hBN Encapsulation Effects on the Phonon Modes of MoS2 with a Thickness of 1 to 10 Layers

Interfaces with surrounding materials, where charged impurities and surface roughness are present, have a significant impact on the electrical and optical properties of 2D materials. In the change of the phonon modes of MoS2 accompanied by thickness variation, the portion caused by intrinsic factors and the portion caused by the interface effect are separated by examining the result of encapsulation with hexagonal boron nitride (hBN). For instance, the frequency of the A1g peak of MoS2 supported by SiO2 decreases by ≈4 cm−1 in air for a thickness reduction from ten layers to monolayer. Of this decrease, roughly 2 cm−1 is attributable to the weakening of the van der Waals interlayer interaction, while the remaining 2 cm−1 is due to the interface effect. The interface state, that is, the types and concentrations of impurities at the interface, between MoS2 and SiO2 is estimated to be similar to that between MoS2 and air because the Raman properties when one surface of MoS2 is in contact with SiO2 and with air are identical within the measurement error. When entirely encapsulated with hBN, the width of the A1g peak of few‐layer MoS2 is significantly reduced, becoming comparable or equal to that of bulk MoS2.


Introduction
In contrast to bulk semiconductors, the surfaces of 2D semiconductor materials composed of one or few atomic layers are atomically flat and free of dangling bonds, and thus can have superior electrical properties such as high mobility and a high on-off current ratio. [1][2][3] The interfaces with the surrounding materials as well as the intrinsic properties of the semiconductor material have a sensitive impact on the electrical and optical properties of 2D semiconductors or van der Waals (vdW) heterostructures based on them because many of the atoms that make up the materials are exposed to the surroundings. [4][5][6] The properties of 2D materials, such as graphene and transition metal dichalcogenides (TMDCs), have mostly been measured using flakes deposited on top of SiO 2 -coated Si substrate (SiO 2 /Si). Consequently, measurements of carrier mobility in 2D materials are significantly lower than theoretical predictions due to interaction with remote phonons of the SiO 2 as well as scattering by many rough surfaces and charged impurities at the interface between the 2D materials and SiO 2 . [1,[7][8][9] Hexagonal boron nitride (hBN), a 2D material with an electronic band gap of around 6.8 eV and free of dangling bonds, forms vdW interfaces that are atomically flat and impurity-free with other 2D materials, and thus hBN is expected to be an insulating material that can be used as a gating/tunneling layer in fabricating stable and high-performing 2D devices as well as to measure the intrinsic properties of 2D materials. [10][11][12][13][14] For instance, the field-effect mobility of SiO 2 -supported trilayer (3L) MoS 2 was found to be 7 cm 2 V −1 s −1 at room temperature, while that of 3L-MoS 2 encapsulated by hBN improved to 69 cm 2 V −1 s −1 with no hysteresis observed in the transfer curve, indicating that the interface between the two materials was free of charged impurities. [12] Moreover, Knobloch et al. predicted that the thickness of hBN should be more than 3.3 nm in order to effectively screen the effects of Coulomb impurities trapped at the interface with SiO 2 on the properties of TMDCs. [11] Interfaces have considerable effects on the optical properties of 2D materials as well. Lee et al. showed that the Raman and photoluminescence properties of hBN-encapsulated MoS 2 remained steady for 8 months in air. [12] Compared to SiO 2 -supported 1L-MoS 2 , previous works have shown that 1L-MoS 2 on Interfaces with surrounding materials, where charged impurities and surface roughness are present, have a significant impact on the electrical and optical properties of 2D materials. In the change of the phonon modes of MoS 2 accompanied by thickness variation, the portion caused by intrinsic factors and the portion caused by the interface effect are separated by examining the result of encapsulation with hexagonal boron nitride (hBN). For instance, the frequency of the A 1g peak of MoS 2 supported by SiO 2 decreases by ≈4 cm −1 in air for a thickness reduction from ten layers to monolayer. Of this decrease, roughly 2 cm −1 is attributable to the weakening of the van der Waals interlayer interaction, while the remaining 2 cm −1 is due to the interface effect. The interface state, that is, the types and concentrations of impurities at the interface, between MoS 2 and SiO 2 is estimated to be similar to that between MoS 2 and air because the Raman properties when one surface of MoS 2 is in contact with SiO 2 and with air are identical within the measurement error. When entirely encapsulated with hBN, the width of the A 1g peak of few-layer MoS 2 is significantly reduced, becoming comparable or equal to that of bulk MoS 2 .
hBN has enhanced tolerance to heat due to improved interfacial binding as well as homogeneous Raman and photoluminescence properties over the entire flake, including the edges with many dangling bonds. [15,16] In the present work, we isolated the interface effect on the Raman peaks of MoS 2 at varying thicknesses by closely examining the effects of hBN encapsulation on the Raman properties of MoS 2 with a thickness of 1L-10L. In addition to providing a means to study the intrinsic properties of various 2D materials by minimizing the interface effect, the findings of this work are expected to offer a diagnostic tool for exploring the status of various interfaces existing in heterostructure devices based on those materials.

Results and Discussion
The frequencies and widths of the Raman peaks of TMDCs are sensitively influenced by the local environment, such as polymer/solvent residues, defect density, impurities trapped at the interface, strain and other factors. [5,16,17] Therefore, even for flakes of the same thickness, the Raman spectrum may differ depending on the particular flake or the measurement point. Both the frequency and width of the A 1g peak of MoS 2 , in particular, fluctuate significantly depending on the interface state at a thickness of 4L or less.
In this work, measurement was carried out in the following manner in order to minimize the sample dependence of the Raman spectrum. As the first sample, we deposited MoS 2 flakes with a thickness of 1L to 6L on SiO 2 and partially encap-  Figures S1 and S2, Supporting Information). The same MoS 2 flakes were utilized to minimize the effects of sample dependence and to ascertain the hBN encapsulation effect. Bottom-and top-hBN are thick enough-at least 3.7 nm-to effectively screen out the effects of Coulomb impurities at the interfaces with air and SiO 2 . [11] The Raman spectra of MoS 2 flakes with a thickness of 1L-4L are displayed in Figure 2. Two first-order Raman modes of MoS 2 are observed: the in-plane E 2g 1 peak between 380 and 390 cm −1 , and the out-of-plane A 1g peak between 400 and 410 cm −1 . The Raman strength of hBN-supported MoS 2 is significantly reduced compared to that of SiO 2 -supported MoS 2 due to the destructive interference of both excitation and Raman scattered light within the bottom-hBN layer. [5,16] In particular, 1L-and 2L-MoS 2 show a more pronounced reduction in Raman peak intensity. As more MoS 2 surfaces are encapsulated with hBN, we notice that the A 1g peak further blueshifts and narrows in width.
All Raman spectra were analyzed by fitting them with Lorentzian functions ( Figure S3, Supporting Information); the results are summarized in Figure 3. Encapsulation with hBN, which forms a flat and impurity-free interface with MoS 2 , causes noticeable changes in both the frequency and width of the A 1g peak. However, the E 2g 1 peak exhibits only a slight reduction in width and little change in frequency. The A 1g peak frequency and width change more noticeably when both sides of MoS 2 are encapsulated with hBN than when only one side is encapsulated up to approximately 6L. We note that our results can be used to measure the number of layers of hBNencapsulated MoS 2 . The hBN encapsulation effect almost completely vanishes as the MoS 2 gets thicker. This tendency is well aligned with the general properties of MoS 2 in terms of thickness change; its 2D properties are clearly discernible up to 6L, but as it thickens further, bulk properties take control. [18,19]  www.advmatinterfaces.de Figure 3a,b,d,e shows that the frequency and width of the A 1g and E 2g 1 peaks of hBN/MoS 2 /SiO 2 and air/MoS 2 /hBN at each thickness are identical within the measurement error. In other words, MoS 2 exhibits the same phonon properties whether one side is in contact with SiO 2 or with air, which is consistent with the finding of Lee et al. that both SiO 2 -supported MoS 2 and air-suspended MoS 2 exhibit the same Raman properties under laboratory conditions. [20] It is expected that an atmosphere-like molecular composition is transferred onto the SiO 2 surface as oxygen, hydrogen, and water molecules from the atmosphere become trapped by the numerous dangling bonds on the SiO 2 surface. As a result, it is believed that the concentration and distribution of charged impurities at the interface between MoS 2 and hBN are approximately identical to those at the interface between MoS 2 and air. Figure 3 indicates that although the optical phonons and surface roughness of SiO 2 reduce the mobility of MoS 2 through scattering with carriers, [12,[21][22][23][24] their effects on the phonon properties of MoS 2 are negligible.
The A 1g peak of MoS 2 blueshifts and its width reduces by hBN encapsulation. For 1L-MoS 2 , compared to that of air/MoS 2 / SiO 2 , the frequency of the A 1g peak of both hBN/MoS 2 /SiO 2 and air/MoS 2 /hBN increases by ≈1.2 cm −1 and its full width at half maximum (FWHM) decreases by 0.8 cm −1 , whereas hBN/MoS 2 /hBN shows an increase in the frequency of the A 1g peak of 2.1 cm −1 and a decrease in the FWHM of 1.0 cm −1 . Below 6 layers, the width of the A 1g peak of hBN/MoS 2 /hBN reduces significantly, approaching or equaling that of bulk MoS 2 . However, even when entirely encapsulated with hBN, the frequency of the E 2g 1 peak remains constant, and only its width finely decreases in the 2D region as shown in Figure 3c,f.
Encapsulation with hBN causes changes in the electron concentration and interface state of MoS 2 . It has been reported that MoS 2 on SiO 2 is doped with electrons due to charged impurities inhomogeneously trapped at the interface. [25] In addition, previous electrical and optical experiments have shown that MoS 2 on hBN is electron depleted and much more uniform than when placed on SiO 2 due to the significantly reduced charged impurities at the interface. [5,9,15,16] Strain may also be applied to MoS 2 during the hBN encapsulation procedure. [26] In this case, the frequency and width of the A 1g peak exhibit little change, but the E 2g 1 peak is reported to redshift, broaden and then split into two peaks under intense strain because in-plane symmetry is broken, [17,27] which is contrary to the results shown here in Figure 3, so excluding the strain effect.
The A 1g mode of MoS 2 , in which only sulfur (S) atoms vibrate out-of-plane, and the E 2g 1 mode, in which both molybdenum (Mo) and S atoms vibrate in-plane, exhibit quite different dependence on thickness and doping. First, as vdW interlayer interaction increases with more layers, the A 1g peak blueshifts and becomes saturated due to an increase in the effective restoring forces acting on S atoms. However, the effect of the short-range vdW interlayer interaction is overtaken by the reduction in long-range Coulomb interaction, especially for the Mo atoms, which causes the E 2g 1 peak redshift. [28] Second, Chakraborty et al. found that at an electron doping of 1.8 × 10 13 cm −2 , the A 1g peak of 1L-MoS 2 was redshifted by 4 cm −1 and its FWHM was broadened by 6 cm −1 due to the increased electron-phonon interactions; however, the E 2g 1 peak did not show a noticeable change because of the weak dependence of electron-phonon interactions on doping concentration. [29] We note that the chemical doping of graphene with AuCl 3 was shown to result in a blueshift of the G-mode of 1L-and 2L-graphene as the doping concentration increased, whereas that of multilayer graphene remained unchanged. [30] This behavior is consistent with the observation that 2D carrier systems exhibit electrical properties that differ from bulk transport behavior because they are sensitive to extrinsic scattering mechanisms, such as charged impurities trapped at the interfaces. [9]

www.advmatinterfaces.de
For a thickness change from 10L to 1L, the A 1g peak of hBN/ MoS 2 /hBN redshifts by around 2 cm −1 in Figure 3c, which is consistent with a theoretical calculation of the frequency change of the A 1g peak due to the weakening of the vdW interlayer interaction. [28] Figure 3c strongly indicates that the main factor modifying the frequency of the A 1g peak of electrondepleted hBN/MoS 2 /hBN is the weakening of vdW interlayer interaction brought on by the reduction in thickness. This is consistent with the results of Figure 3f, which show that the FWHM of the A 1g peak does not change much over the thickness of 1L to 10L because the interface has little effect. Air/ MoS 2 /SiO 2 exhibits a softening of the A 1g peak of ≈4 cm −1 for a thickness reduction from 10L to 1L, of which 2 cm −1 is due to a decrease in vdW interlayer interaction and the remaining 2 cm −1 is attributable to the electron doping due to charged impurities at the interface. Contrary to air/MoS 2 /SiO 2 , hBN/ MoS 2 /hBN exhibits less fluctuation in the frequency of the A 1g peak depending on the measurement point due to the little effect of the inhomogeneous interface state. Figure 3 shows that hBN encapsulation significantly reduces both the average FWHM and its fluctuation of the A 1g peak at each thickness below 6L. The width of a Raman peak is given by the energy-time uncertainty principle, E  · /2 τ ∆ ≥ , where τ is the effective lifetime of a phonon and can vary in various parts of a sample due to inhomogeneity. [31] In the case of air/MoS 2 / SiO 2 , both electrons and the Coulomb potentials of charged impurities that are randomly trapped at both interfaces scatter with phonons, resulting in a shorter average lifetime, a larger FWHM, and marginally different values for each measurement point. However, in the case of the electron-depleted hBN/ MoS 2 /hBN, the scattering of phonons by electrons and Coulomb potentials is suppressed because the effect of the charged impurities at the interfaces almost disappears. As a result, the average FWHM of the A 1g peak and the fluctuation depending on the measurement point are both reduced. But as shown in Figure 3f, the FWHM and fluctuation of the A 1g peak for MoS 2 with a thickness of 1L-3L are slightly larger than those of the bulk, so it is considered that there is still a very small amount of charged impurity at the interfaces between hBN and MoS 2 . We believe that the FWHM of the A 1g peak from 1L-MoS 2 to bulk MoS 2 will remain constant if the impurities at the interface between hBN and MoS 2 are entirely eliminated. Although the frequency of the E 2g 1 peak is unaffected by hBN encapsulation, its FWHM is somewhat reduced for MoS 2 with a thickness of a few layers, indicating that the E 2g 1 peak also has a weak interaction with the charged impurities at the interface. The lifetimes of the E 2g 1 and A 1g phonons of bulk MoS 2 obtained using the energy-time uncertainty principle are τ(E 2g 1 ) ≥ 1.9 and τ(A 1g ) ≥ 1.3 ps, respectively.
We showed above that, despite the presence of trace impurities between hBN and MoS 2 , the intrinsic effects essentially govern the change in the frequency and width of the phonon  Figure 4. At a thickness of less than approximately 6L, the frequency of the A 1g peak decreases, its width widens, and the gap between the A 1g and E 2g 1 peaks narrows by the interface effect. It is clear that the interface effect is larger when MoS 2 is in contact with SiO 2 or air on both sides as compared to just one side.
As a reliable "thickness indicator" of MoS 2 flakes over a wider range over 10 layers, Li et al. proposed using the intensity ratio of the Raman peaks of MoS 2 to that of the underlying Si substrate. [18]  and sample thickness is not established for the hBN-supported samples, because the bottom-hBN layer between MoS 2 and Si has different interference effects for the Raman scattered light of MoS 2 and Si. [5,18] In other words, I I / MoS S i 2 can be used as a "thickness indicator" only when MoS 2 is in direct contact with SiO 2 /Si.

Conclusion
We examined the effects of hBN encapsulation on the phonon modes of MoS 2 flakes with a thickness of 1L−10L. The interface state between MoS 2 and SiO 2 is estimated to be quite similar to that between MoS 2 and air because the Raman properties when  www.advmatinterfaces.de one surface of MoS 2 is in contact with SiO 2 or air are identical within the measurement error. Both the vdW interlayer interaction and the interface modify the frequency and width of the A 1g peak of MoS 2 as the number of layers changes. When entirely encapsulated with hBN, the width of the A 1g peak of few-layer MoS 2 is significantly reduced, becoming comparable or equal to that of bulk MoS 2 , indicating that the interface effect almost completely disappears. In the change of the MoS 2 phonon modes accompanied by a thickness change, the portions caused by the vdW interlayer interaction and by the interface are isolated successfully. We have also shown that the ratio of the Raman intensity of MoS 2 to that of the underlying Si substrate can be used as a reliable "thickness indicator" only when MoS 2 is in direct contact with SiO 2 /Si, but not if other layers are interposed between them.

Experimental Section
MoS 2 or hBN flakes were mechanically exfoliated from their corresponding single crystals on polydimethylsiloxane (PDMS) films (Gel-Pak) and dry-transferred onto Si substrates capped with a 90 nm thick SiO 2 layer. The samples were then annealed at 180 °C for 2 min to improve the adhesion between the flakes and substrates. Some MoS 2 or hBN flakes exfoliated on PDMS films were transferred onto pre-located hBN or MoS 2 flakes on SiO 2 utilizing a dry transfer method in order to obtain heterostructures of MoS 2 and hBN. For full encapsulation of MoS 2 , top-hBN flakes were positioned on stacks of MoS 2 /hBN. Finally, the samples were annealed at 180 °C for 2 min. In this work, the entire sample preparation procedure was carried out in laboratory conditions. The thickness of the flakes was measured using contact mode AFM topography images and Raman spectra. Micro-Raman spectra were collected under laboratory conditions in backscattering geometry. A 532 nm laser line from a diode pumped solid state laser (Cobolt, Samba) was used as an excitation light. Scattered light was analyzed using a Horiba Jobin Yvon LabRAM HR spectrometer equipped with a grating of 1800 grooves/mm and a cooled chargecoupled device of 1024 × 256 pixels. The FWHM of the laser spot focused with a long working distance 100× objective lens (numerical aperture = 0.8) was ≈0.5 µm, and the laser power on the samples was kept at ≈0.3 mW to prevent laser irradiation effects. [32] The effect of sample heating was negligible as the Raman spectrum remained unchanged even after multiple measurements at the same point. Raman spectra between 100 and 600 cm −1 were taken at the same time, that is, the Raman peaks of the MoS 2 and the underlying Si were measured simultaneously. The T 2g peak of Si at 520.7 cm −1 was used as an internal reference to calibrate the frequencies of the MoS 2 Raman peaks. All Raman spectra were measured under the same conditions.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.