In situ exploration of the thermodynamic evolution properties in the type II interface from the WSe₂–WS₂ lateral heterojunction

The 2D monolayer WSe₂–WS₂ heterojunction was grown by a two-step epitaxial chemical vapor deposition method on an Si/SiO₂ substrate. Two quartz boats with WSe₂ and WS₂ powder were loaded into a horizontal tube furnace with a quartz tube. A clean Si/SiO₂ substrate was placed at the downstream end of the furnace as the growth substrate. The system was pump-purged with argon (Ar) gas and then ramped to the desired growth temperature for the growth of the WS₂ nanosheets at 1150 °C for 20 min under ambient pressure and a constant flow of 300 sccm Ar worked as the carrier gas. Then the boat with WS₂ powder was pushed out of the hot zone without breaking the vacuum, while the boat with WSe₂ powder was simultaneously pushed into the hot zone of the tube in situ. The temperature was then ramped to 1190 °C for the lateral epitaxial growth of the WSe₂ for about 20 min under 500 sccm Ar flow. Then the growth was terminated by shutting off the power of the furnace, and the sample was naturally cooled down to ambient temperature.

The Raman scattering and photolumiscence (PL) measurements were implemented on a confocal microscope system (Jobin-Yvon LabRAM HR Evolution spectrometer) with a 532 nm excited laser. A grating of 1800 grooves mm⁻¹ for Raman and 600 grooves mm⁻¹ grating for PL was used to collect the spectral data. A THMSE 600 heating/cooling stage (Linkam Scientific Instruments) was used in the temperature range from 305 K–405 K with a resolution of 0.1 K. The topography and surface potential images of the sample were simultaneously obtained from the KPFM technique (Bruker, Dimension Icon SPM) in LiftMode, with dual-pass scanning. To avoid oxidation of the sample during the heating process, we performed all of the above in a dry N₂ atmosphere. Further, to arrive at the thermal equilibrium state of the charge transferring across the interface of the heterojunction, the temperature ascending speed was controlled at a rate of 5 K min⁻¹ and temperature plateaus of 60 min. The tip bias was adjusted to cancel out the capacitive force by the KPFM system, according to the relation ϕ_sample = ϕ_tip  − eCPD, where CPD represents the contact potential difference between the tip and sample and e is the elementary charge, respectively. The parameters ϕ_sample and ϕ_tip are the workfunctions of the sample and tip, respectively. The workfunction of the Pt/Ir-coated tip (ϕ_tip) was calibrated to be 4.48 eV using a freshly exfoliated surface of highly oriented pyrolytic graphite (HOPG), of which the workfunction (ϕ_HOPG) is assumed to be 4.47 eV [21]. The dimension heater/cooler could heat the sample from 300 K to 405 K with an accuracy of 1 K.

Figure 1 shows the spatial distribution structures and optical properties of the planar WSe₂–WS₂ lateral heterojunction. The optical contrast between the inner and outer triangle shows the core–shell structure of the WSe₂–WS₂ lateral heterojunction. The AFM image (85 × 85 μm²) in figure 1(b) confirms that the WSe₂–WS₂ heterojunction is monolayer, with a thickness of approximately 0.8 nm for the shell region and 0.6 nm for the core region, respectively. Figures 1(c)–(e) show the Raman images that map at 257.2 cm⁻¹ corresponding to the ${{\rm{A}}}_{1g}$ resonant mode of WSe₂ and 338.1 cm⁻¹ corresponding to the E¹_2g resonant mode of WS₂, respectively. PL intensity maps at 1.69 eV and 1.87 eV are further applied to characterize the distinct core–shell structure. We can clearly see that the PL peak around 1.87 eV is mainly localized at the inner layer (WS₂) while 1.69 eV for the outer layer (WSe₂). From figures 1(c)–(h), the spatial composition distribution within the WSe₂–WS₂ heterojunction further verifies that the monolayer WSe₂ periphery region is laterally epitaxially connected with the edge of the monolayer WS₂ core region.

Zoom In Zoom Out Reset image size

Figure 2(a) shows the PL spectra taken from the point labeled in the inset of figure 1(a). The band structures at different positions in the heterojunction are distinct in the Gaussian fitting. The photon-excited electron–hole pairs will separate and relax to the conduction band edge of WSe₂ and the valence band edge of WS₂, as shown in the schematic diagram of figure 3(a). The distinct peaks at 1.68 and 1.87 eV correspond to the interband excitation for the pristine WSe₂ and WS₂ with a direct band gap, respectively. There is a gradual blueshift of the PL exciton position of pristine WSe₂ closing to the interface. Considering that the PL intensity of WS₂ is stronger than that of WSe₂, this behavior in WS₂ is negligible. The profile shows that the exciton energy at the interface is approximately 60 meV higher than that of the pristine WSe₂ district. However, the emission energy of pristine WS₂ is approximately 8 meV higher than that of the interface region. Previous works have demonstrated that photon–electron interactions in semiconductor materials can be seriously affected by chalcogen composition [22–24]. The optical properties of 2D TMDC alloy nanosheets, such as the band gap or the Raman frequency modes, can be varied linearly with stoichiometric compounds of a tunable sulfur ratio [23, 24]. Here, the prominent deviation of the PL emission energy between the pure and interface region in the WSe₂–WS₂ heterojunction is presumably due to the mixed alloying of the chemical composition around the interface region. In other words, the S atoms are believed to coexist with Se atoms with a gradually changing stoichiometric ratio from pure WSe₂ or WS₂ to the alloying interface. Moreover, figure S1 (available online at stacks.iop.org/NANO/29/435703/mmedia) depicts the Raman spectra of the sample, which shows the peak positions shifting from pure WSe₂ or WS₂ to the interface region when the value of the ${\rm{\Delta }}{\omega }_{{A}_{1g}}$ mode is 3.07 cm⁻¹ for WSe₂ and the ${\rm{\Delta }}{\omega }_{{E}_{2g}}$ mode is 3.44 cm⁻¹, and the ${\rm{\Delta }}{\omega }_{{A}_{1g}}$ mode is 1.99 cm⁻¹ for WS₂, respectively. Therefore, the shifting behavior in both the PL and Raman spectra can confirm the hypothesis of the existence of a chemical composition transition region around the interface.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The detailed temperature evolution of the electron transition for the covalently bonded interface in the WSe₂–WS₂ heterojunction is shown in figure 2(b). The photon energy of WSe₂ (P${}_{{{WSe}}_{2}}$), WS₂ (P${}_{{{WS}}_{2}}$) is fitted with the integrated intensity peak, on purpose to describe the intensity variation trend with temperature. Figure 2(b) displays the luminescence intensity evolution of the heterointerface as a function of temperature from 305 K–405 K with a stability of about 0.1 K. The PL spectra can be deconvoluted into two major peaks. Along with the variation of the luminescent intensity, the direct transition in the heterostructures experiences a gradual redshift as the temperature increases, as shown in figure 2(c). Such behavior is similar to the response of conventional semiconductors under high temperature, which can be attributed to increased phonon–exciton interactions and slight changes in bonding lengths. Thus, it provides a route to evaluate temperature correlations to the semiconductors. The temperature dependence of the PL peak position is fitted by employing the modified Varshni relationship:

$$\begin{eqnarray}&&E(T)={E}_{0}-S\langle {\hslash }{\rm{\Delta }}\rangle \left[\cot h\displaystyle \frac{\langle {\hslash }{\rm{\Delta }}\rangle }{2{k}_{B}T}-1\right]\end{eqnarray} \tag{ 1 }$$

where E₀ is the emission energy at zero absolute temperature, S is the Huang–Rhys factor which can reflect the coupling strength between the exciton and phonon, $\langle {\hslash }{\rm{\Delta }}\rangle$ is the average phonon energy, and ℏ and k_B are the Plancks and Boltzmann constant, respectively. The fitting values of E₀, S, and $\langle {\hslash }{\rm{\Delta }}\rangle$ of WSe₂, WS₂ are extracted in table 1. By comparing the fitting parameters of the TMDCs at the interface, such as S, the contribution of the interface coupling can be informed. In addition, the temperature difference between WSe₂, WS₂ at the interface can also be derived by comparing the measured difference between the emission PL energy. The variation with temperature of the peak ${{\rm{P}}}_{{{WSe}}_{2}}$ and ${{\rm{P}}}_{{{WS}}_{2}}$ at the interface do not disply a redshift as obvious as in the pristine WSe₂, WS₂ monolayer, which is a reflection of the strong coupling interaction between the exciton and phonon at the covalently bonded interface. Furthermore, the intensity ratio (I${}_{{{WS}}_{2}}$/I${}_{{{WSe}}_{2}}$) has been plotted to reflect the competing evolution of the exciton quantities between WSe₂ and WS₂ in variable temperature processes. The decrease of the intensity ratios, as shown in figure 2(d), also represents the direct–indirect spectral-weight conversion with the increasing temperature.

Table 1.  The parameter values of the modified Varshni model for peak energy at the interface as a function of temperature in figure 2(c).

Sample	Exciton	E0 (eV)	S	$\langle {\hslash }{\rm{\Delta }}\rangle$ (meV)
WSe2–WS2	P${}_{{{WS}}_{2}}$	1.96	3.43	8.13
Interface	P${}_{{{WSe}}_{2}}$	1.74	2.78	6.23

Figure 3(a) schematically depicts the type II band structures of the covalently bonded interface. From the surface potential image in figure 3(b), the workfunction of the shell WSe₂ is lower than that of the core WS₂. The discriminating difference between the Fermi levels of these two materials, which forms the built-in potential barrier, can facilitate the photo-excited electron–hole separation and verify the type II band alignment with an atomically sharp heterointerface. Neglecting the hybridization of electronic states in the WSe₂ and WS₂ layers, the lowest conduction band resides in WSe₂ and the highest valence band resides in WS₂, respectively [12, 16, 18, 20, 25]. Likewise, in the WSe₂–WS₂ heterostructure, due to type-II band alignment, photo-excited electrons and holes will relax (dashed lines in figure 3(a)) to the conduction band edge of WSe₂ and the valence band edge of WS₂, respectively.

As shown in figure 3(c), the value of the surface potential is averaged along the y-axis direction in the bottom panel of figure 3(b) under the thermal equilibrium condition, from which we can conclude that the WSe₂–WS₂ heterojunction serves as a quasi planar n–p junction, where the WSe₂ works for the n-side and WS₂ for the p-side. Treating the surface potential distribution with a sigmoid fitting function (S-fitting) below: [26, 27]

$$\begin{eqnarray}&&y=\displaystyle \frac{{A}_{1}-{A}_{2}}{1+{e}^{(x-{x}_{0})/{dx}}}+{A}_{2}.\end{eqnarray} \tag{ 2 }$$

Here, A₁ = 0.281 and A₂ = −1.433; dx = 1.083 and x₀ = 9.242 μm form the frontier between the quasi p-type region and n-type region, respectively. The distinction between the inherent Fermi levels of WSe₂ and WS₂ (△E_F) can be calculated by estimating the CPD difference between the WSe₂ and WS₂ sides (△CPD${}_{{{WSe}}_{2}-{{WS}}_{2}})$:

$$\begin{eqnarray}&&{\rm{\Delta }}{E}_{F}={E}_{{F}_{{{WSe}}_{2}}}-{E}_{{F}_{{{WS}}_{2}}}={\phi }_{{{WS}}_{2}}-{\phi }_{{{WSe}}_{2}}=e{\rm{\Delta }}{{CPD}}_{{{WSe}}_{2}-{{WS}}_{2}}.\end{eqnarray} \tag{ 3 }$$

Together with the S-fitting results, we can evaluate the value △E_F and the depletion region width to be approximately 1.73 eV and 11.67 μm, respectively. Meanwhile, the potential drop at the WSe₂/SiO₂ interface can be attributed to the abrupt height change, or the work function differences between semiconductor WSe₂ and insulator SiO₂ [28].

Furthermore, by differentiating the sigmoid fitting profile, we obtain the built-in electric field distribution, as shown in figure 3(d). The formed built-in electric field locates at the interface zone that impedes the dynamic diffusion of carriers. We use the below equations to define the built-in electric field distribution (ξ(x)) of the 2D abrupt junction: [29, 30]

$$\begin{eqnarray}&&{\xi }_{{{WSe}}_{2}}(x)\sim \displaystyle \frac{{{en}}_{D}(x-{W}_{n})}{{\epsilon }_{{{WSe}}_{2}}{\epsilon }_{0}{\rho }_{{{WSe}}_{2}}},(0\lt x\lt {W}_{n})\end{eqnarray} \tag{ 4 }$$

$$\begin{eqnarray}&&{\xi }_{{{WS}}_{2}}(x)\sim -\displaystyle \frac{{{en}}_{A}(x+{W}_{p})}{{\epsilon }_{{{WS}}_{2}}{\epsilon }_{0}{\rho }_{{{WS}}_{2}}},(-{W}_{p}\lt x\lt 0).\end{eqnarray} \tag{ 5 }$$

It is noteworthy that ${}_{{{WS}}_{2}}$ = 6.1 and ${}_{{{WSe}}_{2}}$ = 11.7 are the dielectric constant of monolayer WS₂ and WSe₂ [16, 19], respectively. The parameter ₀ is the vacuum permittivity. ρ${}_{{{WSe}}_{2}}$ and ρ${}_{{{WS}}_{2}}$ are the screening lengths, given by ρ${}_{{{WSe}}_{2}}$ = 2πχ${}_{{{WSe}}_{2}}$ and ρ${}_{{{WS}}_{2}}$ = 2πχ${}_{{{WS}}_{2}}$, where χ${}_{{{WSe}}_{2}}$ = 0.718 nm and χ${}_{{{WS}}_{2}}$ = 0.603 nm are polarizabilities of monolayer WSe₂ and WS₂, respectively [31]. The maximum electric field is estimated to be 32.52 V cm⁻¹, almost locating at the midpoint of the depletion region. Moreover, the amount of the intrinsic doping concentration is derived from the slope of the linear fitting of the electric field distribution. It can be concluded that n_D ≈ 2.4 × 10⁷ cm⁻² in n-type WSe₂ and n_A ≈ 1.04 × 10⁷ cm⁻² in p-type WS₂, respectively. This result is reasonable when compared with the intrinsic carrier concentrations (∼10¹⁰ cm⁻²) reported in previous work [1, 16, 19]. Among which, W_n and W_p are the depletion width in WSe₂ and WS₂, respectively. The total depletion width W = W_n + W_p, which is the varying part of the surface potential profile along the x-axis direction, can be obtained by the p–n junction modeling below: [19, 32]

$$\begin{eqnarray}&&{W}_{p}=\sqrt{\displaystyle \frac{2{N}_{D}{\epsilon }_{0}{\epsilon }_{n}{\epsilon }_{p}{V}_{{bi}}}{{{eN}}_{A}({\epsilon }_{n}{N}_{A}+{\epsilon }_{p}{N}_{D})}}\end{eqnarray} \tag{ 6 }$$

$$\begin{eqnarray}&&{W}_{n}=\displaystyle \frac{{W}_{p}{N}_{A}}{{N}_{D}}.\end{eqnarray} \tag{ 7 }$$

According to the relations N_D[A] = n_D[n_A]/t_n[p], t is the thickness of the monolayer in-plane heterojunction and can be read directly from the AFM image (t${}_{{{WSe}}_{2}}$ ≈ 0.8 nm, t${}_{{{WS}}_{2}}$ ≈ 0.6 nm). So N_D ≈ 2.99 × 10¹⁴ cm⁻³, N_A ≈ 1.73 × 10¹⁴ cm⁻³. In addition, the modeling predicts the depletion width W_n and W_p to be approximately 1.43 and 2.48 μm respectively, compared with the estimated value (W ∼ 11.67 μm). This implies that the theoretical p–n heterojunction modeling, which is based on the assumption of no free carriers and constant dopant concentration for the intentionally doped inorganic semiconductors with high-carrier mobility [32], is not completely suitable for describing electronic transport in the 2D covalently bonded heterojunction, and remains to be explored.

Considering the heating phenomenon in the practical integrated circuit of the photoelectric devices, we performed temperature-dependent (305 K–405 K) characterization utilizing the KPFM technique, which was conducted in a N₂ atmosphere. Figure 4(a) displays the surface potential mappings of the WSe₂–WS₂–WSe₂ interface at different temperatures. It distinctly shows that the surface potential of WSe₂ declines with temperature, and the WS₂ manifests an opposing tendency. The variation of the surface potential reflects the thermodynamic charges transferring across the WSe₂–WS₂–WSe₂ interface. An n–p to p–n junction transition emerges at the critical point of temperature (T_c, 365 ∼ T_c ∼ 375 K) [33]. For clarity, we plotted the horizontal profiles of the surface potential distribution at the WSe₂–WS₂–WSe₂ interface, as shown in figure 4(b). It is noteworthy that each of the curves is averaged along the y-axis direction in figure 4(a) at the corresponding temperature. A flat band condition is reached at T_c, where the value of the workfunctions from the WSe₂ and WS₂ sides stay very close, implying that a zero-barrier ohmic contact between WSe₂ and the WS₂ side is formed.

Zoom In Zoom Out Reset image size

To understand the intrinsic mechanism of the electronic states during the period of the temperature increase, the surface potential profiles of the WSe₂–WS₂ interface in the whole temperature range are plotted in figure 5(a). Each curve is fitted by the sigmoid function as discussed above. The closed Fermi levels of WSe₂ and WS₂ at T_c benefit the thermodynamic behaviors of the carriers traveling through the WSe₂–WS₂ interface. For a detailed view of the dynamic variation of the electric contact between WSe₂ and WS₂, the built-in potential barrier with temperature is derived in figure 5(b). Obviously, the positive value of the contact barrier tends to decline with an increase in temperature, indicating a negative temperature correlation of the built-in contact barrier. This shows a decrease in the contact resistance with an increase in the temperature of semiconductor devices, unlike the resistance in the semimetal devices of the 1T' phase [34]. The mostly weakened barrier (∼0 V) emerges, implying that the electric contact is effectively tailored at T_c. When the temperature is beyond T_c, the contact barrier between the WSe₂ and WS₂ side increases with a negative value. Consequently, the charge injection from WSe₂ to WS₂ is hindered, rendering a relatively opposing diffusion from WS₂ to WSe₂.

Zoom In Zoom Out Reset image size

Since the specific resistivity inside the semiconductors can be strongly affected by the doping level, we have explored the variation of the intrinsic carrier concentration, which is indirectly displayed by the workfunction profile. As shown in figure 5(c), the workfunction variation of WSe₂ and WS₂ with temperature is deduced. Note that the Fermi levels shift in the opposite direction with the workfunction; the profile shows that both the amount of carriers in the n-type WSe₂ and p-type WS₂ sides decrease with temperature. This phenomenon is inconsistent with the view that the higher the temperature, the greater the chance of the carriers being thermally excited. Bearing in mind that the environment, such as substrate and ambient, has a strong impact on the doping concentration of carriers [35], it can be concluded that the drop in the amount of carriers can thus be attributed to the electron-withdrawing effect induced by the Si/SiO₂ substrate of the WSe₂–WS₂ heterojunction. [9, 36].

Considering that the electric field poling is sensitive to temperature [37], the spatial distribution of the built-in electric field at different temperatures is obtained by differentiating the work function, as shown in figure 5(d). The pronounced transformation of the built-in electric field can be due to the thermal transferring behaviors that result in the change of electron location at the lattice, which is very susceptible to structural strain along the lattice distribution at the interface. The direction of the built-in electric field gradually reverses at T_c. This evolution of the built-in electric field with the increasing temperature process is assumed to be the difference in the thermal expansion coefficient (TEC) between the WSe₂ and WS₂ materials, which can induce a structural mismatch of the lattice [33]. Previous work has reported that the TEC mismatch would induce a strain effect in the 2D materials [38]. To verify the strain effect during the temperature varying process, the temperature-dependent Raman spectroscopy at the interface of the WSe₂–WS₂ lateral heterojunction is utilized, as shown in figure S2(a). The peak position behaves as a function of temperature for the frequency modes of WS₂ and WSe₂, as shown in figure S2(b)–(d), which we conclude to be the contribution from the thermal anharmonic effect and the thermal expansion [38–41]. In addition, there is a discrepancy in the extracted slope of the linear fitting from the Raman modes versus temperature, which is consistent with previous work [42]. These results accordingly confirm that the strain effect is a consequence of the difference between the thermal coefficient of these materials.

Based on equations (5) and (6), the depletion width at variable temperature is deduced in figure 5(f). The width shrinks till it reaches the minimum value at T_c. When the temperature is beyond the T_c, the width of the depletion region gradually increases, which indicates that the unit cells recover their cubic shape. We concluded that this is due to the thermal diffusion of the intrinsically excited carriers. Due to the fact that the generated holes and electrons move towards the opposite direction and approach the border of the depletion region, the depletion width declines.