Tunability of electrical and thermoelectrical properties of monolayer MoS 2 through oxygen passivation
Swarup Deb, Pritam Bhattacharyya, Poulab Chakrabarti, Himadri Chakraborti, Kantimay Das Gupta, Alok Shukla, Subhabrata Dhar
TTunability of electrical and thermoelectrical properties of monolayer MoS throughoxygen passivation Swarup Deb, ∗ Pritam Bhattacharyya, Poulab Chakrabarti, HimadriChakraborti, Kantimay Das Gupta, Alok Shukla, and Subhabrata Dhar † Department of Physics , Indian Institute of Technology Bombay , Powai , Mumbai 400076 , India (Dated: March 10, 2020)Electric and thermoelectric properties of strictly monolayer MoS films, which are grown usinga novel micro-cavity based CVD growth technique, have been studied under diverse environmentaland annealing conditions. Resistance of a thermoelectric device that is fabricated on a continuousmonolayer MoS layer using photolithography technique has been found to reduce by about sixorders of magnitudes upon annealing in vacuum at 525 K. Seebeck coefficient of the layer alsoreduces by almost an order of magnitude upon annealing. When the sample is exposed to oxygenatmosphere, these parameters return to their previous values. In fact, it has been found that theelectron concentration, mobility as well as the thermoelectric power of the material can be tunedby controlling the temperature of annealing and oxygen exposure. Once established, these valuesare maintained as long as the layer is not exposed to oxygen environment. This can offer a uniqueway to control doping in the material provided an effective encapsulation method is devised. Suchcontrol is an important step forward for device application. The effect has been attributed to thepassivation of di-sulfur vacancy donors present in the MoS film by physisorbed oxygen molecules.Band structural calculations using density functional theory have been carried out, results of whichindeed validate this picture. Two dimensional (2D) materials particularly tran-sition metal dichalcogenides (TMDs), such as MoS ,MoSe , WS etc. have emerged as the materials forthe next generation logic, electronic, opto-electronic de-vices and energy-related technologies [1–4]. Mechanicalexfoliation[5] and chemical vapour deposition (CVD)[6,7] are two most widely accepted techniques for preparingsingle layer 2D samples. Both of these techniques havecertain positive and negative sides. For example, exfoli-ated samples are superior in quality but they fail in termsof scalability. Typical size of monolayer regions in theseflakes is not more than a few tens of micrometer[8, 9].Therefore, fabrication of large scale integrated circuits isnot possible on such layers. On the other hand, CVD canprovide monolayers with much larger area coverage[10–12]. But, these films often suffer from high density ofdefects and grain boundaries[13, 14]. Irrespective of thepreparation technique, increased surface to volume ratiohas made these materials vulnerable to the ambiance. Itshould be noted that chalcogen vacancies, which act asdonors, are omnipresent in these materials[15, 16]. Inprinciple, these defects can influence the surface adsorp-tion, which in turn can passivate these donors affectingboth the concentration and the mobility of the carriersin the layer. If the density of the adsorbates, which pas-sivate chalcogen vacancy donors, can be stabilized on thefilm surface by certain means, one can tune the dop-ing level of the material, an important step towards de-vice application. It should be mentioned that the op-tical properties of the film can also be affected by sur-face adsorption[17]. In fact, there are recent studies on ∗ [email protected] † [email protected] the effect of defect passivation on the optical propertiesin CVD grown monolayer TMD films[18–21]. However,the influence of adsorption on the electronic propertiesof monolayer TMDs has hardly been studied so far.Thermoelectrics have drawn a great deal of attentionfor the last several decades as they offer a nature-friendlyway to convert heat to electricity and to use electricityfor refrigeration. Low dimensional systems have come upas viable option[22–25]. Monolayer TMDs are predictedto be excellent thermoelectric materials because of theirrelatively high effective mass and the property of valleydegeneracy[26, 27]. The Seebeck coefficient, which is oneof the key parameter that decides the figure of merit of athermoelectric material, has been experimentally foundto be as high[26] as 30 mV K − in 1L-MoS that is waymore than any other nano-scaled materials[24, 25]. Yet,it should be mentioned that only a handful of experimen-tal studies are carried out on thermoelectric propertiesof 1L-TMDs[26–28]. It will be interesting to explore howthe surface adsorption affects the Seebeck coefficient ofthese materials.Recently, we have developed a novel micro-cavity basedCVD growth technique, where strictly monolayer MoS (1L-MoS ) film can be grown on c-sapphire substratescovering an area as large as a few mm [29, 30]. Thesefilms show several GΩ of resistance in ambient condi-tions. Here, we have studied the thermoelectric proper-ties of these films at different environmental conditionsbefore and after annealing it in vacuum. A thermoelectricdevice is fabricated on a continuous film using the pho-tolithography technique. It has been found that vacuumannealing at a temperature of about 525 K can reducethe resistance of the device by about six orders of magni-tudes and at the same time thermoelectric power also getsreduced by almost an order of magnitude. Interestingly, a r X i v : . [ c ond - m a t . m e s - h a ll ] M a r 𝑹 𝒕𝒉 ′ Micro-Heater 𝑹 𝒕𝒉 Open Circuit
Condition S T CBDriftCold Hot e - Diffusion c)b) 𝑹 𝒕𝒉 𝑹 𝒕𝒉 ′ D - D O S Fermi
Level -+ MoS a) FIG. 1. (a) Scanning electron micrograph of an actual device. (b) Schematic of the full device (not scaled). R th ( R (cid:48) th ) isthe four probe resistance of the metal line that connects contact pads 1 and 2 (1 (cid:48) and 2 (cid:48) ) with 3 and 4 (3 (cid:48) and 4 (cid:48) ). Heaterdimension in the actual device is kept much larger than the length of the MoS region to ensure temperature uniformity acrossthe cross-section. (c) Mechanism of charge flow and potential build up across the device in the presence of a thermal gradientfor open circuit configurations is schematically illustrated. The (+) and (-) sign indicates the polarity of the voltmeter probes.In the given configuration measured thermoelectric voltage for n-type semiconductor will be ( − ) ve . both the resistance ( R ) and the Seebeck coefficient ( S ) re-turn to their previous values when the sample is exposedto oxygen environment. It has been observed that R , aswell as S of the device, can be tuned by controlling thetemperature of annealing and oxygen exposure. Thesevalues hardly change as long as the device is not exposedto oxygen environment. The study attributes the effectto the passivation of sulfur vacancy donors by oxygenmolecules, which are physisorbed on the surface. Bandstructural calculations within the framework of densityfunctional theory have been performed to check the va-lidity of the model. Theory shows that the adsorption ofoxygen molecules at sulfur vacancy sites is indeed ener-getically stable. This results in the formation of energylevels 250 meV below the donor states arising from thesulfur vacancies. It has also been shown that these levelscan capture electrons, which leads to the passivation ofthe donors.Strictly monolayer MoS films were grown on c -sapphire substrates using a microcavity based CVD tech-nique. Prior to the growth, substrates were cleanedsubsequently in TCE, acetone, and methanol and fi-nally dipped in H O:HF (10:1) solution for 40 sec. Moredetails of the growth procedure have been discussedelsewhere[29, 30]. Standard optical lithographic tech-nique was used for device fabrication. Layers of two dif-ferent metals viz. titanium (Ti ∼
20 nm) and gold (Au ∼
100 nm) were thermally deposited in a thermal evapora-tor at a background pressure less than 1 × − mbar.The device was subjected to rapid thermal annealingat a temperature of 300 ° C for 1 min. The unwanted areas,which create electrical contact between the micro-heaterand the active area, were then selectively etched throughoxygen plasma ashing. Positive photoresist S1813 wasused as etch mask for this dry etching process. Fig-ure 1(a) shows the scanning electron microscopic imageof a part of the device in the vicinity of the micro-heater.Panel (b) of figure 1 shows the schematic depiction of thedevice. Probe 1 and 4 (1 (cid:48) and 4 (cid:48) ) were used as currentprobes while 2 and 3 (2 (cid:48) and 3 (cid:48) ) are the voltage probe for the four-probe resistivity measurement for the twometal lines, which serve as resistive-thermometers at thetwo locations. Contact pads 1 and 1 (cid:48) were also usedto perform 2-probe current ( I ) vs voltage ( V ) measure-ments. I-V profiles were recorded using Keithley-6487picometer-voltage source. Two lock-in amplifiers, Sig-nal Recovery-7225 and Stanford Research-SR830 (phase-locked with each other) were used to measure the 4-proberesistances (say, R th and R (cid:48) th ) of the thermometers. Astorage-type liquid nitrogen cryostat was used to per-form electric and thermoelectric measurements at differ-ent temperatures ranging from 80 to 420 K. The chamberwas also utilized for in situ annealing of the device in vac-uum. Resistance vs temperature calibration of both themicro-thermometers was carried out by slowly increasingthe cryostat temperature from 80 K to 420 K. Ramp rateof approximately 1 Kmin − was used during calibrationto avoid temperature lag between the sensor located inthe cold finger (Pt-100) and the device. After the calibra-tion one of the lock-in amplifier (Signal Recovery-7225)was reconfigured to measure the temperature differencebetween the micro-thermometers of the device. Specialcare was taken to minimize the common-mode gain acrossthe on-chip thermometers. Keithley-6221 current sourcewas used to excite the micro-heater. Thermoelectric volt-age across the device was picked up using Keithley-2182ananovoltmeter, across terminal 1 and 1 (cid:48) .Room temperature I versus V profiles for the deviceare shown in figure 2(a). Black symbols represent thedata recorded before evacuating the sample space of thecryostat. Resistance is measured to be ∼
26 GΩ. Theprofile represented by red symbols is obtained after evac-uating the sample space to ∼ × − mbar. Clearly,evacuation at room temperature has no significant ef-fect on the resistance of the device. Also, note that I - V profiles are quite linear at such a highly resistive state.Next, the sample is heated using a cartridge heater em-bedded in the sample holder of the cryostat. Panel (b)shows a set of resistance ( R ) versus temperature ( T ) datarecorded during successive annealing of the sample. Ev-
300 350 400 450 50010 th rd nd Additional Micro-heater R ( ) Temperature (K)
Only CryostatHeater st th a) b) -0.6 -0.3 0.0 0.3 0.6-40-2002040 I ( - a m p ) V (volt) st nd -4-2024 I ( - a m p ) FIG. 2. (a) Current-voltage characteristics recorded at roomtemperature for the sample in different resistive states. (b)Resistance ( R ) versus temperature plots recorded after dif-ferent stages of annealing. Number written beside respectiveprofiles represents the order of the annealing cycle. Temper-ature variation up to the dotted mark was achieved by anexternal heater while the on chip micro-heater was addition-ally used to reach higher temperature values. idently, at every step, R decreases with increasing T ,which is expected for semiconductors. However, inter-estingly R does not go back to the same value when thetemperature is reduced to room temperature after everystep of annealing. Rather it follows a lower path to reacha smaller value. For example, room temperature resis-tance of the device before initiating any annealing pro-cess is recorded to be ∼
230 MΩ. After the 1 st and 2 nd runs, R decreases to 128 MΩ and 44 MΩ, respectively.At the forth annealing step, once the sample tempera-ture is reached 420 K, the highest value achievable usingthe embedded heater, the micro-heater fabricated on thechip is switched on. Current through the micro-heateris increased to 75 mA in steps of 1 mA . These data aremarked as 4 th . It has been found that the micro-heatercan rise the device temperature to 525 K. Interestingly,the resistance of the device shows several orders of mag-nitude reduction after this high temperature annealing. R vs T profile recorded at 5 th step is also shown in fig-ure 2(b). Clearly, at this step the rate of reduction of R with increasing T is much less than that is found inall the previous annealing steps. Room temperature I - V profile for this lowest resistive state is also plotted infigure 2(a). At this state, resistance is measured to be32.8 kΩ, which is about six orders of magnitude less thanthe resistance measured before the sample goes throughany annealing treatment. It has been further noticedthat the resistance is practically unchanged even afterweeks in vacuum. These findings provide an opportunityto arrest the resistive state of the device at any desiredvalue.Change in resistance after annealing can occur eitherdue to the variation in carrier concentration or mobilityor both. In order to track the change of these parame-ters, we perform thermoelectric measurements at roomtemperature after taking the device to different resistivestates. We start with Seebeck coefficient ( S ) measure-ment of the sample when it is at the lowest resistancestate. Later, the sample space is successively purged withair to take the resistance to a higher value. Once a de- sired resistive state is achieved, the chamber is evacuatedto ∼ × − mbar before carrying out the thermoelec-tric measurement. S is found to be negative in all cases,which suggests that electrons are the majority carriers inthis system (n-type). Magnitude of S as a function of theresistance is plotted in figure 3(a). Clearly, | S | increaseswith R . Current under a temperature gradient in a semi- | S | ( m V K - ) R ( ) n D ( c m - ) R ( ) ( c m V - s e c - ) Resistance( ) n D ( c m - ) a) b) FIG. 3. (a) Magnitude of Seebeck coefficient ( | S | ) as a func-tion of device resistance ( R ). (b) Electron concentration( n D ) obtained from S as a function of R . Inset shows thevariation of n D and mobility ( µ ) in log-log plots. conductor can be expressed as I = G ∆ V + G S ∆ T [31],where G the conductance of the sample, ∆ T and ∆ V are the temperature and potential differences betweenthe contacts. In open circuit configuration, since I =0, above equation leads to ∆ V = -( G s /G )∆ T = S ∆ T .Seebeck coefficient S can thus be defined as the ratio be-tween the diffusion current ( G S ∆ T ) and the product ofthe conductance G and ∆ T . While the diffusion currentdepends upon the difference of carrier densities ( δn ) atthe hot and the cold ends of the device, the conductance G is proportional to the average carrier concentration n .Therefore, it can be said that the Seebeck coefficient issomewhat proportional to the ratio of these two quanti-ties ( δn and n ). In an open circuit configuration, the dif-fused charges accumulate on the colder side, which resultsin an upward shift of the Fermi level with respect to thatof the warmer end. However, such a redistribution of car-riers also leads to a potential difference between the twoends (cooler-end turns higher in potential for the elec-trons as compared to the hotter-end). A steady-state isfinally achieved when this built-in potential restricts anyfurther diffusion to take place. This situation is schemat-ically depicted in figure 1(c). The ratio δn / n is expectedto decrease as the background carrier concentration n inthe layer increases. As a result, | S | decreases with resis-tance. An analytical expression for S can be obtained bysolving linearized Boltzmann equation for G s and G [32],which for 2D-semiconductors takes the form[27]: S = − k B q e (cid:34) η − (2 + r ) (cid:82) ∞ f o (cid:15) r +1 d(cid:15) (1 + r ) (cid:82) ∞ f o (cid:15) r d(cid:15) (cid:35) (1)where, k B the Boltzmann constant and q e the elec-tron charge, f o the Fermi distribution function [ f o =1 / exp ( (cid:15) − η )]. r is known as the scattering expo-nent that depends upon the type of scattering mechanism Flushed with respective gases O N R ( ) Time (min) Ar E v a c u a t e d FIG. 4. Change in Resistance of the device after exposingthe sample in different environments. Conductivity of thechannel reduces drastically after introduction of O to thesample space. limiting the carrier relaxation time. In MoS , acous-tic phonons are found to be the most dominant scat-ters for the electrons at room temperature. Note thatfor acoustic phonon scattering, r = 0[27] in 2D. Equa-tion 1 is solved to obtain η = ( E f − E c ) /k B T and subse-quently, 2D electron concentration n D is estimated from n D = 8 πm ∗ k B T h − ln (1 + e η ), where m ∗ and h repre-sent the electronic effective mass and Planck’s constant,respectively. Figure 3(b) shows the variation of n D asfunction of device resistance R . As expected the carrierconcentration of the sample decreases with increasing re-sistance. A lower bound estimate of mobility, µ can bemade from n D and R through µ = l/ ( bRn D q e ), where l and b are the length and width of the device. Inset offigure 3(b) portrays both µ and n D as functions of R inlogarithmic scale. It is interesting to note that beyond athreshold value of R , both the parameters show a sharpchange. n D decreases by several orders of magnitudeand at the same time µ increases significantly.All these observations point towards the fact that en-vironment surrounding these 1L-MoS films plays an im-portant role in governing their resistances. For betterunderstanding, we have measured resistance of the de-vice under different controlled environments. Resultsare shown in figure 4. The sample is first annealed andtaken to a state where the resistance is approximately50 KΩ. Sample space is then flushed with argon (Ar) gas(99.999% pure, oxygen ≤ ≤ ) (99.999% pure, oxy-gen ≤ ≤ ∼ ≤ ≤ R is increased by an order of magnitude and after 5.5 hr, R enhances by about two orders of magnitude.Observations of figure 3 and 4 clearly suggest that ad-sorption/desorption of oxygen molecules at the surface must be the reason for the dramatic change in carrierdensity of the layer. It should be noted that sulfurvacancies are shown to be the most abundant type ofpoint defects, especially in CVD grown MoS . Amongdifferent types of sulfur vacancies in a monolayer film,same side di-sulfur vacancies, SV (absence of two sul-fur ions from the same side of a unit cell) are expectedto act as donors[15, 16]. We believe that at the ambi-ent condition, adsorption of oxygen molecules at thesevacancy sites passivates the donors resulting in the re-duction of the electron concentration in the conductionband. Upon vacuum annealing, these molecules are re-moved from the surface and the electron concentrationrecovers. The range of temperature involved in theseexperiments as well as the reversible nature of the phe-nomenon suggests that the oxygen molecules must be ph-ysisorbed (not chemisorbed) on the surface. In order tocheck the feasibility of this hypothesis theoretical calcula-tions have been carried out within the framework of den-sity functional theory[33] by employing the Vienna Ab-initio Simulation Package (VASP)[34, 35] software. Forthe purpose, we used projector augmented wave (PAW)pseudo-potentials[34, 36] with a kinetic energy cut-off of450 eV, and Perdew-Burke Ernzerhof (PBE) exchange-correlation functional[37]. To optimize the structures,we employed a k-mesh of 5 × ×
1, while for the densityof state (DOS) calculations, 21 × × -6 eV and 0.02 eV˚A − , respectively. The effectsof spin-polarization and spin-orbit coupling were also in-cluded in our calculations. In this work, the van derWaals force correction was incorporated using DFT-D3method[38]. For the relaxed structure, the pressure inthe supercell was less than 0.2 Kbar. At least 20 ˚A vac-uum was introduced along the c -direction to minimizethe spurious interactions. To perform the band structurecalculations, 60 k-points were considered in the recip-rocal space. A 4 × moleculeswith the sulfur-vacancies in MoS monolayer. Two sul-fur atoms are removed from same side of the unit cellas shown in figure 5(a). Calculations show that the va-cancy SV acts as a donor state and give rise to severalnew energy states in the forbidden gap as presented inpanel (b) of figure 5. Note that the top of the donorband lies approximately 0.25 eV below the conductionband minimum (CBM). These donor states appear dueto the hybridization of weak sulfur 3p and strong molyb-denum 4d orbitals. It should be noted that the find-ing is in good agreement with the theoretical predictionsmade earlier[39]. In order to look into the effect of O adsorption at the surface, we first carried out geometryrelaxation with the O molecule placed at an initial guessposition. The energetically most favorable site for attach-ment comes out to be on the top of any vacancy site ofthe defect and ∼ with SV +O ) are depicted in figure 5(c). In a)b) c)d) e) Top viewSide view -1.0-0.50.00.51.01.5 E ne r g y ( e V ) Γ Γ
M K -1.0-0.50.00.51.01.5 E ne r g y ( e V ) Γ Γ
M K
FIG. 5. (a) The relaxed geometry (4 × monolayer with SV . (c) The relaxedgeometry and (d) band structure of the complex system (O adsorbed MoS monolayer with SV2). The red-colored bands inthe gap region are predominantly contributed by the O molecule. (e) Side view of the differential charge density plot of thecomplex system with an isovalue of 9.8 × − e˚A − , where the yellow color indicates the electron-rich region and the cyan colordenotes the loss of electrons. figure 5(d), band structure calculated for SV +O com-plex is shown. Attachment of O creates two flat bands(red lines) that are approximately 250 meV below thedonor levels. Rest of the band structure remains to bealmost the same as that is obtained for 1L-MoS with SV . The charge density difference in O adsorption canbe described as, ρ ad = ρ MoS + SV + O − ρ MoS + SV − ρ O ,where ρ MoS + SV + O , ρ MoS + SV , and ρ O are the chargedensities of the 1L-MoS with SV +adsorbed O , 1L-MoS with SV , and isolated O molecule, respectively.The differential charge density plot of the complex systemis presented in figure 5(e). Note that upon adsorption ofO molecule, the 1L-MoS loses electrons, while theseare accumulated around O molecule suggesting electrontransfer from SV defects to O molecule, which is alsoconsistent with the appearance of energy levels (red lines)below the SV donor states when O is adsorbed at the SV defect site as shown in figure 5(d). Theory thus sup-ports the picture that the adsorption of O molecules pas-sivates SV donor states by introducing a lower energystate that traps these electrons. The formation energy of SV +O complex is estimated to be ∼
340 meV suggest-ing a weak bonding of the O molecules with the surface.This the reason why at elevated temperatures and in high vacuum, O molecules are efficiently detached from thesurface that results in a drastic increase in its conductiv-ity as observed experimentally.Adsorption of oxygen molecules at the sulfur vacancysites results in the formation of electron traps, which lie250 meV below the donor levels introduced by these va-cancies in 1L-MoS layers. As a result, these donorsare passivated and the as-grown layer that is exposedto the ambient condition becomes highly resistive. It hasbeen found that the resistance, as well as the thermo-electric power of the material, can be tuned to a largeextent by controlling the temperature of annealing andoxygen exposure. Once established, these parameters re-main unchanged unless the layer is exposed to oxygenenvironment. Annealing followed by an effective encap-sulation method can offer a unique way to control dopingin the material, which is an important milestone towardstechnology application. Moreover, these findings gener-ate excitements to use 1L-MoS for oxygen gas sensingapplications.We acknowledge the financial support by Council ofScientific & Industrial Research of Government of Indiaunder the project Code:03(1403)/17/EMR II. We thankCentre of Excellence in Nanoelectronics -CEN of IITBombay, for providing various experimental facilities. [1] K. Roy, T. Ahmed, H. Dubey, T. P. Sai, R. Kashid,S. Maliakal, K. Hsieh, S. Shamim, and A. Ghosh, Ad- vanced Materials , 1704412 (2018).[2] D. Unuchek, A. Ciarrocchi, A. Avsar, Z. Sun, K. Watan- abe, T. Taniguchi, and A. Kis, Nature Nanotechnology , 1104 (2019).[3] K. K. H. Smithe, C. D. English, S. V. Suryavanshi, andE. Pop, Nano Letters , 4516 (2018).[4] D. J. Hynek, J. V. Pondick, and J. J. Cha, APL Mate-rials , 030902 (2019).[5] Y. Tang, K. F. Mak, and J. Shan, (2019).[6] J. V. Pondick, J. M. Woods, J. Xing, Y. Zhou, and J. J.Cha, ACS Applied Nano Materials , 5655 (2018).[7] D. Dumcenco, D. Ovchinnikov, O. Sanchez, P. Gillet,D. Alexander, S. Lazar, A. Radenovic, and A. Kis, 2DMater. , 044005 (2015).[8] P. Budania, P. T. Baine, J. H. Montgomery, D. W. Mc-Neill, S. J. N. Mitchell, M. Modreanu, and P. K. Hurley, , 025022 (2017).[9] G. Z. Magda, J. Pet, G. Dobrik, C. Hwang, L. P. Bir,and L. Tapaszt, Scientific Reports , 14714 (2015).[10] S. H. Baek, Y. Choi, and W. Choi, Nanoscale Res. Lett. , 388 (2015).[11] K. K. H. Smithe, S. V. Suryavanshi, M. Muoz Rojo, A. D.Tedjarati, and E. Pop, ACS Nano , 8456 (2017).[12] D. Dumcenco, D. Ovchinnikov, K. Marinov, P. Lazi,M. Gibertini, N. Marzari, O. L. Sanchez, Y.-C. Kung,D. Krasnozhon, M.-W. Chen, S. Bertolazzi, P. Gillet,A. Fontcuberta i Morral, A. Radenovic, and A. Kis,ACS Nano , 4611 (2015).[13] S. Najmaei, M. Amani, M. L. Chin, Z. Liu, A. G. Bird-well, T. P. ORegan, P. M. Ajayan, M. Dubey, and J. Lou,ACS Nano , 7930 (2014).[14] O. V. Yazyev and Y. P. Chen, Nature Nanotechnology ,755.[15] V. K. Sangwan, D. Jariwala, I. S. Kim, K.-S. Chen, T. J.Marks, L. J. Landhon, and M. C. Hersam, Nature Nan-otechnology , 403 (2015).[16] V. K. Sangwan, H.-S. Lee, H. Bergeron, I. Balla, M. E.Beck, K.-S. Chen, and M. C. Hersam, Nature , 500(2018).[17] S. Tongay, J. Suh, C. Ataca, W. Fan, A. Luce, J. S. Kang,J. Liu, C. Ko, R. Raghunathanan, J. Zhou, F. Ogletree,J. Li, J. C. Grossman, and J. Wu, Scientific Reports ,2657 (2013).[18] R. Rao, V. Carozo, Y. Wang, A. E. Islam, N. Perea-Lopez, K. Fujisawa, V. H. Crespi, M. Terrones, andB. Maruyama, 2D Materials , 045031 (2019).[19] K. Park, H. Kang, S. Koo, D. Lee, and S. Ryu, NatureCommunications , 4931 (2019).[20] P. K. Gogoi, Z. Hu, Q. Wang, A. Carvalho, D. Schmidt, X. Yin, Y.-H. Chang, L.-J. Li, C. H. Sow, A. H. C. Neto,M. B. H. Breese, A. Rusydi, and A. T. S. Wee, Phys.Rev. Lett. , 077402 (2017).[21] S. Tongay, J. Zhou, C. Ataca, J. Liu, J. S. Kang, T. S.Matthews, L. You, J. Li, J. C. Grossman, and J. Wu,Nano Letters , 2831 (2013).[22] S. Goswami, C. Siegert, M. Baenninger, M. Pepper,I. Farrer, D. A. Ritchie, and A. Ghosh, Phys. Rev. Lett. , 026602 (2009).[23] C.-H. Lee, G.-C. Yi, Y. M. Zuev, and P. Kim, AppliedPhysics Letters , 022106 (2009).[24] P. Wei, W. Bao, Y. Pu, C. N. Lau, and J. Shi, Phys.Rev. Lett. , 166808 (2009).[25] S. Roddaro, D. Ercolani, M. A. Safeen, S. Suomalainen,F. Rossella, F. Giazotto, L. Sorba, and F. Beltram, NanoLetters , 3638 (2013).[26] J. Wu, H. Schmidt, K. K. Amara, X. Xu, G. Eda, andB. zyilmaz, Nano Letters , 2730 (2014).[27] K. Hippalgaonkar, Y. Wang, Y. Ye, D. Y. Qiu, H. Zhu,Y. Wang, J. Moore, S. G. Louie, and X. Zhang, Phys.Rev. B , 115407 (2017).[28] M. Buscema, M. Barkelid, V. Zwiller, H. S. J. van derZant, G. A. Steele, and A. Castellanos-Gomez, NanoLetters , 358 (2013).[29] P. K. Mohapatra, S. Deb, B. P. Singh, P. Vasa, andS. Dhar, Appl. Phys. Lett. , 042101 (2016).[30] S. Deb, P. Chakrabarti, P. K. Mohapatra, B. K. Barick,and S. Dhar, Applied Surface Science , 542 (2018).[31] I. Terasaki, in Materials for Energy Conversion Devices ,edited by C. C. Sorrell, S. Sugihara, and J. Nowotny(Woodhead Publishing, 2005) pp. 339 – 357.[32] X. Zhang and Y. Pei, npj Quantum Materials , 68(2017).[33] W. Kohn and L. J. Sham, Phys. Rev. , A1133 (1965).[34] G. Kresse and J. Furthm¨uller, Phys. Rev. B , 11169(1996).[35] G. Kresse and J. Furthm¨uller, Computational MaterialsScience , 15 (1996).[36] P. E. Bl¨ochl, Phys. Rev. B , 17953 (1994).[37] J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev.Lett. , 3865 (1996).[38] S. Grimme, J. Antony, S. Ehrlich, and H. Krieg, TheJournal of Chemical Physics , 154104 (2010).[39] H. Qiu, T. Xu, Z. Wang, W. Ren, H. Nan, Z. Ni, Q. Chen,S. Yuan, F. Miao, F. Song, G. Long, Y. Shi, L. Sun,J. Wang, and X. Wang, Nature Communications4