**Environmental Effects on the Electrical Characteristics of Back-Gated WSe2 Field-Effect Transistors**

**Francesca Urban 1,2, Nadia Martucciello 2, Lisanne Peters 3, Niall McEvoy <sup>3</sup> and Antonio Di Bartolomeo 1,2,\***


Received: 4 October 2018; Accepted: 1 November 2018; Published: 3 November 2018

**Abstract:** We study the effect of polymer coating, pressure, temperature, and light on the electrical characteristics of monolayer WSe2 back-gated transistors with Ni/Au contacts. Our investigation shows that the removal of a layer of poly(methyl methacrylate) (PMMA) or a decrease of the pressure change the device conductivity from p- to n-type. From the temperature behavior of the transistor transfer characteristics, a gate-tunable Schottky barrier at the contacts is demonstrated and a barrier height of ∼70 meV in the flat-band condition is measured. We also report and discuss a temperature-driven change in the mobility and the subthreshold swing that is used to estimate the trap density at the WSe2/SiO2 interface. Finally, from studying the spectral photoresponse of the WSe2, it is proven that the device can be used as a photodetector with a responsivity of <sup>∼</sup> 0.5 AW−<sup>1</sup> at 700 nm and 0.37 mW/cm2 optical power.

**Keywords:** 2D materials; field effect transistors; PMMA; tungsten diselenide

#### **1. Introduction**

The continuous downscaling of the channel length and thickness in modern field effect transistors (FETs) has increased the need for atomically-layered materials to minimize short channel effects at extreme scaling limits [1,2]. Layered transition metal dichalcogenides (TMDs), owing to their two-dimensional structure, reasonable charge-carrier mobilities, and the absence of dangling bonds can enable extreme channel length scaling and have recently emerged as promising materials for future electronic and optoelectronic devices [3–7]. These graphene-like materials offer the advantages of sizeable and non-zero bandgap, high on/off ratio and quasi-ideal subthreshold swing, mechanical flexibility, and thermal and chemical stability. Similar to graphene, their electronic transport properties are strongly influenced by the choice of the metal contacts [8–10], by interface traps and impurities [11,12], as well as by structural defects and environmental exposure [13–17]. These effects need to be understood and controlled for technological applications.

Molybdenum disulfide (MoS2) has been one of the most heavily investigated systems from the TMD family to date [18–25]. Similar to MoS2, tungsten diselenide (WSe2), whose electrical and optical properties have been relatively less explored [26], is characterized by an indirect bandgap (1.0–1.2 eV) in the bulk form and shows a transition to a direct gap of 1.6 eV when it is thinned to monolayer [27]. Recent reports on WSe2 FETs have demonstrated a relatively high field-effect mobility controllable by temperature and bias voltage [26,28], an ideal subthreshold swing ∼ 60 mV/dec [29] and an *on/off* ratio up to 108. The ambipolar behavior, controllable using different metal contacts, like In or Pd [8], which favor electron and hole injection, respectively [26,30,31], makes mono- and few-layer WSe2

an interesting material for complementary logic applications; indeed, a stable WSe2-based CMOS technology has been demonstrated [32–34].

A great challenge for electronic integration of WSe2 is the development of low-resistance ohmic contacts, a task often complicated by the appearance of Schottky barriers due to the occurrence of Fermi level pinning [35,36]. Accordingly, several studies have aimed to clarify the role of the contacts, focusing on the carrier transport at the WSe2/metal interface [30,37,38].

In this paper, we study back-gated monolayer WSe2 devices with Ni contacts, measuring their electrical characteristics under different conditions, considering, for instance, the effect of a poly(methyl methacrylate) (PMMA) coating layer, and the dependence on the chamber pressure and the sample temperature. Similar to graphene [39,40], we observe that PMMA strongly influences the electrical transport, in this case to the extent that the polarity of the device changes from p-type to n-type conduction when the PMMA layer is removed. We demonstrate that lowering the pressure on air-exposed WSe2 FETs affects their characteristics in a similar way to PMMA, turning the conduction from p- to n-type. Furthermore, from the current-voltage (I-V) characteristics measured at different temperatures, we prove a gate modulation of the Schottky barrier (SB) at the contacts.

In addition, we study the temperature dependence of the carrier mobility and the subthreshold swing and show that both undergo a change of behavior with increasing temperature. From the subthreshold swing data, we derive the interface trap density, which affects the photoresponse of the device. The monolayer *WSe*<sup>2</sup> device, characterized at several laser wavelengths, achieves a responsivity as high as ∼ 0.5 *AW*−<sup>1</sup> at 700 nm, i.e. at a photon energy close to the WSe2 bandgap.

#### **2. Experimental**

The WSe2 flakes were grown in a two-zone heating furnace. Selenium pellets (Sigma-Aldrich Inc, St. Louis, MO, USA) were evaporated at 250 ◦C in the lower-temperature, upstream heating zone, while in the high-temperature, downstream zone the tungsten precursor (20 nm sputtered and subsequently oxidized tungsten) was placed with the growth substrate. A highly p-doped Si (silicon) substrate covered by 300 nm of SiO2 (silicon dioxide) was placed top down on the tungsten precursor, which was heated up to 850 ◦C. The tungsten precursor/growth substrate stack forms a microreactor, which increases the reactivity due to the close proximity between the precursor and growth substrate, requiring a lower amount of precursor and minimizing the contamination of the furnace. This is a similar approach to previous reports on the growth of MoS2 in a microreactor [41] but in this case different chalcogen and transition metal precursors are used. Both furnaces were kept at the reaction temperatures for 40 min under a flow of 50 sccm forming gas (*H*2/Ar 1:9) at a pressure of 6 Torr, after which the furnace was cooled down.

A schematic of the back-gated FET device and a scanning electron microscope top-view of a WSe2 monolayer with evaporated Ni/Au (5/50 *nm*) contacts, made by use of e-beam lithography, are shown in Figure 1a,b. In the following, the transistor characterization refers to contact 1 and 2, which define a device with channel length L ∼ 2 μm and mean width W ∼ 22 μm (Figure 1b). The electrical analyses are performed using a Keithley 4200 SCS (semiconductor characterization system, Tektronix Inc., Beaverton, OR, USA) connected with a Janis ST-500 probe station (Janis Research Company LLC, Woburn, MA, USA), equipped with four probes used for the electrical connection to the drain and source Ni/Au terminals and to the Si back-gate of the device.

Raman and photoluminescence (PL) spectra were acquired using a WITec Alpha 300 tool (WITec GmbH, Ulm, Germany) with a 532 nm excitation laser. The Raman spectrum of the WSe2, displayed in Figure 1c, exhibits two peaks around ∼ <sup>250</sup> cm−<sup>1</sup> and ∼ <sup>260</sup> cm<sup>−</sup>1, corresponding to an overlapping contribution from the in-plane vibrations of W and Se atoms (E<sup>1</sup> 2g) and out-of-plane vibrations of Se atoms (A1g), and to a second-order resonant Raman mode (2 *LA* (*M*)) due to LA phonons at the M point in the Brillouin zone [42,43], respectively. The peak frequency positions are typical of a WSe2 monolayer of thickness *d* ∼ 0.7 nm [29]. The monolayer structure of the flake is further confirmed by the PL spectrum of Figure 1d, which shows an intense and narrow peak with maximum at ∼ 778 nm

and FWHM of ∼ 21 *nm*. Such a peak corresponds to a bandgap of ∼ 1.59 eV, a value closer to that of a monolayer than to that of a bilayer. Hence, both Raman and PL spectra indicate that the flake is a monolayer.

**Figure 1.** Schematic diagram (**a**) and optical microscope image (**b**) of the *WSe*<sup>2</sup> back gate FET transistor. Raman (**c**) and photoluminescence (**d**) spectra of the WSe2 flake.

#### **3. Results and Discussion**

We start the transistor characterization by comparing the device I-V curves with and without a PMMA coating layer, which was used to protect the transistor channel from residue and adsorbates [44,45]. It has been observed that a PMMA film, or even only residue of it, can cause p-type doping of graphene and other 2D channels due to the presence of oxygen. Here, we report a similar effect for CVD-grown WSe2 FETs, measured at *T* = 293 K, and *P* ∼ 2 mbar.

The PMMA-covered devices behave like p-type transistors, as can be seen from the *Ids* − *Vgs* transfer curves of Figure 2a which show high channel current *Ids* (*on*-state of the FET) at negative gate voltages, *Vgs*. The p-type conduction is explained considering the charge transfer to oxygen which, acting as electron capture center, suppresses the free electron density and enhances the hole concentration in the channel. Furthermore, the transition of the channel to p-type could cause a depinning of the Fermi level and facilitate hole injection at the contacts (indeed, in TMDs, Fermi level pinning often occurs close to the minimum of the conduction band) [36,46–48]. After the removal of the PMMA by immersion in acetone, a dramatic change to n-type behavior appeared, with the *on*-state at *Vgs* < 0 V, as shown in Figure 2b. A similar effect has been reported in literature [32,49] for exfoliated *WSe*<sup>2</sup> flakes on an SiO2/Si substrate covered by F4PCNQ-doped PMMA.

**Figure 2.** Transfer characteristics (*Ids* − *Vgs* curves) obtained at a drain voltage bias *Vds* = 2 *V* for the device covered with PMMA (**a**) and after the removal of PMMA (**b**) at *Vds* = 5 *V*. The inset shows a complete cycle with the gate voltage *Vgs* swept forward and backward. Output characteristics (*Ids* − *Vds* curves) at different gate voltages for the device with (**c**) and without (**d**) PMMA. For the uncovered device, the drain bias was increased from *Vds* = 2 V to *Vds* = 5 V to better characterize the above-threshold region.

The corresponding *Ids* − *Vds* output characteristics are reported in Figure 2c,d. Both plots show non-linear behavior for the device in the *on* state with increasing positive-negative asymmetry when the device approaches the *off* state. This points to the presence of Schottky barriers at the Ni/WSe2 contacts, possibly with slightly different heights [37,50].

For increasing *Vgs*, the channel current at constant *Vds* shows an exponential dependence (below the threshold region) followed by a linear or power law behavior (above the threshold region).

A quadratic behavior is particularly evident in the transfer characteristic of Figure 2a, even though the transistor is operated in the triode region. The parabolic dependence of *Ids* on *Vgs* can be ascribed to the linear gate-voltage dependence of the mobility *μ* [51,52], which defines the drain current as:

$$I\_{ds} = \frac{\mathcal{W}\mathcal{C}\_{\text{ox}}}{L} \left(V\_{\text{\%}} - V\_{th}\right) V\_{ds} \tag{1}$$

with:

$$
\mu = \mu\_B \left( V\_{\mathbb{S}^\mathfrak{s}} - V\_{th} \right) \tag{2}
$$

in which *μ<sup>B</sup>* represents the mobility per unit gate voltage and *Vth* is the threshold voltage. The *Vgs*-dependent mobility can be explained by considering that the increasing carrier density becomes more effective at screening Coulomb scattering or in filling trap states at higher *Vgs*, thus resulting in enhanced mobility.

The dependence of the mobility on the gate voltage can be established by extracting it in the usual way using:

$$
\mu = \frac{L}{\mathcal{W}} \frac{1}{\mathcal{C}\_{ox}} \frac{1}{V\_{ds}} \frac{dI\_{ds}}{dV\_{\mathcal{S}^s}} \tag{3}
$$

Figure 3a,b show the *μ* − *Vgs* curves on logarithmic and linear (insets) scales obtained from Equation (3) and the data of Figure 2, for the devices with and without PMMA, respectively. These confirm a linear dependence of *μ* on *Vgs* over a certain range. Remarkably, for the device with removed PMMA, the mobility shows the typical decrease observed in common FETs due to increased scattering suffered by carriers attracted at the channel/dielectric interface at higher gate voltages.

**Figure 3.** Mobility versus gate voltage on a logarithmic scale for the *WSe*<sup>2</sup> flake covered (**a**) and uncovered (**b**) by PMMA. The inset graphs show the mobility on a linear scale.

By neglecting the *Vgs* dependence of the mobility, as is usually done in the literature, *μ* can be obtained by fitting a straight line to the transfer characteristics, as shown in Figure 2a,b.

By this method, we estimate an electron mobility of ∼ 0.04 *cm*2*V*−1*s*−<sup>1</sup> for the n-type transistor without PMMA, consistent with other works with WSe2 on SiO2 [41], and a hole mobility of <sup>∼</sup> 0.1 cm2V−1s−<sup>1</sup> for the PMMA-covered p-type transistor. We notice that, although a different channel carrier concentration might contribute to this difference, these values are consistent with the higher hole mobility in WSe2 reported elsewhere [29,32,45]. The low mobility indicates a high density of trap states, which are also responsible for the hysteretic behavior of the transfer characteristic shown in the inset of Figure 2b. The hysteresis is caused by trapping and detrapping of charge carriers, whose potential adds to that of the back-gate [51–54].

The subthreshold swing, *SS* = *dVgs*/*dlog*(*Ids*), is 4 *V*/*decade* and 1.5 *V*/*decade*, for the p-type and n-type transistor, respectively. The different *SS* results from a different trap density at the WSe2/dielectric interface, implying a higher trap density when the WSe2 channel is covered by PMMA, which acts as a second interface [55,56].

After the removal of the polymeric film and exposure of the device to air for a few days, we observed a restoration of a prevailing p-type behavior due to O2 and water adsorption on the WSe2 surface and possible depinning of the Fermi level. We then studied the effect of dynamic pressure by increasing the vacuum level of the probe station chamber from atmosphere (∼ <sup>1</sup> *bar*) to ∼ <sup>10</sup>−<sup>5</sup> mbar. As reported in Figure 4a, the transistor transfer characteristic changed again from p- to n-type with a gradual decrease of the subthreshold swing and an increase of the *on/off* ratio, as shown in Figure 4b. We attribute the polarity change to the desorption of adsorbed O2 and H2O and to the consequent possible pinning of the Fermi level close to the minimum of the conduction band.

**Figure 4.** (**a**) Transfer characteristics at different pressures from atmospheric value (blue curve) to <sup>∼</sup> <sup>10</sup>−<sup>5</sup> mbar (light green curve) (**b**) *On/Off* ratio (full marks, left scale) and subthreshold swing (empty marks, right scale) as a function of the chamber pressure.

We then examined the temperature (*T*) dependence of the transfer characteristics of the PMMA-free, n-type transistor at low pressure, which can be conveniently used to investigate the Schottky barrier for electrons at the contacts.

We extract the Schottky barrier at the flat-band condition from a plot of the Schottky barrier height as a function of *Vgs* for the device at a source-drain bias of 5 V (Figure 5). Given that the device is n-type, such a barrier refers to electron injection from the contacts and it is caused by the aforementioned pinning of the Fermi level close to the minimum of the conduction band. Measuring the *Ids* − *Vgs* characteristics of the device at several temperatures (Figure 5a) and extracting *Ids* − *T* datasets at given gate voltages (examples are marked by the vertical lines in Figure 5a), we constructed the Arrhenius plot of Figure 5b, showing the *ln Ids*/*T*3/2 <sup>−</sup> <sup>1</sup> *<sup>T</sup>* curves at a representative subset of *Vgs* values. We assumed that the contacts behave as two back-to-back Schottky junctions, where the current is controlled by the reverse-biased junction and is written as:

$$I\_{ds} \sim T^{3/2} \exp\left(-\frac{\Phi\_B}{kT}\right) \tag{4}$$

where *k* is the Boltzmann constant, *T* is the absolute temperature, and Φ*<sup>B</sup>* is the Schottky barrier height [30,57,58]. According to equation (4), a linear fit of ln *Ids*/*T*3/2 vs 1/*T* for each *Vgs* dataset in Figure 5b yields a Schottky barrier Φ*B*. The so-obtained Φ*<sup>B</sup>* − *Vgs* relationship is displayed in Figure 5c and can be divided into three zones, each one corresponding to a different transport regime, consistent with the behavior of the transfer characteristics of Figure 5a.

**Figure 5.** *Cont.*

**Figure 5.** (**a**) Transfer characteristic at several temperatures. (**b**) Arrhenius plot of the current at different temperatures corresponding to a subset of the gate voltages (two of these *Vgs* values are represented by the vertical lines in (**a**)). (**c**) Apparent Schottky barrier as a function of the gate voltage; the insets show the band alignment and the transport regimes at the Ni/WSe2 reverse-biased contact. (**d**) Threshold voltage *Vth* as a function of the temperature; the inset shows, as an example, the parabolic fit of the *Ids* − *Vgs* curve at *T* = 273 *K*.

At low gate voltage the device is set in the *off* state and the transport is due to the thermal excitation of electrons over the barrier. The WSe2 conduction-band level is gradually lowered by the increasing gate voltage, as sketched in the insets of Figure 5c. This results in a lowering of the barrier with a subsequent steep exponential rise of the current in the transfer characteristic (with 60 *mV*/*decade* slope in the ideal case). When the gate voltage is further increased the device reaches the flat band condition (*Vgs* = *VFB*), which sometimes appears in the subthreshold part of the transfer characteristics as a sudden change of slope; for *Vgs* > *VFB* the device enters the so-called Schottky regime which includes part of the downward bended region of *Ids* − *Vds* curves and is characterized by thermionic emission and field emission. Finally, at higher *Vgs*, tunneling through the thinned Ni/WSe2 barrier becomes the dominant conduction mechanism and the device reaches the above threshold region with a linear, or power-law, *Ids* − *Vgs* dependence.

The gate voltage that corresponds to *VFB* is identified by the change of slope in the Φ*<sup>B</sup>* − *Vgs* plot at lower *Vgs*. The Φ*<sup>B</sup>* corresponding to *Vgs* = *VFB* is the so-called Schottky barrier height at the flat-band (or simply Schottky barrier). From the plot in Figure 5c, its value is ∼ 70 *meV*, confirming the presence of a barrier at the Ni-WSe2 contacts, inferred from the asymmetric output characteristics of Figure 2.

Figure 5d shows the temperature-dependent behavior of the threshold voltage *Vth*, which has been extracted assuming a quadratic *Ids* − *Vgs* law as expressed by Equation (1) and (2). The decrease in *Vth* is easily explained by considering that the increasing temperature accelerates the transition from the Schottky to the power-law (above threshold) regime; furthermore, the plot seems to indicate a change of slope above room temperature.

Figure 6a reports the temperature-dependent behavior of the mobility, *μ*, at *Vgs* = 10 *V* obtained from the quadratic fit of the *Ids* − *Vgs* curves.

**Figure 6.** Temperature dependence of (**a**) mobility per unit voltage *μ<sup>B</sup>* and (**b**) subthreshold swing.

The mobility increases for *T* < 250 K and decreases for *T* > 250 K, behavior typical of semiconductor materials, indicating that charged-impurity Coulomb scattering dominates at lower temperatures, while phonon scattering becomes the conduction-limiting mechanism at high temperature [44].

The subthreshold swing has a dependence on temperature that can be simplified with the following expression:

$$SS = n\frac{kT}{q}ln10\tag{5}$$

where *n* is the body factor which is related to the interface trap (*Cit*), *SiO*<sup>2</sup> (*CSiO*2) and channel depletion layer (*Cdl*) capacitances by:

$$m = 1 + \frac{\mathbb{C}\_{it} + \mathbb{C}\_{dl}}{\mathbb{C}\_{ox}} \tag{6}$$

Figure 6b confirms the linear *SS* − *T* dependence (Equation 5) but shows an unexpected rise above room temperature. The deviation from Equation (5) behavior at high temperature is a consequence of the low Schottky barrier which becomes less effective above room temperature (*kT* = 26 meV), resulting in an increase of the subthreshold current leakage.

Assuming that the WSe2 monolayer channel is fully depleted, i.e. that *Cdl* ≈ 0, from the fit of the experimental data with Equation (5), we obtain a *<sup>n</sup>* <sup>≈</sup> 48 and an interface trap density *Nit* <sup>=</sup> *Cit <sup>q</sup>*<sup>2</sup> ≈ 1.3 × <sup>10</sup><sup>13</sup> *eV*−1*cm*<sup>−</sup>2, which is consistent with previous results reported in the literature [59].

The presence of such a density of trap states explains the observed hysteretic behavior of the transfer characteristic, displayed in the inset of Figure 2b [52]. It also affects the electrical response of the device under illumination.

We performed photocurrent measurements with light at different wavelengths, selected by filtering a supercontinuous laser source (NKT Photonics, Superk Compact, wavelength ranging from 450 nm to 2400 nm, total output power of 110 mW) using pass-band filters with 50 nm bandwidth. Figure 7a shows the photoresponse of the WSe2 FET to laser light pulses of 30 s for five different wavelengths.

**Figure 7.** (**a**) Drain-source current measured under 30 s laser pulses at different wavelengths (*Vds* = + <sup>5</sup> *<sup>V</sup>*,*Vgs* <sup>=</sup> <sup>0</sup> *<sup>V</sup>*, *<sup>P</sup>* <sup>∼</sup> <sup>10</sup>−<sup>4</sup> *mbar*). (**b**) Photocurrent generated by a 30 *<sup>s</sup>* laser pulse at the wavelength of <sup>∼</sup> <sup>700</sup> *nm* and optical power <sup>∼</sup> 0.37 *mW*/*cm*<sup>2</sup> with exponential fits.

The photocurrent exhibits a higher peak at the wavelength of 700 nm (photon energy 1.7 eV), which is slightly above the bandgap of a WSe2 monolayer, supporting the Raman and the PL spectroscopy assignment of the single-layer nature of the WSe2 channel.

Figure 7b reports the photocurrent, *Iph* = *Ilight* − *Idark*, obtained in response to a laser pulse of 30 *s* at a wavelength of ∼ <sup>700</sup> *nm*, and an optical power ∼ 0.37 *mW*/*cm*2. It corresponds to a peak with rising time *τ*<sup>0</sup> ∼ 9 *s* and a double exponential decay with times *τ*<sup>1</sup> ∼ 2 *s* and *τ*<sup>2</sup> ∼ 36 *s*, indicating the presence of faster and slower traps [60]. Such features are consistent with a photoresponse decay longer than 5 s for quasi-ohmic contacts measured on similar WSe2 FETs [61,62]. Indeed, we notice that the contact type can play an important role in the response time of WSe2 phototransistors and that reduced times have been reported for Schottky contacts [61,62].

Furthermore, we estimate a photoresponsivity (Figure 7b):

$$R = \frac{I\_{\rm ph}}{\mathcal{W}\_{\rm opt}} \approx 0.5 \frac{A}{\mathcal{W}} \tag{7}$$

where *Wopt* is the incident power. This is in good agreement with the previously reported value of 0.6 *A*/*W* obtained at 750 nm [62]. Such a responsivity is competitive with solid state devices on the market and, despite the ultrathin nature of the absorber, confirms the excellent photoresponse of monolayer WSe2 due to its direct bandgap [63,64].

#### **4. Conclusions**

We showed that different environmental conditions can have dramatic effects on the electrical properties of back-gated transistors with monolayer WSe2 channels. In particular, we demonstrated that the removal of a polymer coating layer, as well as of oxygen and water adsorbates, can change the conduction from p- to n-type. From I-V characterization at different temperatures, we extracted the Ni/WSe2 Schottky barrier height, which we studied as a function of the back-gate voltage. We reported and discussed a change in the temperature behavior of the mobility and the subthreshold swing. Finally, we studied the photoresponse of the device to selected laser wavelengths achieving a responsivity competitive with solid-state devices on the market.

**Author Contributions:** F.U. and A.D.B. wrote the paper; A.D.B. and F.U. conceived and designed the experiments; F.U. and L.P. performed the experiments; F.U., N.M.E., and A.D.B. analyzed the data; and N.M., A.D.B., and N.M.E. contributed reagents/materials/analysis tools.

**Funding:** This research was funded by Regione Campania grant "POR Campania FSE 2014–2020, Asse III Ob. specifico l4, D.D. n. 80 del 31/05/2016", by CNR-SPIN grant "SEED Project 2017" and by Science Foundation Ireland grant number: 15/SIRG/3329.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
