Visualizing the metal-Mo S2 contacts in two-dimensional field-effect transistors with atomic resolution

Ryan J. Wu, Sagar Udyavara, Rui Ma, Yan Wang, Manish Chhowalla, Turan Birol, Steven J. Koester, Matthew Neurock, K. Andre Mkhoyan

Research output: Contribution to journalArticlepeer-review

8 Scopus citations

Abstract

Two-dimensional MoS2 is an excellent channel material for ultrathin field-effect transistors, but high contact resistance across the deposited metal-MoS2 interface continues to limit its full realization. Using atomic-resolution scanning transmission electron microscopy and first-principles calculations, we showed that deposited metals with a high affinity for sulfur could have a fundamental limitation. Ti-MoS2 contact shows a destruction of the MoS2 layers, a formation of clusters and void pockets, and penetration of Ti into MoS2, resulting in many localized pinning states in the band gap. InAu-MoS2 contact shows that it is possible to achieve a van der Waals-type interface and dramatically reduced pinning states.

Original languageEnglish (US)
Article number111001
JournalPhysical Review Materials
Volume3
Issue number11
DOIs
StatePublished - Nov 8 2019

Bibliographical note

Funding Information:
et al. Wu Ryan J. 1 * Udyavara Sagar 1 * Ma Rui 2 Wang Yan 3 Chhowalla Manish 3,4 Birol Turan 1 Koester Steven J. 2 Neurock Matthew 1 † https://orcid.org/0000-0003-3568-5452 Mkhoyan K. Andre 1 ‡ Department of Chemical Engineering and Materials Science, 1 University of Minnesota , Minneapolis, Minnesota 55455, USA Department of Electrical and Computer Engineering, 2 University of Minnesota , Minneapolis, Minnesota 55455, USA Department of Materials Science and Engineering, 3 Rutgers University , Piscataway, New Jersey 08854, USA Department of Materials Science and Metallurgy, 4 University of Cambridge , Cambridge CB3 0FS, United Kingdom * These authors contributed equally to this work. † mneurock@umn.edu ‡ mkhoyan@umn.edu 8 November 2019 November 2019 3 11 111001 27 April 2019 ©2019 American Physical Society 2019 American Physical Society Two-dimensional Mo S 2 is an excellent channel material for ultrathin field-effect transistors, but high contact resistance across the deposited metal- Mo S 2 interface continues to limit its full realization. Using atomic-resolution scanning transmission electron microscopy and first-principles calculations, we showed that deposited metals with a high affinity for sulfur could have a fundamental limitation. Ti- Mo S 2 contact shows a destruction of the Mo S 2 layers, a formation of clusters and void pockets, and penetration of Ti into Mo S 2 , resulting in many localized pinning states in the band gap. InAu- Mo S 2 contact shows that it is possible to achieve a van der Waals-type interface and dramatically reduced pinning states. National Science Foundation 10.13039/100000001 DMR-1420013 National Institute of Standards and Technology 10.13039/100000161 Ultrathin field-effect transistors (FETs) using Mo S 2 as the channel material have shown excellent performance, making them viable for sub-10-nm node devices [1–5] . However, the contact between the two-dimensional (2D) Mo S 2 and an evaporated three-dimensional (3D) metal electrode remains a challenge due to the high contact resistance [6–9] attributed to the Fermi level pinning at the interface [10–13] . Although alternative approaches [9,14–19] to depositing the metal onto Mo S 2 have been reported [20–22] , including a direct metal film transfer which results in van der Waals (vdW) bonding [23,24] , the origin of high contact resistance using evaporated metals remained unclear. Previous reports have attributed the Fermi level pinning to the formation of surface states created by adsorbed contaminants [25] or damage by kinetic energy transfer from metal deposition [23] . While these factors could play a role, they should be resolvable by improving the metal deposition. On the other hand, the actual bonding between a metal and Mo S 2 layer is a more fundamental issue. The recent report by Wang et al. [26] shows that when In or I n x A u y is used as the contact metal, low-resistant vdW contact can be achieved even with evaporated metal deposition. Therefore, understanding of the metal- Mo S 2 interface from direct atomic-scale observations can be instrumental in mitigating Fermi level pinning. To elucidate the structure of the metal- Mo S 2 interface, we first studied deposited Ti contacts. In addition to Ti [12,13,27,28] , other metals, such as Au, Pt, Ag, Sc, Pd, Ni, and Cu, were also used as the contact [4,23,27,29] , but Ti provides a good case for a metal- Mo S 2 interface with strong bonds, as Ti possesses a very high affinity for sulfur (the Ti-S bond dissociation energy is 4.35 eV compared to Mo-S at 3.69 eV) [30] . We also studied the I n x A u y - Mo S 2 interface, where vdW-type bonding was reported [26] . Interfaces are studied using a combination of atomic-resolution scanning transmission electron microscopy (STEM) imaging with electron energy-loss spectroscopy (EELS). Density functional theory (DFT) calculations are carried out to further clarify the structural and electronic changes occurring at the deposited metal- Mo S 2 interface. The STEM specimens were prepared by focused ion beam (FIB) thinning of working FETs [see the Methods and Materials section, Fig. S1, in the Supplemental Material (SM) [31] ] with characteristics similar to those reported in the literature [32,33] . Figures  1(a) and 1(b) show the schematic layout of the FETs and a low-magnification cross-sectional annular dark-field (ADF)-STEM image where the Au/Ti contact and Si / Si O 2 substrate bookend the Mo S 2 layers (see SM Fig. S2 [31] ). A high-magnification ADF-STEM image [Fig.  1(c) ] shows that along the Ti- Mo S 2 interface, there are areas where the Ti contacts the Mo S 2 and alters the structure of the topmost layer. Next to these areas, small void pockets are visible which leaves the topmost Mo S 2 layer pristine. While there are reports suggesting that Ti forms strong bonding with Mo S 2 [25,28] , this is a direct observation of (i) the degradation of the topmost Mo S 2 layer and (ii) the presence of nm-size void pockets along the interface. The Ti atoms tend to not only bond but also cluster on the Mo S 2 surface during deposition. Such Ti clustering behavior was predicted [25] but not observed experimentally (see SM Fig. S3 for additional images [31] ). It should be noted that under the conditions of Ti deposition used here, there should be no damage to the Mo S 2 layers. 10.1103/PhysRevMaterials.3.111001.f1 1 FIG. 1. (a) Schematic layout of the FET with Mo S 2 channel and metal contacts. (b) Low-magnification cross-sectional ADF-STEM image of the FET. The protective amorphous C/Pt layers are also visible here. Scale bar is 0.2 μm. (c) High-magnification image of the Ti- Mo S 2 interface from the boxed area in (b). An area where Ti is clustered is indicated by a yellow arrow and areas with void pockets by white arrows. Scale bar is 2 nm. (d) Atomic-resolution ADF-STEM image of the Ti- Mo S 2 interface. The horizontally averaged ADF intensity is shown on the right. A ball-and-stick model of Mo S 2 is overlaid on the image. Scale bar is 6 Å. An atomic-resolution ADF-STEM image [Fig.  1(d) ] obtained from one of these Ti-clustered areas shows that the topmost Mo S 2 layer is degraded and barely identifiable. The compromised integrity of the topmost Mo S 2 layer indicates that the bonding between Ti and S is strong enough to restructure the contacting Mo S 2 layer. This Ti-driven restructuring of the topmost Mo S 2 layer is far beyond the effects of carbon or oxide contaminants on the surface of the Mo S 2 as (i) no degradation is observed in the sections of the Mo S 2 layers under the void pockets which have the same level of contaminants, and (ii) in areas of the device not directly below the Ti contacts, the Mo S 2 remains pristine (see SM Fig. S3 [31] ). A slightly lower intensity of the ADF signal in the Ti region directly above the first Mo S 2 layer is likely due to void pockets in the projection, or an nm-thick Ti sublayer with a lower atomic density, or both. To measure the changes in the electronic structure of Mo S 2 caused by the Ti contact, a layer-by-layer EELS analysis was performed. Core-level EELS edges measure localized changes in the element-specific electronic density of states (DOS) of the conduction band [34] . Figure 2(a) shows two S L 2 , 3 edges measured from the Mo S 2 channel: one from layer 1, the topmost Mo S 2 layer directly in contact with Ti, and from layer 5. The dominating features of the S L 2 , 3 fine structure, peaks I and II, composed of S 3 s and 3 d partial DOS [35] , are different in these spectra. The peaks are more subdued in layer 1, which is consistent with the observed loss of crystallinity of the top Mo S 2 layer [34] . Additional S L 2 , 3 edge EELS measurements from the Mo S 2 not in contact with Ti showed no differences between the first and fifth Mo S 2 layers (see SM Fig. S4 [31] ). 10.1103/PhysRevMaterials.3.111001.f2 2 FIG. 2. (a) EELS S L 2 , 3 edges measured from the first (contacting to Ti) and the fifth Mo S 2 layers. The differences between the two spectra are shown below. (b) Atomic-resolution ADF-STEM image of the Mo S 2 layers (left) and EELS S L 2 , 3 edge measured from the layers (right). Scale bar is 5 Å. Measured spectra are shown as scatter points and fitted spectra are shown as lines. (c) The fractions of the interfacial (layer 1) character in each S L 23 edge. (d) The fraction of Ti in Mo S 2 layers. A y = erf ( x ) fit through the data points in (c) and (d). Figure 2(b) shows a set of core-level EELS measurements from each of the first seven Mo S 2 layers and from the Ti contact layer directly above the Mo S 2 . Changes in peaks I and II of the S L 2 , 3 edge were quantified by fitting each spectrum to a linear superposition of the two reference spectra shown in Fig.  2(a) [36] , and the fractions of the spectrum with interfacial and bulk characters were evaluated [Fig.  2(c) ]. Layer 2 and even layer 3 of the Mo S 2 have considerable interfacial character, indicating that the effects of the Ti contact go beyond the surface layer (for minor effects of probe broadening, see the Methods and Materials section in the SM [31] ). The presence of Ti atoms in deeper Mo S 2 layers was also considered. The Ti L 2 , 3 edge was measured across the first seven Mo S 2 layers [Fig.  2(d) ]. Layers 2 and 3 also showed an appreciable amount of Ti present. The amounts of Ti present in layers 2 and 3 could be a factor in the observed changes in the fine structure of the S L 2 , 3 edge at these depths. However, they are not high enough to affect the pristinelike view of the atomic structure imaged in the projection [Fig.  1(d) ]. DFT calculations were carried out to understand the interactions of Ti atoms with Mo S 2 layers. In these simulations, individual Ti atoms were systematically added onto the surface of a monolayer Mo S 2 to mimic experimental deposition. This “single-atom-addition” approach provides insight into the atomic processes occurring at the metal- Mo S 2 interface during deposition, and complements the “metal- Mo S 2 -slab” approach [37–40] , which models metal contacts directly transferred onto Mo S 2 [23] . Simulations were performed without the effects of temperature and Ti atoms were added without kinetic energy (for details, see the Methods and Materials section in the SM [31] ). Figures  3(a) and 3(b) show the calculated lowest-energy structures of the Ti- Mo S 2 interfaces after the addition of one to five Ti atoms. As will be clear later, a five-atom cluster of Ti was sufficient to explain the main STEM observations discussed earlier. The interactions between the Ti and S are indeed strong enough to disrupt the Mo S 2 by pulling S atoms out of the Mo S 2 surface. Calculations also show the formation of a cluster, which degrades the pristine crystal structure of Mo S 2 . Furthermore, even with only five Ti atoms, a relatively large opening in the Mo S 2 layer is formed [Fig.  3(b) ]. This “nanopore” in the Mo S 2 is large enough to allow the sixth Ti atom to penetrate and cause more structural modifications [Fig.  3(c) ]. These results match with the experimentally observed structural degradation of the topmost Mo S 2 layer and provide a pathway to how appreciable amounts of Ti could penetrate into the second and third layers of Mo S 2 . They also imply that the disruption of the Mo S 2 crystal is inherent to Ti- Mo S 2 bonding and cannot be avoided. Additional ab initio molecular dynamics (AIMD) calculations for the 5-Ti-atom cluster confirm its thermal stability at room temperature ( T = 300 K ) (see SM Video S1 [31] ). 10.1103/PhysRevMaterials.3.111001.f3 3 FIG. 3. (a) Models showing the lowest-energy optimized structures of the Ti- Mo S 2 system following additions of one, two, and three Ti atoms (Mo, purple; S, yellow; Ti, blue). (b) Side view (above) and top view (below) of the Ti- Mo S 2 system with five added Ti atoms. An opening in the Mo S 2 layer is highlighted as the gray shaded region. (c) Side view (above) and top view (below) of the Ti- Mo S 2 system showing the penetration of the sixth Ti atom into the opening highlighted in (b). The electronic band structures and DOS for a pristine monolayer Mo S 2 and for the Mo S 2 with five bonded Ti atoms, shown in Fig.  3(c) , were calculated (Fig.  4 ). As can been seen, a cluster of five Ti atoms will force the system to have many localized states in the band gap, some of which will act as pinning states for charge carriers. These band-gap states appear to be located on all three atomic species (Mo, S, and Ti) and likely pin the Fermi level at the interface [11,41] . The degradation of the crystallinity of the Mo S 2 also results in the removal of degeneracies in the band structure and the flattening of the bands leading to dispersed DOS, which was observed in S L 2 , 3 EELS measurements [Fig.  2(a) ]. 10.1103/PhysRevMaterials.3.111001.f4 4 FIG. 4. (a) The atomic structure of bilayer Mo S 2 with distances between sulfur layers indicated. (b) Electronic band structure and element-specific DOS for pristine monolayer Mo S 2 . (c)–(h) The atomic and electronic band structure of monolayer Mo S 2 with five atoms of Ti, Sc, Cu, Au, In, and I n x A u y (with three In and two Au). The charge densities for all structures are shown in transparent green on the top of the corresponding band structure. An isosurface value of 0.042 was chosen conservatively to ensure that the charge densities of the Cu atoms overlap. The minimum of the conduction band and maximum of the valence band are highlighted by maroon colors. The DFT calculations using the single-atom-addition approach were extended to other metals: Sc, Cu, Au, In, and I n x A u y (Fig.  4 ). The results show that Sc, a transition metal as Ti, clusters and disrupts the Mo S 2 layer. Both Sc and Ti disperse the valence and conduction bands and create a high number of pinning states in the band gap. In contrast, Cu and Au bind weakly to the surface sulfur and do not disrupt the structure of the Mo S 2 layer. Their effects on the electronic band structure of Mo S 2 , including a number of new band-gap states, are also weaker. Of the metals studied, In and I n x A u y are the least disruptive. The charge density plots in Fig.  4 show that In and I n x A u y bind to Mo S 2 through vdW-type bonds, and leave the Mo S 2 structure intact (for the thermal stability of the 5-In-atom cluster at T = 300 K , see SM Video S2 [31] ). This is also reflected in the electronic states, where the number of band-gap states introduced by In or I n x A u y is very low. The calculations suggest that In or I n x A u y should show minimum pinning and very low contact resistance, which is consistent with the measured low contact resistance in the FETs reported by Wang et al. [26] . In the case of the I n x A u y alloy, In provides vdW-type bonding to Mo S 2 with a very low number of gap states, while Au provides bonds with a light ionic character without considerably modifying the surface of Mo S 2 and, therefore, low-barrier electron transition paths across the vdW gap. Due to similarities between the Au and Cu contacts to the Mo S 2 layer [Figs.  4(e) and 4(f) ], the I n x C u y alloy could also be a good candidate for a low-cost, higher melting point metal with vdW-type low contact resistance. Calculations show that even in the presence of an S vacancy at the Mo S 2 surface, In atoms bind with the vdW-type bond, while Ti atoms, as before, disrupt the Mo S 2 with strong Ti-S bonds (see SM Fig. S5 [31] ). STEM-EELS experiments were performed for the I n x A u y − Mo S 2 interface [26] to compare with the results of the DFT calculations based on the single-atom-addition approach. An atomic-resolution ADF-STEM image and corresponding EELS characterization of the Mo S 2 layers are presented in Fig.  5 . As can be seen, the I n x A u y - Mo S 2 contact is nondisruptive and the topmost Mo S 2 layer appears completely intact. The distance between the top sulfur layer and the first atomic layer I n x A u y is 2.7 ± 0.1 Å, which matches well with the DFT predicted distance of 2.69 Å. EELS measurements of the S L 2 , 3 edge show no detectable differences between spectra from layer 1, the topmost Mo S 2 layer directly in contact with I n x A u y , and the bulklike Mo S 2 layer 5 [Fig.  5(b) ], which is also consistent with the results of DFT calculations showing very minor changes in conduction band DOS. 10.1103/PhysRevMaterials.3.111001.f5 5 FIG. 5. (a) Atomic-resolution ADF-STEM image of the I n x A u y - Mo S 2 interface. The horizontally averaged ADF intensity is shown on the left. Scale bar is 5 Å. (b) EELS S L 2 , 3 edge measured from Mo S 2 layers 1 to 5. The difference spectrum between the first (contacting to the I n x A u y ) and the fifth Mo S 2 layers is shown below. In conclusion, the atomic-resolution STEM-EELS study of a metal-deposited metal- Mo S 2 interface shows that for Ti the strength of the metal-sulfur interaction is sufficient to result in the degradation of the surface Mo S 2 layer, the penetration of Ti into deeper layers, the clustering of Ti atoms, and the formation of void pockets, making the interface inhomogeneous. DFT calculations suggest such structural modifications of the metal- Mo S 2 interface are inherent for systems where the metal has a very high affinity for sulfur (Ti, Sc, etc.). Band-structure calculations suggest that the Fermi level pinning in such systems is likely unavoidable. In contrast, for a metal with a low affinity for sulfur, such as In, the resulting interface can become vdW type. While the Fermi level pinning might still be present for a wide variety of metals, including Au or Cu, they can be dramatically mitigated with the proper selection of a metal or alloy, such as In, I n x A u y , or I n x C u y . Introduced here, the single-atom-addition approach in DFT calculations can be an effective method to evaluate the effects of deposited metals on Mo S 2 and other layered materials. The STEM analysis was performed in the Characterization Facility of the University of Minnesota, which receives partial support from the NSF through the MRSEC program. This project was partially supported by NSF MRSEC Program Grant No. DMR-1420013, the C-SPIN, one of the SRC STARnet centers, and SMART, one of seven centers of nCORE, a SRC program sponsored by NIST.

How much support was provided by MRSEC?

  • Partial

Reporting period for MRSEC

  • Period 6

Fingerprint

Dive into the research topics of 'Visualizing the metal-Mo S2 contacts in two-dimensional field-effect transistors with atomic resolution'. Together they form a unique fingerprint.

Cite this