III. RESULTS AND DISCUSSION

Results of structural characterization of one of the heterostructures, 20MoS₂/35BN, are presented in Fig. 1. The XRR reflectivity curve of this sample is shown (solid black). We would like to note that for quantitative roughness estimation we have utilized XRR for the following reasons. XRR is one of the widely used techniques for providing roughness information for a large area specimen quickly, which typically is not possible using atomic force microscopy (AFM) or scanning electron microscopy (SEM). Further, interface roughness can also be characterized using XRR, which is a major advantage compared to both AFM and SEM.

FIG. 1.

(a) XRR data of the 20MoS₂/35BN sample, showing XRR fringes indicative of the presence of sharp interfaces in these heterostructures. The inset shows the optical image of the deposited 20MoS₂/35BN heterostructure and blank Si substrate. (b) Possible existence of these interfaces.

Clear XRR fringes are observed in the data, which indicates that these multi-layers have reasonably sharp interfaces. A fit analysis of these data was performed to extract the thickness, roughness, and density of all the constituent layers (red curve). Apart from BN and MoS₂ layers, we found evidence of a rough, chemically mixed layer at the Si–BN interface in the samples. Without considering the existence of these layers, the goodness-of-fit was much higher than desirable. Even though the chemical composition of these layers cannot be quantified through this analysis, the fit density and scattering factor value are consistent with an oxy-nitride layer of Si and B at the Si–BN interface. This is not surprising since a native oxide layer of Si, and the presence of N₂ during BN deposition accentuates the nitride phase. The values of thickness and roughness of these interfacial layers obtained from our XRR investigations for both samples are provided in Table I. Apart from the presence of these interfacial layers, we also find a lower MoS₂ average density (compared to the bulk value) and small deviations from the nominal 1:2 MoS₂ stoichiometry within the sample area. However, Raman spectroscopy measurements performed on our samples showed clear evidence that the MoS₂ phase was present.

TABLE I.

Parameters extracted from XRR reflectivity measurements are listed.

Raman spectroscopic investigations are routinely used to identify layered MoS₂ structures. ^{Reference Windom, Sawyer and Hahn22–Reference Boukhicha, Calandra, Measson, Lancry and Shukla25} The data from the Raman measurements performed on 20MoS₂/35BN and 20MoS₂/80BN heterostructures are shown in Figs. 2(a) and 2(b), respectively. Multiple Raman spectra, with a laser exposure time of ∼120 s, were collected from several areas of the samples to validate the reproducibility of the data as well as the uniformity of the deposited samples. The Raman spectra showed signatures of vibrational modes related to MoS₂. Two of the most prominent peaks, around 381 and 407 cm⁻¹, were observed in the Raman data for both the samples and can be, respectively, assigned to the first-order ${\rm{E}}_{2{\rm{g}}}^1$ and A_1g modes at the Γ-point of the hexagonal Brillouin zone of 2D-MoS₂. ^{Reference Ghosh, Winchester, Muchharla, Wasala, Feng, Elias, Krishna, Harada, Chin, Dani, Kar, Terrones and Talapatra11,Reference Bilgin, Liu, Vargas, Winchester, Man, Upmanyu, Dani, Gupta, Talapatra, Mohite and Kar13,Reference Windom, Sawyer and Hahn22–Reference Boukhicha, Calandra, Measson, Lancry and Shukla25} A broad peak around 450 cm⁻¹ was also observed, which could be due to the second order longitudinal acoustic phonons near the M-point of the Brillouin zone [2LA(M)], often observed in multiple layers of MoS₂. Further, a sharp peak, around 285 cm⁻¹, was observed [see the inset of Figs. 2(a) and 2(b)]. This peak is generally associated with E_1g phonons near the Γ-point and was recently observed in multilayer samples grown using direct vapor-phase sulfurization of MoO₂ on SiO₂/Si as well as quartz substrates. ^{Reference Bilgin, Liu, Vargas, Winchester, Man, Upmanyu, Dani, Gupta, Talapatra, Mohite and Kar13}

FIG. 2.

(a) Raman spectrum obtained from the 20MoS₂/35BN sample. Several vibrational modes akin to MoS₂ layers are identified in this spectrum. The inset shows the presence of E_1g mode ∼290 cm⁻¹ in these samples. (b) The Raman spectrum obtained from the 20MoS₂/80BN sample. The inset shows the presence of E_1g mode ∼285 cm⁻¹ in these samples. The peak observed at 520 cm⁻¹ in both these spectra belongs to the Si substrate on which these heterostructures were fabricated.

We next investigated the photo-current response properties of the MoS₂/BN heterostructures. Before performing the photo-switching experiments, current–voltage (I–V) characteristics of these devices were measured in a two probe configuration. The samples were between the two contact electrodes, biased up to ±1 V, for performing the I–V measurements. Figure 3(a) shows the I–V curves obtained from typical heterostructures (without any light illumination). I–V curves for both the structures were found to be linear in nature. This indicated that good ohmic contacts persisted between the MoS₂ channel and the metal electrodes within the applied voltage regime. The linearity of I–V curves without illumination was tested to show that within the applied voltage regime, under which the measurements were performed, the contacts were able to replenish the carriers when they are drawn out from the material under an applied electric field. We found that the channel current is slightly increased with the increase in the thickness of the bottom BN layer.

FIG. 3.

(a) Linear current–voltage (I–V) response obtained from 20MoS₂/35BN and 20MoS₂/80BN samples is shown. (b) Room temperature photo-switching behavior of 20MoS₂/35BN under laser illumination of λ = 658 nm at different laser powers ∼0.8–10 μW is shown. (c) Similar data, as shown in (b) are presented for 20MoS₂/80BN. (d) Shows the variation of I _ph with laser intensity (at 10 μW) for both these samples. The dotted line in this figure indicates power fit to the data plotted on a double logarithmic scale. The measurements for all the data presented in this figure were performed at room temperature and under high vacuum (∼10⁻⁶ Torr).

In Figs. 3(b) and 3(c), the photo-switching characteristics of MoS₂/BN heterostructures at different laser intensities are shown. A source–drain voltage of 1 V was applied between the electrodes of the devices being measured and the laser intensity was varied from ∼1 to 10 μW (scaled value of laser power according to the device area). The photo-switching measurements were performed in the following way. The device was continuously biased with 1 V without turning the laser ON and the current (I _dark) was recorded. Subsequently, the laser intensity was fixed at a specific power and was switched ON for about 10 s, during which the evolution of the current (I _ill) was recorded. This sequence was followed for the photo-response of these devices for several laser intensities. Between each “laser ON” sequence we waited until the channel current reached its initial I _dark value (about 20–30 s). The photo-current (I _ph) is the current obtained by subtracting the current under illumination from the current obtained under dark conditions at different laser intensities (I _ph = I _ill − I _dark).

In Fig. 3(d), dependence of I _ph on laser power (P) is presented. In this figure, I _ph as a function of P is plotted in a double logarithmic scale. The data obtained seems to fit very well with a power law dependence of the form I _ph ∝ P ^γ, where γ = 1.04 and γ = 1.07 for 20MoS₂/35BN and 20MoS₂/80BN heterostructures, respectively. This behavior is surprisingly different from most of the studies reported so far in the literature. Typically, in highly crystalline materials, the value of γ = 0.5 corresponds to bimolecular processes, or γ = 1 corresponds to monomolecular processes. ^{Reference Rose26} These processes are often observed in semiconducting systems with few trap states. Power exponents with values of γ ranging between 0.5 and 1 (0.5 < γ < 1.0) are also very common in disordered semiconductors and in chalcogenide glasses. ^{Reference Rose26–Reference Mott and Davis30} In the absence of the photo-phenomenon arising from photo-gating artifacts, thermoelectric effects and so on, fractional dependence of the power exponent seen in a variety of systems was explained on the basis of a theoretical model which assumes an exponential distribution of trap states in the mid-gap state. ^{Reference Rose26} Such fractional distribution was recently observed for thin film photo-detectors fabricated using liquid phase exfoliated MoS₂ materials. ^{Reference Ghosh, Winchester, Muchharla, Wasala, Feng, Elias, Krishna, Harada, Chin, Dani, Kar, Terrones and Talapatra11}

However, a few reports on the super linearity of power exponent γ in several systems have been reported in the past. ^{Reference Kushwaha, Kushwaha, Shukla and Kumar31,Reference Bakr32} Though to date a clear theoretical understanding of such a phenomenon is still lacking, a few have tried to explain it on the basis of the presence of recombination centers with different energies and capture cross-sections. ^{Reference Rose26} More recently, such super linear dependence of I _ph over a broad range of laser power was also observed in a variety of compositionally different MoS₂ and MoSe₂ systems ^{Reference Klee, Preciado, Barroso, Nguyen, Lee, Erickson, Triplett, Davis, Lu, Bobek, McKinley, Martinez, Mann, Talin, Bartels and Léonard33} and was explained on the basis of a model, which considered the presence of three types of recombination centers in the samples studied. Whether a similar model can be applied to explain the behavior seen in our experiments or some other, more complicated, phenomenon is responsible for the super linearity of the γ observed in the MoS₂/BN heterostructures, is, at present, beyond the scope of this article and will be investigated in greater detail in the future.

We also calculated the responsivity values of these samples at different laser powers. Responsivity (R) is defined as R = I _ph/P _light (where P _light is the power of the incident light), and is an important performance factor used for quantifying the effectiveness of photoactive materials. Figure 4(a) shows the room temperature responsivity of the MoS₂/BN heterostructures measured under high vacuum (∼10⁻⁶ Torr) at different laser powers. The responsivity at 10 μW was found to be ∼2.7 and ∼12 mA/W for 20MoS₂/35BN and 20MoS₂/80BN heterostructures, respectively. These responsivity values are several orders of magnitude greater than the responsivity of liquid exfoliated MoS₂ thin film photodetectors ∼50 μA/W (Ref. Reference Ghosh, Winchester, Muchharla, Wasala, Feng, Elias, Krishna, Harada, Chin, Dani, Kar, Terrones and Talapatra11) as well as chemical vapor deposition (CVD) grown multilayer WS₂ based devices ∼22 μA/W. ^{Reference Perea-López, Elías, Berkdemir, Castro-Beltran, Gutiérrez, Feng, Lv, Hayashi, López-Urías, Ghosh, Muchharla, Talapatra, Terrones and Terrones27} Thus a very convincing evidence of the enhancement of photocurrent was observed with the increase in the underlying BN layer thickness in these heterostructure. Though, several factors can affect photocurrent generation in optically active semiconductors (e.g., presence of trap states), recent investigations show ^{Reference Buscema, Steele, van der Zant and Castellanos-Gomez14,Reference Lien, Kang, Amani, Chen, Tosun, Wang, Roy, Eggleston, Wu, Dubey, Lee, He and Javey15} that underlying dielectric layer has profound effects on the optical properties of TMDCs. For example, it was hypothesized ^{Reference Buscema, Steele, van der Zant and Castellanos-Gomez14,Reference Lien, Kang, Amani, Chen, Tosun, Wang, Roy, Eggleston, Wu, Dubey, Lee, He and Javey15} that due to the several layers present in these hetero-structures (e.g., MoS₂, BN, and Si) absorbed light can undergo multiple reflections between the layer boundaries. As a result of these multiple reflections, constructive interference between absorption and emission of light occurs. Such effects can substantially alter the optical properties of the top layer materials. We believe that a similar phenomenon is perhaps responsible in the present case as well.

FIG. 4.

(a) Responsivities obtained for the samples 20MoS₂/35BN and 20MoS₂/80BN as a function of laser power is shown. These measurements were performed at room temperature under a high vacuum of ∼10⁻⁶ Torr. (b) Responsivities obtained for five different 20MoS₂/80BN devices under ambient conditions. (c) Shows an optical micrograph of 20MoS₂/80BN with contact pads on top along with a schematic of the two point probe arrangement used for the measurements to obtain the data shown in (b).

We further found that the responsivity values for these heterostructures drastically reduce when they are exposed to ambient conditions. For example, the measurement on several 20MoS₂/80BN samples in the presence of air and with a laser illumination of 10 μW showed roughly an order of magnitude reduction in their responsivity (between 1 and 0.5 mA/W for five different devices measured) values. These results are presented in Fig. 4(b). Such reduction in responsivities is not surprising and is associated with the fact that MoS₂ surfaces are extremely sensitive to the surrounding environmental conditions. ^{Reference Qiu, Pan, Yao, Li, Shi and Wang34,Reference Park, Park, Jang, Lee, Jeong, Cho, Hong and Lee35} Specifically, channel current in MoS₂ FETs deteriorates significantly under ambient conditions ^{Reference Qiu, Pan, Yao, Li, Shi and Wang34} as well as in the presence of oxygen. ^{Reference Park, Park, Jang, Lee, Jeong, Cho, Hong and Lee35} Such deterioration is caused by the electron capturing nature of adsorbed oxygen on the sulfur or defect sites of MoS₂, which results in the reduction of the channel current. ^{Reference Qiu, Pan, Yao, Li, Shi and Wang34} We anticipate similar processes occurring in our case as well.