Subscriber access provided by UNIV OF WESTERN ONTARIO
Communication
Intrinsic disorder in graphene on transition metal dichalcogenide heterostructures Matthew Abraham Yankowitz, Stefano Larentis, Kyoung Kim, Jiamin Xue, Devin McKenzie, Shengqiang Huang, Marina Paggen, Mazhar Nawaz Ali, Robert J. Cava, Emanuel Tutuc, and Brian J LeRoy Nano Lett., Just Accepted Manuscript • Publication Date (Web): 09 Feb 2015 Downloaded from http://pubs.acs.org on February 10, 2015
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Nano Letters is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 16
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
Intrinsic disorder in graphene on transition metal dichalcogenide heterostructures Matthew Yankowitz,† Stefano Larentis,‡ Kyounghwan Kim,‡ Jiamin Xue,‡ Devin McKenzie,† Shengqiang Huang,† Marina Paggen,†,¶ Mazhar N. Ali,§ Robert J. Cava,§ Emanuel Tutuc,‡ and Brian J. LeRoy∗,† Physics Department, University of Arizona, Tucson, AZ 85721, USA, Microelectronics Research Center, The University of Texas at Austin, Austin, TX 78758, USA, Physics Department, University of Texas at El Paso, El Paso, TX 79968, USA, and Department of Chemistry, Princeton University, Princeton, New Jersey 08544, USA E-mail: [email protected]
Abstract Semiconducting transition metal dichalchogenides (TMDs) are a family of van der Waals bonded materials that have recently received interest as alternative substrates to hexagonal boron nitride (hBN) for graphene, as well as for components in novel graphene-based device heterostructures. We elucidate the local structural and electronic properties of graphene on TMD heterostructures through scanning tunneling microscopy and spectroscopy measurements. We find that crystalline defects intrinsic to TMDs induce substantial electronic scattering and charge carrier density fluctuations ∗
To whom correspondence should be addressed University of Arizona ‡ University of Texas at Austin ¶ University of Texas at El Paso § Princeton University †
1 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
in the graphene. These signatures of local disorder explain the significant degradation of graphene device mobilities using TMD substrates, particularly compared to similar graphene on hBN devices.
Keywords scanning tunneling microscopy/spectroscopy, graphene, transition metal dichalchogenides, defects, scattering
The electronic properties of two-dimensional materials such as graphene are extremely sensitive to their environment, especially the underlying substrate. Since the isolation of graphene in 2004, considerable effort has been put into finding the best substrates, both from a device standpoint and for inducing novel physical phenomena. Perhaps the most common substrate, SiO2 , causes out-of-plane ripples and locally dopes the graphene due to trapped charged impurities. 1 Suspended graphene devices, fabricated by etching away the SiO2 layer, offer the best intrinsic graphene quality, although their fabrication is far too challenging to scale to industrial levels. Hexagonal boron nitride has emerged as a very promising substrate; as an insulating crystal it not only flattens the graphene but screens underlying charge impurities from the base substrate. 2–5 Careful device fabrication techniques can yield devices of graphene quality nearing that of suspended graphene, and these heterostructures are more friendly for industrial scaling. Since hBN has a similar lattice constant to graphene, when the two lattices are in near perfect alignment interactions between the crystals strongly renormalize the graphene band structure. 5,6 This opens an avenue for the study of interesting phenomena, such as the Hofstadter quantization, 7–9 and also provides a route to make graphene insulating. 5 However, it may not be ideal for large scale graphene device applications where fabrication leads to a random alignment between the crystals, and the intrinsic 2 ACS Paragon Plus Environment
Page 2 of 16
Page 3 of 16
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
graphene band structure needs to be preserved. Recently, transition metal dichalcogenides have made a strong resurgence in materials research, as these crystals can be exfoliated to atomic scale thicknesses and stacked via van der Waals interactions similarly to graphene and hBN. 10,11 A subset of the TMDs exhibit similar semiconducting behavior, with indirect band gaps in bulk ranging from ∼ 1 - 1.4 eV. 10 Na¨ıvely, these materials, when insulating, should offer comparable quality to hBN as substrates for graphene, but without the possibility of band structure modification due to their considerably different lattice constants. 12,13 Additionally, their semiconducting nature permits dynamic gate voltage control over the interaction strength with graphene, 12 and they offer the potential for the study of new physical phenomena (for example, potential spin-orbit coupling induced in the graphene layer due to the heavy metal atoms of the TMD 14 ). From a device standpoint, there are numerous potential applications involving heterostructures between graphene and TMDs; for example, as tunneling transistors, 15–18 highly efficient flexible photovoltaic devices, 19,20 or nonvolatile memory cells. 21,22 Unfortunately, graphene on TMD devices have thus far been of significantly lower mobility than comparable device structures utilizing an hBN substrate, 14,23,24 and a local understanding of this behavior is lacking. In this Letter, we show via local scanning probe measurements that graphene on TMD devices suffer an unavoidable degradation in electronic quality due to defects in the TMD crystals. We study graphene on substrates belonging to the MX2 family, where M is a transition metal (Mo, W) and X is a chalcogen atom (S, Se, Te). The specific TMDs examined here are MoS2 , WS2 , WSe2 , and MoTe2 . We have also examined graphene on SnS2 , which is not technically a TMD but shares the same crystal structure and is also a semiconductor. 25,26 Figs. 1(a) and (b) show topography images of graphene on MoS2 and WSe2 obtained via scanning tunneling microscopy (STM). Numerous defects can be observed in the MoS2 sample, characterized by highly localized topographic height variations (both positive and negative, depending on tunnel parameters). The graphene lattice is continuous over these defects,
3 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
indicating the defects reside in the MoS2 and are presumably either atomic substitutions or vacancies. The strength and appearance of these defects can be tuned by the sample and gate voltages, and their appearance in other TMDs is found to depend on those factors as well as crystal thickness. The graphene on MoTe2 and SnS2 samples exhibit significantly worse topography, marked by islands of well-adhered regions surrounded by significantly rougher regions where the graphene did not appear to be adhered to the TMD, suggesting that these crystals are not stable enough in air to allow consistently good adhesion with the graphene 11 (see Supporting Information for further discussion of defects and topography observations). The insets of Figs. 1(a) and (b) show atomic resolution images of both samples, and additionally display hexagonal superlattices due to the interference pattern formed by the graphene and TMD lattices. As is the case for graphene on hBN, a moir´e pattern is expected to form between the graphene and TMD lattices due to their relative rotation φ and lattice mismatch δ. The moir´e wavelength is
λ= q
(1 + δ)a 2(1 + δ)(1 − cos φ) + δ 2
where a is the graphene lattice constant. 6 Because the lattice mismatch is much larger between graphene and TMDs than graphene and hBN, the range of possible moir´e wavelengths is much smaller. The inset of Fig. 1(d) shows the dispersion of moir´e wavelengths as a function of angle φ for graphene on MoS2 and WSe2 (the dispersions for the other TMDs studied here are very similar). The lattice constants of the TMDs in this study range from 3.15 to 3.64 ˚ A, 13,25 and thus the possible moir´e wavelengths are on the order of 0.5 nm to just over 1 nm. As a result of these short moir´e wavelengths, no modification of the graphene band structure is expected at low energies. 6 Figure 1(c) shows a Fourier transform of the inset of Fig. 1(a), with the superlattice (yellow/green) and graphene lattice (red) hexagons highlighted. All samples studied show good agreement between the moir´e wavelength and the angle between the lattices.
4 ACS Paragon Plus Environment
Page 4 of 16
(a)
(b)
(c)
(d) 4 Conductivity (mS)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
Wavelength (nm)
Page 5 of 16
3
1.2 1.0 0.8 0.6 0.4 -30
0 ø (degrees)
30
2 G/WSe2 G/MoS2
1 0
-40
-20 0 20 Gate Voltage (V)
40
Figure 1: (a) Topography of graphene on MoS2 . Arrows mark some of the defects buried in the MoS2 . Inset: Atomic resolution with a moir´e pattern of 0.65 nm. (b) Same as (a) for WSe2 . No defects are visible. Inset: Atomic resolution with a moir´e pattern of 0.91 nm. For (a) and (b), the scale bar for the main figure is 10 nm and for the insets it is 1 nm. Typical imaging parameters are sample voltages between Vs = 0.1 V and 0.3 V and tunnel currents between It = 100 pA and 300 pA. (c) Fourier transform of (a). The atomic lattice is highlighted with a red hexagon, and the moir´e with a yellow hexagon. The moir´e points surrounding the atomic lattice are highlighted with green hexagons. All extra resonances are the result of noise in the system. The scale bar is 10 nm−1 . (d) Global conductance measurements of graphene on MoS2 and WSe2 devices. The MoS2 becomes conducting at gate voltages just above 0 V, after which point the device exhibits negative compressibility. 23 The WSe2 crystal is always biased within the band gap. Inset: Dispersion of possible moir´e wavelengths for graphene on MoS2 and WSe2 as a function of the relative rotation between the lattices. In addition to their similar crystal structure, these TMDs also share similar electronic properties. They are indirect gap semiconductors with band gaps in bulk ranging from about 1.0 eV to 1.4 eV 10 (and 2.2 eV for SnS2 26 ). In proximity to graphene, the difference in energy between the graphene work function and the electron affinity of the TMD determines the relative band alignment. The threshold gate voltage at which the TMD begins to conduct depends on this band alignment as well as the band bending related to the TMD crystal thickness. 23 For comparably thin crystals (∼5 - 15 nm), we observe that the Dirac point is
5 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
aligned very near the conduction bands of MoS2 , MoTe2 and SnS2 , and closer to the middle of the gap for WS2 and WSe2 . Figure 1(d) shows global conductance measurements at 4.2 K for graphene on MoS2 and WSe2 . The graphene on WSe2 device is nearly symmetric around charge neutrality (indicating the WSe2 always remains insulating), whereas the MoS2 device has a saturating conductance at positive gate voltages as the MoS2 becomes conducting. All devices studies exhibit similar transport characteristics to those in Fig. 1(d). We find field-effect mobilities of 5,000 - 10,000 cm2 /Vs in these devices, which is over an order of magnitude lower than of comparable graphene on hBN devices. We restrict our study to thin TMD substrates as devices with thicker crystals become strongly hysteretic with gate voltage. Figures 2(a) and (b) show normalized (dI/dV)/(I/V) spectroscopy measurements as a function of gate voltage for graphene on MoS2 and WSe2 . For MoS2 , the Dirac point (marked with a white dotted line) essentially stops moving in energy once the gate voltage is large enough to induce charge carriers in the TMD. In some samples we have additionally observed the Dirac point move with positive dVs /dVg when the MoS2 is conducting, indicative of the negative compressibility of the system 23 (see Supporting Information). For graphene on WSe2 , the Dirac point moves with the standard square root of gate voltage dispersion expected for graphene on an insulating substrate, 27 indicating the WSe2 substrate is never biased to the valence or conduction bands. In this case, the WSe2 should behave very similarly to an hBN substrate, flattening the graphene and screening charged impurities in the underlying SiO2 . For all the graphene on TMD devices, the movement of the Dirac point is well fit by a Fermi velocity of 0.95 ± 0.05 x 106 m/s. We additionally observe weak nondispersive features characteristic of phonon-mediated tunneling into graphene (most notably the depressed dI/dV surrounding zero sample bias) as well as features related to electronplasmon interactions (dispersive inflection point at energies slightly above the Dirac point). 28 The Supporting Information provides further identification and discussion of these features. In addition to the spectroscopy features discussed above, we observe two types of features
6 ACS Paragon Plus Environment
Page 6 of 16
Page 7 of 16
(b) Sample Voltage (V)
0.2 0.1 0.0 -0.1 -0.2 -0.3
(c)
-50
0
Gate Voltage (V)
2.0 G/MoS2 G/WSe2
(dI/dV)/(I/V)
1.8 1.6 1.4 1.2 1.0
-0.1
0.0
0.1
0.2
Sample Voltage (V)
0.2 0.1 0.0 -0.1 3
-0.2
0.3
-50
0
2
50
Gate Voltage (V)
(d)
0.8 -0.2
0.3
-0.3
50
Sample Voltage (V)
Sample Voltage (V)
(a) 0.3
0.3
1
(dI/dV)/(I/V)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
0.2 0
0.1 0.0 -0.1 -0.2 -0.3
0
10 20 30 Position (nm)
40
Figure 2: (a) Normalized (dI/dV)/(I/V) spectroscopy as a function of gate voltage for graphene on MoS2 . The Dirac point is highlighted with a dotted white line. The Dirac point remains nearly stationary with gate voltage above ∼+10 V as the MoS2 becomes conducting and the charge density in the graphene is constant. States due to charging of defects in MoS2 are marked with black arrows. (b) Same as (a) for graphene on WSe2 . In both (a) and (b), extra peaks of varying strength due to scattering surround the Dirac point at both positive and negative energies. The nearly vertical bright features near zero gate voltage result from confinement by p-n junctions. 29 (c) Cuts of (a) and (b) (marked by the dashed vertical lines) at Vg = -20 V. For both, the Dirac point is around Vs = +0.15V (marked with black diamonds). Extra states due to scattering are marked with tick marks. (d) Normalized dI/dV spectroscopy as a function of position in graphene on MoTe2 at Vg = -9 V. The approximate position of the Dirac point is marked with a dotted white line. All other resonances are attributed to intravalley scattering, and their energy spacing varies randomly with position. which are not characteristic of pristine graphene. Ubiquitous amongst all MX2 substrates examined in this study, we observe extra resonances surrounding and moving roughly in parallel with the Dirac point. Fig. 2(c) shows a line cuts of Figs. 2(a) and (b) at Vg = -20 V, indicating clearly the presence of these extra states, marked with black ticks (see the Supporting Information for a derivative of Fig. 2(b), which also highlights these satellite states). Additionally, unique to MoS2 substrates, we observe strong features which move oppositely of the Dirac point with gate voltage (marked with black arrows in Fig. 2(a)) which we will discuss later.
7 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
The resonances which track in parallel the Dirac point are typically characteristic of electronic scattering in graphene, 29 although we have considered other origins for the extra states we observe, such as new features of the band structure due to interactions with the substrate or the excitation of phonons in the TMD substrate by electrons tunneling into the graphene. Such features should have well-defined energies which do not vary spatially or between different samples, depending on the exact nature of their origin. To test this hypothesis we take line maps of normalized dI/dV spectroscopy, as shown in Fig. 2(d) for graphene on MoTe2 . We see clearly that the energy spacing of these states varies spatially, and in some cases the states split or merge. While this behavior is inconsistent with the alternative explanations considered above, it is consistent with electronic scattering in graphene from point and line defects in the TMD substrate and/or the remnant disorder potential resulting from those defects. 29 To further investigate the intravalley scattering in graphene on TMD heterostructures, we take large area dI/dV maps of graphene on WSe2 at various gate voltages (see Figs. 3(a) - (c) for three examples). The coherent features in the maps change size with gate voltage, characteristic of intravalley scattering from the Coulomb potentials of buried defects. Fig. 3(d) shows the Fourier transform of Fig. 3(c). The disk at the center of the image arises from 2k scattering across a single graphene Dirac cone. For each map, we take a circular average of the Fourier transform and extract the corresponding wave vector k as the half-width at half-maximum of the best-fit Lorentzian. These wave vectors are plotted as a function of energy relative to the Dirac point in Fig. 3(e). The extracted points agree with a linear dispersion characterized by a Fermi velocity of ∼0.95 x 106 m/s. Fig. 4(a) shows a similar dI/dV map of graphene on MoS2 , which exhibits numerous ring features due to the charging or discharging of defect states resulting from the interaction of MoS2 defects with graphene (similar to those seen in artificial impurities on graphene 30 ). These charging rings are the same features which run oppositely of the Dirac point in the gate map (features marked with black arrows in Fig. 2(a)). The size and nature of the rings
8 ACS Paragon Plus Environment
Page 8 of 16
Page 9 of 16
(a)
(d)
(b)
(c)
(e)
0.4
0.2
Energy (eV)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
0.0
-0.2
-0.4 -0.4
-0.2
0.0 -1 k (nm )
0.2
0.4
Figure 3: (a) - (c) Maps of dI/dV spectroscopy for graphene on WSe2 at Vg = -60 V, -5 V, and +60 V, respectively. The local charge neutrality point is around Vg = -15 V. Each map is taken with Vs = 50 mV and It = 150 pA. The coherent structure is due to intravalley scattering, and becomes shorter wavelength at gate voltages further from the Dirac point. The scale bars in (a)-(c) are 50 nm. (d) Fourier transform of (c), exhibiting a disk-like feature at the center corresponding to the scattering wave vector. The circular average is plotted in red. The scale bar is 0.2 nm−1 . (e) Graphene energy versus momentum dispersion extracted from maps similar to those shown in (a) - (c). The wave vector k is extracted from the half-width at half-maximum of the Lorentzian fit of the Fourier transforms of each map. The energy is determined by converting the sample voltage of each measurement to energy from the Dirac point using vF = 0.95 x 106 m/s. The solid blue lines plot the dispersion E=h ¯ vF k. The error in k represents the uncertainty of the Lorentzian fit, and the error in E represents the uncertainty in identifying the Fermi velocity. can be tuned with back gate and sample voltage (similar ring structures have previously been observed in bare TMDs, but lacked this degree of tunability 31,32 ). Of all the TMDs studied, these charging rings are unique to MoS2 (both naturally occurring and synthetic). While no ring features are observed in graphene on WSe2 , there are numerous weak wandering line features which run through high resolution dI/dV maps (Fig. 4(b)). Since the graphene lattice is smooth over these line features and they are independent of sample and gate voltage, we attribute these to line dislocations in the WSe2 crystals. Similar features are observed in graphene on WS2 as well. Fig. 4(c) shows the Fourier transform of an atomically resolved dI/dV map of graphene on MoS2 . The Fourier transform exhibits the usual resonances due to
9 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Page 10 of 16
the superlattice (yellow/green hexagons) and atomic lattice (yellow hexagon), as well as the long-wavelength intravalley scattering (red circle). In addition, we observe weak resonances at the K and K’ points of graphene (white circles). These resonances are indicative of intervalley scattering in graphene, which result from the presence of the atomic-scale point and line defects in the TMD substrates (for comparison, no such scattering is observed in clean graphene on hBN devices as the defect density in hBN is orders of magnitude smaller 33 ). (a)
(b)
(c)
(d)
K
K
K’
-4
-2 0 2 11 2 Δn (10 /cm )
4
Figure 4: (a) dI/dV spectroscopy map of graphene on MoS2 . The numerous ring-like features are attributed to the charging of defect states by the STM tip. The map is taken with Vs = 0.025 V, It = 150 pA, and Vg = 0 V. (b) Similar map of graphene on WSe2 . The faint wandering line features running through the map are attributed to line dislocations in the WSe2 crystal. The map is taken with Vs = 0.05 V, It = 150 pA, and Vg = +50 V. The scale bar is 10 nm in (a) and 20 nm in (b). (c) Fourier transform of an atomically resolved dI/dV map of graphene on MoS2 . In addition to the superlattice (yellow/green hexagons), atomic lattice (red hexagon), and long-wavelength intravalley scattering (red circle), there are also weak resonances due to intervalley scattering at the graphene K and K’ points (white circles). All extra resonances are the result of noise in the system. The scale bar is 10 nm−1 . (d) Spatially resolved map of the charge fluctuations in graphene on WSe2 at Vg = 0 V. The overall charge background is removed. The scale bar is 50 nm. To address the possible influence of these defects on the charge environment in graphene, 10 ACS Paragon Plus Environment
Page 11 of 16
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
we take spatially resolved maps of the Dirac point energy. By converting the Dirac point energy ED to charge carrier density n via n = (1/π)(ED /¯hvF )2 , we find charge fluctuations of δn ∼ 1.3 ± 0.2 x 1011 cm−2 for graphene on MoS2 , WS2 , and WSe2 (see Fig. 4(d) for graphene on WSe2 ). While a few times better than the fluctuations observed in graphene on SiO2 , 34 we find these to be around an order of magnitude larger than in comparable graphene on hBN devices. 3,4 These charge fluctuations are consistent both with the defect density observed in bare MoS2 35 as well as with the density of defects seen in these graphene on TMD samples, suggesting that while the TMDs may be screening trapped charges in the SiO2 substrate, defects in the TMDs still create significant static charge disorder in the graphene. Even when the bottom few layers of the MoS2 substrate are conducting and thus fully screening the SiO2 interface (i.e. at large positive gate voltages), the observed charge fluctuations are only reduced by about a factor of two, which is still considerably larger than those observed with hBN. This further implies that the fluctuations are primarily due to TMD defects (the observed reduction is likely due to enhanced screening from the conducting MoS2 layers). Finally, we do not observe any significant dependence on anneal temperature (between 150 ◦
C and 300 ◦ C), suggesting the defects in the TMD substrates are intrinsic to these crystals
and are therefore unavoidable with current synthesis techniques. Contrary to prior reports, 12 we consistently observe a lower electronic quality of graphene on TMD devices than those using hBN. The dirtier local charge environment and prevalence of scattering is consistent with the lower mobility in our devices as well as those from prior reports. 14,23,24 As heterostructures of graphene and TMDs grow quickly in popularity, it is critical to understand their intrinsic limitations. Unless new methods are developed for reducing defects in TMD crystals, the quality of these heterostructures will continue to be inferior to those of graphene on hBN.
11 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Methods Samples are fabricated by transferring either exfoliated or CVD-grown graphene flakes onto either naturally occurring (MoS2 ) or synthetic (MoS2 , WS2 , WSe2 , MoTe2 , SnS2 ) TMD flakes. TMD flakes are exfoliated directly onto a Si substrate capped with 285 nm of thermally grown SiO2 . Exfoliated graphene samples are transferred using the wet transfer technique utilized in Ref. 23. We have not observed any difference in STM measurements between samples using exfoliated or CVD graphene, or between naturally occurring or synthetic MoS2 , so these distinctions are ignored. Samples are annealed in vacuum below 10−5 mbar at 300 ◦
C for MoS2 , WS2 , and WSe2 (or 150 ◦ C where otherwise noted), and 250 ◦ C for MoTe2 and
SnS2 . We have examined about 15 samples with various TMD substrates in total. Naturally occurring bulk MoS2 crystals were purchased from SPI. Synthetic MoS2 crystals were purchased from 2D Semiconductors. WSe2 crystals were purchased from Nanoscience Instruments. 2H-WS2 crystals were grown by the direct vapor transport method of Ref. 36. α-MoTe2 crystals were synthesized following the method of Ref. 37. SnS2 crystals were synthesized following the method of Ref. 26. All the STM measurements were performed in ultrahigh vacuum at a temperature of 4.5 K. dI/dV spectroscopy measurements were acquired by turning off the feedback circuit and adding a small (5-10 mV) a.c. voltage at 563 Hz to the sample voltage. The current was measured by lock-in detection.
Acknowledgement The authors thank Oliver Monti, David Racke, Laura Sharp, David Soltz, and Bruce Parkinson for supplying the SnS2 crystals, as well as Allan H. MacDonald for valuable theoretical discussions. M.Y., S.H. and B.J.L. were supported by NSF Career Award No. DMR-0953784. S.L., K.K., J.X. and E.T. were supported by the Nanoelectronics Research Initiative and Intel.
12 ACS Paragon Plus Environment
Page 12 of 16
Page 13 of 16
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
D.M. was supported by the NSF REU program award No. PHY-1156753. The authors declare no competing financial interests.
Supporting Information Available Schematic of the general device structure; further discussion on topography of graphene on WS2 , MoTe2 , and SnS2 ; dependence of defects on imaging parameters and TMD thickness; gate maps of graphene on WSe2 , WS2 , MoTe2 , and SnS2 as well as negative compressibility in graphene on MoS2 ; charge fluctuations in WSe2 , WS2 , and MoS2 ; dependence on anneal temperature. This material is available free of charge via the Internet at http://pubs.acs.org/.
References (1) Das Sarma, S.; Adam, S., Hwang, E. H.; Rossi, E. Rev. Mod. Phys. 2011, 83, 407-470. (2) Dean, C. R.; et al. Nature Nanotech. 2010, 5, 722-726. (3) Xue, J.; Sanchez-Yamagishi, J.; Bulmash, D.; Jacquod, P.; Deshpande, A.; Watanabe, K.; Taniguchi, T.; Jarillo-Herrero, P.; LeRoy, B. J. Nature Mater. 2011, 10, 282-285. (4) Decker, R.; Wang, Y.; Brar, V. W.; Regan, W.; Tsai, H.-Z.; Wu, Q; Gannett, W.; Zettl, A.; Crommie, M. F. Nano Lett. 2011, 11, 2291-2295. (5) Yankowitz, M.; Xue, J.; LeRoy, B. J. J. Phys.: Condens. Matter 2014, 26 303201. (6) Yankowitz, M.; Xue, J.; Cormode, D.; Sanchez-Yamagishi, J. D.; Watanabe, K.; Taniguchi, T.; Jarillo-Herrero, P.; Jacquod, P.; LeRoy, B. J. Nature Phys. 2012, 8 382-386. (7) Ponomarenko, L. A.; et al. Nature 2013, 497, 594-597. (8) Dean, C. R.; et al. Nature 2013, 497, 598-602. 13 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
(9) Hunt, B.; et al. Science 2013, 340, 1427-1430. (10) Wang, Q. H.; Kalantar-Zadeh, K.; Kis, A.; Coleman, J. N.; Strano, M. S. Nature Nanotech. 2012, 7, 699-712. (11) Geim, A. K.; Grigorieva, I. V. Nature 2013, 499, 419-425. (12) Lu, C.-P.; Li, G.; Watanabe, K.; Taniguchi, T.; Andrei, E. Y. Phys. Rev. Lett. 2014, 113, 156804. (13) Wilson, J. A.; Yoffe, A. D. Adv. Phys. 1969, 18, 193-335. (14) Kretinin, A. V.; et al. Nano Lett. 2014, 14, 3270-3276. (15) Britnell, L.; et al. Science 2012, 335, 947-950. (16) Yu, W. J.; Li, Z.; Zhou, H.; Chen, Y.; Wang, Y.; Huang, Y.; Duan, X. Nature Mater. 2013, 12, 246-252. (17) Georgiou, T.; et al. Nature Nanotech. 2013, 8, 100-103. (18) Moriya, R.; Yamaguchi, T.; Inoue, Y.; Morikawa, S.; Sata, Y.; Masubuchi, S.; Machida, T. Appl. Phys. Lett. 2014, 105, 083119. (19) Britnell, L.; et al. Science. 2013, 340, 1311-1314. (20) Yu, W. J.; Lui, Y.; Zhou, H.; Yin, A.; Li, Z.; Huang, Y.; Duan, X. Nature Nanotech. 2013, 8, 952-958. (21) Bertolazzi, S.; Krasnozhon, D.; Kis. A. ACS Nano 2013, 7, 3246-3252. (22) Roy, K.; Padmanabhan, M.; Goswami, S.; Sai, T. P.; Ramalingam, G.; Raghavan, S.; Ghosh, A. Nature Nanotech. 2013, 8, 826-830. (23) Larentis, S.; Tolsma, J. R.; Fallahazad, B.; Dillen, D. C.; Kim, K.; MacDonald, A. H.; Tutuc, E. Nano Lett. 2014, 14, 2039-2045. 14 ACS Paragon Plus Environment
Page 14 of 16
Page 15 of 16
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
(24) Tan, J. Y.; et al. Appl. Phys. Lett. 2014, 104, 183504. (25) Greenaway, D. L.; Nitsche, R. J. Phys. Chem. Solids 1965, 26, 1445-1458. (26) Sharp, L.; Soltz, D.; Parkinson, B. A. Crystal Growth & Design 2006, 6, 1523-1527. (27) Zhang, Y.; Brar, V. W.; Wang, F.; Girit, C.; Yayon, Y.; Panlasigui, M.; Zettl, A.; Crommie, M. F. Nature Phys. 2008, 4, 627-630. (28) Brar, V. W.; et al. Phys. Rev. Lett. 2010, 104, 036805. (29) Jung, S.; Rutter, G. M.; Klimov, N. N.; Newell, D. B.; Calizo, I.; Hight-Walker, A. R.; Zhitenev, N. B.; Stroscio, J. A. Nature Phys. 2011, 7, 245-251. (30) Brar, V. W.; et al. Nature Phys. 2011, 7, 43-47. (31) Heckl, W. M.; Ohnesorge, F.; Binning, G.; Specht, M.; Hashmi, M. J. Vac. Sci. Technol. B 1991, 9, 1072-1078. (32) Magonov, S. N.; Cantow, H.-J.; Whangbo, M.-H. Surface Science 1994, 318, L1175L1180. (33) Wong, D.; et al. arXiv:1412.1878, 2014. (34) Zhang, Y.; Brar, V. W.; Girit, C.; Zettl, A.; Crommie, M. F. Nature Phys. 2009, 5, 722-726. (35) Lu, C.-P.; Li, G.; Mao, J.; Wang, L.-M.; Andrei, E. Y. Nano Lett. 2014, 14, 4628-4633. (36) Agarwal, M. K.; Nagi Reddy, K.; Patel, H. B. J. Crystal Growth 1979, 46, 139-142. (37) Al-Hilli, A.; Evans, B. L. Journal of Crystal Growth 1972, 15, 93-101.
15 ACS Paragon Plus Environment
Nano Letters
Graphical TOC Entry 2.0
Tip
G/MoS2 G/WSe2
1.8
Au SiO Si
G/WSe2
(dI/dV)/(I/V)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Page 16 of 16
2
1.6 1.4 1.2 1.0 0.8
G on MoS2
10 nm
-0.2
-0.1
0.0
0.1
0.2
Sample Voltage (V)
16 ACS Paragon Plus Environment
0.3
Communication
Intrinsic disorder in graphene on transition metal dichalcogenide heterostructures Matthew Abraham Yankowitz, Stefano Larentis, Kyoung Kim, Jiamin Xue, Devin McKenzie, Shengqiang Huang, Marina Paggen, Mazhar Nawaz Ali, Robert J. Cava, Emanuel Tutuc, and Brian J LeRoy Nano Lett., Just Accepted Manuscript • Publication Date (Web): 09 Feb 2015 Downloaded from http://pubs.acs.org on February 10, 2015
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Nano Letters is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 16
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
Intrinsic disorder in graphene on transition metal dichalcogenide heterostructures Matthew Yankowitz,† Stefano Larentis,‡ Kyounghwan Kim,‡ Jiamin Xue,‡ Devin McKenzie,† Shengqiang Huang,† Marina Paggen,†,¶ Mazhar N. Ali,§ Robert J. Cava,§ Emanuel Tutuc,‡ and Brian J. LeRoy∗,† Physics Department, University of Arizona, Tucson, AZ 85721, USA, Microelectronics Research Center, The University of Texas at Austin, Austin, TX 78758, USA, Physics Department, University of Texas at El Paso, El Paso, TX 79968, USA, and Department of Chemistry, Princeton University, Princeton, New Jersey 08544, USA E-mail: [email protected]
Abstract Semiconducting transition metal dichalchogenides (TMDs) are a family of van der Waals bonded materials that have recently received interest as alternative substrates to hexagonal boron nitride (hBN) for graphene, as well as for components in novel graphene-based device heterostructures. We elucidate the local structural and electronic properties of graphene on TMD heterostructures through scanning tunneling microscopy and spectroscopy measurements. We find that crystalline defects intrinsic to TMDs induce substantial electronic scattering and charge carrier density fluctuations ∗
To whom correspondence should be addressed University of Arizona ‡ University of Texas at Austin ¶ University of Texas at El Paso § Princeton University †
1 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
in the graphene. These signatures of local disorder explain the significant degradation of graphene device mobilities using TMD substrates, particularly compared to similar graphene on hBN devices.
Keywords scanning tunneling microscopy/spectroscopy, graphene, transition metal dichalchogenides, defects, scattering
The electronic properties of two-dimensional materials such as graphene are extremely sensitive to their environment, especially the underlying substrate. Since the isolation of graphene in 2004, considerable effort has been put into finding the best substrates, both from a device standpoint and for inducing novel physical phenomena. Perhaps the most common substrate, SiO2 , causes out-of-plane ripples and locally dopes the graphene due to trapped charged impurities. 1 Suspended graphene devices, fabricated by etching away the SiO2 layer, offer the best intrinsic graphene quality, although their fabrication is far too challenging to scale to industrial levels. Hexagonal boron nitride has emerged as a very promising substrate; as an insulating crystal it not only flattens the graphene but screens underlying charge impurities from the base substrate. 2–5 Careful device fabrication techniques can yield devices of graphene quality nearing that of suspended graphene, and these heterostructures are more friendly for industrial scaling. Since hBN has a similar lattice constant to graphene, when the two lattices are in near perfect alignment interactions between the crystals strongly renormalize the graphene band structure. 5,6 This opens an avenue for the study of interesting phenomena, such as the Hofstadter quantization, 7–9 and also provides a route to make graphene insulating. 5 However, it may not be ideal for large scale graphene device applications where fabrication leads to a random alignment between the crystals, and the intrinsic 2 ACS Paragon Plus Environment
Page 2 of 16
Page 3 of 16
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
graphene band structure needs to be preserved. Recently, transition metal dichalcogenides have made a strong resurgence in materials research, as these crystals can be exfoliated to atomic scale thicknesses and stacked via van der Waals interactions similarly to graphene and hBN. 10,11 A subset of the TMDs exhibit similar semiconducting behavior, with indirect band gaps in bulk ranging from ∼ 1 - 1.4 eV. 10 Na¨ıvely, these materials, when insulating, should offer comparable quality to hBN as substrates for graphene, but without the possibility of band structure modification due to their considerably different lattice constants. 12,13 Additionally, their semiconducting nature permits dynamic gate voltage control over the interaction strength with graphene, 12 and they offer the potential for the study of new physical phenomena (for example, potential spin-orbit coupling induced in the graphene layer due to the heavy metal atoms of the TMD 14 ). From a device standpoint, there are numerous potential applications involving heterostructures between graphene and TMDs; for example, as tunneling transistors, 15–18 highly efficient flexible photovoltaic devices, 19,20 or nonvolatile memory cells. 21,22 Unfortunately, graphene on TMD devices have thus far been of significantly lower mobility than comparable device structures utilizing an hBN substrate, 14,23,24 and a local understanding of this behavior is lacking. In this Letter, we show via local scanning probe measurements that graphene on TMD devices suffer an unavoidable degradation in electronic quality due to defects in the TMD crystals. We study graphene on substrates belonging to the MX2 family, where M is a transition metal (Mo, W) and X is a chalcogen atom (S, Se, Te). The specific TMDs examined here are MoS2 , WS2 , WSe2 , and MoTe2 . We have also examined graphene on SnS2 , which is not technically a TMD but shares the same crystal structure and is also a semiconductor. 25,26 Figs. 1(a) and (b) show topography images of graphene on MoS2 and WSe2 obtained via scanning tunneling microscopy (STM). Numerous defects can be observed in the MoS2 sample, characterized by highly localized topographic height variations (both positive and negative, depending on tunnel parameters). The graphene lattice is continuous over these defects,
3 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
indicating the defects reside in the MoS2 and are presumably either atomic substitutions or vacancies. The strength and appearance of these defects can be tuned by the sample and gate voltages, and their appearance in other TMDs is found to depend on those factors as well as crystal thickness. The graphene on MoTe2 and SnS2 samples exhibit significantly worse topography, marked by islands of well-adhered regions surrounded by significantly rougher regions where the graphene did not appear to be adhered to the TMD, suggesting that these crystals are not stable enough in air to allow consistently good adhesion with the graphene 11 (see Supporting Information for further discussion of defects and topography observations). The insets of Figs. 1(a) and (b) show atomic resolution images of both samples, and additionally display hexagonal superlattices due to the interference pattern formed by the graphene and TMD lattices. As is the case for graphene on hBN, a moir´e pattern is expected to form between the graphene and TMD lattices due to their relative rotation φ and lattice mismatch δ. The moir´e wavelength is
λ= q
(1 + δ)a 2(1 + δ)(1 − cos φ) + δ 2
where a is the graphene lattice constant. 6 Because the lattice mismatch is much larger between graphene and TMDs than graphene and hBN, the range of possible moir´e wavelengths is much smaller. The inset of Fig. 1(d) shows the dispersion of moir´e wavelengths as a function of angle φ for graphene on MoS2 and WSe2 (the dispersions for the other TMDs studied here are very similar). The lattice constants of the TMDs in this study range from 3.15 to 3.64 ˚ A, 13,25 and thus the possible moir´e wavelengths are on the order of 0.5 nm to just over 1 nm. As a result of these short moir´e wavelengths, no modification of the graphene band structure is expected at low energies. 6 Figure 1(c) shows a Fourier transform of the inset of Fig. 1(a), with the superlattice (yellow/green) and graphene lattice (red) hexagons highlighted. All samples studied show good agreement between the moir´e wavelength and the angle between the lattices.
4 ACS Paragon Plus Environment
Page 4 of 16
(a)
(b)
(c)
(d) 4 Conductivity (mS)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
Wavelength (nm)
Page 5 of 16
3
1.2 1.0 0.8 0.6 0.4 -30
0 ø (degrees)
30
2 G/WSe2 G/MoS2
1 0
-40
-20 0 20 Gate Voltage (V)
40
Figure 1: (a) Topography of graphene on MoS2 . Arrows mark some of the defects buried in the MoS2 . Inset: Atomic resolution with a moir´e pattern of 0.65 nm. (b) Same as (a) for WSe2 . No defects are visible. Inset: Atomic resolution with a moir´e pattern of 0.91 nm. For (a) and (b), the scale bar for the main figure is 10 nm and for the insets it is 1 nm. Typical imaging parameters are sample voltages between Vs = 0.1 V and 0.3 V and tunnel currents between It = 100 pA and 300 pA. (c) Fourier transform of (a). The atomic lattice is highlighted with a red hexagon, and the moir´e with a yellow hexagon. The moir´e points surrounding the atomic lattice are highlighted with green hexagons. All extra resonances are the result of noise in the system. The scale bar is 10 nm−1 . (d) Global conductance measurements of graphene on MoS2 and WSe2 devices. The MoS2 becomes conducting at gate voltages just above 0 V, after which point the device exhibits negative compressibility. 23 The WSe2 crystal is always biased within the band gap. Inset: Dispersion of possible moir´e wavelengths for graphene on MoS2 and WSe2 as a function of the relative rotation between the lattices. In addition to their similar crystal structure, these TMDs also share similar electronic properties. They are indirect gap semiconductors with band gaps in bulk ranging from about 1.0 eV to 1.4 eV 10 (and 2.2 eV for SnS2 26 ). In proximity to graphene, the difference in energy between the graphene work function and the electron affinity of the TMD determines the relative band alignment. The threshold gate voltage at which the TMD begins to conduct depends on this band alignment as well as the band bending related to the TMD crystal thickness. 23 For comparably thin crystals (∼5 - 15 nm), we observe that the Dirac point is
5 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
aligned very near the conduction bands of MoS2 , MoTe2 and SnS2 , and closer to the middle of the gap for WS2 and WSe2 . Figure 1(d) shows global conductance measurements at 4.2 K for graphene on MoS2 and WSe2 . The graphene on WSe2 device is nearly symmetric around charge neutrality (indicating the WSe2 always remains insulating), whereas the MoS2 device has a saturating conductance at positive gate voltages as the MoS2 becomes conducting. All devices studies exhibit similar transport characteristics to those in Fig. 1(d). We find field-effect mobilities of 5,000 - 10,000 cm2 /Vs in these devices, which is over an order of magnitude lower than of comparable graphene on hBN devices. We restrict our study to thin TMD substrates as devices with thicker crystals become strongly hysteretic with gate voltage. Figures 2(a) and (b) show normalized (dI/dV)/(I/V) spectroscopy measurements as a function of gate voltage for graphene on MoS2 and WSe2 . For MoS2 , the Dirac point (marked with a white dotted line) essentially stops moving in energy once the gate voltage is large enough to induce charge carriers in the TMD. In some samples we have additionally observed the Dirac point move with positive dVs /dVg when the MoS2 is conducting, indicative of the negative compressibility of the system 23 (see Supporting Information). For graphene on WSe2 , the Dirac point moves with the standard square root of gate voltage dispersion expected for graphene on an insulating substrate, 27 indicating the WSe2 substrate is never biased to the valence or conduction bands. In this case, the WSe2 should behave very similarly to an hBN substrate, flattening the graphene and screening charged impurities in the underlying SiO2 . For all the graphene on TMD devices, the movement of the Dirac point is well fit by a Fermi velocity of 0.95 ± 0.05 x 106 m/s. We additionally observe weak nondispersive features characteristic of phonon-mediated tunneling into graphene (most notably the depressed dI/dV surrounding zero sample bias) as well as features related to electronplasmon interactions (dispersive inflection point at energies slightly above the Dirac point). 28 The Supporting Information provides further identification and discussion of these features. In addition to the spectroscopy features discussed above, we observe two types of features
6 ACS Paragon Plus Environment
Page 6 of 16
Page 7 of 16
(b) Sample Voltage (V)
0.2 0.1 0.0 -0.1 -0.2 -0.3
(c)
-50
0
Gate Voltage (V)
2.0 G/MoS2 G/WSe2
(dI/dV)/(I/V)
1.8 1.6 1.4 1.2 1.0
-0.1
0.0
0.1
0.2
Sample Voltage (V)
0.2 0.1 0.0 -0.1 3
-0.2
0.3
-50
0
2
50
Gate Voltage (V)
(d)
0.8 -0.2
0.3
-0.3
50
Sample Voltage (V)
Sample Voltage (V)
(a) 0.3
0.3
1
(dI/dV)/(I/V)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
0.2 0
0.1 0.0 -0.1 -0.2 -0.3
0
10 20 30 Position (nm)
40
Figure 2: (a) Normalized (dI/dV)/(I/V) spectroscopy as a function of gate voltage for graphene on MoS2 . The Dirac point is highlighted with a dotted white line. The Dirac point remains nearly stationary with gate voltage above ∼+10 V as the MoS2 becomes conducting and the charge density in the graphene is constant. States due to charging of defects in MoS2 are marked with black arrows. (b) Same as (a) for graphene on WSe2 . In both (a) and (b), extra peaks of varying strength due to scattering surround the Dirac point at both positive and negative energies. The nearly vertical bright features near zero gate voltage result from confinement by p-n junctions. 29 (c) Cuts of (a) and (b) (marked by the dashed vertical lines) at Vg = -20 V. For both, the Dirac point is around Vs = +0.15V (marked with black diamonds). Extra states due to scattering are marked with tick marks. (d) Normalized dI/dV spectroscopy as a function of position in graphene on MoTe2 at Vg = -9 V. The approximate position of the Dirac point is marked with a dotted white line. All other resonances are attributed to intravalley scattering, and their energy spacing varies randomly with position. which are not characteristic of pristine graphene. Ubiquitous amongst all MX2 substrates examined in this study, we observe extra resonances surrounding and moving roughly in parallel with the Dirac point. Fig. 2(c) shows a line cuts of Figs. 2(a) and (b) at Vg = -20 V, indicating clearly the presence of these extra states, marked with black ticks (see the Supporting Information for a derivative of Fig. 2(b), which also highlights these satellite states). Additionally, unique to MoS2 substrates, we observe strong features which move oppositely of the Dirac point with gate voltage (marked with black arrows in Fig. 2(a)) which we will discuss later.
7 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
The resonances which track in parallel the Dirac point are typically characteristic of electronic scattering in graphene, 29 although we have considered other origins for the extra states we observe, such as new features of the band structure due to interactions with the substrate or the excitation of phonons in the TMD substrate by electrons tunneling into the graphene. Such features should have well-defined energies which do not vary spatially or between different samples, depending on the exact nature of their origin. To test this hypothesis we take line maps of normalized dI/dV spectroscopy, as shown in Fig. 2(d) for graphene on MoTe2 . We see clearly that the energy spacing of these states varies spatially, and in some cases the states split or merge. While this behavior is inconsistent with the alternative explanations considered above, it is consistent with electronic scattering in graphene from point and line defects in the TMD substrate and/or the remnant disorder potential resulting from those defects. 29 To further investigate the intravalley scattering in graphene on TMD heterostructures, we take large area dI/dV maps of graphene on WSe2 at various gate voltages (see Figs. 3(a) - (c) for three examples). The coherent features in the maps change size with gate voltage, characteristic of intravalley scattering from the Coulomb potentials of buried defects. Fig. 3(d) shows the Fourier transform of Fig. 3(c). The disk at the center of the image arises from 2k scattering across a single graphene Dirac cone. For each map, we take a circular average of the Fourier transform and extract the corresponding wave vector k as the half-width at half-maximum of the best-fit Lorentzian. These wave vectors are plotted as a function of energy relative to the Dirac point in Fig. 3(e). The extracted points agree with a linear dispersion characterized by a Fermi velocity of ∼0.95 x 106 m/s. Fig. 4(a) shows a similar dI/dV map of graphene on MoS2 , which exhibits numerous ring features due to the charging or discharging of defect states resulting from the interaction of MoS2 defects with graphene (similar to those seen in artificial impurities on graphene 30 ). These charging rings are the same features which run oppositely of the Dirac point in the gate map (features marked with black arrows in Fig. 2(a)). The size and nature of the rings
8 ACS Paragon Plus Environment
Page 8 of 16
Page 9 of 16
(a)
(d)
(b)
(c)
(e)
0.4
0.2
Energy (eV)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
0.0
-0.2
-0.4 -0.4
-0.2
0.0 -1 k (nm )
0.2
0.4
Figure 3: (a) - (c) Maps of dI/dV spectroscopy for graphene on WSe2 at Vg = -60 V, -5 V, and +60 V, respectively. The local charge neutrality point is around Vg = -15 V. Each map is taken with Vs = 50 mV and It = 150 pA. The coherent structure is due to intravalley scattering, and becomes shorter wavelength at gate voltages further from the Dirac point. The scale bars in (a)-(c) are 50 nm. (d) Fourier transform of (c), exhibiting a disk-like feature at the center corresponding to the scattering wave vector. The circular average is plotted in red. The scale bar is 0.2 nm−1 . (e) Graphene energy versus momentum dispersion extracted from maps similar to those shown in (a) - (c). The wave vector k is extracted from the half-width at half-maximum of the Lorentzian fit of the Fourier transforms of each map. The energy is determined by converting the sample voltage of each measurement to energy from the Dirac point using vF = 0.95 x 106 m/s. The solid blue lines plot the dispersion E=h ¯ vF k. The error in k represents the uncertainty of the Lorentzian fit, and the error in E represents the uncertainty in identifying the Fermi velocity. can be tuned with back gate and sample voltage (similar ring structures have previously been observed in bare TMDs, but lacked this degree of tunability 31,32 ). Of all the TMDs studied, these charging rings are unique to MoS2 (both naturally occurring and synthetic). While no ring features are observed in graphene on WSe2 , there are numerous weak wandering line features which run through high resolution dI/dV maps (Fig. 4(b)). Since the graphene lattice is smooth over these line features and they are independent of sample and gate voltage, we attribute these to line dislocations in the WSe2 crystals. Similar features are observed in graphene on WS2 as well. Fig. 4(c) shows the Fourier transform of an atomically resolved dI/dV map of graphene on MoS2 . The Fourier transform exhibits the usual resonances due to
9 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Page 10 of 16
the superlattice (yellow/green hexagons) and atomic lattice (yellow hexagon), as well as the long-wavelength intravalley scattering (red circle). In addition, we observe weak resonances at the K and K’ points of graphene (white circles). These resonances are indicative of intervalley scattering in graphene, which result from the presence of the atomic-scale point and line defects in the TMD substrates (for comparison, no such scattering is observed in clean graphene on hBN devices as the defect density in hBN is orders of magnitude smaller 33 ). (a)
(b)
(c)
(d)
K
K
K’
-4
-2 0 2 11 2 Δn (10 /cm )
4
Figure 4: (a) dI/dV spectroscopy map of graphene on MoS2 . The numerous ring-like features are attributed to the charging of defect states by the STM tip. The map is taken with Vs = 0.025 V, It = 150 pA, and Vg = 0 V. (b) Similar map of graphene on WSe2 . The faint wandering line features running through the map are attributed to line dislocations in the WSe2 crystal. The map is taken with Vs = 0.05 V, It = 150 pA, and Vg = +50 V. The scale bar is 10 nm in (a) and 20 nm in (b). (c) Fourier transform of an atomically resolved dI/dV map of graphene on MoS2 . In addition to the superlattice (yellow/green hexagons), atomic lattice (red hexagon), and long-wavelength intravalley scattering (red circle), there are also weak resonances due to intervalley scattering at the graphene K and K’ points (white circles). All extra resonances are the result of noise in the system. The scale bar is 10 nm−1 . (d) Spatially resolved map of the charge fluctuations in graphene on WSe2 at Vg = 0 V. The overall charge background is removed. The scale bar is 50 nm. To address the possible influence of these defects on the charge environment in graphene, 10 ACS Paragon Plus Environment
Page 11 of 16
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
we take spatially resolved maps of the Dirac point energy. By converting the Dirac point energy ED to charge carrier density n via n = (1/π)(ED /¯hvF )2 , we find charge fluctuations of δn ∼ 1.3 ± 0.2 x 1011 cm−2 for graphene on MoS2 , WS2 , and WSe2 (see Fig. 4(d) for graphene on WSe2 ). While a few times better than the fluctuations observed in graphene on SiO2 , 34 we find these to be around an order of magnitude larger than in comparable graphene on hBN devices. 3,4 These charge fluctuations are consistent both with the defect density observed in bare MoS2 35 as well as with the density of defects seen in these graphene on TMD samples, suggesting that while the TMDs may be screening trapped charges in the SiO2 substrate, defects in the TMDs still create significant static charge disorder in the graphene. Even when the bottom few layers of the MoS2 substrate are conducting and thus fully screening the SiO2 interface (i.e. at large positive gate voltages), the observed charge fluctuations are only reduced by about a factor of two, which is still considerably larger than those observed with hBN. This further implies that the fluctuations are primarily due to TMD defects (the observed reduction is likely due to enhanced screening from the conducting MoS2 layers). Finally, we do not observe any significant dependence on anneal temperature (between 150 ◦
C and 300 ◦ C), suggesting the defects in the TMD substrates are intrinsic to these crystals
and are therefore unavoidable with current synthesis techniques. Contrary to prior reports, 12 we consistently observe a lower electronic quality of graphene on TMD devices than those using hBN. The dirtier local charge environment and prevalence of scattering is consistent with the lower mobility in our devices as well as those from prior reports. 14,23,24 As heterostructures of graphene and TMDs grow quickly in popularity, it is critical to understand their intrinsic limitations. Unless new methods are developed for reducing defects in TMD crystals, the quality of these heterostructures will continue to be inferior to those of graphene on hBN.
11 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Methods Samples are fabricated by transferring either exfoliated or CVD-grown graphene flakes onto either naturally occurring (MoS2 ) or synthetic (MoS2 , WS2 , WSe2 , MoTe2 , SnS2 ) TMD flakes. TMD flakes are exfoliated directly onto a Si substrate capped with 285 nm of thermally grown SiO2 . Exfoliated graphene samples are transferred using the wet transfer technique utilized in Ref. 23. We have not observed any difference in STM measurements between samples using exfoliated or CVD graphene, or between naturally occurring or synthetic MoS2 , so these distinctions are ignored. Samples are annealed in vacuum below 10−5 mbar at 300 ◦
C for MoS2 , WS2 , and WSe2 (or 150 ◦ C where otherwise noted), and 250 ◦ C for MoTe2 and
SnS2 . We have examined about 15 samples with various TMD substrates in total. Naturally occurring bulk MoS2 crystals were purchased from SPI. Synthetic MoS2 crystals were purchased from 2D Semiconductors. WSe2 crystals were purchased from Nanoscience Instruments. 2H-WS2 crystals were grown by the direct vapor transport method of Ref. 36. α-MoTe2 crystals were synthesized following the method of Ref. 37. SnS2 crystals were synthesized following the method of Ref. 26. All the STM measurements were performed in ultrahigh vacuum at a temperature of 4.5 K. dI/dV spectroscopy measurements were acquired by turning off the feedback circuit and adding a small (5-10 mV) a.c. voltage at 563 Hz to the sample voltage. The current was measured by lock-in detection.
Acknowledgement The authors thank Oliver Monti, David Racke, Laura Sharp, David Soltz, and Bruce Parkinson for supplying the SnS2 crystals, as well as Allan H. MacDonald for valuable theoretical discussions. M.Y., S.H. and B.J.L. were supported by NSF Career Award No. DMR-0953784. S.L., K.K., J.X. and E.T. were supported by the Nanoelectronics Research Initiative and Intel.
12 ACS Paragon Plus Environment
Page 12 of 16
Page 13 of 16
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
D.M. was supported by the NSF REU program award No. PHY-1156753. The authors declare no competing financial interests.
Supporting Information Available Schematic of the general device structure; further discussion on topography of graphene on WS2 , MoTe2 , and SnS2 ; dependence of defects on imaging parameters and TMD thickness; gate maps of graphene on WSe2 , WS2 , MoTe2 , and SnS2 as well as negative compressibility in graphene on MoS2 ; charge fluctuations in WSe2 , WS2 , and MoS2 ; dependence on anneal temperature. This material is available free of charge via the Internet at http://pubs.acs.org/.
References (1) Das Sarma, S.; Adam, S., Hwang, E. H.; Rossi, E. Rev. Mod. Phys. 2011, 83, 407-470. (2) Dean, C. R.; et al. Nature Nanotech. 2010, 5, 722-726. (3) Xue, J.; Sanchez-Yamagishi, J.; Bulmash, D.; Jacquod, P.; Deshpande, A.; Watanabe, K.; Taniguchi, T.; Jarillo-Herrero, P.; LeRoy, B. J. Nature Mater. 2011, 10, 282-285. (4) Decker, R.; Wang, Y.; Brar, V. W.; Regan, W.; Tsai, H.-Z.; Wu, Q; Gannett, W.; Zettl, A.; Crommie, M. F. Nano Lett. 2011, 11, 2291-2295. (5) Yankowitz, M.; Xue, J.; LeRoy, B. J. J. Phys.: Condens. Matter 2014, 26 303201. (6) Yankowitz, M.; Xue, J.; Cormode, D.; Sanchez-Yamagishi, J. D.; Watanabe, K.; Taniguchi, T.; Jarillo-Herrero, P.; Jacquod, P.; LeRoy, B. J. Nature Phys. 2012, 8 382-386. (7) Ponomarenko, L. A.; et al. Nature 2013, 497, 594-597. (8) Dean, C. R.; et al. Nature 2013, 497, 598-602. 13 ACS Paragon Plus Environment
Nano Letters
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
(9) Hunt, B.; et al. Science 2013, 340, 1427-1430. (10) Wang, Q. H.; Kalantar-Zadeh, K.; Kis, A.; Coleman, J. N.; Strano, M. S. Nature Nanotech. 2012, 7, 699-712. (11) Geim, A. K.; Grigorieva, I. V. Nature 2013, 499, 419-425. (12) Lu, C.-P.; Li, G.; Watanabe, K.; Taniguchi, T.; Andrei, E. Y. Phys. Rev. Lett. 2014, 113, 156804. (13) Wilson, J. A.; Yoffe, A. D. Adv. Phys. 1969, 18, 193-335. (14) Kretinin, A. V.; et al. Nano Lett. 2014, 14, 3270-3276. (15) Britnell, L.; et al. Science 2012, 335, 947-950. (16) Yu, W. J.; Li, Z.; Zhou, H.; Chen, Y.; Wang, Y.; Huang, Y.; Duan, X. Nature Mater. 2013, 12, 246-252. (17) Georgiou, T.; et al. Nature Nanotech. 2013, 8, 100-103. (18) Moriya, R.; Yamaguchi, T.; Inoue, Y.; Morikawa, S.; Sata, Y.; Masubuchi, S.; Machida, T. Appl. Phys. Lett. 2014, 105, 083119. (19) Britnell, L.; et al. Science. 2013, 340, 1311-1314. (20) Yu, W. J.; Lui, Y.; Zhou, H.; Yin, A.; Li, Z.; Huang, Y.; Duan, X. Nature Nanotech. 2013, 8, 952-958. (21) Bertolazzi, S.; Krasnozhon, D.; Kis. A. ACS Nano 2013, 7, 3246-3252. (22) Roy, K.; Padmanabhan, M.; Goswami, S.; Sai, T. P.; Ramalingam, G.; Raghavan, S.; Ghosh, A. Nature Nanotech. 2013, 8, 826-830. (23) Larentis, S.; Tolsma, J. R.; Fallahazad, B.; Dillen, D. C.; Kim, K.; MacDonald, A. H.; Tutuc, E. Nano Lett. 2014, 14, 2039-2045. 14 ACS Paragon Plus Environment
Page 14 of 16
Page 15 of 16
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Nano Letters
(24) Tan, J. Y.; et al. Appl. Phys. Lett. 2014, 104, 183504. (25) Greenaway, D. L.; Nitsche, R. J. Phys. Chem. Solids 1965, 26, 1445-1458. (26) Sharp, L.; Soltz, D.; Parkinson, B. A. Crystal Growth & Design 2006, 6, 1523-1527. (27) Zhang, Y.; Brar, V. W.; Wang, F.; Girit, C.; Yayon, Y.; Panlasigui, M.; Zettl, A.; Crommie, M. F. Nature Phys. 2008, 4, 627-630. (28) Brar, V. W.; et al. Phys. Rev. Lett. 2010, 104, 036805. (29) Jung, S.; Rutter, G. M.; Klimov, N. N.; Newell, D. B.; Calizo, I.; Hight-Walker, A. R.; Zhitenev, N. B.; Stroscio, J. A. Nature Phys. 2011, 7, 245-251. (30) Brar, V. W.; et al. Nature Phys. 2011, 7, 43-47. (31) Heckl, W. M.; Ohnesorge, F.; Binning, G.; Specht, M.; Hashmi, M. J. Vac. Sci. Technol. B 1991, 9, 1072-1078. (32) Magonov, S. N.; Cantow, H.-J.; Whangbo, M.-H. Surface Science 1994, 318, L1175L1180. (33) Wong, D.; et al. arXiv:1412.1878, 2014. (34) Zhang, Y.; Brar, V. W.; Girit, C.; Zettl, A.; Crommie, M. F. Nature Phys. 2009, 5, 722-726. (35) Lu, C.-P.; Li, G.; Mao, J.; Wang, L.-M.; Andrei, E. Y. Nano Lett. 2014, 14, 4628-4633. (36) Agarwal, M. K.; Nagi Reddy, K.; Patel, H. B. J. Crystal Growth 1979, 46, 139-142. (37) Al-Hilli, A.; Evans, B. L. Journal of Crystal Growth 1972, 15, 93-101.
15 ACS Paragon Plus Environment
Nano Letters
Graphical TOC Entry 2.0
Tip
G/MoS2 G/WSe2
1.8
Au SiO Si
G/WSe2
(dI/dV)/(I/V)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
Page 16 of 16
2
1.6 1.4 1.2 1.0 0.8
G on MoS2
10 nm
-0.2
-0.1
0.0
0.1
0.2
Sample Voltage (V)
16 ACS Paragon Plus Environment
0.3
