Atomic layer deposition of molybdenum disulfide films using MoF₆ and H₂S

Research on layered materials, such as two-dimensional (2D) transition metal dichalcogenides (TMDCs), has grown rapidly in the past decade. After the mechanical exfoliation of graphene was reported in 2004,¹ numerous new 2D materials have been synthesized and researched.^2,3 Prior to this recent interest in TMDCs, 2D materials such as molybdenum disulfide (MoS₂) and tungsten disulfide (WS₂) were used primarily as solid state lubricants.⁴ These materials have layered structures and were heavily researched in their bulk (multilayer), nanotube, and fullerene structural forms.⁵ More recently, the unique optical and electrical properties of these materials^2,6 have attracted much attention, with MoS₂ quickly becoming the prototypical TMDC. Whereas bulk MoS₂ has a band gap of 1.3 eV, single monolayer MoS₂ has a unique indirect-to-direct band gap transition of 1.8 eV.⁶ This presents a unique opportunity for semiconductor device manufacturing compared to the more widely studied graphene which is metallic (no bandgap) in its native state. MoS₂ has also been used as a cathode and anode barrier layer in lithium ion batteries^7,8 and as a catalyst for hydrogen production.^9,10

Transition metal dichalcogenides (MX₂ where M is a transition metal and X is a chalcogenide) have a layered structure where each layer consists of an X-M-X unit. The TMDC crystals are typically described as having trigonal or octahedral prismatic coordination to help describe the material in a single layer, where each M atom has six X atoms forming a hexagon above and below it.^11,12 These atomic trilayers feature strong in-plane covalent bonding, but weak van der Waals bonding between layers. This dichotomy of bonding characteristics facilitates the synthesis and isolation of single layer TMDCs. MoS₂ in the bulk crystalline form has three stable phases under standard conditions: 2H, 3R and 1T.^13,14 While the 2H and 3R are of the most interest because of their indirect-to-direct bandgap transition, the 1T phase has received less attention in the electronics field because its metallic properties are not suitable for devices such as transistors.

Two-dimensional MoS₂ has been synthesized using a variety of top-down and bottom-up methods. Typical top-down methods are mechanical exfoliation,¹ liquid exfoliation,¹⁵ and ion intercalation.¹⁶ These methods can yield high quality monolayer films up to 2.25 μm² from bulk crystals.¹⁷ Liquid exfoliation has been used to create dispersions of monolayer MoS₂ for inkjet printing flexible electronics.¹⁸ Many bottom-up approaches have been used for coating materials, including MoS₂. Muratore et al. deposited three- and five-layer MoS₂ at 350 °C using physical vapor deposition showing little to no oxygen contamination during growth.¹⁹ Wu et al. synthesized high quality MoS₂,WS₂, and multilayer films by the sequential sulfurization of Mo and W metal for vertical heterostructures.²⁰ Using chemical vapor deposition (CVD), Vilá et al. controlled the orientation of the MoS₂ growth by varying the MoO_x:S₂ ratio while sulfurizing MoO₃ at 750 °C.²¹ Feng et al. controlled the grain size of single layer MoS₂ by controlling the H₂ with MoO₃ and elemental sulfur.²² In addition to the oxide, MoCl₅ and elemental sulfur have been used to grow monolayer MoS₂ at high temperatures.²³ Among the various bottom-up approaches for growing MoS₂ thin films, atomic layer deposition (ALD) holds great appeal since the layer-by-layer growth has the potential to create uniform monolayer MoS₂ over large areas.

ALD is a chemical vapor process that involves sequential pulses of precursor vapors and is defined by self-limiting surface half-reactions.²⁴ Tan et al. showed ALD of MoS₂ using MoCl₅ and H₂S, which yielded a continuous film after 10 cycles at 300 °C.²⁵ Crystalline monolayers were obtained after annealing at high temperatures. Using this same chemistry, Browning et al. showed similar growth and used this method to fabricate back-gated field effect transistors. They found a relatively low mobility of 1 cm²V⁻¹s⁻¹, which they attributed to their device geometries.²⁶ Valdivia et al. reported the ability to grow MoS₂ on 150 mm Si/SiO₂ substrates.²⁷ Molybdenum carbonyl [Mo(CO)₆] has been used by multiple groups in conjunction with H₂S.^8,28,29 Jin et al. grew amorphous MoS₂ films with Mo(CO)₆ and dimethyldisulfide at 100 °C.³⁰ Recently, Jurca et al. reported a low-temperature ALD process using Mo(NMe₂)₄ and H₂S between 60 and 120 °C, while Mattinen et al. reported a process using a new Mo precursor with H₂S to deposit films between 250 and 350 °C that exhibited a range of morphologies and crystallinity.^31,32 For both of these studies, deposition at higher temperatures resulted in decomposition of the Mo precursor and non-self-limiting growth. Among the ALD MoS₂ publications to date, those that report 2D films are Refs. 25–30.

Scharf et al. reported that WS₂ ALD using tungsten hexafluoride (WF₆) and H₂S did not grow on the native oxide of silicon but readily nucleated on ALD ZnS.³³ Interestingly, the ALD WS₂ growth per cycle (GPC) decreased steadily with increasing WS₂ thickness on the ALD ZnS surface, but the higher initial WS₂ GPC was easily restored using a single pulse of diethyl zinc (DEZ).³³ Delabie et al. obtained WS₂ films by depositing a Si sacrificial layer on top of Al₂O₃ to aid in the reduction of the WF₆.³⁴ In addition to the sacrificial layer, a H₂ plasma was used after the WF₆ dose, which yielded a crystalline WS₂ film without the need for postdeposition annealing.³⁴ 

In this study, we explored MoS₂ ALD using molybdenum fluoride (MoF₆) and H₂S. MoF₆ has been used previously for Mo ALD using disilane,³⁵ and for MoS₂ CVD with H₂S.^36,37 MoF₆ is reduced readily by both Si and H₂ (Ref. 38),

The free energy changes for these reactions are −450 and −237 kJ/mol Mo, respectively, at 200 °C, indicating that both reactions are thermodynamically highly favorable.³⁸ In the previous report of Mo ALD using MoF₆ and disilane, the authors reported self-limiting behavior but measured a higher than predicted GPC which they attributed to CVD (i.e., MoF₆ → Mo + 3F₂) promoted by local, transient heating from the very exothermic ALD surface reactions.

In, this work x-ray amorphous molybdenum sulfide films were grown by ALD using MoF₆ and H₂S. In situ quartz crystal microbalance (QCM) measurements revealed that both half-reactions are self-limiting at 200 °C. Crystalline films were achieved after annealing at 350 °C in a hydrogen environment. The growth rate could be enhanced using diethyl zinc without changing the optical band gap of the material.

MoS₂ ALD was performed using a custom viscous flow, hot-walled reactor, which was detailed previously.³⁹ Deposition was performed on ∼1 × 1 in. coupons of Si with the native oxide intact and fused silica. The reactor temperature was maintained at 200 °C for all samples. During growth, ultrahigh purity N₂ (99.999%) was adjusted so the system process pressure was approximately 1 Torr. Molybdenum hexafluoride (MoF₆ 98%, Sigma Aldrich) and hydrogen disulfide (H₂S 99.5% Matheson Trigas) were sequentially pulsed into the reactor with purges of N₂ between each exposure. The MoF₆ and H₂S partial pressures were 20 and 150 mTorr during dosing of the respective precursor. The delivery pressure in the reactor for both precursors was regulated with both an inline 200 μm aperture (Lenox Laser) and a metering valve. Both gases are extremely dangerous and special precautions are needed due to the flammability/toxicity of H₂S and the corrosive nature of MoF₆. The ALD timing can be described as t₁-t₂-t₃-t₄, where t₁ and t₃ are the MoF₆ and H₂S exposure times, respectively, and t₂ and t₄ are the corresponding purge times with all times in seconds (s). For the MoS₂ growth, t₁ and t₃ were both 1 s, while the purge times (t₂ and t₄) were kept at 5 s. In some experiments, the samples were annealed in situ after deposition on a temperature-controlled hot stage. The sample annealing was performed in ultrahigh purity hydrogen at 350 °C holding for 15 min. The samples were then cooled quickly back to room temperature. In addition to the binary chemistry of MoF₆ and H₂S, the MoS₂ ALD was promoted/doped with ZnS using two successive ZnS ALD cycles composed of DEZ (99% Sigma Aldrich) and H₂S.

The MoS₂ ALD was investigated by in situ QCM measurements using a modified Maxtek Model BSH-150 sensor head. An RC cut quartz crystal (Phillip Technologies) with an alloy coating was used as the sensor due to its broad temperature range. To prevent deposition on the back side of the crystal, silver paste was used to seal the crystal and sensor head.³⁹ A backside N₂ purge was adjusted to approximately 0.5% of the process pressure.

X-ray photoelectron spectroscopy (XPS) measurements were carried out at the KECKII/NUANCE facility at Northwestern University on a Thermo Scientific ESCALAB 250 Xi (Al Kα radiation, hν = S5 1486.6 eV) equipped with an electron flood gun. Lower resolution survey scans and high resolution scans of the 3d, 2s and 2p electron energies were performed. The XPS data were analyzed using thermo avantage 5.97 software and all spectra were referenced to the C1s peak (284.8 eV). Peak deconvolution in the high-resolution spectra (Mo 3d, S 2p) was performed using the Powell fitting algorithm with 30% mixed Gaussian–Lorentzian fitted peaks in all cases. Fitting procedures were based on constraining the spin-orbit split doublet peak areas and FWHM according to the relevant core level (e.g., 3d_5/2 and 3d_3/2 is constrained to 3:2 peak area).

Raman spectroscopy (inVia, Renishaw) was used to probe the layered structure. The E_2g and A_1g vibrational modes arise from the in-plane and out-of-plane modes, respectively.⁴⁰ Measurements were performed in reflection using an excitation wavelength of 514 nm on all samples. To prevent sample damage, a neutral density filter of 5%–10% transmission was used. A D2 Phaser x-ray diffractometer (XRD) (Bruker) using a Cu Kα source in Bragg-Brentano geometry was used to probe the crystallinity and crystal structure of the MoS₂. A J.A. Woollam, Inc. α-SE Ellipsometer (Lincoln, NE) was used to measure the thickness of the bulk films using a Cauchy model.

The optical properties of the ALD molybdenum sulfide were measured using a Cary 5000 spectrophotometer (Varian) in transmission mode on films deposited on fused silica substrates. Kapton tape was placed on the backside of the quartz substrates during ALD and removed prior to measurement to mask off the region probed by the Cary 5000 beam. Prior to each measurement, a background reference was recorded to ensure accuracy. Linear regression of Tauc-plots was used to determine the optical band-gap of the films.

Thermodynamic calculations (HSC Chemistry, Outotec Oy) of the Gibbs free energies of reaction (ΔG) were performed to evaluate possible chemical reactions occurring during the molybdenum sulfide ALD. Two plausible chemical pathways were identified: direct and indirect. In the direct pathway, MoF₆ and H₂S react to form MoS₂, HF, and elemental S [Eq. (3)], with ΔG = −379 kJ/mol at 200 °C. In the indirect pathway, the initial solid-phase product is MoS₃ [Eq. (4)], with ΔG = −402 kJ/mol at 200 °C. Subsequent H₂ reduction [Eq. (5)] yields MoS₂, with ΔG = −24 kJ/mol at 350 °C. [We compute ΔG at 350 °C for Eq. (5) to match the experimental conditions used in the postdeposition annealing]. We note that the indirect pathway has a greater thermodynamic driving force (ΔG = −426 kJ/mol) compared to the direct pathway (ΔG = −379 kJ/mol). Moreover, the direct pathway might have a larger activation energy given the requirement for Mo reduction (+6 to +4) in Eq. (3), and so the indirect pathway might be kinetically favored as well. These mechanistic considerations will come into play later in our data analysis.

Our initial QCM studies were performed to probe the degree of self-limitation for the MoF₆ and H₂S half-reactions. Figure 1(a) shows the in situ QCM measurements (red circles) recorded using the timing sequence x–5-1-5 where the MoF₆ exposure time was varied between x = 0–2 s. The growth per cycle values were calculated assuming a bulk density of MoS₂, d = 5.06 g/cm³. The QCM data were fit with a Langmuir adsorption curve (line) and demonstrate saturation after ∼1 s MoF₆ exposure time at 0.4 Å/cycle. These measurements were repeated using the timing sequence 1-5-x-5 where the H₂S exposure time was varied between x = 0–2 s and revealed saturation after ∼1 s H₂S exposures at 0.40 Å/cycle [Fig. 1(b)]

Next, we used in situ QCM to monitor the mass changes versus time during molybdenum sulfide ALD using the timing sequence 1.5–15-1.5–15 at 200 °C [Fig. 2(a)]. The MoF₆ and H₂S dosing periods are indicated by the red and blue traces, respectively, at the bottom of the graph. We found linear growth at a GPC of 0.46(±0.01) Å/cycle using d = 5.06 g/cm³. This value is slightly higher than the GPC value of 0.40 Å/cycle from Fig. 1 and this discrepancy may relate to the longer dose times of 1.5 s used for both precursors in Fig. 2(a) while in Fig. 1 one of the precursor dose times was fixed at 1.0 s.

The data in Fig. 2(a) show a regular staircase pattern that coincides with the precursor dosing times. Figure 2(b) shows a magnified view of the QCM data for the first two ALD cycles. We see that the thickness (or mass) increases abruptly during the MoF₆ exposures and decreases abruptly during the H₂S exposures. Furthermore, the mass decreases continuously during the MoF₆ purge periods and appears to approach a steady-state value. In contrast, the mass is constant during the H₂S purge times. The slight delay between the QCM response and the precursor dosing traces results from the transit time of ∼0.5 s required for the precursors to travel from the dosing valves to the QCM in our ALD system.

The relative mass changes produced by the MoF₆ and H₂S exposures can be used to evaluate the molybdenum sulfide growth mechanism. If we assume that the molybdenum sulfide ALD proceeds via the direct route [Eq. (3)] and, furthermore, that the sulfur product sublimes from the surface, then we can propose the following surface reactions:

where surface species are designated with “*,” and all other species are in the gas phase. In Eq. (6), MoF₆ reacts with x surface thiol (SH) groups liberating x HF molecules, so that (6 − x) F atoms remain bound to the Mo. In Eq. (7) the new surface reacts with H₂S to release the remaining (6 − x) F atoms as HF as well as solid S. We hypothesize sulfur subsequently becomes a volatile species, probably in the form of S₈, while the surface has the newly formed MoS₂ species and is terminated with x SH groups so that the original surface functionality is restored. We note that the hypothesis of S sublimation is reasonable given that the vapor pressure of S is ∼2 Torr at 200 °C.⁴¹ We can define the QCM step ratio as R= Δm_A/Δm, where Δm_A is the mass change from reaction Eq. (6) and Δm is the mass change for one complete ALD cycle minus the sulfur species we assumed has entered the gas phase after the reaction. Given the atomic weights of the surface species, we can write

The average step ratio from the QCM data in Fig. 2(a) is R = 1.32(±0.05). From Eq. (8), this implies that x = 0, meaning that there are no surface thiols involved in the ALD process, but rather the MoF₆ reacts leaving all 6 F atoms on the surface (some of which may bond to other, nearby Mo atoms).

Alternatively, the molybdenum sulfide ALD may proceeds via the indirect route [Eq. (4)], which suggests the following half-reactions:

These reactions are identical to Eqs. (6) and (7) with the exception that all of the S remains on the surface and the ALD film has the composition MoS₃. Equations (9) and (10) yield the following QCM mass ratio:

Equation (11) predicts R = 1.09 for x = 0, and R = 0.47 for x = 6. In other words, there is no x value that yields the experimental QCM step ratio R = 1.32(±0.05), implying that the indirect pathway [Eq. (4)] is not correct. Given that the QCM data are consistent with the direct pathway [Eq. (3)], then a plausible interpretation for the gradual mass loss during the MoF₆ purge time is the slow sublimation of S from the surface.

A series of films were deposited on silicon and fused silica substrates using the 1–5-1–5 timing sequence at 200 °C varying the number of ALD cycles between 100 and 1000. The thicknesses of the films deposited on silicon were determined using spectroscopic ellipsometry, and these data are shown as the solid symbols in Fig. 3(a). The line in Fig. 3(a) is a quadratic fit to the thickness data and demonstrates that the thickness does not increase linearly with cycles as expected for a layer-by-layer ALD process. The thickness after 100 cycles is 60 Å and the thickness after 1000 cycles is 750 Å so that the corresponding growth per cycle values are 0.60 and 0.75 Å/cycle, respectively. Both of these values are significantly higher than the value derived from in situ QCM in Fig. 2(a) of 0.46 Å/cycle where only 19 cycles were performed. It is evident that the growth per cycle for the ALD MoS₂ increases with thickness. One explanation for this phenomenon can be found in the SEM image for the 600 ALD cycle film on silicon shown in Fig. 3(b). This plan-view SEM image shows a rough morphology composed of flakes with lateral dimensions of 50–100 nm and thicknesses of ∼10 nm. Additional SEM images recorded at an angle of 20° to the substrate normal (not shown) indicate that these flakes extend approximately vertically from the silicon substrate surface. We hypothesize that these flakes create a surface area that increases with increasing ALD cycles and this leads to greater MoS₂ deposition. This unusual morphology might result from faster growth at the edges of the MoS₂ sheets where the ALD precursors react with defect sites.

Raman spectroscopy is a common method for identifying and characterizing MoS₂.⁴⁰ Figure 4(a) shows Raman spectra recorded from the 600 ALD cycle MoS₂ film on silicon using a 514 nm excitation laser as deposited (black line) and after the H₂ anneal (red line). The annealed MoS₂ film shows the characteristic peaks of the 2D material at 380 cm⁻¹ (E_2g, in-plane) and 410 cm⁻¹ (A_1g, out-of-plane). In contrast, the as-deposited film does not show these characteristic 2D MoS₂ peaks, but rather shows a weak peak at ∼435 cm⁻¹. This suggests that the as-deposited films are not stoichiometric MoS₂, or that they lack short-range order.

Preannealed samples showed an amorphous film when measured by XRD; however, after annealing, MoS₂ could be seen, featuring the (002) reflection, which arises from the layered structure. The XRD data from MoS₂ are consistent with the Inorganic Crystal Structure Database PDF 01–073-1508 for the interplanar spacing.

XPS measurements were performed on both the as-deposited films and the films annealed in H₂ to investigate the chemical composition. These films were prepared using 600 ALD cycles with a thickness of 45 nm. Figure 5 shows low-resolution survey scans for the as-deposited [Fig. 5(a)] and annealed [Fig. 5(b)] films. These spectra show that F, O, Mo, and S were present in the films and the relative concentrations are given in Table I. The F detected by XPS likely results from unreacted MoF_x species in the film. We believe that the O results from exposing the substrates to air. The MoS₂ films were removed from the ALD reactor at 200 °C, and oxidation of monolayer MoS₂ has been shown to occur as low as 100 °C.⁴² The concentrations of both the F and O decrease during the H₂ anneal.

High-resolution XPS data from the Mo 3d and S 2p regions before and after H₂ anneal are shown in Fig. 6. The energies of the Mo 3d and S 2p peaks were consistent with chemically exfoliated MoS₂ samples.⁴³ These spectra indicate that there exists MoS₂ and substoichiometric (MoS_x) environments due to the presence of an oxide phase after exposure of the samples to ambient air. The cause for the substoichiometric MoS₂ (MoS_x; where x < 2) is associated with MoS₂ dilution in an oxidized Mo (MoO_x) phase, generating a possible mixed (Mo-O_x-S_x) molybdenum oxysulfide species. In fact, oxygen is known to have a high reactivity with the edge defects found in MoS₂ nanoflakes, which are also found in the present work [Fig. 3(b)].⁴⁴ Based on the overall composition of Mo and S (Table I), the presence of a mixed Mo(IV) and Mo(V/VI) environment contributes to the small S/Mo ratios of 1.1 and 1.35 for the as-deposited and annealed samples, respectively. Again, this is not surprising given the strong precedence for the stability of MoO₂ and MoO₃ species at temperatures below 550 °C compared to MoS₂, even when sulfurization is undertaken.⁴⁵ After analysis of the high-resolution Mo 3d peak envelope [Fig. 6(a)], the integrated peak areas of the peaks corresponding to the spin-orbit split 3d_5/2 and 3d_3/2 contributions for MoS₂ (∼228 and ∼231, respectively) relative to the neighboring S-Mo-O_x/Mo-O_x doublet (∼229 and ∼232 eV, respectively) in the as-deposited and annealed samples increased by 36% and 50%, respectively. Thus, the higher relative amount of MoS₂ after annealing the films in H₂ at 350 °C corroborates the appearance of the (002) diffraction peak in the XRD pattern [Fig. 4(b)], which is characteristic of layered MoS_2. Examination of the high-resolution S 2p XPS peak envelope for both the as-deposited [Fig. 6(c)] and annealed samples [Fig. 6(d)] demonstrates the presence of only S²⁻ with spin-orbit split 2p_1/2 and 2p_3/2 contributions arising at 162.9 and 161.7 eV, respectively.^45,46 It can be concluded that, in addition to removing residual F arising from the MoF₆ precursor, the postannealing procedure was effective in removing some of the oxygen from the stable Mo-S_x-O_x phase, which yielded a purer distribution of MoS₂ with more dominant contributions attributed to Mo(IV) in the Mo 3d XPS region.

To summarize, XRD suggests that the as-deposited film is amorphous whereas the SEM image shows what appear to be nanocrystals. It is likely that the diffraction peaks from these nanocrystals are too weak or broad to be detected by our XRD. The Raman measurements do not indicate crystalline MoS₂ as-deposited, and this may result from the residual F detected in the films by XPS. Finally, XPS indicates predominantly Mo(IV), and this agrees with the in situ QCM measurements that suggested MoS₂ is the reaction product.

To measure the optical properties of the films, we used fused silica substrates that were masked with Kapton tape preventing deposition from occurring on one side. This process simplified the optical measurements since the beam was only interacting with a single film. Figure 7(a) shows transmission data for ALD MoS₂ films of 12, 32, and 60 nm thickness measured over the visible spectrum in specular transmission mode. Figure 7(b) shows a fit to the 12 nm sample using a Tauc model assuming a direct, allowed transition, and the intercept of this line yields a band-gap of approximately 1.34 eV. This value agrees with literature values for bulk MoS₂.⁴⁷ 

As previously discussed, WS₂ ALD using WF₆ and H₂S can be accelerated by periodically dosing DEZ and H₂S to form a monolayer of ALD ZnS.³³ To explore whether this same phenomenon occurs during MoS₂ ALD using MoF₆ and H₂S, we performed in-situ QCM measurements. Figure 8(a) shows the mass versus time recorded by in situ QCM using the timing sequence 1–10-2–10 for the MoS₂ ALD and 1–10-2–10 for the ZnS ALD. In these experiments, two ZnS ALD cycles were performed followed by 20 MoS₂ ALD cycles, and this pattern was repeated several times. As shown in Fig. 8(a), each time the ZnS ALD is performed, the subsequent MoS₂ ALD is enhanced for the next 5–6 cycles and then decreases back to a steady state GPC value of ∼0.4 Å/cycle. We note that this behavior is virtually identical to that seen for WS₂ ALD, and the authors speculated that surface Zn might act as a catalyst to reduce the WF₆ and form ZnF, which is converted back to ZnS by exposure to H₂S.

Additional details can be gained from Fig. 8(b) which shows an expanded view of Fig. 8(a) between 9 and 13 cycles as a plot of thickness versus time. The first two ALD ZnS cycles deposit ∼2 Å/cycle as expected, but the very first MoS₂ ALD cycle following the ZnS deposits essentially zero mass. The second and third MoS₂ ALD cycles deposit 3.7 and 2.8 Å/cycle, respectively, which are seven to nine times higher than the steady-state value of 0.4 Å/cycle. Moreover, the step shape for these initial MoS₂ ALD cycles are much different from the steady-state step shape shown in Fig. 2(b) indicating different surface chemistry for the MoS₂ ALD half-reactions on the ZnS surface. Previous studies of WS₂ ALD found ZnF₂ crystals in the WS₂ coating and speculated that reduction of the WF₆ by surface Zn might accelerate the WS₂ ALD.⁴⁸ However, this hypothesis fails to account for the first WS₂ ALD cycle, which showed no growth. One explanation for the previous work and the present study is that the initial MoF₆ exposure both adds and removes material from the surface such that the net mass change is nearly zero. For instance, the MoF₆ might react with surface ZnS to yield a volatile sulfide-fluoride compound:

This reaction produces only a 6 amu mass change. Although very little is known about MoSF₄, the analogous reaction with ZnO to form ZnOF₄ is thermodynamically favorable (−116 kJ/mol at 200 °C), and the ZnOF₄ is highly volatile (>1000 Torr at 200 °C). This etching reaction would explain the ZnF₂ residue in the previous papers. Additional in-situ measurements including quadrupole mass spectrometry to identify the gas phase products and Fourier transform infrared (FTIR) absorption spectroscopy to evaluate the surface functional groups directing the surface reactions would help to understand better the surface chemistry for the ZnS-MoS₂ ALD.

We have shown using MoF₆ and H₂S as ALD precursors, self-limiting growth of x-ray amorphous MoS₂ is attainable. Two routes of growth were proposed: indirect (MoS₃) and direct (MoS₂). The MoS₃ route is thermodynamically favorable; however, QCM measurement showed that the direct route was the most plausible route. Moreover, XPS data confirmed the as-deposited films were MoS_2. While molybdenum oxide was present, this was attributed to air exposure of the samples upon removal from the reactor at elevated temperature. After hydrogen annealing, crystalline MoS₂ x-ray peaks and Raman peaks were visible.

Similar to the WS₂ accelerated growth on ZnS, DEZ substitution of MoF₆ pulses accelerated the MoS₂ growth. Our QCM measurements suggest that an etching reaction involving volatile MoSF₄ species may occur and could explain earlier reports of ZnF₂ residues in WS₂. This work offers an alternative halogen-based process for carbon-free atomic layer deposition of MoS₂ at relatively low temperatures.

This work made use of the XPS facility of the NUANCE Center at Northwestern University, which has received support from the Soft and Hybrid Nanotechnology Experimental (SHyNE) Resource (NSF NNCI-1542205). S.L. acknowledges support from the U.S. Department of Energy, Office of Science, Office of Workforce Development for Teachers and Scientists, Office of Science Graduate Student Research (SCGSR) program. The SCGSR program is administered by the Oak Ridge Institute for Science and Education for the DOE under Contract No. DE‐SC0014664. Electron microscopy was performed at the Electron Microscopy Center for Materials Research (EMCMR) at Argonne National Laboratory. Use of the Center for Nanoscale Materials, including resources in the Electron Microscopy Center, was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-06CH11357.