Effects of atomic layer deposition on the optical properties of two-dimensional transition metal dichalcogenide monolayers

Two-dimensional semiconducting transition metal dichalcogenides (TMDs) have attracted significant interest due to their unique optoelectronic properties. More often, these materials are enclosed inside a dielectric layer that can work as an insulator for field-effect transistors. The insulating layer is typically grown with atomic layer deposition (ALD). Here, we study the effects on bare and hBN-covered monolayer MoS2 and WSe2 flakes with ALD TiO2 films. Our results reveal a significant shift and decrease in intensity in photoluminescence and Raman signals of the monolayer TMDs. Further analysis suggests that these changes are caused by chemical doping, strain, and dielectric screening after the ALD. Our study not only sheds light on the impact of ALD on the properties of TMDs, but also indicates ALD can be an alternative method to engineer the doping, strain and dielectric environment for potential optoelectronics and photonics applications.

Atomic layer deposition (ALD) is a thin film deposition method that allows highly accurate control of film thickness and composition. Growing dielectric films by ALD on semiconductor materials is a standard industrial process in the semiconductor industry. The dielectric coating provides chemical and thermal stability that enhances the performance of semiconductors and extends their lifetime [19,20]. Indeed, ALD has been widely used for the present research of 2D layered materials [13][14][15][16]18]. However, the influence of ALD on the optoelectronic performance of TMDs remains exclusive. Although TMDs possess the qualities for future optoelectronic applications, they are known to be chemically unstable at normal ambient conditions, and their optical characteristics depend significantly on their dielectric environment [21,22]. Therefore, it is of great significance to study the effects of ALD on the optical properties of TMDs.
Here, we focus on three possible mechanisms in the ALD coating process that can affect the optical and structural properties of TMD monolayers: chemical effect (e.g. chemical doping or oxidation), dielectric screening, and mechanical strain. To fully understand these mechanisms, we study the optical and structural properties of TMD monolayers under different coating conditions that are compared with bare TMD monolayers and hexagonal boron nitride encapsulated TMDs. Two TMD monolayers (i.e. WSe 2 and MoS 2 ) are used in this study. We select ALD-grown TiO 2 as an example by following a standard industrial procedure. We reveal that by intercalating a dielectric hBN, the effect of chemical doping can be significantly reduced during the ALD process, enhancing the chemical and thermal stability of the TMD monolayers for better integration into the standard ALD process.

Methods
WSe 2 and MoS 2 are mechanically exfoliated and transferred by the dry stamp method using polydimethylsiloxane as the host polymer [23] on silicon substrates with a 300 nm SiO 2 layer on top. Similarly, used hBN flakes are exfoliated and transferred using polypropylene carbonate as the host polymer [24] to cover the TMD flakes. To ensure reliable results, multiple samples are prepared, with a minimum sample area of 10 µm 2 utilized for each composition case. A typical TMD flake and its resulting hBN/TMD heterostructure are presented in figure 1(a). The TiO 2 films are grown by ALD in the Picosun SUNALE R-200 Advanced ALD system. The growing temperature is kept at 200 • C, and the precursors are water and TiCl 4 . The sample is preconditioned in the ALD system by using four pulses of water to improve the adhesion on 2D materials [25]. This is followed by 500 alternating cycles of water and TiCl 4 to deposit a ∼37 nm thick TiO 2 film. The film thickness and refractive index are measured with an ellipsometer (Semilab Plasmos SE-2000), and the results are obtained by fitting a linearly combined Tauc-Lorentz and Gaussian dispersion models to the measured data.
The optical measurements are performed with a Raman spectroscopy (WITec alpha300 RA+). With this system, we measure the photoluminescence (PL) of WSe 2 samples using a laser at 633 nm wavelength. The PL of MoS 2 and Raman scattering of both TMDs (i.e. WSe 2 and MoS 2 ) are measured using a 532 nm excitation laser. The excitation power for both lasers is fixed at ∼500 µW, while the spot size is estimated to be ∼510 nm for the 532 nm laser and ∼610 for the 633 nm laser. For time-resolved PL (TRPL) measurements, an excitation laser with a wavelength of 532 nm and a repetition rate of 80 MHz is employed. A signal is detected with MPD single-photon avalanche photodiode, and Picoquant PicoHarp 300 is used as a counter.

Results and discussion
The optical characterization of the ALD-deposited TiO 2 films reveals a combination of amorphous and crystalline phases. The measured film refractive index of ∼2.31 (at a wavelength of 632.8 nm) is consistent with the previously reported results for amorphous TiO 2 [26]. The morphology of the TiO 2 films is studied by scanning electron microscopy (SEM) and atomic force microscopy (AFM). SEM and AFM images reveal small grain-like structures on the surface of the TiO 2 film (figures S2 and 1(b)). These structures are most likely polycrystalline grains grown on an amorphous film, which agrees well with the previous ALD TiO 2 results grown at 200 • C [27]. These polycrystalline grains have most likely anatase crystalline structures due to an observed Raman scattering mode at ∼142 cm −1 (figure S1) [28]. Figure 1(b) also shows that these crystalline grains do not grow as much on the TMD crystals as on a SiO 2 substrate. A grayscale intensity distribution analysis reveals that the grains cover ∼22.3% of the surface of TiO 2 grown on TMD compared to ∼53.6% of TiO 2 grown on the SiO 2 substrates (figure S1). Note that there are also a few pinholes in the TiO 2 film visible on the surface of TMD. Nevertheless, ALD TiO 2 grown on TMDs has a reasonably high quality, comparable to the previously reported ALD films grown on top of 2D materials [29,30].
To study the effects of ALD coating on the optical properties of WSe 2 and MoS 2 , we measure the PL and Raman scattering before and after the ALD at room temperature. We compare these two cases to the WSe 2 monolayer encapsulated with an hBN flake, on top of which we grow the TiO 2 film. We appoint the observed optical changes to three mechanisms: chemical effects (including chemical doping and oxidation), dielectric screening, and strain. These different mechanisms are illustrated in figure 1(c).
ALD relies on different precursors attaching to the sample surface. Since TMD layers have no dangling bonds and are typically attached to nearby materials via van der Waals forces, they are unideal for ALD coating. In our growing process, we introduce water pretreatment to improve precursor attaching on the surface of TMDs [25]. At high growing temperatures involved in the ALD process, the water molecules can oxidize the TMDs, as illustrated in figure 1(c). Thus, the intensities of the Raman scattering and PL are expected to drop. Additionally, TiO 2 is typically an ntype semiconductor [31], and therefore, it can introduce n-type doping to TMD when grown on top. As the PL of TMDs is typically exciton based, the excess electrons alter the PL to become more trion based. The doping of the TMD has also been shown to influence the Raman scattering modes [32]. An increase in the surrounding dielectric constant typically lowers the binding energy of excitons [22,33,34]. Since the PL of TMDs is exciton based, a notable change in the transition energy of the PL can be observed when a material of a higher dielectric constant is deposited on top. In a combined coating of hBN and TiO 2 , if the hBN layer is thin enough, the dielectric screening caused by TiO 2 can still influence the underlying TMD. Dielectric screening can also shift the Raman modes [35]. ALD TiO 2 films have shown quite strong residual stress due to higher growth temperature. It has been reported that this residual stress can reach values of 200-600 MPa [36]. If we consider this stress to transfer as such to the substrate silicon biaxially, we get a biaxial strain of roughly ∼0.1%-0.4%. The confirmation of the strain induced crystalline change can be facilitated with transmission electron microscopy (TEM). However, as the strain mainly affects the substrate and, consequently, the TMD flake, the response of the sample prepared on a TEM grid may exhibit notable variations compared to that on silicon substrate. Nevertheless, it has been demonstrated that the PL and Raman modes are highly affected by strain [37][38][39][40][41]. Therefore, even though the strain is not that significant, its influence should be observed in the optical properties. Figure 2(a) shows the measured PL data and the Gaussian fit for each data set. From the Gaussian fit, we extracted the PL intensity, the transition energy, and the PL linewidth. These values are shown in figure 2(b). The most prominent change of TMDs after the ALD of TiO 2 is the PL quenching (i.e. PL signal intensity decrease). There is also a notable redshift in PL emission center after the ALD of TiO 2 . In addition, the linewidth of the PL peak is greatly increased. A similar redshift of the PL of WSe 2 has been previously observed [42] and has been signed to doping, which causes the emission to transfer to more trion based as the typical PL of uncoated WSe 2 is based more on neutral excitons [43]. This would also explain widening of the PL as it is now contributed by two peaks with similar level of intensity. Increasing the thickness of TiO 2 ALD from ∼37 to 80 nm (by increasing the number of ALD cycles to 1000) shows a notable intensity reduction in PL as more light is scattered by the ALD film (figure 2(c)). Interestingly, there was no further redshift, nor widening observed. This further supports the idea of TiO 2 coating-based doping via chemical reactions at the surface of WSe 2 . Additionally, we grow the film at a higher temperature of 300 • C, and the results show no significant difference from those grown at 200 • C ( figure S2).
On the other hand, after hBN encapsulation, the PL intensity is slightly increased, a small redshift can be observed, and linewidth is marginally reduced. Covering WSe 2 with hBN has been shown to reduce the PL linewidth before, which has been assigned to reduced interaction with the ambient environment [44]. However, covering the hBN/WSe 2 heterostructure with TiO 2 , a noticeable redshift is also observed. This shift is not nearly as high as without hBN, and additionally, the linewidth is a little narrower. In addition, the PL intensity does not drop as significantly in comparison to TiO 2 directly coated WSe 2 flakes. As the surface-based chemical reaction between TiO 2 and WSe 2 should be mitigated due to the protective hBN layer, the redshift cannot be assigned to those. Previous studies show a redshift of ∼50-100 meV per strain percentage [37]. Thus, a shift of ∼5-40 meV is expected for the approximated strain values. As these values are quite similar to the observed ∼10 meV for the hBN encapsulated, we can assign the shift to the strain effect. Although, it has been demonstrated that when the WSe 2 monolayer is subjected to strain, the PL peak should get narrower by ∼9.16 meV/% [45]. This is not observed for our TiO 2 -coated samples, which are widened due to the doping effect. However, the linewidth gets slightly narrower for the hBN-covered sample. The observed narrowing of ∼1.5 meV, would correspond to a strain of ∼0.16%. Therefore, we can assign the main observed shift and narrowing of the PL peak of the hBN-encapsulated TiO 2 -coated sample to the strain. For non-encapsulated sample (i.e. pure TMD flakes), the observed shift and widening are mainly due to doping, but they should be partly affected by the strain.
As mentioned before, the other significant change observed is the drastic decrease in the PL intensity. The ellipsometry gives a refractive index of ∼2.3. By Fresnel equations, we get a reflection of ∼15.5% for normal incidence at the TiO 2 -air interface. Therefore, when taking into account a transmission of ∼84.5% for both, the excitation and PL, the signal intensity is reduced by ∼28.6% with the addition of the TiO 2 film interface. Furthermore, the crystalline grains on the surface of films scatter the excitation and PL, reducing intensity even further. This effect is even more pronounced for the thicker and grainier TiO 2 film (figures 1(c) and S1). As the TiO 2 is a semiconductor, charge transfer between TiO 2 and WSe 2 can also reduce the PL. For pure WSe 2 (non-hBN-encapsulated sample), the intensity is decreased to ∼0.5%. For hBN encapsulated samples, the resulting intensity is ∼3% of the original (i.e. before ALD deposition). Therefore, we can estimate chemical reactions, defects and charge transfer to lower the intensity nearly by an order of magnitude.
To analyze exciton dynamics of our samples, we perform TRPL measurements for bare WSe 2 , TiO 2 coated WSe 2 and TiO 2 coated and hBN covered WSe 2 sample. The results are shown in figure 2(d). The fitting the data for the bare WSe2 sample results in an exponential fit with time constant 0.459 ns, which is consistent with previously reported values [46,47]. For TiO 2 covered sample, biexponential fit gave time constants ∼1.162 and 9.037 ns. Similarly, the TiO 2 -coated and hBN-covered sample was also fitted with a biexponential function, giving time constants of ∼0.734 ns and 5.731 ns. It is evident that TiO 2 coating not only significantly increased the time decay, which is more than double that of the bare WSe 2 , but also introduced a slower decay component. Both decay times are smaller for the TiO 2 coated hBN/WSe 2 heterostructure sample. Previous studies have demonstrated that defects can extend the decay time [48,49], and the relatively slower component may be attributed to additional recombination levels instructed by these defects. Furthermore, as mentioned before, the addition of TiO 2 and the resulting scattering have decreased the excitation power reaching the WSe 2 , and it has been established that lower excitation power leads to longer decay times [50]. It is highly likely that both factors contribute to the observed results. Nevertheless, covering WSe 2 with hBN, has decreased the decay times, possibly due to a reduction in defects. Nonetheless, the decay times are still longer than those of bare WSe 2, which might be due to the aforementioned excitation intensity reduction. It has also been presented that strain can decrease decay times [51]. However, considering the more significant factors contributing to recombination in our samples, the effect of strain is likely negligible.
To explore the possible crystalline and chemical changes induced by ALD of TiO 2 , we measure the Raman scattering spectra of the TiO 2 -covered WSe 2 . The resulting spectra are shown in figure 3(a). The average values of intensity, shift, and linewidth are illustrated in figure 3(b). For WSe 2 , there is a Raman peak at ∼251.7 cm −1 , which is assigned to merged E 1 2g and A 1g vibrational modes [52,53]. This Raman peak is downshifted ∼2.22 cm −1 after ALD of TiO 2 . On the other hand, covering WSe 2 with hBN also downshifts the E 1 2g /A 1g peak by ∼0.75 cm −1 . The E 1 2g /A 1g peak of the hBN-covered WSe 2 sample is further downshifted by ∼1.08 cm −1 when the TiO 2 film is deposited. However, the final downshift of the peak (that is ∼1.83 cm −1 ) is smaller than that of the TiO 2 -covered bare WSe 2 . The observed downshift of the E 1 2g /A 1g peak of WSe 2 after the ALD of TiO 2 can be caused by n-type doping which has been perceived before [54,55]. This is supported by the fact that TiO 2 is typically an n-type semiconductor [31].
The E 1 2g and A 1g modes have been reported to shift ∼0.4 and 0.5 cm −1 per strain percentage when uniaxial strain is applied [38]. Even though the ALDinduced strain is biaxial, and thus the strain effect can be assumed to be higher than the uniaxial, the shift seen for TiO 2 covered hBN/WSe 2 sample is large enough that there should be other factors at play. The downshift caused by hBN encapsulation has been observed before and assigned to the change of dielectric environment [56]. As the effects of doping should now be minimized, the majority of the TiO 2 -induced shift should be caused by the dielectric screening. This theory is supported by the fact that the final shift is reduced by adding the hBN layer. Even though the screening caused by hBN is still present, the increased distance between TiO 2 and WSe 2 lowers the overall dielectric screening effect. The 2LA Raman peak follows the observed changes seen for E 1 2g /A 1g peak (figure 3). For TiO 2 covered WSe 2 samples, this peak sees higher uncertainty for the linewidth, but this is mostly caused by the signal intensity approaching the noise level and thus making the fitting more uncertain.
We also measure the effects of TiO 2 ALD films on monolayer MoS 2 . As with WSe 2 , we measure the effects of a thin hBN encapsulation layer. Here for analysis, we fit two Gaussian peaks for A and B excitons. Figure 4(a) shows measured PL signals and corresponding fits. From the fit, we obtain the average intensities, transition energies, and linewidth.
These values are shown in figure 4(b). The TiO 2 layer shifts the A and B exciton-related PL peaks by ∼21 and 14 meV, respectively. Small widening of both peaks can also be observed as the full-width at half maximum of A (B) PL peak increases from ∼94.6 (178.6) meV to 102.6 (188.0) meV. The A exciton shifting more than the B exciton indicates that the A exciton-related peak has become more trion based.
When the MoS 2 flake is encapsulated with hBN, A and B PL peaks are blueshifted by ∼21 meV and 24 meV, respectively. When the hBN encapsulated MoS 2 is covered with ALD TiO 2 , the peaks redshift ∼32.7 and 40.1 meV, which is a stronger shift compared to the bare MoS 2 flake. The shift is more notable for the B PL peak. Additionally, this shift is slightly higher than the blueshift caused by hBN, giving the peaks' final center lower energy than the original bare MoS 2 . The blueshift caused by hBN covering is most likely coming from reduced oxygen and water doping arising from ambient conditions [22,57]. This, again, makes the PL more neutral exciton based. For uniaxial strain, ∼32-65 meV redshift of the A exciton PL peak per strain percentage has been demonstrated [37,39,40], and for biaxial, the shift effect can be ∼2.3 times higher [58]. Therefore, for the estimated strain of ∼0.1%, the calculated shift can be as high as ∼15 meV, which is roughly half of the observed shift for the ALD-coated heterostructure. The remaining shift arises presumably from the dielectric screening effect.
The measurement data of the Raman scattering and the Lorentzian fit for A 1g and E 1 2g modes of the MoS 2 flakes before and after ALD are presented in figure 4(c). Figure 4( [41]. With the acquired values, we can estimate the strain to be roughly ∼0.1%, which agrees well with our estimate. The A 1g Raman peak is downshifted when TiO 2 is added to the bare MoS 2 flake but remains basically unchanged for the hBN-covered TiO 2 . This shift is presumably caused by doping, as the A 1g is quite susceptible to it [32]. This is further supported by the widening of the A 1g peak that has also been incorporated with doping. Unlike the A 1g peak, E 1 2g peak is basically unaffected by doping but, on the other hand, typically shifts more than A 1g when strain (whether uni or biaxial) is applied [41]. Therefore, as the A 1g peak shifts more than E 1 2g , it seems that the resulting shift arises from both factors. The hBN-induced shift of A 1g peak is similar to the previously reported that has been assigned to interlayer van der Waals interactions [35,59].

Conclusions
We study the effects of ALD TiO 2 on the optical properties of WSe 2 and MoS 2 . A significant reduction of PL intensity is observed, especially for WSe 2 . This intensity decrease can be mainly pointed to the absorption and added reflection interfaces of the ALD TiO 2 films. Deposition of the TiO 2 film also introduces chemical doping and defects, which reduces the intensity even further. The film also introduces a dielectric screening effect that has an impact, especially on the PL of MoS 2 . An addition of the hBN layer reduces the doping-based quenching of PL intensity but also introduces its own dielectric screening effect while also slightly negating the one resulting from the ALD TiO 2 film. From the observed shifts and narrowing of the PL signal in hBN-covered samples, we can estimate our TiO 2 film to induce biaxial strain of ∼0.1%. Our results give more insights into the effects of the typically used processes in the fabrication of 2D material-based devices.

Data availability statements
The data cannot be made publicly available upon publication because they are not available in a format that is sufficiently accessible or reusable by other researchers. The data that support the findings of this study are available upon reasonable request from the authors.