Realization of vertical metal semiconductor heterostructures via solution phase epitaxy

The creation of crystal phase heterostructures of transition metal chalcogenides, e.g., the 1T/2H heterostructures, has led to the formation of metal/semiconductor junctions with low potential barriers. Very differently, post-transition metal chalcogenides are semiconductors regardless of their phases. Herein, we report, based on experimental and simulation results, that alloying between 1T-SnS2 and 1T-WS2 induces a charge redistribution in Sn and W to realize metallic Sn0.5W0.5S2 nanosheets. These nanosheets are epitaxially deposited on surfaces of semiconducting SnS2 nanoplates to form vertical heterostructures. The ohmic-like contact formed at the Sn0.5W0.5S2/SnS2 heterointerface affords rapid transport of charge carriers, and allows for the fabrication of fast photodetectors. Such facile charge transfer, combined with a high surface affinity for acetone molecules, further enables their use as highly selective 100 ppb level acetone sensors. Our work suggests that combining compositional and structural control in solution-phase epitaxy holds promises for solution-processible thin-film optoelectronics and sensors.

H eterostructures constructed from layered materials such as graphene, metal chalcogenides and black phosphorus (BP) have aroused particular interest due to their combined advantageous features and the emergence of unusual properties/functions [1][2][3][4][5][6] . As a group of the mostly studied layered materials, transition metal chalcogenides (TMCs) like MoS 2 and WS 2 exist in different crystal phases such as the 2H and 1T polytypes with distinct electronic properties. Their semiconductor-to-metal transition (i.e., 2H-to-1T) can be realized via Li-intercalation, 7 strain engineering, 8 e-beam/laser irradiation 2 or doping/alloying 2,[7][8][9] . This attractive feature allows for the formation of 2H/1T (semiconductor/metal) phase junctions that exhibit much reduced contact resistance compared to that of using noble metal contacts (e.g., Au) which normally interface with semiconducting TMCs with large Fermi level misalignment 2,10 . Besides TMCs, post-TMCs such as SnS 2 and InSe are another important group of layered materials, exhibiting attractive electronic and optoelectronic properties for a wide range of applications, including transistors, photodetectors, and sensors 11,12 . Similar to many other semiconducting materials, the type of metal contact with post-TMCs plays a critical role in tuning their functional performance 13 . However, due to the less metallic nature of post-transition metals, post-TMCs are generally semiconductors regardless of their crystal phases, such as the 2H, 4H, and 1T polytypes 14 . Consequently, contacting post-TMCs with metallic layered materials to achieve low interfacial resistance remains a big challenge.
To date, much effort has been devoted to the preparation of various heterostructures based on layered materials, which display different geometric arrangements such as the lateral and vertical heterostructures 1,3,15 . In particular, to prepare vertical heterostructures in which dissimilar layered crystals are stacked one above the other in a pre-designed sequence, solid-state procedures, including dry transfer, chemical vapor deposition (CVD) and chemical vapor transport (CVT) method have mostly been applied 3,16,17 . This is because these methods allow good control over the spatial arrangement of the layers as well as the deposition sequence. Compared to these solid-state methods, solutionphased approaches are advantageous in terms of relatively easier procedures, low-cost setups, and most importantly, scalability 18,19 . However, direct wet-chemical growth of vertical heterostructures of layered metal chalcogenides has thus far been challenging.
In view of their potential applications, heterostructures/heterojunctions such as InSe/graphene, MoTe 2 /MoS 2 , MoS 2 /perovskite, and graphene/MoS 2 /graphene have recently shown promising performance in photodetectors, due to the improved charge separation/transport and enhanced light adsorption [20][21][22][23] . Meanwhile, development of gas sensors for detection of volatile organic compounds are important in applications such as environmental monitoring and non-invasive diagnosis of diseases based on breath analysis 24,25 . Chemiresistive sensors based on metal oxides/sulfides have been used for detection of volatile organic compounds, however, high operating temperatures (typically ≥150°C) are usually required to achieve good sensitivity and selectivity 24,26 . Very recently, layered materials such as SnS 2 , WS 2 and Ti 3 C 2 T x have demonstrated great potential for roomtemperature gas detection [27][28][29] . It is expected that creation of heterostructures may realize further improved sensing performance [30][31][32] .
In this contribution, nanoplates of SnS 2 , a typical n-type semiconductor, are used as synthesis templates for the surface deposition of layered Sn 0.5 W 0.5 S 2 nanosheets, which show 83% metallic phase, leading to the formation of metal/semiconductor vertical heterostructures. Kelvin probe force microscope (KPFM) and tunneling atomic force microscopy (TUNA) analyses suggest the formation of ohmic-like contact at the Sn 0.5 W 0.5 S 2 /SnS 2 interface. The resultant heterostructures are fabricated into chemiresistive sensors to detect acetone at room temperature and exhibit a good selectivity and a minimum detectable concentration down to 100 ppb. The good sensing performance could be attributed to the low charge transfer resistance at the Sn 0.5 W 0.5 S 2 / SnS 2 interface that enables a much increased signal-to-noise ratio, and the alloying induced enhancement in surface gas adsorption.

Results
Synthesis and characterizations of Sn 0.5 W 0.5 S 2 /SnS 2 . As a representative post-TMC, SnS 2 has been widely studied and applied in phototransistors and sensors for its favorable band structure and relatively high surface electronegativity 12,33 . Typically, SnS 2 nanoplates were synthesized via a hydrothermal reaction with thiourea (CS(NH 2 ) 2 ) and tin tetrachloride hydrate (SnCl 4 •5H 2 O) as the precursors for S and Sn, respectively (see the Methods section for the detailed procedure) 34 . As shown in the scanning electron microscope (SEM), transmission electron microscope (TEM) and atomic force microscopy (AFM) images in Supplementary Figs. 1 and 2, hexagonal nanoplates with edge lengths ranging from 200 to 700 nm and an average thickness of 43 nm were obtained. As confirmed by the selected area electron diffraction (SAED) and X-ray diffraction (XRD) analyses, the nanoplates are α-SnS 2 with a 1T structure (space group P 3m1), where a = 3.65 and c = 5.90 (ICSD no. 42566) 12 6 to the precursors of the aforementioned synthesis solution, alloyed Sn 1-x W x S 2 nanosheets were in-situ synthesized and hybridized with SnS 2 nanoplates as shown in Fig. 1a, b. These heterostructures show an average lateral size of 750 nm (Fig. 1a) and an average thickness of 60 nm ( Fig. 1c and Supplementary Fig. 3). The side-view TEM image in Fig. 1d clearly shows a SnS 2 nanoplate covered by Sn 1x W x S 2 nanosheets on both its basal faces forming a vertical heterostructure. The deposited Sn 1-x W x S 2 nanosheets are typically 6-9 nm in thickness ( Supplementary Fig. 4). Energy dispersive X-ray spectroscopy (EDX) mapping of a typical heterostructure reveals the distribution of Sn, W, and S elements (Fig. 1e), in which the center of the heterostructure shows a higher concentration of Sn compared to the edge. EDX line analysis of the cross-section of a typical heterostructure, which was prepared by cutting with focused ion beam (FIB), further indicates that the SnS 2 nanoplate was covered by Sn 1-x W x S 2 nanosheets (Fig. 1f). EDX spot analyses on edges of Sn 1-x W x S 2 nanosheets suggest x ≈ 0.5, confirming Sn 0.5 W 0.5 S 2 nanosheets were obtained ( Supplementary Fig. 5).
The structural properties of the heterostructures were investigated with XRD and high resolution TEM (HRTEM) analysis. In the XRD pattern in Fig. 2a, besides the characteristic peaks for 1T-SnS 2 , two relatively broader peaks observed at 8.9°and 17.8°could be attributed, respectively, to the (001) and (002) planes of Sn 0.5 W 0.5 S 2 nanosheets with an enlarged interlayer spacing of about 1.0 nm. The enlarged spacing, which was also observed in side-view HRTEM images ( Supplementary Fig. 6), may result from the use of CS(NH 2 ) 2 and (NH 4 ) 10 H 2 (W 2 O 7 ) 6 in our synthesis, leading to the intercalation of NH 4 + ions in between the adjacent layers 35 . This was also confirmed by X-ray photoelectron spectroscopy (XPS) analysis ( Supplementary Fig. 7). The crystal structure of a typical heterostructure was further investigated by taking its SAED pattern. As shown in Fig. 2b, two sets of patterns both along the [001] zone axis show the epitaxial registry. The hexagonal pattern with six inner and six outer spots can be assigned to 1T-SnS 2 , and the measured (100) lattice spacing is 3.16 Å, in good agreement with the XRD result (Fig. 2a). The other set of pattern for Sn 0.5 W 0.5 S 2 shows elongated spots, forming discontinued ring segments. This may result from the curled structure of the Sn 0.5 W 0.5 S 2 nanosheets as well as possible misorientation. The measured lattice spacing for Sn 0.5 W 0.5 S 2 (100) planes was 3.0 Å, corresponding to a lattice parameter of a = 3.46 Å. This reasonably falls in between that of WS 2 (a = 3.16 Å) and SnS 2 (a = 3.65 Å). Note that the mismatch between the (100) planes of SnS 2 and Sn 0.5 W 0.5 S 2 is 5%, which can be tolerated in van der Waals epitaxial growth of layered materials 36 . The top-view HRTEM image in Fig. 2c distinctly shows the relatively darker region for the SnS 2 nanoplate covered with Sn 0.5 W 0.5 S 2 , and the brighter region for the periphery Sn 0.5 W 0.5 S 2 nanosheet. The lattice fringes extend continuously from the SnS 2 /Sn 0.5 W 0.5 S 2 center to the Sn 0.5 W 0.5 S 2 edge, further confirming the epitaxial growth mode. Moiré patterns could be observed in some areas due to the overlap of Sn 0.5 W 0.5 S 2 and SnS 2 at small misorientation angles. For example, as shown in Fig. 2d, a Moiré pattern with a periodicity of 4.0 nm was observed when the Sn 0.5 W 0.5 S 2 overlayer made a misorientation angle of 3°with SnS 2 (see the detailed analysis in Supplementary Fig. 8). Note that the overlapping of two hexagonal lattice patterns normally produces a hexagonal Moiré pattern 37 , which was not observed in the present work. This may be due to the fact that the Sn 0.5 W 0.5 S 2 nanosheets showed lattice distortion with varied interlayer spacings (0.6-1.0 nm, Supplementary Fig. 6), and thus were deviated from being perfectly flat on the SnS 2 nanoplate. As a result, only short-range line-like Moiré patterns were observed 38 . High resolution scanning TEM (STEM) images of the edge area of a heterostructure show the 1T-phase-like atomic arrangement of Sn 0.5 W 0.5 S 2 ( Supplementary  Fig. 9) 39 . It is interesting that zigzag lattice patterns for 1T' or T d phases which have been previously observed in TMCs such as WS 2 and WTe 2 40,41 were not observed in Sn 0.5 W 0.5 S 2 . Based on our density functional theory (DFT) calculation results shown in Supplementary Table 1 and Supplementary Fig. 10, the W-S bonds tend to be shorter than the Sn-S bonds in the alloyed system, resulting in a distortion from the perfect in-plane symmetric 1T lattice. The distorted 1T structure was also reflected in the Raman spectrum in Supplementary Fig. 11, where in addition to the peaks at 310, 351, and 414 cm −1 that correspond to the SnS 2 -like A 1g , WS 2 -like E 2g and WS 2 -like A 1g modes, respectively, the active modes observed at 171 cm −1 and 224 cm −1 in the lower frequency region could be attributed to distorted 1T-WS 2 12,35 . The dominant distorted 1T-phased W-S coordination was further confirmed by XPS analysis (Fig. 2e). It has been reported that the XPS band positions of a metal element are sensitive to their oxidation states, coordination geometries, and Fermi levels 42,43 . Normally in the Wf spectrum, the doublet peaks (32.1 and 34.2 eV) associated with W in the 1T-phased structure are downshifted by 0.6 eV relative to those associated with W in the 2H structure (32.7 and 34.8 eV). The deconvolution of the W 4f bands thus could enable the quantitative estimation of the 1T and 2H phase concentrations 7 . In our case, from the deconvoluted peak areas (Fig. 2e), the concentrations of the distorted 1T and 2H phases were calculated to be 83% and 17%, respectively. For the Sn 3d spectrum, the peak positions match well with the 1T-SnS 2 structure (Fig. 2f) 44 .
The formation of 1T or distorted 1T structures have been observed previously in TMCs when they were intercalated with alkali metal ions (e.g., Li + , K + etc.) 39,45 Fig. 12) 48 , which were subsequently heated up to 220°C under hydrothermal conditions. After the reaction had proceeded for 12 h at 220°C, nanorods with lengths of 10-100 nm were observed together with SnS 2 nanoplates in the solution (Fig. 3a, b). These nanorods are alloyed oxide of Sn and W with a formula of Sn 0.17 WO 3 based on XRD pattern (ICSD No. 38043 [https://icsd.fiz-karlsruhe.de], Fig. 3c) 49 , EDX analysis ( Supplementary Fig. 13a, b) and HRTEM imaging ( Supplementary Fig. 13c). The reason why WS 2 was not produced at this stage is that the bond energy of Sn-S was likely to be lower than that of W-S due to the larger ionic radius of Sn 4+ compared to that of W 4+50,51 . As the reaction proceeded further, the amount of the Sn 0.17 WO 3 nanorods decreased, and nanosheets started to form on the surfaces of the SnS 2 nanoplates (Supplementary Fig. 14). Evidently, the evolution of the XRD patterns of the intermediate products indicates a decrease of the Sn 0.17 WO 3 amount over time (Fig. 3c), which is accompanied with an increase of the Sn and W ion concentration in the solution based on the inductively coupled plasma mass spectrometry (ICP-MS) measurements (Supplementary Table 2). Based on our control experiments, Sn 0.17 WO 3 nanorods could form at 180°C and decompose at temperatures above 200°C ( Supplementary  Fig. 15). This suggests that the Sn 0.17 WO 3 nanorods forming at the beginning of the hydrothermal reaction gradually decomposed at 220°C (step 1 in Fig. 3d), providing additional Sn and W ions with a high W/Sn ratio (>30) (Supplementary Table 2). Such a high W/Sn ratio could drive the growth of alloyed Sn 0.5 W 0.5 S 2 nanosheets on the surface of SnS 2 even though the formation of W-S bond is less favored compared to that of the Sn-S bond (step 2 in Fig. 3d). In addition, we also tried to extend the Sn-W binary system to Sn-Mo system, but found that, under certain conditions, Sn 1-x Mo x S 2 nanosheets grew epitaxially on SnS 2 mainly via the edge growth. This phenomenon might be due to the different synthesis energies required for basal growth or edge growth 3 , which requires our further investigation.
Electronic properties of Sn 0.5 W 0.5 S 2 /SnS 2 . Alloying has been a powerful approach to tune the bandgaps of TMCs or to achieve their semiconductor-to-metal transitions 9,52 . However, alloying between layered TMC and post-TMC has not been reported so far, and thus the electronic properties of such alloys could be of great interest for potential applications. DFT calculations were performed to understand the electronic properties of the distorted 1T-Sn 0.5 W 0.5 S 2 deposited on 1T-SnS 2 with and without NH 4 + intercalation. The detailed results of the optimized crystal structures, calculated Bader charges, band structures and density of states (DOS) are shown in Fig. 4 1T-SnS 2 exhibits intrinsic metallic behavior with observable DOS at the Fermi level, dominantly contributed from the W and S atoms and slightly from the Sn atoms (Fig. 4d). The calculated Bader charge of W atoms in distorted 1T-Sn 0.5 W 0.5 S 2 is 0.20 e higher compared to that in 1T-WS 2 , whereas the Bader charge of Sn atoms in distorted 1T-Sn 0.5 W 0.5 S 2 is 0.10 e lower compared to that in 1T-SnS 2 (Fig. 4a, c and Supplementary Fig. 16). This suggests a charge redistribution in the Sn 0.5 W 0.5 S 2 alloy by charge transfer from W to Sn atoms 53 . Similarly, the distorted 1T-Sn 0.5 W 0.5 S 2 with intercalated NH 4 + ions (20 mol%) also shows the metallic behavior ( Supplementary Fig. 17). To experimentally verify the calculated results, the electronic properties of the SnS 2 nanoplates and Sn 0.5 W 0.5 S 2 /SnS 2 heterostructures were measured by fabrication of back-gated thin film field effect transistors ( Supplementary Fig. 18). The drain current (I d ) vs. drain-source voltage (V ds ) curves at varied gate voltages reveal that the SnS 2 nanoplates are typical n-type semiconductors ( Supplementary  Fig. 18a, b) 12 . In contrast, the I-V curves of films prepared from Sn 0.5 W 0.5 S 2 /SnS 2 heterostructures are almost insensitive to gate voltages ( Supplementary Fig. 18c, d), suggesting the metallic charge transport through the Sn 0.5 W 0.5 S 2 nanosheets, consistent with the theoretical predictions.
To exmamine the interface property between a SnS 2 nanoplate and the surface deposited Sn 0.5 W 0.5 S 2 , its surface potential, which corelates to its work function, was analyzed by KPFM in air 21 . A PtIr tip was used as the probe, and a SiO 2 /Si substrate sputtered with Au/Cr with a theoretical work function (ɸ Au ) of 5.100 eV was used as the potential reference. The 2D potential image of a typical Sn 0.5 W 0.5 S 2 /SnS 2 heterostructure is shown in Fig. 5a, where the color variation reflects the local surface potential difference (ΔV = ɸ Au − ɸ sample ) (details on the calculation of the surface potentials are given in the Methods section). It can be seen that the surface potential of the Sn 0.5 W 0.5 S 2 /SnS 2 heterostructure is 0.029 V lower than that of Au to give an estimated work function of 5.071 eV (Fig. 5a). On the other hand, an SnS 2 nanoplate exhibits a work function of 5.110 eV (Fig. 5b). The decrease in surface potential after deposition of Sn 0.5 W 0.5 S 2 on SnS 2 suggests that the metallic Sn 0.5 W 0.5 S 2 possesses a lower work function than that of the ntype SnS 2 . Although based on DFT calcuations, the work function of distorted 1T-Sn 0.5 W 0.5 S 2 is 5.62 eV which is much higher than that of SnS 2 , it can be substantially lowered to 2.87 eV by introducing an NH 4 + intercalation with a molar concentration of about 20% (Supplementary Table 1). Such work function modulation induced by doping or chemical absorbates has also been reported previously 54 . This also explains why the experimentally measured work function of Sn 0.5 W 0.5 S 2 nanosheets with partial NH 4 + intercalation was lower than that of SnS 2 . Therefore, an ohmic contact should form at the Sn 0.5 W 0.5 S 2 /SnS 2 heterointerface (Fig. 5c), affording a low charge transfer resistance. This was further confirmed by measuring I-V curves on individual SnS 2 nanoplates or Sn 0.5 W 0.5 S 2 /SnS 2 heterostructures deposited on highly oriented pyrolytic graphite (HOPG) with TUNA (Fig. 5d, e). The I-V curve for a SnS 2 nanoplate is highly asymmetric and the onset of the current rectification is at 3.50 V (Fig. 5d), indicating the presence of Schottky barrier at the PtIr tip/n-type SnS 2 interface, provided that the work function for PtIr, HOPG, and SnS 2 is 5.50, 4.60, and 5.11 eV, respectively 55,56 . In sharp contrast, the I-V curve for Sn 0.5 W 0.5 S 2 /SnS 2 is almost linear and symmetric with respect to 0 V, suggesting that the contact at SnS 2 /Sn 0.5 W 0.5 S 2 interface is ohmic-like (Fig. 5e) 57 .
Sn 0.5 W 0.5 S 2 /SnS 2 for photodetectors. The advantage of the facile charge transport across the ohmic-like heterointerfaces was demonstrated in fabrication of thin film photodetectors based on the Sn 0.5 W 0.5 S 2 /SnS 2 heterostructures. Figure 6a shows the I-V curves of the device under 405 nm laser illumination with power  intensity varied from 0.45 to 1.05 mW. A clear rise of the photocurrent with increasing illumination intensity was observed, indicating the effective conversion of photon flux to photogenerated carriers. In addition, the Sn 0.5 W 0.5 S 2 /SnS 2 photodetector showed symmetric and linear I-V plots, which is in sharp contrast to the non-linear I-V curves observed for SnS 2based device (Supplementary Fig. 19a). This further indicates the low resistance contact formed between the semiconducting and metallic components in Sn 0.5 W 0.5 S 2 /SnS 2 . The temporal photoresponse of the photodetectors was measured as well as shown in Fig. 6b, c and Supplementary Figs. 19b, c. The Sn 0.5 W 0.5 S 2 /SnS 2 photodetector showed an abrupt rise of photocurrent with a fast response time of 42.1 ms (defined as the time required to increase 90% from the minimum to maximum current density), which is comparable and outperforms some previously reported TMC based photodetectors 58,59 . This value is also about 50 times shorter than that of the SnS 2 -based photodetector (2.10 s) (Supplementary Fig. 19c). Such markedly shortened response time suggests the rapid transport of charge carriers across the Sn 0.5 W 0.5 S 2 /SnS 2 heterointerfaces 60,61 . Note that a relatively large dark current and thus a much reduced on/off ratio were observed for the Sn 0.5 W 0.5 S 2 /SnS 2 -based device as compared with the SnS 2 device. This was due to the metallic nature of the Sn 0.5 W 0.5 S 2 nanosheets. The similar phenomenon was reported previously in photodetectors based on graphene composites 62,63 .
Sn 0.5 W 0.5 S 2 /SnS 2 for gas sensing. The presence of a metallic component in a thin film channel may pose some limitation to its optoelectronic performance, such as the relatively large dark current observed in the aforementioned Sn 0.5 W 0.5 S 2 /SnS 2 photodetector. However, the metallic structure might become an advantage for applications like sensors. As another proof of concept demonstration of the advantage of solution-processible functional materials, the Sn 0.5 W 0.5 S 2 /SnS 2 heterostructures were deposited on Au interdigitated electrodes to fabricate chemiresistive gas sensors for detection of volatile organic compounds such as acetone, which is a potential biomarker for diabetes and lung cancer 64 . For comparison, SnS 2 -based gas sensors were also fabricated. The response-recovery curves of the gas sensors were measured under gas flows with increasing concentrations (typically 0.1-50 ppm) at room temperature (Fig. 7a, b and Supplementary Figs. 20 and 21). Taking sensing of acetone for example, the resistance of the Sn 0.5 W 0.5 S 2 /SnS 2 sensor decreased upon exposure to acetone and the decrease in resistance (ΔR = R a -R 0 ) with increasing acetone concentration (Fig. 7a, b and Supplementary Fig. 20). A minimum detectable concentration of 0.1 ppm was achieved. This is 20 times lower compared to that of the sensor based on SnS 2 which only afforded a minimum detectable concentration of 2 ppm ( Supplementary Fig. 21), and to the best of our knowledge, outperforms other reported metal sulfide/oxide chemiresistive acetone sensors. More importantly, our sensor showed the best sensitivity (i.e., sensing response, ΔR/R 0 ) at 100 ppb levels among all reported chemiresistive sensors operating at room temperature (Supplementary Table 3). Furthermore, as shown in Supplementary Fig. 22, a typical Sn 0.5 W 0.5 S 2 /SnS 2 sensor was repeatedly exposed to 1 ppm acetone and then back to N 2 gas for 10 cycles and showed an almost constant sensing response of 1.88 ± 0.07% (by taking standard deviation of the results from 10 cycles), indicating its good repeatability.   The improved sensitivity after deposition of Sn 0.5 W 0.5 S 2 nanosheets on SnS 2 nanoplates could be attributed to the following reasons. First, the formation of an ohmic-like contact between Sn 0.5 W 0.5 S 2 and SnS 2 that allowed rapid charge transfer across the metal-semiconductor interface partly led to a 35 times reduction in background noise and thus a much higher signal to noise (S/N) ratio ( Supplementary Fig. 23) 27 . Second, it can be noted that, the response/recovery time of the Sn 0.5 W 0.5 S 2 /SnS 2 gas sensor was longer than that of the SnS 2 gas sensor ( Fig. 7a and Supplementary Fig. 21), pointing to a chemical adsorption-related sensing pathway 64,65 . To confirm this, we calculated the adsorption energy of acetone on the different sensing materials (Fig. 7c). The adsorption energy of acetone on Sn 0.5 W 0.5 S 2 is 2.37 eV is much larger compared to that on SnS 2 (0.40 eV). This indicates that acetone molecules interact more strongly with Sn 0.5 W 0.5 S 2 . It is worth noting that the NH 4 + -intercalation did not significantly change the adsorption ability of Sn 0.5 W 0.5 S 2 nanosheets towards acetone, with the adsorption energy slightly reduced by 0.09 eV. As further shown in Fig. 7d, there is an obvious charge accumulation on Sn 0.5 W 0.5 S 2 due to charge transfer from the absorbed acetone molecule. Third, one of the advantages of using low-dimensional materials in gas sensing as compared with bulk crystals is their large specific surface areas, which are beneficial for providing large interfaces for channel-gas interaction 25 . Indeed, the deposition of wrinkled Sn 0.5 W 0.5 S 2 nanosheets on SnS 2 nanoplates increased the specific surface areas from typically 6.25 to 11.37 m²/g based on the Brunauer-Emmett-Teller (BET) measurements as shown in Supplementary Fig. 24. However, in spite of the beneficial effects from the large specific surface area of the wrinkled Sn 0.5 W 0.5 S 2 nanosheets and their strong interaction with acetone, increasing the amount of Sn 0.5 W 0.5 S 2 nanosheets on SnS 2 did not further improve the sensing performance, but on the contrary, led to poorer sensitivity with a minimal detectable concentration of only 1 ppm (Supplementary Fig. 25). This suggests that the concentration of the metallic phase present in the hybrid sensing film should not be too high, otherwise, the gas-induced doping effect on the semiconducting SnS 2 would be substantially weakened. Therefore, the presence of the semiconductor/metal heterostructures with a low charge transfer barrier, combined with sufficient active surfaces for strong gas adsorption are important in achieving low sensitivity in our thin film based gas sensors.
The selectivity of the Sn 0.5 W 0.5 S 2 /SnS 2 sensor was also investigated by comparing its sensing response towards acetone with other electron donating gases like diethyl ether and propanal, neutral gas like toluene, and electron withdrawing gas like NO 2 (Fig. 7e, Supplementary Fig. 26). The response of the Sn 0.5 W 0.5 S 2 /SnS 2 sensor towards diethyl ether and propanal was 10 times lower compared to that towards acetone at 2 ppm. This is largely due to the weaker electron donating ability of diethyl ether and propanal compared to that of acetone (Fig. 7e). Under the exposure of toluene, the sensor showed no response. In sharp contrast, when responding to NO 2 , an electron withdrawing gas, the sensor showed increased resistance (Fig. 7e, Supplementary  Fig. 26e), agreeing with the fact that SnS 2 is an n-type semiconductor (Supplementary Fig. 18b).

Discussion
The epitaxial growth of metallic Sn 0.5 W 0.5 S 2 nanosheets on the surfaces of SnS 2 nanoplates were realized via a solution-phase epitaxial deposition process. Importantly, an alloyed metal oxide (i.e., Sn 0.17 WO 3 ) was identified as an intermediate product which formed at 180°C and decomposed subsequently at 220°C to provide additional Sn and W ions with a high W/Sn ratio of about 30. We proposed that this high W/Sn concentration could drive the formation of alloyed Sn 0.5 W 0.5 S 2 nanosheets despite the fact that the formation of Sn-S bond requires less energy compared to that of the W-S bond. Although 1T-SnS 2 is a semiconductor, alloyed Sn 0.5 W 0.5 S 2 showed 83% distorted 1T structure, which is metal-like as predicted by theoretical calculations. KPFM and TUNA measurements on a Sn 0.5 W 0.5 S 2 /SnS 2 heterostructure suggested the formation of ohmic-like contact at the heterointerface, resulting in a low charge transfer resistance. The rapid charge transport at the Sn 0.5 W 0.5 S 2 /SnS 2 heterointerface allowed for the fabrication of fast photodetector with a short response time of 42 ms. Additionally, when the Sn 0.5 W 0.5 S 2 /SnS 2 heterostructures were fabricated into thin films for gas sensing, a much enhanced signal-to-noise ratio was achieved partly due to the presence of metallic Sn 0.5 W 0.5 S 2 layers. Furthermore, Sn 0.5 W 0.5 S 2 showed much enhanced surface adsorption of acetone than SnS 2 based on theoretical calculations. As a consequence, selective and sensitive detection of acetone was achieved with an ultralow minimum detectable concentration of 100 ppb at room temperature. This, combined with large sensing responses at 100 ppb levels outperforms previously reported room-temperature chemiresistive sensors. Our use of semiconducting SnS 2 nanoplates as synthesis templates for the solution-phase epitaxial growth of metallic Sn 0.5 W 0.5 S 2 nanosheets demonstrates a promising way towards the facile, economic and high-yield preparation of functional hybrid nanomaterials. Synthesis of layered materials through alloying among elements with distinct electronic and chemical properties is expected to bring about more materials and unusual phenomena. Teflon-lined stainless steel autoclave, heated to 220°C in an electrical oven and then maintained at this temperature for 60 h before being cooled down naturally to room temperature. The obtained product was centrifuged at 8000 rpm for 10 min, and the precipitate was washed with DI water for three times before further characterization.

Methods
Characterizations. Scanning electron microscope (SEM, JEOL JSM-7800F, Japan), transmission electron microscope (TEM, JEOL 2100Plus, Japan) and high resolution transmission electron microscope (HRTEM, JEOL 2100 F, Japan) coupled with energy dispersive X-ray (EDX) spectroscope were used to investigate the compositional, morphological and structural features of the samples. X-ray diffraction (XRD, Rigaku SmartLab, Japan) was performed using CuKα radiation (λ = 1.54 Å). X-ray photoelectron spectroscopy (XPS, PHI 5000 VersaProbe, Japan) measurements were conducted on the different metal sulfide nanostructures, and the binding energies were corrected for specimen charging effects using the C 1 s level at 284.6 eV as the reference. Raman spectra (Horiba HR800, France) of the samples were collected with a 532 nm laser. Semiconductor parameter analyzer (Tektronix Keithley 4200, America) and probe station (Lake Shore TTPX, America) were used to investigate the semiconductor properties of the samples. A commercial atomic force microscope (AFM, Dimension ICON with Nanoscope V controller, Bruker) was used to investigate the electrical properties of the individual nanostructures in air. Inductively coupled plasma mass spectrometry (ICP-MS, Agilent 7700×, America) was used to measure the concentration of Sn and W ions in the synthesis solution. Brunauer-Emmett-Teller (BET, Micromeritics, ASAP2460, USA) measurements were carried out to determine the specific surface area and pore size distribution of various samples.
Semiconducting property characterization. After Au (50 nm)/Cr (30 nm) drain and source electrodes were deposited onto a SiO 2 (285 nm)/Si substrate via thermal evaporation through a shadow mask, SnS 2 nanoplates or Sn 0.5 W 0.5 S 2 /SnS 2 heterostructures in water were drop-casted onto the substrate, acting as the channel to connect the drain and source electrodes with a channel length of 11 μm. The semiconducting properties of the channel materials were then characterized using a Keithley 4200 semiconductor characterization system operating at 77 K in vacuum (5 × 10 −5 Torr).
KPFM and TUNA measurements. After Au (50 nm)/Cr (30 nm) was coated onto a SiO 2 (285 nm)/Si substrate via thermal evaporation, SnS 2 nanoplates (or Sn 0.5 W 0.5 S 2 /SnS 2 heterostructures) in water were drop-casted onto the substrate. A KPFM (Dimension ICON with Nanoscope V controller, Bruker) was then used to characterize the surface potential of the SnS 2 nanoplates (or Sn 0.5 W 0.5 S 2 /SnS 2 heterostructures) at ambient conditions. The contact potential difference between the tip (PtIr) and the sample surface (V CPD ), which is also referred to as the surface potential can be calculated by using the following equations: where φ t , φ s , and φ f represent the work functions of the probe tip, the substrate, and the sample film, respectively. PeakForce Tunneling atomic force microscopy (TUNA, Dimension ICON with Nanoscope V controller, Bruker) was used to investigate the current-voltage (I-V) characteristics of individual SnS 2 nanoplates or Sn 0.5 W 0.5 S 2 /SnS 2 heterostructures. During the measurement, the PtIr tip was pressed against the sample with a constant force, feedback was switched to contact mode, and the voltage was linearly ramped up and down while the current signal was collected. Analysis of the I-V curves was performed with the Nanoscope Analysis software.
Fabrication of photodetector and photoresponse measurements. Au (15 nm in thickness) interdigitated electrodes with a 10 nm spacing were deposited onto a SiO 2 (300 nm)/Si substrate via magnetron sputtering through a shadow mask. After that, 0.5 μL of a concentrated dispersion (6.0 mg/mL) of the Sn 0.5 W 0.5 S 2 /SnS 2 heterostructures or SnS 2 nanoplates was drop-casted on the electrodes.
The current-voltage (I-V) and the current-time (I-t) curves of the photodetectors were measured on a semiconductor characterization system (Keithley 4200, USA) in air at room temperature. A 405 nm laser was used for all the measurements. The actual power intensity was measured by a power meter (LP1, Sanwa Electric Instrument Co., Ltd., Japan).
Fabrication of chemiresistive sensors and gas sensing tests. Chemiresistive gas sensors were fabricated based on SnS 2 nanoplates or Sn 0.5 W 0.5 S 2 /SnS 2 heterostructures for sensing of various gases, including CH 3 COCH 3 , (CH 3 CH 2 ) 2 O, CH 3 CH 2 CHO, C 6 H 5 CH 3 , and NO 2 . Typically, a drop of 100 μL aqueous solution containing 10 mM as-prepared SnS 2 or Sn 0.5 W 0.5 S 2 /SnS 2 was drop-casted onto an Au interdigitated electrode (with 0.1 mm spacing over a 2 × 1 cm 2 area, Changchun Mega Borui Technology Co., Ltd) and then dried in oven at 60°C. The gas sensing test was performed in an air-tight chamber with electrical feedthroughs at room temperature (25°C). A constant current was applied to the sensor electrode, and the variation of the sensor resistance was monitored and recorded with the changes in the gas environment using a data acquisition system (34972A, Agilent) with a 20 channel multiplexer (34901A, Agilent). A typical sensing measurement cycle consisted of three sequential steps: (1) a dried N 2 flow was introduced into the chamber to record a baseline resistance (R 0 ); (2) a target gas, e.g., acetone, balanced in N 2 was introduced, and the concentration increased (0.1-50 ppm) with progressive cycles; and (3) when the resistance of the sensor reached equilibrium in the target gas (R a ), the target gas was replaced by N 2 to allow the resistance of the sensor to return to R 0 . All gas flows were set at 500 sccm, precisely controlled by using mass-flow controllers.
Computational methods. All the computations were performed with Vienna Ab initio simulation package (VASP) which is based on the density functional theory (DFT) 66,67 . The exchange-correlation interaction uses the general gradient approximation (GGA) formulated by Perdew-Burke-Ernzerhof (PBE) 66 . All electron interactions were described with projector augmented wave (PAW) pseudo potentials. Long-range dispersion corrections have been considered within the DFT-D2 method. The dispersion coefficient C 6 and van der Waals radius R 0 for H, C, N, O, S, Sn and W used in our DFT-D2 method were taken from previous reports 68,69 . The scale factor S 6 was set to 0.75 because the GGA-PBE function was employed. An 11 × 11 × 1 k-point mesh was used for the interaction of the Brillouin-zone. The cutoff energy for the plane wave basis set was restricted to 400 eV, and a vacuum region of at least 12 Å was used in building the slab models. The convergence threshold was set as 10 −4 eV in energy and 0.02 eV/Å in force, respectively. As shown in Fig. 4c, a four-layer 1T-Sn 0.5 W 0.5 S 2 on a monolayer 1T-SnS 2 was constructed with a 2 × 2 supercell, which contains 24 Sn, 16 W and 80 S atoms, respectively. Three NH 4 + ions were introduced in the four-layer 1T-Sn 0.5 W 0.5 S 2 on a monolayer 1T-SnS 2 system (Supplementary Fig. 10). All atoms in the structure were fully relaxed to optimize without any restriction, and the convergence threshold was set as 10 −4 eV in energy and 0.02 eV/Å in force, respectively. The optimized lattice constants and work functions (Φ) were summarized in Supplementary Table 1. The experimental lattice constant (a = 3.46 Å) was used in DFT calculations for the in-plane periodicity of the four-layer 1T-Sn 0.5 W 0.5 S 2 on a monolayer 1T-SnS 2 without and with intercalated NH 4 + ions. To evaluate the stability of the adsorption of an acetone molecule on a three-layer 1T-SnS 2 and three-layer 1T-Sn 0.5 W 0.5 S 2 without and with intercalated NH 4 + ions (each system contains a 4 × 4 supercell), the adsorption was defined by △E a = E acetone + E substrate − Eacetone-substrate (where E acetone-substrate is the total energy of the acetone/ substrate compound systems, whereas E acetone and E substrate are the energy of the isolated acetone molecule, and the total energies of upper two-layer relaxed and bottom-layer fixed 1T-SnS 2 or distorted 1T-Sn 0.5 W 0.5 S 2 without or with intercalated NH 4 + ions systems, respectively. VESTA was used for preparation of the structure models 70 .

Data availability
The data that support the findings of this study are available from the corresponding author on request.