Introduction

Interest in atomically thin two-dimensional (2D) layered compounds is growing due to unique physical properties found for monolayer (ML) structures1,2. One such material, molybdenum disulfide (MoS2), has generated particular interest due to the presence of an indirect-to-direct band gap transition and observation of photoluminescence (PL)3,4,5 and electro-luminescence6 in the visible range up to room temperature. A high on/off ratio (exceeding 108) has suggested a potential use in field effect transistors7, while a strong valley polarization is likely to be used in the development of future valleytronics applications8,9,10,11,Fig. 2. A 100 nm thick layer of either SiO2 or SixNy is deposited using PECVD on top of the MoS2/SiO2/Si samples for both PECVD and thermal SiO2/Si substrates. Here we observe even less variation in lineshapes between the films. A further suppression of the low energy shoulder L and neutral exciton peak A0 is found for films capped with SixNy (a,b,e,f) on both types of substrates and with SiO2 on thermally grown substrates. In contrast, L and A0 peaks are pronounced when cap** with SiO2 is used for MoS2 films on PECVD substrates. Further to this, from comparison of spectra in (a,b,c,d) and (e,f,g,h), we find that the PL linewidths of films deposited on the PECVD oxide are notably broader than for those on the thermal oxide substrates.

Figure 2
figure 2

PL spectra measured for individual mechanically exfoliated MoS2 films capped by a 100 nm PECVD layer of dielectric material.

The effect of cap** is shown for films deposited on PECVD grown SiO2 substrates for SixNy (a), (b) and SiO2 (c), (d) cap** layers and also for films deposited on thermally grown SiO2 and capped with SixNy (e), (f) and SiO2 (g), (h).

Full size image

An interesting trend in all spectra presented in Figs. 1 and 2 is a correlation between the intensities of the features L and A0: the two peaks are either both rather pronounced or suppressed in any given spectrum relative to the trion peak A−. This may imply that peak L becomes suppressed when the film captures an excess of negative charge.

Analysis of PL peak energies

A statistical analysis of PL peak energies for films deposited on the two types of substrates is presented in Fig. 3. Fig. 3(a,b) show that the average values for the PL peak energies, , for uncapped films are for the PECVD substrates and for thermal oxide substrates, with an almost two times larger standard deviation, σEmax for the former (18 versus 11 meV). The data collected for the capped films (shaded for SixNy and hatched for SiO2) are presented in Fig. 3(c) and (d) for the thermal and PECVD oxide substrates, respectively. Significant narrowing of the peak energy distribution is found in all cases: σEmax ≈ 6 meV has been found. The average peak energies are very similar for both SiO2 and SixNy cap** on the thermal oxide substrates (), but differ for PECVD substrates: and 1.870 eV for SiO2 and SixNy cap**, respectively.

Figure 3
figure 3

(a–d) PL peak energies for A exciton complex in MoS2 thin films.

Data for films deposited on thermally (PECVD) grown SiO2 substrates are shown in top (bottom) panels. Panels (a)–(b) and (c)–(d) show PL peak positions for uncapped and capped films, respectively.

Full size image

From previous reports4, for films with thicknesses in the range 2 to 5 MLs, one can expect the PL peak shift on the order of 20 meV. In addition, PL yield was reported to be about 10 times higher for 2 ML films compared with 4 ML and for 3 ML compared with 5 ML4. In our study, the integrated PL signal shows a large variation within about one order of magnitude between the films. The dependence of the PL yield on the type of the substrate and cap** is not very pronounced. While our data for PL intensities is consistent with the reported in the literature for the range of thicknesses which we studied, the PL peak energy distribution shows the unexpected broadening for uncapped samples: for example, deviations from by ± 20–30 meV are evident in Fig. 3(a,b). For the capped samples, new trends are observed: the significant narrowing and red-shift of Emax distributions. As shown below, these effects reflect changes in the PL lineshapes between the capped and uncapped samples, which in their turn reflect changes in the relative intensities of the A−, A0 and L peaks.

We note that the new experimental trends observed in our PL studies do not depend on the exact distribution of thicknesses in the ensembles of the investigated films, provided these distributions are similar for all types of samples studied. The latter is the case in our study, as the films were produced using the same method, show similar range of the colour-contrasts under optical examination and exhibit similar ranges of PL yield.

Analysis of PL lineshapes

In this section we will present the lineshape analysis for the A exciton PL based on the measurement of full width at half maximum (FWHM) in each PL spectrum. This approach allows one to account for contributions of the three PL features, L, A0 and A−. The data are summarized in Fig. 4 and Table 1.

Table 1 Mean values, standard deviations and coefficients of variation for full width at half maximum of PL spectra measured for thin MoS2 films
Full size table
Figure 4
figure 4

PL FWHM of exciton complex A in thin MoS2 films.

Data for MoS2 films deposited on thermally and PECVD grown SiO2 substrates is shown with blue and red, respectively. (a) PL FWHM of uncapped MoS2 films. (b) PL FWHM of SixNy capped MoS2 films. (c) PL FWHM of SiO2 capped MoS2 films.

Full size image

PECVD grown SiO2 substrates

These data are presented in Fig. 4 in red. Data for uncapped films are shown in Fig. 4(a), from where it is evident that the lineshapes vary dramatically from film to film within a range from 50 to 170 meV. FWHM for uncapped films on PECVD grown substrates is on average with a large standard deviation σFWHM = 33 meV. This gives a rather high coefficient of variation .

The non-uniformity of lineshapes of the PL spectra is significantly suppressed by cap** the films with SixNy and SiO2 (shown with red in Fig. 4(b) and (c), respectively). This is evidenced from the reduction of the coefficient of variation in the FWHM values by a factor of 4 in capped films compared with the uncapped samples (in Table 1). Despite the narrowed spread of ΔEFWHM values, the average FWHM in SiO2 capped films is rather high, 109 meV, which reflects relatively strong contribution of L and A0 PL features. Contributions of A−, L and A0 features vary very considerably in the uncapped samples, leading to on average smaller linewidths but a very significant spread in FWHM values. In contrast, in SixNy capped films, A− peak dominates and both L and A0 features are relatively weak, which effectively results in narrowing of PL.

Thermally grown SiO2 substrates

These data are presented in Fig. 4 in blue. It can be seen that uncapped films deposited on the flatter thermal oxide substrates appear to have significantly narrower distributions of linewidths compared to uncapped films on PECVD substrates: coefficient of variation of ΔEFWHM is by a factor of 2 smaller for films on the thermally grown substrates [see Fig. 4(a) and Table 1]. In addition, compared with the films deposited on PECVD grown SiO2, FWHM is also reduced by about 20% to 79 meV. Such narrowing reflects weaker contribution of L and A0 peaks in PL spectra.

The non-uniformity of the PL spectra still present in uncapped films deposited on thermally grown SiO2 is further suppressed by cap** the films with SixNy and SiO2 [shown with blue in Fig. 4(b) and (c), respectively]. In general, the coefficients of variation for FWHM of the capped films are rather similar for both substrates and are in the range of 0.06–0.09, showing significant improvement of the reproducibility of PL features compared with the uncapped samples (see Table 1). For SixNy capped films on thermally grown SiO2, we also observe narrowing of PL emission to . This reflects further suppression of L and A0 peaks relative to A−, the effect less pronounced in SiO2 capped films.

AFM and UFM measurements

To further understand the interactions between MoS2 films and the substrate/cap** materials, we carried out detailed AFM and UFM measurements of our samples (Fig. 5). AFM measurements of films deposited on PECVD grown substrates Fig. 5(a) show that the film is distorted in shape and follows the morphology of the underlying substrate. The root mean square (rms) roughness Rrms of these films is 1.7 nm with a maximum height Rmax = 11 nm, similar to the parameters of the substrate, Rrms = 2 nm and Rmax = 15 nm. Such Rmax is greater than the thickness of films (<3 nm), leading to significant film distortions. UFM measurements of these films [Fig. 5(b)] show small areas of higher stiffness (light colour, marked with arrows) and much larger areas of low stiffness (i.e. no contact with the substrate) shown with a dark colour. This shows that the film is largely suspended above the substrate on point contacts.

Figure 5
figure 5

AFM (left column) and UFM (right column) images for MoS2 thin films deposited on PECVD and thermally grown SiO2 substrates.

(a), (b) PECVD substrate, uncapped MoS2 film; (c), (d) thermally grown substrate, uncapped MoS2 film; (e), (f) PECVD substrate, MoS2 film capped with 15 nm of SiO2 grown by PECVD; (g), (h) thermally grown substrate, MoS2 film capped with 15 nm of SiO2 grown by PECVD.

Full size image

AFM measurements of films deposited on thermally grown SiO2 substrates [Fig. 5(c)] show a much more uniform film surface due to the less rough underlying substrate. This is reflected in a significantly improved Rrms = 0.3 nm and Rmax = 1.8 nm. These values are still higher than those for the bare substrate with Rrms = 0.09 nm and Rmax = 0.68 nm. A more uniform stiffness distribution is observed for these films in UFM [Fig. 5(d)], although the darker colour of the film demonstrates that it is much softer than the surrounding substrate and thus still has relatively poor contact with the substrate. A darker shading at film edges demonstrates that they have poorer contact than the film center and effectively curl away from the substrate.

AFM and UFM data for films capped with 15 nm SiO2 after deposition on PECVD and thermally grown SiO2 are given in Fig. 5(e, f) and (g, h), respectively. For the PECVD substrate, the roughness of the MoS2 film is similar to that in the uncapped sample in Fig. 5(a): Rrms = 1.68 nm and Rmax = 10.2 nm. From the UFM data in Fig. 5(f), it is evident that although the contact of the MoS2 film with the surrounding SiO2 is greatly improved compared with the uncapped films, a large degree of non-uniformity is still present, as concluded from many dark spots on the UFM image. In great contrast to that, the capped MoS2 film on thermally grown SiO2 is flatter [Fig. 5(g, h)], Rrms = 0.42 nm and Rmax = 6.1 nm, with the roughness most likely originating from the PECVD grown SiO2 cap** layer. The UFM image in Fig. 5(h) shows remarkable uniformity of the stiffness of the film similar to that of the capped substrate, demonstrating uniform and firm contact (i.e. improved mechanical coupling) between the MoS2 film and the surrounding dielectrics.

Discussion

There is a marked correlation between the PL properties of the MoS2 films and film stiffness measured by UFM. The stiffness reflects the strength of the mechanical coupling between the adjacent monolayers of the MoS2 film and the surrounding dielectrics. The increased bonding and its uniformity for films deposited on less rough thermally grown SiO2 substrates and for capped MoS2 films manifests in the more reproducible PL characteristics, leading to reduced standard deviations of the peak positions and linewidths. These spectral characteristics are influenced by the relative intensities of the three dominating PL features, trion A−, neutral exciton A0 and low energy L peak, which are influenced by the charge balance in the MoS2 films sensitive to the dielectric environment. The efficiency of charging can be qualitatively estimated from the relative intensities of A− and A0 peaks. In the vast majority of the films, A− dominates. As noted above, the intensity of A0 directly correlates (qualitatively) with that of the relatively broad low energy PL shoulder L (see Fig. 1 and 2), previously ascribed to emission from surface states. The lineshape analysis presented in Fig. 4 and Table 1 is particularly sensitive to the contribution of peak L.

The PL lineshape analysis and comparison with the UFM data lead to conclusion that negative charging of the MoS2 films is relatively inefficient for partly suspended uncapped films on rough PECVD substrates. Both in SiO2 and SixNy capped films on PECVD substrates, the charging effects are more pronounced. However, both A0 and L features still have rather high intensities. The relatively low charging efficiency is most likely related to a non-uniform bonding between the MoS2 films and the surrounding dielectric layers as concluded from from UFM data [see Fig. 5(f)]. The charging is more pronounced for uncapped MoS2 films on thermal oxide substrates and is enhanced significantly more for capped films: for SixNy cap** A0 and L peaks only appear as weak shoulders in PL spectra.

It is clear from this analysis that the charge balance in the MoS2 films is altered strongly when the films are brought in close and uniform contact with the surrounding dielectrics, enabling efficient transfer of charge in a monolithic hybrid heterostructure. Both n-type4,7,17 and p-type17,18 conductivities have been reported in thin MoS2 films deposited on SiO2. It is thus possible that the sign and density of charges in exfoliated MoS2 films may be strongly affected by the properties of PECVD grown SiO2 and SixNy, where the electronic properties may vary depending on the growth conditions19,20,21. It is notable, however, that for a large variety of samples studied in this work, the negative charge accumulation in the MoS2 films is pronounced and is further enhanced when the bonding of the films with the dielectric layers is improved.

In order to estimate the density of the accumulated charges we refer to Ref. 16, where PL spectra as a function of electron density in the film were measured. The neutral exciton PL peak A0 becomes less intense than the trion peak A− at the electron density n ≈ 2 × 1012 cm−2. Since in our experiments in many films A0 peak is relatively pronounced, we conclude that we have studied the regime where the electron densities are of the order of 1012 cm−2 or less.

The band-structure of MoS2 and hence its optical characteristics can also be influenced by strain22,Fig. 3). On the other hand, do**-dependent Stokes shifts of the trion PL have been found recently16, which may explain the behavior we find in charged MoS2 sheets. One would expect a more uniform strain distribution in the case of uniform mechanical properties of the sample, which as shown by UFM is achieved for capped MoS2 films on flat thermally grown SiO2 substrates. In order to roughly estimate a possible magnitude of strain in our films we refer to recent work in Refs. Micro-photoluminescence experiments

Low temperature (10 K) micro-PL was carried out on a large number of thin films in a continuous flow He cryostat. The signal was collected and analyzed using a single spectrometer and a nitrogen-cooled charged-coupled device. The sample was excited with a laser at 532 nm. All PL spectra presented in this work were measured in a range of low powers where no dependence on power of PL lineshape was found.

AFM/UFM experiments

As shown elsewhere33, the ultrasonic force microscopy (UFM) allows imaging of the near-surface features and subsurface interfaces with superior nanometre scale resolution compared to AFM techniques34. In the sample-UFM modality used in this paper35, the sample in contact with the AFM tip is vibrated at small amplitude (0.5–2 nm) and high frequency (2–10 MHz), much higher than the resonance frequencies of the AFM cantilever. The resulting sample stress produces a reaction, that is modified by the voids, subsurface defects or sample-substrate interfaces and can be detected as an additional ‘ultrasonic’ force. A unique feature of UFM is that it enables nanometre scale resolution imaging of morphology of subsurface nano-structures and interfaces of solid-state objects. In order to interpret the images of a few layer films presented in Fig. 5, one can note that the bright (dark) colors correspond to higher (lower) sample stiffness.