You are on page 1of 4

Impact of impurities, interface traps and contacts on

MoS2 MOSFETs: modelling and experiments


Gioele Mirabelli, Farzan Gity, Scott Monaghan, we report the electrical characterization of back-gated MoS2
Paul K. Hurley, Ray Duffy FETs. Based on the experimental device characteristics a
model and parameter set for MoS2 was implemented in the
Tyndall National Institute, University College Cork continuum-based Synopsys TCAD simulator Sentaurus
Lee Maltings Complex, Cork, Ireland Device, which showed consistent behavior with the
experimental data. Charged impurity concentrations, interface
Abstract— Device modelling is a key enabling capability for states and MoS2/metal contact barriers were varied to explore
the semiconductor industry, especially for process optimisation, the parameters which influence carrier transport in MoS2
and for insight into the physics of novel architectures and FETs.
materials that are difficult to access experimentally. Despite
much innovative experimental work, device modelling II. EXPERIMENT
capabilities for field effect devices based on Transition Metal The experimental procedure is reported in Fig. 1.
Dichalcogenide (TMD) channel materials are at an early stage of Mechanical exfoliation with scotch tape was used to obtain thin
development. Properly formulated physics-based models would
flakes from a MoS2 bulk crystal. The flakes were transferred on
give a substantial improvement for time- and cost-effective
development of TMD devices. In this work, experimental device
a substrate of 85 nm of dry thermally grown SiO2 and a highly-
data was used to develop models and parameter sets in the boron doped Si handle wafer. The height of the flakes was
continuum-based Synopsys Sentaurus Device software. established by optical color-contrast to be 3-4 layers [6]. Ti/Au
Specifically, few-layer MoS2 Field-Effect-Transistors (FETs) metal contact pads and electrodes were defined by electron-
were systematically electrically characterized, and the modelling beam lithography, followed by metal evaporation and lift off
of the experimental data focused on the impact of impurities, process. 15 kV beam exposures were performed with a Zeiss
interface traps, and contact barriers. Furthermore, the SUPRA SEM with a Raith Elephy Plus blanker. An adhesion
experimental MoS2 FETs device characteristics, combined with layer of 5 nm of Ti and 45 nm of Au were deposited with e-
the physics based transport models, suggests that the low beam evaporation. For structural analysis, cross-section
experimental electron mobility values are a result of a high samples were obtained by FEI’s Dual Beam Helios Nanolab
density of charge impurity defects in the MoS2 channel. To the 600i system using Ga ion beam. Cross-sectional Transmission
best of our knowledge continuum-based TCAD device models did Electron Microscopy (XTEM) imaging was carried out using a
not previously exist for MoS2, or TMD-semiconductors in JEOL 2100 HRTEM operated at 200 kV in Bright Field mode
general. using a Gatan Double Tilt holder. Electrical characterization
was performed on a wafer at 25oC in a dry/dark environment
Keywords—MoS2; Device Modeling; Impurities; Electrical using the HP4156C parameter analyzer.
characterization; Physics;
I. INTRODUCTION III. MATERIAL AND ELECTRICAL ANALYSIS
In recent years the interest in 2D-semiconductors, and Fig. 1 shows the Scanning Electron Microscope (SEM) and
Transition-Metal-Dichalcogenides (TMDs) in particular, has cross-sectional Transmission Electron Microscope (TEM)
increased noticeably. Due to their intrinsic 2D nature the
electrostatic control of the gate is higher with respect of 3D-
semiconductors, making them more immune to Short Channel
Effects [1]. One of the issues associated with 2D-materials is
the lack of a suitable large-area growth technique, which
limits the systematic study TMD materials grown over large
are substrates. Often, TMD films are characterized by defects
or grain boundaries, which can limit their electrical
performances. Therefore, most of the experimental results in
literature have been obtained from flakes mechanically
exfoliated from bulk crystals. To date these have shown the
best electrical results, with high on/off ratios, low inverse
subtreshold slopse and carrier mobilities in the range of tens of
cm2V-1s-1 [2].
Nevertheless, a large concentration of impurities was
reported in TMD crystals [3], and if in a charged state, these
impurities will degrade the carrier mobility and resistivity of Fig. 1: Top: summary of the experimental process flow and optical image
the material [4] especially in TMD films comprised of a small of the contacted MoS2 flake. Bottom: (a) SEM top-view of the MoS2 device
(1-10) number of monolayers [5]. In the first part of this work, after contact deposition. (b) Representative high-resolution TEM image
showing the layered structure of the sample.

978-1-5090-5978-2/17/$31.00 ©2017 IEEE 288

Authorized licensed use limited to: Indian Instt of Engg Science & Tech- SHIBPUR. Downloaded on August 23,2021 at 11:19:21 UTC from IEEE Xplore. Restrictions apply.
Ma and Jena [11] reported a systematic study of carrier
transport in 2D crystals, which are found to be highly
dependent on the dielectric environment and ionized impurity
density. In particular, the mobility versus electron density for
low carrier concentration follows the relationship

(1)

where µCI is mobility, NCI is ionized impurity concentration


and ACI is a fitting parameter that depends on the dielectric
environment. Considering that ne can be obtained
experimentally, impurity concentration can be extracted by
fitting the equation with the field-effect mobility of the
Fig. 2: Transfer characteristics of the Trapezium for different drain-source devices. Fitted data in [11] shows an impurity concentration of
voltages. Inset: transfer characteristics of the Kite and the Triangle at Vds=1V. 1013 cm-2, while in [12], where the same method is used,
images of the “Trapezium” device, so named because of its different levels of impurity concentration (1011 and 1013 cm-2)
shape. The device width is approximately 3.5 µm along the have been reported, probably due to the high variability of
length of the sample and the channel length is confirmed to be TMDs and film/device preparation conditions. This is also
1 µm as determined after the metal deposition. The TEM cross consistent with the relationship of Hall effect mobility to
section of the same device, confirms the 3-layers thickness
doping concentration observations in natural and synthetic
implied by optical contrast.
TMD crystals [4].
Fig. 2 shows the transfer characteristics of the device for
Fig. 3b shows the impurity concentration level for each
different Vds. The transfer characteristics at Vds=1V are
device, extracted from fitting the mobility using eq. (1). The
presented in the inset of Figure 2 for two other back gated
average impurity concentration is 4×1013 cm-2. Similar levels
MoS2 FETs, termed the “Kite” and the “Triangle” based on
of impurity were found before by ICMPS analysis on MoS2
the MoS2 flake shapes on which the devices were made. These
crystals [3]. The target in the ITRS roadmap [13] for
devices are also 3L thick. The electrical results among the
unintentional impurity concentration in semiconductors
three devices are highly consistent. The drain current at a back
intended for logic device application is for values less than
gate voltage of 7.5 V and a drain-source voltage of 1V is ~50
5×1010 cm-2.
nA/µm, limited probably by contact resistance and/or low
mobility. The off-current is limited by the sensitivity of the
IV. TCAD SIMULATIONS
measurement system, hence could potentially be lower. The
To understand the implications on the electrical performance of
MoS2 flakes reported here were exfoliated from the same bulk
such a high impurity concentration the Synopsys TCAD
material onto the same SiO2/Si substrate and experienced the
simulator Sentaurus Device [14] was used to develop a model
same process steps and environment-related exposure. As a and parameter set for MoS2. To date a clear model for MoS2 is
consequence, the slight variability among these devices is still missing. Due to the lack of a proper growth process, the
most probable related to intrinsic factors only, such as defects low yield of the mechanical exfoliation and the high variability
[7], grain boundaries and/or impurities [8], as opposed to of TMDs it is difficult to obtain consistent results and have
process-related factors such as metal contacting [9] or air systematic experiments. The use of a TCAD simulator allows
exposure [10]. an investigation of what the physical processes are currently
To compare the characteristics of the devices field-effect limiting device performance.
mobility values were extracted from the derivative of the
A schematic cross section of the structure used for
transfer characteristics. Fig. 3a shows the field effect mobility
simulation studies is shown in Fig. 4. This replicates the actual
values plotted against carrier concentration, evaluated
structure used in the experimental section. The main
considering a linear charge dependence on the gate voltage
parameters for MoS2 are obtained from literature. Some of the
overdrive. Due the low back-gate voltage (maximum value is parameters, such as band-gap [15] and in-plane dielectric
7.5 V) and the relatively thick back gate oxide (85 nm of SiO2) constant [16] are dependent on the actual thickness of the
the vertical field contribution to mobility degradation is not MoS2. Then, to match the experimental results of the
expected to be a significant contribution. This is confirmed in Trapezium a mobility model dependent on impurities, defective
Fig. 3, where the field effect mobility increases with the
calculated carrier density.

Fig. 3: (a) Field effect mobility versus carrier density extracted from the Fig. 4: Schematic of the device structure implemented in Sentaurus device
transfer characteristic at Vds=1V. (b) Impurity concentration extracted from and the main parameters. The symbol * denotes thickness-dependent
the mobilities in part (a) using eq. (1). parameters.

289

Authorized licensed use limited to: Indian Instt of Engg Science & Tech- SHIBPUR. Downloaded on August 23,2021 at 11:19:21 UTC from IEEE Xplore. Restrictions apply.
contacts and interface traps were implemented. The Unified
Philips mobility model was used to take into account the
impurity dependency [17]

(2)

where µCI is the mobility dependent on coulomb impurities,


and
(3)

(4)
where ND and NA are the ionized donor and acceptor Fig. 6: Simulated transfer characteristics varying the barrier at the source and
concentrations, p is the hole concentration, G(Pe) and F(Pe) are drain contacts. Inset: same curves in linear scale.
analytical functions describing minority impurity and electron states are introduced close to the MoS2 conduction band edge
hole scattering. fe is a fitting parameter equal to 1. µe,N and µe,C [19]. In order to match the subthreshold behavior of the
depend on the maximum and minimum mobility values, set as Trapezium 4 trap energies are introduced, with energy levels at
1000 and 9 cm2V-1s-1 respectively [10] Et1=1.25kT, Et2=2.5kT, Et3=5kT and Et4=10kT from the
conduction band and with concentration of 8×1011, 3×1011,
(5) 2.5×1011 and 1×1011 cm-2, respectively.
Fig. 7 illustrates the difference in the subthreshold behavior
of the device without any interface traps, with Et1 and Et2 only
(the traps closest to the conduction band), and Et3 and Et4 only
(6) (the traps closest to the mid-gap). A schematic of the trap
energy levels is presented as inset in Fig. 7. As expected the
The model proposed in [11] was used as reference and the traps closest to the mid-gap affect more the subthreshold
main parameters, i.e., Nref and α, are found to be 4×1018 cm-3 behavior, while the ones close to the conduction band affect the
and 1.0 for monolayer MoS2. As reported in Fig. 5, these on-current. Fig. 8 compares the experimental and the simulated
values of Nref and α reproduce the magnitude and trend of the data, with an impurity concentration of 4×1013 cm-2.
relationship between mobility versus charged impurity density Since the model and the parameter set are in good
reported for MoS2 in Ref. [11], especially for high impurity agreement with the experimental data, it can be used to gain a
concentration, which is the region of most interest for this deeper understanding of the main physical processes which are
study. Impurities are introduced as mid-gap “fixed charges”, limiting the drain to source current of the back gated MoS2
which in Sentaurus are traps always completely occupied. MOSFETs, focusing on the impact of the impurity
Furthermore, barriers were used at the contact to take into concentration. Fig. 9a shows the field-effect mobility of MoS2
account the high Schottky barriers usually found in TMD- FETs (Vds=0.1V) varying impurity concentration for a
FETs. The value used to match the experimental results of the different number of layers (no interface traps or contact
Trapezium is 0.17 eV, which is close to the experimentally barriers are considered). As can be seen, the mobility is highly
reported value [18]. Fig. 6 illustrates the impact of the Schottky dependent on charged impurity concentration and it is
barrier height at the metal/MoS2 interface on the transfer probable that the experimental field effect mobility values
characteristics. extracted to date have been significantly influenced (degraded)
The subthreshold behavior of the device is also dependent due to the high impurity concentration.
on interface traps located at the SiO2/MoS2 interface of the Experimentally, mobility shows a peak between 5-10
back gated device structure. To model the stretch out of the layers [20], but are highly sensitivity to adsorbates on the
experimental transfer characteristic, acceptor like interface
1000

100
Mobility (cm2/V.s)

Nref=6e18
α=1.2
α=1.0
10 Nref=4e18
α=1.2
α=1.0
Model from [11]
1
109 1010 1011 1012 1013
Impurity Concentration (cm-2) Fig. 7: Simulated transfer characteristics using acceptor traps at different
energies. Inset: schematic of the trap energy levels with respect of the MoS2
Fig. 5: Mobility versus impurity concentration considering different
bandgap. The transfer characteristics are shifted negatively along the back
parameters implemented using the Unified Philips mobility model. The
gate voltage by changing the gate work function. This shift is to match the
model is based on theoretical calculations for MoS2.
shift in the experimental characeristics.

290

Authorized licensed use limited to: Indian Instt of Engg Science & Tech- SHIBPUR. Downloaded on August 23,2021 at 11:19:21 UTC from IEEE Xplore. Restrictions apply.
imperfections, which can be tackled by optimizing the growth
and device processing conditions.
ACKNOWLEDGMENTS
We acknowledge the support of SFI through the US-Ireland
Partnership Programme Grant No. SFI/13/US/I2862 (UNITE)
and of the IRC through the Postgraduate Scholarship
EPSPG/2015/69. The research was supported in part by the
HEA Programme for Research in Third Level Institutions in
Ireland under Grant Agreement No. HEA PRTLI5.

REFERENCES
[1] F. Schwierz, J. Pezoldta and R. Granznera, “Two-dimensional materials
Fig. 8: Transfer characteristics of the experimental and the simulated devices in and their prospects in transistor electronics”, Nanoscale, 7, 8261-8283
semi-log and linear scales considering an impurity concentration of 4x1013 cm-2, (2010).
0.17eV of contact barriers and interface traps at energy levels at Et1=1.25kT, [2] A. Ayari, W. Cobas, O. Ogundadegbe, and M. S. Fuhrer, “Realization
Et2=2.5kT, Et3=5kT and Et4=10kT from the conduction band and with and electrical characterization of ultrathin crystals of layered transition-
concentration of 8×1011, 3×1011, 2.5×1011 and 1×1011 cm-2, respectively. metal dichalcogenides”, J. Appl. Phys. 101, 014507 (2007).
[3] R. Addou, et al., “Impurities and Electronic Property Variations of
Natural MoS2 Crystal Surfaces”, ACS Nano 9 (9), 9124-9133 (2015).
[4] S. Monaghan, et al., “Hall-effect Mobility for a Selection of Natural
and Synthetic 2D Semiconductor Crystals”, Session 2: More than
Moore devices and applications, Heterogeneous Integration, Other,
EUROSOI-ULIS 2017, 3-5 April 2017, Athens, Greece.
[5] Z. Y. Ong, and M. V. Fischetti, “Mobility enhancement and
temperature dependence in top-gated single-layer MoS2,” Phys. Rev.
B., vol. 88, (2013).
[6] Hai Li, et al., “Rapid and Reliable Thickness Identification of Two-
Dimensional Nanosheets Using Optical Microscopy”, ACS Nano 7
(11), 10344-10353 (2013).
[7] S. McDonnell, R. Addou, C. Buie, R. M. Wallace, and C. L. Hinkle,
“Defect-Dominated Doping and Contact Resistance in MoS2”, ACS
Fig. 9: (a) Field-effect mobility versus number of MoS2 layers considering Nano 8 (3), 2880-2888 (2014).
different impurity concentration. (b) Countour plot of the electron density near [8] S. J. McDonnell, and R. M. Wallace, “Atomically-thin layered films for
the drain contact for 2 layers of MoS2. Both the plots are evaluated at Vds=0.1V device applications based upon 2D TMDC materials”, Thin Solid Films
and Vbg=7.5V. 616, 482–501 (2016).
[9] A. Allain, J. Kang, K. Banerjee, and A. Kis, “Electrical contacts to two-
surface, especially for few layers [21]. Therefore, due to the dimensional semiconductors”, Nature Materials 14, 1195–1205 (2015).
several imperfections of the material itself, it is possible that [10] G. Mirabelli, et al., “Air sensitivity of MoS2, MoSe2, MoTe2, HfS2, and
these trends might differ when testing purer MoS2. Also, the HfSe2”, J. Appl. Phys., 120, 125102 (2016).
experimental values of mobility we reported are lower with [11] N. Ma, and D. Jena,” Charge Scattering and Mobility in Atomically Thin
Semiconductors”, Phys. Rev.X 4 011043 (2014).
respect to what is given by the model.
[12] T. Mori, et al., "Characterization of Effective Mobility and Its
To investigate the field effect mobility further the electron Degradation Mechanism in MoS2 MOSFETs," in IEEE Transactions on
density near the drain contact was considered. Fig. 9b shows Nanotechnology, vol. 15, no. 4, pp. 651-656, (2016).
the electron contour plot around the drain contact for a 2 [13] International Technology Roadmap for Semiconductors (ITRS, 2015);
http://www.itrs.net/
layers MoS2 back gated device. Since it is a back-gated FET a [14] Synopsys Inc., CA, USA, Sentaurus Device User Guide (2015).
change in carrier density from the SiO2/MoS2 interface to the [15] H. Zhong, et al., “Interfacial Properties of Monolayer and Bilayer MoS2
top metal contact exists, resulting in carrier transport Contacts with Metals: Beyond the Energy Band Calculations”, Sci. Rep.,
perpendicular to the MoS2 planes under the contact region. 6, 21786 (2016).
[16] X. Chen, et al., “Probing the electron states and metal-insulator
Thus, the mobility versus thickness trend might be different transition mechanisms in molybdenum disulphide vertical
considering top-gated or back-gated FETs. Also achieving a heterostructures”, Nat. Commun. 6, 6088 (2015).
highly doped MoS2 under the drain and source metal contacts, [17] D. B. M. Klaassen, “A Unified Mobility Model for Device Simulation—
I. Model Equations and Concentration Dependence,” Solid-State
will significantly reduce contact resistance [22]. Electronics, vol. 35, no. 7, pp. 953–959 (1992).
[18] C. Kim, et al., “Fermi Level Pinning at Electrical Metal Contacts of
V. CONCLUSIONS Monolayer Molybdenum Dichalcogenides”, ACS Nano 11 (2), 1588-
In conclusion, MoS2 FETs were electrically characterized 1596 (2017).
[19] W. Cao, J. Kang, W. Liu, and K. Banerjee, “A compact current–voltage
from which continuum-based models and parameter sets were model for 2D semiconductor based field-effect transistors considering
developed. It has been show that the experimentally extracted interface traps, mobility degradation, and inefficient doping effect,”
field-effect mobility values are strongly dependent on the IEEE Trans. Electron Devices, vol. 61, no. 12, pp. 4282–4290, (2014).
impurity concentration present in the material. The trend was [20] M. Lin, et al, “Thickness-dependent charge transport in few-layer MoS2
field-effect transistors”,Nanotechnology 27 165203 (2016).
confirmed by TCAD simulations, which also showed the [21] D. Lembke, A. Allain, A. Kis, “Thickness-dependent mobility in two-
impact of interface trap density and contact barriers on the dimensional MoS2 transistors”,Nanoscale, 7, 6255–6260 (2015).
electrical device behavior. Even if promising results are shown [22] G. Mirabelli, et al., “Back-gated Nb-doped MoS2 junctionless field-
effect-transistors”, AIP Advances 6, 025323 (2016).
to date the true TMD device performance is still masked by

291

Authorized licensed use limited to: Indian Instt of Engg Science & Tech- SHIBPUR. Downloaded on August 23,2021 at 11:19:21 UTC from IEEE Xplore. Restrictions apply.

You might also like