Enhancing Structural Properties and Performance of Graphene-Based Devices Using Self-Assembled HMDS Monolayers.

The performance of graphene devices is often limited by defects and impurities induced during device fabrication. Polymer residue left on the surface of graphene after photoresist processing can increase electron scattering and hinder electron transport. Furthermore, exposing graphene to plasma-based processing such as sputtering of metallization layers can increase the defect density in graphene and alter the device performance. Therefore, the preservation of the high-quality surface of graphene during thin-film deposition and device manufacturing is essential for many electronic applications. Here, we show that the use of self-assembled monolayers (SAMs) of hexamethyldisilazane (HMDS) as a buffer layer during the device fabrication of graphene can significantly reduce damage, improve the quality of graphene, and enhance device performance. The role of HMDS has been systematically investigated using surface analysis techniques and electrical measurements. The benefits of HMDS treatment include a significant reduction in defect density compared with as-treated graphene and more than a 2-fold reduction of contact resistance. This surface treatment is simple and offers a practical route for improving graphene device interfaces, which is important for the integration of graphene into functional devices such as electronics and sensor devices.

Introduction

Graphene has demonstrated excellent electrical, mechanical, and optical properties that make it ideal as a building block for numerous device applications.[1,2] Near ballistic transport and high carrier mobility make it a suitable candidate for the next generation of nanoelectronics, especially high-frequency devices.[3,4] Furthermore, graphene has been proposed in many other device applications such as terahertz,[5−7] nano-electromechanical systems,[8,9] solar cells,[10] gas sensors,[11−13] and biosensors.[14−16] However, the device performance is often limited by structural defects and damage introduced during different stages of fabrication and processing. This results in increased device scattering and contact resistance, affecting device performance. Modern device fabrication involves photolithography, etching, and thin-film deposition. However, implementing these steps on two-dimensional (2D) materials has been associated with many issues, such as the tearing of graphene, introduction of unintentional doping, increased damage to devices, and higher defect density in graphene.[17] For instance, the use of plasma and ion bombardment during sputtering has been frequently reported to increase damage to graphene,[18−20] degrade device properties, and increase contact resistance.[21] Moreover, polymer residues left on the surface of graphene after photolithography are difficult to be removed completely and can degrade device performance.[22,23] Several approaches have been implemented to reduce the defect density and degradation of devices during lithography or sputtering for the improvement of device performance. Some approaches include annealing[24] or the insertion of sacrificial layers of AlOx[25,26] and Au,[27] aiming to passivate the device during the fabrication process.

Self-assembled monolayers (SAMs) have been widely used to tune the electronic properties of 2D materials and to improve device performance.[28,29] For instance, inserting hydrophobic SAMs as a buffer layer between the graphene and SiO2/Si substrate has been reported to control the electrical properties of graphene-field-effect transistors (GFETs) and to reduce hysteresis behavior in GFET characteristics.[30,31] Furthermore, the application of perfluorodecyltrichlorosilane (FDTS)[32] or hexamethyldisilazane (HMDS)[33] prior to chemical vapor deposition (CVD) and atomic layer deposition (ALD) processes has been demonstrated to enhance the coating of different dielectric films on graphene. Finally, it has been demonstrated that the mechanical properties of graphene can be modified by the hydrophobicity of the surface introduced by SAMs.[34]

Although some work has been performed to improve the GFETs using HMDS,[35,36] its role on the surface of graphene has not yet been fully investigated. This paper describes a systematic study of graphene modified with HMDS using surface analysis techniques, showing how its electronic properties are altered after HMDS treatment. We also show how treatment conditions can affect the properties of graphene.

This paper demonstrates the importance of incorporating self-assembled monolayers of HMDS at the graphene interface during device manufacturing and their role in improving the structural properties of graphene and device performance. We show how this can affect each step during device fabrication and the property of a device. By applying HMDS to graphene prior to the sputtering process, we observed a significant reduction in defect density of graphene after film deposition and a reduction of contact resistance. We also found an improvement in the surface quality of graphene when HMDS was applied prior to photoresist coating. This approach is simple and can be incorporated during the manufacturing process to reduce damage to graphene and to enhance the performance of graphene-based devices.

Results and Discussion

Effect of HMDS on Contact Resistance and GFET Device Performance

The circular transmission line method (CTLM) is used to extract contact characteristics and values of the sheet resistance of graphene.[37,38] The method is chosen because of its simplicity and the fact that it does not require the fabrication of mesa structures. Therefore, there is no need for any further lithography step, thus avoiding problems of doping and the effect of polymer residues on the surface associated with photolithography. Measurements have been obtained for more than 12 devices for each type of treatment. Figure S1 shows a representative plot of I–V at different spacings for both as-transferred and HMDS-treated samples. The HMDS treatment was performed for 15 h in the exposed graphene region after patterning and before the deposition of the contacts. The total resistance RT(W) as a function of spacing distance L is plotted in Figure a for devices with and without HMDS. The contact resistance is obtained from the y-axis intercept of the CTLM. A more than 2-fold reduction in contact resistance is observed after HMDS treatment, as can be seen from the histogram in Figure b.

Figure 1

(a) Total contact resistance for CTLM devices with and without HMDS treatment (the inset is a magnification of the interception with y-axis). (b) Contact resistance for devices with and without HMDS treatment (error bars represent the data range within 1.5* interquartile range).

This improvement in contact resistance after HMDS treatment gives an indication that HMDS acts as a protective layer and reduces the damage caused during sputtering and the contamination introduced during device processing. Another possible reason for the improvement in specific contact resistivity is an increase in work function difference between graphene and metal contacts with HMDS treatment compared with as-transferred graphene. This may lead to an increase of the charge transfer from the graphene to metals due to an increased density of the states of graphene underneath the contact area.[39]

Further improvement in contact resistance is achieved after postprocessing annealing at 210 °C in Ar ambient for 3 h. The contact resistance of HMDS-treated samples was reduced from (1200 ± 90) to (810 ± 79) Ω·μm. This value is much lower than the reported values of contact resistance for metal contacts deposited on graphene by sputttering.[40−43]

An improvement in GFETs characteristic after HMDS treatment has been reported recently.[35,36] HMDS can remove oxygen and water molecules absorbed on graphene surface and some polymer residues that act as charged impurity scattering center and degrading device performance.[35,36] Here, we repeatedly observed a Dirac point near zero after HMDS treatment (Figure b,c), compared with around 15 V for bare graphene. This shift in Dirac voltage can be used to calculate the shift in Fermi energy in graphene induced after HMDS. The Fermi energy shift is calculated according to the approximation assuming a linear band structurewhere vF is the Fermi velocity in monolayer graphene of 1.1 × 106 m/s, n is the intrinsic carrier concentration, which can be written as n = η(VDHMDS – VDBARE)[44] where η ≈ 7.21010 cm–2·V–1 is the intrinsic carrier concentration per volt,[44,45] and VDHMDS and VDBARE are the Dirac voltages with and without HMDS treatment, respectively. The observed shift in Fermi level after HMDS is around 0.14 eV. More than 30% increase in hole mobility and around 100% increase in electron mobility were observed after HMDS treatment (Figure S2). The mobility is calculated from the transfer curve in the linear regime using the following equationwhere in the linear region of the measure IV characteristic, Cox oxide capacitance, VSD is the source–drain voltage, and L and W are the length and the width of graphene structure, respectively.

Figure 2

(a) Optical image of the GFET structure being used. (b) Transfer characteristics of GFETs with different treatment times of graphene with HMDS. (c) Histogram shows the Dirac point of GFETs with and without HMDS treatment. The scale bar in (a) is 50 μm.

Effect of Sputtering on As-Transferred and HMDS-Treated Graphene

Figure a shows a representative Raman spectrum of as-transferred graphene after 3 nm titanium sputtered on bare graphene and on HMDS-treated graphene. The D-peak located at 1346 cm–1 is observed after titanium deposition and film deposition. The D-peak indicates a structural disorder induced in sp2 graphene due to defects. However, a significant reduction in D-peak intensity is visible for the sample treated with HMDS. The ID/IG extracted from the Raman map is plotted for both cases in Figure b. The values for position (G), full width at half-maximum (FWHM)(G), and ID/IG for the as-transferred sample after Ti deposition are 1589 cm–1, 24.4 ± 2.5 cm–1, and 0.75, respectively. The value of ID/IG is reduced after HMDS treatment to 0.2 and FWHM(G) is 18.1 ± 2.2 cm–1, while its position (G) is 1587 cm–1. The increase in ID/IG and FWHM(G) for untreated samples is a sign of an increase in disorder of graphene for the as-transferred sputtered sample compared with the HMDS-treated one. The histogram of FWHM(G) is shown in Figure c. We also determined if there is any stain or doping induced in the film after metal deposition by calculating the ratio

Figure 3

(a) Raman spectra of graphene with and without HMDS after DC sputtering deposition of Ti compared with as-transferred graphene. (b) ID/IG of graphene after Ti deposition for unprotected and protected samples with HMDS. (c) FWHM of (G) peak of graphene with and without HMDS after DC sputtering deposition of Ti compared with as-transferred graphene.

The values of Raman frequencies of 2D (ω2D) and G (ωG) modes are sensitive to charge impurities and strain in graphene.[46] We found that the average value of for as-transferred devices is 0.78 ± 0.2, which corresponds to hole doping, while the average values of after Ti deposition for as-transferred and HMDS-treated samples are 0.85 ± 0.3 and 0.47 ± 0.2, respectively. Here, the intrinsic frequencies (ωG0 and ω2D0) are considered the values of freestanding graphene 1581.6 and 2676.9 cm–1, respectively.[46] Since the value of , this indicates that no strain has been induced after Ti deposition. However, these values show that graphene in the HMDS-treated samples is n-doped, while in the as-transferred samples it is p-doped.

We further calculated the defect concentration induced during the sputtering process for both cases. Considering samples with low degree of disorder (LD > 10 nm), the defect concentration is given by[47,48]where EL is the excitation energy of the Raman laser (EL = 2.33 eV) and LD is the distance between two defects in nm. The defect density in graphene induced by sputtering is determined to be nD_HMDS = 1.1 × 1011 cm–2 for the HMDS-treated sample and nD_as_tr = 3.9 × 1011 cm–2 for the as-transferred sample without treatment. These values correspond to LD_HMDS = 30 nm and LD_as_tr = 16 nm. The values of nD and LD observed after HMDS treatment are typical values reported in CVD pristine graphene.[49,50] The reduction of disorder in graphene after HMDS treatment indicates that HMDS acts as a protective layer for graphene that reduces damage during the fabrication process. Reductions of damage are observed for both rf and dc sputtering and different films, as shown in Figure S3 (see the Supporting Information). This approach can be incorporated during sputtering thin-film deposition of metals and dielectric materials to graphene FETs while preserving the high quality of underlying graphene. The HMDS can be easily removed from the graphene interface by a simple annealing process of around 200 °C, and the quality of graphene is further improved. This can be observed by a significant reduction in the defect in graphene after annealing (Figure S4).

Effects of Photolithography on As-Transferred and HMDS-Treated Graphene

As-transferred graphene and HMDS-treated graphene substrates were patterned using AZ5214E photoresist before the photoresist was removed with acetone. AZ5214E was chosen as it is widely employed in microfabrication and can be used as a positive and image reversal photoresist. Figure shows the Raman map of as-transferred graphene, as-transferred graphene after photolithography, and graphene treated with HMDS after the lithography process. The Raman scan area is 20 μm × 20 μm. Both the first-order G peak and the second-order 2D peak are present in the Raman spectra; the D-peak is absent for as-transferred samples. This indicates that graphene has a very low defect density. The G and 2D peaks of the as-transferred samples are centered at 1587 and 2683 cm–1, respectively, with an FWHM value of the G peak of 18 cm–1. This indicates that as-transferred graphene exhibits hole doping, which could be due to the presence of water absorbents or substrate impurities. The hole doping is reduced after HMDS, which can be observed as downshifts of 2 and 3 cm–1 for the 2D and G peaks in HMDS samples. The influence of doping can be observed from the ratio between 2D/G peaks. The average I2D/IG peaks ratio for nontreated and HMDS-treated samples are 2.18 and 2.42, respectively. The increase in I2D/IG ratio after HMDS treatment indicates that doping is reduced with HMDS treatment. Photolithography does not introduce doping in graphene after HMDS, which may be because the HMDS acts as a protective layer preventing any direct contact between graphene and photoresist. Another feature that can be observed from the Raman spectrum is the presence of the D-peak at ∼1340 cm–1, illustrating defects introduced into graphene during the photolithography process. An increase in ID/IG peak intensity of as-transferred graphene after photoresist was observed without HMDS treatment, the HMDS-treated graphene shows a negligible D-peak, indicating that the HMDS protects the graphene layer and prevents direct contact between graphene and photoresist. HMDS is compatible with photoresist, and therefore leaving the sample in HMDS with photoresist does not degrade or damage the photoresist.

Figure 4

(a) G peak position of as-transferred and HMDS-treated graphene after photoresist removal. (b) Two-dimensional position of as-transferred and HMDS-treated graphene after photoresist removal. (c) Distribution of D, G, and 2D peak intensities of the as-transferred graphene sample after photoresist. (d) Distribution of D, G, and 2D peak intensities of the HMDS-treated graphene sample after photoresist.

Effect of HMDS Treatment Time on the Surface of Graphene

The presence of HMDS on the graphene surface was first characterized using atomic force microscopy (AFM). The surface roughness of graphene increased from 0.3 nm without HMDS to 0.5 nm after HMDS treatment (Figure a,b). The thickness of HMDS on the graphene surface was approximately determined using line profile to be around 1.3–1.8 nm, as shown in Figure c,d. We further performed X-ray photoelectron spectroscopy (XPS) and water contact angle (WCA) measurements to characterize the surface of graphene after HMDS treatment to investigate the effect of HMDS treatment on the electronic properties of graphene and to determine the optimum treatment time. XPS was used first to analyze surface properties and the presence of HMDS on the surface of Gr/SiO2/Si substrates. The N 1s peak at a binding energy of ∼400 eV is significantly enhanced after the HMDS treatment of graphene compared to as-transferred samples (Figure a). The presence of nitrogen in the as-transferred samples could be introduced during the graphene transfer process. This could be due to the reaction of ammonium persulfate (APS) or RCA1 cleaning in the form of ammonium hydroxide with defects and edges in graphene.[51,52] XPS spectra of the two samples are shown in Figures S5 and S6. To reduce the effects of the substrate and the transfer process, we have performed the XPS analysis by coating HMDS on Gr/Cu and by conducting a comparison with as-grown graphene on copper.

Figure 5

AFM characteristics of bare Gr and Gr modification with HMDS. (a) Bare Gr. (b) and (c) Gr after HMDS treatment. (d) Line profile shows the height of the HMDS layer on Gr. Scale bars = 1 μm.

Figure 6

(a) N 1s XPS spectrum of as-transferred and HMDS-treated graphene on the SiO2/Si substrate. (b) N 1s XPS spectrum of as-grown and HMDS-treated graphene on Cu. (c) Evolution of the C 1s peak as a function of the HMDS treatment of Gr on Cu. (d) XPS spectra of Gr, Gr-HMDS, and Au at the secondary electron energy region with substrate bias at −30 V.

The N 1s spectrum of as-grown Gr on Cu shows no presence of nitrogen species on graphene, as shown in Figure b. On the other hand, HMDS-treated graphene samples show an N 1s peak at ∼399.9 eV. The longer the samples were immersed in HMDS, the stronger the N 1s signal is with no obvious improvement after 12 h treatment (Further XPS analysis is shown in Figures S7 and S8). Furthermore, we compared the peak position of C 1s in bare Gr/Cu with HMDS-Gr/Cu with different HMDS treatment durations. It is interesting to note that the position of the peak is blue-shifted toward higher binding energy with increasing exposure to HMDS, as shown in Figure c. The maximum shift observed is ∼0.15 eV after 12 h HMDS treatment (The deconvolution of C 1s peak and the position of the C–C sp2 peak position as a function of HMDS treatment time are shown in Figure S9), which suggests a change in Fermi level in graphene as a reduction of p-doping due to removal of water after HMDS treatment.[29,44,53]

We measured the change in the work function of as-transferred graphene with that of HMDS-treated graphene to determine the effect of HMDS treatment on its electronic properties. The change in work function is measured using XPS with Al irradiation source of the photon energy of hν = 1486.6 eV. The sample was biased with a negative potential of −30 V applied to the substrate. The work function is then calculated using the following equation:where ECut-off is the secondary electron energy cutoff. Figure d shows the XPS spectra for as-transferred graphene, HMDS-treated graphene, and Au. The work function of graphene is reduced by ∼0.20 eV after coating with HMDS. This reduction in work function of graphene after HMDS treatment is in close agreement with the value of Fermi energy shift of 0.14 eV as observed in the transfer characteristics of the GFETs after HMDS treatment extracted from eq . The small difference between the two values could be attributed to the effect of contacts to graphene that cause doping in graphene underneath the contact region. The origin of this decrease in work function could be partly attributed to the removal of water and oxygen absorbents from its surface after HMDS treatment. The untreated graphene exhibits p-doping characteristics due to water molecules at its surface due to its electron-withdrawing character. The dipole moment in water can induce a shift in Fermi level in graphene.[54] We calculated the change in work function due to the contribution of water dipole moment by[44,55]where Γ ≈ 5 × 1013 cm–2 is the density of water molecules accumulate on the graphene surface; ε0 is the permittivity of the free space; ε ≈ 2 is the relative dielectric constant of a few layers of water;[56] μ = 1.85 D is the dipole moment of water molecule normal to the surface;[57] and θ is the angle of the dipole from the surface normal. The approximate change in work function induced by water dipole moment is estimated to be 0.12 eV.

The dipole moment in HMDS is also expected to affect the graphene electronic properties by screening the impurities directly underneath graphene. However, such role is less significant because HMDS has weaker dipole moment compared with water.[58] Furthermore, our results show that HMDS reduces the traps/defects induced during the processing of graphene. Our results are also supported by the observation of negligible hysteresis in graphene transistors after HMDS treatment as an indication of the role of HMDS in removing water molecules and screening out the charge impurities in graphene[58,59] (Figure S10). It was noticeable that the interaction between graphene and water changes after the application of HMDS due to the hydrophobic properties of HMDS layer, and the surface of the graphene becomes more hydrophobic with HMDS modifications. Figure S11 shows the evolution of the water contact angle (WCA) of graphene on the SiO2/Si substrate after different durations of HMDS treatment. The WCA was recorded over 10 s with 330 frames and a droplet size of ∼30 μL. The WCA of graphene changes by about 12° after 12 h treatment with HMDS.

In this study, a solution-processed HMDS was used to modify graphene with SAMs. However, for long-term stability and shorter processing time, vapor-processed HMDS could be more beneficial.

Conclusions

In summary, we have demonstrated that the passivation of the graphene surface with the self-assembled monolayers of HMDS has significantly improved the quality of graphene after thin-film deposition and device manufacturing. Through the application of HMDS to the graphene surface prior to lithography and contact deposition, a direct contact between the graphene and photoresist is prevented. The HMDS functions as a protective layer for the graphene during plasma processing. As a result, the concentration of the D-peak is reduced more than 2-fold after HMDS treatment and the contact resistance of graphene has been improved more than doubled compared with that of as-transferred graphene. Therefore, the HMDS protection of graphene is expected to enable the deposition of gate dielectric and metal electrodes on graphene with preserving the high quality of graphene. This approach is compatible with microfabrication processing and offers a simple solution for the fabrication of high-quality graphene-field-effect devices.

Experimental Section

Graphene preparation: Monolayer graphene grown by CVD on Cu foil was coated with 300 nm poly(methyl methcrylate) (PMMA) to support the graphene transfer. The PMMA was cured in ambient conditions for 12 h. Graphene at the backside of Cu foil was etched using oxygen plasma for 1 min at power 75 W. The Cu was then etched using ammonium persulfate (APS) (0.01 g/mL in H2O) overnight. The PMMA/graphene stack is then floated on two consecutive ultra DI water baths to rinse the graphene surface for up to 1 h per bath. The PMMA/graphene was then transferred onto modified RCA1 solution (20:1:1 H2O/H2O2/NH4OH) for 1 h at room temperature to remove Cu residues followed by two steps of ultra DI water bath for 1 h each. The freestanding PMMA/graphene layer was transferred to a SiO2/Si substrate. The PMMA/graphene on the substrate was baked on a hotplate for 2 h at 150 °C to improve the adhesion of graphene to the substrate. PMMA was then removed using acetone for 5 h followed by isopropanol (IPA) for 5 min. The sample was then annealed at 200 °C for 6 h to remove PMMA residues on the graphene surface.

Raman spectroscopy measurements were performed using a confocal Witec spectrometer and exited with a laser wavelength of 532 nm (excitation energy EL = ℏw = 2.33 eV) through an optical fiber, and an objective lens of 100X, NA = 0.8, and laser spot of 0.4 μm. The laser power was kept below 2 mW and a spectral resolution was ∼3 cm–1; the Raman peak position was calibrated based on the Si peak position at 520.7 cm–1. The D, G, and 2D peaks were fitted with Lorenzian functions.

X-ray photoelectron spectroscopy (XPS) experiments and measurements were performed with Kα+ and an Al radiation source (hν = 1486.6 eV) in an ultrahigh vacuum chamber for spectroscopic analysis with a base pressure of 5 × 10–8 mbar.

Fabrication and electrical measurements of graphene contact structures: For metal contact structures, graphene on Si/SiO2 substrates was patterned using optical lithography, then sets of samples were immersed in HMDS for 12 h in the exposed contact region and the other set kept in a desiccator. Typically, 5/50 nm Ti/Au was deposited on the sample. Finally, the lift-off process was performed using acetone to remove the photoresist. Two-probe measurements were used to measure the contact characteristics of the CTLM structures. The measurements were performed at room temperature using B1500 Semiconductor Analyzer.

For GFETs, graphene on Si/SiO2 substrates was patterned using optical lithography and oxygen plasma to define the device structure. Another lithography step followed by the sputtering process and lift-off were used to form 5/50 nm Ti/Au. Three-probe measurements were used to measure the transfer characteristics of GFETs with back-gate.

Contact angle measurements: a Theta optical tensiometer was used to measure the static contact angle. A droplet of water 30 μL in size was deposited onto the surface of the graphene and the measurements were taken within a 10 s frame. The contact angle was then determined by fitting the Young–Laplace equation around the droplet. Graphene samples clean and free of defects were selected for contact angle measurements to reduce the error that could be caused by a defective surface.

DC and rf sputtering experiments: two sets of Gr/Si/SiO2 substrates were prepared. One was immersed in fresh HMDS solution overnight and the other was kept in ambient conditions. The substrates were placed in a sputter chamber at ultrahigh vacuum conditions of 1 × 10–7 mbar. The sputtering process was performed at pressure levels between 6.0 and 8.5 × 10–3 mbar in the presence of Ar. In the case of DC sputtering, the power was set to 60 W and the deposition rate was 0.2 Å/s, while power levels of 90 and 100 W were used for rf sputtering with deposition rates of 0.07 and 0.08 Å/s.

Photoresist experiments: two sets of Gr/Si/SiO2 substrates were prepared. One was immersed in fresh HMDS solution for 6 h and the other was kept in ambient conditions. AZ5214E was spun-coated on the substrates, and the samples were soft baked on a hotplate at 90 °C for 3 min. Both sets of samples were exposed to UV light with an intensity of around 80 mJ/cm2 before they developed using an MF319 developer.