Extending the colloidal transition metal dichalcogenide library to ReS2 nanosheets for application in gas sensing and electrocatalysis

Among the large family of transition metal dichalcogenides (TMDCs), recently ReS2 has stood out due to its nearly layer-independent optoelectronic and physicochemical properties. These are related to its 1T distorted octahedral structure, which leads to strong in-plane anisotropy and the presence of active sites at its surface, which makes ReS2 interesting for applications such as gas sensors and catalysts for H2 production. However, the current fabrication methods for ReS2 use chemical or physical vapor deposition (CVD or PVD) processes that are costly and involve complex and lengthy fabrication procedures, therefore limiting their large-scale production and exploitation. To address this issue, we developed a colloidal synthesis approach, which allows the production of ReS2 to be attained at temperatures below 360 Celsius degrees and with reaction times<2 h, resulting in a more cost-efficient strategy than the CVD and PVD methods. By combining the solution-based synthesis with surface functionalization strategies, we demonstrate the feasibility of colloidal ReS2 nanosheet films for gas sensing of different toxic gases, moisture and other volatile compounds with highly competitive performance in comparison with devices built with CVD-grown ReS2 and MoS2. In addition, the integration of the ReS2 nanosheet films in assemblies, in which they are deposited on top of networks of carbon nanotubes, allowed us to fabricate electrodes for electrocatalysis for H2 production in both acid and alkaline conditions. Results from proof-of-principle devices show an electrocatalytic overpotential that is competitive with devices based on ReS2 produced by CVD, and even with MoS2, WS2 and MoSe2 electrocatalysts.

These are related to its 1T distorted octahedral structure, which leads to strong in-plane anisotropy and the presence of active sites at its surface, which makes ReS2 interesting for applications such as gas sensors and catalysts for H2 production. However, the current fabrication methods for ReS2 use chemical or physical vapor deposition (CVD or PVD) processes that are costly and involve complex and lengthy fabrication procedures, therefore limiting their large-scale production and exploitation. To address this issue, we developed a colloidal synthesis approach, which allows the production of ReS2 to be attained at temperatures below 360 °C and with reaction times < 2 h, resulting in a more cost-efficient strategy than the CVD and PVD methods. By combining the solution-based synthesis with surface functionalization strategies, we demonstrate the feasibility of colloidal ReS2 nanosheet films for gas sensing of different toxic gases, moisture and other volatile compounds with highly competitive performance in comparison with devices built with CVD-grown ReS2 and MoS2. In addition, the integration of the ReS2 nanosheet films in assemblies, in which they are deposited on top of networks of carbon nanotubes, allowed us to fabricate electrodes for electrocatalysis for H2 production in both acid and alkaline conditions. Results from proof-ofprinciple devices show an electrocatalytic overpotential that is competitive with devices based on ReS2 produced by CVD, and even with MoS2, WS2 and MoSe2 electrocatalysts.

Introduction
Transition metal dichalcogenides (TMDCs) are highly interesting and versatile materials due to their physicochemical properties that can be modified by exfoliation into single-or few-layered structures. [1][2][3][4] Due to quantum confinement and/or surface effects, such single and few-layered structures can behave differently from their bulk phase and manifest photoluminescence, [5] catalytically active sites [6] or strong light-matter coupling. [7] Moreover, depending on their composition and structure, they can be semiconductors (e.g. MoS2, WS2), metals (e.g. NbS2, VSe2) or even superconductors (e.g. NbSe2, TaS2). [1,8] Among the TMDCs, rhenium disulfide (ReS2) has recently emerged as an interesting material. [9][10][11] In contrast to the well-known group-VIB TMDCs (MoS2 or WS2), which present a hexagonal crystalline structure, ReS2 has an unusual distorted octahedral (1T) triclinic structure, leading to a strong in-plane anisotropy as the black phosphorous (BP). [9,10] However, unlike BP, ReS2 is stable in air. [9,10] The ReS2 structure consists of a P 1 ̅ symmetry with Re atoms forming Re4 clusters, interconnected in zig-zag chains, and distorted S6 octahedra that are formed upon cooperative atomic displacement (Scheme 1). [9], [11], [12,13] Moreover, ReS2 exhibits both metal-chalcogen and metal-metal bonds due to the extra valence electron of Re atoms, since Re belongs to the group-VII of elements. [9,[11][12][13] This structural anisotropy translates to electrical, optical, vibrational, [15] thermal [16] and physicochemical properties that do not depend significantly on the number of layers. For example, ReS2 is a direct-gap semiconductor (1.4-1.5 eV) from bulk to monolayer (0.7 nm) thicknesses, [15] unlike MoS2 or WS2 that instead present an indirect to direct band gap (Eg) transition when the thickness is reduced from bulk to a monolayer. [9,10] The predictability of the physical properties of ReS2 has made it attractive for a large variety of applications, in which it can be integrated as monolayer or few layer crystals in polarization-sensitive photodetectors, [17][18][19] field effect or heterojunction transistor structures, [20,21] and gas sensors, [25] [26] or as thick nanocrystals or nanocrystal sheets with some tens up to hundred layers in batteries, [24][25][26] solar cells, [27] and electrocatalysts. [28,29] Despite its favorable properties, the feasibility of ReS2 in technological and economic terms will strongly depend on the development of a scalable synthesis that can be readily integrated with the current device technology, being compatible with economic constrictions. In this regard, there are several challenges. First, Re is not an earth-abundant element, and therefore, its price (around 2800 €/kg [30] ), determined by its availability and the market demand, is high in comparison to Mo and W (less than 30 €/kg [31] ). Moreover, current methods for the production of ReS2 are mainly based on chemical vapor deposition [19,[32][33][34] (CVD), epitaxial growth [35] and the Bridgman method, [20] and rely on high process temperatures (from 450° to 1100 o C) for the precursor decomposition (e.g., the melting points of Re powder -3180 o C; Re-Te eutectoid 430 o C; ReO3 400 o C). [9,10] Moreover, they also imply the use of halogen vapor transport in the case of CVD [32] ; or HF treatments for cleaning in the Bridgman method [20] and as etchant to delaminate the ReS2 from the mica substrate in the epitaxial growth [35] . In addition, the aforementioned methods require long processing times from several hours for CVD [9,10] and epitaxial growth [9,10] up to several weeks for the Bridgman method. [9,10,20] Solution-based techniques such as chemical-intercalation [36] and liquid phase exfoliation (LPE) [37,38] have been developed as large-scalable routes for the ReS2 flakes production, but in most cases CVD-grown ReS2 is used as starting material, and therefore these methodologies inevitably still rely on expensive and demanding processing. To face the challenges related to the fabrication of ReS2, colloidal synthesis can be an alternative method for its large scale production, [39][40][41] achieving a compromise between crystal quality and physical properties that are needed for device applications. In fact, colloidal synthesis has already been successful in the development of other TMDCs, such as MoS2, [2,[42][43][44][45] MoSe2, [2,46,47] MoTe2, [48] WS2, [49] WSe2, [46] ZrS2, [50] TiS2 [50] and HfS2 [50] . Scheme 1. Sketches of the distorted 1T crystal structure of the ReS2 from different perspectives. Lattice orientations are indicated by the a-b-c axes. (a) A general view of a three-layer system, (b) side view in-plane of the layers; (c) top view of the three layers crystals where the Re4 clusters can be identified, (d) enlarged top view indicating the distances between the atoms. To obtain the crystal structure we used the crystallographic data from refs. [12,13] and the VESTA v.3.4.6 software [14] .
Transition metal dichalcogenides have already been explored as gas sensing materials showing high sensitivity, room temperature operation, facile processing and high resistance to degradation, in order to address the main drawbacks of the materials that are currently used in environmental monitoring such as metal oxides [51] , conducting polymers [51] and carbon nanotubes. [51] Recently, an interest in the use of ReS2 for the fabrication of gas sensors [22,23] has emerged due to its strong interaction with non-metal adatoms (H, N, P, O, S, F, etc.). In transistors built from Scotch-tape exfoliated ReS2, [22] a key role of the sulfur vacancies and interlayer interactions has been demonstrated to enhance the gas sensing capabilities, leading, for example, to better performance than MoS2- [52] and graphene-based detectors [53] for NH3 detection. In addition, ReS2 is also sensitive to other gases such as O2 or air, and can be applied in humidity sensors. [23] The presence of active sites at the surface promoted by the stable and distorted 1T structure of the ReS2 can be beneficial not only for the gas sensing application but also in electrocatalysis for the molecular hydrogen (H2) production from electrochemical water splitting, [25,36,54] an important and growing field due to the high energy density of ~120-140 MJ kg -1 and sustainability of H2. [55][56][57] Actually, before ReS2, other TMDCs have been reported as high-performance hydrogen evolution reaction (HER)-electrocatalysts. [6,[58][59][60] In particular, the group-VIB TMDCs have reached an advanced stage of development, [6,[58][59][60] exhibiting overpotentials inferior to 0.1 V at a cathodic current density of 10 mA cm -2 (ƞ10) in both acid [61,62] and alkaline [63], [64,65] electrolytes. However, only the metallic edge states of their trigonal prismatic (2H) phase can absorb H + with a small Gibbs free energy (ΔGH 0 ~ 0.08 eV), acting as active site for HER, [66], [67,68] while the basal planes are electrocatalytically inactive. [67,68], [69] Recent advances have shown that the HER activity of group-VIB TMDCs can be significantly enhanced when the semiconducting 2H phase of MoS2 is converted into metallic 1T (octahedral) phase. [70], [71] In this context, the stable and distorted 1T structure of the ReS2 represents an interesting electrocatalyst model, which can be advantageous compared to group-VIB TMDCs due to its metal-metal bonds. These bonds create a superlattice structure of Re chains that distort the octahedral structure of 1T phase (C3v symmetry). [15,28] Consequently, the Gibbs free energy of the hydrogen adsorption (ΔGH 0 ) on ReS2 basal planes can be as low as ~0.1 eV. [28] In fact, recent theoretical and experimental studies have shown that Re-Re bonds serve as electron reservoirs to originate an intrinsic charge distribution regulation, which tunes the Gibbs free energy of the hydrogen adsorption (ΔGH 0 ) on ReS2 basal planes towards the optimal thermoneutral value (i.e., 0 eV). [54] This value is comparable to that of metallic edge states of 2H-MoS2 (~0.08 eV). [28,72] Thanks to the HER-activity of its basal planes, the ReS2 should exhibit a higher number of active sites than group-VIB TMDCs. [28] In addition, in TMDCs such as MoS2 or WS2, the HER-activity strongly depends on the number of layers due to the dependence with the Eg, and the HER-activity weakens by passing from monolayer to few-layer materials as a consequence of an inefficient intra-flake electron transport via a hopping mechanism. [6] [69] In contrast, the nearly layer-independent properties of ReS2 make it a suitable catalyst, either monolayer or multilayer structures, [9,15] as it has been demonstrated in 3D structures composed by multilayer (≥17L) ReS2 flakes [28,29] .
In this work, we report a colloidal approach for the synthesis of ReS2 nanosheets, starting from ReCl5 and elemental sulfur as precursors. The obtained ReS2 sheets are tested as active material in gas sensors and HER. For the colloidal synthesis of the ReS2 nanosheets, we use elemental sulfur instead of CS2 [49][73] [42] or dodecanethiol (DDT) [73] [43] that are used as sulfur source for the colloidal synthesis of other TMDCs. [49][73] [42][73] [43] The advantage of using elemental sulfur relies on the fact that it is much less hazardous and/or toxic than the aforementioned precursors. We compare the properties of the colloidal ReS2 with their bulk and LPE counterparts, and find that the colloidal ReS2 nanosheets have a similar composition and Eg, but lower crystallinity. We fabricated gas sensors for a variety of agents by electrically contacting homogenous films of colloidal ReS2 sheets. With suitable ligands for the surface functionalization of the colloidal ReS2 nanosheets, we are able to enhance the gas sensor sensitivity to toxic gases (NH3) and humidity, compared to their non-functionalized counterparts. Sensitivity, recovery and time-response of the functionalized gas sensors based on colloidal ReS2 can compete with gas sensors fabricated from CVD-produced ReS2 reported in literature. [23] Concerning the HER activity in acidic media, a mixture of colloidally synthesized ReS2 in combination with carbon nanotubes, resulted in electrodes that can compete with the state-of-art CVD (KI)-aided ReS2. [54]

Results and discussion
The colloidal ReS2 sheets were synthesized from ReCl5 and elemental sulfur by a syringe pump method, [49] [47][42] [48] [44] in which the Re-precursor is added dropwise in a Soleylamine (S-OlAm) solution at 350°C. This procedure follows the use of chloride metal precursors [49][73] [42] [48] and the progressive injection method [49][47] [42][48] [44] for the colloidal synthesis of TMDCs, but it introduces elemental S as the sulfur source. The chalcogen source can play a role in the growth kinetics, in particular in the formation of H2S. Cheon's group demonstrated that sudden H2S influx, triggered by CS2, favors the formation of nanodisks, while continuous H2S formation with DDT leads to flakes. [73] With elemental sulfur as chalcogen source we obtained nanosheets, and therefore assume slow H2S kinetics in the oleylamine medium, with a release of ~75% over a 3h heating period. [74] For the nucleation and growth of the sheets, we add the ReCl5-oleic acid (OA) precursor dropwise, with a syringe pump, to the hot solution. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) images in Figure 1a-b (see also Figure S1a in Supporting Information, SI) show the formation of colloidal ReS2 nanosheets (hereafter, c-ReS2) with variable dimensions, formed by individual domains with a lateral size of ~4±1 nm. Moreover, some c-ReS2 nanosheets are found lying perpendicular to the support film (Figure 1c-d). From these images a thickness of ~0.4±0.1 nm is estimated for individual c-ReS2 nanosheets. This value is lower than the 0.7 nm reported for mechanically exfoliated monolayers in ReS2. [75] The discrepancy can be attributed to the heavy strain building up in these highly anisotropic colloidal 2D structures, leading to in-plane expansion of the lattice and contraction along the out-of-plane direction. [76] The obtained c-ReS2 nanosheets are comparable in morphology to other colloidal TMDC materials such as MoS2, [42] [43][45] WS2 [50] or WSe2 [46] reported in literature. The ReS2 sheets prepared by LPE, as control material, do not show aggregation as demonstrated in the SI, Figure S2a-b. Regarding the nanosheet formation, previous studies in the CVD production of ReS2 proved that there are two growth mechanisms: a fast one in (100) direction, and a slow one in (020) direction. To obtain nanosheets with a 1T distorted octahedral structure, the growth rate in both (100) and (020) directions should be comparable, [19] which may be achieved by oleic acid, oleylamine and Cl -(from the ReCl5) acting as ligands that passivate different crystal facets. In order to gain insight into the crystal structure, we compare the selected-area electron diffraction (SAED) patterns from c-ReS2 nanosheets and LPE-ReS2 flakes showing comparable 2θ values for Bragg peak positions (see SI, Figure S3 for the TEM-SAED patterns), which correspond to the 1T distorted octahedral structure. Furthermore, broader diffraction features characterize the c-ReS2 nanosheets, due to the much smaller size of single-crystal domains (few nm or less in the colloidal sample vs. hundreds of nm for LPE flakes). The composition of the as-prepared c-ReS2 was investigated by X-Ray photoelectron spectroscopy (XPS), as shown in Figure 2a-b, revealing an atomic stoichiometry Re:S of 1:1.3, which is estimated from the Re 4f and S 2p spectra. This ratio is slightly smaller than would be expected from the 1:2 ReS2 stoichiometry. However, also the LPE fabricated samples and bulk ReS2 yield lower Re:S stoichiometric values of around 1:1.6 and 1:1.7, respectively (see also SI, Figure S1e for complementary elemental analysis). It is interesting to note that in the LPE and bulk cases, the S 2p signal can be decomposed in two different doublets, corresponding to two different chemical environments of sulfur. Similar results have been reported in Ref. [77] , and the two S components have been assigned to ReS2 (doublet at lower binding energy, with S 2p3/2 component at approx. 161.4 eV) and to S atoms that are not connected to Re atoms (doublet at higher binding energy, with S 2p3/2 component at approx. 162.1 eV). However, we cannot discard that the two S components come from the existence of different Re-S bond lengths in the distorted ReS2 structure, as illustrated in Scheme 1d. [12,13] For colloidal ReS2, a good fitting of the S 2p profile was obtained by using two doublets as for the LPE and bulkcounterparts, with S 2p3/2 components at ∼161.6 eV and ∼162.2 eV. Moreover, XPS peaks in c-ReS2 are broader than the ones for the LPE and bulk materials likely indicating a less crystalline structure in the colloidal case. [78] To support this statement, we also evaluated the extinction coefficient and Raman spectra that are plotted in Figure 2c and d-e. As expected for ReS2, [9,10][79] the material shows a strong and broad extinction from 300 nm to almost 1000 nm. However, in contrast to its LPE equivalent, there is no clear excitonic peak at ~810 nm that corresponds to the Eg of the ReS2. [79] The Eg was obtained from the (αh) n vs. h (Tauc plot) analysis (see Figure S1d and Figure S2c in SI) using the Tauc relation Ah = Y(h-Eg) n , in which A is the absorbance, h is Planck's constant, ν is the photon's frequency, and Y is a proportionality constant. [80] The value of the exponent denotes whether it is a direct transition (n = 2) or an indirect one (n = 0.5). [81] Since ReS2 should be a direct bandgap semiconductor, we applied n = 2, resulting in an Eg of ∼1.41 eV (colloidal) and ∼1.43 eV (LPE). Both values are in agreement with the theoretical (1.41 eV) [82] and experimental values (1.4-1.5 eV) from bulk to monolayer (0.7 nm) thicknesses. [15] The analysis of the Raman spectra in Figure 2d-e has been performed following the detailed reports on ReS2 carried out by Balicas' and Terrones' groups, [83,84] which identified 18 first-order modes in the 100-450 cm -1 range (see SI for Raman peaks interpretation). Since in the randomly oriented assembly of the nanosheets we cannot control the crystal orientation, we performed a comparative Raman analysis with excitation at two different wavelengths, i.e., 532 and 785 nm. Independently of the excitation wavelength and synthesis batch, the Raman peaks from the colloidal sample are broader than the corresponding ones in the LPE sample. Such broader linewidth in the Raman signal of the colloidal sample indicates a lower degree of crystallinity compared to the LPE one, [85] which can be related to the difference in the temperature that is needed for their preparation. Highly crystalline ReS2 fabricated by CVD requires a temperature > 600 o C, [35] while the colloidal synthesis is performed at 350°C. The low crystallinity of the colloidal ReS2 nanosheets might also be the cause for the absence of the excitonic peak in the spectrum in Figure 2c. For both colloidal and LPE samples, the following Raman modes are enhanced at 785 nm excitation compared to the 532 nm excitation, Ag 9 (273 cm -1 ), Ag 10 (279 cm -1 ), Ag 12 (305 cm -1 ), Ag 13-14 (322 cm -1 ), Ag 16 (366 cm -1 ), Ag 17 (373 cm -1 ), Ag 18 (405 cm -1 ) and Ag 3 (417 cm -1 ), while the Ag 15 (350 cm -1 ) and Ag 2 (440 cm -1 ) modes disappear, indicating that the samples are composed by few-layered (2L-4L) ReS2, in agreement with the data reported by Terrones' group. [84] In order to prove that our colloidal synthetic approach can be versatile also for other materials, we fabricated ReSe2 using Se powder as chalcogen source (see SI, Figure S4 for synthesis details and material characterization).
In order to test the suitability of the ReS2 sheets as films for gas sensing, and to compare the performance of the colloidal and LPE samples, we prepared films on glass substrates by drop-casting (Figure 3a). For the c-ReS2 nanosheets, we used the original dispersion for dropcasting, followed by annealing at 300°C in a glovebox (3 h) to remove the organic ligands from synthesis (oleic acid and oleylamine) with the aim to make the film electrically conductive. Films from LPE samples were fabricated by drop-casting without any further treatment. In this way, we obtained continuous and compact films with thickness of ~400 nm, as shown in the SEM images in Figure 3b-c (see Experimental Section for fabrication details). The devices were completed by the evaporation of Au electrodes as contacts. All devices manifest ohmic behavior (see SI, Figure S5a for conductance measurements under inert -N2-atmosphere), however the resistance of the devices made from the colloidal ReS2 is much lower (460 MΩ) than the one of devices fabricated with LPE-ReS2 (5.9 GΩ). For gas sensing via the detection of conductance changes of the film, we positioned the samples inside a chamber with controlled atmosphere connected to a probe station (see Figure 3a and the Experimental Section for details on the gas filling of the chamber and measurement protocols). From the gas flow rates we estimate the filling and purging time of the chamber to be around 1s, which is shorter than the response times that we measured on our devices. We tested the response to humidity, to NH3 that is a toxic agent, and to ethanol (EtOH) and acetone as representative volatile organic compounds. Figure 3d-f shows the representative response (ℝ) for different gases of the devices obtained via the relative changes of the film conductance (G) [51] in the presence of the target gas compared to inert (N2) atmosphere, ℝ = ( − )⁄ . The numerical values are reported in the SI in Table S1, in which also the response in terms of resistance is given for comparison with literature.
For NH3 and H2O the response of the gas sensors is positive, i.e. the conductance under the target gas is higher compared to that under the inert N2 atmosphere. This can be explained by physisorption of the gas molecules that act as electron donors. [22,23,51,86,87] The higher response for gas sensors based on ReS2 under NH3 compared to air and O2 was also reported in theoretical studies. [22] For NH3 detection, the relative response of the devices built from LPE-ReS2 flakes (hereafter, 'LPE devices') ( Figure 3f) is much higher compared to the ones fabricated from annealed c-ReS2 nanosheets (in the following noted as 'colloidal devices'). However, this result is mainly related to the very low electrical conductivity of the LPE device under inert atmosphere. In fact, the recovery time of the LPE devices to the initial currents (i.e., in the nA range) is of the order of hours, while the recovery time of the colloidal devices is of the order of seconds (see SI, Table S2 for rise and fall times of conductance response of the different devices; and Tables S3 and S4 for estimated minimum amounts detectable with the technique). In practical terms this means that for detection with response times of few seconds the baseline of the LPE device is at a value of ~1000, and therefore the relative response to NH3 at such time scales is ~10, while that of the colloidal device is ~2.5 -3. Concerning humidity detection, the response of the devices made from colloidal ReS2 (ℝ = 1.2; rise time (τR = 8.7 s); fall time (τF = 0.3s)) outperforms clearly the ones based on LPE flakes (ℝ = 8.8; τR = 5.2 s; τF = 31s). In fact, the recovery time to return to 10% of the response after the gas flow stopped is one order of magnitude shorter for c-ReS2 compared to the LPE flakes one. For EtOH, acetone and compressed air the conductance of the devices made from c-ReS2 decreases compared to one achieved in inert N2 atmosphere (Figure 3e). This result can be rationalized by oxidation of the film due to the incorporation of electron acceptor molecules. [51,86,87] Furthermore, the relative conductivity change of the devices made from c-ReS2 exposed to EtOH, acetone and compressed air is much weaker (ℝ = -0.15; ℝ = -0.12; and ℝ = -0.06) and slower in recovery (EtOH: τR = 5.2 s; τF = 25s; Acetone: τR = 1 s; τF = 56s; and air τR = 2.3 s; τF = 12s) as compared to their exposure to NH3 (ℝ = 2.4; τR = 9.3 s; τF = 1.3 s) or H2O (ℝ = 1.2; τR = 8.7 s; τF = 0.3 s). Representative SEM images of the colloidal and LPE-ReS2-based gas sensors, respectively. The films (~400 nm thickness) were prepared by drop-casting the corresponding dispersions on glass substrates. (d) Representative gas induced-time response of the device built with ann-ReS2 in two consecutive cycles exposed to NH3 (left) and H2O (right). (e) Representative gas induced-time response of the device built with ann-ReS2 exposed to different gases: NH3, H2O (left), EtOH, acetone and dry air (right). (f) Representative gas induced-time response of the device built with LPE-ReS2 exposed in two consecutive cycles to NH3 (left) and H2O (right). In all the cases, the gas-induced response was determined from the film conductance variation as detailed in the text.
The circumstance that the colloidal ReS2 nanosheets are passivated by organic ligands opens the possibility to manipulate the film conductivity by ligand exchange with other molecules. We therefore use this approach to improve the performance of the c-ReS2 based sensors, in terms of recovery time and response. In particular, we performed a ligand exchange process that replaces the long-chain organic ligands used in the synthesis (oleic acid and oleylamine, both with a C18 aliphatic chain) with shorter molecules that contain Cx aliphatic chains with x<4 or an aromatic ring (see Experimental Section for more details). Shorter ligands lead typically to stronger coupling of the nanomaterial in a compact film, which increases the charge carrier mobility. [88,89] Moreover, chemical modification of the nanostructure surface by ligand exchange can modify the electronic properties of the material, as well as their reactivity with other functional groups such as gas molecules. [51,86] We tested different short chain molecules in this respect: 3-mercaptopropionic acid (MPA), 1,4benzenedithiol (BDT) and 4-aminobenzoic acid (ABA), and performed the ligand exchange in solution. The films from the ligand exchanged solutions were prepared with the protocol described before, which ensured a compact film with a thickness of ~400 nm (Figure 4a-b, see Figure S6 in the SI for additional characterization). A larger S contribution in the XPS signal (compared to the samples before ligand exchange) that results from the thiol groups of the exchanged molecules confirmed the presence of MPA and BDT molecules. Concerning the exchange with ABA, the success of the ligand exchange process can only be assessed from the increase of the conductivity, since the characteristic N 1s XPS peak (NH2, at 400.5 eV) also originates from oleylamine. The ligand-exchanged films also show ohmic conductivity, and their resistance strongly depends on the individual ligands (see SI, Figure S5b for conductance measurements under inert -N2-atmosphere). Here, devices made from MPA-exchanged c-ReS2 nanosheets manifested the highest conductance (R = 120 MΩ), followed by ABA (R=510 MΩ) and BDT (R = 200 MΩ). Compared to the original device discussed in Figure 3, we achieved an increase in electrical conductivity by a factor of 4 with the MPA ligands. We tested the ligand-exchanged films for gas sensing with the same procedure as described before. Overall, we observe similar behavior before and after the ligand exchange process, with an electrical current increase under NH3, H2O and CO2, and, instead, decrease under EtOH, acetone and air Figure 4c depicts the response of the device built from MPA-exchanged c-ReS2 nanosheets to different gases: NH3, H2O and CO2 that lead to current increase, and EtOH, acetone and air that result in a small current decrease. Therefore, we can assume that in all the cases, colloidal, LPE and ligand exchange fabricated devices, the sensing mechanism is based on a charge transfer between the physisorbed gas molecules and the ReS2 structure. [51,87] Figure 4d shows the response of the gas sensors made with ligand-exchanged c-ReS2 nanosheet films when exposed to NH3. In this context, the device made from MPA-exchanged c-ReS2 nanosheets shows the highest response (ℝ=31, with a response time of 3.3 s), followed by the ones built from BDTand ABA-exchanged c-ReS2 nanosheets. The functionalization of c-ReS2 nanosheets with BDT yields a higher sensitivity (ℝ=24 response for NH3 detection) than ABA (ℝ=16 response for NH3 detection), although the conductance of ABA (RABA=510 MΩ vs RBDT=200 MΩ) is higher under inert atmosphere. The different results in terms of sensitivity and responsivity obtained without and with the functionalization of c-ReS2 nanosheets, and even more, by varying the molecules used for the functionalization, point to the importance of: (i) possible sites for the binding of gas molecules; (ii) the efficiency of charge transfer for the gas sensing sensitivity, and (iii) a good film conductivity, which is beneficial for a fast response and recovery. Here, we can expect a trade-off between a film with high surface area that is beneficial for interaction with the gas molecules, and a compact film that results in high conductivity and from which a fast response can be expected. In particular, the high sensitivity obtained with the BDT and MPA ligand exchange for NH3 (ℝBDT=24 and ℝMPA=31, respectively) and H2O (ℝBDT=2.4 and ℝMPA=4.5, respectively) points to a beneficial role of the SH groups of MPA and BDT for physisorption of the gas molecules and the electron transfer towards the ReS2 film. We note that the performance of the device made from MPA-exchanged c-ReS2 nanosheets is superior to the one built with LPE-ReS2 in Figure 3 in terms of sensitivity and response time (e.g. for NH3 detection, ℝ=9.9, with a response time of 5.2 s for LPE-ReS2 and ℝ=31, with a response time of 3.3 s for MPA-exchanged c-ReS2).
The gas sensor built from MPA-exchanged c-ReS2 outperforms, to the best of our knowledge, those reported in literature fabricated with CVD-grown ReS2. [23] The humidity sensor reported by Yang et al. [23] made by CVD-grown ReS2 has a resistance variation of ~-60% at a relative humidity of 70 %, with a response time of the order of tens of seconds, while our gas sensor fabricated with MPA-exchanged c-ReS2 reaches a variation of 80% with much faster response time (4.1 s) and comparable recovery time (1 s). Gas sensing (NH3, O2, air) has been also reported from phototransistors made from micro-mechanically exfoliated ReS2 flakes, [22] with a response time in the order of ms. However, this response time is related to the illumination of the device and therefore does not directly compare to our devices that work in dark condition. Concerning the gas sensors for NH3 and EtOH based on other 2D materials, we note that our MPA-functionalized sensor has a faster response time and better recovery of the initial conductance state than devices that use micro-mechanically exfoliated or CVD-grown MoS2. [  Finally, we tested c-ReS2 films as an electrocatalytic film for HER (Figure 5a). To make the c-ReS2 more electrochemically active, we isolated the nanosheets from the dispersion by centrifugation and solvent evaporation. The obtained powder was then treated by thermal annealing (Ar flow, 500 °C) to remove the organic ligands from synthesis, and thereby expose the catalytically active surface of the nanosheets. After the annealing, the ReS2 powder was dispersed in N-methyl-2-pyrrolidone (NMP) by ultrasonication, and the obtained dispersion was used for the deposition on glassy carbon (GC) rigid electrodes and single-walled carbon nanotubes (SWCNTs)-based flexible papers (i.e., buckypapers) (Figure 5b) (see Experimental Section for more details). The choice of the SWCNTs-based paper as support relies on our recent works, [61], [63] in which we demonstrated a long-range (≥ 1m) electrochemical coupling between HER-active TMDCs and SWCNTs for increasing the HER-activity of TMDCs. [61,63] Moreover, the porosity of such substrate promotes the adhesion of the TMDC films without the need of ion conducting catalyst binders [61], [63], [92] [93] (e.g., Nafion in acid solution [94] and Tokuyama AS-4 in the alkaline one [95] ), which can be detrimental to the electrocatalytic activity of the catalyst. [96], [97] The GC-based electrodes were obtained by drop-casting the c-ReS2 dispersions at a mass loading of 0.13 mg cm -2 . The hybrid SWCNTs/c-ReS2 electrodes were produced through a sequential vacuum filtration deposition of the material dispersions onto nylon membranes (material mass loading of 1.5 mg cm -2 for both SWCNTs and c-ReS2 (5 mg of each material)). Figures 5c-d display top-view and cross-sectional SEM images of a representative SWCNTs/c-ReS2 electrode. The c-ReS2 nanosheets form a film atop the SWCNTs network, and the electrode shows a bilayer architecture with ~20 m-thick SWCNTbased collector and a thin c-ReS2-based active film (thickness in the order of 1 m). The highmagnification SEM image (Figure 5c) reveals the presence of ReS2 aggregates with various dimensions, in the 0.5-10 µm range. The smallest ReS2 aggregates (lateral size < 500 nm) penetrate the mesoporous SWCNT network (see SI, Figure S9). For such configuration, SWCNTs increase the electron accessibility to the HER-active sites of c-ReS2 nanosheets, speeding up the HER-kinetics compared to flat GC. [98][61] [92][99] Figure 5e shows a top-view photograph of the as-produced SWCNTs/c-ReS2 (electrode area of ~3.5 cm 2 ), in which the presence of ReS2 aggregates is visible by eye. Such aggregates are also evident in SEM images (Figure 5f) with a lateral size ranging from 50 to 200 m. The HER-activity of the SWCNTs/c-ReS2 was tested in both in acidic (0.5 M H2SO4) and alkaline (1 M KOH) media. In principle, the HER in acidic solution proceeds with an initial discharge of the hydronium ion (H3O + ) and the formation of atomic H adsorbed on the electrocatalyst surface (Hads), in the so-called Volmer step (H3O + + e -⇄ Hads + H2O), followed by either an electrochemical Heyrovsky step (Hads + H3O + + e -⇄ H2 + H2O) or a chemical Tafel recombination step (2Hads ⇄ H2). In alkaline media, the Hads is formed by discharging H2O (H2O + e -⇄ Hads + OH -). Then, either a Heyrovsky step (H2O + Hads + e -⇄ H2 + OH -) or a chemical Tafel recombination step (2Hads ⇄ H2) occurs. Apart from the cathodic current density ƞ10, the Tafel slope is also an important Figure of Merit to assess the HER-activity. [100] However, a rigorous kinetic analysis of the HER establishing the Tafel slope was not carried out for our electrodes, since SWCNTs hold a high surface area that leads to a remarkable capacitive current density (in the order of 1 mA cm -2 ) even at a low LSV sweep voltage rate (≤5 mV s -1 ). This can be the cause of misleading interpretations of the estimated kinetic parameters. [63], [101] Figure 6a-b show the current-resistance (iR)-corrected polarization curves of GC/c-ReS2 and SWCNTs/c-ReS2 electrodes in acidic (0.5 M H2SO4) and alkaline (1 M KOH) aqueous solutions, respectively. The series resistance that arises from the electrical resistance of the working electrode and the electrolyte resistance, i.e., R, was extrapolated by single frequency electrochemical impedance spectroscopy (EIS) (see Experimental Section for further details). The curves measured for the substrates, i.e., GC and SWCNTs, are also shown as references. The curve measured for commercial platinum on carbon (Pt/C) is reported as benchmark for HER. The c-ReS2-based electrodes show an enhanced HER-activity compared to those of the substrates, i.e., GC and SWCNTs. Noteworthy, the use of SWCNTs as the substrate significantly increases the cathodic current densities, i.e., the HER-activity, compared to the one shown by CG/c-ReS2. Thus, in acidic media, ƞ10 decreases from 0.465 V for the GC/c-ReS2 to 0.196 V for SWCNTs/c-ReS2. In alkaline solutions, GC/c-ReS2 electrode is poorly HERactive (ƞ10 > 0.7 V). Interestingly, SWCNTs/c-ReS2 electrode exhibits a ƞ10 of 0.299 V, indicating that the SWCNTs and c-ReS2 nanosheets interact to synergistically enhance the HER-activity of the electrode. This effect is attributed to the activity of the SWCNTs for initiating the H2O discharge, thus accelerating the Volmer reaction on the c-ReS2 nanosheets in alkaline conditions. [102] In addition, it is worth noticing that SWCNTs/c-ReS2 displays an HERactivity in acidic media comparable to the one shown by SWCNTs/LPE-ReS2 (ƞ10 = 0.192 V), which, however, shows a better HER-activity in alkaline media (ƞ10 = 0.238 V). The morphological and electrochemical characterization of SWCNTs/LPE-ReS2 electrode is reported in the SI (see Figure S10). Electrochemical impedance spectroscopy measurements do not show any significant differences between SWCNTs/c-ReS2 and SWCNTs/LPE-ReS2 (see Nyquist plot in Figure S11). Moreover, we estimated the double-layer capacitance (Cdl) of the c-ReS2 and LPE-ReS2 films deposited on GC from cyclic voltammetry (CV) measurements (see SI for further details, Figure S12a) Beyond the electrocatalytic activity, the durability is another important criterion for the exploitation of an electrocatalyst. Figure 6c-d shows the chronoamperometry measurements (current retention vs. time) for the SWCNTs/c-ReS2, in acidic and alkaline solutions, respectively. A constant overpotential was applied in order to provide the same starting cathodic current density of 20 mA cm -2 for HER. Glassy carbon rod has been used as counter-electrode to avoid Pt dissolution/re-deposition effects altering the HER-activity of working electrode in presence of Pt-based counter electrodes. In both acidic and alkaline media, SWCNTs/c-ReS2 exhibit a nearly stable behavior over 48 h (current retention of 94.3% and 98.0%, respectively). The slight fluctuation of the HER-activity can be tentatively attributed to morphological changes of the electrode film during the HER process. Similar effects have been previously observed in electrocatalysts based on metallic TMDCs, such as TaS2 [103] and NbS2 [93,104] nanosheets, as well as in other 2D material-based electrodes, including graphene-based electrocatalysts. [105] In fact, the mechanical stresses originated by H2 bubbling can cause a reorientation/fragmentation of the 2D electrocatalysts, which, consequently, shows higher electrochemically accessible surface area. [103,104,106] Moreover, such mechanical stresses occasionally caused the formation of cracks in the active material films, as observed by SEM image analysis of additional electrodes (Figures S13). This led to a significant change of the cathodic current of the electrodes over time in both acidic and alkaline media. For example, in acidic media some electrodes reported an increase of the initial HER-activity over 24 h, achieving a ƞ0 = 0.162 V vs. RHE (see Figure S14a). This behaviour might be attributed to the progressive increase of the H + accessibility to the HER-active sites of the electrodes. [103,104,106] In alkaline media, the electrodes often displayed an increase of the electrical resistance (e.g., from 4.4 Ω to 45.9 Ω, measured by single frequency EIS), which caused a decrease of 21.5% of the initial current density (see Figure S14b). Despite the decrease of the HER-activity of the electrode, the iR-corrected polarization curves shows an electrochemical activation of the c-ReS2 nanosheets after 24 h (ƞ10 reduced from 0.327 V to 0.181 V), similarly to the electrodes tested in acidic media. The use of electrocatalyst binders, such as sulfonated tetrafluoroethylene-based fluoropolymer copolymers (e.g., Nafion), could prospectively "freeze" an optimized electrode morphology of the current electrodes, as well as an increase of the electrode durability under HER-operation.   [67,98,[107][108][109] WS2 [110] and MoSe2 [61] ) and phosphides (NiCo2Px). [111] In particular, in acidic solution, ƞ10 of SWCNTs/ReS2 (0.196 V, 0.162 V after 24 h-stability test) approaches the best values obtained by ReS2-based electrocatalysts (0.147 V), [54] which come from a defect-activated monolayer ReS2 grown by CVD. These results indicate that there is still room for improvement in our system by properly optimizing the material surface in future experiments, as it has also been demonstrated for other TMDCs [107,112,113] . Additionally, the HER-performance of our electrocatalysts are comparable or superior to those previously reported for electrocatalysts based on other TMDCs (ƞ10 values between 0.170 and 0.372 V), concretely, in absence of either mechanical strain or doping effects, including the most studied MoS2, [67,98,[107][108][109] WS2 [110] and MoSe2 [61] . The relevant electrocatalytic properties of ReS2 have been recently attributed to the presence of metal-metal bonds (not present for group-VIB TMDCs), [9,[11][12][13] which are highly HER-active sites in presence of vacancies causing an intrinsic HER-promoting optimization of the electronic charge distribution. [54] Moreover, we reported here also the HER-activity of ReS2-based electrocatalyst in alkaline conditions, providing new insight for the development of pH-universal electrocatalysts to be exploited as cathode materials in current large-scale H2 production technologies, [114], [115], [116] e.g., chloro-alkaline systems [117] or alkaline zero-gap water electrolysis units, [118] and proton exchange membrane (PEM) electrolysis. [119], [120] [36] Defects-activated monolayer ReS2 0.147 0.5 M H2SO4 Monolayer [54] Vertically Oriented Arrays of ReS2 Nanosheets ~0.3 0.5 M H2SO4 ~0.67 [25] Lithiated Vertically Oriented Arrays of ReS2 Nanosheets ~0.2 0.5 M H2SO4 ~0.67 (excluding Li) [25] ReS2 3D reticulated vitreous carbon foams 0.336 a) 0.5 M H2SO4 Not reported (ReS2 grown onto carbon foam) [28] One-pot synthetized ReS2 >0.35 0.5 M H2SO4 0.02 [121] MoS2 > 0.35 0.5 M H2SO4 Monolayer [107] MoS2 with S vacancies 0.25 0.5 M H2SO4 Strained MoS2 with S vacancies 0.17 0.5 M H2SO4 Co-doped MoS2 0.159 0.5 M H2SO4 0.5 [112] Reduced graphene oxide:MoS2 hybrid on GC ~0.15 0.5 M H2SO4 0.285 [113] Reduced  [108] Solvothermal produced MoS2 on GC 0.252 0.5 M H2SO4 ~0.5 [109] Liquid-phase exfoliated MoSe2 0.37 0.5 M H2SO4 2 [61] SWCNTs/MoSe2 0.17 0.5 M H2SO4 2 [61] WS2 nanosheets > 150 0.5 M H2SO4 0.35 [110] NiCo2Px Nanowires 0.104 0.5 M H2SO4 5.9 [111] a) Data not iR-corrected

Conclusions
We report a bottom-up approach for the synthesis of colloidal ReS2 (c-ReS2) nanosheets that relies on low fabrication temperatures (below 360 °C) and process times inferior to 4 h. We use elemental sulfur as source, which is more accessible and environmentally friendly than other precursor materials currently used in the colloidal synthesis of different TMDCs. Dropcast c-ReS2-based films are tested as gas sensors for a variety of agents, achieving highly competitive performance after annealing or ligand exchange in comparison with devices built with CVD-ReS2 or -MoS2. Furthermore, the colloidal ReS2 nanosheets were tested as HERelectrocatalysts operating both in alkaline and acidic media. In particular, electrodes made by c-ReS2 films deposited on single-walled carbon nanotubes (SWCNT) (SWCNTs/c-ReS2) exhibit overpotentials at a cathodic current density of 10 mA cm -2 (ƞ10) of 0.196 V in 0.5 M in H2SO4 and 0.299 V in 1 M KOH. In acidic media, the ƞ10 of SWCNTs/c-ReS2 approaches the best values obtained by CVD-grown defect-activated ReS2-based electrocatalysts. In comparison with non-optimized electrocatalysts based on other group-VIB TMDCs, the ReS2based electrocatalysts here developed exhibit similar or even superior HER-activity. The c-ReS2 nanosheets can therefore present a promising solution for the fabrication of gas sensors and HER electrocatalysts, with still considerable room for improvement by exploiting device designs and finely tuned surface modifications.

Colloidal synthesis of ReS2 nanosheets.
The general procedure to obtain c-ReS2 consists in the use of two separated precursor solutions: one in the round-bottom flask and the second one added by syringe pump method. For the reaction, we used a metal:chalcogenide molar ratio of 1:6. The dispersion that contains the Re precursor (ReCl5 67 mg) is prepared by sonication in a bath at 60 o C (1h) with 500 µL OA + 2 mL ODE. This will be added by syringe pump (rate 2 mL h -1 ) in a hot medium under Ar flow (350 o C) of S-OlAm. The synthesis medium itself was prepared by mixing 35 mg S with 7 mL OlAm and degassing under Ar during 2h at 250 o C in a Schlenk line until a clear orange-brownish solution is achieved. The reaction temperature was set to 350 o C. The suspension as obtained from synthesis was purified twice (in air) by adding toluene, acetone and isopropanol (15:10:5 mL), followed by centrifugation (2599 g, Sigma 3-16P centrifuge, rotor 19776). The sheets were finally dispersed in toluene using a vortex.

TEM.
The morphology of the synthesized c-ReX2 nanosheets and LPE-ReX2 (X = S or Se) flakes was evaluated by transmission electron microscopy in a JEOL JEM-1011 microscope (W filament), operated at 100 kV. Samples were drop-cast on carbon-coated copper grids. Selected area electron diffraction patterns and overview TEM images were also acquired with a JEOL JEM-1400Plus TEM, with a thermionic source (LaB6 crystal), operated at 120 kV. Selected area electron diffraction pattern processing (azimuthal integration, background subtraction) was done using the PASAD plugin for Digital Micrograph. [122] High-angle annular dark-field scanning TEM and bright-field TEM imaging at higher magnification of the c-ReX2 nanosheets (X = S or Se) samples were carried out by using a FEI Tecnai G 2 F20, with Schottky emitter, operated at 200 kV. Statistics for the estimation of the lateral size and thickness were carried out considering 20 individual domains in the nanosheets for each sample.

AFM.
Atomic force microscopy (AFM) images were acquired by using a Nanowizard III (JPK Instruments, Germany) mounted on an Axio Observer D1 (Carl Zeiss, Germany) inverted optical microscope. The samples were prepared by drop-casting cReS2 flakes dispersion onto a freshly cleaved mica substrate (G250-1, Agar Scientific Ltd., Essex, U.K.) and the AFM measurements were carried out by using PPP-NCHR cantilevers (Nanosensors, USA) with a nominal tip diameter of 10 nm. Intermittent contact mode AFM images of 5×5 µm 2 and 1.5×1.5 µm 2 were collected with 1024 data points per line and the working set point is kept above 70% of the free oscillation amplitude. The scan rate for the acquisition of images was 0.9 Hz. Height profiles were processed by using the JPK Data Processing software (JPK Instruments, Germany) and the data were analyzed with OriginPro 9.1 software. 4.5. XPS. X-ray photoelectron spectroscopy was carried out in a Kratos Axis UltraDLD spectrometer using a monochromatic Al K a source (15 kV, 20 mA). High-resolution scans were performed at a constant pass energy of 10 eV and steps of 0.1 eV. The photo-electrons were detected at a take-off angle φ = 0° with respect to the surface normal. The pressure in the analysis chamber was kept below 7 × 10 -9 Torr. The binding energy scale was internally referenced to the Au 4f 7/2 peak at 84 eV. The spectra were analyzed using the CasaXPS software (version 2.3.16). 4.6. Raman spectroscopy characterization. Raman measurements were performed in a Renishaw inVia micro-Raman microscope equipped with a 50× (0.75 N.A.) objective with excitation wavelengths of 532 and 785 nm and an incident power ≤ 1 mW to avoid heating and damage of the samples. The samples were prepared by drop casting the diluted materials dispersion onto a Si wafer covered with 300 nm thermally grown SiO2 (LDB Technologies Ltd.). For each sample, at least 50 spectra were collected.

Fabrication of the gas sensors.
A ligand exchange procedure in solution was carried out with the as-prepared OA/OlAm-capped ReS2 nanosheets dispersed in toluene. The different ligands (MPA, ATP and BDT) were dissolved in MeOH at a concentration of 1 mM. The c-ReS2 dispersion in toluene (50 mg mL -1 ) and the ligand solution were mixed in a volume ratio of 1:1 and stirred 2 min in a vortex, following by centrifugation (5000 rpm) and removal of the supernatant. Then, procedure is repeated once more, with a subsequent purification step with toluene: MeOH twice and centrifugation. Finally, the ligand exchanged ReS2 nanosheets were dispersed in isopropanol at a concentration of 50 mg mL -1 , and immediately used for the preparation of the films on cleaned glass slides (ca. 1.5×1.5 cm 2 ) by drop-casting (100 µL). Prior to the deposition, the glass substrates were cleaned in an ultrasonic bath (8 min each step), first with acetone, followed by isopropanol, dried with a N2 flow, and plasma cleaning step under N2 (100W, 2 min). For the Au contacts deposition, we used a shadow mask with square holes of 1×1 mm 2 area, separated by 100 µm. The Au film with a thickness of 80 nm was evaporated in a Kenosistec® e-beam evaporator at a deposition rate of 0.3 Å s -1 and base pressure of about 1.0 10 -6 mbar. The thickness of the different films was measured with a Veeco Dektak 150 profilometer.

Electrical characterization of the gas sensors.
We carried out the electrical tests in a chamber (see Figure 3a) equipped with two inlets, one for the N2 and the other for the target gases. Side ports allows the access of micromanipulators connected to a Keithley 2612 sourcemeter. The chamber (~643 cm 3 ) has a transparent top lid, such that positioning of the tips is possible by eye. A N2 line, at a pressure slightly above ambient pressure (1.2 bar, flow of ~ 667 cm 3 s -1 determined with a flowmeter (Yokogawa ® )), provided the inert atmosphere used as reference and allows to fill and purge the chamber in ~1 s. A glass bubbler was connected and used for containing the water (MilliQ ® ), ammonia (2% vol aq. solution), EtOH and acetone while bubbling with the N2 flow used as carrier to create the saturated atmosphere. To ensure uniform evaporation of the volatile gases, the bubbler was kept in a warm water bath (60°C). For CO2 and compressed dry air, reservoirs connected to the experimental chamber were used. The experiments were performed according to the following procedure: current was measured as a function of time, first under N2 flux, then the test gas was injected in the chamber for 10s, then the flux was switched back to N2 until recovery of the base current. The procedure was repeated multiple times for each device to ensure reproducibility. 4.9. Fabrication of the electrocatalyst electrodes. The solvent was removed from the asprepared (c-ReS2) sample by bubbling with nitrogen. The obtained powder was annealed in a tubular furnace (PSC 12/--/600H, Lenton, UK, 25 mm inner tube diameter, Ar 100sccm) at 500 °C for 8 h. Then, the powder was dispersed in NMP at a concentration of 1 mg mL -1 and sonicated for 7 h. We prepared the GC/c-ReS2 electrodes by drop-casting the c-ReS2 NMP dispersion (0.2 mg) on GC substrates (Sigma Aldrich ® ) with geometrical area of 1×1.5 cm 2 (c-ReS2 mass loading ~0.13 mg cm -2 ). Electrodes of Pt/C were produced as benchmark for HER by depositing the Pt/C dispersion onto GC substrates. The Pt/C dispersion was produced by dissolving 4 mg of Pt/C (5 wt.% Pt loading, Sigma Aldrich ® ) and 80 µL of Nafion solution (5 wt.%, Sigma Aldrich ® ) in 1 mL of 1:4 v/v ethanol/water. The Pt/C mass loading of the electrodes was 0.262 mg cm -2 , in agreement with protocols developed previously. [102,105] . The SWCNTs/c-ReS2 electrodes were fabricated by sequential vacuum filtrations of 5 mg of SWCNTs (>90%, Cheap Tubes, sonic-tip de-bundled in NMP at a 0.4 mg mL -1 concentration [61,63] ) and 5 mg of c-ReS2 on nylon filter membranes (0.2 μm pore size, 25 mm diameter from Sigma Aldrich ® ). The as-produced films (geometrical area = 3.5 cm 2 , and c-ReS2 mass loading ~1.59 mg cm -2 ) spontaneously peels off the nylon membrane during the drying, resulting in self-standing electrodes ready to be used. The SWCNTs/LPE-ReS2 electrodes were produced using the same protocols described for SWCNTs/c-ReS2, except for the use of the asproduced LPE-ReS2 dispersion in isopropanol instead of c-ReS2 one. Additional electrodes were produced by depositing the c-ReS2 and LPE-ReS2 nanosheet dispersions onto GC by drop casting method (catalyst mass loading = 0.2 mg cm -2 ). All the electrodes were dried overnight at room temperature before their electrochemical characterization. 4.10. Electrochemical measurements of the electrodes. Measurements were carried out at room temperature in a flat-bottom quartz cell under a three-electrode configuration using an Ivium ® CompactStat potentiostat/galvanostat station controlled via IviumSoft ® . A Pt wire or a GC rod were used as the counter electrode and a KCl-saturated Ag/AgCl was used as the reference electrode. As aqueous medium, 200 mL of two aqueous solutions were used: 0.5 M H2SO4 (99.999%, Sigma-Aldrich) and 1M KOH (90%, Sigma Aldrich). MilliQ ® water was used to prepare the solutions. Oxygen was purged from electrolyte solutions by flowing N2 throughout the aqueous medium using a porous frit for 30 min before starting the measurements. A constant N2 flow was kept afterward for the whole duration of the experiments to avoid re-dissolution of O2 in the electrolyte. The potential difference between the working electrode and the Ag/AgCl reference electrode was converted to the RHE scale using the Nernst equation: = / + 0.059 + / 0 , where ERHE is the converted potential vs. RHE, EAg/AgCl is the potential experimentally measured against the Ag/AgCl reference electrode, and E 0 Ag/AgCl is the standard potential of Ag/AgCl at 25 °C (0.1976 V). The LSV curves were acquired at the scan rate of 5 mV s −1 . Polarization curves were iR-corrected, in which i is the measured working electrode current and the R is the series resistance that arises from the working electrode substrate and electrolyte resistances. R was extrapolated by the real part of the impedance (Re[Z]) measured by single frequency EIS at open-circuit potential and at the frequency of 100 kHz. Electrochemical impedance spectra of the SWCNTs/c-ReS2 and SWCNTs/LPE-ReS2 were acquired at open circuit potential at frequencies between 0.1 Hz and 200 kHz. Stability tests were carried out by chronoamperometry measurements (j-t curves) by measuring the current in the potentiostatic mode at a fixed overpotential in order to provide the same starting cathodic current density of 20 mA cm -2 for HER. The Cdl of the c-ReS2 and LPE-ReS2 films deposited onto GC were estimated by CV measurements in a non-Faradaic region of potential (i.e., potential between 0.30 and 0.45 V vs. Ag/AgCl) at various potential scan rates (from 20 to 800 mV s -1 ) in 0.5 M H2SO4.

SEM/EDX characterization.
We performed SEM/EDX measurements on the sensor devices and electrode samples using a Helios Nanolab 600 (FEI Company) microscope. SEM measurements were performed at 5 kV and 0.2 nA. For the cross-section images, we carefully cut the electrodes with a scalpel and measured in 90° tilted sample holder.

The table of contents entry:
Colloidal synthesis of rhenium disulfide nanosheets enables a simple and cost-effective exploitation of its peculiar layer-independent properties for gas-sensing and electrochemical H2 production. The surface functionalization of the nanosheets leads to sensitive and fast response gas sensors, while their assembly with carbon nanotubes enhances its electrocatalytic activity, making both device performance competitive with CVD rhenium disulfide.

Supporting Information Additional c-ReS2 material characterization
In contrast to the high-magnification transmission electron microscopy (TEM) images displayed in Figure 1a-b of the main text, which allow us to discern individual colloidal ReS2 (c-ReS2) nanosheets, low-magnification TEM images in Figure S1a, only show assemblies of nanosheets. We also found these assemblies formed by the c-ReS2 nanosheets as well as small aggregates (lateral size <100 nm), with thickness from 6 to 3 nm by atomic force microscopy (AFM) analysis, reported in Figure S1b-c. The height-profile statistical analysis of the small assemblies of c-ReS2 nanosheets (performed by excluding big aggregates) shows a mean value of 12.2 nm and a median of 7.8 nm. The data follow a lognormal distribution peaked at 4.3 nm with 23.8 % of nanosheets. Figure S1d shows the estimation of the band gap (Eg) from (αh) n vs. h (Tauc plot) analysis using the Tauc relation Ah = Y(h-Eg) n . With n=2 (ReS2 is a directgap semiconductor), we find an Eg value of about 1.41 eV for the colloidal nanosheets.
We carried out elemental analysis of the c-ReS2 by means of scanning electron microscopy (SEM) and energy dispersive X-Ray spectroscopy (EDS) in a Helios Nanolab 600 (FEI Company) microscope combined with an X-Max detector and INCA ® system (Oxford Instruments). For the EDS spectra acquisition and analysis measurement conditions were set at 15kV and 0.8nA. Results shown a stoichiometry ratio Re:S of 1:1.4 from the analysis of the Re and S peaks ( Figure S1e). As comparison, we also performed the SEM/EDS analysis in the liquid phase exfoliated (LPE) and bulk counterparts obtaining a stoichiometry ratio Re:S of 1:1.5 and 1:1.6, respectively. The values estimated in all cases are smaller than the expected 1:2 ratio for ReS2, and in agreement with the X-Ray photoelectron spectroscopy (XPS) data presented in the main text (Figure 2a-b).

Liquid phase exfoliation of ReS2 and characterization
ReS2 flakes (Figure S2a-b) were produced from bulk ReS2 crystals (HQGraphene ® ) by liquid phase exfoliation in isopropanol (IsOH) followed by sedimentation-based separation. [1][2][3] Briefly 20 mg of bulk ReS2 was added to 5 mL of IsOH and then, ultrasonicated (Branson ® 5800 cleaner, Branson ultrasonics) for 6h. The resulting dispersion was centrifuged for 20 min at 935g (Sigma 3-16P centrifuge, rotor 19776), in order to separate the exfoliated ReS2 flakes as supernatant from the un-exfoliated and thick ReS2 crystals remaining on the bottom of the vial. Only 80% of the supernatant volume is collected by pipetting, avoiding possible contamination from the precipitated powder. Figure S2c shows the estimation of the Eg from (αh) n vs. h (Tauc plot) analysis, [4] resulting in an Eg value of about 1.43 eV.

Supplementary SAED information
As mentioned in the main text, we evaluate the crystal structure of the samples from the TEM/selected-area electron diffraction (SAED) analysis. From the TEM images and the corresponding selected-area diffraction (SAED) patterns shown in Figure S3a-d for c-ReS2 and LPE-ReS2 samples, we obtained the background-subtracted and azimuthally integrated SAED patterns displayed in Figure S3e.

Discussion of the ReS2 Raman spectra
As mentioned in the main text regarding Figure 1d-e , the analysis of the Raman spectra has been carried out following the detailed reports on ReS2 performed by Balicas' and Terrones' groups. [5,6] Therefore, the in-plane mode near 150 cm −1 is labeled as Ag 1 , the out-of-plane mode around 437 cm −1 , where just the S atoms vibrate, is assigned as Ag 2 , and a quasi-out-of-plane mode that is denoted as Ag 3 can be identified around 418 cm −1 , in which the S vibrations occur at a small angle with respect to the basal plane. The remaining modes, in which the atoms vibrate at different angles due to the low symmetry of ReS2, are denoted as Ag 4 to Ag 18 with increasing frequency. With a 532 nm excitation wavelength, Figure 1e (main text), we observe the main Raman modes. [6] In the c-ReS2 , the Re in-plane vibrations (Ag 1 151 cm -1 , Ag 7 212 cm -1 , Ag 8 237 cm -1 ) and S vibrations (Ag 15 349 cm -1 , Ag 18 408 cm -1 , Ag 2 436 cm -1 ) are clearly visible. The LPE sample shows a similar behavior, in which Re in-plane vibrations (Ag 1 152 cm -1 ; Ag 6 162 cm -1 ; Ag 7 213 cm -1 ; Ag 8 235 cm -1 ) and S vibrations (Ag 15 346 cm -1 , Ag 18 at 407 cm -1 , Ag 3 418 cm -1 , Ag 2 437 cm -1 ) are clearly detected, but shows also two additional modes (Ag 6 162 cm -1 ; and Ag 3 418 cm -1 ) compared to the c-ReS2 that are probably due to the higher crystallinity of the LPE sample.

Characterization of colloidal ReSe2 nanosheets
As introduced in the main text, colloidal ReSe2 nanosheets (c-ReSe2) can be obtained from the same synthetic protocol than for the c-ReS2 by changing S powder for Se powder (86 mg, 99.99%, Strem Chemicals ® ) using the same ReCl5 (67 mg) as Re precursor, and setting the temperature to 350°C. As happened for ReS2, this temperature is lower than the 625°C established for the respective CVD approaches. [7] For comparative purposes, we prepared the corresponding LPE counterpart in IsOH following the same protocol described for ReS2 exfoliating the ReSe2 crystal supplied from HQGraphene ® .
The ReSe2 nanosheets obtained are assembled in form of flowers as can be observed in the corresponding low-magnification TEM images (Figure S4a). From high-magnification TEM images, a lateral size of ca. 4 ± 1 nm and thickness of 0.4 ± 0.1 nm is estimated for the tilted c-ReSe2 nanosheets. As happened for c-ReS2 nanosheets, this value is lower than the one reported for CVD-growth hexagonal ReSe2 monolayer (0.7 nm) [7] or CVD-growth distorted 1T ReSe2 monolayer (0.9 nm), [8] probably due to the heavy strain building up in these highly anisotropic 2D colloidal structures. [9] Moreover, the SAED patterns from c-ReSe2 nanosheets and LPE-ReSe2 flakes ( Figure S4c) indicate comparable 2θ values for Bragg peak positions, corresponding to the hexagonal structure. As for the c-ReS2, much broader peaks characterize the c-ReSe2 nanosheets, due to the much smaller size of single-crystal domains (few nm in the colloidal sample vs. hundreds of nm for LPE flakes). From the SEM/EDS analysis, Figure S4d, we estimated a Re:Se ratio of 1:1.5 (c-ReSe2), 1:1.5 (LPE-ReSe2) and 1:1.6 (bulk-ReSe2) from the Re and Se peaks. This ratio is slightly smaller than would be expected from the 1:2 ReSe2 stoichiometry, and in accordance with the elemental analysis data obtained also for the ReS2 samples. We also evaluated the extinction coefficient and Raman spectra that are plotted in Figure S4e and f. As expected for ReSe2, the material shows a strong and broad extinction from 300 nm to almost 1000 nm. However, in contrast to its LPE equivalent, there is no clear excitonic peak at ~920 nm that corresponds to the Eg of the ReSe2. [10] The Raman spectra in Figure S4e shows the presence of the characteristics Raman modes of ReSe2 in the 100 to 300 cm -1 range. Since as reported by Wolverson and collaborators, [11] the relative intensities of the bands are dependent on the orientation of the flakes and do not scale proportionately with thickness, we just carried out a qualitative interpretation of the Raman spectrum of c-ReSe2 in comparison with its LPE and bulk counterparts. As can be seen in Figure S4e, the Raman peaks from the colloidal sample are broader than the corresponding ones in the LPE sample, indicating, as for the c-ReS2 case, a lower degree of crystallinity compared to the LPE one. [12] The low crystallinity of the c-ReSe2 nanosheets might also be responsible for the absence of the excitonic peak in the spectrum in Figure S4d, and attributed to the lower synthesis temperature in the colloidal approach (350°C) in comparison with CVD approach (625°C [7] ).

Gas sensors: I-V curves, SEM-EDX supplementary characterization, device performance for functionalized c-ReS2
For comparison, the I-V characteristics under inert N2 of the different gas sensing devices prepared were collected as shown in Figure S5a  In order to obtain films without holes or cracks for the fabrication of the gas-sensing devices, we tested different spin-coating and drop-casting procedures. Scanning electron microscopy (SEM) images in Figure S6a show the formation of holes in the film prepared with MPA-exchanged c-ReS2 by using spin coating, which can affect the sensor performance. Therefore, we prepared the films by drop-casting. With this approach we obtained a more homogeneous films independent of the functionalization molecule as shown in the SEM images in Figure S6b-c for devices built with ABA-and BDT-exchanged c-ReS2, respectively.  Table S1 summarizes the performance of the different gas sensors in terms of conductance (G) [13] and resistance (R) response. Table S2 reports the response and recovery times determined as the time needed to reach 90% of response in the presence of gases, and to return to 10% of the response, respectively. [13] For the recovery time measurement, a continuous N2 flow was applied to purge the gas chamber. In order to estimate the minimum amount of gas (NH3 or H2O) detectable according with the design of the technique, we have calculated the minimum concentration (amount of gas) that we can detect taking into account the noise level of our device in terms of current read (Inoise). Therefore, we consider that, for the current corresponding to the minimum amount of gas Igas,min, we have I gas,min − I inert ≥ I noise I noise , was estimated by measuring the current under inert atmosphere for 10 s (Iinert) and calculating the standard deviation. Then, we assume a linear trend of the response with the concentration of gas (Q); that is, I gas − I inert = αQ and the sensitivity α can be calculated from the measurement reported in Figures 2d-f and 3cd and in the SI, Table S1 as: α = I gas − I inert Q 0 Where Q0 is the reference gas concentration, which for H2O is 95% relative humidity (measured with a humidity sensor at the inlet, and equivalent to 23080 ppm on volume [14] ), while for we use a NH3 aqueous solution at 2% in volume (thus, ca. 500 ppm). With these assumptions, the minimum gas quantity Qmin can be determined as: In the Tables S3 and S4, it is reported the Qmin values for NH3 and H2O detectable for the different devices according with the design of the technique. For the stability-retention tests, the measurements comprised 3 steps: (i) inert (N2) gas flow; (ii) target gas flow of about 10s; and (iii) flow paused during 10-15s, while the device remains exposed under an environment of the target gas (indicated as "env"). As can be seen in Figure S7 for the device fabricated with MPA-exchanged c-ReS2 under different gases: NH3, H2O, EtOH and acetone, the conductance increases under the presence of NH3 and H2O (or decreases for EtOH or acetone) with the gas inlet, keeping the modification during the first seconds of the flow paused entering in a slight fluctuation due to the absorption/desorption equilibrium. As soon as the N2 flow is again introduced, the desorption of the gas molecules occurs, reflected in the decrease for NH3 and H2O (or increase for EtOH or acetone) of the conductance up to get back the initial state. To demonstrate the sensitivity of the devices built from ABA-and BDT-exchanged c-ReS2, Figure S8 includes the gas-response determined from the conductance under the exposure of NH3 and H2O. As pointed out in the main text and Table S1, the higher response is achieved under the exposure to NH3 gas, in agreement with theoretical calculations on sensors fabricated with ReS2 reported in literature [15] .

Complementary electrodes characterization and electrochemical tests
As indicated in the main text, the SWCNTs/c-ReS2 assemblies for the preparation of the electrodes show a heterogeneous size dispersion of the ReS2 crystals from 0.5 to 200 µm. The big crystals detach from the membrane after the immersion of the electrodes in the electrolyte, not contributing to the HER-activity. However, the ReS2 crystals (< 1 µm) are able to penetrate the SWCNT network as shown in Figure S9 and participate in the HER.     Double-layer capacitance (Cdl) of the c-ReS2 and LPE-ReS2 films were estimated from cyclic voltammetry (CV) measurements in a non-Faradaic region of potential (potential between 0.30 and 0.45 V vs. Ag/AgCl) at various potential scan rates (ranging from 20 to 800 mV s -1 ) in 0.5 M H2SO4. Double-layer capacitance provides information regarding the electrochemically accessible surface area of the electrode. Such physical quantity depends on both the physical surface area and the porosity of the electrode. The double-layer capacitances (Cdl) of the catalyst films were estimated. To estimate experimentally Cdl, the catalyst films were produced by drop casting the corresponding dispersions on flat substrates of glassy carbon (GC) (catalyst mass loading of 0.2 mg cm -2 ). The use of flat GC as the substrate allows the Cdl contribution of the substrate to be limited compared to the case of catalyst films deposited on SWCNTs (as for the electrodes investigated in the main text). By plotting the difference between the anodic and the cathodic current densities (∆j = (ja-jc)) at 0.375 V vs. RHE as a function of the scan rate (SR) (Figure S12a), the Cdl was calculated from the slope of linear fit of ∆j vs. SR, assuming Cdl = (∆j)/2(SR). The calculated Cdl of the c-ReS2 film is ~0.162 mF cm -2 , which is ~1.5 time the one of the LPE-ReS2 film. These results indicate that the electrochemical accessible surface area of c-ReS2 film is superior to the LPE-ReS2 film. This could be due to the small lateral size (~ 4±1 nm determined by TEM) and partial vertical orientation of the c-ReS2 nanosheets relatively to the substrate creating a porous network as can be seen in the high magnification SEM image in Figure S12b, while the large LPE-ReS2 flakes (> 100 nm determined by TEM) are randomly stacked, Figure S12c. Moreover, we performed stability tests during 24 and 48 h in both acidic and alkaline solutions analyzing also how these conditions affect to the morphology of the SWCNTs/c-ReS2 electrodes. These long electrochemical experiments induced the formation of cracks, but the SWCNTs prevented the electrodes from breaking (see Figure S13 for acid and alkaline tests after 24h (a-f) and 48h (g-l)). As discussed in the main text, these changes are induced by the mechanical stresses caused by H2 bubbling through the layered structure of the electrodes [16], [17] and active material reorganization during HER-activation. Figure S13. Top-view representative SEM images at different magnification of the SWCNTs/c-ReS2 electrodes after stability test 24 h (red square, a-f) and 48 h (blue square, g-l) for HER in 0.5 M H2SO4 (a, b, c, g, h, i) (the enlargement increases from right panels to left panels) and in 1 M KOH (d, e, f, j, k, l) (the enlargement increases from right panels to left panels). Figure S14a-b show the chronoamperometry measurements (current retention vs. time) for representative SWCNTs/c-ReS2 (beyond those reported in the main text), over a 24 h in acidic and alkaline solutions, respectively. As mentioned in the main text, a constant overpotential was applied in order to provide the same starting cathodic current density of 20 mA cm -2 for HER, and GC rod was used as counter-electrode. As shown in Figure S14a, in acidic solution, the electrode has shown a significant increase (of 10.7%) of the cathodic current density after 24 h. This increase of the activity in acidic condition can be ascribed to favourable morphological changes of the electrode film during the HER process, already observed in metallic TMDCs [18] [19,20] or even in graphene-based electrocatalysts. [21] In alkaline solution (Figure S14b), the electrode retains ~79.5% of its initial cathodic current density. Despite the decrease of the HER-activity of the electrode observed in alkaline solutions, the iR-corrected polarization curves show that the c-ReS2 nanosheets are electrochemically activated in both acidic and alkaline solutions after 24 h. In fact, ƞ10 is reduced from 0.195 V to 0.162 V in 0.5 M H2SO4, and from 0.327 V to 0.181 V in 1 M KOH. The decrease of the cathodic current density in alkaline solution is ascribed to the increase of the electrical resistance of the electrodes (from 4.4 Ω to 45.9 Ω). In fact, cracks are observed in the electrodes after the electrochemical measurements by SEM (Figure S13a-f). The morphological changes, which are more pronounced in alkaline solutions compared to those in acidic ones, can originate from both the mechanical stresses caused by H2 bubbling through the layered structure of the electrodes [16], [17] . These stresses can cause a reorientation and/or fragmentation of the 2D electrocatalysts, as often observed in literature. [18][19][20][21][22] These morphological changes can be also responsible for the different electrochemical stability behavior of the HER activity observed for 24 h and 48 h tests (Figure 6c-d, main text). This can be avoided by using electrocatalyst binders, such as sulfonated tetrafluoroethylene-based fluoropolymer copolymers (e.g., Nafion).