Cinzia CasiraghiSchool of Chemistry, University of Manchester

Light-Matter Interaction in Heterostructures made of

2D Crystals

!"#$%&'(#)*$!"#$%&'$

!"#$%&'$(")!*+'*##&'*+)!

,-#.'/%&0!!

12$"#(&!!*+'*##&'*+)

!

3-0/'$/)!

!"#$%&!'(")*+,#-.%&/+ 0-#12*&*+3+4!+,-5/6#(/+

-#$#&+/1*,6-%/,%15+

!"#$%&'(()*+%&!,%&-.*/%

!"#$%&'%(""#$%#&'()*!+&$$(#$!

!",-!+)''.#/!

"$)'#0'(1#!2&$$(#$!

340.#&'()*!

!"#$%&&'($)*+,-*,$%&&.$!)*"+,-.-,/01"

$$

5#'#$)/'$40'4$#/!

!6(&%)*78.(9#!0&$+)*/!:;6<!!"/$%&&0($223"+,-.-,/01""!"#$%&'&()*+,(

12,-3+4+*53+6-($)375+-($8,4,*39$86:+-;($)<:,7=533+*,9($,3*$*243"+,-.-,/01"

etc

)*+,-*,$%&>%($554"+,-.-,/01$$

!"#$%0-1#".+%2340%05"-67#8"9%

=$&>?#*#!3&*)$(++)*/!:@6<!?53<7,$@A,B+937C$%&>'($D@)$?5-6$%&>'($,3*$

2-6."-9:.+%;##<-=")+%&-*=:)65)"%

A,BA!

OVERVIEW

The Family of 2D-Crystals

PRODUCTION

Novoselov et al, Nature 2012

will probably be limited to niche applications.High-frequency transistorsbased on SiC-grown graphene34 may well find applications within adecade when the existing technology, based on III–V materials (such asInGaAs, GaN, and so on) reaches its limit at about 1 THz. The short gatetransistors that are currently widely used make even the 20-mm sizedomains (currently achieved in graphene grown on SiC) suitable for suchapplications. Another very attractive, though niche, application of thistype of graphene is inmetrological resistance standards35, where samplesof graphene grown on SiC have already been demonstrated to deliverhigher resistance accuracy at higher temperatures thando conventionallyused GaAs heterostructures.Apart from the high temperature required for growth, which cur-

rently seems to be an insurmountable problem, the other issues thatneed to be addressed in the next decade are the elimination of terraces,the growth of the second or third layers at the edges of the terraces(which also strongly contribute to carrier scattering), an increase inthe size of the crystallites and control of unintentional doping fromthe substrate and buffer layers.

Other growth methodsAlthough there are a number of other growthmethods, it is unlikely thatthey will become commercially viable in the next decade. Nevertheless,some of these methods have certain advantages and should beresearched further. Surface-assisted coupling of molecular monomerprecursors into linear polyphenylenes with subsequent cyclodehydro-genation is an exciting way to create high-quality graphene nanoribbonsand evenmore complex structures (like T- and Y-shaped connections)36

using a chemistry-driven bottom-up approach. Molecular beam epitaxyhas been used to grow chemically pure graphene37, but it is unlikely to beused on a large scale because of its much higher cost thanCVDmethods.

Laser ablation is a potentially interesting growth technique allowing thedeposition of graphene nanoplatelets on arbitrary surfaces38. This rela-tively expensive method is in direct competition with the spray-coatingof chemically exfoliated graphene, so it is unlikely to be widely used.

Graphene electronicsIt is unlikely that graphene will make it into high-performance inte-grated logic circuits as a planar channel material within the next decadebecause of the absence of a bandgap. However, many other, less strin-gent, graphene electronic applications are being developed, using theavailable (probably not ideal in terms of quality) material. Figure 2 andTable 2 list some of the possible applications and the time that it maytake for graphene-based prototypes to be demonstrated.

Flexible electronicsTransparent conductive coatings are widely used in electronic productssuch as touch screen displays, e-paper (electronic paper) and organiclight-emitting diodes (OLEDs) and require a low sheet resistance withhigh transmittance (of over 90%) depending on the specific application.Graphene meets the electrical and optical requirements (sheet resistancereaching 30V per square of 2D area in highly doped samples) and anexcellent transmittance of 97.7%per layer8, although the traditionally usedindium tin oxide (ITO) still demonstrates slightly better characteristics.However, considering that the quality of graphene improves every year(already making the difference in performance marginal), while ITO willbecome more expensive and ITO deposition is already expensive, gra-phene has a chance of securing a good fraction of the market. Graphenealso has outstandingmechanical flexibility and chemical durability—veryimportant characteristics for flexible electronic devices29, in which ITOusually fails.The requirements of electrical properties (for example, sheet resist-

ance) for each electrode type differ from application to application.Depending on the production methods, various grades of transparentconductive coating could be produced from graphene. Thus, electrodesfor touch screens (although requiring an expensive CVD method ofproduction) tolerate a relatively high sheet resistance (50–300V persquare) for a transmittance of 90%. The advantage of graphene electro-des in touch panels is that graphene’s endurance far exceeds that of anyother available candidate at the moment. Moreover, the fracture strainof graphene is ten times higher5 than that of ITO, meaning that it couldalso successfully be applied to bendable and rollable devices.Rollable e-paper is a very appealing electronic product. It requires a

bending radius of 5–10mm, which is easily achievable by a grapheneelectrode. In addition, graphene’s uniform absorption across the visiblespectrum8 is beneficial for colour e-papers. However, the contact resist-ance between the graphene electrode and the metal line of the drivingcircuitry is still a problem. A working prototype is expected by 2015, butthe manufacturing cost needs to decrease before it will appear on themarket.OLED devices have become an attractive technology and the first

(non-graphene) products are expected on the market by 2013. Besides

Table 1 | Properties of graphene obtained by different methodsMethod Crystallite size (mm) Sample size (mm) Charge carrier mobility (at ambient

temperature) (cm2V21 s21)Applications

Mechanical exfoliation .1,000 .1 .2 3105 and .106 (at lowtemperature)

Research

Chemical exfoliation #0.1 Infinite as a layer ofoverlapping flakes

100 (for a layer of overlapping flakes) Coatings, paint/ink, composites, transparentconductive layers, energy storage, bioapplications

Chemical exfoliation viagraphene oxide

,100 Infinite as a layer ofoverlapping flakes

1 (for a layer of overlapping flakes) Coatings, paint/ink, composites, transparentconductive layers, energy storage, bioapplications

CVD 1,000 ,1,000 10,000 Photonics, nanoelectronics, transparentconductive layers, sensors, bioapplications

SiC 50 100 10,000 High-frequency transistors and other electronicdevices

Price (for mass production)

Qua

lity

(coating, composites,inks, energy storage,bio, transparent conductive layers)

Liquid-phase exfoliation

(coating, bio, transparent conductive layers, electronics, photonics)

SiC

Mechanical exfoliation(research,

prototyping)

(electronics,RF transistors)

(nanoelectronics)

Molecularassembly

CVD

Figure 1 | There are several methods of mass-production of graphene,which allow a wide choice in terms of size, quality and price for anyparticular application.

RESEARCH REVIEW

1 9 4 | N A T U R E | V O L 4 9 0 | 1 1 O C T O B E R 2 0 1 2

Macmillan Publishers Limited. All rights reserved©2012

• LATERAL SIZE and COMPOSITION strongly depends on the experimental conditions.

NEVER 100% single-layer

• Exfoliation of their bulk counterparts via chemical wet dispersion followed by ultra-sonication.

• Low-cost and mass scalable

• Able to produce high QUALITY graphene (i.e. Oxygen free)

Works only with ORGANIC SOLVENTSHernandez et al, Nature Nanotech 2008

LIQUID-PHASE EXFOLIATION

NATUREMATERIALS DOI: 10.1038/NMAT3944 ARTICLES

c

500 nm'

Rotor/stator

500 nm

g

Rotor Stator

200 nm

1,000 2,000 3,000

D'

G

D

Inte

nsity

Raman shift (cm−1)

2D

D G

2D

284 288 292

C OC N

Inte

nsity

Binding energy (eV)0 5 10 15 20 25

0

4

8

12

16N high(NMP)

N high(NMP)

Num

ber o

f obj

ects

Flake thickness, N

N high(NaC)

C i lo

wD

low

N lo

wt h

igh

V lo

wD

high

V hi

ght l

owC i

hig

hN

high

0.1

1.0

10

!NG" (NMP, Raman)

ID /IG (NMP)!NG", I D

/IG

, C

C co

nten

t !NG" (NaC, AFM)

2,600 2,800

1 2 3(103 cm−1)

e f h k l

a b d i j

C HC C

C C fraction (NMP)C C fraction (NMP)

Figure 1 | Production of graphene by shear mixing. a, A Silverson model L5M high-shear mixer with mixing head in a 5 l beaker of graphene dispersion.b,c, Close-up view of a D=32mmmixing head (b) and a D= 16mmmixing head with rotor (left) separated from stator (c). d, Graphene–NMP dispersionsproduced by shear exfoliation. e, Wide-field TEM image of SEG nanosheets (after centrifugation). f–h TEM images of individual nanosheets (f), amultilayer (g, bottom left) and monolayer (g, right) as evidenced by its electron di�raction pattern (g, inset) and a monolayer (h; imaged by high-resolutionscanning TEM). i, Histogram of nanosheet thickness as measured by AFM on a surfactant-exfoliated sample. The presence of monolayers was confirmedby Raman characterization (inset). j,k, XPS (j) and Raman (k) spectra (NMP-exfoliated samples) measured on thin films. AFM, Raman and XPS analysiswere performed on dispersions made using both high and low values of a given processing parameter while keeping others constant (SupplementarySection 3). The dispersion type is indicated in the panel. l, Information extracted from Raman, XPS and flake thickness data plotted versus dispersion type.Blue—mean flake thickness �NG� measured by AFM for a surfactant-exfoliated dispersion and inferred from Raman for a NMP-exfoliated dispersion(Supplementary Section 3); black, fraction of XPS spectrum associated with C–C bonds; red, ratio of intensities of Raman D and G bands. The error barsrepresent the standard error associated with multiple measurements (∼100 for AFM and ∼10 for Raman). Unless noted otherwise, all data are reported forNMP dispersions.

Shear exfoliationShown in Fig. 1a is a Silversonmodel L5Mmixer that generates highshear using a closely spaced (∼100 µm) rotor/stator combination(Fig. 1b and Supplementary Section 1), and is available with a rangeof rotor diameters (Fig. 1b,c). Initial trials involved the shear-mixingof graphite both in the solvent N -methyl-2-pyrrolidone (NMP)and in aqueous surfactant solutions10 (sodium cholate, NaC),resulting in large-volume suspensions (Fig. 1d and SupplementarySections 1–3, initial processing parameters: rotor diameter,D=32mm; initial graphite concentration, Ci =50mgml−1;mixing time, t=20min; liquid volume, V = 4.5 l; rotor speed,N = 4, 500 r.p.m.). After centrifugation these suspensions containlarge quantities of high-quality graphene nanosheets, includingsome monolayers (Fig. 1e–h).

To test the effect of mixing parameters on the quality of shearexfoliated graphene (SEG), we prepared a range of dispersionsusing both NMP and water/NaC, keeping all but one of the mixingparameters constant (as above) but maximizing and minimizingthe remaining one (Supplementary Section 3). These dispersionswere studied by transmission electron microscopy (TEM) andatomic force microscopy (AFM) to measure the flake size andthickness before filtering to form ∼100-nm-thick films that werecharacterized by X-ray photoelectron spectroscopy (XPS) andRaman spectroscopy (Fig. 1i–k and Supplementary Section 3). TEMmeasurements showed nanosheet sizes in the 300–800 nm rangeand AFM of surfactant-exfoliated samples gave typical thicknesses,NG, of less than 10 monolayers per nanosheet (�NG�∼5–8)(Supplementary Section 3). The presence of monolayers wasconfirmed by Raman spectroscopy (Fig. 1i, inset). For filmsprepared from NMP-exfoliated graphene, XPS showed no evidenceof oxidation and Raman spectroscopy reproducibly showed 2D

bands (Fig. 1k, inset) consistent with NG between 4 and 7 anda relatively weak, narrow D band (Supplementary Section 3).The Raman D/G band intensity ratio is proportional to inversenanosheet length22 and the D/D� band intensity ratio23 is ∼4(Supplementary Section 3). Taken together, these data show the Dband to be dominated by nanosheet edge contributions and confirmthat no basal-plane defects are introduced during exfoliation22,23

(Supplementary Section 3). As shown in Fig. 1l, these propertieswere relatively invariant with mixing parameters indicating thatwell-exfoliated, non-oxidized, defect-free graphene can be producedusing a broad range of mixing conditions. We note that these flakesare virtually indistinguishable from those produced by sonicationboth in terms of size and quality10.

The exfoliation mechanismConsidering the exfoliation mechanism, our initial expectation wasthat localized, turbulent, highly dissipative regions were responsiblefor exfoliation24,25. However, we found turbulent energy dissipationto be unnecessary. Figure 2a maps the combinations of N andD that result in exfoliation: graphene is produced not only forturbulent, high-Reynolds-number (Re) scenarios, but also forcombinations that give ReMixer =ND2ρ/η<104, where turbulence isnot fully developed26 (ρ and η are the liquid density and viscosityrespectively). To determine whether graphene could be producedin the complete absence of turbulence, that is, under high-shearlaminar flow, we shear-mixed graphite and NMP in a Couette(a concentric cylinder-based rheological cell, radius R = 14mm,thickness d = 0.1mm, rotation frequency of inner wall ω: shearrate γ ≈ Rω/d , Supplementary Section 7). TEM confirmed thatgraphene was produced, with concentration increasing with time ast 0.69 (Fig. 2b). Interestingly, we found graphene in the Couette only

NATUREMATERIALS | VOL 13 | JUNE 2014 | www.nature.com/naturematerials 625

© 2014 Macmillan Publishers Limited. All rights reserved

EXFOLIATION with SURFACTANTS

Lotya et al, JACS 2009

Surfactant is not easy to remove, yield of monolayer is < 10%

EXFOLIATION IN WATER

Parviz et al, ACS Nano, 2012 Yang et al, Carbon, 2012

OUR WATER-BASED INKS

Up to 60-70% high-quality single-layers

Reduced amount of stabilizer

EXFOLIATION with STABILIZERS

XPSRaman spectroscopy

PS2 PS4

Schlierf et al, Nanoscale 2013

OUR WATER-BASED INKS

In collaboration with Vincenzo Palermo, CNRS Bologna (Italy) and David Bejonne, University of Mons (Belgium)

Funded by the European Science Foundation

!"#$%&'&()'*+(#',(-#.)'#/&+(

!"#$%&'&()'*+(,"0$12#+/(0'($-#+/)2(

$-#+/)2( 3#.)'#/&($"0,42&,(56(7)-/"#/)0'8(

7)-/&"(

97(/%&(-#.)'#/&()+(/%)2*(&'04:%;()/(2#'(5&("&.0<&,(7"0.(/%&(7)-/&"8(=4/()/()+(<&"6(5")//-&(

UK Patent Application No 1401721.4

OUR WATER-BASED INKS

! ""!

0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6-30

-20

-10

0

Py-4SO3/h-BN Py-2SO3/h-BN Py-1SO3/h-BN

Distance (nm)

PM

F (k

cal/m

ol)

F igure S17 Potential of mean Force (PMF) curves of Py-1SO3, Py-2SO3, and Py-4SO3

adsorbtion on h-BN.

Figure S18 shows the coulombic and van der Waals contributions to the interaction energy of the

pyrene-core- for each of the pyrene derivatives at the global minima (fully adsorbed states). One

can see that with increasing number of SO3 groups the pyrene- interaction

with the solvent increases, which is due to the induced charge on the pyrene-core making it less

hydrophobic. The trend, however, is different upon going from Py-1SO3 to Py-2SO3, which

shows that the introduction of -OH groups neutralizes the induction of charge from the SO3

groups. The same qualitative behavior has been observed in the exfoliation of graphite (see Fig.

9b in Ref. 3b).

Extended to other 2D crystals

Exfoliation dominated by interaction molecule- solvent

OTHER 2D CRYSTALS - TEM

FABRICATION

drop castingvacuum filtration

Withers et al, Nano Lett 2014

P≅ 0.5 mW

Photoresponsivity≅0.1 mA/Winset to left panel), demonstrating that such structures can beindeed used as photovoltaic devices.We recorded zero-bias photocurrent as a function of the

position of the laser spot (less than 1 !m in diameter) on thedevice by using a 100! microscope objective. Zero-biasphotocurrent maps, taken at di!erent back gate voltages,demonstrate that the photocurrent is produced only in theregions where all three layers (BGr, TMDC, and TGr) overlap.We notice that the edges of the sample provide a slightly o!setgate dependence of the photocurrent, which we attribute to thee!ect of environmental doping of the TMDC not covered byTGr. Similar to the case of the transistor, the back gate voltagecontrols the value and the direction of the electric "eld acrossWS2 and thus the magnitude and the polarity of thephotocurrent (Figure 4). For the largest electric "eld acrossWS2 (at Vg = 60 V) used in these measurements, we achievedphotoresponsivity values of !0.1 mA/W.The e#ciency could still be increased by using larger $akes;

electron and hole scattering and localization on the impuritiesand edges are reduced, which, in turn, would reduce thecontribution of the recombination mechanisms.32 We wouldlike to stress that our devices do not require exactly monolayerTMDC to be used, which simpli"es the procedure even further.Using even thicker $akes (by reducing the sonication time)means that those would also be larger laterally and allow moree#cient charge transfer between the layers, thus allowing formore e#cient e"h separation. Also, the use of thick TMDC$akes ensures that the band structure of TMDC used has aindirect band gap,33,34 thus reducing the probability ofrecombination. Note that the photoabsorption for TMDC(per layer) practically does not change with the number oflayers.13

Although the photoresponsivity of our devices is signi"cantlysmaller than that obtained in current state of the artphotovoltaic devices35 or in similar heterostructures based onmonocrystalline WS2,

13 the advantage of our structures is thatthey can be produced by a variety of low-cost and scalablemethods and are compatible with $exible substrates (note thatthe use of CVD graphene as a back electrode for solar cellapplication is compatible both with the $exible substrates andwith this low cost method). To this end we fabricated PET/BGr/WS2/TGr heterostructures on a $exible PET substrate(thickness 0.2 mm) (Figure 5A). We tested two di!erentmethods for sample fabrication: BGr and WS2 layers wereproduced by either drop-casting or vacuum "ltering (withsubsequent wet transferring) of the respective LPE dispersion.Both layers were shaped into strips by mechanical removal ofthe unnecessary material (Figure 5A). We used CVD grapheneas TGr to achieve maximum optical transparency. A four-pointbending rig was utilized to apply uniaxial strain to theheterostructure (Figure 5D).As in the previous experiment we scanned a laser across the

sample while simultaneously recording the photocurrent(Figure 5B,C). The photocurrent is only observed whenilluminating the area where all three layers (BGr, WS2, andTGr) overlap. After bending, some local variation in thephotocurrent was detected. However, the overall pattern(Figure 5B, C), the integral value of the photocurrent (Figure5E), and the overall resistance of the device (Figure 5F) remainpractically independent of the strain, demonstrating thepossibility to use such heterostructures for $exible electronics.Finally, we demonstrate a di!erent type of heterostructure

where LPE h-BN is used as a gate dielectric. The dielectric

properties of h-BN,19,28 added to its excellent chemical andthermal stability, mechanical and thermal properties,1 make h-BN thin "lms a promising dielectric alternative in the nextgeneration of nanodevices.36 Here we tested Si/SiO2/BGr/h-BN/Au devices, where LPE h-BN (prepared through "ltering ofa h-BN suspension, with subsequent transfer of the h-BN paperfrom the "lter to the device) served as transparent dielectricbetween the channel (BGr, CVD graphene) and the gate (Au)(Figure 6A, inset). We also used mechanically exfoliated single-and few-layer CVD and LPE graphene as a top electrode.The resistivity of the BGr channel as a function of top gate

voltage Vgt is presented in Figure 6A. The contour plot of theresistivity as a function of Vg and Vgt demonstrates the usualresistivity maximum shifting across a diagonal of the plot(Figure 6B, the dashed line). The slope of the line allows us toestablish the ratio of the capacitances to Si and top gate (herewe ignore the "nite compressibility of 2D electron gas ingraphene). Knowing the thickness of h-BN from the AFMstudy (600 nm for this particular sample) allows us to estimatethe e!ective dielectric constant of LPE h-BN to be !1.5. Thesigni"cant deviation from the bulk value (!4, as established inrecent tunnelling experiments37) is due to lose packing of h-BNlaminates. This low value of the dielectric constant of the LPEh-BN could be an advantageous property when considering itsincorporation in densely packed electronic elements, where lossneeds to be kept to a minimum. Indeed air gaps in conventionalinsulators are deliberately induced to reduce the overalle!ective dielectric constant.38,39 Knowing the capacitance tothe top gate allows us to estimate the mobility of the BGr to beof the order of 3 ! 103 cm2/V·s, which is typical of CVDgraphene.4 This clearly indicates that deposition of LPE h-BNdoes not deteriorate the properties of graphene. We have alsotested the breakdown voltage for our LPE h-BN (SupportingInformation), which turned out to be 0.25 V/nm. This is

Figure 5. (A) Optical micrograph of a LPE PET/BGr/WS2/TGrheterostructure. The yellow dotted lines indicate the boundaries ofLPE BGr; the green dotted lines CVD TGr; the red square shows thearea investigated by photocurrent mapping (size 70 !m ! 70 !m).The brownish stripe which covers the BGr is 60 nm LPE WS2.Photocurrent maps (70 !m ! 70 !m) taken at an incident power of190 !W and energy of 1.96 eV at two di!erent curvatures: 0 mm"1 (B,corresponds to zero strain) and 0.15 mm"1 (C, corresponds to 1.5%strain). (D) Schematic representation of our bending setup. (E)Average photocurrent obtained from the photocurrent maps as afunction of the applied strain. (F) I"Vb characteristics with (red) andwithout (blue) illumination for the strained (solid curves) andunstrained (dashed curves) cases. The illumination (power 190 !W)was focused into a !1 !m2 spot.

Nano Letters Letter

dx.doi.org/10.1021/nl501355j | Nano Lett. 2014, 14, 3987"39923990

HETEROSTRUCTURES

inset to left panel), demonstrating that such structures can beindeed used as photovoltaic devices.We recorded zero-bias photocurrent as a function of the

position of the laser spot (less than 1 !m in diameter) on thedevice by using a 100! microscope objective. Zero-biasphotocurrent maps, taken at di!erent back gate voltages,demonstrate that the photocurrent is produced only in theregions where all three layers (BGr, TMDC, and TGr) overlap.We notice that the edges of the sample provide a slightly o!setgate dependence of the photocurrent, which we attribute to thee!ect of environmental doping of the TMDC not covered byTGr. Similar to the case of the transistor, the back gate voltagecontrols the value and the direction of the electric "eld acrossWS2 and thus the magnitude and the polarity of thephotocurrent (Figure 4). For the largest electric "eld acrossWS2 (at Vg = 60 V) used in these measurements, we achievedphotoresponsivity values of !0.1 mA/W.The e#ciency could still be increased by using larger $akes;

electron and hole scattering and localization on the impuritiesand edges are reduced, which, in turn, would reduce thecontribution of the recombination mechanisms.32 We wouldlike to stress that our devices do not require exactly monolayerTMDC to be used, which simpli"es the procedure even further.Using even thicker $akes (by reducing the sonication time)means that those would also be larger laterally and allow moree#cient charge transfer between the layers, thus allowing formore e#cient e"h separation. Also, the use of thick TMDC$akes ensures that the band structure of TMDC used has aindirect band gap,33,34 thus reducing the probability ofrecombination. Note that the photoabsorption for TMDC(per layer) practically does not change with the number oflayers.13

Although the photoresponsivity of our devices is signi"cantlysmaller than that obtained in current state of the artphotovoltaic devices35 or in similar heterostructures based onmonocrystalline WS2,

13 the advantage of our structures is thatthey can be produced by a variety of low-cost and scalablemethods and are compatible with $exible substrates (note thatthe use of CVD graphene as a back electrode for solar cellapplication is compatible both with the $exible substrates andwith this low cost method). To this end we fabricated PET/BGr/WS2/TGr heterostructures on a $exible PET substrate(thickness 0.2 mm) (Figure 5A). We tested two di!erentmethods for sample fabrication: BGr and WS2 layers wereproduced by either drop-casting or vacuum "ltering (withsubsequent wet transferring) of the respective LPE dispersion.Both layers were shaped into strips by mechanical removal ofthe unnecessary material (Figure 5A). We used CVD grapheneas TGr to achieve maximum optical transparency. A four-pointbending rig was utilized to apply uniaxial strain to theheterostructure (Figure 5D).As in the previous experiment we scanned a laser across the

sample while simultaneously recording the photocurrent(Figure 5B,C). The photocurrent is only observed whenilluminating the area where all three layers (BGr, WS2, andTGr) overlap. After bending, some local variation in thephotocurrent was detected. However, the overall pattern(Figure 5B, C), the integral value of the photocurrent (Figure5E), and the overall resistance of the device (Figure 5F) remainpractically independent of the strain, demonstrating thepossibility to use such heterostructures for $exible electronics.Finally, we demonstrate a di!erent type of heterostructure

where LPE h-BN is used as a gate dielectric. The dielectric

properties of h-BN,19,28 added to its excellent chemical andthermal stability, mechanical and thermal properties,1 make h-BN thin "lms a promising dielectric alternative in the nextgeneration of nanodevices.36 Here we tested Si/SiO2/BGr/h-BN/Au devices, where LPE h-BN (prepared through "ltering ofa h-BN suspension, with subsequent transfer of the h-BN paperfrom the "lter to the device) served as transparent dielectricbetween the channel (BGr, CVD graphene) and the gate (Au)(Figure 6A, inset). We also used mechanically exfoliated single-and few-layer CVD and LPE graphene as a top electrode.The resistivity of the BGr channel as a function of top gate

voltage Vgt is presented in Figure 6A. The contour plot of theresistivity as a function of Vg and Vgt demonstrates the usualresistivity maximum shifting across a diagonal of the plot(Figure 6B, the dashed line). The slope of the line allows us toestablish the ratio of the capacitances to Si and top gate (herewe ignore the "nite compressibility of 2D electron gas ingraphene). Knowing the thickness of h-BN from the AFMstudy (600 nm for this particular sample) allows us to estimatethe e!ective dielectric constant of LPE h-BN to be !1.5. Thesigni"cant deviation from the bulk value (!4, as established inrecent tunnelling experiments37) is due to lose packing of h-BNlaminates. This low value of the dielectric constant of the LPEh-BN could be an advantageous property when considering itsincorporation in densely packed electronic elements, where lossneeds to be kept to a minimum. Indeed air gaps in conventionalinsulators are deliberately induced to reduce the overalle!ective dielectric constant.38,39 Knowing the capacitance tothe top gate allows us to estimate the mobility of the BGr to beof the order of 3 ! 103 cm2/V·s, which is typical of CVDgraphene.4 This clearly indicates that deposition of LPE h-BNdoes not deteriorate the properties of graphene. We have alsotested the breakdown voltage for our LPE h-BN (SupportingInformation), which turned out to be 0.25 V/nm. This is

Figure 5. (A) Optical micrograph of a LPE PET/BGr/WS2/TGrheterostructure. The yellow dotted lines indicate the boundaries ofLPE BGr; the green dotted lines CVD TGr; the red square shows thearea investigated by photocurrent mapping (size 70 !m ! 70 !m).The brownish stripe which covers the BGr is 60 nm LPE WS2.Photocurrent maps (70 !m ! 70 !m) taken at an incident power of190 !W and energy of 1.96 eV at two di!erent curvatures: 0 mm"1 (B,corresponds to zero strain) and 0.15 mm"1 (C, corresponds to 1.5%strain). (D) Schematic representation of our bending setup. (E)Average photocurrent obtained from the photocurrent maps as afunction of the applied strain. (F) I"Vb characteristics with (red) andwithout (blue) illumination for the strained (solid curves) andunstrained (dashed curves) cases. The illumination (power 190 !W)was focused into a !1 !m2 spot.

Nano Letters Letter

dx.doi.org/10.1021/nl501355j | Nano Lett. 2014, 14, 3987"39923990

RAPID GROW: $5 billion in 2012 -> $35 billion in 2020 -> $300 billion by 2030

smart packaging

active clothingpharmacy

PRINTED ELECTRONICS

Finn, D. J.; et al. J. Mater. Chem. C, 2, 925 (2014)

TORRISI ET AL . VOL. 6 ’ NO. 4 ’ 2992–3006 ’ 2012

www.acsnano.org

2998

indicating that the O2-treated substrate SE is modi!edfollowing HMDS treatment. !LV was measured as!73 mJ m!2 in ref 120 for DI water, whereas !SV "116.5mJm!2 and" 40mJm!2 were reported for O2

121

and HMDS-treated122 Si/SiO2 substrates. Consequently,!SL " 43.9 mJ m!2 and!9.1 mJ m!2 for O2 and HMDS-treated Si/SiO2. A higher !SL implies a higher SE.123

Indeed, our !SL correspond to SEs !73.9 and !39.1 mJm!2 for O2 and HMDS-treated Si/SiO2. A small "c resultsin the rapid drop spreading on the substrate,86 as forO2-treated SiO2. HMDS provides higher "c, since itlowers !SL (thus the substrate SE), therefore reducingthe wettability.87,124

When inkjet printing stripes, the interdrop (i.e.,center-to-center) distance is an important para-meter.125 When the distance is large, individual dropsare deposited.76,78,125 As the interdrop distance de-creases, these merge into a line.125 Thus, in order toobtain a continuous line, we need an interdrop dis-tance smaller than the drop diameter.125 On the otherhand, refs 76 and 126 reported that a very smallinterdrop distance can result in particle aggregationon the substrate, thus a nonuniform stripe (i.e., irregularedges). We select an interdrop distance suitableto have continuous lines, avoiding at the same timenonuniformities and irregular edges.

Shown in Figure 8a,b,c are optical images of printedstripes on pristine, O2-plasma-treated and HMDS-treated Si/SiO2, whereas Figure 8d,e,f plot the respec-tive atomic force microscope (AFM) topographies. Thestripe in Figure 8a is !100!110 #m wide, havingan average thickness of !70 nm and an irregular"ake distribution, with aggregation of "akes. That inFigure 8b is wider (!130!140 #m), with aggregates atthe edges and an average thickness of !55 nm. The

stripe in Figure 8c has a more uniform and regulardistribution of "akes, having a !85!90 #m width and!90 nm average thickness. The width narrows goingfrom the O2-plasma-treated to the HMDS-treatedSi/SiO2, due to the SE decrease. Figure 8d,e showstripes with voids and irregular "ake distribution, withRz " 30!40 nm. Figure 8f presents a more homoge-neous network with Rz" 15 nm. Thus, Rz is lower when"c is higher, because the poor wettability of drops withhigher "c reduces the stripe width (as shown inFigure 8a,b,c), con!ning the "akes onto a smaller area.The uniformity of stripes printed on the HMDS-treatedsubstrate corroborates the above considerations onthe SE changes. In fact, the presence of silane groupsin the molecular structure of HMDS91 acts as promoterof metallic particle adhesion to the substrate.91,127

Analogously, HMDS may promote the adhesion ofgraphene "akes to the substrate, thus favoring theformation of a regular network.

Figure 9a compares a typical Raman spectrum of a"ake in the ink, with a measurement on the !rst stripeand on a stripe 90 nm thick, after 30 printing repeti-tions. Figure 9b,c,d,e,f,g and Figure 10 compare thePos(2D), FWHM(2D), and Disp(G) distributions. Thedata show that the !rst stripe has very similar char-acteristics to the ink, as expected. However, the spectraafter 90 repetitions show Pos(2D) and FWHM(2D) dis-tributions more typical of a multilayer sample, havinglost any direct signature of SLG. Note, however, that the2D peak shape, even for the 90 nm stripe, remainsdistinctly di#erent from that of graphite. A similaraggregation of "akes was previously observed forthick !lms derived from graphene dispersions.56 In allcases Disp(G) remains similar and very low, again show-ing the lack of largeamounts of defectswithin the"akes.

Figure 8. Optical micrograph of inkjet-printed stripes on (a) pristine, (b) O2-treated and (c) HMDS-treated substrates. (d, e, f)AFM images of a, b, c, respectively.

ARTIC

LE

Torrisi et al, ACS Nano, 6, 2992 (2012)

50 µm

INKJET PRINTING

- Coffee ring effect- Post-processing- Residuals

NMP-BASED INKS

- Drop unstable- Inks unstable in cartdrige as made

WATER-BASED INKS

modified NMP

ONE POT method: Stable Graphene inks printed on paper, glass, plastic, silicon, etc

UK Patent Application No 1401721.4

4035302520151051

On PELTM

• Successful printing of BN, MoS2, WS2 inkson PET on Silicon

6010 50 40 30 20 10 5 1

30 20 10

30 20 10

30 20

OTHER 2D-INKS

100 µm

RAMAN SPECTROSCOPY OF GRAPHENE

❒ Identification Ferrari, Casiraghi et al PRL 2006

❒ Doping Pisana, Casiraghi et al. Nature Mat. 2006 Casiraghi et al. APL 2007

❒ DisorderCasiraghi RRL-PSS 2009Eckmann et al, NL 2012Otto et al, Nanoresearch 2014Kim et al, ACS Nano 2012

❒ Chemical DerivativesEckmann et al, NL 2012; PRB 2013Felten et al, Nanotechnology 2013; Small 2012

❒ Heterostructures, SuperlatticesEckmann et al, NL 2013Woods et al, Nature Physics 2014Zhou et al, ACS Nano 2014

❒ StrainZabel et al NL 2012

❒ EdgesCasiraghi et al., NL 2009

❒ e-p, e-e interactionsPiscancec et al PRL 2004Basko, PRB 2008Casiraghi, PRB 2009Klar et al, PRB 2013

... AND MORE!

DIMER

TRIMER

RIBBON

Synthesis of structurally well-defined and liquid-phase-processable graphene nanoribbonsAkimitsu Narita1, Xinliang Feng1*, Yenny Hernandez1, Søren A. Jensen1,2, Mischa Bonn1,Huafeng Yang3, Ivan A. Verzhbitskiy4, Cinzia Casiraghi3,4, Michael Ryan Hansen1,5, Amelie H. R. Koch1,George Fytas1,6, Oleksandr Ivasenko7, Bing Li7, Kunal S. Mali7, Tatyana Balandina7,Sankarapillai Mahesh7, Steven De Feyter7 and Klaus Mullen1*

The properties of graphene nanoribbons (GNRs) make them good candidates for next-generation electronic materials.Whereas ‘top-down’ methods, such as the lithographical patterning of graphene and the unzipping of carbon nanotubes,give mixtures of different GNRs, structurally well-defined GNRs can be made using a ‘bottom-up’ organic synthesisapproach through solution-mediated or surface-assisted cyclodehydrogenation reactions. Specifically, non-planarpolyphenylene precursors were first ‘built up’ from small molecules, and then ‘graphitized’ and ‘planarized’ to yield GNRs.However, fabrication of processable and longitudinally well-extended GNRs has remained a major challenge. Here wereport a bottom-up solution synthesis of long (>200 nm) liquid-phase-processable GNRs with a well-defined structure anda large optical bandgap of 1.88 eV. Self-assembled monolayers of GNRs can be observed by scanning probe microscopy,and non-contact time-resolved terahertz conductivity measurements reveal excellent charge-carrier mobility withinindividual GNRs. Such structurally well-defined GNRs may prove useful for fundamental studies of graphenenanostructures, as well as the development of GNR-based nanoelectronics.

Graphene nanoribbons (GNRs), defined as nanometre-widestrips of graphene, are attracting increasing attention ashighly promising candidates for next-generation semicon-

ductor materials1–4. Quantum confinement effects impart GNRswith semiconducting properties, namely with a finite bandgapthat critically depends on the ribbon width and its edge structure1,3.Fabrication of GNRs has been carried out primarily by ‘top-down’approaches, such as lithographical patterning of graphene5,6 andunzipping of carbon nanotubes7,8, to reveal their semiconductingnature and excellent transport properties1. However, thesemethods are generally limited by low yields and lack of structuralprecision, which leads to GNRs with uncontrolled edge structures.

In contrast, a ‘bottom-up’ chemical synthetic approach based onsolution-mediated9–13 or surface-assisted14 cyclodehydrogenation,namely ‘graphitization’ and ‘planarization’, of tailor-made three-dimensional (3D) polyphenylene precursors offers an appealingstrategy for making structurally well-defined and homogeneousGNRs. The polyphenylene precursors are built up from small mol-ecules, and thus their structures can be tailored within the capabilitiesof modern synthetic chemistry15. On the one hand, GNRs (.30 nm)produced by solution-mediated methods are precluded from unam-biguous structural characterization, that is, microscopic visualization,because of their limited processability9,12. On the other hand, GNRsproduced by the surface-assisted protocol are characterized to beatomically precise using scanning tunnelling microscopy (STM)14.Nevertheless, this method can only provide a limited amount ofGNR material, which is further bound to a metal surface and soimpedes wider applications in electronic devices.

To date, there is no report on the large-scale fabrication of long(.100 nm) and processable GNRs with high structural definition.Here we demonstrate an efficient bottom-up solution synthesis oflongitudinally well-extended (.200 nm) GNRs with excellentliquid-phase processability based on the cyclodehydrogenation ofsemirigid polyphenylene precursors of high molecular weight(Fig. 1a). The narrow GNRs had an optical bandgap of 1.88 eV,and their structure was supported by infrared, Raman, ultraviolet–visible (UV-vis) absorption and nuclear magnetic resonance(NMR) spectroscopies. Scanning probe microscopy (SPM) analysisof GNRs deposited from dispersions on graphite substrates demon-strates an exquisitely ordered self-assembled monolayer of GNRs,which further corroborates their well-defined structure. Moreover,such GNRs offer the chance for the first liquid-phase investigationof their electronic properties by employing non-contact, time-resolved terahertz (THz) conductivity measurements16 to revealexcellent intramolecular charge-carrier mobility of the GNRs.

Results and discussionSolution synthesis of liquid-phase-processable GNRs with highlongitudinal extension. First, a novel type of polyphenyleneprecursor, 2, with extremely high molecular weight was synthesizedby employing Diels–Alder polymerization (Fig. 1a)17,18. Themonomer building block 1, which consists of a cyclopentadienoneas the conjugated diene and an ethynyl group as the dienophile,functions as a precursor for the AB-type Diels–Alderpolymerization to give 2 in high yield (Supplementary Methods).Dodecyl chains were installed on the periphery of 2 to enhance

1Max Planck Institute for Polymer Research, Ackermannweg 10, D-55128 Mainz, Germany, 2FOM Institute AMOLF, Science Park 104, 1098 XG Amsterdam,The Netherlands, 3School of Chemistry and Photon Science Institute, Manchester University, Oxford Road, Manchester, M139PL, UK, 4Department ofPhysics, Free University Berlin, Arnimalle 14, 14195 Berlin, Germany, 5Interdisciplinary Nanoscience Center (iNANO) and Department of Chemistry, AarhusUniversity, Gustav Wieds Vej 14, DK-8000 Aarhus C, Denmark, 6Department of Materials Science, University of Crete and FORTH, Heraklion,Greece, 7Division of Molecular Imaging and Photonics, Department of Chemistry, KU Leuven Celestijnenlaan, 200 F, B-3001 Leuven, Belgium.*e-mail: feng@mpip-mainz.mpg.de; muellen@mpip-mainz.mpg.de

ARTICLESPUBLISHED ONLINE: 8 DECEMBER 2013 | DOI: 10.1038/NCHEM.1819

NATURE CHEMISTRY | VOL 6 | FEBRUARY 2014 | www.nature.com/naturechemistry126

Bandgap= 1.88 eVMobility= 150 − 15, 000 cm2V−1s−2

BOTTOM-UP ApproachLIQUID-PHASE ASSISTED Funded by the European Science Foundation

GRAPHENE NANORIBBONSRibbon vs dimer and trimer

NATURE CHEMISTRY | www.nature.com/naturechemistry 42

SUPPLEMENTARY INFORMATIONDOI: 10.1038/NCHEM.1819

42

Supplementary Figure 16. Raman spectra of GNR 3, dimer 4, and trimer 5 measured at 514.5 nm on

powder samples with laser power below 0.1 mW.

Characterization of polyphenylene precursor S18 and GNR 6 Linear-mode MALDI-TOF MS analysis of a fraction of polyphenylene precursor S18

obtained by fractionation with recycling preparative SEC (Mw: (8.4±0.8)–(12±1) kgmol–1, Mn:

(7.6±0.8)–(11±1) kgmol–1, and PDI of 1.1 based on SEC analysis) showed a regular pattern of

signals reaching m/z = ~25,000 with an interval of ~1,052, which was in agreement with the

molecular weight of one repeating unit, i.e. 1,054 (Supplementary Fig. 17). Fragmentation of

the alkyl chains was observed as small peaks in the spectrum58,59. Reflectron-mode

MALDI-TOF MS analysis of the same sample of S18 displayed peaks at m/z = 5,294, 6,348,

7,402, 8,455, 9,510, corresponding to the molecular weight of cyclic oligomers with sodium

ion (Supplementary Fig. 18). This MALDI-TOF MS result as well as the SEC profiles of

precursor S18 (Supplementary Fig. 19) suggested formation of cyclic oligomers by

intramolecular Diels–Alder cycloaddition in a similar manner to that of precursor 2

(Supplementary Fig. 4)19.

300 900 1,500 2,100 2,700Raman shift (cm–1)

3,300

Inte

nsity

(a.u

.) 220 340 460 580 700Raman shift (cm–1)

Inte

nsity

(a.u

.)

GNR 3

Dimer 4Trimer 5

SWNT

Narita,..,Casiraghi, Feng, Müllen, ACS Nano, 2014

GRAPHENE NANORIBBONS

Aim: understanding origin of the RLBM and its dependence on structure in collaboration with Ferrari, Cambridge and Molinari, Modena

Collaborators

A. K. GeimK. S. NovoselovF. Withers*A. FeltenA. C. FerrariK. MuellenX. FengE. MolinariD. Prezzi

A. Eckmann*H. Yang*D. McManusY. ShinS. K. SonC. HolroydK. Zhou*R. Sorrentino*I. Verzhbitskiy*

Acknowledgements

THANKS FOR YOUR ATTENTION!

MY GROUP

GRAPHENE WEEK 2015

http://graphene-flagship.eu/?page_id=554