Area‐Selective Growth of HfS2 Thin Films via Atomic Layer Deposition at Low Temperature

The transition‐metal dichalcogenide HfS2 is a promising alternative semiconductor with adequate band gap and high carrier mobility. However, a controllable growth of continuous HfS2 films with selectivity for specific surfaces at a low temperature on a large scale has not been demonstrated yet. Herein, HfS2 films are grown at 100 °C by atomic layer deposition (ALD) based on the precursors tetrakis(dimethylamido)hafnium and H2S. In situ vibrational spectroscopy allows for the definition of the temperature range over which (Me2N)4Hf chemisorbs as one monolayer. In that range, sequential exposures of the solid surface with (Me2N)4Hf and H2S result in self‐limiting reactions that yield alternating surface termination with dimethylamide and thiol. Repeating the cycle grows smooth, continuous, stoichiometric films of thicknesses adjustable from angstroms to >100 nm, as demonstrated by spectroscopic ellipsometry, XRR, AFM, UV–vis and Raman spectroscopy, XPS, and TEM. The well‐defined surface chemistry enables one to deposit HfS2 selectively using, for example, patterns generated in molecular self‐assembled monolayers. This novel ALD reaction combines several attractive features necessary for integrating HfS2 into devices.


Introduction
Since the exceptional physical properties of graphene have been discovered, atomically thin 2D materials have been intensively studied as a promising alternative to classical semiconductors in electronic and optoelectronic applications. [1] Among them, the transition-metal dichalcogenides (TMDCs) have drawn significant research attention recently, in particular, because they feature an adjustable band gap. [2] As a representative of this family, MoS 2 is the most extensively studied TMDC. [3] Transistors based on this benchmark TDMC have shown a remarkably high on/off ratio due to its sizeable band gap. [4] However, the high performance of MoS 2 -based devices is limited by the material's relatively low charge carrier mobility. [5] Hafnium disulfide (HfS 2 ) is an emerging alternative, which is predicted to have a particularly The transition-metal dichalcogenide HfS 2 is a promising alternative semiconductor with adequate band gap and high carrier mobility. However, a controllable growth of continuous HfS 2 films with selectivity for specific surfaces at a low temperature on a large scale has not been demonstrated yet. Herein, HfS 2 films are grown at 100 °C by atomic layer deposition (ALD) based on the precursors tetrakis(dimethylamido)hafnium and H 2 S. In situ vibrational spectroscopy allows for the definition of the temperature range over which (Me 2 N) 4 Hf chemisorbs as one monolayer. In that range, sequential exposures of the solid surface with (Me 2 N) 4 Hf and H 2 S result in self-limiting reactions that yield alternating surface termination with dimethylamide and thiol. Repeating the cycle grows smooth, continuous, stoichiometric films of thicknesses adjustable from angstroms to >100 nm, as demonstrated by spectroscopic ellipsometry, XRR, AFM, UV-vis and Raman spectroscopy, XPS, and TEM. The well-defined surface chemistry enables one to deposit HfS 2 selectively using, for example, patterns generated in molecular selfassembled monolayers. This novel ALD reaction combines several attractive features necessary for integrating HfS 2 into devices. Y high charge carrier mobility and a proper band gap. [5,6] So far, experimental research on HfS 2 has allowed devices made from it to reach performances matching that of MoS 2 . [7] To date, only a few methods are available to fabricate ultrathin HfS 2 films and flakes, such as chemical vapor deposition (CVD) and micromechanical exfoliation of bulk crystals. [8] Exfoliation is, however, limited by solvent contamination and the fact that it produces discrete flakes. CVD yields full-area films, but the high volatilization and reaction temperatures needed of up to 1000 °C restrict the choice of substrate and lithographic patterning methods. [8b,c] In neither case, ultrathin, continuous and pinholefree layers on wafer scale have been demonstrated as needed for (opto)electronic device integration. [7c,9] Atomic layer deposition (ALD) is a thin film deposition technique based on sequential self-saturating reactions of gaseous precursors separated by purges. It works at low to moderate temperatures (50-350 °C typically) and allows to deposit highly uniform films on large scales (up to square meters) with an accurate thickness control down to the angstrom level. Moreover, its unique capability of coating structured substrates including deep pores, tubes, and trenches conformally gives ALD unparalleled advantages in the design of (opto)electronic devices. [10] To the best of our knowledge, only two ALD processes have been reported for HfS 2 by Mattinen and Chang. [11] Mattinen and colleagues performed the reaction with HfCl 4 and H 2 S at 400 °C to yield uniform, crystalline HfS 2 films. The high reaction temperature is rendered necessary by the poor nucleation behavior of the chloride precursor, which also generates corrosive HCl as the by-product and tends to generate chlorine contaminations in the bulk or at the interface. [12] Chang and coworkers used a halide-free precursor in combination with H 2 S plasma. The plasma enables for less elevated, yet not very low temperatures of 150-350 °C to be used; however, no oxidefree films were demonstrated.
Herein, we establish the low-temperature ALD growth of HfS 2 thin films using tetrakis(dimethyl-amido)hafnium, (Me 2 N) 4 Hf, and H 2 S as precursors. We first use vibrational spectroscopy to observe the surface reactions of both precursors, we then define the ALD temperature window and the self-limiting character of the growth, we next characterize the chemical identity and morphology of the deposits extensively, and we finally demonstrate area-selective growth on the selfassembled monolayer (SAM) surface.

Fundamental Observations of the Precursor Surface Chemistry
Alkylamide-transition metal(IV) complexes are rapidly developing as a family of ALD precursors due to their high reactivity and volatility. In this paper, we choose tetrakis(dimethylamido)hafnium, (Me 2 N) 4 Hf, for the HfS 2 film growth. We start the study by investigating the adsorption behavior of (Me 2 N) 4 Hf under ultra-high vacuum (UHV) conditions on a model substrate, namely Co 3 O 4 (111), using infrared reflection absorption spectroscopy (IRAS). In the presence of water traces in the UHV environment, reactive OH groups are formed on the Co 3 O 4 (111) surface. [13] However, the detection of such OH groups in IR is rather difficult, as those vibrations usually give rise to very broad and weak signals.

Multilayer Adsorption of (Me 2 N) 4 Hf at 150 K and Theoretical Calculation
At low coverages, the IR spectra recorded during the deposition of (Me 2 N) 4 Hf onto a Co 3 O 4 (111) sample maintained at 150 K ( Figure S1, Supporting Information, compared with DFT-computed spectra of (Me 2 N) 4 Hf as summarized in Table S1, Supporting Information) exhibits features at 2848, 2814, 2760, 1461, 1249, 1155, and 954 cm −1 . The dimethylamine by-product generated upon the chemisorption of (Me 2 N) 4 Hf at the hydroxylterminated surface is difficult to detect however, because its characteristic bands are overlapping with those of the (Me 2 N) 4 Hf. [14] With increasing coverage, the (Me 2 N) 4 Hf bands grow and additional absorption signals emerge at 3001, 2925, 1438, 1424, 1397, 1130, 1114, 1061, 1015, 975, 964, and 888 cm −1 . Comparison with theoretical results allows us to ascribe the new bands to a (Me 2 N) 4 Hf dimer due to the formation of a condensed multilayer.

Temperature-Programmed IRAS of (Me 2 N) 4 Hf at 150-450 K
When the (Me 2 N) 4 Hf multilayer is heated from 150 to 450 K (Figure 1), the intensity of all peaks decreases from 150 to 220 (±5) K (see Figure S2, Supporting Information). The system transforms from multilayer to monolayer at this temperature, which is characterized by sharper signals and shifts observable especially well below 1000 cm −1 . The low-temperature band at 954 cm −1 with two shoulders at 975 and 964 cm −1 , respectively, not only decreases in intensity upon heating, it also shifts to 944 cm −1 , while the band at 964 cm −1 picks up in intensity. These bands assigned to ν(Hf-N) or ν(N-C) vibrations evidence a preferential orientation of the chemisorbed monolayer on the surface. Indeed, surface selection rules [15] encode orientation information because only vibrations with a dynamic dipole moment perpendicular to the surface can be probed by IRAS. [15]

Temperature-Programmed IRAS of Alternating Precursors' Pulses at 300-650 K
In order to investigate the temperature-dependent behavior of the individual precursor reactions, we first prepared a thiol-terminated surface by exposing the Co 3 O 4 (111) substrate to 1 × 10 −6 mbar of D 2 S for 5 min. Subsequent exposure to 5 × 10 −7 mbar of (Me 2 N) 4 Hf during a heating ramp (2 K min −1 , Figure 2) causes spectral changes in the form of difference spectra.
New bands, marked in blue (Figure 2b,c; see also Figure S3a and Table S1, Supporting Information: 956, 1142, 1253, 2779, 2828, and 2869 cm −1 ), are formed during the precursor uptake and are similar to those observed above and attributed to the chemisorbed monomer ( Figure S1, Supporting Information). Bands marked in red at 1468 and 888 cm −1 represent species lost upon (Me 2 N) 4 Hf exposure. Interestingly, these two bands grow during D 2 S exposure (see Figure S3b, Supporting Information) and are related to either the thiol termination of the surface or a physisorbed dimethylamine by-product formed during D 2 S dosage. [16] The data collected at increasing temperatures show that cyclic reactions occur up to almost 570 K, although the integrated band areas (Figure 2c) suggest the window between 380 and 430 K as the most reactive one. In summary, the IRAS data collected under high vacuum conditions represent a guideline concerning the temperature window to be expected for the ALD process, and are consistent with the presence of well-defined acid-base reactions on the surface with generation of dimethylamine as the by-product.

ALD Process Study
When the surface reactions studied in an UHV chamber are transferred to a commercial ALD reactor, the film growth rate measured by spectroscopic ellipsometry as a function of the (Me 2 N) 4 Hf pulse duration (Figure 3a) exhibits saturating behavior. This property confirms self-limiting surface chemistry, and therefore, ALD growth. In the undersaturated regime, that is (Me 2 N) 4 Hf pulses from 0.1 to 1.2 s, the film is highly inhomogeneous. In saturating conditions, the inhomogeneity measured over the size of the chamber (10 cm) is on the order of 10% for a thickness of around 6 nm. For H 2 S, saturation is observed from 0.1 s pulse duration, whereas 0.2 s pulse is taken as standard. The determination of the ALD temperature window is presented in Figure 3b. From 65 °C as the reaction chamber temperature (lower limit imposed by the evaporation temperature), the growth rate remains stable within experimental uncertainty at about 1 Å per cycle until 100 °C. At 120 °C and beyond, the deposition rate is significantly lower, consistent with the results gathered via the TP-IRAS analysis. In the standard conditions chosen, that is, 100 °C growth temperature, Figure 3c demonstrates a linear dependence of the film thickness on the number of ALD cycles performed with a growth determined (more accurately than in Figure 3b) to be 1.2 Å per cycle, and the inhomogeneity over the chamber decreases to less than 5%. Figure 4 confirms the accuracy of this value. Taking a sample (Si/200 nm SiO 2 ) with 50 ALD cycles, Figure 4a exhibits the spectroscopic ellipsometry data recorded before and immediately after deposition. The curve fits yield a thickness of 6.3 nm for the HfS 2 layer. Independently of it, an X-ray reflectometry measurement ( Figure 4b) yields a thickness of 6.2 nm and a roughness of 0.5 nm.

Characterization of the ALD Film
The HfS 2 material is known to be unstable to aerobic oxidation and affords the highly stable hafnium dioxide at ambient conditions. [17] Indeed, the prompt oxidation of the ALD film is observed in a transmission electron microscopy (TEM) experiment shown in Figure 5a. After 50 ALD cycles, the chamber is opened to air at room temperature for approximately a minute, reheated to 100 °C, and submitted to a further set of 100 cycles; a cross-section of the film exhibits a clear contrast at two-third of the depth. Further characteristics of the film include its outstanding smoothness demonstrated in Figure 5a, and its essentially amorphous character. The selected area electron diffraction (SAED) radial integrated intensity in Figure 5c reveals order on a very short length scale, which we would refrain from calling polycrystalline. This extraordinary low roughness of the ALD-grown HfS 2 is quantified by AFM ( Figure S4, Supporting Information), with a root-mean-square value of 0.19 nm for an 18 nm thick film, close to the value of the substrate. The superb smoothness of the film observed by STEM and AFM agrees well with the roughness determined in XRR.
The chemical identity (elemental composition) of this film can only be determined after encapsulation with a protective layer. Although h-BN flakes, [17a] polymer, [18] , and ALD  [7c,11a,19] have been reported, we find that ZnS (by ALD) performs best, as a thin layer of it (7 nm) is sufficient to prevent the formation of oxide at room temperature for weeks. Energy-dispersive X-ray spectroscopic microanalysis performed of a ZnS-capped HfS 2 ALD coating (150 nm) on a porous alumina substrate (Figure 5d) yields an atomic composition of 27.5% S, 0.6% Zn, and 12.6% Hf. This yields an essentially perfect Hf/S stoichiometry of 1:2.1. We note that no other encapsulating layer that we have tested provides efficient protection from air. In our hands, all ALD processes using water, oxygen or ozone as the source of oxides end up forming substoichiometric HfS 2 (see Figure S5, Supporting Information), consistent with the partial conversion of HfS 2 into the corresponding oxide caused by the highly oxophilic character of hafnium(IV).
Further confirmation of the chemical composition, stoichiometry, and valence state of the ALD HfS 2 is provided by X-ray photoelectron spectroscopy (XPS) analysis: Figure 6 presents the ZnS-capped film ( Figure S6, Supporting Information, shows the oxidized Al 2 O 3 -capped film for comparison). The survey spectra and the S 2p and Hf 4f region all concur to show exclusively ZnS at the surface of the pristine sample, whereas a short sputtering treatment exposes the underlying HfS 2 . The S 2p doublet at 161.5 and 162.7 eV attributed to ZnS is now complemented by an additional contribution from HfS 2, at lower energies, 161.1 and 162.3 eV. Correspondingly, the Hf 4f region exhibits a beautiful doublet at 16.4 and 18.1 eV, matching reported HfS 2 data. [8a,d,20] Raman spectroscopy is a versatile method to determine structural properties such as thickness, crystallographic structure, and  7presence of defects in thin layered materials. [21] The Raman spectrum recorded of a HfS 2 coating (on a porous alumina substrate) upon excitation at 532 nm is presented in Figure 7a. The measurement has been taken immediately after growth. The main peak at 343.8 cm −1 corresponds to the A 1g mode of HfS 2 and is due to out-of-plane vibration of S atoms. The weaker peak at 257 cm −1 is attributed to the E g mode and relates to S atom vibration in the basal plane; it is, in full agreement with the peaks obtained from the bulk HfS 2 material and the previous experimental reports. [8b,11a,22] The small peak observed at 487 cm −1 is from the substrate. The fact that the observation of Raman modes corresponds to those known from crystalline HfS 2 indicates a high structural quality of HfS 2 domains within the film. This confirms the borderline (amorphous/polycrystalline) character observed by electron diffraction. Its optical properties corroborate a band gap of 2.5 eV, which is estimated from the UV-vis absorption spectrum, Figure 7b, and which is in line with literature data. [8b,d]

Area-Selective ALD Growth
The well-defined surface chemistry characterized so far suggests that ALD growth should be selective on substrates or areas that carry suitable surface reactive groups. Here, patterned selfassembled monolayer (SAM) substrates are used, as the ALD selective growth is designed on thiol-terminated molecular monolayers, whereas areas carrying SAMs with perfluoroalkyl groups represent the inert areas ( Figure S7, Supporting Information). Figure 8a shows the results of ALD growth of HfS 2 on the patterned SAM surface, with a stark contrast between the thiol-terminated and inert perfluorinated areas on the left and right sides of the sample, respectively. The EDX measurements in Figure 8b clarify that the Hf signal is detected from the thiol terminated sites but not on the inert perfluorinated areas after ALD growth. Furthermore, the SAM approach allows the experimentalist to tune the density of reactive groups in a mixed system. Figure 9 displays how the morphology of the HfS 2 layer obtained after 30 ALD cycles is affected by dilution of the active thiol sites in the inert perfluoroalkyl (see Figure S8, Supporting Information, for the rest of the AFM images). As quantified by atomic force microscopy (AFM), substrates carrying at least 50% thiols all yield a similar root-mean square roughness below 0.5 nm, whereas the diminishing density of nucleation sites gives rise to an increase in roughness to around 3.5 nm. In this case of individual nucleation sites essentially isolated from each other, the roughness corresponds perfectly to the  nominal film thickness measured on a Si wafer or expected based on the number of cycles. Ideally, the thiol-free substrate should give rise to no nucleation at all. Instead, the area investigated here of 2 µm × 2 µm exhibits approximately 380 nuclei (counted by an image processing software). Given the density of SAM of four molecules per square nanometer (see Figure S7, Supporting Information, for the molecules' structures), these 380 nuclei are to be compared to the 1.6 × 10 7 molecular sites in this same 2 µm × 2 µm region. In other words, the nucleation probability on unreactive sites (or the defect rate) is 8 × 10 −7 , or less than one part per million. This number remains essentially constant upon changes in substrate temperature or purge duration ( Figure S9, Supporting Information). These observations speak for the level of control achievable by well-defined surface chemistry.

Conclusions
In summary, the present study provides the first low-temperature (≤100 °C), plasma-free ALD process for generating thin coatings of 2D semiconducting HfS 2 . A very well behaved surface chemistry is consistently observed from UHV conditions to the low-vacuum conditions of applied ALD reactors. Overall, it is highly advantageous for an area-selective deposition.
Temperature-programmed infrared reflection absorption spectroscopy evidences the transition from condensation to the formation of monolayer and finally the thermal decomposition of the (Me 2 N) 4 Hf precursor on an oxidic surface over a wide temperature range. The insight obtained in UHV is consistent with the temperature window established in a lowvacuum reactor, where the low-temperature growth of continuous, smooth HfS 2 films is successfully established with a precise thickness control from (Me 2 N) 4 Hf and H 2 S as precursors. The film grows in a self-limiting manner at a rate of ca. 1.2 Å per cycle. The extremely air-sensitive HfS 2 is efficiently protected by an ALD deposited zinc sulfide layer of 7 nm thickness. Oxides, in particular Al 2 O 3 , are inadequate even if deposited directly after HfS 2 ALD without a vacuum break. Extensive material characterization by atomic force microscopy, Raman spectroscopy, X-ray photoelectron spectroscopy, UV-vis absorption spectroscopy, and transmission electron microscopy, yield a consistent picture of this borderline (amorphous/polycrystalline) characterized yet highly stoichiometric and pure material.
The choice of the highly reactive and volatile, homoleptic dialkylamido metal complex is instrumental in enabling film growth at low temperatures. In combination with the unique Figure 7. a) Raman spectra of the HfS 2 coated anodic aluminum oxide (AAO) membrane (in gray), crystalline HfS 2 bulk material (in gray blue), and the bare AAO substrate (in brown). b) UV-vis absorption spectrum of HfS 2 deposited on quartz. The absorption spectra of the substrates and the capping material ZnS was also recorded. versatility of the ALD technique, this allows the experimentalist to coat various substrates and even generate patterns based on area-selective growth. The selectivity can be as high as 0.9999992. Overall, this low-temperature ALD HfS 2 chemistry combines several advantages for its application in electronic and optoelectronic devices.

Experimental Section
Chemicals: All chemicals are of analytical reagent grade and were used as received without any further treatment. Tetrakis(dimethylamido) hafnium and H 2 S (3% in N 2 ) were ordered from Strem and Air Liquide, respectively. Methanol (MeOH, ≥99%) and isopropanol (i-PrOH, ≥99.5%) were purchased from Carl ROTH, and phosphonic acids  were purchased from Sikemia. The standard SiO 2 /Si wafer and quartz substrates were purchased from Silicon Materials, Inc., and Plan Optik AG, respectively. The specialized substrates anodic aluminium oxide (AAO) and Co 3 O 4 (111) thin films were fabricated according to literature procedures. [23] Characterization: The thickness of the deposited HfS 2 layers was determined on Si (100) wafers with a spectroscopic ellipsometer Sentech SENpro equipped with a halogen lamp and a monochromator. Energy dispersive X-ray spectroscopy (EDX) was measured on a JEOL JSM 6400 PC equipped with a LaB 6 cathode and SDD X-ray detector. X-ray photoelectron spectroscopy (XPS) spectra were recorded with monochromatized Al Kα radiation (PHI Quantera II, Japan). Raman spectrum of the thin film was measured in ambient conditions on a WITec alpha-R Confocal Raman Microscope in backscattering with 532 nm excitation wavelength. Raman spectrum of HfS 2 bulk in backscattering geometry was recorded using a Horiba Jobin Yvon LabRAM Aramis Raman spectrometer. Neon lines were used to calibrate the spectrum. The optical absorption spectra were measured with an ultraviolet-visible spectrophotometer (Ocean Optics) equipped with a DH-2000-L light source, a HR40000 spectrometer, and an ISP-50-8-R integrating sphere. X-ray reflectometry measurements were performed at the Rigaku SmartLab Diffractometer with Cu Kα source (λ = 1.54 Å) operating at 45 kV and 160 mA. Experimental data was background-corrected and a geometric footprint was applied due to the beam overspilling effects at small incident angles. The data was fitted using the reflectivity fitting package of the software GenX. [24] Secondary electron microscope (SEM) images were acquired with an AURIGA featuring a GEMINI column from Zeiss. Atomic force microscope (AFM) imaging was carried out using a NanoMan VS (Bruker) with Bruker silicon probes (OTESPA, spring constant of 26 N m −1 , resonance frequency of 300 kHz), operated in a tapping mode. 1024 × 1024 pixels resolution was used for measured surface of 2 × 2 µm 2 and processed data were acquired with Gwyddion 2.42.
IRAS Measurements: The measurements were performed in an UHV setup with a base pressure of 1.5 × 10 −10 mbar which is described in detail elsewhere. [25] The (Me 2 N) 4 Hf was deposited onto the Co 3 O 4 (111) surface from a well-sealed glass crucible maintained at room temperature. The crucible was separated from the UHV chamber by a gate valve, and (Me 2 N) 4 Hf was pumped into the UHV chamber via a separate bypass system. To avoid possible deposition on the chamber walls, the precursor was deposited via a stainless steel tube, which was placed directly in front of the sample surface. All spectra were recorded on a Fouriertransform infrared spectrometer (Bruker Vertex 80v) equipped with an external liquid nitrogen-cooled mercury/cadmium telluride detector. Both the spectrometer and detector were connected to the UHV chamber via differentially pumped KBr windows. For the isothermal IRAS measurement at 150 K, the IR spectra were continuously recorded (4 cm −1 , 60 s/ spectrum) during the exposure of (Me 2 N) 4 Hf (1.0 × 10 −8 mbar, 20 min). The TP-IRAS spectra (4 cm −1 , 60 s) were continuously recorded during the heating of the sample with an applied heating rate of 2 K min −1 . To correct the spectra for intensity losses caused by a decreased reflectivity at elevated temperatures during the heating process, all TP-IRAS spectra were scaled to the spectrum prior to the sample heating. [26] The spectrum of the clean as-prepared Co 3 O 4 (111) surface was used as a reference spectrum (4 cm −1 , 10 min) for all spectra recorded.
Temperature-Programmed IRAS Spectra during Alternating Exposures: First, the as-prepared sample was exposed to D 2 S for 5 min (1.0 × 10 −6 mbar). Subsequently, the sample was exposed to 20 alternating cycles of (Me 2 N) 4 Hf/pumping and D 2 S/pumping during sample heating, applying the following procedure for one cycle: Exposure of (Me 2 N) 4 Hf (5.0 × 10 −7 mbar) for 3 min, then 2 min of pumping. Exposure of D 2 S (1.0 × 10 −6 mbar, Sigma-Aldrich, 97 atom% D) for 3 min, then 2 min of pumping. For data analysis, again the procedure by Xu et al. was applied in this experiment. [26] DFT Calculations: DFT calculations were performed using the TURBOMOLE suite v7.2. [27] The exchange correlation functional of Perdew, Burke and Ernzerhof [28] was used with the def2-TZVP basis set of Weigend and Ahlrichs. [29] Dispersion interactions were accounted for applying the D3 correction scheme of Grimme. [30] Calculations were accelerated through the RI-J approximation. [31] Vibrational frequencies were calculated within the harmonic approximation.
Patterning of SAM Substrate: 10 nm thick aluminium oxide (AlO x ) was deposited on the polished Si wafer via an ALD process. The substrate was treated with O 2 plasma (Pico, Diener Electronic GmbH) under 0.2 mbar of O 2 pressure for 5 min and subsequently immersed into 0.2 mm ethanolic solution of PAC 12 HS for 18 h, yielding thiol functionalized selfassembled monolayer (SAM) formation. Then, photoresist (MICROPOIT S1813) patterns were created via conventional photolithography process, followed by O 2 plasma etching (0.2 mbar, 5 min) to remove unprotected HS-SAM area. Immediately after plasma treatment, the substrate was immersed into a solution of PAC 12 F 21 (0.2 mm in i-PrOH, 18 h) to fill up the created vacancy area of the substrate with fluorinated SAM. Treatment of the substrate with an acetone for 10 min removed the protective photoresist pattern and yielded substrate with a collinear defined pattern of SH and F terminated SAM molecules.
Growth of HfS 2 : The HfS 2 films for characterization were grown on a homemade ALD reactor. (Me 2 N) 4 Hf was delivered from a cylinder maintained at 60 °C and was pulsed into the reaction chamber (heated to 100 °C) with a continuous N 2 flow as the carrier gas. The pulse, exposure, and purge durations were 1.4, 20, and 30 s for (Me 2 N) 4 Hf and 0.2, 20, and 30 s for H 2 S (3% in N 2 ). This cycle was repeated until the desired thickness was reached.
Caution: H 2 S is a flammable, highly toxic gas that requires particular safety measures including, but not limited to, working in a ventilated fume hood, scrubbing excess of it on the exhaust line, and monitoring air quality with a dedicated sensor!

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.