Electronic field effect tuning of the electronic properties of fluorinated epitaxial graphene Free

Graphene is a single layer planar sheet of sp²-hybridized carbon atoms arranged in a hexagonal lattice and is one of the crystalline forms of carbon alongside diamond, carbon nanotubes, fullerenes, and graphite.^1,2 It is currently one of the great finds of the 20th century and may impact our quality of life through its various applications. It is the subject of much discussion due to its high tensile strength,³ high carrier mobility,⁴ and thermal conductivity,⁵ just to name a few of its unique properties. The ability to tune the band structure and control the properties of graphene is an integral part of ensuring its use in numerous applications.

Functionalization of graphene by chemical doping has risen as an appealing method in controlling graphene's physical, electronic, and chemical properties.⁶ The ability to tune the bandgap in graphene makes it an exciting material to implement in optoelectronic devices. Surface chemical functionalization is attractive because graphene consists of only surface atoms. Studies have shown that chemical functionalization can be used as a means to control graphene's properties, in particular, the bandgap. Conventional chemical functionalization using oxygen, hydrogen, and fluorine has been used to open a bandgap.⁷ Fluorinated graphene has emerged as a wide bandgap semiconductor as well as a high quality insulator.⁸ Fluorine is highly electronegative and fluorination improves the surface reactivity of graphene's relatively inert sp² bonding, thereby enabling a wide range of modifications.⁷ Through fluorination conditions, the C and F atoms yield a C–F bond that is ionic, semi-ionic, and covalent.^9,10 Preliminary studies have shown that SF₆ reactive ion etching plasma can fluorinate multilayer and single-layer epitaxial graphene films while the integrity of the sp²-hybridized structure remains intact. This chemical bonding could be used to adjust the properties of the fluorinated graphene.¹⁰ Altering the fluorine content can be controlled by exposing the fluorine for a longer or shorter period of time; this process is restricted to one or two surface layers.⁶ 

A fundamental understanding of the electronic properties of graphene and the interlayer interactions in multilayer graphene is critical to its incorporation into nanoelectronic devices. The planar structure of graphene makes it an ideal system for the utilization of traditional surface science techniques. In the current work, the evolution of electronic properties of fluorinated multilayer epitaxial graphene (MEG) is investigated as a function of an applied electrostatic field using ultraviolet photoemission spectroscopy (UPS).

MEG (∼20 layers) was grown by the de Heer group at the Georgia Institute of Technology by confinement controlled sublimation¹¹ at 1600°C on the C-face of semi-insulating 4H-SiC substrate (Cree, Inc.). Unlike graphite, which grows as AB stacked layers, MEG layers grown on the C-face contain a high density of rotational stacking faults. These faults cause the adjacent layers to decouple electronically, resulting in density of states (DOS) spectra of a single layer graphene sheet.¹¹ The MEG samples were fluorinated by exposing the samples to a SF₆ plasma generated in a Plasma-Therm reactive ion etcher for 30 and 60 s at a radio-frequency of 13.56 MHz, a power of 50 W and a SF₆ partial pressure of 100 mTorr at room temperature. Analysis of the resulting fluorinated MEG [fluorinated epitaxial graphene (FEG)] structures was performed with a Kratos Analytical Axis Ultra DLD x-ray photoemission spectroscopy system. Fluorine concentrations in these samples were determined to be 5.2 ± 0.1 at. % and 14.1 ± 0.15 at. %, respectively.

The FEG samples were mounted onto a molybdenum wafer block in nitrogen filled glove box at atmospheric pressure using indium (In) (99.9999% purity) at 160 °C. The In surrounded the outer edge of the FEG samples and extended to the top surface to eliminate charging effects during UPS. The wafer block was placed into a sealed container to minimize contamination and removed from the glove box for transport into the UPS load lock. The block was placed in the nitrogen purged load lock in less than one minute, sealed, the pressure reduced to 5.5 × 10⁻⁹Torr, and then transferred to the UPS chamber. The base pressure of the UPS chamber was less than 5 × 10⁻¹⁰Torr. The sample was outgassed at 160 °C under ultrahigh vacuum conditions using radiative heating from a resistive filament mounted behind the wafer block.

The angle integrated kinetic energy UPS of the samples were measured with a PHI 15–255 GAR double pass cylindrical mirror analyzer operated in the retarded mode with an instrumental resolution of ±0.105 eV. The kinetic energy distribution of the photoemission electrons provides a surface sensitive (4–5 Å) measurement of the joint density of states of the filled electronic states in the valence band of the material. The analysis was performed using the He I (21.2 eV) line from a VSW UV-10 discharge lamp. The UPS spectra were acquired at ground potential (zero bias) and at various negative sample bias potential differences with respect to the grounded face of the analyzer. Specifically, the bias was increased in even voltage steps up to −10 V then decreased in odd voltage steps for each successive spectrum. The semi-insulating substrate and In coated edges of the samples ensures that the field is established across the thickness of the samples' MEG layers.

The linear bands at the band edge of the DOS for single layer graphene correspond to transport by massless Dirac fermions. A comprehensive understanding of the interlayer coupling of MEG is thus of central importance to understand and tune the electronic properties of FEG. In previous work, MEG was modeled as a commensurate twisted bilayer with a rotational angle of 21.8° which has regions of AA and AB stacking.¹² With the application of an electric field applied normal to the surface, the model showed that the band structure of the individual layers hybridized from the single Dirac cone of AB stacking into two Dirac cones reflective of AA stacking.¹³ The experimental evidence of the decoupling was a transition of the electronic DOS from a linear dispersion at the band edge to a quadratic dispersion with increasing bias and its attendant Dirac cone separation.¹² 

To trace the effect of interlayer coupling, UPS spectra of FEG samples were collected under an applied electric field and the results compared with our prior MEG study.¹² The normalized spectra in Fig. 1(a) (30 s exposure), Fig. 1(b) (60 s exposure), and Fig. 1(c) (MEG) have been shifted in energy so that the photoemission thresholds of MEG and the 30 s (60 s) exposure coincide with that of the −1.0 V (−2.0 V) biased spectrum.

The photoemission thresholds were determined by linearly extrapolating the low kinetic energy edge of each electron distribution curve (EDC) from the full width at half maximum of the lower energy secondary electron peak to the spectral baseline. It is obvious that with the addition of a perpendicularly applied electric field, the corresponding DOS of the FEG exposed to SF₆ for 30 and 60 s changed tremendously. As with MEG, the first phenomenon noted is that the spectral shifts are nonrigid, i.e., the EDCs do not shift rigidly by the amount of the applied voltage as for a typical semiconductor. This indicates a redistribution of charge carriers (i.e., doping) as the Fermi level for the system was tuned with respect to the point of charge neutrality.¹³ Concurrently, the shape and widths of the EDCs are different from that of the unbiased samples. There is also an increase of the DOS starting at the valence band maximum (VBM). The peak associated with zone folding and the fusion of split bands at −10.5 eV in the MEG spectra¹² is not present in the 30 s FEG spectra and only becomes apparent in the 60 s FEG spectra after the −7.00 V bias.

The changes in the VBM of the UPS spectra between 0 and 3 eV below the Fermi level are shown in Fig. 2. The slope of the FEG spectrum increases between −1 and −2 eV for both exposures and the linear dispersion changes to quadratic with a bias greater than −3 V. This effect in the UPS investigation of MEG states concludes that the applied electric bias affected the interlayer coupling¹² and altered the DOS contribution of its stacking orders. The electronic structure variation of FEG can perhaps be attributed to the contribution of its stacking orders with applied bias, provided that it has mixed phase like MEG. No theoretical studies have been found to corroborate this as of yet.

Figures 3(a) and 3(b) show the deviation of the spectral shifts (Δshift) of the photoemission threshold with applied voltage for the FEG samples for the 30 and 60 s plasma exposures, respectively. Δshift of the EDCs is indicative of carrier distribution modification with bias.

There are three identified sources of charge injection in the FEG samples. One is due to the chemical bonding of the fluorine molecule¹⁴ to carbon in the graphene layers, the next due to the application of the electric field, and the last to spontaneous polarization of the substrate.^14–16 

Ideally, the adsorption of the electronegative fluorine creates an electrical double layer at the surface with a negatively charged outer surface. This surface dipole retards electron emission and increases the work function of the FEG relative to MEG.¹⁷ It should be noted that the strength of the dipole varies with the polarity of the F–C bond, which is controlled by the nature of F–C chemical bonding.⁶ The latter depends on experimental conditions such as fluorine concentration, bond orientation, defects, and surface topography.¹⁸ The negative bias applied to the samples results in a field with an orientation opposite to the F-C surface dipole of the FEG surface. The strong C–F bond initially distorts the MEG band such that the application of the bias compensates the opposite field of the dipole in Fig. 3(a) for the FEG 30 s exposure sample. With increasing bias, the dipole field appears to be balanced in the vicinity of −4.5 V. Δshift then increases with increasing bias. This effect is completely reversible with decreasing voltage (red dots in figure). At the minimum, Δshift reflects the rigid shift of a semiconductor. It would be interesting to know if the material transitions from a semimetal to a semiconductor at this point.¹⁹ The behavior of the 60 s FEG in Fig. 3(b) is significantly shifted with respect to that of the 30 s FEG. Its minimum does not appear to occur until a bias range of −9 to −10 V and even so does not reflect a rigid shift of the threshold with the bias.

The change in work function of FEG is extrapolated from the change in the EDC width. The results for each sample are shown in Fig. 4. The width is the result of subtracting the photoemission threshold from the VBM.¹⁷ The change in EDC width is inversely proportional to the change in the work function, e.g., an increase in the spectral width corresponds to a decrease in the work function of the material. The valence band width of the 30 s FEG sample shown in Fig. 4(a) hits a maximum at a bias of −3 V before decreasing. This transition coincides with the transition from linear to quadrature dispersion in Fig. 2(a).

The maximum width for the 60 s FEG sample shown in Fig. 4(b) occurs at a bias of −2 V.

The widths of both initially increase implying decreased work functions then there is an increase in work function with bias voltages beyond −3 V. By contrast, the work function of MEG increases almost linearly with bias greater that −1 V.

In the latter case of the semimetallic MEG, the application of the negative bias voltage to the sample establishes a normal electrostatic dipole field with the positive end oriented away from the sample. This field injects carriers into the conduction band with increasing bias thus decreasing the work function. The major trends noted in the FEG samples can be adequately explained with the compensation of the surface F-C dipole by increasing the strength of an opposite polarity electrostatic field applied normal to the substrate and the establishment of a bandgap in the material. The F-C dipole opposes electron emission and consequently yields a higher work function for FEG compared to MEG. This dipole is fully compensated with the initial application of the bias voltage field resulting in dramatic increases in spectral widths and corresponding reduction in the work functions. The work function of the unbiased FEG samples are 0.64 and 0.71 eV higher than MEG for the 30 s FEG and 60 s FEG samples, respectively. In contrast to MEG, the formation of a bandgap in the biased FEG (Ref. 19) samples precludes the injection of carriers into the conduction band. The increases in bandgap with the applied electrostatic field strength result in an increasing work function shown by the decreasing spectral widths in Figs. 4(a) and 4(b).

Previous investigations conclude that FEG has ferromagnetic properties. Hysteresis occurs in some ferromagnetic materials and ferroelectric materials, as well in the deformation of materials in response to varying force. The presence of hysteresis effects is examined by monitoring the change in work function as a function of the direction of the applied electric field in Figs. 4(a) and 4(b). The electric field was applied with an increase in even values of voltage from 0 to −10 V, and then reversed from −9 to −1 V in odd value steps to identify any hysteresis effects. Electron spins could be aligned with the E-field of the F-C dipole then flipped as the applied electrostatic field increases and compensates the dipole field. The application of the electric field floods the valence band with electrons as shown in Fig. 1, the spin of those electrons would be aligned with the direction of the electric field. Maximum saturation occurs when all electrons are aligned with the electric field. With an increase in bias from 0 to −10 V, an increase in valence band width occurred, which reached a maximum at −3 V for the 30 s exposure and −2 V for the 60 s exposure. Hysteresis is observed in the widths of the FEG samples particularly the 30 s exposure sample in Fig. 4(a) in the range of −3 to −10 V and also in the corresponding photoemission intensities in Fig. 5. Hysteresis is present as the bias is increased evenly and decreased oddly with a marginal offset in the width value. The hysteresis in the raw intensities is actually the opposite of what was expected. The biasing sequence leads to an increased width (lower work function) for the ramp up in bias voltage [see Fig. 4(a)] and a decreased width for the ramp down. The expectation was that the lower work function of the upper loop would yield a higher UPS intensity but the corresponding photoemission intensity in Fig. 5(a) was suppressed and resulted in a loop inversion.

The inversion of the photoyield is not understood and will require further study. The spontaneous polarization and depolarization of the semi-insulating substrate as the sample bias is increased and decreased, respectively, also presents itself as a plausible explanation for the hysteresis and cannot be ruled out at this juncture.

The ability to tune the band structure and control the properties of graphene is an integral part into fulfilling its use for application purposes. Functionalization of graphene as a result of chemically doping has risen as an appealing method in controlling graphene's physical, electronic, and chemical properties. Surface chemical functionalization is attractive because graphene consists of only surface atoms. Studies have shown that chemical functionalization can be used as a means to controlling graphene's properties.

Fluorinated graphene has emerged as a stand out star as a possible wide band gap semiconductor as well as a high quality insulator.^7,11,12 The addition of fluorine allows the ability to engineer the electronic properties of graphene. Fluorination improves graphene's surface reactivity of its relatively inert sp² bonded graphene. A controllable work function is an important step in integrating graphene into electronic devices. In this work, MEG fluorinated by reactive ion etching and subjected to an electric field applied normal to the surface is studied by UPS.

We demonstrate a noteworthy control of the low-energy electronic states produced by tuning the interlayer interactions in the FEG samples. This condition is similar to the effects of an applied bias on MEG samples. The control of the electronic properties of FEG as a function of electric bias is a necessary application if to be used in future electronic devices which may be under an electric field or strain. Interlayer coupling may also play a role in different graphene-based systems such as heterobilayers.

In summary, the work establishes multiple claims. Plasma technology is a practical and facile method for the functionalization of graphene without destroying the integrity of its lattice and surface structure. Work function changes in fluorinated graphene occur using plasma fluorination and a work function increase is dependent on the polarity of the C–F bonds. The band structure and work function of FEG change with an addition of an electric field. Spin saturation and dipole strength are factors in creating a continuous increase in work function.

The authors thank Ronald E. Mickens and X. Q. Wang for their critical reading of this manuscript. This work was supported by the National Science Foundation PREM under Grant No. DMR-0934142 and CREST under Grant No. HRD-1137751.