Wafer-Scale Synthesis of WS2 Films with In Situ Controllable p-Type Doping by Atomic Layer Deposition

Wafer-scale synthesis of p-type TMD films is critical for its commercialization in next-generation electro/optoelectronics. In this work, wafer-scale intrinsic n-type WS2 films and in situ Nb-doped p-type WS2 films were synthesized through atomic layer deposition (ALD) on 8-inch α-Al2O3/Si wafers, 2-inch sapphire, and 1 cm2 GaN substrate pieces. The Nb doping concentration was precisely controlled by altering cycle number of Nb precursor and activated by postannealing. WS2 n-FETs and Nb-doped p-FETs with different Nb concentrations have been fabricated using CMOS-compatible processes. X-ray photoelectron spectroscopy, Raman spectroscopy, and Hall measurements confirmed the effective substitutional doping with Nb. The on/off ratio and electron mobility of WS2 n-FET are as high as 105 and 6.85 cm2 V−1 s−1, respectively. In WS2 p-FET with 15-cycle Nb doping, the on/off ratio and hole mobility are 10 and 0.016 cm2 V−1 s−1, respectively. The p-n structure based on n- and p- type WS2 films was proved with a 104 rectifying ratio. The realization of controllable in situ Nb-doped WS2 films paved a way for fabricating wafer-scale complementary WS2 FETs.

Here, in this work, we demonstrate for the first time the wafer-scale synthesis of WS 2 films by ALD with controllable in situ p-type doing, on 8-inch α-Al 2 O 3 /Si wafer, 2-inch sapphire wafers, and 1 cm 2 GaN substrate pieces. The growth mechanisms of ALD WS 2 and in situ Nb doping were analyzed, and the doping concentration is shown to be controllable by altering Nb cycle numbers. Plan-view and crosssectional TEM imaging reveals the layered structure of WS 2 , and Hall effect measurements and TOF-SIMS confirm the effective incorporation of Nb dopants. Moreover, both WS 2 n-FETs and Nb-doped WS 2 p-FETs were fabricated by CMOS-compatible processes from as-prepared ALDgrown n-WS 2 and Nb-doped p-WS 2 films. The on/off ratio and electron mobility of WS 2 n-FET were up to 10 5 and 6.85 cm 2 V -1 s -1 , while the on/off ratio and hole mobility of Nb-doped WS 2 p-FET were 10 1 and 0.016 cm 2 V -1 s -1 , respectively. WS 2 FETs with different concentrations of Nb dopants were also investigated. Our work, by demonstrating in situ controllable Nb-doped WS 2 films and consequently p-FETs, helps establish a path to fabricate complementary WS 2 FETs at wafer-scale volumes. Figure 1(a) illustrates the mechanisms of the ALD process for WS 2 growth and in situ Nb doping. The reactor temperature was 400°C, while the WCl 6 (99.9%), NbCl 5 , and HMDST (98%) were kept at 93°C, 60°C, and room temperature, respectively. One cycle of WS 2 deposition includes 1 s WCl 6 pulse, followed by 8 s purge (Argon, 99.99%), and 1 s HMDST pulse, followed by 5 s purge, sequentially. For Nb doping, NbCl 5 and HMDST are used as precursors. One cycle of NbS 2 deposition includes 1 s NbCl 5 pulse, followed by 8 s purge (Argon, 99.99%), and 1 s HMDST pulse, followed by 5 s purge. The growth rate of WS 2 film was calibrated to about 0.036 nm/ cycle. To realize a controllable in situ doping, WCl 6 pulses were replaced by NbCl 5 pulses, and the doping concentration could thus be adjusted by varying NbCl 5 pulse numbers. Figure 1(b) shows photographs of wafer-scale 400-cycle WS 2 films deposited on 8-inch amorphous-Al 2 O 3 /Si wafer, 2-inch sapphire wafer, and pieced GaN substrate with good uniformity. Raman spectra of 400-cycle annealed WS 2 films at 950°C are shown in Figure 1(c), confirming that high-quality WS 2 could be deposited on all these substrates except for Si with different thickness at 400 cycles. In view of this, we use sapphire as the substrate for this research.

Growth Mechanisms.
2.1.1. ALD-Deposited WS 2 Film. At the initial stage, the WCl 6 and HMDST vapor were exposed directly onto the sapphire substrates and WS 2 layers were formed laterally on sapphire substrates. The subsequent layers were deposited onto the initial WS 2 layer to connect the isolated flakes and form films. Considering this, a postannealing process would be beneficial for improving film quality. The asdeposited WS 2 films were annealed at 950°C for 2 h in sulfur atmosphere. The XPS spectra of as-deposited and annealed WS 2 films are shown in Figure 2(a). The fine spectra of asdeposited WS 2 contained two pairs of W 4f peaks, representing WS 3 and WS 2 , respectively. The higher coordination number of W atom in WS 3 than that in WS 2 results a shift towards higher binding energy, with the binding energies of W 6+4 f 5/2 and W 6+4 f 7/2 being 38.7 eV and 36.68 eV and those of W 4+4 f 5/2 and W 4+4 f 7/2 being 35.22 eV and 33.08 eV, respectively. Similarly, the fine spectra of asdeposited WS 2 showed two pairs of S 2p peaks. The positions of the S 2 2p 1/2 and S 2 2p 3/2 peaks for W 6+ -S bonding were at 164.54 eV and 163.54 eV, while the positions of the S 1 2p 1/2 and S 1 2p 3/2 peaks for W 4+ -S bonding were at 164.02 eV and 163.04 eV, respectively. XPS analysis for asdeposited WS 2 films shows the films to be a mixture of WS 2 and WS 3 , and the stoichiometric ratio of W/S was about 1 : 2.7. A postannealing process in S atmosphere at 950°C for 2 hours improves film crystallinity. After annealing, the fine spectra of W 4f exhibited only one pair of W 4f 5/2 and W 4f 7/2 peaks, indicating WS 3 components decomposed to WS 2 , along with a similar result for S 2p spectra, both without characteristic peaks indicative of W 6+ -S bonding. As a result, the stoichiometric ratio of W/S was reduced to 1 : 2.1, with the help of desulfurization and improved film crystallinity. The full spectra of as-deposited and annealed WS 2 are shown in Fig. S2. To further investigate the crystallinity of as-deposited and annealed WS 2 films, Raman spectroscopy was performed. After annealing, the relative intensity of the A 1g and E 1 2g +2LA(M) peaks for annealed WS 2 was much higher than that of as-deposited WS 2 (Fig. S3), confirming the improved film crystallinity after annealing. Therefore, subsequent WS 2 films in this paper have undergone a postannealing process. In addition, when increasing WS 2 film thickness from 250 cycle to 500 cycle, the separation between the A1g and E 1 2g +2LA(M) peaks increased from 64.2 cm -1 to 69.5 cm -1 , demonstrating good thickness controllability for ALD grown WS 2 , as shown in Figure 2(b). Plan-view and cross-sectional TEM imaging shown in Figure 2(c) reveal a continuous planar film, without warpages or kink formation. The thickness of the annealed 400-cycle WS 2 film was 4.6 nm, and a cross-sectional TEM image of a 3.7 nm WS 2 film is shown in Fig. S4. Preparing monolayer films is very challenging due to the growth mechanism of ALD TMD films. From the plane-view TEM and SAED patterns results, out of 259 WS 2 analyzed grains, the average grain size was 55 nm (details of grain size were shown      Fig. S5), while the largest grain size was as high as 160 nm. The AFM image of 4.6 nm WS 2 film is shown in Fig. S6. 2.1.2. In Situ Niobium-Doped p-Type WS 2 Films. Pure NbS 2 films were deposited by ALD using NbCl 5 and HMDST precursors, and the XPS results of as-deposited NbS 2 films are shown in Fig. S7. The Nb doping process is illustrated in Fig. S8 and Table S1. as-deposited and annealed 400cycle WS 2 films with 30-cycle Nb doping were then investigated by XPS. In the fine spectra of W 4f peaks (Figure 3(a)) of as-deposited Nb-doped WS 2 films, two pairs of characteristic peaks revealing both W 6+ -S bonding and W 4+ -S bonding were observed. However, different from the fine spectra of S 2p of as-deposited WS 2 , a pair of characteristic peaks of Nb-S bonding was also observed, indicating successful Nb substitutional incorporation. The fine spectra of Nb 3d confirmed the presence of NbS 2 as well. After annealing, only W 4+ -S bonding was observed in the W 4f fine spectra (see Figure 3(a)), while W 4+ -S bonding and Nb-S bonding were both observed in the S 2p fine spectra. The Nb 3d fine spectra proved the formation of NbS 2 , indicating that Nb atoms were substituted into WS 2 lattice. The stoichiometric ratio of Nb/S was about 1 : 2.0, while that of W/S was 1 : 2.1. The full spectra of asdeposited and annealed Nb-doped WS 2 are shown in Fig. S9. The Raman spectra of annealed Nb-doped 400cycle WS 2 films with Nb doping varying from 10 cycles to 100 cycles are shown in Figure 3(b). From the spectra, the blue shift of the A 1g peaks was obvious, especially in the Nb-doped WS 2 film with 100-cycle Nb doping, which implies stiffening of the Nb-doped WS 2 lattice with Nb-S bonds [18]. The annealing process was necessary for Nb atoms to be activated and incorporated substitutionally into the WS 2 lattice. The plan-view EDX mapping results are shown in Fig. S10, confirming successful Nb doping of the WS 2 film.
Hall effect measurements of undoped WS 2 and Nbdoped WS 2 with 30-cycle Nb doping were performed at temperatures ranging from 50 K to 300 K. As shown in Figure 3(c), the carrier type of undoped WS 2 was electrons, while the carrier type of Nb-doped WS 2 film was holes, confirming the effective Nb-substitutional doping. The hall mobility of undoped WS 2 was up to 147.9 cm 2 V -1 s -1 at 50 K and 86.3 cm 2 V -1 s -1 at 300 K, while the hall mobility of Nb-doped WS 2 was 12.4 cm 2 V -1 s -1 at 50 K and 3.6 cm 2 V -1 s -1 at 300 K, respectively. The resistivity of Nbdoped WS 2 was 4 orders of magnitude higher than that of WS 2 , which revealed the fact that the Nb atom was effectively doped to substitute W atom in WS 2 lattice.
As shown in Figure 3(d), the Hall mobility and resistivity of Nb-doped WS 2 films with Nb doping of 15, 20, and 100 cycles at 300 K and TOF-SIMS of pristine WS 2 and Nbdoped WS 2 with Nb doping of 20 and 100 cycles were investigated as well. With increasing Nb concentration, the hall mobility decreased from 12.60 cm 2 V -1 s -1 to 5.73 cm 2 V -1 s -1 , while the resistivity of 15-cycle Nb-doped WS 2 film was 3 orders of magnitude higher than that of 100-cycle Nb-doped WS 2 film. This result implied that 100cycle Nb-doped WS 2 was heavily p-doped. Nb secondary ion intensity of pristine WS 2 film was normalized to 1, while the Nb intensity of Nb-doped WS 2 films with Nb doping of 20 and 100 cycles was normalized as 5.13 and 19.25. The increased normalized Nb intensity implied the rising doping concentration with the increase of Nb cycle number. Both Hall effect results and TOF-SIMS gave evidence of in situ controllable and substitutional Nb doping. An accurate quantitative value of concentration of Nb doping could not be obtained due to the poor detection accuracy and low atom collection efficiency. STEM is not applicable for ALD grown Nb-doped WS 2 films, due to the nature of polycrystalline films yielding only the statistical results within few layers. Raw data of Hall measurements of WS 2 and Nb-doped WS 2 with in Figure 3(d) are shown in Table S2.  Figure 4(b), with V d varying from 0.1 V to 0.5 V, while the output characteristics with V g vary from 1 V to 5 V. The transfer on-current of WS 2 n-FET reached as high as 0.4 μA/μm at V d = 0:5 V, and the on-off ratio was up to 10 5 . The detailed mobility of 30 tested WS 2 n-FETs is also plotted in Figure 4(b). The maximum and minimum mobilities of n-FETs were 6.85 cm 2 V -1 s -1 and 0.32 cm 2 V -1 s -1 , respectively, while the median mobility was 3.58 cm 2 V -1 s -1 . The mobility of over 70% of WS 2 n-FETs was in the range of 1 to 5 cm 2 V -1 s -1 .
The transfer characteristic of a 4.6 nm Nb-doped WS 2 p-FET with 15-cycle Nb doping with V d varying from 0.1 V to 0.5 V and the output characteristics with V g varying from -2 V to -6 V are shown in Figure 4(c). Compared to the WS 2 n-FET, the carrier type changed from electron to hole, which proved the Nb substituted for W atom in WS 2 lattice. The on-and off-current of Nb-doped WS 2 p-FET was only 5 × 10 −2 at V d = 0:5 V, far less than that of WS 2 n-FET. However, the hole mobility of Nb-doped WS 2 p-FET was 0.016 cm 2 V -1 s -1 , while the on/off ratio was 10 1 . For Hall effect measurements, the resistivity of 15-cycle Nb-doped WS 2 was 5 orders of magnitude higher than that of undoped WS 2 , and the mobility of 15-cycle Nb-doped WS 2 was far less than that of undoped WS 2 at 300 K. The field-effect mobility of WS 2 FETs was smaller than the Hall effect of WS 2 , due to the influence of transistors' electrical contacts on the underestimation of field-effect mobility. The Hall mobility was roughly estimated through field-effect mobility due to the nonlinear dependence of carrier concentration on gate voltage [41]. Moreover, the stability of our process was inquired through measuring the on-current of Nb-doped WS 2 p-FET with gate length varying from 5 μm to 50 μm. (Figure 4(c)). The distribution of I d,sat (at V g = −4 V, V d = 0:5 V) amongst 132 Nb-doped WS 2 p-FET with 20-cycle Nb doping on the same day was summarized. With increasing gate length, I d,sat decreased, suggesting the fabrication process was wellcontrolled and uniform. To explore the controllability of Nb doping, the transfer characteristics of Nb-doped WS 2 FETs with Nb doping varying from 1 cycle to 20 cycles were measured (Figure 4(d)). Nb-doped WS 2 FET did not show p-type behavior but with a decreased on-and off-current until reaching 15 cycles. When further increasing Nb concentrations, the current of p-FET increased and the on/off ratio decreased in that the resistivity and mobility of Nbdoped WS 2 film decreased, which was identical to the hall effect measurements. The WS 2 FET was heavily p-doped after 20-cycle Nb doping. These results proved the good controllability of in situ Nb doping by ALD.
Due to the lack of dangling bonds at the surface of WS 2 , it was difficult to deposit very high quality high-k dielectrics. Thus, the PBTI of WS 2 n-FET was carried out to analyze the reliability of Al 2 O 3 high-k dielectric. The stress was applied to gate and biased at 5.5 V. DC transfer characteristics at V d = 0:5 V were measured right after the removal of PBTI stress at room temperature. As shown in Figure 4(e), after 1000 s stress, the degradation of on-current was 3.5%, while   5 Research the V th shift was only 300 mV which was 6% of max-applied gate voltage. The results implied the instability of high-k films indeed affected the electrical properties of WS 2 n-FET. Higher quality high-k dielectrics would improve the electrical property of WS 2 n-FET [42]. To investigate the air stability of WS 2 film, the WS 2 n-FET was placed in ambient atmosphere, and the transfer characteristics were tested at V d = 0:5 V after 1 month, 3 months, and 6 months, as shown in Figure 4(f). The on-current of WS 2 n-FET degraded slightly, while the degradation was within one order of magnitude even after 6-month exposure in air. However, despite the fact that the deterioration of off-current was hardly observed after 3-month exposure, the deterioration of off-current was almost one order of magnitude after 6-month exposure. Consequently, the on/off ratio decayed from 10 5 to 10 4 after 6 months in ambient. Furthermore, vertical p-n structure based on WS 2 and Nb-doped WS 2 films was fabricated. The electrical property of p-n structure with rectifying ratio of 10 4 is shown in Figure 4(g), with an ideal factor of 2.3, indicating a conspicuous recombination of electron-hole.
The benchmark of p-type WS 2 transistors is listed in Table 1, including various deposition doping methods. The CVD method could yield the highest I on /I off ratio by Field-effect mobility ( FET )   4.3. Device Measurement. All electrical properties of WS 2 n-FETs and Nb-doped WS 2 p-FETs were measured in ambient room temperature by the Agilent B1500A Semiconductor Device Analyzer in probe station (MPI-TS3000). The field-effect carrier mobility was extracted from the transfer characteristic using the equation μ = ðΔI d /ΔV g Þ × L/ðWC ox V d Þ, and the C ox = 2:656 F/m 2 was the unit gate capacitance between channel and top-gate (C ox = ε 1 ε o /d, ε 1 = 6, and d = 20 nm for Al 2 O 3 dielectric).

Conflicts of Interest
The authors declare no competing financial interest.

Authors' Contributions
Y.W., C.T., and L.J. conceived and designed the experiments. Y.W., C.T., R.B., and H.J.Y. carried out the material deposition, annealing, and device fabrication. Y.W., C.T., and 7 Research Z.C.W. carried out the I-V measurements and reliability measurements. S.H. and X.Z. contributed to material characterizations. All authors contributed to interpreting the data and writing the manuscript. Hanjie Yang, Yang Wang, Xingli Zou, and Li Ji contributed equally to this work.