1. Introduction

Atomically thin two-dimensional (2D) layered semiconductors are exciting and emerging materials due to its unique electronic and optical properties compared to its bulk form [1, 2]. At present, the development of optoelectronics components such as photodetectors based on 2D layered semiconductors has been reported to varying success [3]. Although graphene-based photodetectors present several advantages such as absorption over a very wide energy spectrum, ultrafast carrier dynamics, and high carrier mobilities exceeding 10000 cm²V⁻¹s⁻¹, the bandgap of these photodetectors are not easily tunable. Furthermore, the reported photoresponsivities are generally very low (~mA W⁻¹) due to the fast recombination rates. On the contrary, 2D transition metal dichalcogenides such as molybdenum disulfide (MoS₂) shows a thickness dependent bandgap tunability property ranging from its bulk indirect bandgap of 1.3 eV up to a direct bandgap of 1.8 eV in its monolayer form. Moreover, the inherently high absorption coefficients of exceeding 5x10⁷ m⁻¹ in the visible range makes it a promising material for improving optical detection performance. By leveraging on these superior properties, 2D photodetectors utilizing molybdenum disulfide were demonstrated with improved photodetection performance [4–9] at specific wavelength ranges.

However, one key drawback in most of these reports lies in the fact that the MoS₂ films are produced by scotch-tape based mechanical exfoliation method which is not scalable. Furthermore, the poor uniformity and poor repeatability in obtaining the desired film thickness and geometry demands alternative fabrication methods. In this retrospect, chemical vapor deposition (CVD) methods have been pursued to produce atomically thin MoS₂ layered films on insulating substrates [10–13], with the lateral sizes approaching centimeter scale. Nonetheless, the additional substrate modification steps required prior to the CVD process may make these approaches less attractive for future commercialization. Alternative large-scale synthesis approaches that will allow for a precise control of the MoS₂ film thickness and easy integration into various nanoelectronic and optoelectronic device applications will be an enabling factor for future adoption.

In this work, we demonstrate the integration of large-scale MoS₂ crystals for photodetector applications over a wide spectral range from the visible to the near-infrared. The optimized magnetron sputtering synthesis facilitates a precise control of the number of MoS₂ layers, allowing for a systematic thickness dependent study on the photoresponse performance. Using a gold-free CMOS compatible fabrication process, the 2D MoS₂ photodetectors demonstrated in this work outperforms previous MoS₂ photodetectors based on exfoliation and CVD approaches, as well as graphene-based photodetectors.

2. Experiments and results

2.1 Large-scale synthesis of MoS₂ films by DC magnetron sputtering

The magnetron sputtering synthesis approach [14] involves the growth of atomically thin MoS₂ films by DC magnetron sputtering of Molybdenum target in a vaporized sulfur ambient, using a substrate temperature of 700°C, Sulfur partial pressure of 4.0x10⁻⁷ millibars, Argon pressure of 6.0x10⁻⁴ millibars and a low sputtering power of 6W. Using this low sputtering power, the growth rate can be controlled to a high precision. The sputtered molybdenum atoms react with the vaporized sulfur atoms before landing onto the heated HfO₂/Si substrate to form MoS₂ layers. Although the PVD chamber is currently limited to 1x2 cm² sample size, this deposition technique shows promising scalability to wafer level in meeting the industrial requirements for optical communication applications.

2.2 MoS₂ film uniformity and thickness dependent strain

The uniformity of the MoS₂ films is confirmed using scanning electron microscopy (FEI, NOVA NanoSEM 230) and Raman spectroscopy (WITEC, alpha 300 R), which adopts the confocal Raman imaging approach with a laser wavelength of 532 nm, a spot size of 300 nm, and a spectral resolution as low as 0.02 cm⁻¹. An integration time of 2 seconds is adopted, with an accumulation of ten times to improve the signal to noise ratio.

Figure 1(a) shows the achievement of excellent thickness uniformity for the large scale MoS₂ films down to 2 monolayers (ML), as evident from the matching Raman peaks across three independent locations along the film. The main Raman peak positions corresponding to the in-plane E¹_2g phonon mode and out-of-plane A_1g phonon mode are shown to be consistent with the exfoliated MoS₂ films. Furthermore, Fig. 1(b) shows a precise control of MoS₂ film thicknesses, with an average monolayer thickness of 0.75 nm, and a low deviation of only 5%.

 

Fig. 1 (a) Excellent film uniformity for large scale MoS₂ films as thin as two monolayers, as evident from the matching Raman peaks, (b) Precise control of the number of MoS₂ layers with low thickness deviation (~5%). Inset shows the AFM step height for the 2 monolayers MoS₂ film.

Download Full Size | PDF

Figure 2(a) shows a typical X-ray photoelectron spectroscopy (XPS) measurement for the large scale MoS₂ films on the HfO₂/Si substrate and its deconvolution into their respective Sulphur and Molybdenum core-levels. A thickness dependent analysis of the XPS measurements in Fig. 2(b) showed a clear increase in the S 2s peak binding energy position and a corresponding drop in the linewidth (full width at half-maximum) with increasing number of MoS₂ layers, indicating the presence of thickness dependent film strain, and its relaxation with thicker layers [15]. It is worthy of highlight that the S 2s peak binding energy increases from 225.2 eV at 1 ML up to 226.5 eV for a bulk sample and a corresponding reduction in linewidth from 3.5 eV to 1.2 eV. Hence, the chemical bonding configurations for the atomically thin MoS₂ layers can be significantly different from the bulk case. This can also be observed from the shift of the Raman peaks corresponding to the A_1g and E¹_2g phonon modes in Fig. 2(c). In line with the report by Yang et al. [16] on the study of strain on MoS₂ Raman spectra, the PVD-deposited MoS₂ films in this work are also expected to be under in-plane uniaxial tensile strain as well, evident from the upshift in the A_1g Raman peaks to higher wavenumbers with increasing number of layers. This is coupled with a red shift of its bandgap evident from the extended photoresponsivity of the PVD-deposited MoS₂ photodetectors, as will be shown in the subsequent sections. Strain relaxation eventually happens near the front surface, as evident from the increasing root-mean-square surface roughness in Fig. 2(d).

 

Fig. 2 (a) Deconvolution of the XPS spectrum for the 1 monolayer MoS₂ film, (b) S 2s peak binding energy position increases, and FWHM decreases as the number of MoS₂ layers increases, (c) Shifts of the A_1g and E¹_2g phonon modes with increasing number of MoS₂ layers, (d) Strain relaxation near the front surface as evident from the increasing RMS surface roughness with more MoS₂ layers. Inset: 3D AFM image for the 2 ML MoS₂ film.

Download Full Size | PDF

2.3 Fabrication process flow of large scale MoS₂ photodetectors

Figure 3 shows a gold-free CMOS-compatible fabrication flow to realize the MoS₂ photodetector. First, a double-sided polished planar n-type Si wafer with a resistivity of ~4-8 Ωcm has been utilized, in which the native oxide is removed using a 2% concentration dilute HF dip prior to the deposition of the 5 nm hafnium-dioxide (HfO₂) dielectric by atomic layer deposition (ALD) process (CambridgeNanoTech, Savannah Systems). This is followed by the deposition of large-area MoS₂ films using DC magnetron sputtering approach to obtain a precise number of MoS₂ layers (2 monolayers, 3 monolayers, and 5 monolayers). Details on the sputtering deposition are described in the preceding section. An ultrasonic cleaning in acetone is then performed before the deposition of the metal electrodes (100 nm titanium) by a combination of photolithography (SUSS MicroTec, MA6 mask aligner) and e-beam evaporation (Oerlikon, Univex 450B). The remaining photoresist is then removed by the lift-off process using acetone to obtain the final photodetector device in this work. The resulting large scale MoS₂ photodetector as shown in Fig. 3(e) was evaluated across a wide spectral range from the visible to the near-infrared wavelengths (400 to 1200 nm) for varying MoS₂ film thicknesses and channel biases.

 

Fig. 3 Fabrication process flow of the large scale MoS₂ photodetectors. Starting from a n-type Si wafer (a), a 5 nm HfO₂ dielectric is deposited by ALD approach (b), the large-scale magnetron sputtering deposition of MoS₂ films (c), and the patterning and deposition of the metal electrodes (d). (e) and (f) show the top-view optical image and scanning electron microscopy image of the completed MoS₂ photodetector, respectively.

Download Full Size | PDF

2.4 Large scale MoS₂ photodetector performance

Figure 4(a) shows the achievement of high photoresponsivity in the large scale MoS₂ photodetectors over a wide spectral range from 400 nm to 1200 nm. Interestingly, the photodetectors with 2 ML MoS₂ show an extended photoresponse beyond the cutoff wavelength of 800 nm reported in mechanical exfoliated photodetector [17]. This could be attributed to the inherent higher film strain in thinner MoS₂ deposition layers which causes a red-shift in its optical bandgap, as corroborated by the XPS measurements. Some possible mechanisms underlying this film strain include (a) stresses arising from the different thermal expansion coefficients of the MoS₂ film and the HfO₂/silicon substrate due to a high deposition temperatures of ~700°C; and (b) growth stresses arising from crystal structure changes after deposition.

 

Fig. 4 (a) Increasing MoS₂ thickness enhances photoresponsivity due to an increased photoabsorption. (b) Higher external quantum efficiency is achieved with increasing applied bias across a wide spectral regime.

Download Full Size | PDF

Enhanced photoresponsivity in the visible regime with thicker MoS₂ layers is also observed, which can be attributed to its high absorption coefficients. For an active MoS₂ absorption channel as thin as 2 monolayers, a photoresponsivity of ~0.4 A/W was measured in the visible wavelengths range, which is two orders of magnitude higher than the graphene counterpart. Further increase in the number of absorption layers from two to five increases the photoresponsivity to beyond 1 A/W. In particular, for a 5 ML photodetector, the highest attainable photoresponsivity is 1.8 A/W for a wavelength of 850 nm at a channel bias of 5 V. These results are superior over the monolayer MoS₂ phototransistors [5] (7.5 mA/W) and multilayer MoS₂ phototransistors [18] (120 mA/W) prepared by mechanical exfoliation method as well as the single-layer graphene-based FET [19, 20] (1 mA/W).

Apart from the utilization of thicker absorption layers, optimizing the applied channel bias can also enhance the photodetector performance as shown in Fig. 4(b). Increasing the channel bias not only enhances the charge collection efficiency, but also result in carrier multiplication through the impact ionization process. This leads to an increase in the external quantum efficiencies (EQE). To illustrate this, the measured EQE of a 5 ML MoS₂ photodetector was 30% and 263% at a wavelength of 850 nm for a bias of 1 V and 5 V, respectively. In general, thicker MoS₂ layers allows for more photoabsorption and carriers generation, while an increased channel bias facilitates the collection of photogenerated charge carriers and determines the carrier density dependent impact ionization rates.

However, this has to be tradeoff with a reduced photodetectivity as shown in Fig. 5(a), in which the 2 ML photodetector exhibits photodetectivities exceeding 10⁹ Jones, outperforming both the 3 ML and 5 ML photodetectors. This was primarily attributed to the significantly lower dark current in the 2 ML photodetector (~10⁻⁹ A), which is one order of magnitude lower than the 3 ML photodetectors, and two orders of magnitude lower than the 5 ML photodetectors. Such a phenomenon has been attributed to the different scattering mechanisms with different MoS₂ thicknesses, as reported in [21].

 

Fig. 5 (a) Increased photodetectivity (D*) with thinner MoS₂ layers due to (b) lower dark current level.

Download Full Size | PDF

Figure 6 shows the performance benchmarking of our large-scale MoS₂ photodetector to other reported photodetectors fabricated using mechanically exfoliated MoS₂, CVD MoS₂, and graphene films [5, 6, 18–20, 22]. Clearly, MoS₂ photodetectors enabled by large scale synthesis approach demonstrate promising performance in terms of photoresponsivity and photodetectivity. Such deposition method can be easily scaled up to wafer level to meet the industrial requirements, a distinct advantage over the non-uniform and non-scalable mechanical exfoliation approach.

 

Fig. 6 The PVD-synthesized MoS₂ photodetectors in this work outperform most of the previously reported MoS₂ photodetectors using mechanical exfoliation, and CVD techniques, and also Graphene based photodetectors

Download Full Size | PDF

In terms of switching characteristics, the PVD-deposited MoS₂ photodetector exhibits stable switching characteristics between the “ON” and “OFF” states with a clear saturation current for both states using the selected wavelength of 850 nm, and a channel bias of 5 V. A rise time of 0.3 s and a fall time of 0.36 s is extracted. The measured fall time outperforms that of mechanically exfoliated single layer MoS₂ photodetectors [6,7] which has typical fall time ranging from 2 s to 9 s. While this switching performance pales in comparison to the initial work on single-layer MoS₂ phototransistors (~50 ms) reported by Yin et al. [5], the recent work on Schottky metal-semiconductor-metal MoS₂ photodetectors [23] (rise time of ~70 µs and fall time of ~110 µs) as well as commercially available silicon-based avalanche photodiodes (~few nanoseconds response) [24], a further reduction to the response time in the PVD-deposited MoS₂ photodetectors of this work can be expected with a suitably applied gate bias similar to Ref [6].

3. Conclusion

Two-dimensional transition metal dichalcogenides photodetectors based on large scale MoS₂ crystals were successfully demonstrated using a gold-free CMOS compatible process. Synthesized by magnetron sputtering, a precise control of the MoS₂ film thickness and uniformity down to monolayer (~0.75nm) was achieved over a large area substrate, with potential of scaling up to wafer level. The MoS₂ photodetector was evaluated with a wide spectral response from the visible to the near-infrared wavelengths, which is attributed to its inherently high absorption coefficients. Note worthily, photodetector with five MoS₂ monolayers exhibits a high photoresponsivity of 1.8 A/W, an external quantum efficiency of exceeding 260%, and a photodetectivity of ~5x10⁸ Jones for a wavelength of 850nm, showing promise for optoelectronics applications.