Positive Seebeck coefficient of niobium-doped MoS2 film deposited by sputtering and activated by sulfur vapor annealing

Herein we report on the positive Seebeck coefficient S = 162 μV K−1 of niobium (Nb)-doped MoS2 films prepared by sputtering and activation of Nb atoms by sulfur vapor annealing. The p-type doping achieved via these processes is discussed based on changes in chemical bonding states and resistivity behavior in terms of annealing and measurement temperatures. The results of this study provide a new option for p-type doping of MoS2 films and are expected to contribute to the development of nanoelectronics and a smart society.


Introduction
In recent years, thermoelectric generators (TEGs) have attracted considerable attention as a power source for Internet of Things devices, including wearable electronics for healthcare. [1][2][3][4] However, the power generation performance of TEGs is poor, therefore studies on thermoelectric materials, that can achieve higher performance, are being investigated. Among them, films of molybdenum disulfide (MoS 2 ), which belongs to the family of transition metal dichalcogenides (TMDCs), show promise because of its power factor PF = 8.5 mW m −1 K −2 , which has been reported previously. 5) The power factor PF = S 2 σ, calculated from the Seebeck coefficient (S) and conductivity (σ) of the material, is an indicator of the maximum power output of the TEG. This value is twice as high as that of bismuth telluride (Bi 2 Te 3 ), which has already found practical application as a thermoelectric material.
A p-type semiconductor is required to create the unit structure of a TEG. However, p-type doping of MoS 2 films is difficult due to the sulfur vacancies that act as donors and the Fermi-level pinning. 6,7) For example, in a previous study on nitrogen plasma doping, a positive shift in the threshold voltage of the back-gate field-effect transistor was observed, whereas the conduction characteristics remained as n-type. 8) In addition to plasma doping, [8][9][10][11] other doping methods have been reported, including molecular film coating, [12][13][14][15] gas adsorption, [16][17][18][19] solid-phase oxide stacking, [20][21][22] and lowenergy ion implantation. 23,24) Among these, the bottom-up method, in which doping and film deposition are conducted simultaneously, is considered to have the best long-term stability and can achieve the desired doping concentration without degrading the crystal structure. 25,26) Among the candidates for substitutional dopants, niobium (Nb) is advantageous because of its thermodynamic stability at Mo substitution sites and low acceptor level. [27][28][29] Nbdoped MoS 2 film formations via mechanical exfoliation, [30][31][32] chemical vapor deposition (CVD), 29,33,34) and metal-organic chemical vapor deposition (MOCVD) 35,36) have been reported. The thermoelectric performance of films produced by mechanical exfoliation has been also reported. 32) However, mechanical exfoliation is not suitable for industrial applications due to the difficulty of large-area formation, while CVD and MOCVD are prone to contaminations in precursors and additives. 35,37) On the other hand, physical vapor deposition (PVD) methods, such as sputtering, can be used to deposit films on a large area with good thickness controllability in simple systems, while preventing contaminations. 38,39) Thus far, co-sputtering with dual targets has been reported for the formation of Nb-doped MoS 2 films by PVD. 40,41) However, their thermoelectric properties have not been measured yet. In addition, the carrier concentrations are monotonically controlled by tuning the sputtering power and time on the Nb side, and thus the exponential carrier concentration modulation, which is essential for optimizing thermoelectric properties, has not been achieved. In this study, the formation of ptype Nb-doped MoS 2 films by sputtering an Nb-containing MoS 2 target and the control of the Nb activation rate at Mo sites by annealing are investigated. Furthermore, the thermoelectric properties are characterized.

Experimental method
Nb-containing MoS 2 targets with 0 and 1 wt% Nb were prepared using 99.99% MoS 2 and 99.95% Nb. Silicon substrates covered by 400 nm thick SiO 2 film were cleaned with a piranha solution. Then, 5 nm thick MoS 2 films were formed using a magnetron sputtering tool with an ultrahigh vacuum radio frequency (RF). The sputtering conditions were as follows: RF power = 100 W, distance between the target and the substrate = 200 mm, substrate temperature = 300°C, and argon pressure = 0.5 Pa. The film thickness was obtained by X-ray reflectivity measurement. Some samples were annealed at 700°C for 60 min in a sulfur-containing atmosphere to compensate for the sulfur vacancies and activate the Nb atoms, simultaneously. The electrical properties of the Nb-doped MoS 2 films were measured using the circular transmission line method (CTLM) 42) and van der Pauw method. 43) Titanium nitride with a thickness of 50 nm was deposited by sputtering, followed by wet etching using an ammonia-hydrogen peroxide mixture and photoresist to form the bottom electrode. Then, a 2.5 nm thick MoS 2 film was deposited. The thermoelectric properties of the Nb-doped MoS 2 films were measured by an on-chip thermoelectric device. 43,44) A Nbdoped MoS 2 film of approximately 2.5 nm was formed by RF magnetron sputtering. After annealing in a sulfur-containing atmosphere, the first 10 nm thick HfO 2 insulator was deposited via atomic layer deposition at 250°C with tetrakis(dimethylamino)hafnium (TDMAH) and H 2 O precursors. An active area was defined with lithography and reactive ion etching (RIE), followed by resist removal with O 2 plasma. The second 10 nm thick HfO 2 insulator was then deposited to encapsulate the sidewall of the MoS 2 film. To simultaneously form thermometer/potential probes in contact with the side of the MoS 2 film and a heater electrode on the SiO 2 , the HfO 2 /MoS 2 bilayers were opened by lithography and RIE. After reverse sputtering in Ar plasma, a 40 nm thick W film was deposited by RF magnetron sputtering. Then, the electrodes were formed with the lift-off process. Finally, the channel and heater were isolated by etching the HfO 2 around the heater.

Results and discussion
The X-ray diffraction (XRD) patterns are shown in Fig. 1. The out-of-plane measurements in Fig. 1(a) show that the peak of the (002) plane, which indicates the c-axis orientation of the MoS 2 film, is present in all the samples. Similar shaped peaks with our results have been reported in polycrystalline TMDC thin films formed horizontally on the substrate surface by sputtering, 43,45) therefore the MoS 2 films in this study are considered to have a similar crystal structure whether with or without Nb doping. The in-plane measurements in Fig. 1(b) show the (100) and (110) planes perpendicular to the c-axis in all the samples. The full width at half maximum values of these peaks decreased with sulfur vapor annealing (SVA), which might be caused by a grain size enlargement effect. The grain size of sputtered TMDC film with similar process steps was estimated to be around 10 nm from TEM images. 43) Although the (105) plane was observed between the two peaks without SVA, it became almost invisible after SVA because of the significant improvement in the c-axis orientation of the Nb-doped MoS 2 film. The interplanar distances calculated from Bragg's equation d 2 sin q l = are summarized in Table I. For both (002) and (100) peaks, the surface spacing approached the literature value by SVA. 46) This is considered to be because the sulfur vacancies, that cause the decrease in lattice constant, were compensated for by SVA, and the crystal structure approaches an ideal state. Similar effects have been discussed in previous studies. 43) In general, it can be concluded that Nb doping did not affect the orientation and crystallinity of the MoS 2 films, moreover, these were further improved by SVA.
The X-ray photoelectron spectroscopy (XPS) results of Mo 3d, S 2p, and Nb 3d of each sample, are shown in Fig. 2. In general, the composition ratio is expressed as where C i , A i , and RSF i are the composition ratio, peak area, and relative sensitivity factor of the atom i, respectively. 47) The S:Mo and Nb:Mo ratios of each sample are summarized in Table II. The as-sputtered MoS 2 films without SVA contained approximately 10%-15% sulfur vacancies, which were significantly compensated by SVA. Nb was detected only in the Nb-doped samples. The Nb-to-Mo content ratio was approximately 0.0167, which is the value converted from 1 wt% in the sputtering target. The XPS spectra of the Nbdoped sample with SVA shifted toward the lower energy side compared with those of the others. This indicates that the Fermi level of the MoS 2 film was shifted toward the direction of the valence band, which is consistent with the p-type doping phenomenon of Nb reported in previous papers. 29,48) The XPS spectra of the Nb-doped sample without SVA did not shift, probably because Nb atoms did not act as active dopants due to the poor crystallinity of the MoS 2 film. Nb generates holes in MoS 2 film when it displaces Mo sites bonded with six sulfur atoms; the MoS 2 film before SVA contains a huge amount of sulfur vacancies, resulting in poor crystallinity and a low percentage of Nb in such an arrangement.
The electrical properties measured by CTLM and the van der Pauw method are shown in Fig. 3. Figure 3(a) shows the resistivity at the inverse of the maximum temperature of the process in the CTLM pattern, which corresponds to the sputter temperature of 300°C (black plots) without SVA and with SVA at 500°C-700°C (red plots). It should be supplemented that previous studies have shown that the SVA at temperatures below 500°C does not provide sufficient sulfur compensation or crystallinity improvement. 47)    According to the graph, the resistivity of the Nb-doped sample decreased to approximately one-fifth of that of the undoped sample. However, when SVA was applied at 500°C , the resistivity increased because of the decrease in n-type carrier concentration due to sulfur compensation. 47) As the SVA temperature increased, the resistivity exponentially decreased. The resistivity of the sample subjected to SVA at 700°C eventually became lower than that of the Nb-doped sample without SVA. This can be explained as follows. The resistivity of the MoS 2 film can be expressed as ρ = 1/qnμ. The carrier concentration n and mobility μ are variable, however n is considered the main contributor to the exponential change in resistivity. In this case, SVA is thought to have two effects on the carrier concentration of Nb-doped MoS 2 films: it decreases the n-type carriers because of sulfur compensation and increases the p-type carriers because of Nb activation. The higher resistivity of the sample subjected to SVA at 500°C compared with the sample without SVA is attributed to the greater effect of sulfur compensation than that of Nb activation. On the other hand, the rate of Nb activation exponentially increased as the SVA temperature increased. This is attributed to the effect of Nb activation, which exceeded that of sulfur compensation at 700°C SVA, resulting in the lowest resistivity in this experiment. In other words, the concentration of p-type carriers in Nb-doped MoS 2 films exceeded that of n-type carriers, which suggests that the carrier concentration can be modulated exponentially by controlling the SVA temperature. Figure 3(b) shows the Arrhenius plot of the resistivity obtained from measurements using the van der Pauw method. The two filled rhombus plots were obtained in this study. As seen from the two lines, both samples exhibited semiconductor characteristics in which their resistivity values decreased with an increase in temperature. The slope of the graph appears to change at temperatures around 190 K. The slope at the low-temperature region is smaller than that at the high-temperature region, this is because of the dominance of hopping transport through localized states, as reported in several previous studies. 49,50) On the other hand, in the high-temperature region, the dependence of resistivity on temperature can be modeled with the following thermally activated transport equation: where E a , k B , and σ 0 are the activation energy extracted from the slope of the Arrhenius plot, Boltzmann constant, and prefactor depending on the temperature, respectively. The activation energy values calculated from Eq. (2) were 0.136 eV for the undoped sample without SVA and 0.081 eV for the Nb-doped sample with 700°C SVA. The activation energy corresponds to the energy difference between the Fermi level and valence band maximum (VBM) or conduction band minimum (CBM), whichever is closer in the band structure. The undoped MoS 2 film is an n-type because of sulfur vacancies as reported in a previous study; 51) hence, the activation energy is considered to be the energy difference between the Fermi level and CBM for the undoped sample without SVA. The open symbols explain an Arrhenius plot of an undoped sample with SVA obtained in a previous study. 11) According to this plot, both resistivity and activation energy are higher than those of the undoped sample without SVA. This can be interpreted because of sulfur compensation, which results in weaker n-type characteristics than the undoped sample without SVA. On the other hand, the Nb-doped sample with SVA shows a decrease in both resistivity and activation energy compared to the undoped sample without SVA. Since the Fermi level should be shifted downward from the XPS results, this activation energy is considered to be the energy difference between the Fermi level and the VBM. This can be interpreted as an increase in p-type carriers due to Nb activation exceeding the decrease in n-type carriers due to sulfur compensation, which is consistent with the CTLM results. The changes in band alignment based on the above considerations are shown in Fig. 4. This indicates that the MoS 2 film, which was originally an ntype, was converted into a p-type by Nb doping and SVA.
To demonstrate the p-type properties of the Nb-doped MoS 2 film, the Seebeck coefficient was measured using the on-chip thermoelectric device shown in Fig. 5(a). The temperature coefficients of resistance (TCR) for each thermometer electrode (TM1 and 2) were extracted at cryostat temperatures between 280 and 320 K using the Kelvin probe method with an applied current of 1 μA, as shown in Fig. 5(b). Note that a series of measurements were performed under high vacuum conditions in a cryostat.
Then, the voltage V heater was applied to the heater to generate a temperature gradient across the channel, while the temperature in the measurement chamber was stabilized at 300 K. The relationship between V 2 heater and ΔT was determined using the TCR extracted earlier, as shown in Fig. 6. Finally, the thermoelectric voltage ΔV thermal between the thermometer electrodes were measured by a nanovoltmeter (Keithley 2182A). The result is shown in Fig. 7. The value of ΔV thermal was obtained from approximately 150 measurements, and the offset voltage at ΔT = 0 was subtracted. Eventually, the Seebeck coefficient S = 162 μV K −1 was obtained for the Nb-doped MoS 2 film. This positive value proves a p-type operation in the Nb-doped MoS 2 film, which is consistent with the above discussions. The conductivity σ at 300 K was measured as 79.8 S m −1 , which yielded a calculated PF of 2.09 × 10 −3 mW mK −2 from the expression PF = S 2 σ. As a benchmark, the thermoelectric properties of TMDC films are listed in Table III. The PF value of Nb-doped MoS 2 film was lower than those of previous studies and needs to be improved. Regarding the two components of S and σ to maximize the PF, there is said to be an optimum value for S at 130−187 μV K −1 . 52,53) Since the S value of the Nb-doped MoS 2 film in this study is included in this range, there is little room for improvement in S and the carrier concentration that affects its value. 54) Therefore, to apply Nb-doped MoS 2 films to thermoelectric devices, it is necessary to enhance the conductivity by improving the mobility. Possible techniques include an increase in the grain size by improving the deposition process, reducing the Nb concentration while an increase in the activation rate, and enhancing the quantum effect on the film thickness.

Conclusions
P-type MoS 2 films with a positive Seebeck coefficient were successfully achieved by sputtering with an Nb-containing MoS 2 target and SVA. The p-type doping effect was attributed to the compensation of sulfur vacancies and the activation of Nb by SVA. Because the activation rate of Nb varied exponentially with the change in annealing temperature, the desired carrier concentration can be achieved by  . Temperature T and temperature difference ΔT as a function of V 2 heater at cryostat temperature of 300 K. Temperature measurements were taken 180 s after heater voltage was applied, with a static temperature difference ensured. Fig. 7. Thermoelectric voltage ΔV thermal as a function of temperature difference ΔT across the channel at a cryostat temperature of 300 K. Measurements were taken 180 s after heater voltage was applied. Among approximately 150 plots, those beyond the 3σ range were excluded, and the fitting line was extrapolated using the least squares method.