lable at ScienceDirect

Carbon 106 (2016) 330e335

Contents lists avai

Carbon

journal homepage: www.elsevier .com/locate /carbon

Anomalous photoresponse in the deep-ultraviolet due to resonantexcitonic effects in oxygen plasma treated few-layer graphene

Ram Sevak Singh a, b, **, Dehui Li f, Qihua Xiong f, g, Iman Santoso a, b, c, Xiaojiang Yu c,Wei Chen a, b, e, Andrivo Rusydi a, c, d, *, Andrew Thye Shen Wee a, b, ***

a Department of Physics, National University of Singapore, 117542, Singaporeb Centre for Advanced 2D Materials and Graphene Research Centre, National University of Singapore, 117546, Singaporec Singapore Synchrotron Light Source, National University of Singapore, 5 Research Link, 117603, Singapored NUSNNI-NanoCore, National University of Singapore, 117576, Singaporee Department of Chemistry, National University of Singapore, 1175427, Singaporef Division of Physics and Applied Physics, School of Physics and Mathematical Sciences, Nanyang Technological University, 637371, Singaporeg Division of Microelectronics, School of Electrical and Electronic Engineering, Nanyang Technological University, 639798, Singapore

a r t i c l e i n f o

Article history:Received 3 January 2016Received in revised form20 April 2016Accepted 8 May 2016Available online 13 May 2016

* Corresponding author. Department of Physics, Nat117542, Singapore.** Corresponding author. Present address: Nationalukshetra, 136199, Haryana, India.*** Corresponding author. Department of PhysSingapore, 117542, Singapore.

E-mail addresses: singh915@gmail.com (R.S. S(A. Rusydi), phyweets@nus.edu.sg (A.T.S. Wee).

http://dx.doi.org/10.1016/j.carbon.2016.05.0260008-6223/© 2016 Elsevier Ltd. All rights reserved.

a b s t r a c t

Graphene holds great promise for novel optoelectronic devices due its exceptional optical properties.Here we report an anomalous photoresponse due to resonant excitonic effects in oxygen functionalizedepitaxial graphene (EG) with gain exceeding 104 in the deep-ultraviolet (DUV) region. The method usedto introduce the oxygen adatoms is via mild oxygen plasma treatment. Upon oxygen adsorption, spectralweight transfer occurs and a photoresponse emerges. The photoresponse in the oxygen plasma treatedEG (OPTEG) is attributed to the enhancement of the resonant exciton in the DUV and suppression ofoptical conductivity below it, as evidenced in our spectroscopic ellipsometry measurements. Our resultshows that the OPTEG produces large photoconductive gain, high selectivity, and detectivity in excess of1012 Jones in the DUV spectral range and can be harnessed to fabricate tunable graphene-based high-gainDUV photodetectors.

© 2016 Elsevier Ltd. All rights reserved.

1. Introduction

Graphene has been an attractive research field for fundamentalscience and applications, such as novel photonic and optoelectronicdevices, due to its fascinating electronic and optical properties[1e4]. In particular, epitaxial graphene (EG) prepared on siliconcarbide (SiC) is a promising substrate material for the developmentof carbon-based electronics [5,6] and optoelectronics [7] due to itscompatibility with large-scale semiconductor processing tech-niques. High-speed photodetectors based on EG materials havebeen reported previously [7,8]. Infrared (IR) and ultraviolet (UV)

ional University of Singapore,

Institute of Technology Kur-

ics, National University of

ingh), phyandri@nus.edu.sg

photoconductors based on thin films of reduced graphene oxide(RGO) or graphene nanoribbons have also been demonstrated withdevice photoresponsivity up to 4 mA W�1for IR light and up to0.12 AW�1 for UV light [9]. RGO is a promising candidate in organicor plastic electronics [10,11]. However, compared to pristine gra-phene, the electrical property of RGO is poor, thus limiting its ul-timate utility in high-performance electronics [12,13]. It has beenobserved that the attachment of quantum dots on the graphenesurface enhances the photoconductivity in the visible range[14e16]. An electro-oxidized multilayer EG based device has shownphotoresponsivity up to 200 AW�1 in the UV region [17]. However,the acid solution method used could be detrimental to the deviceperformance and the photoresponsivity mechanism is not known.Therefore, an alternative simple, dry method of oxidation to func-tionalize graphene in devices is needed.

For fundamental science, graphene has been used as a modelsystem to study electronic correlations, i.e. electron-electron (e-e)and electron-hole (e-h) interactions [18]. It has been shown bothexperimentally [19e25] and theoretically [25e28] that the opticalproperties of graphene are determined by strong electron-electron

R.S. Singh et al. / Carbon 106 (2016) 330e335 331

(e-e) and electron-hole (e-h) interactions yielding resonant exci-tonic effects in the deep ultraviolet (DUV). Indeed, resonant exci-tons have been observed in the DUV range and depend on thenumber of layers [19e21], substrate [21e23,25] or doping withadatoms [24]. However, till now there is no mechanistic study ofthe photoresponse of functionalized graphene in devices, particu-larly in the DUV range.

Here, we discover an anomalous photoresponse in the DUV offew layer EG (~4-layer) based devices by modifying its opticalconductivity using mild dry oxygen plasma treatment. Thismethod, namely oxygen plasma treated EG (OPTEG), introduces alow coverage of oxygen adatoms on EG, which then generatesspectral weight transfer in a broad energy range. Such a spectralweight transfer stronglymodifies the resonant excitonic effects andactivates an anomalous photoconductive response in the DUVrange. This new physical phenomenon can be harnessed to producehigh-gain, and large area processable graphene based DUV photo-conductive devices. The controlled low power plasma used in thisstudy does not produce substantial lattice damage in EG. The high-gain, large energy selectivity, and high detectivity (a measure ofsignal to noise ratio) in the DUV spectral range highlights its po-tential advantages over Si or other semiconductor DUVphotodetectors.

2. Results and discussion

The OPTEG samples were investigated using Raman spectros-copy, synchrotron-based high resolution photoelectron spectros-copy (PES), electrical I-V and spectroscopy ellipsometrymeasurements. The photoconductive characteristics of OPTEG de-vices were studied under illumination of white light, and subse-quently the spectral photoresponses were investigated.

In Fig. 1, we show the characterization of pristine EG and OPTEGsamples using atomic force microscopy (AFM), synchrotron-basedphotoelectron spectroscopy (PES), and Raman measurements.AFM surface morphology imaging shows similarity between EGand OPTEG samples with no significant peeling or damage ofOPTEG sample as compared to pristine EG (Fig. 1a-b). Synchrotron-based high-resolution (energy hn ¼ 650 eV) light was used for corelevel PES measurements of EG and OPTEG samples for wide-scan(Fig. 1c) and C 1s spectra (Fig. 1d-e). We observe the emergenceof the O 1s peak (at 532.18 eV) in OPTEG and it indicates the pop-ulation of oxygen species on the surface. Furthermore, detailedanalysis of the C 1s peak (at ~ 284.6 eV) shows chemisorbed CeO (at~286.0 eV) and C]O (at~ 287.3 eV) species [29,30] on the OPTEGsurface. In particular for our OPTEG devices, as representatives forthis study, the oxygen plasma (rf power: 6 W) treatment was per-formed cumulatively for the duration of 20 s (OPTEG1) and 40 s(OPTEG2), respectively. Raman measurements on pristine EG,OPTEG1 and OPTEG2 devices were performed on a Raman micro-scope (Renishaw) with excitation wavelength of 515 nm at roomtemperature and ambient environment as shown in Fig. 1f. Themain features in the Raman spectra are the G-band(at ~ 1581 cm�1), 2D-band (2710 cm�1). Compared to the pristineEG, a slight enhancement in D-band is observed after oxygenplasma exposure. This D-band enhancement is caused by increasedconcentration of sp3 carbon centres in the conjugated sp2 network,attributed to disruption of the graphene lattice including the for-mation of chemical CeO and C]O bonds as observed in the syn-chrotron PES spectra. Based on our characterizations (XPS, AFM andRaman measurements), we have estimated of about ~5% and ~7% Ocoverage on the surface for OPTEG1 and OPTEG2, respectively.

Our main result of an anomalous photoresponse in the DUV isshown in Fig. 2, which shows the schematic of a photodetectordevice fabricated in our study. Interestingly, the photoresponse is

observed in OPTEG1 and OPTEG2 samples and the intensity in-creases significantly at higher energies, in the DUV range (Fig. 2b).Note that the photon energy cut-off at 4.133 eV is the limit of ourinstrumental setup. Nevertheless it is clear that the photoresponsekeeps increasing beyond this energy range. This phenomenon isabsent in pristine EG as shown in inset of Fig. 2b. All the mea-surements were performed at room-temperature and in ambientenvironment at a fixed 2 V bias.

The photoconductive gain of a photodetector device can beestimated using the following equations:

Gain ¼ (Iph/P)(hn/q) ¼ [1240/l (nm)](Iph/P) ¼ [1240/l (nm)](PR)(1)

where, Iph is the photocurrent, P is the effective light power inci-dent on the active area A (here 170 mm � 300 mm) of the photo-detector device, hn is the energy of the incident photon, q is theelectron charge, l is wavelength of light, PR is the photo-responsivity. As shown in Fig. 2b, we find that compared to thepristine EG device, which is nearly insensitive to light (inset ofFig. 2b), an enhanced photoresponse is observed in the devicetreated with oxygen plasma for 20 s (OPTEG1) with a gain of ~104inthe DUV region (4.133 eV). Furthermore, the device (OPTEG2)treated with relatively longer (40 s) exhibits an enhancement inDUV-detection-gain by a factor of ~2.4. The significant performancecharacteristics of OPTEG photodetectors are summarised in Table 1.Our findings suggest that the oxygen plasma treated device is mostsensitive to DUV light. The gain of our OPTEG devices are severalorders of magnitude higher than many photodetectors based onpristine graphene [31], reduced graphene oxide (RGO) materials[32], and many other organic/inorganic semiconducting materials[33e40].

An important figure of merit of a photodetector is detectivity (orsensitivity), which refers to the minimum impinging optical signalthat is distinguishable above noise. It is denoted as D*, and ismeasured in units of Jones (cm Hz1/2 W�1). If noise from the darkcurrent is the dominant contribution, the detectivity can beexpressed as:

D* ¼ (A1/2PR)/(2qId)1/2 (2)

Where Id is the dark current, q is the electron charge. Underillumination at 4.133 eV (DUV light, Fig. 2b), the calculated detec-tivities of OPTEG1 and OPTEG2 devices (at fixed 2 V bias)are ~ 1.3� 1012 Jones and 3.2� 1012 Jones respectively (see Table 1).The detectivity of our OPTEG device is much higher compared tothe detectivity of previously reported photodetectors based onorganic (5TmDCA)/inorganic(ZnO) [40] hybrid systems(1010 Jones),many other organic/inorganic systems [34,41,42], and comparableto a commercial InGaAs (1012 Jones) [43] photodetector or highperformance AlGaN (3 � 1012 Jones) UV detector, however theOPTEG is much thinner (4-layer graphene only) [44].

We turn now to reveal the physical origins of the anomalousphotoresponse observed in the OPTEG devices. The electronicstructure and optical conductivity of graphene, which are deter-mined by the interplay of e-h and e-e interactions, can be revealeddirectly using spectroscopic ellipsometry. We perform spectro-scopic ellipsometry measurements on pristine EG, OPTEG1 andOPTEG2 samples, as shown in Fig. 3. For pristine EG, we observe aprominent peak centered ~4.4 eV, which is attributed to the reso-nant exciton [26].The resonant exciton in EG is found to be red-shifted as compared to free standing graphene which occurs~4.6 eV [19]. This suggests that the strength of e-h interactions forpristine EG is ~800 meV yielding to a larger exciton binding energycompared to ~600 meV for free standing graphene, given theinterband transition of ~5.2 eV [24,26].

Fig. 1. Surface morphology using Atomic Force Microscopy (AFM): (a) an AFM image of pristine epitaxial graphene (EG) (b) an AFM image of oxygen plasma treated EG (OPTEG).Surface similarity with no significant peeling/damage of the film supports that the oxygen plasma used in this study is ‘mild’. (c) Wide-scan PES spectra aquired from pristine EG andOPTEG. (d) C 1s of pristine EG. (e) C 1s of OPTEG showing the content of chemisorbed CeO and C]O species. (f) Raman spectra aquired from pristine EG and OPTEG for two differenttreatment times: 20 s (OPTEG1) and 40 s (OPTEG2). (A colour version of this figure can be viewed online.)

R.S. Singh et al. / Carbon 106 (2016) 330e335332

A new, important phenomenon from spectroscopic ellipsometryis that upon oxygen plasma treatment, we observe spectral weighttransfer in a broad energy range (Fig. 3). Upon oxygen adsorption,spectral weight transfer occurs from below ~4.5eV to higher en-ergies. As a result, the resonant excitonic peak enhances and shiftstoward higher photon energy (or blue-shifted) and the opticalconductivity below the resonant exciton is suppressed. Such ananomalous spectral weigh transfer is a fingerprint of strong elec-tronic correlations [45e47]. This implies that the oxygen adatomsintroduce strong electronic correlations. We note that such opticalproperties are also rule out the role of other impurities such as OH�

because the presence of OH� might shift the resonant exciton peaktoward lower energy accompanied by states at and near Fermienergy as opposed to our observations [48]. Based on our syn-chrotron PES core level spectra (Fig. 1), various C and O bonds are

observed and this may lead to such a spectral weight transfer due tohybridizations [25]. Intriguingly, for the OPTEG2 sample, the opticalconductivity is nearly visible around 1.5 eV and thus resonantexciton excitation is dominant. This itself is a new observation andthe behaviour of optical conductivity in oxygen-modified epitaxialgraphene is different than for single layer graphene previouslyreported [24].Furthermore, in contrast to bandgap in grapheneoxide (GO) material, we realize that graphene layers in our case arepartly oxidized, and it does not result in bandgap creation inOPTEG. Absence of bandgap in OPTEG is further confirmed byelectrical measurements (Fig. S2 and Fig. S3, Supplementary) andultraviolet photoelectron spectroscopy (UPS) measurements(Fig. S4, Supplementary). As such, OPTEG is different from semi-conducting GO and photoconductive mechanism in OPTEG can beexplained as follows.

Fig. 2. (a) A schematic of the device showing the OPTEG active region and pristine EG protected underneath two Cr/Au electrodes. (b) Spectral photoresponse acquired from pristineEG (inset), OPTEG1(red), and OPTEG2(blue). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 1Summary of OPTEG photodetectors’ characteristics under illumination at 4.133 eV (UV light) and its comparison with other UV photodetectors.

Active material Oxygen plasma treatment time (second) Gain Photoresponsivity (A W�1) Detectivity (Jones) Bias voltage (V) Reference

OPTEG1 20 1.00 � 104 2.42 � 103 1.3 � 1012 2 This workOPTEG2 40 2.23 � 104 5.39 � 103 3.2 � 1012 2 This workrGO e 0.40 0.12 e 1 [9]rGO/TiO2 e 0.75 0.18 6.82 � 1011 5 [42]AlGaN e 1.2 � 10�2 1.2 � 10�2 3 � 1012 4 [44]GaN e 0.52 0.15 e 2 [49]SiC e 0.47 0.10 e 50 [50]AlGaN/GaN e 2.78 � 105 7.0 � 104 e 100 [51]

Fig. 3. Optical conductivity data acquired from pristine EG (Black), OPTEG1 (red), andOPTEG2 (blue) materials. (For interpretation of the references to colour in this figurelegend, the reader is referred to the web version of this article.)

R.S. Singh et al. / Carbon 106 (2016) 330e335 333

The conventional photoresponse in semiconductors is known tooccur due to e-h pairing (or exciton formation). When light shineson doped semiconductors, the e-h pairs, which arise from thedoping, are created and quickly dissociated generating (local) cur-rent. This photoresponse is what has been normally detected inconventional doped semiconductors. The anomalous photo-response observed in OPTEG is somewhat different. It can beexplained by the presence of the resonant exciton. Since the reso-nant exciton is an intrinsic property of graphene and occurswithout any additional doping (hole or electron), the resonant

exciton is dominant in optical transitions. However, we do notobserve such a photoresponse in pristine EG. This may be becausethe optical conductivity below the resonant exciton is very high,which may hinder the dissociation of e-h pairs due to interbandtransitions at lower energies. Furthermore, carrier transition andrecombination in pristine graphene is ultrafast (fs or ns), whichresult in high speed but limits the efficiency in graphene basedphotodetectors. The adsorbed oxygen molecules in OPTEG canhinder the ultrafast recombination (extending the life time of car-riers to the order of one second, Fig. 4) and suppress the opticalconductivity below the resonant exciton, leaving the resonantexciton as the dominant optical excitation in the DUV range. Thus,this generates a good platform for the dissociation of resonantexcitons, generating a huge photoresponse. Such a scenario isconsistent with experimental observations. For OPTEG1 (20 s ox-ygen plasma treatment), the resonant exciton shifts toward higherenergy accompanied by suppression of optical conductivity atlower energy (Fig. 3), while an anomalous photoresponse is thengenerated(Fig. 2b). For OPTEG2 (40 s oxygen plasma treatment), theresonant exciton shifts toward even higher energy accompanied byfurther suppression of optical conductivity at lower energy, and thisfurther enhances the photoresponse in the DUV range.

The electrical and photoconductive characteristics of the OPTEGdevices are further elaborated (see Fig. S2, Supporting Information).Our findings show that the device undergoes changes in channelresistance (resistance increases ~3e5 times in treated devices withrespect to the pristine EG channel) after oxygen plasma treatmentsfor 20 s and 40 s, and this is expected due to modification of the sp2

graphene network. Fig. 4a-b show the time response of OPEG1 andOPTEG2 devices (at fixed bias 2 V) illuminated with white light(intensity ~ 30 mW cm�2). We obtain that the time constants for

Fig. 4. (a), (b) Temporal photocurrent Iphunder white light (light intensity ~ 30 mW cm�2, bias voltage ¼ 2 V) obtained from EG devices treated for 20 s (OPTEG1) and 40 s (OPTEG2)respectively. The time constants for rising (tr) and decay (td) by fitting the rising and falling edges with equations I ¼ 1-exp(t/tr) and I ¼ exp(t/td), respectively. Rise (red) and decayexponential (blue) fittings, estimating response times of OPTEG1 (tr ¼ 7 s, td ¼ 12 s) and OPTEG2 (tr ¼ 21 s, td ¼ 41 s) devices respectively. (c) Photocurrent obtained from a OPTEGdevice at different light intensities. (d) Photocurrent vs. intensity (F) of light. The inset shows a plot in logarithmic scales. (For interpretation of the references to colour in this figurelegend, the reader is referred to the web version of this article.)

R.S. Singh et al. / Carbon 106 (2016) 330e335334

rising (tr) and decay (td) are 7 and 12 s for OPTEG1 and increase to21 and 41 s for OPTEG2, respectively (Fig. 4a-b). Our findings showthat OPTEG photodetector is faster as compared to RGO based UVphotodetectors (response time is of the order of minute) reportedpreviously.32 The speed of the OPTEG photodetectors may furtherbe improved by adopting stratergies that include i) reduction ofdevice length scale, and ii) coupling high mobility pristine gra-phene with OPTEG active layers. Fig. 4c shows the photogeneratedcurrent Iph at 2 V bias under different light intensities. The photo-current grows with light intensity, and tends to saturate at higherintensity of ~30 mA (Fig. 4d). The photocurrent Iph and light in-tensity F can be expressed by Iph∝ Fn, where n is an exponent, andhas a fitted value ~0.49 (Fig. 4d) for the OPTEG device. Furthermore,we also observe that the photocurrent in OPTEG devices is satu-rated (not presented here) at relatively low bias voltage ~ (2e2.5) V,and we opt for a 2 V bias as an appropriate fixed value to presentthe photoconductive characteristics throughout the article. Athigher bias voltage, the bias induced generation of dark current isenhanced, and can compete with the photogenerated current tosaturate it.

3. Conclusion

In summary, we have demonstrated that mild oxygen plasmatreatment in epitaxial graphene (OPTEG) based devices can beharnessed to produce high-gain deep-ultraviolet (DUV) photode-tectors. The OPTEG device exhibits photoconductive gain as high as~2� 104at 4.133 eV spectral light. Upon oxygen adsorption, spectralweight transfer occurs. It enhances the resonant exciton and

suppresses optical conductivity below it yielding an anomalousphotoresponse in the OPTEG device. The tunable, large photocon-ductive gain, high selectivity, and detectivity in range of 1 � 1012 to3 � 1012Jones of the OPTEG devices in the DUV spectral rangehighlight their potential advantages over Si and many other semi-conductor DUV photodetectors. The plasma process on large areaprocessable graphene is compatible with CMOS fabrication pro-cesses making it possible to scale up the fabrication of graphene-based high-gain DUV photodetectors.

4. Experimental

4.1. Preparation and characterization of epitaxial graphene

Epitaxial graphene (EG) films (~4-layer) were prepared byannealing chemically etched (10% HF solution) high-purity, semi-insulating 4HeSiC (0001) (resistivity � 105 U cm andthickness ¼ 0.37 mm, CREE Research Inc.) in ultrahigh vacuum(UHV) above 1100 �C. The EG sample preparation and its charac-terization using scanning tunnelling microscopy (STM) and Ramanmeasurements are published elsewhere [7,8]. A thermal evaporatorwas used to deposit Cr(10 nm)/Au(100 nm) metals through ashadow mask onto pristine EG to fabricate two-terminal devices.Subsequently, the EG device with two metal electrodes wasmounted in a sample holder for wire-bonding (Fig. S1, Supplem-etary information). Two copper wires were soldered to the elec-trodes for device measurements.

R.S. Singh et al. / Carbon 106 (2016) 330e335 335

4.2. Oxygen plasma treatment

Oxygen plasma (rf power: 6W, oxygen flow rate: 20 sccm)treatments of EG device were performed cumulatively (for 20 s and40 s) using a reactive ion etching (NTI RIE-2321) system.

4.3. Temporal photocurrent measurements

Temporal photocurrent measurements were performed with anAgilent B2912A precision source/measure unit system, with axenon arc lamp as white light source. The ON and OFF lightswitching was performed at 30 and 40 s intervals in OPTEG1 andOPTEG2 devices respectively.

4.4. Spectral photoresponse measurements

We performed spectral photoresponse measurements using ahome-built photoconductivity measurement setup which consistsof a tungsten halogen lamp coupled with a monochromator(iHR320, HORIBA), mechanical chopper, a pyroelectric detector, apreamplifier and two lock-in amplifier (SR830). The light intensitywas calibrated with a high sensitivity and spectrally flat pyroelec-tric detector (Newport, model 70362). Both the light intensity andphotocurrent signals were collected by two amplifiers. Spectralphotoresponses from OPTEG1 and OPTEG2 devices were measuredat fixed 2 V bias. All the measurements were performed at room-temperature and in ambient environment.

Acknowledgements

This research is supported by the Singapore National ResearchFoundation under its Competitive Research Programme (CRPAward No. NRF-CRP 8-2011-06 and R-143-000-360-281), MOEAcRF Tier-2 (MOE2015-T2-1-099, and MOE2010-T2-2-121), NUS-347 YIA, and FRC (R-144-000-368-112). Q.X. acknowledges a strongsupport by the Singapore National Research Foundation via afellowship grant (NRF-RF2009-06) and Nanyang TechnologicalUniversity through a start-up grant (M58113004).

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.carbon.2016.05.026.

References

[1] A.K. Geim, Science 324 (5934) (2009) 1530e1534.[2] F. Bonaccorso, et al., Nat. Photonics 4 (9) (2010) 611e622.[3] K.I. Bolotin, et al., Solid State Commun. 146 (9e10) (2008) 351e355.[4] J. McClain, et al., J. Phys. Chem. C 114 (34) (2010) 14332e14338.[5] Y.M. Lin, et al., Science 327 (5966) (2010), 662e662.[6] S. Hertel, et al., Nat. Commun. 3 (957) (2012) 1e6.[7] R.S. Singh, et al., Appl. Phys. Lett. 100 (2012) 093116.[8] R.S. Singh, et al., ACS Nano 5 (7) (2011) 5969e5975.[9] B. Chitara, et al., Adv. Mater 23 (45) (2011), 5339e5339.

[10] D.A. Dikin, et al., Nature 448 (7152) (2007) 457e460.[11] S. Park, et al., Adv. Mater 22 (15) (2010) 1736.[12] H.A. Becerril, et al., ACS Nano 2 (3) (2008) 463e470.[13] A. Bagri, et al., Nat. Chem. 2 (7) (2010) 581e587.[14] Y. Lin, et al., ACS Nano 4 (6) (2010) 3033e3038.[15] X.M. Geng, et al., Adv. Mater 22 (5) (2010) 638.[16] G. Konstantatos, et al., Nat. Nanotechnol. 7 (6) (2012) 363e368.[17] M.E. Itkis, et al., Appl. Phys. Lett. 98 (9) (2011).[18] A.H. Castro Neto, et al., Rev. Mod. Phys. 81 (1) (2009) 109e162.[19] V.G. Kravets, et al., Phys. Rev. B 81 (15) (2010) 155413.[20] K.F. Mak, et al., Phys. Rev. Lett. 106 (4) (2011) 046401.[21] I. Santoso, et al., Phys. Rev. B 84 (8) (2011) 081403.[22] P.K. Gogoi, et al., EPL Europhys. Lett. 99 (6) (2012) 67009.[23] Iman Santoso, et al., EPL Europhys. Lett. 108 (3) (2014) 37009.[24] I. Santoso, et al., Phys. Rev. B 89 (7) (2014) 075134.[25] P.K. Gogoi, et al., Phys. Rev. B 91 (3) (2015) 035424.[26] L. Yang, et al., Phys. Rev. Lett. 103 (18) (2009) 186802.[27] P.E. Trevisanutto, et al., Phys. Rev. B 81 (12) (2010) 121405.[28] M.A. Majidi, et al., Phys. Rev. B 90 (19) (2014) 195442.[29] H.Y. Jeong, et al., Nano Lett. 10 (11) (2010) 4381e4386.[30] U. Cvelber, et al., Appl. Surf. Sci. 253 (4) (2006) 1861e1865.[31] T. Mueller, et al., Nat. Photonics 4 (5) (2010) 297e301.[32] B. Chitara, Appl. Phys. Lett. 99 (11) (2011).[33] A. Behnam, et al., J. Appl. Phys. 103 (11) (2008).[34] M.S. Arnold, et al., Nano Lett. 9 (9) (2009) 3354e3358.[35] H.L. Xue, et al., Appl. Phys. Lett. 90 (20) (2007).[36] Y. Lin, et al., J. Electron. Mater 29 (1) (2000) 69e74.[37] H. Ohta, et al., Appl. Phys. Lett. 83 (5) (2003) 1029e1031.[38] I.S. Jeong, et al., Appl. Phys. Lett. 83 (14) (2003) 2946e2948.[39] M. Martens, et al., Appl. Phys. Lett. 98 (21) (2011).[40] M. Sofos, et al., Nat. Mater 8 (1) (2009) 68e75.[41] C.K. Wang, et al., Semicond. Sci. Technol. 20 (6) (2005) 485e489.[42] K.K. Manga, et al., Adv. Mater 22 (46) (2010) 5265e5270.[43] Y. Yamamonto, et al., Science 314 (5806) (2006) 1761e1764.[44] G. Mazzeo, et al., IEEE Sensors J. 6 (4) (2006) 957e963.[45] H. Eskes, et al., Phys. Rev. Lett. 67 (8) (1991) 1035e1038.[46] Y. Ohta, et al., Phys. Rev. B 46 (21) (1992) 14022e14033.[47] P. Phillips, et al., Rev. Mod. Phys. 82 (2) (2010) 1719e1742.[48] J.A. Yan, et al., Phys. Rev. B 82 (2010) 125403.[49] C.K. Wang, et al., Semicond. Sci. Tech. 20 (2005) 485.[50] G.A. Shaw, et al., Proc. SPIE 7320 (2009) 73200J.[51] M. Martens, et al., Appl. Phys. Lett. 98 (2011) 211114.