1000 C, and (f) 1100 C. The E kin utilized is labeled in the upper-right corner of each panel.

FIG. 4 .
FIG. 4. (a) Si 2p spectra collected at 240 eV photon energy from the initial graphene sample, after Al deposition and after annealing at the temperatures specified and (b) Si 2p spectra acquired after annealing at 400 C using three different photon energies.

FIG. 6 .
FIG. 6.The p-bands recorded around the K-point from (a) the initial graphene sample grown on SiC(000 1), and after Al deposition and after annealing at (b) 700 C and (c) 1100 C.