4428 Phys. Chem. Chem. Phys., 2013, 15, 4428--4435 This journal is c the Owner Societies 2013

Cite this: Phys. Chem.Chem.Phys.,2013,15, 4428

Critical parameters in exfoliating graphite intographene†

Matat Buzaglo,a Michael Shtein,a Sivan Kober,a Robert Lovrincic,b Ayelet Vilan*b

and Oren Regev*ac

Dispersing graphite into few-layers graphene sheets (GS) in water is very appealing as an environmental-

friendly, low-cost, low-energy method of obtaining graphene. Very high GS concentrations in water

(0.7 mg mL�1) were obtained by optimizing the nature of dispersant and the type of ultra-sonic generator.

We find that a multi-step sonication procedure involving both tip and bath sources considerably enhances

the yield of exfoliated GS. Raman and transmission electron microscopy indicate few-layers graphene

patches with typical size of B0.65 mm in one dimension and B0.35 mm in the other. These were further

employed in combination with water-dispersed CNTs to fabricate conductive transparent electrodes for a

molecularly-controlled solar-cell with an open-circuit voltage of 0.53 V.

Introduction

Within the ‘‘gold-rush’’ for graphene applications, the questionof large scale availability of graphene and its processibilityis critical for actual technology.1 Flexible and disposable elec-tronics are the technology alternatives to the key role of paperin our society.2 They require conductors, semi-conductors andinsulators that preserve their electronic performance underconsiderable stretching and bending.2 Large-area electronics(displays, photovoltaics) as well as disposable electronics (RFtags, bio-sensors) also require cheap processing proceduressuch as low-temperatures, spin-coating and ink-jet printing.Although conducting polymers basically answer these require-ments, carbon-based nano-materials such as graphene sheets(GS) or carbon nanotubes (CNTs) are superior to conductivepolymers in their electrical conductance as well as stability towardoxidation under ambient conditions.3 For such applications, GSdispersed in water at high concentrations is highly appealing. Thenatural way to obtain such dispersions would be ‘‘top-down’’exfoliation of graphite into GS rather than ‘‘bottom-up’’ growthof single GS films.1,4 Despite the un-matched electronic qualityand uniformity of the latter, it produces much smaller amount

of material at a much higher energy-cost compared to bottom-up exfoliation of graphite.

Exfoliation of graphite into a GS dispersion could be inducedby either chemical or physical forces to overcome the van derWaals force holding the GS in the solid graphite.5 Chemicalexfoliation is based on oxidation of graphite and dissolution ofgraphene-oxide (GO).6,7 Reducing this GO back to GS requiresharsh conditions (either thermal or chemical) and still the finalfilms contain considerable amount of defects that deterioratetheir electrical quality.6,8 Physical exfoliation is driven by ultra-sonication, which forms cavities in the graphite; these are laterfilled by solvent or surfactant molecules that separate the GSbeyond the van der Waals distance of B2 nm and eventually leadto dispersion of GS composed of single to few graphene layers.During sonication, bubbles collapse near the graphite surface andcreate surface pitting.9 The high temperature and pressure peaksat the graphite surface produce surface defects and increase thesurface area of friable solids. The ruptured surface is then easilyexfoliated. The local turbulent flow associated with acousticstreaming improves mass transport between the liquid phaseand the surface, thus enhancing the exfoliation rates. There aretwo commonly used types of sonication source: Tip Sonication(TS, also known as ‘horn’ sonication) and Bath Sonication (BS),which are very different in the power they deliver. Here we showthat the integral energy/volume supplied by sonication is not theonly parameter for efficient exfoliation.

The final GS concentration is a major feature in evaluatingthe efficiency of a dispersion process. The Coleman group usedsurfactant-less solvent, N-methylpyrrolidone (NMP), to disperse2 mg mL�1 GS that was later concentrated to 63 mg mL�1.10

a Department of Chemical Engineering, Ben-Gurion University of the Negev,

Beer-Sheva, Israelb Department of Materials and Interfaces, Weizmann Institute of Science, Rehovot,

Israel. E-mail: ayelet.vilan@weizmann.ac.ilc Ilse Katz Institute for Nanoscale Science and Technology,

Ben-Gurion University of the Negev, Beer-Sheva, Israel. E-mail: oregev@bgu.ac.il

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c3cp43205j

Received 18th June 2012,Accepted 9th January 2013

DOI: 10.1039/c3cp43205j

www.rsc.org/pccp

PCCP

PAPER

Dow

nloa

ded

by U

nive

rsity

of

Abe

rdee

n on

04

Mar

ch 2

013

Publ

ishe

d on

10

Janu

ary

2013

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

3CP4

3205

J

View Article OnlineView Journal | View Issue

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 4428--4435 4429

The absence of surfactants in this method is a clear advantagefor processing, yet NMP is very difficult to remove by itself, dueto its extremely high boiling point (204.3 1C).

Exfoliation by surfactants has the advantage of usingwater as a universal, environmentally friendly solvent, whichis compatible with charged surfaces and solutes. Water has areasonably high vapor pressure, which facilitates its removal inGS processing such as spin coating and casting, though rinsingis required to remove the surfactant from the deposited GSfilms. Earlier reports indicate record GS concentrations of0.3 mg mL�1 in aqueous, surfactant-assisted dispersion.11

The Hersam group dispersed GS in water using Pluronic(non-ionic blockcopolymer) to get a final GS concentration of0.07 mg mL�1.12 We note that the yield of initial graphite thatcan be successfully dispersed into GS increases with dilution ofgraphite content,13 and therefore we prefer to compare finalconcentration rather than yield.

Here we optimize the sonication process toward maximizingthe final GS concentration by manipulating tip and bathsonication cycles. Moreover, we show that the effectiveness ofthe dispersant is dominated by its hydrophobic part, wherearomatic or planar moieties are clearly more efficient thanchain-like hydrophobic tails. We then spin-coated a GS:CNTwater dispersion to fabricate soft, transparent conductor thinfilms, which are applicable in Schottky-junction solar cell andas a top contact for molecular electronics.

ExperimentalMaterials

Graphite flakes (CAS 7782-42-5), sodium dodecylsulfate (CAS151-21-3), MWCNT (Nanocyl 7000), sodium cholate hydrate(CAS 206986-87-0), didodecyldimethylammonium bromide (CAS3282-73-3), and Triton X-100 (CAS 9002-93-1) were purchased fromSigma-Aldrich and used as received. Pluronic P-103, P-123, P-65,P-84, and F-127 (CAS 9003-11-6, Fig. S1 of ESI†) from BASF wereused as received. Deionized water with resistivity of 18.2 MO cmwas used.

Dispersion procedure

Graphite flakes (1 wt%) were mixed with a pre-preparedsolution of surfactant (0.5–1 wt%) in water and sonicated usingcombinations of bath sonication (BS): Elma sonic (model S10;30 W 37 kHz, Singen) and tip sonication (TS): VCX 400 (400 W20 kHz, mtip, Sonics & Materials Inc.). The total energy/volumewas calculated by summing the TS and BS energies and dividingby the solution volume. The exact power delivered by eachinstrument was measured to be 0.30 � 0.05 W for BS (Fig. S2a,ESI†) and 16.0 � 0.3 W or 2.63 � 0.32 W (amplitude of 38% or20% respectively) for the TS (Fig. S2b, ESI†). All sonicationparameters were kept constant. In both TS and BS procedureswe used an ‘‘ice bath’’ in order to keep the temperature constant(B0 1C). The position of the vial in the bath sonicator affectsthe energy it receives. Therefore, we initially located the ‘‘highestenergy’’ spots in the bath sonicator by operating it with analuminum foil covering its surface (without vials). The spots of

maximal sonication energy are observable as clear etching holesin the aluminum foil. All bath-sonication experiments refer tothe vial positioned exactly at this maximal energy location.

For a CNT, maximal exfoliation occurs when the dispersantconcentration is near its CMC.14 We tested few graphite : dispersantweight ratios and found an optimal ratio of 2 : 1 and 1 : 1 forsurfactant or pluronic dispersants, respectively (Fig. S3 of ESI†), inaccordance with an independent study.12 After sonication the dis-persion was centrifuged (Megafuge 1.0, Heraues), for 20 minutes,4000 rpm, and 85 vol% of the supernatant was carefullycollected and kept at ambient conditions.

UV-vis spectra were measured by a Jasco V-530 spectrometerusing plastic cuvettes. Samples were diluted (1 : 6) to avoid absorp-tion saturation. The dispersant absorption over this range isnegligible as evident from a reference spectrum of a dispersant-only solution at high concentration (5 wt%, see Fig. S4 of ESI†).

Thermogravimetric analysis (TGA) for concentration determi-nation is carried out with Mettler-Toledo (TGA/STDA851). Thesupernatant phase was filtered (0.22 mm pores diameter filter)and the solid residue was washed with DI water to remove excessdispersant. The wet powder on the filter was dried at 120 1C for1 hour, weighted and then loaded into 100 mL aluminumcrucibles and measured under N2 at a flow rate of 50 mL min�1

and at a heating rate of 10 1C min�1 from 100 1C to 400 1C,after which the temperature is kept constant at 400 1C for30 minutes. The GS and the dispersant concentrations werethen calculated from the thermograms.

Transmission electron microscopy (TEM) micrographs wereobtained by FEI Tecnai 12 G2 TWIN TEM operated at 120 kV.Dry samples were prepared on holey-carbon-coated coppergrids (300 mesh, lacey carbon, Ted Pella) by placing a drop ofdispersion on a grid and allowing it to dry at ambient condi-tions before storage. The microscope was operated at 120 kV inlow electron dose mode (to reduce radiation damage) and witha few micrometers underfocus to increase phase contrast. Imageswere recorded on a Gatan 794 CCD camera and analyzed byDigital Micrograph 3.6 software.

Atomic force microscope (AFM) images were taken withDimension 3100 SPM (VEECO) in tapping mode using VEECORTESP silicon tips. The dispersion was spin coated on SiO2

wafers and allowed to dry by evaporation at ambient tempera-ture for 24 hours before measuring.

Raman spectra of the GS were measured by Jobin-Yvon HRLabrRam micro-Raman at 514 nm on quartz slide. The sampleswere dried out on the slide from a 200 mL drop before themeasurement.

Mixed GS–CNT films preparation

GS–CNT dispersions were prepared by mixing GS:TX-100 inwater and CNT:TX-100 in water in the desired ratio and sub-sequent bath sonication for 30 minutes to ensure a uniformmixture. Various GS : CNT ratios were tested, but in the rangeof 4 : 1 to 1 : 1 mass ratio no significant difference in terms oftransmittance versus resistance was observed. Spin coating ofGS alone resulted in poor coverage and high resistance.

Paper PCCP

Dow

nloa

ded

by U

nive

rsity

of

Abe

rdee

n on

04

Mar

ch 2

013

Publ

ishe

d on

10

Janu

ary

2013

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

3CP4

3205

JView Article Online

4430 Phys. Chem. Chem. Phys., 2013, 15, 4428--4435 This journal is c the Owner Societies 2013

The films were spin-coated on two different substrates: (i)propyl-trimethoxysilane terminated glass surfaces for the trans-mittance and sheet resistance measurements and (ii) methoxyterminated Si(100) wafers (n-type, B0.1 O cm) for the SEMmeasurements and solar cell tests. Both treated substrateswhere hydrophobic with a water contact angle of around90 degrees. On surfaces with clearly higher or lower contactangles the GS–CNT mixtures showed much poorer adhesion.The GS–CNT dispersion was spin-coated on the substrates at atypical speed of 1500 rpm. To enhance coverage, the spincoating procedure was repeated several times with intermediaterinsing with water to remove the surfactant.

Optical transmittance

Optical transmittance for GS/CNT films on glass was measuredin the wavelength range from 400 to 1000 nm with a WoollamM-2000 spectrometer at normal incidence of light. A bare glasssubstrate was used as reference. The transmittance values givenare the averages over the whole wavelength range.

The sheet resistance Rs was measured on 10 � 10 mm2

samples using a home-made 4-point probe setup with a probespacing of B1 mm. Each value is the average of 5 measure-ments on 5 different spots on one sample surface.

Si/GS solar cell preparation

Schottky-contact solar-cells with 1 � 1 cm2 effective area werefabricated on methoxy-terminated n-Si(100) with GS–CNT film astop transparent electrode. GS–CNT mixtures were spin-coated asabove and subsequently a 200 nm thick silver grid (1 mm linespacing) was evaporated to enhance the fill factor. Photo-voltaicmeasurements were done using a white light source, calibratedto yield roughly 100 mW cm�2.

Results and discussion

Toward achieving high final GS concentration, one can manipulatea few major parameters, including the dispersing agent(dispersant), sonication type, time and power, carbon source,and solvent properties. In this study we choose to focuson sonication procedures and dispersant type. We choosegraphite flakes and water as an abundant carbon source andenvironmentally-friendly solvent.

The sonication is aimed at initiating the exfoliation process.We observe a strong increase in GS concentration by applying asequence of two sonication types: Tip Sonication (TS) and BathSonication (BS) rather than one of them alone. Earlier studiesreport low GS concentration for TS alone,15,16 while BS yieldshigher GS concentration in both pure solvents and dispersant-assisted solution.13,16,17 It was suggested that a combination ofhigh-energy sonication (TS) and low-energy sonication (BS)improves CNT exfoliation.18–20 Here we show that thisapproach is also highly effective for graphite exfoliation. Westudied the relative contributions of different sonication cyclestoward high final GS concentration.

Determination of GS concentration

The dispersant solution is mixed with graphite under sonication.Dark, ink-like slurry is obtained. Following sonication and centrifu-gation, the dispersion is separated (decanted) to exfoliated GS(supernatant) and aggregated graphite (precipitate) phases. Theconcentration of the supernatant phase is our major indication forthe efficiency of the different dispersants and sonication procedures,and the stability of the GS dispersions against precipitation. Theconcentration was routinely determined using UV-vis absorptionbased on thermo-gravimetric (TGA)-determined calibration curve.This careful approach was considered because there is no acceptedvalue for extinction coefficient for GS dispersions.

The calibration curve was established by accurately weightingthe dry content (after filtration and drying steps) of the super-natant (MS, [mg]) and then excluding the residual dispersantweight fraction (jS) by TGA. Finally, the GS concentration in thesupernatant (GSS, [mg mL�1]) is determined by eqn (1):

GSS = MS � (1 � jS)/VS (1)

where VS is the supernatant volume [mL].Fig. 1a shows a typical TGA curve for the GS:TX-100 mixture.

The weight loss between 200–400 1C is due to evaporation ofTX-100. While the boiling point of TX-100 is 270 1C, it requiresheating up to 400 1C for its complete removal, as indicated by athermogram of pure TX-100 (Fig. S5a, ESI,† 98.0 wt% loss).However, heating a pure graphite up to 400 1C shows onlyminor weight reduction (0.8 wt% loss, Fig. S5b, ESI†) due to GSoxidative decarbonylation caused by oxygen impurity in theTGA chamber.

UV-vis absorption was measured in the 400–700 nm range.In this spectral range, all the tested GS dispersions show afeatureless spectrum with negligible dispersant absorption.Using a GS dispersion of calibrated concentration (see above)with several dilutions, we constructed a calibration curve at l =660 nm where GS absorption is rather wavelength independent(Fig. S4 in the ESI†). According to the Beer–Lambert law, theabsorbance increases linearly with the concentration (Fig. 1b),with a slope that equals the extinction coefficient (e). Werepeated this calibration twice (circles and triangles in Fig. 1b)and found an extinction coefficient of 2960 � 40 mL mg�1 m�1.This value is within the range of the reported extinction coeffi-cients for GS and CNT dispersions in water.16 All the followingGS concentration values were determined by using UV-vis andthe above extinction coefficient. Our study is focused on twomajor parameters, which are expected to enhance the exfoliationof graphite into graphene: the type of dispersant and the sourceof energy (sonication). We start by considering the choice ofdispersant using a fixed sonication cycle, and in the followingsection we show the clear advantage of this specific sonicationcycle over other alternatives.

Choice of dispersant

The dispersant stabilizes the GS due to hydrophobic–hydrophilicrepulsion and steric repulsion. In order to maximize the exfolia-tion, we study the effect of the dispersant chemistry on the final

PCCP Paper

Dow

nloa

ded

by U

nive

rsity

of

Abe

rdee

n on

04

Mar

ch 2

013

Publ

ishe

d on

10

Janu

ary

2013

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

3CP4

3205

JView Article Online

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 4428--4435 4431

GS concentration. All the GS dispersions were highly stableagainst de-mixing, as indicated by the minor reduction inUV-vis absorption (1% to 15%) after 30 days of storage underambient conditions without any sonication or other treatmentafter storage (Fig. S6 of ESI†).

We employed a wide range of dispersant types: anionic,cationic, and nonionic surfactants (see schemes in Fig. S1 ofESI†) and found no clear trend in GS concentration with thepolarity of the head-group of the dispersant (Fig. 2) or thesolution surface tension (Fig. S7 of ESI†). In contrast, the natureof the hydrophobic part is critical for an efficient dispersion.

TX-100 gave the highest GS concentration because it includes abenzene ring with a strong p–p interaction with the aromaticstructure of GS.21 Four fused-ring pyrene is an alternativeefficient GS dispersant.22 The aromatic dispersant, TX-100, issuperior to all other dispersants studied (Fig. 2) regardlessof sonication procedure (vide infra). Between the anionicsurfactants, we see that the planar and rigid Sodium Cholate(SC) molecule is a better dispersant than the flexible sodiumdodecylsulphate (SDS) molecule, most probably due to thegeometrical similarity of the SC with the flat GS and its bettersteric repulsion induced by the bulky set of aliphatic rings. Wealso tested GS dispersion by some of the Pluronics23 family ofpolymers, as recently studied by others.12

The Pluronics are triblock copolymers, consisting of hydro-philic/hydrophobic/hydrophilic blocks, with poly ethylene gly-col (PEO) and poly propylene glycol (PPO) for the hydrophilicand hydrophobic blocks, which differ in their length (markedas n and m, Fig. S1 and S8 of ESI†).

Among the Pluronics dispersants we find that Pluronics with40 PO monomers is optimal for GS dispersion with negligibledependence of the EO block’s size, in line with previousexperimental24 and numerical25 findings for CNT dispersions.In contrast, Hersam and coworkers found only a weak depen-dence on the PO chain length and much stronger effect of thehydrophilic chain with an optimum around 75 EO monomers(6.5 kDa).12 In terms of final GS concentration the F127 gave asimilar result (Fig. 2) to that of ref. 12, while the lower molecularweight Pluronics gave 2–3 times higher GS concentration here(Fig. 2 and S8 of ESI†) compared to the previous study.12 Apossible reason could be the different sonication cycles. Wenote though that this observation is based on a rather limitednumber of Pluronics types.

Optimizing the sonication cycle

Former experience with CNT exfoliation suggested that combinedtip and bath sonication cycles considerably enhance the finalCNT concentration. Our results show that this is also the casefor GS. Tip and bath sonication differ considerably in the power

Fig. 1 Experimental determination of GS concentration: (a) thermogram of theGS:TX-100 dispersion prepared by TBT. The dotted-line indicates the sampleweight-loss in the ordinate (left Y-axis) with respect to time or temperature(abscissa). The other curve is the time derivative of the weight-loss curve (mg s�1,right Y-axis). Dashed line is a stepwise approximation to the weight-loss (between 200–400 1C) due to evaporation of the dispersant. The curveshows that an initial weight of 1.82 mg of dried supernatant powder lost0.32 mg. Therefore, the percentage of residual dispersant (jS) is 18.14%. (b)A UV-vis calibration curve, plotting the absorption intensity at l = 660 nm againstthe TGA-determined GS concentration for various dilutions of GS:TX-100 disper-sion prepared by TBT sonication. The extracted extinction coefficient, e (slope) is2960 � 40 mL mg�1 m�1. Circles and triangles correspond to two different stocksolutions. The negligible shift between them indicates minute inaccuracy indetermining the starting GS concentration.

Fig. 2 GS concentration by TBT for various dispersants, using identical TBTsonication treatment (integrated energy 23 kJ; 9 min TS, 3 h BS and 9 min TS).

Paper PCCP

Dow

nloa

ded

by U

nive

rsity

of

Abe

rdee

n on

04

Mar

ch 2

013

Publ

ishe

d on

10

Janu

ary

2013

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

3CP4

3205

JView Article Online

4432 Phys. Chem. Chem. Phys., 2013, 15, 4428--4435 This journal is c the Owner Societies 2013

they supply. Therefore, the trivial assumption would be that thesubstantially higher power of TS would be more efficient than theweak BS. Fig. 3a compares the efficiency of either TS or BS alone,or their combination (TBT). Clearly, in terms of integrated sonica-tion energy, TS alone is the least efficient option, and the weakerBS leads to much higher GS concentration. Still, combining thetwo sources gives the highest GS concentration. This markeddifference is ocularly visible: a much darker supernatant (higherGS concentration) is obtained by TBT than by BS alone (Fig. 3a,inset). Although the BS contribution to the total energy is rathersmall (B1% of the total energy), it has a substantial contributionto the final GS concentration (40%) as shown in Fig. 3b.

Therefore, we further optimized the relative energies of thebath and tip components of the TBT procedure to enhance theGS concentration. Fig. 3b shows an overall linear dependenceof GS concentration on the total sonication energy/volume,which was mostly delivered by the two tip cycles. The effect ofthe bath cycle is presented by three data sets, each for differentbath cycle energy (note that on the abscissa of Fig. 3b, thisdifference in BS energy is almost indistinguishable).

The effect of the BS cycle becomes substantial only above5 kJ mL�1, where doubling the bath energy/volume (from 0.075

to 0.15) increases the GS concentration by B25%. Trying togain insight into the different roles of the bath and tip sonica-tion steps, we note that for the same sonication energy/volumethe much weaker (slower) BS procedure is more efficientthan the stronger (rapid) TS procedure (Fig. 3a). We suggestthat sonication-driven exfoliation involves both cracking thegraphite and peeling of GS. Cracking increases the surface-areaof the graphite, and therefore improves the peeling efficiencybecause the ‘un-zipping’ of the graphite starts from theedges.16,26 A possible explanation to the improved efficiencyof the combined TBT cycle over either BS or TS alone could bethat the tip sonication is more efficient in cracking the graphitewhile bath is better in exfoliation, though some cracking andpeeling probably occur by any sonication source.

The maximal GS concentration received here is 0.7 mg mL�1.This is higher than formerly reported GS concentration in water(Table S1 of ESI†) though lower than obtained in NMP as apure solvent.10 It should be mentioned that further procedureoptimization of TBT time sequence, dispersant-to-graphiteweight ratio, and centrifugation time and rpm could enhancethe final GS concentration even further.

Quality of resulting GS

In this section we evaluate the quality of the resultinggraphene. The thickness of the dispersed GS is characterizedby room temperature transmission electron microscopy (TEM)in imaging and diffraction modes. It reveals the presence ofstacks of a few layers (3–5, Fig. 4a), and some folded individualGS (not shown). The number of GS layers in a patch is analyzedby ‘‘edge counting’’ in TEM images, indicating high electrondensity lines in the edge-on position. The size of the GS wasextracted from TEM images. We define a GS ‘average length’ asthe square root of the patch’s area (min � max lengths). Theaverage length was 400 � 200 nm, regardless of sonicationenergy (Fig. S9(b), ESI†). We further studied several GS disper-sions by cryo-TEM27 (not shown), which preserves the liquidenvironment, and found similar GS size to those imaged by thedry, room temperature TEM technique.

Raman spectroscopy (Fig. 4b) is used to differentiatebetween single to few layers graphene (FLG), and graphite.28

The Raman spectrum is characterized by three major bands: (i)the D band (or peak) at B1350 cm�1 is due to the first-orderphonons and indicates the disorder of the GS; (ii) the G band atB1580 cm�1 is related to the zone center Raman allowed band;and (iii) the 2D band at B2700 cm�1 is due to the second-orderphonons.29 The 2D band splits due to bilayer interaction andthe high energy sub-peak becomes considerably stronger forgraphite, compared to FLG.28 A Lorentzian line fit analysis ofthe 2D band (Fig. S10, ESI†) shows that it has the expected28 twodoublets. For our GS the intensity of the two doublets wassimilar in a marked contrast to the graphite reference (Fig. S10,ESI†). This indicates that the thickness of our GS does notexceed 5 layers. This is in line with our TEM results.

AFM characterization of spin-coated GS films reveals(Fig. S11 of ESI†) isotropic patches with average length ofapproximately 300 nm. The thin films are 1–10 nm thick,

Fig. 3 Optimization of sonication procedure, dispersant and energy/volume. (a)GS concentrations upon bath sonication (BS), tip sonication (TS) and tip–bath–tipsonication (TBT). Inset: image of the supernatant of the GS dispersionsafter centrifugation; (b) concentration of GS (with TX-100) as a function ofintegrated sonication energy/volume, as measured by UV-vis absorption. Thesesolutions were treated by TBT cycles. For both panels the initial concentrationswere 10 mg mL�1 graphite and 10 mg mL�1 TX 100.

PCCP Paper

Dow

nloa

ded

by U

nive

rsity

of

Abe

rdee

n on

04

Mar

ch 2

013

Publ

ishe

d on

10

Janu

ary

2013

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

3CP4

3205

JView Article Online

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 4428--4435 4433

indicating stacks of 1–10 GS (height of a single GS is B1 nm).30

Thus TEM, Raman and AFM all point to few layers GS.

Formation of transparent conducting film

A natural application of GS could be in transparent-conductingfilms. Therefore, we used our aqueous GS solutions to producesuch films and characterize their performance. We found thatour casted GS dispersion has a sheet resistance of 20 kO sq�1,however with only 20% transmittance (cf. B90% desired). Themajor factor limiting the conductivity is the extreme resistivityat grain boundaries between adjacent GS, even for CVD-growngraphene.32 Mixing the GS with a CNT yields conductive films4

by providing percolation paths even at low surface coverage.Fig. 5 shows measured transmittance values, T, versus sheetresistance, Rs, for B20 samples of GS–CNT films directly spin-coated on glass. The samples were prepared at different GS :CNT ratios and spin-coating details (see Experimental). The

inset depicts a typical transmittance spectrum over the coveredwavelength range.

The spectra exhibit only weak wavelength dependence, asexpected from the optical constants of graphene and CNT.While the optical transparency of Fig. 5 is very high, the sheetresistance is B100 times larger than previously reported ones

(200–600 O at 80–90% optical transparency).4,33,34 However, theprocessing conditions of our approach have clear advantagesof being environmental friendly by using acid-free, water dis-persion; energy-cheap in using room-temperature process only;and scalability because it is based on direct spin coating on thedesired substrate. In contrast, previous works used hydrazineas solvent,33 started from CVD graphene,4 or prepared the filmsusing strong acids and vacuum filtration.34 The two mainreasons for the high resistance are insufficient overlap ofthe GS patches and intrinsic low carriers’ density in GS. There-fore, we hope that optimizing the deposition conditionsand chemically doping the GS can further reduce the sheetresistance. We further tested the performance of the GS–CNTtransparent electrode in a simple device. We fabricatedGS–CNT/Si Schottky-contact solar cells. This kind of solar cellshas attracted much attention lately.4,35,36 We note though, thatcurrent-driven devices such as solar-cells might not be themost appropriate applications for solution-cast graphitic films,compared to, e.g., electromagnetic interference shielding ortouch screens. Schottky-contact solar cells require the Si surfaceto be inverted (higher concentration of holes than electrons forthis n-Si substrate).

Inversion is rarely achieved practically; however, thin organicmonolayers (CH3–O–Si in our case) can cause such inversion,provided that they remain intact after depositing the top-contact.37,38

Thus, the soft-deposition by GS–CNT film is expected to bespecifically important for this type of solar cells.

Fig. 6 shows dark and light current for a 1 � 1 cm2 GS–CNT/Si Schottky-contact solar cell. A solar-cell performance is char-acterized by three typical parameters (Fig. 6): the short-circuitcurrent, Jsc, or current at V = 0, the open-circuit voltage, Voc, orthe voltage where the photo-current is nulled and the fill factor,FF, which is the ratio between the maximal power (I � V, grayrectangle in Fig. 6) and the ideal maximum power (Isc � Voc,dashed rectangle in Fig. 6). The Voc is 530 mV, which is fairlyhigh for an un-doped GS–CNT/Si solar cell,4,35 indicating thatthe top-contact deposition does not damage the surface

Fig. 4 Indication for few layers of graphene showing (a) room temperature TEMmicrographs of GS stacks from GS–TX-100. The diffraction pattern (inset) indicatesthat the GS are less than 5 layers thick,31 (b) the Raman spectra of graphene film onquartz substrate at 514 nm. Inset shows a zoom-up of the 2D peak.

Fig. 5 Transmittance (T) vs. sheet resistance (Rs) for mixed GS–CNT films spin-coated on glass substrates. Transmittance values are averaged and use of thebare glass substrate is the reference. The inset shows a typical transmittancespectrum.

Paper PCCP

Dow

nloa

ded

by U

nive

rsity

of

Abe

rdee

n on

04

Mar

ch 2

013

Publ

ishe

d on

10

Janu

ary

2013

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

3CP4

3205

JView Article Online

4434 Phys. Chem. Chem. Phys., 2013, 15, 4428--4435 This journal is c the Owner Societies 2013

passivation. The Jsc is 10 mA cm�2, which is lower than that ofref. 4. We suggest that this originates from shadowing from theAg grid and likely from poor contact between the GS–CNT filmand the Si. The FF is 42% because it is limited by the resistanceof our GS–CNT films, but still higher than in ref. 4, due to theused Ag grid. The fact that the Ag/GS–CNT/CH3–O–Si junctionbehaves as a solar cell with a high Voc implies that a p–njunction is formed within the n-Si surface. Such inversion ofcarriers-type is induced by the dipole of the monolayer.37,38

This is a foremost proof38 that the GS–CNT layer preserves themethoxy monolayer intact and prevents the evaporated Ag gridfrom penetrating to the Si substrate. Therefore, the GS–CNTfilms are not only transparent but also soft-contacts appropri-ate for molecular electronics.39

Conclusions

We report on surfactant-assisted GS dispersion in aqueoussolution which is simple, scalable, and produces stable aqueousdispersions, with GS concentration up to 0.7 mg mL�1. Graphiteexfoliation and stabilization in water was possible using avariety of dispersants. The most efficient of them was the TX-100,which is an aromatic, non-ionic surfactant, most probably dueto the p–p stacking to the GS and closer packing of dispersantover the GS due to the absence of electrostatic repulsion.Combining bath and tip sonication cycles (TBT) is much moreefficient in dispersing graphite into GS than either individualBS or TS treatments.

No significant difference in the structure, size, or exfoliationdegree of GS is detected between the GS produced by differentsonication procedures. The GS dispersions were very stable anddecrease by B10% after a month, regardless of dispersionprocedure. A mixture of GS:CNT water dispersions were usedto prepare transparent conductive electrodes by spin coating on

hydrophobic surfaces. The sheet resistance was 20 kO at 90%transmittance. This high series resistance deteriorates theperformance of inverted Si solar cells made using the GS:CNTfilm as transparent anode. Still, we obtained a considerableopen circuit voltage of 0.53 eV, which indicates that themolecular layer was not damaged by the GS:CNT deposition.

To summarize, the focus and novelty of the presented workare in the striking effect of the sonication cycle (TBT) on theresulting graphene concentration. While the effects of solvent(e.g., viscosity, density) and sonication parameters (e.g., power,total energy/volume and simultaneous tip and bath sonica-tion40) have been studied for SWNTs,41 the study of the TBTcycle has not been previously reported for either graphene orCNTs, to the best of our knowledge. The mechanism of CNTexfoliation and graphene unzipping are not necessarily similardue to different curvature and dimensionality. The applicationof this approach in forming transparent conducting filmindicates that our study can give a major leap in the quest toachieve graphene dispersion of high quality and concentration.

Acknowledgements

We thank Mr Juergen Jopp for conducting the AFM measure-ments, Dr Einat Nativ-Roth for the TEM imaging and Dr LeilaZeiri for performing the Raman spectroscopy measurements.Dr Tsachi Livneh is kindly acknowledged for helping in theanalysis of the Raman spectra results and critical reading.

References

1 B. Luo, S. Liu and L. Zhi, Small, 2012, 8, 630–646.2 B. D. Gates, Science, 2009, 323, 1566–1567.3 C. Biswas and Y. H. Lee, Adv. Funct. Mater., 2011, 21,

3806–3826.4 C. Li, Z. Li, H. Zhu, K. Wang, J. Wei, X. Li, P. Sun, H. Zhang

and D. Wu, J. Phys. Chem. C, 2010, 114, 14008–14012.5 L. X. Benedict, N. G. Chopra, M. L. Cohen, A. Zettl,

S. G. Louie and V. H. Crespi, Chem. Phys. Lett., 1998, 286,490–496.

6 H. A. Becerril, J. Mao, Z. Liu, R. M. Stoltenberg, Z. Bao andY. Chen, ACS Nano, 2008, 2, 463–470.

7 S. Ramesh, L. M. Ericson, V. A. Davis, R. K. Saini, C. Kittrell,M. Pasquali, W. E. Billups, W. W. Adams, R. H. Hauge andR. E. Smalley, J. Phys. Chem. B, 2004, 108, 8794–8798.

8 C. Mattevi, G. Eda, S. Agnoli, S. Miller, K. A. Mkhoyan,O. Celik, D. Mostrogiovanni, G. Granozzi, E. Garfunkel andM. Chhowalla, Adv. Funct. Mater., 2009, 19, 2577–2583.

9 K. S. Suslick, in Handbook of Heterogeneous Catalysis,ed. G. Ertl, H. Knozinger, F. Scuth and J. Weitkamp, VCH,Weinheim, 2nd edn, 1997, ch. 8.6, vol. Part A.

10 U. Khan, H. Porwal, A. O’Neill, K. Nawaz, P. May andJ. N. Coleman, Langmuir, 2011, 27, 9077–9082.

11 M. Lotya, P. J. King, U. Khan, S. De and J. N. Coleman,ACS Nano, 2010, 4, 3155–3162.

12 J.-W. T. Seo, A. A. Green, A. L. Antaris and M. C. Hersam,J. Phys. Chem. Lett., 2011, 2, 1004–1008.

Fig. 6 Current–voltage measurements for a 1 cm2 GS–CNT/Si Schottky-junctionsolar cell in dark and under illumination (B100 mW cm�2). The solar-cell para-meters: fill-factor (FF), open circuit voltage (Voc), and short circuit current (Isc) aregiven in the graph. A silver grid was evaporated on top of the GS–CNT for bettercurrent collection. The inset shows a SEM picture of a mixed GS–CNT film on Si.

PCCP Paper

Dow

nloa

ded

by U

nive

rsity

of

Abe

rdee

n on

04

Mar

ch 2

013

Publ

ishe

d on

10

Janu

ary

2013

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

3CP4

3205

JView Article Online

This journal is c the Owner Societies 2013 Phys. Chem. Chem. Phys., 2013, 15, 4428--4435 4435

13 M. Lotya, Y. Hernandez, P. J. King, R. J. Smith, V. Nicolosi,L. S. Karlsson, F. M. Blighe, S. De, Z. M. Wang,I. T. McGovern, G. S. Duesberg and J. N. Coleman, J. Am.Chem. Soc., 2009, 131, 3611–3620.

14 S. Utsumi, M. Kanamaru, H. Honda, H. Kanoh, H. Tanaka,T. Ohkubo, H. Sakai, M. Abe and K. Kaneko, J. ColloidInterface Sci., 2007, 308, 276–284.

15 A. A. Green and M. C. Hersam, Nano Lett., 2009, 9,4031–4036.

16 U. Khan, A. O’Neill, M. Lotya, S. De and J. N. Coleman,Small, 2010, 6, 864–871.

17 S. Vadukumpully, J. Paul and S. Valiyaveettil, Carbon, 2009,47, 3288–3294.

18 S. Giordani, S. D. Bergin, V. Nicolosi, S. Lebedkin,M. M. Kappes, W. J. Blau and J. N. Coleman, J. Phys. Chem.B, 2006, 110, 15708–15718.

19 B. R. Priya and H. J. Byrne, J. Phys. Chem. C, 2008, 112,332–337.

20 Q. Cheng, S. Debnath, E. Gregan and H. Byrne, J. Phys.Chem. C, 2010, 114, 8821–8827.

21 M. F. Islam, E. Rojas, D. M. Bergey, A. T. Johnson andA. G. Yodh, Nano Lett., 2003, 3, 269–273.

22 J. Malig, C. Romero-Nieto, N. Jux and D. M. Guldi, Adv.Mater., 2012, 24, 800–805.

23 http://worldaccount.basf.com/wa/NAFTA/Catalog/ChemicalsNAFTA/pi/BASF/Brand/pluronic.

24 E. Nativ-Roth, R. Shvartzman-Cohen, C. Bounioux,M. Florent, D. S. Zhang, I. Szleifer and R. Yerushalmi-Rozen,Macromolecules, 2007, 40, 3676–3685.

25 R. Nagarajan, R. A. Bradley and B. R. Nair, J. Chem. Phys.,2009, 131, 104906.

26 G. Zhao, D. Shao, C. Chen and X. Wang, Appl. Phys. Lett.,2011, 98, 183114.

27 Y. Talmon, Ber. Bunsen-Ges. Phys. Chem., 1996, 100, 364–372.

28 A. C. Ferrari, J. C. Meyer, V. Scardaci, C. Casiraghi, M. Lazzeri,F. Mauri, S. Piscanec, D. Jiang, K. S. Novoselov, S. Roth andA. K. Geim, Phys. Rev. Lett., 2006, 97, 187401.

29 N. K. Memon, S. D. Tse, J. F. Al-Sharab, H. Yamaguchi,A. M. B. Goncalves, B. H. Kear, Y. Jaluria, E. Y. Andrei andM. Chhowalla, Carbon, 2011, 49, 5064–5070.

30 K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang,Y. Zhang, S. V. Dubonos, I. V. Grigorieva and A. A. Firsov,Science, 2004, 306, 666–669.

31 J. C. Meyer, A. K. Geim, M. I. Katsnelson, K. S. Novoselov,T. J. Booth and S. Roth, Nature, 2007, 446, 60–63.

32 Q. Yu, L. A. Jauregui, W. Wu, R. Colby, J. Tian, Z. Su, H. Cao,Z. Liu, D. Pandey, D. Wei, T. F. Chung, P. Peng, N. P. Guisinger,E. A. Stach, J. Bao, S.-S. Pei and Y. P. Chen, Nat. Mater., 2011, 10,443–449.

33 V. C. Tung, L.-M. Chen, M. J. Allen, J. K. Wassei, K. Nelson,R. B. Kaner and Y. Yang, Nano Lett., 2009, 9, 1949–1955.

34 G. Xin, W. Hwang, N. Kim, S. M. Cho and H. Chae,Nanotechnology, 2010, 21, 405201.

35 G. Fan, H. Zhu, K. Wang, J. Wei, X. Li, Q. Shu, N. Guo andD. Wu, ACS Appl. Mater. Interfaces, 2011, 3, 721–725.

36 S. Tongay, M. Lemaitre, X. Miao, B. Gila, B. R. Appleton andA. F. Hebard, Phys. Rev. X, 2012, 2, 011002.

37 R. Har-Lavan, I. Ron, F. Thieblemont and D. Cahen,Appl. Phys. Lett., 2009, 94, 043308.

38 H. Shpaisman, R. Har-Lavan, N. Stein, O. Yaffe, R. Korobko,O. Seitz, A. Vilan and D. Cahen, Adv. Funct. Mater., 2010, 20,2181–2188.

39 G. Wang, Y. Kim, M. Choe, T.-W. Kim and T. Lee, Adv.Mater., 2011, 23, 755–760.

40 C. D. G. Caneba, V. Agrawal and M. Rao, J. Miner. Mater.Charact. Eng., 2010, 9, 165–181.

41 Q. Cheng, S. Debnath, E. Gregan and H. J. Byrne, J. Phys.Chem. C, 2010, 114, 8821–8827.

Paper PCCP

Dow

nloa

ded

by U

nive

rsity

of

Abe

rdee

n on

04

Mar

ch 2

013

Publ

ishe

d on

10

Janu

ary

2013

on

http

://pu

bs.r

sc.o

rg |

doi:1

0.10

39/C

3CP4

3205

JView Article Online