Preparation of solution processed photodetectors comprised of two-dimensional tin(ii) sulfide nanosheet thin films assembled via the Langmuir–Blodgett method

We report the manufacture of fully solution processed photodetectors based on two-dimensional tin(ii) sulfide assembled via the Langmuir–Blodgett method. The method we propose can coat a variety of substrates including paper, Si/SiO2 and flexible polymer allowing for a potentially wide range of applications in future optoelectronic devices.

Two-dimensional (2D) materials are condensed matter solids formed of crystalline atomic layers held together via weak van der Waals forces. 1 They have a wide range of applications including use as channel materials in transistors, 2 absorber layers in solar cells, 3 light emission, 4 energy storage 5 and drug delivery 6 among others. 2D materials oen have different properties from their bulk counterparts such as increased strength 7 and electrical conductivity. 8 2D semiconductors may exhibit a change in electronic states from connement in 1D. 9 Thin lms are oen required for the creation of devices from nanomaterials for practical applications and can oen be made into exible devices such as thin lm solar cells 10 or photodetectors. 11,12 Thin lm solar cells in particular have several advantages over conventional solar cells including lower materials consumption and are lightweight, yet have the potential for high power conversion efficiency. 10 Many of the two-dimensional materials produced thus far have been derived from mechanical exfoliation, where Scotch tape or an equivalent is manually used to remove single crystalline layers from a bulk van der Waals solid followed by transfer to a substrate. Whilst this method in general produces extremely high quality crystalline atomic layers, 13 and is therefore oen used to produce prototype devices, it inherently lacks scalabilty. In order to address the problem of mass manufacture of two dimensional materials, liquid phase exfoliation (LPE) was introduced as a cost effective method for producing two dimensional nanomaterials 14 with the possibility of 100 L scales being produced and production rates up to 5 15 This method also does not require the high temperatures needed for methods such as CVD 16 or transfer between the growth and nal substrates. Liquid phase exfoliated nanomaterials are also directly processable from solution. 15 Furthermore, LPE has been shown to be effective for the production of a wide range of 2D materials such as graphene, 15 transition metal dichalcogenides 17 and monochalcogenides such as SnSe. 18 Tin(II) sulde (SnS) is a van der Waals solid with a puckered ab structure consisting of alternating Sn and S atoms, and is isostructural and isoelectronic with black phosphorus. 19 The bulk material has attracted interest due to its indirect band gap energy of 1.07 eV, 20 similar to bulk silicon at 1.14 eV. This band gap energy for SnS is useful for applications such as photodetection 21 and due to its higher theoretical Shockley-Queisser efficiency limit (24%) for solar cells. 22 The liquid phase exfoliation method established by Coleman et al. enables nanosheets to be separated from the bulk into solution utilising matching surface energies of the material and solvent. 23 Liquid phase exfoliation of SnS was rst reported by Lewis et al. it was established that as layer number reduced, band gap energy increased, and by tuning layer number the onset of photon absorption can be tuned over the near infrared 23 to visible range. 24 Overall, LPE is capable of creating large quantities of nanosheets, with potential for industrial scale production. Liquid phase exfoliated SnS has, for example, recently been used in the creation of photoelectrochemical systems with strong stability under both acidic and alkali conditions. 25 Many of the functional devices produced thus far have been derived from micromechanical exfoliation and manual nanomanipulation. A far more elegant solution to producing functional devices is to assemble them from solution, for example Kelly et al. recently reported a transistor based on exfoliated WSe 2 nanosheets. 2 The Langmuir-Blodgett method involves the use of a trough with a layer of water and controllable barriers to compress the lm. Nanomaterials in solution are added to the surface of the water and spread evenly to reduce their surface energy, 26 oen by using a low surface tension spreading solvent such as chloroform. 27 The surface pressure is measured as the lm is compressed with the substrate being withdrawn when the lm becomes solid. 28 The Langmuir-Blodgett method has the advantages of large area deposition and improved control of the lm at the nanoscale in comparison to vacuum ltration as well as the advantage of requiring no volatile solvents in comparison to liquid-liquid assembly methods. The use of movable barriers also allows for greater lm compression. 26 This method has been used to assemble large scale lms of exfoliated MoS 2 by Zhang et al. MoS 2 was exfoliated using nbutyl lithium followed by solvent exchange. MoS 2 was deposited onto the water surface using a 1 : 1 mix of DMF and dichloroethane. Substrates up to 130 cm 2 were coated with a surface coverage of 85-95%. 26 Collapse mechanisms of MoS 2 Langmuir lms have also been studied 29 alongside MoS 2 deposition on the surface of water with an upper hexane layer. 30 Graphene lms have also been prepared using the Langmuir-Blodgett method. 31 The Langmuir-Blodgett method has been used for the assembly of organo-clay hybrid lms via the coating of octadecylammonium chloride in a 4 : 1 chloroform : ethanol solution onto a 2D nanoclay liquid phase exfoliated lm using an electrospray method. 32 A solvent mix of chloroform and NMP has also been utilised for the deposition of nanosheet lms. 33 Recently the Langmuir-Blodget method has been used for the assembly of unmodied clay nanosheets, 34 Ti 3 C 2 Tx MXene nanosheet lms for the removal of Cr(VI) and methyl orange from an aqueous environment 35 as well as for the growth of rGO wrapped nanostructures for use in electrocatalysts. 36 Given the chemical similarity of the basal planes of inorganic 2D materials, we hypothesised that the assembly of group IV-VI nanomaterials such as SnS should also be possible at the air water interface. Due to their interesting semiconducting and properties described, it should also be possible to produce prototype optoelectronic devices from a fully solution processed pathway. In this paper we now communicate a methodology to assemble thin lms comprised of 2D SnS nanosheets using the Langmuir-Blodgett technique (Scheme 1a). We report the use of these lms in simple photodetectors. This represents a scalable methodology to produce fully solution processed devices based on 2D materials.
Scheme 1(a) shows the step by step process of lm preparation. Firstly, bulk SnS is broken down by liquid phase exfoliation from the bulk material to produce a stable dispersion of crystalline nanosheets. Characterisation of the exfoliated nanomaterials was undertaken using atomic force and electron microscopy yielding average sheet dimensions of 23.9 nm height Â 224 nm longest side length ( Fig. S1 †). The nanosheets were then deposited onto the water air interface. The lm is then compressed whilst an immersed substrate is withdrawn, leading to the creation of a densely packed nanosheet lm. Scheme 1(b) shows that SnS can be successfully deposited on the water-air interface via the addition of chloroform as a spreading solvent, as shown previously with other Langmuir based lms. 27 Scheme 1(a) shows a z-type deposition of SnS as the hydrophilic glass and Si with a 300 nm oxide layer is withdrawn through the lm at 1 atm pressure. The lm compression occurred at a rate of 5.88 cm 2 s À1 . No further treatments were performed to change the hydrophilicity of the substrates, the oxide layer present was sufficient to provide hydrophilicity to the substrate. 37 Scheme 1(c) shows a gradual increase in surface pressure as the area was decreased from 1175 cm 2 to 298 cm 2 before a sharp increase in pressure, indicating the lm has reached full compression. The sharp increase in surface pressure during compression is common in Langmuir-Blodgett assembled lms of nanomaterials. 38 In response to compression the surface pressure prole in Scheme 1(c) rises rapidly until it reaches a maximum due to the size of the sheets and the potential difficulty in sliding over each other compared to polymers or smaller nanomaterials. Scheme 1(d) shows that the lm is capable of being coated onto Si/SiO 2 with a mask dening the areas covered.
We characterised the resulting structural and electronic properties of the thin lm of SnS nanosheets deposited via the Langmuir-Blodgett method using a range of techniques. Fig. 1(a) shows a height prole AFM image of a lm edge with an average on-lm roughness (R a ) of 31.9 nm and an average lm thickness of 78.6 nm (Fig S3 † provides an additional lm prole). Previous work on Langmuir-Blodgett deposition has produced thinner lms. The use of high centrifugation speeds yielded 7 nm thick lms for a single deposition 31 whilst the use of lithium ion intercalation before exfoliation enabled lm thicknesses of under 2 nm per layer to be realised. 26 The average lm thickness is above the average sheet thickness, suggesting that the lm is made up of overlapping ake multilayers. However, the thickness of the lms is signicantly lower than those grown via chemical bath deposition (e.g. 290 nm (ref. 39)) indicating that thinner lms can be produced compared to chemical bath methods, and potentially at a much lower cost than methods such as CVD. Images of the lm morphology in plan view SEM ( Fig. 1(b)) suggest no notable alignment of the nanosheets in the lateral dimension as the lm is formed and deposited (see Fig S4 † for statistical analysis of sheet angle measurement). The coverage of the lm is 94.6% as determined by image thresholding using imagej soware to determine the area le uncovered. This gives a coverage of 0.0142 gm À2 as calculated from average thickness, SnS density and % coverage of the substrate. Preliminary SEM results also suggest that the Langmuir-Blodgett method is effective at coating SnS onto a variety of substrates including polyolen lms (Paralm®), aluminium foil and paper (Fig S6 †). We also probed the structure of the thin lms by powder X-ray diffraction (XRD). Aer exfoliation and lm assembly, the diffraction peak associated with the (400) of SnS is still the most intense reection but is characterised by a much larger FWHM compared to that of bulk SnS under the same recording conditions (0.442 AE 18.5% compared to 0.175 AE 5%). This indicates a successful breakdown of the crystal structure and thinning of the material in the (400) plane during exfoliation due to the reduction in long range order 40 (reections for bulk SnS are assigned to orthorhombic SnS and indexed in Fig S2 †). The lack of any additional peaks indicates that there has not been any signicant degradation of the material to the corresponding oxide which is in agreement with previous works. 24,25 The reections at 88 and 94 are unlikely to be from crystalline silicon 41 due to the thick oxide layer and low angle of incidence used. We tentatively ascribe these peaks to the 3,0,À3 and 3,2,4 peaks for SnS. 41 However a condent assignment of this reection requires further studies.
We also characterised the optical properties of the nanosheet thin lms using Raman and UV-Vis-NIR absorption spectroscopy. No shis in the Raman peak positions B 3 g, Ag and B 3 u from bulk SnS to Langmuir-Blodgett lm were observed. The broad feature at around 300 cm À1 for the LB lm may potentially be due to SnS 2 and Sn 2 S 3 impurities. 42 It is predicted that due to the lower density compared to SnS 43 the impurities may increase in concentration compared to the bulk aer centrifugation. These impurities may have signicant effects on the efficiency of the devices produced. 44 A shi in peak positions is typically observed in nanomaterials which exhibit quantum connement, 45 this occurs at 14 nm for SnS. 46 Fig. 1(e) shows a UV-Vis spectra from which the absorption coefficients at xed wavelengths may be obtained, for 350 nm, 405 nm, 450 nm, 500 nm, 600 nm and 800 nm the values obtained were: 2.26 Â 10 5 cm À1 , 2.21 Â 10 5 cm À1 , 2.16 Â 10 5 cm À1 , 2.04 Â 10 5 cm À1 , 1.67 Â 10 5 cm À1 and 1.05 Â 10 5 cm À1 respectively, this matches well to the absorption coefficients of SnS in literature (greater than 10 4 cm À1 ). 47 It also suggests there may be a greater response at shorter wavelengths. Fig. 1(f) shows a band gap of 0.92 eV for the exfoliated SnS in NMP which is below the expected value of 1.07 eV (ref. 20) although lies within the reasonable error introduced by the use of Tauc plots. 48 The band gap also matches well with SnS exfoliated in NMP in previous work. 24 The band gap of the lm appears to change from nanosheet suspension to lm in 1(f). This has been observed previously for Langmuir-Blodgett 49 and other deposited lms. It has also been observed that apparent decreases in band gap may occur due to the presence of scattering artefacts within lms of nanoscale objects. 50 We then produced simple prototype photodetectors via the printing of Ag nanoparticles to form interdigitated electrodes on top of the SnS nanosheet lm. Additionally, SnS lms were deposited onto lithographically dened Au interdigitated electrodes for characterisation and referencing to the printed devices.
Previously SnS photodetectors have been created via methods such as electron beam deposition, 51 thermal evaporation 52 and chemical bath deposition. 53 The Langmuir-Blodgett method allows SnS to be directly processed into a lm from a liquid phase exfoliated solution, allowing them to be produced cheaply and with the potential for scalability.
Inset to Fig. 2(a) is an image of an interdigitated Ag electrode SnS photodetector device with an area of 6.4 Â 10 À5 m 2 . The electrodes can be clearly identied with an average spacing of 99 mm, and an average RMS edge roughness value of 1.89 mm (determined for individual contact lines using the imageJ 'analyze_stripes' plugin 54 (Fig S7 †)). Fig. 2(a) shows an increase in the slope of the I-V curve in the third quadrant indicating a reduction in resistance under 1 sun illumination (1000 W m À2 ) with the AM1.5 spectrum. No short circuit current under illumination was observed indicating that the device functions as a photoconductor. The non-linear response upon negative biasing is due to initial trap lling which once equilibrium has been reached results in linear device operation. Previously it has been shown that silver diffusion into SnS has an interstitial doping effect, neutralising defect states and lowering the lm resistivity. 55,56 It is also possible that the Ag ink morphology and the concentration of nanoparticles in the ink may play an effect on the device properties. 57 A resistivity of 2.85 Â 10 6 U sq À1 was obtained for the device which is signicantly higher than SnS lms prepared by physical vapour deposition (250 U sq À1 ), 58 likely due to poor carrier mobility between akes. Fig. 2(b) indicates that a clear response is present under illumination when an external bias is applied (giving a eld strength of 0.4 V mm À1 ). Closer inspection shows a fast and slow decay component following the illumination being blocked. This biexponential decay indicates the capture of trapped carriers and the presence of trap states within the device. 59,60 This again supports the photoconductive nature of the device operation with a rise time of $0.22 s and a fall time of $2.83 s, 61 both being longer than the shutter closing/opening time of 3.7 ms (which was considered negligible). The rise time is the time taken to get from 10% to 90% of the light current with the fall time being the time taken from 90% of the light current to 10%.
Previous work performed by Jiang et al. has shown a slow fall time in Ag/SnS photoconductor devices arising from carrier trapping. 62 Similarly, in our devices the large rise time may also be due to the presence of a high trap density which must be lled upon light exposure.
The mean dark current is 2.78 Â 10 À10 A with a standard deviation of 2.02 Â 10 À11 A. The mean light current was found to be 3.92 Â 10 À10 A with a standard deviation of 4.03 Â 10 À11 A. A poor signal to noise ratio appears to be present within the device, possibly due to the large number of SnS nanosheets involved in charge carrier transit, leading to a low signal, hence a low signal to noise ratio. The noise could be reduced via surface passivation 63 or the use of a diode like structure to reduce leakage current under reverse bias. 64 A low responsivity of 2.00 Â 10 À9 A W À1 AE 1.5 Â 10 À10 A W À1 was found for energies above the band gap energy of 0.6 eV for the deposited lm.
The low responsivity may be due to poor bridging between individual SnS nanosheets and the poor transport of holes between adjacent akes (hopping) relative to the higher mobility within each ake. 65 There are potentially hundreds of nanosheets between the contacts as determined by the average length obtained (Fig S1 †). To conrm that the optical response was due to the presence of the SnS a reference device was tested (without SnS deposition, Fig S8 †) with no photoresponse observed. Despite the low responsivity, it is notable that the SnS devices fabricated are one of the few examples of a thin lm photodetector device based on 2D materials requiring only solution processing at ambient temperature and atmospheric pressure.
To demonstrate that the observed behaviour originates from the photoresponse of the SnS akes a second device was fabricated by pseudo Langmuir-Blodgett deposition on to lithographically dened Au interdigitated electrodes (15 mm separation) on fused silica (inset Fig. 3(b)). This enabled us to remove any effect of photoinduced Ag migration from the observed behaviour as well as eliminating the issue of potential printing irregularities. Fig. 3(a) shows that the devices display a similar photoresponse to the devices with printed Ag electrodes when exposed to modulated AM1.5 illumination. The dark current remains similar at $0.3 nA, though during illumination the current is higher (0.7 nA vs. 0.4 nA). This increase directly correlates to the higher electric eld strength (0.66 V mm À1 vs. 0.4 V mm À1 ) between the interdigitated electrodes. The responsivity of the device was determined to be 1.79 Â 10 À8 A W À1 , with a photoresponse rise and fall time of 0.77 s and 0.85 s respectively. The responsivity is lower than for photodetectors prepared by Guo  et al. 66 Improvements to the device to improve the responsivity could include methods to improve the lateral size of nanosheets such as intercalation. 67 Other routes to improve the device may include doping 68,69 or a change in architecture to a phototransistor type device. 70 The removal of potential SnS 2 and Sn 2 S 3 impurities via methods such as annealing at 500 C, 500 mbar pressure under argon or the use of higher quality starting material may also be a key route to improve the efficiency of the device. 42 It is also noticeable that the level of noise present in Fig. 3(a) is reduced compared to that in Fig. 2(b), indicating that the Ag electrodes themselves (in addition to the SnS sheets) also affect the performance.
When exciting using AM1.5 illumination it is possible that thermal effects may be present which could give rise to the observed behaviour.
In order to demonstrate a true photoresponse monochromatic illumination was used to determine if illumination energies above the band gap generated a photocurrent response in the device. Fig. 3(b) shows a small response under 350 nm (3.54 eV) illumination. (IV curves for other wavelengths are available in Fig. S9 †). Fig. 3(c) shows an increased response for 350 nm wavelength as determined via the IV curves. This increased response is likely due to increased absorption as shown in the UV-Vis spectra (Fig. 1e), the signal at longer wavelengths is difficult to observe due to the low responsivity. A higher response at lower wavelength has been observed previously for SnS. 53 Fig. 3(d) shows that an increase in current is present under 350 nm and 405 nm illumination which can be cycled on and off. A rise and fall time of 1.09 of 1.44 seconds respectively was observed for 405 nm illumination. A light/dark current ratio of 1.03 was obtained under 405 nm. To account for noise the on and off section had their current averaged using origin soware. A dri in current during measurement was observed, this was considered as the reason for the signicant difference between the dark current for 350 nm and 405 nm. To further reduce noise surface passivation may also be used to improve the device properties. 63 Alternatively, an increase in bias voltage or an increase in monochromatic illumination intensity may improve the signal: noise ratio though may risk damage to the device. A magnied off/on cycle for 405 nm is shown in Fig. S10. † In conclusion, we report here a methodology for the assembly of 2D SnS nanosheets into thin lms using the Langmuir-Blodgett method, and the testing of the lms as prototype all-solution processed photodetectors. Tin(II) sulde was successfully exfoliated with an average sheet thickness of 33 nm with the average longest side length of 224 nm. A nanosheet based lm was coated onto a variety of substrates via the Langmuir-Blodgett method with the addition of chloroform as a spreading solvent. The lms were found to be polycrystalline with an average thickness of 78.6 nm with a high surface coverage up to 94.6% for an Si/SiO 2 substrate. The lms were found to be semiconductive with the ability to respond to light under bias as shown by AM1.5 and monochromatic illumination. Proof-of-concept photodetectors have been successfully produced. It was also conrmed that the response was due to the photoresponse as opposed to a heating effect. This deposition method could potentially be used to create a variety of SnS lms using different exfoliated nanosheet sizes separated via cascade centrifugation as well as the potential for future exible photodetector devices. Despite the low responsivity, large rise and fall times further work could allow the gain to be optimised. We also note that the use of the Langmuir-Blodgett trough is an easily scalable technology and could provide coatings over very large area substrates not only for photodetectors but for other devices such as thin lm solar cells.

Conflicts of interest
The authors declare no conicts of interest.