Photoresponsivity enhancement in monolayer MoS$_2$ by rapid O$_2$:Ar plasma treatment

TL;DR

This study demonstrates that rapid O₂:Ar plasma treatment significantly enhances the photoresponsivity and electrical performance of monolayer MoS₂, offering a fast method to improve 2D material-based optoelectronic devices.

Contribution

It introduces a plasma treatment method that increases photoresponsivity and mobility in monolayer MoS₂ by inducing surface MoOₓ formation, which was characterized by TEM, Raman, and PL.

Findings

Up to ten-fold increase in photoresponsivity.

Enhanced carrier mobility after plasma treatment.

Surface oxidation to MoOₓ improves photocurrent generation.

Abstract

We report up to ten-fold enhancement of the photoresponsivity of monolayer MoS$_2$ by treatment with O$_2$:Ar (1:3) plasma. We characterize the surface of plasma-exposed MoS$_2$ by TEM, Raman and PL mapping and discuss the role of MoO$_x$ in improving the photocurrent generation in our devices. At the highest tested laser power of 0.1 mW, we find ten-fold enhancements to both the output current and carrier field-effect mobility under the illumination wavelength of 488 nm. We suggest that the improvement of electrical performance is due to the surface presence of MoO$_x$ resulting from the chemical conversion of MoS$_2$ by the oxygen-containing plasma. Our results highlight the beneficial role of plasma treatment as a fast and convenient way of improving the properties of synthetic 2D MoS$_2$ devices for future consideration in optoelectronics research.

Full text

Keywords: 2D materials, photodetector, oxygen plasma, field-effect transistors.

††preprint: APS/123-QED

Two-dimensional layered transition metal dichalcogenides (TMDs) have attracted wide research interest due to their intriguing physical properties and potential applications. Molybdenum disulfide (MoS2), a typical layered TMD, is a semiconductor with a direct bandgap of $\sim$ 1.8 eV in the single-layer limit Mak et al. (2010). This allows monolayer MoS2 field-effect transistors (FETs) to achieve high ON/OFF ratios Qiu et al. (2012) (107-109), making them attractive candidates for switching components in future electronics. Recently, optoelectronic devices fabricated from MoS2 have received notable attention Lopez-Sanchez et al. (2013); Chen et al. (2015); Qin et al. (2016); Wang et al. (2018). MoS2 phototransistors are easy to fabricate, respond to a wide range of wavelengths Lopez-Sanchez et al. (2013); Wang et al. (2015a), and exhibit fast DC photoresponses Yore et al. (2017); Wang et al. (2015b). In addition, their photoresponsivity can be tuned by various methods, such as back-gating Yin et al. (2011); Lee et al. (2017), encapsulation in HfO2 Kufer and Konstantatos (2015), strain engineering Wang et al. (2017), layer decoupling Yang et al. (2017) and evaporation of sub-stoichiometric molybdenum oxide overlayers Yoo et al. (2017). Surface sensitization of monolayer MoS2 FETs has also yielded significant enhancements of the measured photocurrent in the case of quantum dots Kufer et al. (2015, 2016); Gough et al. (2018), organic molecules Yu et al. (2014a); Kang et al. (2015); Huang et al. (2016) and metal nanostructures Miao et al. (2015); Jing et al. (2017). However, these methods often involve additional preparation steps in order to fabricate the sensitizing species and deposit it on the MoS2 device. Moreover, the surface-deposited dopants may not be robust to mechanical stressing or further material modification without losing their favorable properties.

Plasma functionalization, in turn, presents a fast and facile way to alter the crystal structure of on-chip layered materials such as MoS2. It facilitates large-scale, multi-sample and rapid tuning of the optoelectronic performance of FETs based on layered semiconductors. In particular, oxygen-containing plasmas tend to form sub-stoichiometric molybdenum oxides on the surface of MoS2 Islam et al. (2014); Giannazzo et al. (2017); Ko et al. (2016). These oxide centres can then act as dopants that alter the charge concentration in the modified MoS2 transistor channel Choudhary et al. (2016); Jadwiszczak et al. (2018a, b), and ultimately govern the electron conduction behavior of the newly-formed oxide/MoS2 heterostructure Khondaker and Islam (2016). In this work, we demonstrate the enhancement of the photoresponsivity of chemical vapour deposition (CVD)-grown monolayer MoS2 by O2:Ar (1:3) plasma treatment. The photoresponsivity is improved ten-fold in gated devices after 2 seconds of exposure to the plasma. At the same time, the field-effect mobility of the device under illumination improves by over one order of magnitude. We carry out transmission electron microscopy (TEM) imaging and spectroscopic mapping to characterize the sample after plasma exposure, and attribute the observed photoresponse to the suppressed charge recombination mediated by surface-bound molybdenum oxides.

MoS2 samples were synthesized on SiO2/Si substrates using the CVD method previously reported O’Brien et al. (2014). The flake thickness was confirmed by optical microscopy and Raman spectroscopy. Standard electron beam lithography was carried out to fabricate the FET devices using PMMA resist and development in MIBK:IPA (1:3) solution. This was followed by metallization with Ti(10 nm)/Au(40 nm) contacts and lift-off in acetone. Plasma treatment was carried out in a Fischione Instruments 1020 plasma cleaner for 2 seconds, utilizing O2:Ar (1:3) gas at a chamber pressure of $\sim$ 5 mbar. The electrical testing was performed at room temperature in a two-probe configuration (Imina miBot) using a source meter unit (Agilent B2912A) in the ambient. The devices were back-gated through the heavily p-doped Si substrate underneath the 285 nm SiO2 overlayer. A 488 nm laser was used for irradiation. Its power density was tuned at five different levels and controlled to ensure no power fluctuation throughout the experiment. The laser was directed through a condenser lens (20$\times$, NA = 0.4) and the spot size was $\approx$ 1.5 $\mu$m. TEM was carried out in a FEI Titan 80-300 system operated at 300 kV, at a chamber pressure of $4\times 10^{-7}$ mbar. Monolayer samples were transferred onto copper TEM grids using the polymer stamp transfer method (Bie et al., 2011). Fabricated devices were imaged in a Zeiss Nanofab helium ion microscope at a beam energy of 25 keV. Raman and photoluminescence (PL) spectra were acquired using a WITec Alpha 300R system ($\lambda$ = 532 nm). Raman spectra were acquired using a spectral grating with 1800 lines/mm. PL spectra were collected using a 600 lines/mm grating. A low laser power ($<$ 100 $\mu$W) was used during mapping to minimize any laser-induced damage or heating of the sample.

Figure 1(a) is a false-color helium ion micrograph of a typical contacted device. The contacts in our devices were always deposited in a parallel geometry, as visible in the image. The transistor channel length was over 5 $\mu$m to confine the laser irradiation solely to the MoS2 region. This was done to avoid any heating effects on the metal/semiconductor interface which are known to induce p-type doping in intrinsic n-type TMDs Seo et al. (2018). We collected the output and transfer characteristics of the device under 5 different illumination powers. As the laser spot avoided the Au electrodes during illumination, we assume that all of the measured photo-generated current originated in the MoS2 semiconductor.

Figure 1(b) shows the output characteristics of the device under laser illumination before any plasma treatment. Prior to any exposure to the plasma, the low-bias IV response of the MoS2 FET shows a well-behaved linear increase with applied bias for both voltage polarities; indicating good ohmic contacts to the semiconductor. Upon successive irradiations with rising laser power, the photocurrent increases, which is typical for semiconducting monolayer MoS2 devices Yin et al. (2011); Lopez-Sanchez et al. (2013); Zhang et al. (2013); Yore et al. (2017); Tang et al. (2017); Park et al. (2018). The output current reaches nearly 10 $\mu$A at $\pm$ 5 V at the highest tested laser power of 72 $\mu$W. Figure 1(c) tracks the IV curves after 2 seconds of exposure to the plasma. We see that the current increases to nearly 25 $\mu$A at the highest illumination power, compared with the untreated sample at the same applied drain-source voltage. This indicates that dopants introduced by the plasma treatment to the MoS2 surface mediate an enhanced charge carrier photo-generation response in the device.

Figure 2(a) shows the transfer curves for the same sample before any plasma treatment. Our as-grown devices perform as standard n-type FETs with a field-effect mobility ($\mu$) of 0.13 cm2 V*-1* s*-1* under no illumination, extracted in the linear region of the transfer curve and at Vds = 1 V. Upon successive laser irradiations we observe a photogating effect, whereby the threshold voltage of the transistor shifts to negative gate biases by more than 10 V due to increased electron doping. This has previously been observed in ultrathin TMD FETs and is attributed to the interaction of photo-generated carriers with charge traps in the transistor channel Xie et al. (2017); Garcia et al. (2018). At the highest incident power, the FET channel is effectively still open at Vg = - 60 V, where the output current stays firmly above 10*-7* A and leads to a large reduction in the ON/OFF ratio of our device.

Figure 2(b) presents the gate curves after plasma treatment. The observed level of output current in the dark transfer curve drops two-fold when evaluated at the gate bias, Vg = 60 V. Meanwhile, the threshold voltage is seen to shift to more positive gate biases by $\sim$ 5 V. This shift indicates oxygen-related p-type doping in the material, consistent with previous works on oxygen plasma-treated MoS2 Giannazzo et al. (2017); Guo et al. (2017); Islam et al. (2014); Jadwiszczak et al. (2018b). In addition, the MoS2 now possesses a weak ambipolar response, indicating hole-branch conduction caused by the likely presence of plasma-created oxides Chuang et al. (2014); McDonnell et al. (2014). After 2 seconds of plasma treatment, the output current in the saturation region of the gate curve improves by one order of magnitude under all illumination powers (note scale on the y-axis). Figure 2(c) tracks the MoS2 channel field-effect mobility before and after chemical reaction with the plasma. Even with no laser illumination, the mobility is seen to improve two-fold in the plasma-treated samples, which we have explored in previous work Jadwiszczak et al. (2018a). After 2 seconds of exposure, the carrier mobility increases over ten-fold as the laser power is turned up. We find no clear relationship between the mobility and the laser power for the untreated sample. However, we obtain a good power law fit to the mobility scaling as $\mu\propto P^{0.64}$ above laser powers of 10*-3* mW. Similarly, in Fig. 2(d), the output current at Vg = 60 V is seen to improve once the device is exposed to the plasma. In both the untreated and treated case, the dependence of the photocurrent on the laser power is sublinear, though the power law response is enhanced by plasma treatment from $\mu\propto P^{0.13}$ to $\mu\propto P^{0.29}$. The scaling exponent in this relationship depends on the charge trapping rate in the MoS2 FET channel Zhang et al. (2013); Massicotte et al. (2016). Our results suggest that the presence of plasma-created oxides on the surface inhibits photo-generated pair recombination via defect sites. We extract both data sets at Vg = 60 V where the FET is moving into depletion, i.e.: the majority carrier concentration in MoS2 induced by gating begins to approach that of the photogenerated carrier density Wu et al. (2013). The slope of the fit to the photocurrent as a function of laser power serves as a measure of the photogating effect seen in the power-graded transfer curves in Figs. 2(a),(b). An increase in the slope after plasma treatment is a direct consequence of the additional charge present in the device.

We plot the DC photoresponsivity, $R_{ph}$, at different gate biases as a function of irradiation power in Figure 3. $R_{ph}$ is the current generated in the device per unit of laser power and is a crucial parameter that quantifies the sensitivity of photodetectors Xie et al. (2017). We obtain good linear fits of $R_{ph}$ as a function of power, $P$, across the whole gate bias range, before and after plasma treatment. The negative slope in the log-log plot indicates the saturation of trap states in the material with increasing incident optical power Lopez-Sanchez et al. (2013); Yu et al. (2014b); Wang et al. (2015a). In Fig. 3(a), we see the 0 V and 60 V gate bias trends exhibiting similar levels of $R_{ph}$, especially at higher laser powers. Upon plasma treatment, in Fig. 3(b), we observe an enhancement of $R_{ph}$ for all tested gate biases and a notable separation of the responsivity as a function of Vg. As Vg is increased, the device becomes more responsive to laser illumination. The temporal response of the device pre- and post-plasma treatment is charted in Fig. 3(c). The photocurrent is seen to improve two-fold for the tested device when the laser irradiation is modulated through 5 s on/off cycles at a power of 36 $\mu$W and Vds = 5 V. The post-sensitization fall ($\tau_{fall}$) and rise ($\tau_{rise}$) times are extracted from single exponential fits in Figs. 3(d) and 3(e) respectively. The time-resolved photoresponses compare favorably with the evaporated MoOx overlayer report Yoo et al. (2017), where our rise time at a much lower irradiation power is 35% shorter.

Figures 5(a), (b) present TEM images of MoS2 flakes before and after 2 s of plasma treatment. Corresponding Fast Fourier Transforms (FFTs) are included as insets. A large change in local contrast on some flake areas can be noted after 2 s of exposure to the plasma. It is evident from the TEM observation that after the plasma treatment, an amorphous oxide of molybdenum builds up on the surface as a consequence of a chemical reaction between the flake and the plasma-created species Ko et al. (2016); Nan et al. (2017); Jadwiszczak et al. (2018a). Spectroscopic mapping of the samples allows for a closer inspection of the chemical state of the MoS2 surface pre- and post-plasma treatment. Figures 6(a), (b) show the spatially-resolved Raman maps of the material corresponding to the in-plane vibrational mode at 385 cm*-1*. We notice a drastic drop in the intensity of the signal at this frequency, indicating a change in the MoS2 lattice which alters the Raman-active modes in the sample. The flake-averaged spectra are presented in Fig. 6(c), demonstrating the quenching effect of plasma treatment on the monolayer MoS2 Raman peaks.

From the spectral component fits (see Supplementary Table 1), the monolayer nature of the sample is confirmed with a wavenumber separation of 20.5 cm*-1* between the A${}^{{}^{\prime}}_{1}$ and E${}^{{}^{\prime}}$ peaks Li et al. (2012). Upon plasma treatment, the intensity of both Raman modes is severely reduced after 2 seconds of exposure, while peak position also shifts and the full-width-at-half-maximum (FWHM) increases. Both the downshift of the E${}^{{}^{\prime}}$ peak and upshift of the A${}^{{}^{\prime}}_{1}$ peak are consistent with reports on molybdenum oxide formation on MoS2, as is the asymmetric broadening of both peaks Choudhary et al. (2016); Ko et al. (2016); Jadwiszczak et al. (2018a). PL maps of the neutral A exciton emission (1.84 eV) of the same flake are presented before and after 2 seconds of plasma treatment in Figs. 6(d), (e). Accompanying spectra averaged across the whole sample are shown in Fig. 6(f). We observe significant quenching of direct excitonic recombination in the sample after the plasma introduces the oxide species on the surface. The emission is also largely blue-shifted to higher energies by $\sim$ 0.1 eV. These observations are also in line with previous studies of oxidized MoS2, where the emission intensity is reduced due to the presence of sub-stoichiometric oxides on the surface Jadwiszczak et al. (2018a).

We suggest that the observed photoresponsivity improvement results from carrier trapping at the MoS2/MoOx interface Yoo et al. (2017). The electron affinity and bandgap of monolayer MoS2 are $\sim$ 4.3 eV and 1.8 eV respectively Liang et al. (2013); Mak et al. (2010). After the rapid plasma treatment, MoOx is generated on the device surface as demonstrated in the previous discussion. Oxides of molybdenum are commonly known as high work function materials (6.8 eV) with a bandgap of 3 eV Kang et al. (2014); Yoo et al. (2017). In this device, plasma-generated oxides and unreacted MoS2 will form an effective medium that spans the FET channel. As the Fermi level of MoS2 is higher than that of MoOx, significant band bending will occur at the interface Yoo et al. (2017). The built-in electric field gradient will be directed from MoS2 towards the oxide. Photo-generated holes will then become trapped at the material interface, inhibiting recombination and thereby enhancing the photocurrent with electrons as majority carriers. This is also supported by the observed photogating effect mediated by the electron-rich surface overlayer. The improved responsivity at higher back-gate fields is a direct consequence of Fermi level alignment which also facilitates easier photocarrier injection into the contacts Yin et al. (2011); Kufer and Konstantatos (2015); Chen et al. (2015) and primes the device for photon detection levels exceeding those of the pristine MoS2.

In conclusion, we have demonstrated that the photoresponsivity of MoS2 monolayer FETs can be enhanced ten-fold by the introduction of surface-bound molybdenum oxides. We confirm their presence via TEM, Raman and PL spectroscopy. The effect of the mobility and photoresponsivity enhancement depends on laser power and is more prominent at powers exceeding several $\mu$Watts. Our work provides insight into heterostructure physics in novel 2D optoelectronic nano-devices.

This research was supported in part by the Science Foundation Ireland (Grant nos: 11/PI/1105, 12/TIDA/I2433, 07/SK/I1220a and 08/CE/I1432) and the Irish Research Council (Grant nos: GOIPG/2014/972 and EPSPG/2011/239).