progress, challenges, and opportunities in two-dimensional...

of 29 /29
BUTLER ET AL. VOL. 7 NO. 4 28982926 2013 www.acsnano.org 2898 March 06, 2013 C 2013 American Chemical Society Progress, Challenges, and Opportunities in Two-Dimensional Materials Beyond Graphene Sheneve Z. Butler, †," Shawna M. Hollen, ‡," Linyou Cao, § Yi Cui, ^ Jay A. Gupta, Humberto R. Gutie ´ rrez, 0 Tony F. Heinz, ) Seung Sae Hong, ^ Jiaxing Huang, z Ariel F. Ismach, # Ezekiel Johnston-Halperin, Masaru Kuno, 4 Vladimir V. Plashnitsa, 4 Richard D. Robinson, 1 Rodney S. Ruoff, # Sayeef Salahuddin, 2 Jie Shan, 3 Li Shi, O Michael G. Spencer, b Mauricio Terrones, 0 Wolfgang Windl, 9 and Joshua E. Goldberger †, * Department of Chemistry and Biochemistry and Department of Physics, The Ohio State University, Columbus, Ohio 43210, United States, § Department of Materials Science and Engineering, North Carolina State University, Raleigh, North Carolina 27607, United States, ^ Department of Materials Science and Engineering, Stanford University, Palo Alto, California 94305, United States, ) Department of Physics and Electrical Engineering, Columbia University, New York, New York 10027, United States, z Department of Materials Science and Engineering, Northwestern University, Evanston, Illinois 60208, United States, # Department of Mechanical Engineering and the Materials Science and Engineering Program, The University of Texas at Austin, Austin, Texas 78712, United States, 4 Department of Chemistry and Biochemistry, University of Notre Dame, Notre Dame, Indiana 46556, United States, 1 Department of Materials Science and Engineering, Cornell University, Ithaca, New York 14853, United States, 2 Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, California 94720, United States, 3 Department of Physics, Case Western Reserve University, Cleveland, Ohio 44106, United States, O Department of Mechanical Engineering, The University of Texas at Austin, Austin, Texas 78712, United States, b Department of Electrical and Computer Engineering, Cornell University, Ithaca, New York 14850, United States, 0 Department of Physics, The Pennsylvania State University, University Park, Pennsylvania 16802, United States, and 9 Department of Materials Science and Engineering, The Ohio State University, Columbus, Ohio 43210, United States. " These authors contributed equally. T wo-dimensional (2D) materials have historically been one of the most ex- tensively studied classes of materials due to the wealth of unusual physical phe- nomena that occur when charge and heat transport is conned to a plane. Many ma- terials with properties dominated by their two-dimensional structural units such as the layered metal dichalcogenides (LMDCs), copper oxides, and iron pnictides exhibit correlated electronic phenomena such as charge density waves and high-temperature superconductivity. 13 The (re)discovery 4,5 of single-layer graphene in 2004 by Novoselov and Geim has shown that it is not only possible to exfoliate stable, single-atom or single-polyhedral-thick 2D materials from van der Waals solids, but that these materials can exhibit unique and fascinating physical properties. In single-layer graphene's band structure, the linear dispersion at the K point gives rise to novel phenomena, such as the anomalous room-temperature quantum Hall eect, and has opened up a new category of Fermi-Diracphysics. Even at one-atom- thick, graphene is a fantastic electronic and thermal conductor, and graphene-based materials have been proposed for a host of * Address correspondence to [email protected] Received for review January 18, 2013 and accepted March 6, 2013. Published online 10.1021/nn400280c ABSTRACT Graphene's success has shown that it is possible to create stable, single and few-atom-thick layers of van der Waals materials, and also that these materials can exhibit fascinating and technologically useful properties. Here we review the state- of-the-art of 2D materials beyond graphene. Initially, we will outline the dierent chemical classes of 2D materials and discuss the various strategies to prepare single-layer, few-layer, and multilayer assem- bly materials in solution, on substrates, and on the wafer scale. Additionally, we present an experimental guide for identifying and characterizing single- layer-thick materials, as well as outlining emerging techniques that yield both local and global information. We describe the dierences that occur in the electronic structure between the bulk and the single layer and discuss various methods of tuning their electronic properties by manipulating the surface. Finally, we highlight the properties and advantages of single-, few-, and many-layer 2D materials in eld-eect transistors, spin- and valley-tronics, thermoelectrics, and topological insulators, among many other applications. KEYWORDS: two-dimensional materials . graphene . nanosheets . graphane . van der Waals epitaxy . van der Waals solid REVIEW

Author: others

Post on 19-Mar-2020

0 views

Category:

Documents


0 download

Embed Size (px)

TRANSCRIPT

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2898

    March 06, 2013

    C 2013 American Chemical Society

    Progress, Challenges, andOpportunities in Two-DimensionalMaterials Beyond GrapheneSheneve Z. Butler,†," Shawna M. Hollen,‡," Linyou Cao,§ Yi Cui,^ Jay A. Gupta,‡ Humberto R. Gutiérrez,0

    Tony F. Heinz, ) Seung Sae Hong,^ Jiaxing Huang,z Ariel F. Ismach,# Ezekiel Johnston-Halperin,‡

    Masaru Kuno,4 Vladimir V. Plashnitsa,4 Richard D. Robinson,1 Rodney S. Ruoff,# Sayeef Salahuddin,2

    Jie Shan,3 Li Shi,O Michael G. Spencer,b Mauricio Terrones,0 WolfgangWindl,9 and Joshua E. Goldberger†,*

    †Department of Chemistry and Biochemistry and ‡Department of Physics, The Ohio State University, Columbus, Ohio 43210, United States, §Department of MaterialsScience and Engineering, North Carolina State University, Raleigh, North Carolina 27607, United States, ^Department of Materials Science and Engineering, StanfordUniversity, Palo Alto, California 94305, United States, )Department of Physics and Electrical Engineering, Columbia University, New York, New York 10027, UnitedStates, zDepartment of Materials Science and Engineering, Northwestern University, Evanston, Illinois 60208, United States, #Department of Mechanical Engineeringand the Materials Science and Engineering Program, The University of Texas at Austin, Austin, Texas 78712, United States, 4Department of Chemistry andBiochemistry, University of Notre Dame, Notre Dame, Indiana 46556, United States, 1Department of Materials Science and Engineering, Cornell University, Ithaca,New York 14853, United States, 2Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, California 94720, United States,3Department of Physics, Case Western Reserve University, Cleveland, Ohio 44106, United States,ODepartment of Mechanical Engineering, The University of Texas atAustin, Austin, Texas 78712, United States, bDepartment of Electrical and Computer Engineering, Cornell University, Ithaca, New York 14850, United States,0Department of Physics, The Pennsylvania State University, University Park, Pennsylvania 16802, United States, and 9Department of Materials Science andEngineering, The Ohio State University, Columbus, Ohio 43210, United States. "These authors contributed equally.

    Two-dimensional (2D) materials havehistorically been one of the most ex-tensively studied classes of materialsdue to the wealth of unusual physical phe-nomena that occur when charge and heattransport is confined to a plane. Many ma-terials with properties dominated by theirtwo-dimensional structural units such asthe layered metal dichalcogenides (LMDCs),copper oxides, and iron pnictides exhibitcorrelated electronic phenomena such ascharge density waves and high-temperaturesuperconductivity.1�3 The (re)discovery4,5 ofsingle-layer graphene in 2004 by Novoselov

    and Geim has shown that it is not onlypossible to exfoliate stable, single-atom orsingle-polyhedral-thick 2D materials fromvan derWaals solids, but that thesematerialscan exhibit unique and fascinating physicalproperties. In single-layer graphene's bandstructure, the linear dispersion at the K pointgives rise to novel phenomena, such as theanomalous room-temperature quantumHalleffect, and has opened up a new categoryof “Fermi-Dirac” physics. Even at one-atom-thick, graphene is a fantastic electronic andthermal conductor, and graphene-basedmaterials have been proposed for a host of

    * Address correspondence [email protected]

    Received for review January 18, 2013and accepted March 6, 2013.

    Published online10.1021/nn400280c

    ABSTRACT Graphene's success has shown that it is possible to

    create stable, single and few-atom-thick layers of van der Waals

    materials, and also that these materials can exhibit fascinating and

    technologically useful properties. Here we review the state-

    of-the-art of 2D materials beyond graphene. Initially, we will outline

    the different chemical classes of 2D materials and discuss the various

    strategies to prepare single-layer, few-layer, and multilayer assem-

    bly materials in solution, on substrates, and on the wafer scale. Additionally, we present an experimental guide for identifying and characterizing single-

    layer-thick materials, as well as outlining emerging techniques that yield both local and global information. We describe the differences that occur in the

    electronic structure between the bulk and the single layer and discuss various methods of tuning their electronic properties by manipulating the surface.

    Finally, we highlight the properties and advantages of single-, few-, and many-layer 2D materials in field-effect transistors, spin- and valley-tronics,

    thermoelectrics, and topological insulators, among many other applications.

    KEYWORDS: two-dimensional materials . graphene . nanosheets . graphane . van der Waals epitaxy . van der Waals solid

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2899

    applications ranging from transparent conductors tothermal interface materials to barristor transistor-likedevices.6�8 Furthermore, as single-layer graphene isentirely surface area, its properties and reactivity pro-foundly depend on the substrate, its local electronicenvironment, and mechanical deformations.Still, there exists an entire periodic table of crystalline

    solid-state materials each having different electronic,mechanical, and transport properties, and the possibi-lity to create single-atom or few-atom polyhedral thick2D layers from any material remains. It was showndecades ago by Frindt et al. that layered van der Waalsmaterials, such as layered metal dichalcogenides,could be mechanically and chemically exfoliated intofew and single layers.9,10 This early work focused onattempts to obtain and characterize these thin layers.9�13

    Experiments probing transport11 only scratched the sur-face of theuniqueproperties these2Dmaterials exhibit. Itwas not until the recent surge of intense research ongraphene that the general potential of 2D materialsbecame apparent. Additionally, the past 8 years of grap-hene research has yielded many methods for synthesiz-ing, transferring, detecting, characterizing, and manipu-lating the properties of layered van der Waals materials.Furthermore, novel synthetic methods including topo-tactic, solution-based, solvothermal, and UHV surfaceepitaxial approaches have unleashed the potential tocreate new van der Waals solids and single-layer-thickmaterials. These established methods have enabled thefield of 2D materials beyond graphene to mature veryquickly.Many novel materials that had been initially consid-

    ered to exist only in the realm of theory have beensynthesized. These include groups IV and II�VI semi-conductor analogues of graphene/graphane (the sp2/H-terminated sp3 derivatives) such as silicene14�17 andgermanane.18 Similar to graphene, the properties atthe single layer are also distinct from the bulk. Further-more, these 2D materials are useful building blocksthat can be restacked and integrated into compositesfor a wide range of applications.Herein, we present a forward-looking review article

    that discusses the state-of-the-art of 2D materialsbeyond graphene. Initially, we will outline the differentchemical classes of 2D materials and discuss thevarious strategies to prepare single-layer and multi-layer assemblies in solution, on substrates, and on thewafer scale. Additionally, we present an experimentalhow-to guide for identifying and characterizing single-layer-thick materials, as well as outline emerging tech-niques that yield both local and global information. Wedescribe the differences that occur in the electronicstructure between the bulk and the single layer anddiscuss various methods of tuning the properties bymanipulating the surface. Finally, we highlight theproperties and advantages of single-, few-, and many-layer 2D materials in field-effect transistors, spin- and

    valley-tronics, thermoelectrics, and topological insula-tors, among many other applications.

    Structure and Synthesis of Two-Dimensional Materials. Thereliable synthesis of single- and few-layer 2D materialsis an essential first step for characterizing the layer-dependent changes in their properties, as well asproviding pathways for their integration into a multi-tude of applications. In general, there are three mainclasses of materials that can be prepared as a single-atom or single-polyhedral-thick layer.

    Classes of Single- and Few-Layer Two-Dimensional

    Materials. Layered van der Waals Solids. The most com-mon class of crystalline structures that can be exfo-liated as stable single layers are the layered van derWaals solids. These crystal structures feature neutral,single-atom-thick or polyhedral-thick layers of atomsthat are covalently or ionically connected with theirneighbors within each layer, whereas the layers areheld together via van der Waals bonding along thethird axis. The weak interlayer van der Waals energies(∼40�70 meV) enable the facile exfoliation of theselayers. The most common approaches for obtainingsingle- and few-layer-thick 2D materials from many ofthese solids include mechanical exfoliation of largecrystals using “Scotch tape”, chemical exfoliation bydispersing in a solvent having the appropriate surfacetension, and molecule/atom intercalation in order toexfoliate these layers and enable their dispersion inpolar solvents. This mechanical exfoliation process hasbeen used to prepare and study the properties of few-layer van der Waals materials, such as MoS2 and NbSe2,since the 1960s.9,11,19 The isolation of individual andfew layers using mechanical exfoliation remains themost powerful approach for studying their propertiessince it is considerably less destructive than the othermethods and has successfully been used to create large,10 μm single-layer flakes on a variety of substrates.

    One of the most well-studied families of van derWaals solids is the layeredmetal chalcogenides (LMDCs),the most common being MoS2. Early transition metal

    VOCABULARY: two-dimensional material - a material

    in which the atomic organization and bond strength along

    two-dimensions are similar and much stronger than along

    a third dimension; van der Waals solid - a material whose

    crystal structure features neutral, single-atom-thick or

    polyhedral-thick layers of atoms with covalent or ionic

    bonding along two dimensions and van der Waals bond-

    ing along the third;nanosheet - stacked assembly or

    hybrid composite formed from single to many layers of

    two-dimensional materials;graphane - a single layer of a

    two-dimensional hexagonal network of sp3-bonded car-

    bon atoms in which every carbon is bonded to a terminal

    hydrogen, alternatingly above and below the layer; van

    der Waals epitaxy - the growth of a thin layer on the

    surface of a substrate in which the material�substrate areheld together by weak van der Waals forces.

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2900

    dichalcogenides with stoichiometry MX2 (M = Ti, Zr, Hf,V, Nb, Ta, Re; X = S, Se, Te) crystallize into layered 2Dstructures inwhich hexagonally packedMX6 octahedra(for d0, d3, and some d1 metals) or trigonal prisms (ford1 and d2 metals) share edges with their six nearestMX6 neighbors within each layer (Figure 1a).

    1 There areover 30 different LMDCs which have many technolo-gically interesting properties, and an emerging body ofexperimental work investigating the structure andproperties of single- and few-layer-thick derivativeshas evolved for many of these compounds (MoS2,WS2, and TiSe2).

    20�22 Other families of van der Waalssolids that have been exfoliated into single layersinclude hexagonal boron nitride,23 vanadium oxidederivatives, and other chalcogenides including Bi2Te3,Sb2Te3, and β-FeSe.

    24,25 Many novel van der Waalscompounds can be created via the topotactic deinter-calation of precursor solids. For example, the layeredCaGe2 and CaSi2 Zintl phases can be topochemicallydeintercalated in aqueous HCl to produce layeredhydrogen-terminated or half-hydrogen-terminated/half-hydroxy-terminated GeH and SiH0.5(OH)0.5, re-spectively (Figure 1b).18,26�28 These group IV graphaneanalogues29�32 are a particularly intriguing class ofsystems due to the possibility of utilizing covalentchemistry to modulate and tune the properties. Asanother example, recently, the exfoliation of metalcarbides such as Ti3AlC2 using HF to produce neutrallayers of Ti3C2(OH)2 has been demonstrated.

    33 Newlayered van der Waals solids are constantly being dis-covered by the solid-state community. In 2012, ReN2was synthesized for the first time using high-pressuretechniques and was found to have the MoS2 crystalstructure type.34

    Neutral, dimensionally reduced hybrid organic/in-organic van der Waals derivatives of nonlayered solidshave also recently been discovered. Dimensional re-duction refers to the creation of novel crystal structuresof metal-anion (M-X) frameworks by the addition of areagent that disrupts the polyhedral connectivityalong one or more dimensions while retaining a de-gree of the metal coordination geometry and generalpolyhedral connectivity.35 In these hybrid van derWaals solids, stoichiometric equivalents of neutral or-ganic ligands bind to the metal and disrupt the M-X-Mframework without changing the electron count andrelative metal-anion radii. For example, it has beenshown that almost every II�VI semiconductor thattypically crystallizes into the three-dimensional spha-lerite or wurtzite lattices, such as ZnS, ZnSe, and ZnTe,can be converted into atomically thin 2D crystallineframeworks when synthesized via solution-phase sol-vothermal techniques in the presence of alkylamineligands.36�40 The bulk sphalerite or wurtzite structureconsists of a corner-sharing metal-anion tetrahedral.These dimensionally reduced structures can be envi-sioned as a single (1�10) plane of corner-sharing

    metal-anion tetrahedra where every metal is bondedto three anions and cappedwith a short-chain alkylamineand every anion is bonded to three different metals(Figure 1c). The exfoliation into single-layer-thick deriva-tives of these hybrid materials, such as ZnSe, has beenrecently demonstrated.41

    Layered Ionic Solids. The second class of materialsthat can be prepared as single or few layers featuresbulk crystal structures with charged 2D polyhedral layersthat are typically held together with strongly electro-positive cationsor strongly electronegativeanions suchashalides, or OH�. To enable their dispersion as singlelayers in solution, these cations/anions are typicallyexchanged with bulky organic cations/anions, such astetrabutylammonium/dodecyl sulfate. These materials canthen be easily dispersed onto substrates, with themajorityof materials depositing as single to few layers. There arenumerous examples of oxide materials that have beenprepared this way including (1) cation-exchanged layersfromRuddlesden�Popperperovskite-type structures, suchas KLn2Ti3O10, KLnNb2O7, RbLnTa2O7, and KCa2Nb3O10

    Figure 1. (a) Crystal structures of the 1T and 2H crystalstructures of theMX2 family (X = yellow sphere). Themetal isin octahedral coordination in the 1T structure, and trigonalprismatic coordination in the two layers per unit cell 2Hcrystal structure. (b) Deintercalation of CaGe2 in aqueousHCl at low temperatures results in GeH (calcium = yellow,germanium = purple, hydrogen = black). (c) Crystal struc-ture of ZnSe(butylamine) (Zn = gray, Se = orange, N = green,C = black). (d) Crystal structure of KCa2Nb3O10, a Ruddle-sden�Popper perovskite phase that can be exfoliated uponreplacing the Kþ cation (orange) with an organic cation.

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2901

    (Figure 1d)42�47 (Ln = lanthanide ion); (2) cation-exchanged layered metal oxides such as LiCoO2 andNa2Ti3O7;

    48,49 (3) halide- or hydroxide-exchanged layersderived from metal hydroxides, such as Ni(OH)2�x orEu(OH)2.5Cl0.5.

    50,51 Additionally, many neutral layeredtransition metal oxides such as MnO2 can undergochanges in oxidation state and become protonated inaqueous acidic solutions.52 This proton can then besubstituted for bulky organic cations. The observedthicknesses from atomic force microscopy (AFM) forthese classes of materials are typically larger than theirexpected values due to the presence of hydrationlayers or intercalating ions, which results in a surface�sheet interface with a thickness between 0.5 and1.0 nm. The lateral sizes of exfoliated layers from ionicmaterials typically depend on the size of the startingcrystalline material, and single-layer flakes in the rangeof tens of micrometers have been observed.48

    Surface Growth of Nonlayered Materials. The de-position of materials on substrates offers the potentialto grow and study the properties of single- to few-atom-thick materials beyond those existing as layeredbulk crystals. For example, recently, it was demonstratedthat monolayers of silicon deposited on Ag(111) or ZrB2organize into a puckered hexagonal graphene-like latticewith sp2 bonding configuration.17,53 This silicenematerialshares a similar band structure to graphene; how-ever, the interactions with the substrate induce a bandgap opening at the K point. Additionally, ultrathin in-sulators such as Cu2N,

    54,55 Al2O3,56 NaCl,57 MgO,58 TiO2,

    59

    andmetal adlayers havebeenprepared in this fashion onmetal substrates. Unlike exfoliation, the choice of sub-strate has been typically limited to metals due to theprevalence of STM as a characterization tool.

    Solution-Phase Growth. Solution-phasemethods suchas solvothermal or colloidal growth reactions offer afacile production method to synthesize gram scalequantities of 2D materials with precise thicknessesand basal-plane sizes.60�64 Recently, general colloidalsynthetic methods have been developed to prepareLMDCs such as TiS2, VS2, ZrS2, HfS2, NbS2, TaS2, TiSe2,VSe2, and NbSe2 via the reaction of metal halides andcarbon sulfide or elemental selenium in the presenceof primary amines.65 The colloidal materials typicallyrange in lateral dimensions from 10 to 100 nm andhave thicknesses from single sheets to tens of nano-meters. As an example, TiS2 nanosheets that havelateral lengths of 500 by 500 nm with thicknesses onthe order of 5 nm have been created (Figure 2). Thesemethods use abundant low-cost precursors and mildcolloidal growth conditions.66 However, strategies forcontrolling the thicknesses and lateral dimensionshave yet to be established.

    One of the key steps in using solution-phase meth-ods to grow transition LMDCs is the generation ofchalcogenide anions in the appropriate oxidation state(S2�, Se2�, and Te2�). To grow sulfides, primary amines

    and sulfur are typical reactants because they formsulfur-containing alkylammonium polythioamine,67

    polythiobisamine,68,69 and alkylammonium poly-sulfide70 complexes. These sulfur-containing species,in turn, exhibit favorable decomposition kinetics andproduce H2Swhen heated, which reacts with themetalprecursor.

    Vapor Deposition. Vapor deposition stands as anappealing, versatile synthetic strategy. However, thedevelopment of a controlled synthetic method of 2Dmaterials such as 2D chalcogenides by vapor deposi-tion requires a better understanding of the fundamen-tals involved. While vapor deposition has beenextensively used for the growth of thin films andnanomaterials such as nanowires,71 nanotubes,72 andgraphene,73 knowledge obtained from these materialsmay not simply apply to 2D materials. Unlike typicalnanomaterials whose growth is primarily governed bya catalyst, the vapor deposition growth of 2D chalco-genides is often noncatalytic.24,74,75 Without the dom-inance of catalysts, the growth of 2D nanosheets issubject to strong influences of many experimentalparameters that may play only a negligible role incatalyzed growths. For example, the diffusion of sourcematerial vapor through the gas flow boundary layerstrongly controls the vapor deposition growth of GeSnanosheets (Figure 3).74

    The growth of single-layer substrate-wide 2D ma-terials is essential for commercialization and wouldalso benefit fundamental studies of single-layer phe-nomena. The key to the preparation of substrate-widesingle-layer van der Waals 2Dmaterials and their hetero-junctions is monolayer (ML) epitaxy. van der Waalsforces have associated energies of 40�70 meV, which

    Figure 2. (a,b) Low-resolution TEM images of TiS2 layers.The inset in (a) shows an image of multiple TiS2 nanosheets,and the inset in (b) a selected area electron diffractionpattern of a single TiS2 layer. (c,d) High-resolution TEMimages showing the edge and basal plane of a single TiS2layers.66

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2902

    are much smaller than covalent bonding energies of200�6000 meV. We infer that epitaxy of crystalline 2Dmaterials on crystalline substrates is defined by strongbonding at the reactive edges of the single-crystaldomains of the material and weak interlayer forcesbetween the sheets. This van der Waals epitaxy occursbetween 2D sheets of the same material (homoepitaxy)aswell as 2D sheets of differentmaterials (heteroepitaxy).Because of the weak interlayer forces, epitaxy is possibleeven if there is significant lattice mismatch between thematerials.25,75 The resulting heterojunctions do not sufferfrom the interfacial defects that are generated during 3Dheteroepitaxy due to the large stresses generated bybent or broken interlayer bonds.76,77 van der Waalsheteroepitaxy has also been shown to be possible with3D substrates that have been suitably passivated.76,77

    Table 1 shows a listing of crystalline 2D heterojunctionsthat have been produced using van der Waals epitaxy.

    Large-Area CVD Growth of Graphene and Hexagonal

    Boron Nitride. It is worthwhile to re-evaluate the suc-cessful vapor deposition growth of single-layer gra-phene in order to understand the extent to which wecan transfer these methods to the substrate-widegrowth of other van der Waals sytems. Thin carbonfilms were originally grown on single-crystal transitionmetals such as platinum by exposing the metal surfaceto a hydrocarbon at high temperature in ultrahighvacuum (UHV) conditions.78�80 LEED patterns observedby the Somorjai group78�80 were assigned in 19694 tobeing from single- and few-layer graphene (SLG and FLG,respectively). The formationof thesegraphene layerswaslater explained by the dissociation of the hydrocarbon onthe metal surface, carbon diffusion into the bulk of the

    metal, and its segregation during cooling or by carbonsupersaturation.81,82 There is an interest in syntheticapproaches for the formation of large-area graphene,for basic research as well as a variety of applications,and this has driven the re-evaluation of FLG growth ontransition metals by the diffusion�segregation tech-nique. The growth of FLG on Ru,83 Ir,84 Co,85,86

    Ni,73,81,82,87 Pt,85 and Pd85 by chemical vapor deposi-tion (CVD) has been reported. Control of the qualityand number of layers grown has proven challenging,and often inhomogeneous films are obtained. FLG wasalso achieved by the sublimation of silicon in SiCcrystals.88

    The successful synthesis of SLG with high homo-geneity and reproducibility was achieved in 2009 bylow-pressure CVD on copper foils with methane as thecarbon source.73 Aspects of the growth kinetics andmechanisms were elucidated using an isotope-label-ing technique in which the Cu is exposed sequentiallyto 13CH4 and then normal methane.

    89 13C-labeledgraphene can be readily distinguished from normalgraphene by Raman spectroscopy and mapping(Figure 4a). A comparison of the growth of monolayergraphene and FLG on Cu versus Ni showed that thegraphene growth on Cu is surface-mediated; that is,dissociation of the hydrocarbon followed by carbonspecies diffusion on the surface leads to nucleation,island growth, and finally completion of amonolayer.89

    This is rationalized by the extremely low carbon solu-bility in Cu even at the growth temperature of about1040 �C that inhibits the diffusion of C into the bulk Cu,making Cu foil an excellent substrate for growth of large-area SLG. Kinetic studies on the growth of graphene ledto the conclusion that the graphene grain size could beincreased by raising the growth temperature and

    TABLE 1. Two-dimensional Heterojunctions Produced by

    van der Waals Epitaxy

    layer substrate ref

    graphene sapphire 322C60 MoS2 323MoS2 graphene 75GaS GaSe, Si(111) 324GeS SiO2/Si 74TaS2 mica 325WS2 graphite 326,327HfS2 WSe2 328CdS InSe/H-S(111) 329SnS2 WSe2, MoS2, MoTe2, GaSe, WSe2 330,331SnSe2 SnS2,WSe2, MoS2, MoTe2, GaSe 330,331Bi2Se3 graphene/SiC, Si(111) 238,332BiTe2Se h-BN 333ZnSe InSe, GaSe 334MoSe2 SnS2, MoS2, GaAs(111) 335GaSe GaAs(111), Si(111) 336,337NbSe2 mica, GaAs(111) 338CdTe MoSe2, WSe2 339PbTe Si(111) 340

    Figure 3. (a) Schematic illustration of experimental setupfor synthesis by a noncatalytic vapor deposition process.The precursor vapor can be introduced from outside orgenerated inside the tube furnace. (b) TEM image of a typicalGeS nanosheet. Inset is an electron diffraction pattern of thenanosheet.74

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2903

    reducing the partial pressure of the hydrocarbon, andgrain size was thereby increased, at first, from a fewmicrometers to tens ofmicrometers.90 The domain sizeof SLG has since been increased to several hundredmicrometers91,92 and even millimeters93 (Figure 4c).Onemeasure of graphene “quality” is themagnitude ofthe carrier mobility, and larger grain size has beencorrelated with higher carrier mobility values.90,93

    “Good” conditions to grow SLG include (1) hydro-carbon dissociation (catalytic or thermal); (2) a catalystfavoring growth of graphene; and (3) nucleation,growth, and completion of the monolayer film byhaving all reactions occurring on the surface of thesubstrate. Indeed, SLG was shown to grow in UHVconditions on single-crystal metals with high carbonsolubility4,82 and more recently by low-pressure CVDon Ni films94 by controlling the kinetic factors duringthe growth and, in particular, having the reaction occur

    only on the metal substrate surface. Other relevantparameters are the sticking coefficient of the carbonspecies on the metal surface and these species' abilityto diffuse on the surface.

    In contrast to graphene, other layered systems arecomposed of two or more elements, making thesynthesis more complex. Hexagonal boron nitride(h-BN) might be the most studied layered material aftergraphene, although reports on the CVD synthesis ofMoS2 are emerging.

    95 The synthesis of single- and few-layer h-BN was achieved first in UHV CVD systems onsingle-crystal metals.96 Recently, the synthesis of sin-gle-layer h-BN and few-layer h-BN was reported in CVDsystems using solid23,97 and gaseous precursors.98

    Submonolayer domains on Cu foils were achievedusing very low partial pressure of ammonia borane23

    (Figure 4d), while control over the number of layerswasachieved using diborane and ammonia on Ni foils,98 as

    Figure 4. (a) Schematic illustration showing the dissociation�dissolution�segregation on Ni and the surface-mediatedgrowth of monolayer graphene on Cu. The Raman mapping of the G band for graphene grown on Cu (bottom left) showsgraphene areas enriched with 13C (dark) and normal C (bright); note that the distribution of 13C is essentially random for thefilmgrownon theNi foil (bottom right);89 scale bars are 5μm. (b) Increasing thedomain sizeby raising the temperature (T) andlowering both the precursor gas flow (JMe) and the methane partial pressure (PMe) as compared to an earlier protocol, SEMimages; scale bars are 10 μm.91 (c) SEM images showing further increase of the graphene domain size whichwas attributed tocontrolling the surface roughness of the Cu foils.93 (d) AFM and SEM images showing submonolayer h-BN transferred to aSiO2/Si and Cu foil, respectively, inwhich triangular domains can be observed. The atomicmodel for such domains is depictedon the right.23 (e) Control of the number of h-BN layers on Ni foils by control of the reaction time. TEM images for samplesgrown for 1, 5, 15, and 30 min.98

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2904

    depicted in Figure 4e. Despite such achievements, thegrowth mechanism for h-BN remains unclear. Thegrowth of SLG was significantly improved in only afew years by understanding some of the key parametersfor growth. Can we apply similar methodologies to thegrowth of single or multiple layers of other layeredmaterials? It calls for further nucleation and growthstudies as well as an adequate set of characterizationtools comparable to thoseused to characterizegraphene.

    Multilayer Assemblies. As will be discussed later,there are numerous applications for multilayer assem-blies and hybrid composites from single- tomany-layer(thickness

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2905

    structural or spectroscopic information with propertiesat the local scale.

    The first challenge in single-layer characterization issingle-layer preparation and detection. Mechanicalexfoliation from bulk crystals is typically the simplestmethod for transferring 2D sheets of layered van derWaals solids such as graphene,5,20,120,121 metal chal-cogenides,122 and GeH18 to any substrate. Exfoliationtypically produces 10 μm sized flakes ranging in thick-ness from single to few layers. Optical microscopy isinitially the most powerful high-throughput methodfor initially identifying single- and multiple-layerflakes.20,58,122,123 Dielectric-coated SiO2/Si substratesare the most common substrates used to visualizeand locate single and few layers (Figure 6a,e,f). Thecolor of the dielectric-coated wafer depends on aninterference effect from reflection off of the two sur-faces of the dielectric. Single- and few-layer flakes onthe surface of the dielectric modify the interferenceand create a color contrast between the flake and thesubstrate.123,124 For optimal contrast, the thickness ofthe dielectric coating needs to be within 5 nm of theideal value. However, since the index of refraction ofmany novel materials is unknown, it is often necessaryto first experimentally exfoliate onto a range of sub-strates having different dielectric thicknesses to initi-ally determine the optimal thickness.

    Visualizing single layers on other substrates withoutthe need for interference methods has also beenexplored. Some metal dichalcogenides (MoS2, WS2)

    have direct band gaps only as single layers, enabl-ing their direct visualization by fluorescence micros-copy.125 However, as the number of layers increases,these materials' band gaps become indirect, signifi-cantly reducing their fluorescence.

    Fluorescence Quenching Microscopy (FQM) usesan organic fluorophore whose fluorescence is quen-ched in the presence of single- or few-layer sheets(Figure 6f). Many 2D materials feature either lowenough band gaps or the appropriate band alignmentto quench nearby fluorophores via a FRET mechanismor charge transfer mechanism. One can locate singlelayers of 2Dmaterials on a substrate bymonitoring thereduction in fluorescence intensity of a thin fluoro-phore layer that was spun-coat on top.126�128

    AFM is a powerful technique to determine layerthickness (Figure 6b,d) with a precision of 5%.114,118,129

    However, discrepancies arise from differences in theinteractions of the tip with the sample and sub-strate.130 For example, the thickness of a single layercan be better determined by measuring the height of asecond layer on a first layer, rather than the heightbetween a single layer and the substrate because in theformer the tip�layer interactions are constant. To controlthese differences and obtain the most accurate heightprofile of a single layer, one must first optimize the tip�surface distance to exclude any hysteretic artifacts.130

    Raman Spectroscopy is a useful method for finger-printing a material and layer-dependent changes ofthe vibrational structure (Figure 6c). It is necessary to

    Figure 6. (a) Optical micrograph of thin films of MoS2150 (b) AFM of MoS2.

    150 (c) Raman spectra of single-layer, few-layer, andbulk MoS2 films.

    150 (d) AFM of graphene,126 (e) optical micrograph of graphene,126 (f) FQM image of graphene using PVP/fluorescein.126 Scale bars are 10 μm.

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2906

    use low power (e.g., 16 kW/cm2 for Bi2Te3,131 or

    40 kW/cm2 for GeH) when obtaining Raman spectraof single layers to prevent sample decomposition. TheStokes/anti-Stokes peak ratio can be used to determinethe local temperature.132 Raman spectra can alsodetect vibrational modes that are active due to sym-metry breaking in single-layer films and enhancedvibrations in out-of-plane modes when the single layeris suspended.131 When varying the layer number be-tween bulk, few, and single layers, the Raman spectradiffer in spectral width and intensity due to differinginterlayer interactions.122,129 Substrates that have vibra-tional modes that overlap with those in the material ofinterest should be avoided. Vibrational spectroscopy,in general, and Raman spectroscopy, in particular,could be used to detect isotopic-enriched 2Dmaterialsto study the growth mechanism as demonstrated andthus be used to study the growth mechanism.89

    Transmission Electron Microscopy (TEM) can pro-vide detailed information on the nature of crystallinity,layer sizes, interlayer stacking relationships, and ele-mental composition (Figure 7a,b).63,114,119,122,133�135 Inthe case of graphene, selected area electron diffrac-tion can distinguish between monolayer and multi-layer graphene when the intensity ratio of the (100)to (110) diffraction spots is larger or smaller than 1,respectively. This deviation in diffraction intensity canalso distinguish between single- and multilayer flakesof other 2D materials when there is a change ininterlayer registration. Recent advances have shownthat dark-field microscopy can quantitatively deter-mine the layer number and stacking order ofmultilayergraphene.136 Scanning transmission electron micro-scopy (STEM), using high-angle annular dark-field de-tector coupled with electron energy loss spectroscopycan be used to visualize individual atoms in an isolatedlayer. Here a sub-angstrom electron beam is rasteredacross a sample and is scattered when it hits a parti-cular atom. Figure 7c shows a STEM image of carbonand oxygen substitutions in a single layer of boronnitride. Low accelerating voltages (

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2907

    nanosheets, the presence of low-angle basal reflectionsin the pattern indicates an ordered arrangement of re-stacked nanosheets along the stacking axis, and like-wise, the absence of intense basal reflections indicatesrandom orientations for aggregated nanosheets.111

    The orientation of films on substrates favors the align-ment of the thin axis. Broadening signatures of specificpeaks under this orientation can indicate the thinaxis.147 To gain more insight, the nanosheet XRDpattern can be modeled by the structure factor, theLorentz polarization factor, the Laue interference func-tion, and the temperature factor.114,146 Structure factorsfor single and multiple layers can be calculated assum-ing atomic positions, unit cell structure, and orientationto a substrate.114 Results of this modeling can helpdetermine the fundamental units of the nanosheets(e.g., atomic make up of termination layers), the spacingbetween sheets, and if the nanosheets are restacked. In-plane XRD can be used successfully to observe hkreflections (if l is the thickness axis) of the nanosheets'2D unit cell.114,119 This is a useful method to observemissing peaks from the naturally textured orientation ofnanosheets on a substrate. Unit cell parameters can beextracted from refinements of these data and comparedto bulk values.

    Characterization Opportunities. The developmentand implementation of new characterization techni-ques that can rapidly and nondestructively probe theatomic structure, defects, and properties of single-layermaterials is necessary to advance the field. Since the

    electronic properties of materials are often dominatedby minority defects, the discovery, availability, andimplementation of techniques that can identify min-ority species, defects, and edge states, such as next-generation aberration-corrected STEMaswell as STM,willprove incredibly valuable. The emergence of combinedspectroscopic and imaging techniques such asAFM-TERScan significantly aid in the identification and character-ization of single-layer-thick sheets while minimizing sub-strate signals. Furthermore, to refine the growth of high-quality single-layer materials for electronic applications,the implementation of feedback loops between synth-esis and device fabrication will be necessary.

    Properties, Integration, and Devices. The unique quali-ties of 2D materials, such as their reduced dimension-ality and symmetry, lead to the appearance ofphenomena that are very different from those of theirbulk material counterparts. This difference is perhapsmost glaring in the transformation of the band struc-ture as the single layer is approached. Additionally,since a 2D material is entirely made up of its surface,the interface between the surface and the substrateand the presence of adatoms and defects can drama-tically alter the material's inherent properties. The two-dimensional nature of these materials also plays anentirely mechanical role as they are inherently flexible,strong, and extremely thin. These materials exhibitother unique and potentially useful properties includ-ing high electron mobilities, topologically protectedstates, tunable band structures, and high thermalconductivities. The development of 2D materials isexpected to improve current device technology, andtheir novel transport and topological properties mightprovide additional opportunities for spintronic devicesand quantum computing.148 Even multilayer assem-blies of 2D materials have a broad base of potentialapplications including batteries, supercapacitors, andp-n junctions. This section provides a detailed over-view of the novel properties of 2D materials beyondgraphene and discusses their potential for deviceapplications and integration.

    Changes in Electronic Structure at the Single Layer:

    MoS2, a Case Study. The layeredmetal dichalcogenides(LMDCs) constitute one of the most studied families ofvan der Waals solids beyond graphite.148 These mate-rials display remarkably diverse electronic properties,ranging from metals to semiconductors to insulators.They also exhibit interesting strongly correlated elec-tron phenomena, such as charge density waves andsuperconductivity. Among them,molybdenumdisulfide(MoS2; semiconductor), has attracted particular atten-tion. Several distinctive properties, related both to itsoptical149�161 and transport characteristics,20,148,162�173

    have been demonstrated inMoS2 at themono- and few-layer level. Examples of these properties include theexistence of a direct gap,125,149 unlike the bulkmaterial,with strong excitonic effects156 and the possibility of

    Figure 8. XRD patterns of niobate nanosheets of (top) bulk,(middle) (solid line) wet aggregates of nanosheets and(dashed line) calculated diffractionpattern for a single layer,(bottom) and dried aggregates of nanosheets.146 Reprintedwith permission from ref 146. Copyright 2002 Elsevier.

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2908

    full control of the valley and spin occupation by opticalpumping with circularly polarized light.153�155,158 MoS2has also been shown to yield field-effect transistorswith very high current on�off ratios at monolayerthickness163 and to function within a vertical graphene-MoS2-graphene tunneling transistor strucure.

    171 Theseproperties, in combination with the promise of synthesisof large-area high-quality samples (notably by recentlydeveloped CVDmethods174,175), suggest interesting pos-sibilities for applications of this and related materials inelectronics and optoelectronics. Here we summarizesome of the recent progress in understanding theelectronic properties of ultrathin MoS2 layers.

    The overall physical structure of the MoS2 mono-layer is similar to that of the graphene honeycomblattice. For MoS2, however, the A and B sublattices,rather than both being occupied by C atoms, areoccupied either by Mo atoms or by a pair of S atoms(Figure 12c), with the entire MoS2 monolayer structureforming a triple layer of atoms. The difference betweenthe A and B sublattices lifts the degeneracy in theelectronic structure at the K (K0) point in the Brillouinzone. This degeneracy is responsible for the distinctiveDirac cone dispersion relation in graphene. In contrast,inmonolayerMoS2, a substantial energy gap is present.

    The bulk MoS2 crystal, consisting of MoS2 mono-layers arranged with Bernal stacking, is known to be anindirect gap semiconductor, with a band gap of 1.29eV.176 The calculated band structure for bulk MoS2shows that the valence and conduction band extremaoccur away from the K (K0) point (Figure 9a).125 Throughcharacterizationbyabsorption, photoluminescence (PL),and photoconductivity spectroscopy (Figure 9c�e) asa function of thickness, Mak and co-workers149 havetraced the effect of quantum confinement on theelectronic structure of MoS2 layers. With decreasingthickness, the indirect band gap, which lies below thedirect gap in the bulk material, shifts upward in energybymore than 0.6 eV. This leads to a crossover to a directgap material in the limit of monolayer thickness sincethe electronic states near the K point are more localizedwithin the layer and shift up in energy only modestlywith decreasing layer thickness (Figure 9a,b). The resultof this crossover in the band structure is the emergenceof bright photoluminescence, which is absent for thickerfilms, in the monolayer material125,149 (Figure 9c). Thedirect gap character of monolayer MoS2 has also beenconfirmed by ab initio calculations.125,177�181

    The ability to control the material's electronic struc-ture through the sample thickness is not unique toMoS2.

    Figure 9. Calculated band structures of (a) bulk and (b) monolayer MoS2.125 The solid arrows indicate the lowest-energy

    transitions. (c) PL spectra for mono- and bilayer MoS2 samples. Inset: PL quantum yield of thin layers of MoS2 for number oflayers N = 1�6 in log scale. (d) Absorption spectra (left axis) normalized by the layer number N in the photon energy rangefrom 1.3 to 2.4 eV for mono- and bilayer MoS2. The corresponding PL spectra (right axis, normalized by the intensity of thepeak (A) are included for comparison. The spectra are displaced along the vertical axis for clarity. (e) Bandgapenergy of thin layersof MoS2 forN = 1�6. The band gap values were inferred from the energy of the indirect gap PL feature I for N = 2�6 and from theenergy of the PL peak A for N = 1. As a reference, the (indirect) band gap energy of bulk MoS2 is shown as dashed line.

    149

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2909

    Similar changes in the electronic band structure as afunction of sample thickness have been predicted forother semiconducting LMDCs compounds MX2(M = Mo, W and X = S, Se, Te).182 Furthermore, recentcalculations have shown that the electronic and opticalproperties of monolayer and few-layer MoS2 can becontrolled through strain engineering.183�186

    Another important aspect of the properties of 2Dsemiconducting LMDCs such as MoS2 is the importantrole of many-body electronic interactions. In particular,the optical properties are dominated by excitonictransitions. The enhanced Coulomb interactions exist-ing in these materials can be attributed to reduceddielectric screening in these atomically thin systems.The absorption spectrum of the undoped MoS2 mono-layer (Figure 9d) shows pronounced absorption peaksfrom excitonic transitions, rather than the steps thatwould be expected for absorption arising from band-to-band transitions in a 2D system with a direct gap. Thetwo features, known as the A (∼1.9 eV) and B (∼2.1 eV)excitons,125,149,187 arise from transitions from the twohighest-lying spin-split valence bands to the lowestconduction bands.

    The influence of doping on the optical properties ofmonolayer MoS2 has been studied recently by Maket al. using FET structures156 for electrostatic control ofthe doping level. By analyzing the absorption andemission line shapes at low temperature, Mak et al.have identified, in addition to the presence of the usualneutral excitons, negative trions, quasi-particles com-posed of two electrons and a hole. The trion bindingenergy has been determined to be∼20 meV, which farexceeds that observed in conventional 2D semiconduc-tor quantumwells and renders trions significant even atroom temperature. Similar phenomena have also re-cently been observed in MoSe2 FETs by Ross et al.

    157

    Because of symmetry, crystals frequently have in-dependent, degenerate valleys in the conduction orvalence band. Access to this so-called “valley” degreeof freedom has attracted interest both from the pointof view of fundamental physics and a possible newcontrol variable for new classes of electronic devices,commonly called “valley-tronics” and discussed inmore detail below. To date, however, it has provendifficult to selectively populate a given valley to pro-duce valley polarization. MoS2 monolayers with adirect energy gap at the K and K0 points in the Brillouinzone have recently been predicted to be suitable forthe control of valley population using optical pumpingwith circularly polarized light.154,188 Because of thebroken inversion symmetry, strong spin�orbit interac-tions split the spin-degenerate valence bands by∼160meV in MoS2 monolayers. This broken spin degener-acy, in combination with time-reversal symmetry, im-plies that the valley and spin of the valence bands areinherently coupled. When optical transitions betweenthe upper valence band and the conduction band

    (corresponding to the A exciton) are induced by circu-larly polarized photons, excitation occurs only at eitherthe K or K0 valley, depending on the helicity of the light.This excitation process, because of the coupled valleyand spin degree of freedom, also gives rise to photo-generated particles of a well-defined spin state. Thesetheoretical predictions have been recently verified byseveral experimental groups153�155,158 through a de-monstration of the retention of the circular polarizationstate of light emission induced by resonant excitationwith circularly polarized light. The important role ofinversion symmetry in establishing valley polarizationwas further confirmed by a comparison of the behaviorof MoS2 bilayers in which inversion symmetry, unlikefor the case of the monolayer, is present. More generaldiscussion of 2Dmaterials for spin- and valley-tronics isprovided in later sections of this review.

    Edge Effects on the Electronic Structure. As we dis-cuss in the previous section, LMDCs exhibit novelphysicochemical properties when the bulk (3D) formis reduced to a monolayer (2D). The most notoriouschange is related to the electronic band structure,which can change from an indirect band gap in 3D toa direct band gap in a 2Dmonolayer.135,149 If themono-layer has finite lateral dimensions, contributions fromthe edges might become important and new phenom-ena are expected.135,189Althoughamonolayer of LMDCsis actually composed of three atomic layers in whicheach layer of metal atoms is sandwiched between twolayers of chalchogenatoms (X-M-X), theprojection in theplane exhibits a hexagonal or honeycomb-like crystalstructure similar to graphene; hence the edge termina-tions canbezigzag, armchair, chiral, or amixture of thesetypes.

    Both the structural and chemical terminations ofthe edge are crucial to determine the local physico-chemical properties of these systems. First principlescalculations have predicted that diverse edge passiva-tion in metal dichalcogenides could produce differentspins states that modify the edge electronic and mag-netic properties.135,189,190 In addition, high spin densitycould be localized surroundingmetal vacancies.191 Theexperimental exploration of edges in inorganic 2Dmaterials is still at an early stage. From the experi-mental point of view, triangular clusters of MoS2 on Ausubstrates have been studied by STM,139,192,193 reveal-ing metallic electronic states localized close to theedges. For perfect continuous edges, those metallicstates can be viewed as one-dimensional conductingwires.192 The same group studied the size dependenceof the cluster morphology and electronic structure andconsistently found sulfur-terminated edges in single-layer clusters.139 Enhanced catalytic activity has alsobeen correlated with the density of edge sites in MoS22D clusters.194

    To date, MoS2 has been the testing groundfor many experiments in monolayers of metal

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2910

    dichalcogenides.135,139,149,150,152,155,163,173,174,194,199�202

    However,othercompoundsof this familyarealsogoodcan-didates to study new and exciting properties.135,191,199�202

    Recently, the optical behavior of WS2 triangular mono-layers was investigated by Raman spectroscopy andmicro-PL.135 Monolayers of WS2 exhibited a single PLpeak between 1.9 and 2.0 eV, corresponding to a 2Dexciton associated with the direct band gap transition. Aremarkable enhancement of the PL intensity was ob-served in regions close to the edges of the 2D trian-gular clusters (Figure 10a). PL mappings of the WS2triangular island (Figure 10b) showing the peak posi-tion and absolute maximum intensity, respectively, areshown in Figure 10c,e. In general, the spectrum closerto the edge shows both a significantly stronger PLsignal when compared to the central region, as well asa red shift of the PL peak. TEM analysis revealed that thetriangular islands possess mainly zigzag edges, and thesynthesis conditions provided a sulfur-rich environment.Determining the exact mechanism for PL enhancementdemands a one-to-one correlation between the chemicalstructure of the edge and its optical response. The mainchallenge consists of implementing experiments thatreveal the exciton dynamics, such as temperature-dependent PL and/or time-resolved PL, with spatialresolution of the order of tens of nanometers.

    Surface Tuning of Band Structure. The need to tunethe optical and electronic properties of graphene forapplications has created a large number of proposedmethods to engineer its band structure by introducingdefects and/or dopants.203�205 Both p-doping andn-doping of graphene have been reported, the latterby doping with nitrogen species204,206,207 and the formerby functionalization of graphenewith hydrogen, oxygen,hydroxyl groups, or carboxylic groups.208 However,

    adsorbed molecules have been found to have anextremely high mobility and low adhesion in ambientatmosphere,209 which makes reliable device fabrica-tion challenging. Chemisorbed molecules change theband structure dramatically from a near-zero gapsemimetal to a wide gap semiconductor, turning gra-phene into a highly effective insulator.210,211 The car-rier mobility has also been shown to be decreased bymore than 3 orders of magnitude to ∼10 cm2/V 3 s.Overall, tuning the electronic and carrier properties ofgraphene in a useful range has been a very challengingtask, stimulating the search for alternative 2D semi-conductors with better tunability.

    Even though the graphene-like band structure ofsilicene and germanane and the explanation for puck-ered versus flat 2D sheets was reported many yearsago,212 much less effort has been spent in examiningthe tunability of these alternative elemental 2D mate-rials, which have been successfully synthesized re-cently.15,17,18,53,211 Traditional density functional theo-ry (DFT) suggests that a 2.75 eV direct band gap isformed after complete hydrogenation of silicene,213�218

    whereas other types of functional groups (F, Cl, Br, andI) lead to different bandgap values.219 H-terminatedGegraphane analogues have been calculated to also havea direct band gap of 1.7 eV.220 Here we summarizerecent theoretical results that explain how to control-lably manipulate the relative band energies and localextrema of a covalently modifiable group IV silicongraphane analogue by termination with different or-ganic substituents.

    To develop the relationship between the bandstructure and the terminating functional group onsilicene, Restrepo et al.221 used DFT calculations withspin�orbit coupling and a hybrid HSE06 exchange-correlation functional,222,223 which features exact ex-change not included in previous calculations. Silicane(H-terminated silicene) is found to have an indirectband gap of 2.94 eV with the conduction band mini-mum at theM point and a direct gap of 3.14 eV at the Γpoint (Figure 11). The local conduction bandminimumat Γ originates from the Si�H σ*-bonds, whereas theglobal minimum at M corresponds to the Si�Si σ*-bonds.221 These two conduction band minima reactdifferently to manipulation of the silicane atomicstructure. Strain that increases the Si�Si bond lengthslowers the minimum at Γ while having a nearly neg-ligible effect on the minimum at M. This leads to achange from indirect to direct band gap when theground-state Si�Si bond length of 2.36 Å is increasedto 2.40 Å. This change also causes the effective mass atthe conduction band minimum to decrease from 0.18to 0.087 me (48%). Changes in the Si�H bond lengthalso mostly affect the Γ and not the M point states butare smaller than those induced by varying the Si�Sibond length. This high sensitivity of the Γ point mini-mum to changes in the bond between the Si atoms

    Figure 10. (a) Room-temperature PL spectra at two differ-ent positions (small circles in c and e) on the island shown in(b). (b) SEM images of a large WS2 triangular island. (c,e)Corresponding micro-PL mappings of the PL peak positionand the absolute intensity at the peak maximum, respec-tively. (d) Top-view atomistic model of a small WS2 clusterwith sulfur-saturated zigzag edges.135

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2911

    and the terminating hydrogen atom suggests that it ispossible to tune the band gap using different termi-nating substituents. Small organic moieties such asmethyl or ethynyl groups have also been calculated todecrease the conduction band energies and result indirect band gaps or equal conduction band minimumenergies at M and Γ (Figure 11), with varying effectivemasses.221 These results show that in contrast to gra-phene, where functionalization leads to drastic changesin the semimetallic band structure, silicene functiona-lization should allow for a wide continuous tunabilityin band and conduction properties and offer excit-ing prospects for the engineering of new electronicmaterials.

    Two-DimensionalMaterials for Spin- andValley-tronics.

    Expanding charge-based microelectronics to take ad-vantage of the spin and “valley” degrees of freedom isexpected to increase computing power, reduce energyconsumption, and enable the development of entirelynovel devices.224 Graphene is an attractive material forspintronic applications because of its tunable carrierconcentration and intrinsically low spin�orbit interac-tion and hyperfine couplings, which are expected tolead to long spin lifetimes.225 Here, graphene serves asan important test case for the expansion of spintronicsto other 2D materials beyond graphene.

    Many benchmarks toward spin functional gra-phene devices have already been met, the most im-portant of which is the demonstration of room-temperature gate-tunable spin transport, as shown inFigure 12a.226,227 These experiments used lateral spinvalves (Figure 12a, inset), which have ferromagneticelectrodes to preferentially inject spin-up or spin-down

    electrons. The spin diffuses from the injection site andcan be detected nonlocally by an electrode outside ofthe current path. Record spin lifetimes in graphene of6.2 ns228 at 20 K have been observed (Figure 12b).Essential to this achievement was overcoming the so-called conductance mismatch problem at the injectioninterface with the incorporation of a high-quality, half-monolayer tunnel barrier of MgO.229 Despite thesesuccesses, measured spin lifetimes in graphene areorders of magnitude lower than those expected fromab initio calculations.230 A central challenge in gra-phene spin-transport studies is to find the source ofthis additional spin scattering. Recent measurementsof spin transport in bilayer graphene spin valves as wellas those in single-layer graphene spin valves dopedwith adatoms indicate that local inhomogeneities inthe electronic environment are primarily responsiblefor the low measured spin lifetimes. If true, then whileunfortunate for applications requiring robust spincoherence this extreme sensitivity of spin transportto adsorbates and screening might be intentionallyexploited for novel functionality.

    The spin properties of graphene provide a frame-work for the exploration of spin functionality in other2D materials. MoS2 is a specific example of how thatfunctionality can be expanded. MoS2 exhibits strongspin�orbit coupling, which can significantly suppressspin scattering for a certain spindirection (Figure 12c).231

    Though similar to graphene in other ways, the strongspin coupling in MoS2 makes this material insensitive tospin scattering where graphene is sensitive. Achievingtunable spin scattering through creative combination ofgraphene and MoS2 could provide a spectrum of new

    Figure 11. DFT results calculated with HSE06 hybrid potentials for silicene functionalized with (a) H, (b) half H and halfhydroxyl (OH), (c) methane CH3, and (d) acetylene C2H.

    221

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2912

    spin functionality. Even more, MoS2 has the potential tolink spin and orbital degrees of freedom, combining twoof the most exciting new ideas for novel approaches toelectronics: spintronics and valley-tronics.

    Valley-tronics, as described above, is the manipu-lation of the population of degenerate low-energy

    valleys that are exhibited in the band structure of somematerials, including graphene and MoS2. Of particularrelevance for electronic transport, these valleys areseparated widely enough in momentum space thatintervalley scattering is strongly suppressed.231,232 Thepotential for computing with valley-tronics was first

    Figure 12. (a) Nonlocal magnetoresistance scans of single-layer graphene spin valves measured at room temperature. Theblack (red) curve shows the nonlocal resistance as the in-plane magnetic field is swept up (down). The nonlocal MR (ΔRNL) of130Ω is indicated by the arrow. Inset: Nonlocal spin transport measurement on this device with a spacing of L = 2.1 μm andSLGwidth ofW=2.2 μm.229 (b) Hanlemagnetoresistancemeasurement (with field applied out-of-plane) of a bilayer graphenespin valve at the charge neutrality point gate voltage measured at a temperature of 20 K (inset at 300 K). The spin lifetime(τs = 6.2 ns) anddiffusion constant (D=0.0047m

    2/s) are determinedby afit to theHanle curve.228 (c) (Left, top) Representationof the trigonal prismatic structure of monolayer MoS2 where each layer is a honeycomb lattice structure with each sublatticeoccupied by a Mo and two S atoms. (Left, bottom) The lowest-energy conduction bands and the highest-energy valencebands labeled by the z-component of their total angular momentum near the K and K0 point of the Brillouin zone. The spindegeneracy at the valence band edges is lifted by spin�orbit interactions. The valley and spin degrees of freedom arecoupled. Under left-circularly polarized excitation, only the K-valley is populated, whereas under right-circularly polarizedexcitation, only the K-valley is populated.156 (Right) Illustration of polarized spins (blue arrows) excited by right-circularlypolarized light (red beam) populating the K0-valley in an MoS2 sample. (d) Band lines near the Fermi level of BN/grapheneheterobilayers with specific stacking patterns (shown on the right), calculated using local density approximation functional.The interlayer distances are x = 5, 3.3, 3, 2.7, and 2.4 Å and the Fermi level is set to 0.240 Panel d reprintedwith permission fromref 233. Copyright 2011 American Institute of Physics.

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2913

    discussed for 2D electron gas devices, like AlAs andSi,233 but may be better realized with 2Dmaterials withappropriate band structures. Because inversion sym-metry is broken in MoS2, the valley and spin degrees offreedom are coupled and new optical selection rulesenable excitation of specific spin carriers into a specificvalley, as illustrated in Figure 12c.156,231 This spin-valleycoupling forbids independent spin or valley flipping,ensuring robust indices useful for spintronic and valley-tronic applications. There is likely a lot of functionalityyet to be discovered in 2Dmaterials beyond graphene.Not only in MoS2, just one of many transition metaldichalcogenides,148 but similar phenomena may alsobe exploited in non-MX2 materials such as silicene orgermanane.

    Finally, mechanically assembled heterostructuresof van der Waals bonded 2D materials might beexpected to show combined functionality of the in-dividual layers.234 The unique properties of the inde-pendent layers, which could include spin or chargedensity waves, superconductivity, and carefully chosenband structures, coupled with an anticipated high rateof electron tunneling between layers could lead toemergent properties.235 Not being limited by epitaxy,like traditional thin film heterostructures, a nearlyinfinite number of 2D material heterostructures withvarious properties can be imagined. Historically, hetero-structures such as semiconductor superlattices have ledto the development of novel technologies such asdielectric mirrors,236 microwave metamaterials,237 andnegative refraction index materials.238

    Emergent functionality in 2D material heterostruc-tures is anticipated theoretically. Density functionaltheory calculations predicted a band gap in gra-phene/BN bilayers.239 More recently, first principlescalculations show that the band gap and effectiveelectron mass in graphene/BN heterostructures mightbe tuned by the interlayer spacing and stacking ar-rangement of the individual layers, as shown inFigure 12d.240 Experimentally, graphene/BN hetero-structures have proved important for achieving highelectron mobility in graphene since the underlying BNlattice creates a better dielectric environment andreducing sensitivity to charged point defects in theunderlying SiO2.

    241,242 These heterostructures havealso led to an understanding of the observed “minimalconductivity” in graphene,243 which results fromcharge puddles created by the nearby SiO2 sub-strate.244,245 Additionally, tunneling field-effect transis-tors made from graphene/BN or graphene/MoS2 het-erostructures have been demonstrated to have highon/off ratios (up to 10 000).171 While these heterostruc-tures do not yet show emergent phenomena, theyrepresent important technological advances.

    Topological Insulators. On the surface of bulk ma-terials, either dangling bonds or band bending by anelectrostatic potential can result in the appearance of

    2D electronic states. Topological insulators (TIs) haveunusual bulk properties that allow them to accommo-date 2D electronic states on their surface.246 In TIs,strong spin�orbit effects create metallic electronicstates on the materials' surfaces, while the bulk of thematerial remains insulating. Furthermore, the surfacebands of TIs have an interesting electronic property inwhich the spin of a surface electron is always perpendi-cular to its momentum.247 This extraordinary spin prop-erty does not allow electrons to be backscattered byimpurities,248 suggesting electronic channels of highmobility.

    Following the first observation of this surface bandfrom a Bi�Sb alloy,249 many materials, especiallylayered van der Waals materials, have been reportedto be topological insulators. Layered binary chalcogen-ides, such as Bi2Se3, Bi2Te3, and Sb2Te3, are particularlyinteresting candidates for applications250 due to theirrelatively large bulk band gap (∼0.2 eV) and simpleband structure, which is similar to the Dirac band ofgraphene (Figure 13a,b). The existence of surfacestates in this family of materials is confirmed experi-mentally251�254 (Figure 13c). However, the surfaceelectron transport is not yet straightforward to observemainly due to the large contribution of bulk electrons.Imperfections in the bulk crystal structure create vastamounts of additional electrons, overwhelming thesurface electron transport.253,254 Therefore, the TIsneed to be engineered to reduce the bulk contributionin order to achieve transport dominated by surfaceelectrons.

    Lowering the dimensionality of these TIs mightprovide a route to study their surface state. Reducingsample size can suppress bulk carriers by increasing asurface-to-volume ratio.254 Also, low-dimensional TIswill allow the electron chemical potential to be modu-lated by an electric field, which can also reduce bulkcarriers. TI nanodevices can be prepared from CVD-grownnanostructures254�256 (Figure 13d,e) and via themechanical exfoliation of single crystals (Figure 13f).257

    Field-effect measurements have successfully demon-strated themodulationof thecarrier typesbygate voltage,showing a potential for electronic applications.255�257

    To control bulk electrons in TIs, compensationaldoping253,255,256 and proper surface encapsulation256

    are crucial.The unusual properties of the TI surface states

    suggestmany applications not feasible in conventionalmaterials. The intrinsic spin properties of TIs are attrac-tive for spintronics applications.258 Also, superconduct-ing TIs are expected to generate exotic quantummodes, potentially useful for quantum computa-tion.258,259 Additionally the surface states can play arole in energy science by enhancing thermoelectriceffects in Bi2Te3 and Bi2Se3, which are well-knownthermoelectric materials.260 Finally, this newly discov-ered electronic state, often called “graphene with a

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2914

    spin texture”, is a platform for a number of interestingstudies and applications.

    Thermal and Thermoelectric Properties. Recent ad-vances in the synthesis and isolation of graphene andother atomic layered materials have enabled funda-mental studies of unique thermal and thermoelectrictransport properties of 2D systems, as well as thepotential applications of these 2Dmaterials for thermalmanagement and thermoelectric energy conversion.Because of the absence of interlayer phonon scatter-ing, the thermal conductivity of an ideal 2D system,such as suspended SLG or h-BN, is expected to behigherthan the basal-plane values of their three-dimensionalstacks, namely, graphite and bulk h-BN. Although theresults from several micro-Raman measurements261,262

    of suspended SLG are in agreement with these theore-tical predictions, other Raman measurements263,264 ofsuspended SLG samples have yielded values compar-able to or lower than the basal-plane values of graphite,as shown in Figure 14. In addition, further experimentsare needed to verify recent theoretical predictions thatflexural phonons (out-of-plane modes) make a largecontribution to the thermal conductivity in suspendedfew-layer graphene and h-BN.265,266

    In addition, both theoretical and experimental stud-ies have foundconsiderable reductionof thebasal-planethermal conductivity in graphene or h-BN supportedon or encased in an amorphous material267�269 orcontaminated by polymer residue,270 as shown inFigure 14. In particular, the interface interaction withan amorphous material has been found to decreasethe basal-plane thermal conductivity with decreasinglayer thickness of FLG268,271 or suspended few-layerh-BN with polymer residue.272 This trend is opposite tothat suggested for suspended FLG or h-BN.261,265,266

    These findings indicate that the basal-plane values forhigh-quality graphite and h-BN bulk crystals present apractical limit in the thermal conductivity of few-layergraphene and h-BN when these 2D layered materialsare used as heat spreaders in contact with a heat sourceand a heat sink or as nanofillers to increase the effectivethermal conductivity of a composite. However, evensupported SLG can possess a basal-plane thermal con-ductivity higher than that of copper.267 Moreover,further theoretical and experimental research may po-tentially find new approaches to increasing the basal-plane thermal conductivity of supported graphene andh-BN by choosing the right support material or tuning

    Figure 13. (a) Electronic structure of Bi2Se3 reconstructed by angle-resolved photoemission spectroscopy (ARPES), showingbulk conduction band (upper pocket), bulk valence band (lower pocket), and surface-state band (linear band).251 Reprintedwith permission from ref 251. Copyright 2009Nature PublishingGroup. (b) Crystal structure of Bi2Se3 (layered structure). Eachquintuple layer is made of five atomic layers of bismuth (Bi) and selenium (Se).254 (c) Magnetoresistance of a Bi2Se3nanoribbon in parallel magnetic field, showing periodic oscillations from the Aharonov�Bohm interference of the surfaceelectrons.254 (d) Optical microscopy image of topological insulator (BixSb2�xTe3) nanoplates (5�10 nm thickness) on SiO2substrate.255 The scale bar indicates 5 μm. (e) Scanning electronmicroscopy image of topological insulator (Bi2Se3) nanowiresand nanoribbons (50�200 nm thickness).256 The scale bar indicates 10 μm. (f) Optical microscopy image of a topologicalinsulator (Bi2Se3) flake on SiO2 substrate from mechanical exfoliation of a bulk crystal.

    257 The scale bar indicates 2 μm.Reprinted with permission from ref 257. Copyright 2011 Nature Publishing Group.

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2915

    the interface roughness and interaction. In particular,superior electron mobility has been found in graphenesupported on h-BN compared to graphene supportedon amorphous SiO2.

    241 It remains to be investigatedwhether graphene supported on atomically smoothh-BNmay possess superior thermal transport propertiescompared to graphene supported on SiO2.

    The interface thermal resistance is as important inthermal transport as the basal-plane thermal conduc-tivity of 2D layeredmaterials. Although the cross-planethermal conductivities of graphite and h-BN are muchsmaller than their basal-plane values, the cross-planediffusive thermal resistance of few-layer graphene andh-BN is expected to be smaller than the interfacethermal resistance because of the small thickness.The thermal interface resistance of graphene in contactwith a dielectric or metal has been measured to be inthe range of (1�5) � 10�8 K m2 W�1 at room tem-perature.273�276 These values are not very large for thevan der Waals interface, likely because graphene israther conformal to the surface roughness.277 In addi-tion, interface phonon transport between highly ani-sotropic graphene and a dissimilar material remains aninteresting subject for theoretical study.278

    Compared to graphene that is a semimetal, h-BNis an insulator, which can be advantageous for itsapplications as a heat spreader in direct contact withhigh-power density semiconductor nanodevices. More-over, other 2D layered materials such as MoS2,

    133

    silicene,53 and germanane provide different bandgap values suitable for various functional devices. Inthese functional devices made of 2D layered materials,local heating can become an important issue similar tothe challenge faced by silicon nanoelectronic devices,especially if the in-plane thermal conductivity of these2D materials is suppressed by phonon scattering byinterface disorder or interactions with another materi-al. However, little is currently known about the thermaltransport properties of these 2D materials and hetero-structure devices171 made of multiple 2D materials.

    The possibly suppressed thermal conductivity (κ) of2D layered materials with surface disorder or perturba-tion is desirable for thermoelectric materials. Duringthe past two decades, nanostructures have been ac-tively investigated for suppressing the phonon con-tribution to the thermal conductivity of thermoelectricmaterials.279 In such efforts, it is critical to control theinterface charge states and scattering in the nanos-tructured materials, so that these effects do not resultin a considerable reduction of the charge mobility,power factor, S2σ, and figure of merit, Z = S2σ/κ, whereS is the Seebeck coefficient and σ is the electricalconductivity.280 Moreover, research along this direc-tion was initially stimulated by theoretical predictionsof power factor enhancement in 2D and one-dimen-sional quantum structures.281,282 Recent theoreticalstudies have further suggested that the power factorcan be enhanced by the coupling of the two topolo-gical surface states in few quintuple layers of Bi2Te3and Bi2Se3.

    283 A careful control of the surface qualityand surface band bending will be required for theseeffects due to quantum confinement and topologicalsurface states to be verified in experiments. Thesechallenges suggest abundant opportunities for re-search in thermal and thermoelectric properties of2D materials beyond graphene.

    Applications to Transistor Scalability. Progress inthe semiconductor industry has been driven by acontinuous reduction of the transistor geometry. Cur-rent silicon-based field-effect transistors (FETs) feature22 nm gate lengths, and further size reductions moti-vate the need to search for alternative device conceptsand materials. A conventional FET consists of a semi-conducting channel that is connected to source anddrain electrodes and can be switched on and off by theapplication of a field via a gate electrode that isseparated by a dielectric material (Figure 15a). Desir-able features of switching transistors include high on/off current ratios, low power consumption, and fastswitching times.

    Two-dimensional materials have an innate advan-tage over bulk materials, like Si, for device scalabilitybecause of their favorable electrostatic properties. Inparticular, due to their dimensionality, these materialscan approach the ideal effective screening length, λ, andcan be operated beyond the quantum capacitance

    Figure 14. Comparison of reported experimental thermalconductivity values of graphite, graphene, and h-BN. Thesuspended SLG data are from Balandin et al.,319 Chenet al.,262 Lee et al.,264 and Faugeras et al.263 The data forSLG and 8-layer graphene (8LG) supported on SiO2 are fromSeol et al.267 and Sadeghi et al.,271 respectively. The data forpyrolytic graphite (PG) are the recommended values byTouloukian.320 The bulk h-BN data are from Sichel et al.321

    The data for the 11- and 5-layer suspended h-BN withpolymer residue are from Jo et al.272

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2916

    limit (QCL). To reduce the channel length in today'stypical silicon-on-insulator transistor, the semiconduc-tor thickness must also be reduced to ensure that 2Delectrostatic screening effects are minimized.284 Elec-trostatic potential variations along the channel arescreened out within a material-dependent λ.285,286 Toprevent the electric field of the drain from being totallyscreened in a small channel, like those of ultrascaledMOSFETs, it is desirable for λ to be as small as possible.While the exact form of λ depends on the details of thedevice structure, for single and double gate geome-tries, it can be approximated by285

    λ ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiεsdsdoxεox

    s(1)

    Here, εs and ds denote the dielectric constant andthickness of the semiconductor, respectively, and εoxand dox are the respective quantities for the dielectricoxide. In either case, minimizing ds will minimize λ.Thus, single-layer-thick 2Dmaterials represent the bestpossible scenario that nature has to offer.

    A second advantage of 2Dmaterial-based transistors istheir ability to operate beyond the QCL, the limit at whichthe oxide capacitance is equal to the quantum capaci-tance, the capacitance due to gate-voltage-induced freecharges in the channel.287,288 In the QCL regime, theintrinsic gate delay is reduced (τ), and the P 3 τ (where P isthe dynamic power) nowdecreases linearly with channel

    length in the QCL regime.289,290 Thus the energy con-sumption required for switching isminimized in the QCL.In addition, many 2D materials from the dichalcogenidefamily, and even silicane andgermanane, possess a largerband gap than bulk Si. This property could result in amuch lower direct source-to-drain leakage current, whichis especially important for ultrascaled transistors.291

    Recognizing the potential impact of 2D materialsfor transistor applications, a significant amount of workhas been devoted to making these transistors in thepast few years.163,167,292�296 Recently, the performanceof MoS2 transistors has been theoretically examined

    164

    using real space, self-consistent, and fully parallel non-equilibrium Green's function transport simulationwithin a k 3 p approach and band structure fitted to firstprinciples calculations. Figure 15b shows the draincurrent (ID) versus gate voltage (VG) characteristic fora simulated MoS2 transistor. The maximum ON currentcan be as high as 2 mA/μm, and the OFF current couldbe as small as 10 fA/μm, thereby giving an incrediblylarge ON/OFF ratio (∼1010). Figure 15c shows thesubthreshold swing as a function of gate oxide thick-ness. These simulation results show fantastic immunityto short channel effects. Note that almost an idealsubthreshold swing (60 mV/decade) is achievable witha reasonably large gate oxide thickness of around 5 nmeven for a 30 nm channel length device. This result canbe attributed to the fact that the screening length λ isvery small as expected for a 2D material. Figure 15d

    Figure 15. (a) Schematic illustration of cross section of MoS2 FET.163 Reprinted with permission from ref 163. Copyright 2011

    Nature Publishing Group. (b) Transfer characteristic of a monolayer MoS2 transistor in both log and linear scales. (c)Subthreshold swing as a function of gate oxide thickness for two different gate lengths. (d) Capacitance vs voltagecharacteristics. Inset shows the change in surface potential ψs with gate voltage Vg.

    164

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2917

    shows the capacitance�voltage characteristic. Notethat the total capacitance (shown in blue) is muchsmaller than the oxide capacitance (black dashed line)and closer to the quantum capacitance (shown in solidred). This means that the device is working close to theQCL. These simulations confirm the fantastic electro-static control, QCL operation, and low leakage currentfor a 2D material.

    Multilayer Assemblies and Their Applications. Nano-sheets (Figure 5), referring to “restacked” or assembled2D materials, provide new opportunities and advan-tages for devices. For many applications, these nano-sheets and their composites have inherent and pre-dicted advantages: they can be solution processed,they can display new behavior due to their extremelyhigh surface area, and as free-standingmultilayer assem-blies, they are flexible119,297�299 and can bemanipulatedeither mechanically or by deposition on templates. Mix-ing of nanosheets with other materials, including metalnanoparticles, polymers, and small molecules, can be aroute to interesting solution-phase nanocomposite hy-brids, such as superlattice-like assemblies.99�101

    Reassembly of nanosheets into multilayer films canlead to improved properties in the areas of super-capacitors,300 pseudocapacitors,301 photoconductivematerials,302,303 and heterojunction photodiodes304

    as new magnetic materials305 and as magneto-opticalcomponents.306,307 Additional applications include thefollowing:

    Batteries: The morphological and size advantagesof nanosheets have shown improved performance asbattery electrodes. For example, reassembled octatita-nate nanosheets have better reversible capacities thanelectrodesmade frombulk octatitanate, presumably dueto the ability of these nanosheets to better withstanddamage caused by lithium insertion and extraction.300,308

    Magnetic Properties: Ferromagnetic nanosheetscan display unique properties such as anisotropicmag-netization and stronger influences fromsurface atoms.309

    For instance, gigantic magneto-optical effects have beenobserved in Ti0.8Co0.2O2 nanosheets. Because thesenanosheets are composed entirely of surface atoms,surface cobalt spins and their local ferromagnetic cou-plings are much stronger than those that occur in dilutemagnetic semiconductors and bulk magnets.309

    Photoconductors: Restacked nanosheets havehigher surface area than their bulk precursors, and thisaspect, coupled with the finite thickness effects onelectronic structure, the reactivity, and the open frame-work, can lead to enhancements in nanosheet photo-activity in solar energy conversion111 and in capac-itance.301 Furthermore, it has been recently demon-strated that a vertically oriented 2D organic/inorganiclamellar photodetector architecture can have bet-ter detectivities than sensitized nanoparticle thinfilms.310 This is because the excellent 2D conductionchannel allows for reduced majority carrier transit

    times, and the atomic-scale thickness of the 2D ma-terial enables a high sensitization ratio.

    High Dielectric Constant Materials: As high-k di-electrics, nanosheets can be used where thin filmsfrom vapor deposition cannot. Metal oxide thin filmsgrown by conventional vapor deposition techniquesoften show decreasing dielectric constants with de-creasing film thickness.118,311,312 One reason for thedegradation in dielectric properties is the high tem-peratures that complex vapor processing requires, oftenabove 600 �C, which reduces the polarizability due tononstoichiometry.312 Free-standing and stacks of nano-sheets, however, have been shown to have low leakagecurrent densities and high dielectric constants for thick-nessesdownto10nm(Figure16a).117,118,312Reassemblednanosheets also have the advantage of electronic isola-tion of each nanosheet in these multilayer assemblies.309

    Liquid Crystals: Inorganic magnetic nanoparticlesof magnetite have been mixed with niobate nano-sheets and placed in a magnetic field to guidethe nanoparticles' orientation.111,313 Light scatteringshows orientational dependence on themagnetic field(Figure 16b). Potassium niobate nanosheets have beensuccessfully functionalizedwith silane and dispersed inorganic solvents to create liquid crystals.314 The nano-sheet colloids were dispersed in chloroform and ex-hibited birefringence characteristics of a nematicphase.111,314 Liquid-crystal properties of nanosheets havealso been reported for nanosheets made of niobates315

    andKCa2Nb3O10,316 amongothers. The lateraldimensions,

    aspect ratio, and colloid concentration of the nano-sheets strongly dictate the liquid-crystal properties.315

    OPPORTUNITIES AND OUTLOOK

    In this nascent field of 2D materials, the salientdifferences between bulk and few or single layers arejust starting to be understood. New fundamentalstudies can address single-layer scale differences inmany-body interactions; phonon transport, such asflexural phonon modes; interfacial electron�electron,electron�phonon, electron�magnon, etc. coupling;excitonic and other quasiparticle properties; the nature

    Figure 16. (a) High-k properties of titania nanosheet assem-blies compared to conventional thin films. The values forthe multilayered titania nanosheets are larger than whathas been reported for any other high-k material.312 Rep-rinted with permission from ref 312. Copyright 2006 JohnWiley and Sons. (b) Optical micrograph of a liquid crystalmade fromFe3O4�Ca2Nb3O10 nanosheets in pyridine underan inhomogeneous magnetic field.313

    REVIEW

  • BUTLER ET AL. VOL. 7 ’ NO. 4 ’ 2898–2926 ’ 2013

    www.acsnano.org

    2918

    and impact of defects and the substrate; the influenceof high doping, strain effects, and electric fields; me-chanical properties; quantum size effects; and edgeeffects in transport. At the single-layer scale, interac-tions with the surroundings often dominate experi-mental observations, and consequently, there are amultitude of unexpected experimental challenges inisolating the intrinsic material properties. There arenumerous exciting opportunities in developing thegrowth of high-quality large-area materials with con-trollable layer thicknesses. Such systems would not onlyexpand our understanding of the underlying physicsbut potentially lead to the discovery of unanticipatedphenomena andapplications. Already, newunderstand-ing of the 2D materials has contributed to entirely newscientific frontiers such as spin- and valley-tronics. Theability to harness such unique properties and phenom-ena will surely lead to exciting technological advances.

    Conflict of Interest: The authors declare no competingfinancial interest.

    Acknowledgment. The topics mentioned here were dis-cussed at a workshop entitled “2DMaterials beyond Graphene”held at the Ohio State University in August 2012 and sponsoredby the Army Research Office (under Award Number W911NF-12-1-0226) and the Institute for Materials Research at The OhioState University. J.E.G. additionally wishes to acknowledge SteveRingel and Chakrapani Varanasi for conceptual and logisticalsupport. J.E.G. also wishes to acknowledge support from theArmy Research Office (W911-NF-12-1-0481) and National ScienceFoundation (DMR 1201953). In the preparation of this review, wehighlight a recently published review focused on the electronicsof single-layer metal dichalcogenides.148

    REFERENCES AND NOTES1. Wilson, J. A.; Yoffe, A. D. Transition Metal Dichalcogen-

    ides Discussion and Interpretation of Observed Optical,Electrical, and Structural Properties. Adv. Phys. 1969, 18,193–335.

    2. Bednorz, J. G.; Muller, K. A. Possible High-Tc Supercon-ductivity in the Ba-La-Cu-O System. Z. Phys. B 1986, 64,189–193.

    3. Kamihara, Y.; Hiramatsu, H.; Hirano, M.; Kawamura, R.;Yanagi, H.; Kamiya, T.; Hosono, H. Iron-Based LayeredSuperconductor: LaOFeP. J. Am. Chem. Soc. 2006, 128,10012–10013.

    4. May, J. W. Platinum Surface LEED Rings. Surf. Sci. 1969,17, 267–270.

    5. Novoselov, K. S.; Geim, A. K.; Morozov, S. V.; Jiang, D.;Zhang, Y.; Dubonos, S. V.; Grigorieva, I. V.; Firsov, A. A.Electric Field Effect in Atomically Thin Carbon Films.Science 2004, 306, 666–669.

    6. Stankovich, S.; Dikin, D. A.; Dommett, G. H. B.; Kohlhaas,K. M.; Zimney, E. J.; Stach, E. A.; Piner, R. D.; Nguyen, S. T.;Ruoff, R. S. Graphene-Based Composite Materials. Nature2006, 442, 282–286.

    7. Kim, K. S.; Zhao, Y.; Jang, H.; Lee, S. Y.; Kim, J. M.; Kim, K. S.;Ahn, J.-H.; Kim, P.; Choi, J.-Y.; Hong, B. H. Large-ScalePattern Growth of Graphene Films for Stretchable Trans-parent Electrodes. Nature 2009, 457, 706–710.

    8. Yang, H.; Heo, J.; Park, S.; Song, H. J.; Seo, D. H.; Byun, K. E.;Kim, P.; Yoo, I.; Chung, H. J.; Kim, K. Graphene Barristor, aTriode Device with a Gate-Controlled Schottky Barrier.Science 2012, 336, 1140–1143.

    9. Frindt, R. F. Single Crystals of MoS2 Several Molec