Characteristics of p-Type Conduction in P-Doped MoS2 by Phosphorous Pentoxide during Chemical Vapor Deposition

We demonstrated p-type conduction in MoS2 grown with phosphorous pentoxide via chemical vapor deposition (CVD). Monolayer MoS2 with a triangular shape and 15-µm grains was confirmed by atomic force microscopy. The difference between the Raman signals of the A1g and E12g modes for both the pristine and P-doped samples was 19.4 cm−1. In the X-ray photoelectron spectroscopy results, the main core level peaks of P-doped MoS2 downshifted by about 0.5 eV to a lower binding energy compared to the pristine material. Field-effect transistors (FETs) fabricated with the P-doped monolayer MoS2 showed p-type conduction with a field-effect mobility of 0.023 cm2/V⋅s and an on/off current ratio of 103, while FETs with the pristine MoS2 showed n-type behavior with a field-effect mobility of 29.7 cm2/V⋅s and an on/off current ratio of 105. The carriers in the FET channel were identified as holes with a concentration of 1.01 × 1011 cm−2 in P-doped MoS2, while the pristine material had an electron concentration of 6.47 × 1011 cm−2.


Introduction
Recently, various studies have analyzed two-dimensional (2D) materials, such as graphene, MoS 2 , and WSe 2 , because of their critical properties and abundant potential for use in optical and electrical applications [1][2][3]. Graphene has a zero band gap structure, but has not been able to replace semiconductor-based devices [4,5]. Additionally, layered transition metal dichalcogenides (TMDs) such as MoS 2 and WSe 2 have received enormous attention as promising materials and layer structures, in which transition metals are sandwiched between two chalcogen atom layers by a covalent force. Moreover, there are Van der Waals (VdW) forces interacting in individual layers, which make exfoliation easily. Interestingly, these materials have a unique property; their band gap structure varies depending on the thickness. In the case of MoS 2 , the band gap of a monolayer has a direct band gap of 1.8 eV, while a few layers of MoS 2 and bulk MoS 2 have an indirect band gap structure with a band gap of about 1.2 eV [6].
The chemical vapor deposition (CVD) method has several advantages compared to other methods, such as mechanical and liquid exfoliation methods [7,8]. The disadvantages of the Scotch tape-based mechanical method are its difficulty in controlling the flake thickness, size, and uniformity, which makes it inappropriate for large-scale applications. The liquid method still needs to be developed for applications, while the CVD method has been used to prepare ultrathin monolayers or few-layer MoS 2 films over large areas [9]. Transistors have been fabricated via CVD growth of monolayer MoS 2 . These have been reported to exhibit good properties, including a high on/off current ratio and high mobility [10]. To realize detailed applications, this method needs to be able to produce a junction composed of n-and p-type materials. Although there have been many challenges to p-type doping of MoS 2 using niobium (Nb) or phosphorous (P) atoms [11,12], and it remains difficult to successfully dope ultrathin MoS 2 . According to a previous report, P atoms seem to be the most suitable acceptors among group V elements [13].
In this paper, we report on the CVD growth and characteristics of monolayer MoS 2 with and without the addition of phosphorous pentoxide (P 2 O 5 ) powder. The thickness and grain size of the MoS 2 layer were measured using non-contact-mode atomic force microscopy (AFM) and Raman spectroscopy. To confirm the electrical characteristics of MoS 2 , back-gated field-effect transistors (FETs) were fabricated. The p-type conduction from monolayer MoS 2 grown with P 2 O 5 powder was confirmed and compared to pristine MoS 2 with n-type behavior.

Experimental
To synthesize an MoS 2 layer by the CVD method, molybdenum trioxide (MoO 3 , CERAC Inc, Milwaukee, WI, USA) powder with 99.999% purity as a precursor material and sulfur (iTASCO Inc, Seoul, Korea) powder of 99.999% purity as a reactant material were used. For p-type doping of MoS 2 in this experiment, 98.99% purity P 2 O 5 (SIGMA-ALDRICH, St. Louis, MO, USA) powder was added as a dopant material. SiO 2 /Si substrates (2 × 2 cm 2 ) with a SiO 2 thickness of 270 nm and three alumina boats were used. The alumina boats were filled with 10 mg of MoO 3 powder, 300 mg of S powder, and 1 mg of P 2 O 5 powder, respectively. During CVD growth of MoS 2 , the furnace was heated to 750 • C with a heating rate of 30 • C/min under argon gas flowing at 100 sccm. The role of argon gas was to transport S and P 2 O 5 when they were vaporized. During the growth of MoS 2 , the gas flow and furnace temperature were kept constant for 30 min, and then the furnace was quickly cooled down to room temperature.
The MoS 2 thickness and grain size were analyzed by using non-contact-mode atomic force microscopy (AFM) (XE-100, Park's Systems, Seoul, Korea) and optical microscopy. X-ray photoelectron spectroscopy (XPS) (K-Alpha+, Thermo Fisher Scientific, Waltham, MA, USA) under~4 × 10 −10 Torr and Raman spectroscopy (NRS-3100, JASCO, Tokyo, Japan) with a λ = 532 nm laser were measured at room temperature to identify the doping characteristics. To confirm the electrical characteristics of doped monolayer MoS 2 , back-gated FETs were fabricated. In this process, photolithography was used for patterning source and drain electrodes of Ni/Au (5 nm/50 nm) metals. Figure 1a shows a simplified schematic diagram for the synthesis of MoS 2 , with and without phosphorus doping, using the CVD system. Here, MoO 3 powder was placed in the middle of the furnace, slightly away from the S and P 2 O 5 powders. The P 2 O 5 powder for P doping was located about 7.5 cm from the MoO 3 powder. The ratio of S to Mo atoms is an important point for growing monolayer MoS 2 flakes. We used a face-down substrate approach, where the SiO 2 substrate is positioned vertically facing the MoO 3 -containing alumina boat. Unlike previous doping studies that used a two-furnace system [14][15][16], this method used in situ doping with a one-furnace CVD system. The CVD process for MoS 2 growth can be divided into two steps: Nucleation and growth. Figure 1b shows the temperature profile of the reaction furnace and pressure variation in the quartz tube with Ar gas flow, respectively, as a function of time. In this figure, S atoms are introduced at 650 • C which is 100 • C lower than the growth temperature (750 • C). When S atoms are introduced, the growth of MoS 2 starts and then monolayer MoS 2 flakes appear [17].  Here, the MoS2 layers grown on the SiO2/Si substrate under a sufficient S atmosphere were observed to have a triangular shape [17,18]; this is the same shape as pristine MoS2. The grain size of doped MoS2 on the SiO2/Si substrate was about 15 µm. To confirm the formation of a monolayer of P-doped MoS2, Raman spectroscopy and AFM measurements were performed, as shown in Figure 2b,c, respectively. The thickness of an MoS2 flake measured by AFM was about 0.6 nm to 0.9 nm; this layer thickness is the same as a previous result [10]. This measurement value corresponds to the interlayer spacing of a monolayer of S-Mo-S bonding in the MoS2 crystal. Two characteristic Raman peaks, i.e., E 1 2g and A1g from in-plane and out-of-plane modes, respectively, were measured by a laser with an excitation wavelength of 532 nm at room temperature, as shown in Figure 2c. The in-plane E 1 2g mode presents the vibration of one Mo atom and two S atoms in opposite directions, while the out-of-plane A1g mode vibrates only S atoms in opposite directions (as shown in the inset of Figure 2c). From reported results that describe the dependence of the Raman peaks on the number of layers [19,20], we know that the difference between two Raman peaks depending on the number of MoS2 layers is larger than 20 cm −1 for thicknesses above a bilayer (2 L). As shown in Figure 2d, Raman peaks from P-doped MoS2 were located at 384.5 cm −1 (E 1 2g mode) and 403.9 cm −1 (A1g mode). On the other hand, the E 1 2g and A1g signals of the pristine monolayer MoS2 were observed at around 384.6 cm −1 and 405 cm −1 , respectively. The difference between the two Raman modes for P-doped and pristine MoS2 (Figure 2d) appear to be about 19.4 cm −1 and 20 cm −1 , respectively; these values indicate a single layer of MoS2.  Figure 2a shows an optical microscopy image of P-doped MoS 2 grown via CVD. Here, the MoS 2 layers grown on the SiO 2 /Si substrate under a sufficient S atmosphere were observed to have a triangular shape [17,18]; this is the same shape as pristine MoS 2 . The grain size of doped MoS 2 on the SiO 2 /Si substrate was about 15 µm. To confirm the formation of a monolayer of P-doped MoS 2 , Raman spectroscopy and AFM measurements were performed, as shown in Figure 2b,c, respectively. The thickness of an MoS 2 flake measured by AFM was about 0.6 nm to 0.9 nm; this layer thickness is the same as a previous result [10]. This measurement value corresponds to the interlayer spacing of a monolayer of S-Mo-S bonding in the MoS 2 crystal. Two characteristic Raman peaks, i.e., E 1 2g and A 1g from in-plane and out-of-plane modes, respectively, were measured by a laser with an excitation wavelength of 532 nm at room temperature, as shown in Figure 2c. The in-plane E 1 2g mode presents the vibration of one Mo atom and two S atoms in opposite directions, while the out-of-plane A 1g mode vibrates only S atoms in opposite directions (as shown in the inset of Figure 2c). From reported results that describe the dependence of the Raman peaks on the number of layers [19,20], we know that the difference between two Raman peaks depending on the number of MoS 2 layers is larger than 20 cm −1 for thicknesses above a bilayer (2 L). As shown in Figure 2d, Raman peaks from P-doped MoS 2 were located at 384.5 cm −1 (E 1 2g mode) and 403.9 cm −1 (A 1g mode). On the other hand, the E 1 2g and A 1g signals of the pristine monolayer MoS 2 were observed at around 384.6 cm −1 and 405 cm −1 , respectively. The difference between the two Raman modes for P-doped and pristine MoS 2 (Figure 2d  Here, the MoS2 layers grown on the SiO2/Si substrate under a sufficient S atmosphere were observed to have a triangular shape [17,18]; this is the same shape as pristine MoS2. The grain size of doped MoS2 on the SiO2/Si substrate was about 15 µm. To confirm the formation of a monolayer of P-doped MoS2, Raman spectroscopy and AFM measurements were performed, as shown in Figure 2b,c, respectively. The thickness of an MoS2 flake measured by AFM was about 0.6 nm to 0.9 nm; this layer thickness is the same as a previous result [10]. This measurement value corresponds to the interlayer spacing of a monolayer of S-Mo-S bonding in the MoS2 crystal. Two characteristic Raman peaks, i.e., E 1 2g and A1g from in-plane and out-of-plane modes, respectively, were measured by a laser with an excitation wavelength of 532 nm at room temperature, as shown in Figure 2c. The in-plane E 1 2g mode presents the vibration of one Mo atom and two S atoms in opposite directions, while the out-of-plane A1g mode vibrates only S atoms in opposite directions (as shown in the inset of Figure 2c). From reported results that describe the dependence of the Raman peaks on the number of layers [19,20], we know that the difference between two Raman peaks depending on the number of MoS2 layers is larger than 20 cm −1 for thicknesses above a bilayer (2 L). As shown in Figure 2d, Raman peaks from P-doped MoS2 were located at 384.5 cm −1 (E 1 2g mode) and 403.9 cm −1 (A1g mode). On the other hand, the E 1 2g and A1g signals of the pristine monolayer MoS2 were observed at around 384.6 cm −1 and 405 cm −1 , respectively. The difference between the two Raman modes for P-doped and pristine MoS2 (Figure 2d Here, the Raman signal peak of the A1g mode was found to be shifted by about 1.1 cm −1 , while the signal peak of the E 1 2g mode was almost unchanged. Azcatl et al. reported that a strain induced by dopants can generate contractions of the MoS2 lattice structure [21]; this phenomenon occurs due to the longer bond length of Mo-S atoms than that of Mo-P atoms. It was also reported that the A1g mode is often more influenced by doping effects than other modes (e.g., the E 1 2g mode); this is due to its strong coupling with electrons [22,23]. Therefore, the Raman active signal with A1g has a shift larger than the other Raman active signal because this peak of the Raman mode is quite sensitive to the doping effect. We confirmed that the Raman shifts in Figure 2d agreed with previous results [24]. The full width at half maximum (FHWM) of the E 1 2g peak was investigated to characterize the crystalline quality of MoS2 obtained by the CVD synthesis method. The FWHM result of the CVDgrown monolayer flake is 3.8 cm −1 , which is similar to a recently reported value of a CVD-synthesized single-layer flake [18].

Results
The energy peaks appearing in XPS were also analyzed to confirm the doping properties in monolayer MoS2 crystals. Figure 3a-c show the comparative XPS core level analyses of pristine and doped monolayer MoS2. In Figure 3a, the P 2p binding energy peak, which clearly appears only at 134.3 eV, is associated with a doped flake feature. It is worth mentioning that the existence of this peak provides apparent evidence that P2O5 takes its position before the introduction of S. In addition, the Mo 3d and S 2p core levels indicated that the phenomenon causes a uniform shift of 0.5 eV, from 229.6 eV to 229.1 eV and from 162.4 eV to 161.9 eV, respectively (Figure 3b,c). That is, each peak moved toward a lower binding energy after P-doping, which is very similar to the reported results for Nb-doped MoS2 [24]. This study reported that the Fermi level (EF) of pristine MoS2 is located close to the conduction band (Ec) edge, while an Nb-doped p-type sample has a Fermi level near the valence band edge. The work function and electron affinity of the pristine monolayer MoS2 is 5.1 eV and 4.28 eV, respectively [25]. The pristine MoS2 Fermi level is 0.82 eV, which is the Ec-EF result, and the doping sample Fermi level is 1.32 eV, 0.82 + ΔBE (measured from XPS data). Therefore, it is suggested that doping with P2O5 leads to a downshift in the Fermi level of about 0.5 eV, close to the valence band maximum. Here, the Raman signal peak of the A 1g mode was found to be shifted by about 1.1 cm −1 , while the signal peak of the E 1 2g mode was almost unchanged. Azcatl et al. reported that a strain induced by dopants can generate contractions of the MoS 2 lattice structure [21]; this phenomenon occurs due to the longer bond length of Mo-S atoms than that of Mo-P atoms. It was also reported that the A 1g mode is often more influenced by doping effects than other modes (e.g., the E 1 2g mode); this is due to its strong coupling with electrons [22,23]. Therefore, the Raman active signal with A 1g has a shift larger than the other Raman active signal because this peak of the Raman mode is quite sensitive to the doping effect. We confirmed that the Raman shifts in Figure 2d agreed with previous results [24]. The full width at half maximum (FHWM) of the E 1 2g peak was investigated to characterize the crystalline quality of MoS 2 obtained by the CVD synthesis method. The FWHM result of the CVD-grown monolayer flake is 3.8 cm −1 , which is similar to a recently reported value of a CVD-synthesized single-layer flake [18].
The energy peaks appearing in XPS were also analyzed to confirm the doping properties in monolayer MoS 2 crystals. Figure 3a-c show the comparative XPS core level analyses of pristine and doped monolayer MoS 2 . In Figure 3a, the P 2p binding energy peak, which clearly appears only at 134.3 eV, is associated with a doped flake feature. It is worth mentioning that the existence of this peak provides apparent evidence that P 2 O 5 takes its position before the introduction of S. In addition, the Mo 3d and S 2p core levels indicated that the phenomenon causes a uniform shift of 0.5 eV, from 229.6 eV to 229.1 eV and from 162.4 eV to 161.9 eV, respectively (Figure 3b,c). That is, each peak moved toward a lower binding energy after P-doping, which is very similar to the reported results for Nb-doped MoS 2 [24]. This study reported that the Fermi level (E F ) of pristine MoS 2 is located close to the conduction band (E c ) edge, while an Nb-doped p-type sample has a Fermi level near the valence band edge. The work function and electron affinity of the pristine monolayer MoS 2 is 5.1 eV and 4.28 eV, respectively [25]. The pristine MoS 2 Fermi level is 0.82 eV, which is the E c -E F result, and the doping sample Fermi level is 1.32 eV, 0.82 + ∆BE (measured from XPS data). Therefore, it is suggested that doping with P 2 O 5 leads to a downshift in the Fermi level of about 0.5 eV, close to the valence band maximum. Here, the Raman signal peak of the A1g mode was found to be shifted by about 1.1 cm −1 , while the signal peak of the E 1 2g mode was almost unchanged. Azcatl et al. reported that a strain induced by dopants can generate contractions of the MoS2 lattice structure [21]; this phenomenon occurs due to the longer bond length of Mo-S atoms than that of Mo-P atoms. It was also reported that the A1g mode is often more influenced by doping effects than other modes (e.g., the E 1 2g mode); this is due to its strong coupling with electrons [22,23]. Therefore, the Raman active signal with A1g has a shift larger than the other Raman active signal because this peak of the Raman mode is quite sensitive to the doping effect. We confirmed that the Raman shifts in Figure 2d agreed with previous results [24]. The full width at half maximum (FHWM) of the E 1 2g peak was investigated to characterize the crystalline quality of MoS2 obtained by the CVD synthesis method. The FWHM result of the CVDgrown monolayer flake is 3.8 cm −1 , which is similar to a recently reported value of a CVD-synthesized single-layer flake [18].
The energy peaks appearing in XPS were also analyzed to confirm the doping properties in monolayer MoS2 crystals. Figure 3a-c show the comparative XPS core level analyses of pristine and doped monolayer MoS2. In Figure 3a, the P 2p binding energy peak, which clearly appears only at 134.3 eV, is associated with a doped flake feature. It is worth mentioning that the existence of this peak provides apparent evidence that P2O5 takes its position before the introduction of S. In addition, the Mo 3d and S 2p core levels indicated that the phenomenon causes a uniform shift of 0.5 eV, from 229.6 eV to 229.1 eV and from 162.4 eV to 161.9 eV, respectively (Figure 3b,c). That is, each peak moved toward a lower binding energy after P-doping, which is very similar to the reported results for Nb-doped MoS2 [24]. This study reported that the Fermi level (EF) of pristine MoS2 is located close to the conduction band (Ec) edge, while an Nb-doped p-type sample has a Fermi level near the valence band edge. The work function and electron affinity of the pristine monolayer MoS2 is 5.1 eV and 4.28 eV, respectively [25]. The pristine MoS2 Fermi level is 0.82 eV, which is the Ec-EF result, and the doping sample Fermi level is 1.32 eV, 0.82 + ΔBE (measured from XPS data). Therefore, it is suggested that doping with P2O5 leads to a downshift in the Fermi level of about 0.5 eV, close to the valence band maximum.   Here, Ohmic metals of Ti and Ni were used for n-type pristine and p-type P-doped MoS2 FETs, respectively, to match the metal work functions [26]. Figure 4c shows the transfer characteristics of these devices fabricated on pristine and P-doped monolayer MoS2 flakes. The inset image of Figure  4d is the optical microscopy image of MoS2 FETs which was fabricated on the doped monolayer MoS2. The pristine MoS2 FETs demonstrates n-type conduction with a high on/off current ratio of ~10 5 [27]. The threshold voltage VT value extracted by the linear extrapolation method was about −8.1 V. In the case of P-doping, the transfer curve indicated p-type conduction with an on/off current ratio of ~10 3 , and the VT was −6.9 V at a drain-source voltage of 0.1 V. The field-effect mobilities of these FETs were calculated by the following relation: where L and W are the channel length and width, respectively. The back-gate capacitance (Cox = ε0εr/d) was ~1.28 × 10 −8 F/cm 2 , where ε0X is the dielectric constant and d is the thickness of silicon oxide. Using the transconductance value obtained by the relation of gm = dIDS/dVBG, the field-effect mobilities were determined to be about 29.7 cm 2 /Vs and 0.023 cm 2 /Vs for the pristine and P-doped MoS2 FETs, respectively. The carrier concentration in the FET channel could also be estimated by using the following relation: Here, Ohmic metals of Ti and Ni were used for n-type pristine and p-type P-doped MoS 2 FETs, respectively, to match the metal work functions [26]. Figure 4c shows the transfer characteristics of these devices fabricated on pristine and P-doped monolayer MoS 2 flakes. The inset image of Figure 4d is the optical microscopy image of MoS 2 FETs which was fabricated on the doped monolayer MoS 2 . The pristine MoS 2 FETs demonstrates n-type conduction with a high on/off current ratio of~10 5 [27]. The threshold voltage V T value extracted by the linear extrapolation method was about −8.1 V. In the case of P-doping, the transfer curve indicated p-type conduction with an on/off current ratio of~10 3 , and the V T was −6.9 V at a drain-source voltage of 0.1 V. The field-effect mobilities of these FETs were calculated by the following relation: where L and W are the channel length and width, respectively. The back-gate capacitance (C ox = ε 0 ε r /d) was~1.28 × 10 −8 F/cm 2 , where ε 0X is the dielectric constant and d is the thickness of silicon oxide. Using the transconductance value obtained by the relation of g m = dI DS /dV BG , the field-effect mobilities were determined to be about 29.7 cm 2 /V·s and 0.023 cm 2 /V·s for the pristine and P-doped MoS 2 FETs, respectively. The carrier concentration in the FET channel could also be estimated by using the following relation: where e is the electron charge [28]. The electron concentration in pristine MoS 2 was 6.47 × 10 11 cm −2 , whereas the hole concentration in P-doped MoS 2 was 1.01 × 10 11 cm −2 . Based on these results, the complete p-type conduction of MoS 2 with the addition of P 2 O 5 was demonstrated in this study.

Conclusions
We have demonstrated the p-type conduction of P-doped MoS 2 by P 2 O 5 via a CVD process. Based on AFM and Raman measurements, pristine and P-doped MoS 2 were confirmed to have monolayer thickness with grain sizes in the order of 15 µm. From XPS data, it was suggested that the Fermi level of P-doped MoS 2 shifted by about 0.5 eV toward the valence band compared to the pristine MoS 2 . FETs with P-doped monolayer MoS 2 showed p-type conduction with a field-effect mobility of 0.023 cm 2 /V·s and an on/off current ratio of 10 3 , while pristine MoS 2 FETs had n-type behavior with a field-effect mobility of 29.7 cm 2 /V·s and an on/off current ratio of 10 5 . The carriers in the FET channel were identified to be holes with a concentration of 1.01 × 10 11 cm −2 in P-doped MoS 2 and electrons with a concentration of 6.47 × 10 11 cm −2 in the pristine material. This phosphorous doping technique should be applicable to other TMD materials.