Effect of Post-Synthesis Treatments on the Properties of ZnS Nanoparticles: An Experimental and Computational Study

This work deals with the characterization of ZnS NanoParticles (NP), prepared by precipitation employing thioacetamide as sulfur source at different reaction time length. The attention is focused on the modification induced on structural, surface and electronic properties of ZnS NP by post-synthesis treatments. These were aimed at removing from the samples surface adsorbed reactants, by washing or thermal treatments, both in air or vacuum. The effect of these parameters is followed by X-Ray Diffraction (XRD), Transmission Electron Microscopy (TEM), Fourier Transform InfraRed (FTIR), gas-volumetric and ThermoGravimetric Analysis (TGA). Moreover, the effect of nanostructuration on the semiconducting material band gap is evaluated by Diffuse Reflectance UV-Vis (DR UV-Vis) spectroscopy. Density Functional Theory (DFT) calculations have been employed to clarify the role of the adsorbed reactants on the surface stability and to assess the relationship between particle size and band gap value.


INTRODUCTION
Since the first discovery by Honda and Fujishima, photocatalyst hydrogen production has attracted strong consideration [1]. Many heterogeneous photocatalysts have been found, mostly working in ultraviolet region, which accounts for only 5% of the solar spectrum [2][3][4]. Taking this into account, it is understandable the existing wide interest in searching new visible-light-driven photocatalysts that could replace the "old" ones and extend their response to the solar spectrum. Up to now, the number of such photocatalysts is still limited [5][6][7]. Although CdS is reported as an effective photocatalyst with a bang gap suitable for visible light exploitation, its activity depends on expensive noble metals deposited on its surface, and environmental and safety concerns related to the well-know cadmium toxicity are serious limitations to applications in large scale [8,9]. As an alternative ZnS can produce H 2 effectively without noble metal co-catalyst, due to its high conduction band position, but the wide band gap (3.7 eV) implies that only ultraviolet radiation can be employed to produce electron/hole couples [10]. This is the reason why band gap engineering has been proposed to modify the electronic properties of ZnS and improve its visible light harvesting efficacy [11][12][13].
Apart from the electronic properties (band gap, conduction and valence band position) the main requirements for an efficient photocatalysts are high crystallinity (to avoid the presence of defects, acting as electron/hole recombination traps), coupled to high surface area and small particle size (to increase the efficacy of electron/hole migration to the surface). This implies that a careful control of these parameters is crucial in the synthesis of a nanostructured photocatalyst.
Different chemical routes have been used to prepare small semiconducting NanoParticles (NP). The most common way is to use stabilizing agents which prevent growth of particles and their agglomeration by electrostatic and steric repulsion [14][15][16][17][18][19]. But at the same time they modify the structural, morphological and optical properties of nanoparticles. Growth of strongly confined nanoparticles without a capping agent is difficult to achieve and rarely reported. Among other chemical routes the use of reversed micellars solutions [20,21], thin film deposition process, inorganic templates were reported.
For materials based on CdS and ZnS an alternative route for preparation of nanoparticles is a hydrothermal synthesis procedure [22]. The use of ThioAcetAmide (TAA) as a sulfur source allows obtaining well mono-dispersed particles, whose size depends upon the reaction conditions. The formation of small spherical CdS NP, which are then transformed into cylindrical structures while the reaction proceeds was reported [23]. The early work where TAA was applied in ZnS synthesis reported the formation of submicrometric particles [24]. Later, the effect of sulfur source (TAA in comparison with Na 2 S) in the synthesis of ZnS based system (ZnS/CdS, ZnS/ZnO) was studied by different authors [25,26]. The detailed investigation of the precipitation dynamic of ZnS from TAA decomposition in conditions close to those applied in other works was carried out with attention on product agglomeration [27]. The authors showed a crucial role of hydrodynamics in the determination of size and agglomeration, and studied the influence of parameters such as pH, electrical conductivity, concentration of sulfide ions and turbidity. The obtained results could be very helpful in control of morphology and associated properties.
Recently, we also demonstrated that this synthesis procedure works well in case of ZnS [28]. In particular, we showed that the reaction carried out at room temperature employing a methanol/ethanol mixture as a solvent was suitable for the formation of small ZnS NP with much higher yield with respect to what obtained in water. In this contribution, the attention is focused on the effect of post-synthesis treatments, aimed at removing the reactants and impurities adsorbed on the ZnS photocatalyst surface, on the material properties. This is studied on samples prepared at different reaction time length in methanol/ethanol mixture with TAA. The work reports a detailed physico-chemical characterization of the materials, concerning their morphology, structure, textural properties, adsorbed species and electronic properties. Moreover, Density Functional Theory (DFT) calculations have been employed to gain hints on the size dependence of surface and electron properties of the ZnS NP. The adsorption of formic acid (as a simple model of surface species adsorbed during the synthesis process) has been simulated. The interaction with the most relevant ZnS surfaces (110 and 111) [29] has been considered. In a typical procedure TAA and ZnAc were dissolved separately in ethanol/methanol 1/1 mixture at 313 K, to give 0.2 M solutions [28,29]. The TAA solution was then slowly added to the ZnAc one under continuous stirring. The resulting mixture (50 mL total volume) was left under stirring for 6 h or 24 h. Afterwards the resulting white powder was separated by centrifugation (4 000 rpm), and washed first with the alcohol mixture and then trice with distilled water at Room Temperature (RT) or 353 K. The cleaned product was dried for about 3 h in oven at 353 K. The obtained samples were carefully milled before characterization. The labeling of the samples in the text is related to reaction time length (ZnS_6h or ZnS_24h) and to the water temperature in washing procedure: ZnS_xh_HW (x = 6 or 24) are the samples treated with hot water. The same synthetic procedure was carried out at 6 hours reaction by employing Na 2 S as sulfur source instead of TAA (sample ZnS_Na 2 S). This resulted in a much lower yield, so that these results are only reported for comparison.
Thermal activation in oven or in vacuum was carried out on sample ZnS_6h_HW. The resulting samples are labeled mentioning the activation temperature and conditions (ZnS_573 activated in oven at 573 K, sample ZnS_673_V activated at 673 K in vacuum).

Characterization
Powder X-Ray Diffraction (XRD) patterns were measured with PW3050/60 X'Pert PRO MPD diffractometer (Panalytical) working in Bragg-Brentano geometry using Cu Ka radiation (k = 1.5406 Å) and operated at 45 kV, 40 mA with a scan speed of 0.01°min À1 and step size of 0.0170°. The crystallite size D = 4/3 L was calculated by applying the Scherrer's equation to the (111) peak at 28.7°: where k is the X-ray wavelength, b the measured broadening of the diffraction peak and h the diffraction angle [30]. The measured b value was corrected by subtracting the instrumental line broadening, estimated in 0.02°. The error on particle size estimation was calculated by including standard deviation on 5 measurements and instrumental resolution.
Transmission Electron Microscopy (HRTEM) measurements were performed on JEM 3010-UHR microscope (JEOL Ltd.) operating at 300 kV. For the measurements ZnS powder was dispersed on a copper grid coated with a perforated carbon film.
Gasvolumetric nitrogen adsorption/desorption isotherms were measured at Liquid Nitrogen Temperature (LNT) with ASAP 2020 physisorption analyzer (Micromeritics). Prior to analyses, samples were evacuated at 423 K over 2 h. The Specific Surface Area (SSA) was calculated by the Brunauer-Emmett-Teller (BET) method in the p/p 0 range from 0.05 to 0.2 [31]. The average pore size and volume were calculated on the adsorption branch of the isotherms according to the Barrett-Joyner-Halenda (BJH) method [32,33]. ThermoGravimetric Analysis (TGA) was carried out on a TAQ600 (TA Instruments) heating the samples at a rate of 10 K min À1 from 303 to 1123 K in air flow. Before starting the measurements, samples were equilibrated at 303 K.
Fourier Transform InfraRed (FTIR) spectra were recorded with an IFS66 spectrophotometer (Bruker) equipped with a Deuterated TriGlycine Sulfate (DTGS) detector, working with resolution of 4 cm À1 over 32 scans. Samples were in the form of self-supporting pellets suitable for transmission infrared experiments and were placed in a quartz cell equipped with KBr windows, designed for in situ studies in vacuum and controlled atmosphere. Before infrared analysis the samples were evacuated at the desired temperature (RT or 673 K) for 1 hour.
Diffuse Reflectance (DR) UV-Vis measurements were recorded on powdered samples on a Cary 5 000 spectrophotometer equipped with a reflectance sphere. Spectra were recorded at a scan rate of 300 nm/min with a step size of 1 nm and fixed Spectral Band Width (SBW) of 2 nm. The measured intensities were converted with the Kubelka-Munk function. The fundamental absorption, which corresponds to electron excitation from the valance band to conduction band, can be used to determine the value of the optical band gap of the synthesized ZnS nanoparticles [34][35][36][37]. The relation between the incident photon energy (hm) and the absorption coefficient (a) is given by the following relation: (ahm) 1/n = A(hm À E g ), where A is constant and E g is the band gap energy of the material and the exponent n depends on the type of transition [38,39]. ZnS as a semiconductor is characterized by direct allowed transition where n = 1/2 [40], therefore the direct band gaps of the samples are calculated by plotting (ahm) 2 versus hm and then extrapolating the straight portion of the curve on hm axis at a = 0.

DFT Modeling
Results presented are obtained with periodic CRYSTAL09 code [41], using the hybrids B3PW (the combination of the Becke's three parameter functional [42] for the exchange and the Perdew Wang functional [43] for the correlation) and PBE0 [44]. Basis sets of triple-zed TZ quality for the valence electrons have been employed: the relativistic Steven-Krauss small core RECP with a 4211-41 [4sp2d] basis set for valence electrons for Zn atoms [45] combined to a 8-6311-1 [1s-4sp-1d] for S [45]. Within this computational set up, B3PE has been found to provide accurate estimate of the ZnS band gap (E g = 3.71 eV) very close to the experimental value (E g = 3.65-3.7 eV), while PBE0 was found to accurately accounts for surface properties [29,46,47].
In the CRYSTAL code, the level of accuracy in evaluating the Coulomb and the exchange series is controlled by five parameters [41] for which 10 À7 , 10 À7 , 10 À7 , 10 À7 , 10 À18 values have been used for all calculations. The convergence threshold for SCF energy was set to 10 À9 Ha. The reciprocal space was sampled according to a regular sublattice determined by the shrinking factor IS [41]. A value of IS = 8 has been adopted which corresponds to 34 independent k-points in the irreducible part of the Brillouin zone.
Nanoparticles have been modeled with bidimensional slabs of increasing thickness d (from 0.5 to about 6 nm), characterized by two infinite dimensions (x, y) and a finite thickness, and cut from the bulk so as to expose the most stable ZnS surface, the (110) [29]. Stoichiometric ZnS slabs (Zn/S ratio equal to 1/1) have been considered, in agreement with the experimental and theoretical reports [28,[48][49][50][51]. Surface Energy (E s ) is calculated as: where E slab is the energy of an n-layered 2D slab used to model the material that exposes the surface of interest, E bulk is the energy of the bulk material and A is the area of the 2D surface cell. The factor 1/2 accounts for the existence of two limiting surfaces. E s is then the energy per unit area required to form the surface from the bulk and it is reported in J/m 2 .
The interaction with formic acid has been computed considering a coverage (h = 0.5 (simulated adopting a 2 9 1 supercell). Slabs of thickness of 1.2 and 1.5 nm have been selected for the (110) and (111) surfaces (corresponding to 7 and 6 ZnS layers, respectively). The Adsorption Energy (ÁE ads ) per molecule is calculated as: where E m is the energy of the optimized isolated molecule, E slab is the energy of the optimized isolated slab modeling the surface and n is the number of molecules per cell. The basis set superposition error BSSE has been estimated, by employing the counterpoise method [52], to account for no more than 0.07 eV and results are corrected accordingly.
For all the systems discussed in this study, geometry has been optimized relaxing all the atoms in the cell and keeping fixed the lattice parameters at the bulk values. All the symmetry constraints have been removed. Adsorbate frequencies have been computed at the U point, within the harmonic approximation, by diagonalizing the massweighted Hessian matrix only for the adsorbate fragment, once checked that its vibrational modes are not coupled with other crystal phonons.

Effect of Synthesis and Post-Synthesis Conditions on Crystallite Size and Textural Properties
The general properties of the prepared samples are listed in Table 1, with those of ZnS_bulk (commercial sample) for comparison. These include the crystallite size calculated by XRD (Scherrer equation applied to the 111 peak) and the textural properties, that is SSA and pore volume. The reported band gap values were calculated from DR UV-Vis spectra, as discussed in the following (Sect. 2.3). A selection of XRD patterns of the materials is reported in Figure 1. All synthesized ZnS samples have three main diffraction peaks at 28.7, 47.9 and 56.8°, corresponding to the (111), (220) and (311) lattice planes of ZnS cubic sphalerite structure (JCPDS 010772100). Minor features can be observed at 32.9, 70.3 and 77.5°, due to (200), (400) and (331) reflections. Sample ZnS_6h shows quite broad peaks, corresponding to an average crystallite size (Scherrer equation) of 1.8 nm. The peak width decreases both in relation to longer reaction time (ZnS_24h) and after washing with hot water. On the contrary, hot washing did not particularly affect the XRD patterns and related crystallite size of sample ZnS_24h.
The estimated crystallite size of all samples is listed in Table 1. The smallest crystallite size was obtained for ZnS_6h. Changing all other parameters leads to slight growing of particles. Moreover the use of Na 2 S as sulfur source instead of TAA resulted in relatively big particles. As mentioned in the experimental section, this procedure gives a very low yield, and is only reported for comparison.
The changes in SSA and porosity of the samples as a function of synthesis parameters are not directly correlated to the crystallite size estimated by XRD. For instance, washing with hot water causes an increase in SSA and pore volume in sample ZnS_6h. This could be easily explained by considering that the samples porosity is due to interparticle voids [29]: the removal of adsorbed products from the sample surface by hot washing increases the accessible volume and surface area of the particles. However, the same effect is not evident in sample ZnS_24h, where washing causes a slight decrease in both crystallite size and SSA.
On the whole, all prepared samples show relatively high SSA, which represents an important point for application in photocatalysis.
The prepared samples were also subjected to thermal treatments, as an alternative way (often employed in catalysis) to remove adsorbed impurities and obtain a clean surface. The effect of thermal treatments in vacuum was already described in Reference [29], by employing both XRD and TEM techniques. As reported elsewhere, thermal treatments in air of ZnS causes the transformation to ZnO [53][54][55]. This was confirmed on our samples by both activation in static oven and in N 2 /O 2 flow. In both cases, the transformation to ZnO started around 673 K (data not reported), while after activation at 573 K only the typical XRD peaks of cubic ZnS were observed (Fig. 1, grey dashed curve c). This pattern is compared in the same figure with that obtained after activation in vacuum at 673 K [29]. The crystallite size of both samples is similar to what measured before thermal activation (Tab. 1), but a decrease in SSA and pore volume is observed, suggesting further agglomeration of the primary ZnS NP.
TEM micrographs of the prepared ZnS samples revealed the presence of dense agglomerates of spheroidal particles (Fig. 2). It is however possible to observe light regions among the particles, which could be related to interparticle voids resulting in high surface area and pore volume. The high agglomeration of the material did not allow obtaining high resolution images of single particles, so that a reliable particle size analysis was not possible. The average value of ca. 2.6 (±0.4) nm (measured on 20-50 particles) was found for the sample ZnS_6h, in good agreement with XRD. It is important to underline the fact that the changes in crystalline domains size estimated by XRD cannot be appreciated by TEM, since the micrographs of samples ZnS_6h_HW are hardly distinguishable from those of ZnS_6h (Fig. 2, compare top panels with the left bottom one). In the case of sample ZnS_24h, the particles look more defined and crystallized with respect to the former (right bottom panel of Fig. 2). Particle size analysis was almost impossible in this case, due to very dense agglomeration of the primary particles, which however showed a primary size around 2-3 nm, which is similar to that of ZnS_6h. The different size estimated by XRD can thus be explained on the basis of higher crystallinity (less defective surfaces) and/or more dense aggregation.
Lattice fringes are visible in the TEM micrograph (bottom right panel of Fig. 2), indicating the crystalline nature of the particles. Fourier Transform (FT) analysis of the images resulted in a diffraction ring with average d-spacing of 3.0 Å, characteristic of the (111) peak of ZnS cubic structure.
On the whole, the results discussed in this paragraph suggest that the particles obtained at 6 h synthesis are stabilized by adsorbed reactants, so that hot washing causes a partial aggregation of the primary particles. At longer reaction time length, more crystalline domains are formed, which are densely interconnected, as shown by both TEM and surface area analysis. In this case, the role of adsorbed species in stabilizing the particles is less important.

Quantitative Analysis
The nature and amount of species adsorbed on the surface of the prepared ZnS sample was studied by thermogravimetric analysis. To this aim, the attention was focused on sample ZnS_6h and following modifications (ZnS_6h_HW, ZnS_573 and ZnS_673_V). These are reported in Figure 3 together the results obtained on ZnS_Na 2 S for comparison. All experiments were carried out in air flow, and an analysis on ZnS_6h was repeated in nitrogen. The TG curve of ZnS_6h is characterized by three distinct steps of weight loss accompanied by definite peaks on the Differential Thermal Analysis (DTA) curve (not shown). The first minor weight loss (4.2%) is centered around 363 K and corresponds to desorption of water molecules weakly bonded to the surface. The next two steps with larger weight losses started around 473 K (maximum of DTA at 620 K) and 773 K (DTA maximum around 830-850 K). To understand the nature of these steps it is useful to compare the profiles of the different samples and corresponding amounts listed in Table 2.
First, the amount of adsorbed water slightly changes upon synthesis and treatment conditions, in a range from 1.3 to 6.6%. This could suggest an effect of treatments or reactants on surface hydrophilicity. For instance the highest amount is observed on samples washed with hot water (ZnS_6h_HW) and prepared with Na 2 S. This could be related to the fact that the surface of these samples is less covered by reactants and it is thus more available to water interaction. This hypothesis is confirmed by comparing the TG profiles of ZnS_6h, ZnS_Na 2 S and ZnS_6h_HW. The first looses 18 wt% in step II, while this is reduced to 2.1% in  ZnS_6h_HW and to practically zero in ZnS_Na 2 S. This clearly indicates that hot water treatment removes weakly adsorbed reactants from the synthesis, in particular TAA, which was not employed to prepare ZnS_Na 2 S. Noticeably, a small amount of adsorbed species is still present after hot washing, thermal activation at 573 K in air or 673 K in vacuum (2.4 and 0.4 wt%, respectively). In this case, the desorption starts at higher temperature (533-593 K) with respect to untreated sample (453 K) indicating a stronger interaction of the related adsorbed species. These are explained as residual acetate groups from the synthesis, as will be discussed in the following.
Finally, step III is similar in all samples, but is not observed when the analysis is carried out in nitrogen (curve not reported, data summarized in Tab. 2). This obviously allows assigning this phenomenon to an oxidative process, that is the transformation of ZnS to ZnO, which in these conditions starts above 800 K irrespective of the samples synthesis or treatment, in agreement with XRD findings (not reported).

Infrared Spectroscopy (FTIR)
The nature of adsorbed species was also studied by infrared spectroscopy, again focusing on sample ZnS_6h and its transformation upon post-synthesis treatments (Fig. 4). The as prepared sample (curve a) shows very intense bands in the whole mid infrared region, which can be easily assigned to adsorbed species, reactants or contaminants from the atmosphere. We observed a broad absorption in the OH stretching region (mOH) between 3 600 and 2 500 cm À1 , with superimposed bands between 3 000 and 2 800 cm À1 related to aliphatic mCH 2 and mCH 3 modes. At lower frequency, weak bands at 2 940 and 2 365 cm À1 are assigned to mS-H groups [56]. The more intense spectral features (out of scale) are found in the 1 750-1 250 cm À1 range, where both N-H bending (dNH, found at 1 400 cm À1 for TAA in KBr, spectra not reported) and mCO modes are found. Other peaks at lower frequency can be assigned to bending modes of CH 3 in CH 3 -S groups, namely symmetric vibrations at 1 385 and rocking at 1 110 cm À1 , and to mC-S modes (611 cm À1 ) as confirmed by comparison with KBr spectra of TAA (not reported). These bands confirm the presence of TAA molecules adsorbed on the sample surface, which could be present as CH 3 -S-NH 2 in tautomeric equilibrium with CH 3 -SH-NH form [57]. On the contrary, the two strong bands at 1 564 and 1 441 cm À1 can be ascribed to Step II Step I Step III FTIR spectra of ZnS_6h sample as such a) and after postsynthesis treatments: b) ZnS_6h_HW; c) ZnS_573 and d) ZnS_673_V. All samples were outgassed at RT before measurements, expected ZnS_673_V which was outgassed at 673 K.  [58] is less evident, since it should be characterized by the typical amide I band (mCO), which is expected with high intensity around 1 690-1 694 cm À1 [56]. After hot water treatment (sample ZnS_6h_HW, curve b) most of the bands related to TAA (mS-H, dCH 3 -S, mC-S) disappear, in agreement with TG analysis. What is observed in this case is a broad adsorption in the mOH region (3 600-2 500 cm À1 ) suggesting the presence of OH groups involved in hydrogen bonding. This is confirmed by the observation of a broad band centered at 884 cm À1 , related to the corresponding dOH. This could be related to partial surface hydroxylation (Zn-OH groups) and water molecules strongly adsorbed on the surface (not removed by RT outgassing), in agreement with the relatively high affinity for water indicated by TGA. The typical dOH mode of H 2 O, expected around 1 620-1 640 cm À1 is observable as a shoulder on the tail of the intense band at 1 564 cm À1 . Apart from the bands described above, the main features of sample ZnS_6h_HW are two intense and well defined bands at 1 564 and 1 441 cm À1 , with weaker components at 1 348 and 1 027 cm À1 , which could be mainly assigned to adsorbed acetate groups (the typical amide I band of acetamide, expected around 1 690 cm À1 is missing). The interaction energy of these species with the ZnS NP surface is clearly high, since they are not removed after treatment at 573 K in air (curve c). In this case however, the bands related to OH groups in hydrogen bonding decreases, in agreement with TG analysis.
Finally, an almost totally cleaned surface is obtained only by thermally activating the sample in vacuum at 673 K. In this case, as already pointed out in Reference [29], only small amounts of acetate bands are observed, which could correspond to the weight loss of step II in TG curves (residual amount of 0.4 wt%, Tab. 2) and could be adsorbed on defective sites or polar surfaces of the material [29].

DFT Modeling
DFT modeling has been employed to understand the surface characteristics of the ZnS NP. The surface energies E s of the lowest index ZnS surfaces, the (110), (111) and (100) are 0.6, 1.0 and 1.7 J/m 2 , [29]. Dehydroxylated and "clean" surfaces were considered, which should model the situation corresponding to sample ZnS_673_V, that is of a sample activated in vacuum, with only a small fraction of reactancts adsorbed on reactive surfaces [29]. The size dependence of E s for the (110) and (111) surfaces is reported in Figure 5 (hereafter the 100 surface will be not considered anymore because of its very large E s ). Data reported in the figure, collected for d ranging from 0.5 to 6 nm, show that, in ZnS nanoparticles, the size of the particles does not affect the surface stability. The (110) surface is by far the most stable surface in nanocrystals as it is in microcrystal [55,59]. The large differences of the surface energies of the (110) and (111) surfaces is related to their geometrical structure. According to Tasker's electrostatic model [60], in zincblende systems only the (110) surface is non polar, whereas (111) and (100) are polar surfaces of type III. Polar surfaces are characterized by two not equivalent layers of opposite charged ions that alternate along the normal to the surface; each repeat unit bears a dipole moment that increases with the number of the layers and determines the larger instability of these surfaces. Considering the large difference in E s between the two surfaces, the (111) polar one is not likely to be exposed at the ZnS nanoparticles to a sizable extent, regardless of their size, Figure 5.
In order to shed light on the whole scenario laid out by infrared and TGA, suggesting the presence of a small fraction of strongly adsorbed species, the interaction with the (111) and (110) surfaces of formic acid has been considered. This molecule was employed as a simplified model to represent the acetate species present in solution during the synthesis, which are supposed to be the main species adsorbed on the dehydrated and cleaned ZnS surface on the basis of infrared spectroscopy. Adsorbed species are schematically depicted in Figure 6 and their main adsorption features collected in Table 3. On the (110) surface one dissociative  Table 3. Computed vibrational frequencies provide a fingerprint for the recognition of the adsorbed species, Table 3. For the 110-ND1 carboxylic group, the asymmetric stretching m as (O-C-O) = 1 700 cm À1 and the bending d(C-OH) = 1 237 cm À1 are found, while for the 111-D carboxylate group the asymmetric stretching m as (O-C-O) = 1 570 cm À1 is indeed shifted to lower wavenumbers with respect to the undissociated form. Only this latter vibration, the asymmetric stretching of dissociate formiate species, is compatible with FTIR data recorded upon the thermal treatment at 573 K (the peak at 1 564 cm À1 , Fig. 4c). The significant reduction of this peak only for treatment at very high T (Fig. 4d) indicates an irreversible character of the adsorption, in agreement with the strong computed adsorption energy for 111-D. This finding suggests the presence in the ZnS NP of small extension of the (111) polar surface stabilized by the adsorption of foreign species from the synthesis solution. To this extent, it is worth mentioning a recent report that combining FTIR observations and DFT calculations, has investigated the adsorption properties of CO molecules, used as a probe for acidic Lewis surface sites; this study has demonstrated that CO experienced only one type of sites belonging to the (110) surface [29].
The results discussed in the present work, confirm the finding that ZnS nanoparticles, regardless of their size, expose to a largely wider extent the most stable (110) surface;

Electronic Properties
The electronic properties of the prepared ZnS NP samples were investigated by DR UV-Vis spectroscopy (Fig. 7). This is a simple and widespread technique for the estimation of band gap energy (E g ) in semiconductors [34-37, 62, 63], by the measurement of the absorption edge position, through the Tauc plot elaboration (see Experimental for details) [38,39]. This parameter is quite important for applications in photocatalysis, and in semiconducting materials is dependent on particle size, due to quantum size effect [34-37, 40, 63]. Notwithstanding the small size of the primary ZnS NP prepared in this work, these cannot be defined as Quantum Dots (QD), since the UV-Vis spectra of the latter are characterized by the presence of discrete bands (described by the position of their maximum) instead of a continuum absorption (defined by its edge position) [37]. The spectrum of as prepared ZnS_6h is characterized by a relatively large E g value (4.22 eV, Tab. 1) with respect to the bulk material (3.68 eV) [40] as a result of nanostructuration [34]. On the contrary, longer reaction time and/or hot water washing result in a shift of the edge position towards higher wavelength values. Indeed, the corresponding E g values are close to that of bulk material for most of the other ZnS samples with crystallite size (estimated by XRD peak widths) between 3.4-4.9 nm (Tab. 1, entries 2-5). Interestingly, a shift to E g values higher with respect to bulk ZnS are observed after thermal activation, both in air (ZnS_573) or in vacuum (ZnS_673_V). This could be explained by a higher defectivity of the treated samples surface, influencing their electronic structure.
The considerable dependence of E g on the size of the ZnS nanoparticles is also confirmed by DFT calculations. Figure 8 reports E g as a function of the size of the nanoparticles: NP with d ! 4.5-5.5 nm have an energy gap that is actually as large as the bulk value, E g = 3.7 eV, in excellent agreement with observations [40]. In smaller particles, due to quantum size confinement effects, the energy gap increases monotonically by decreasing the size of the NP. For instance E g = 4.1 eV for d = 1.5 nm and 3.7 eV for d = 5 nm, in very good agreement with XRD estimation (notoriously underestimating particle size) and UV data (Tab. 1). Despite the sensitivity of E g to the size of the NP, the electron structure is almost unaffected by d. As reported in the insert of Figure 8, ZnS is characterized by a direct gap in C (see also [40]); the top of the valence band is due to S 3sp states and the bottom of the conduction band mainly from Zn 3sp states. The Zn 3d levels are very deep in energy (about À15 eV) and do not contribute to the valence.

CONCLUSIONS
ZnS NP photocatalysts were prepared by precipitation with TAA and acetate as sulfur and zinc sources, respectively, by modifying reaction time length. The obtained materials were then treated with hot water and thermally, both in air and in vacuum, to remove the reactants adsorbed on the surface of NP during the synthesis. The effect of these procedures on the properties of the ZnS NP was studied by combining different techniques and aspects.
The results show that this synthetic methodology allows to prepare nanostructured ZnS agglomerates, in the form of white powders, with good yield and ease of preparation. Low reaction time (6 h) resulted in agglomerated crystallites with primary size around 2 nm (as determined by XRD and TEM). Particle size was found to increase after washing with hot water, as a result of removal of large amounts of adsorbed chemicals. This was also evidenced by thermogravimetric analysis and infrared spectroscopy, indicating the presence of large amounts of TAA (around 18 wt%) weakly interacting with the surface. The sample prepared at longer reaction time (24 hours) was characterized by larger (around 4 nm), more densely aggregated and crystalline NP, which were less sensitive to the treatment with hot water.
Further analysis of the data showed that a small amount of acetate species (around 2 wt%) was strongly adsorbed on the ZnS NP surface, and could be removed almost totally only after thermal activation in vacuum at 673 K. This behavior was explained on the basis of DFT calculations, showing the strong adsorption of dissociate formic acid (chosen as a model molecule to simulate the acetate salt employed in the reaction) on the (111) polar surface. This surface is unstable with respect to the apolar (110) one, irrespective to the size of the NP, (Fig. 5), but could be stabilized in a small amount by the adsorption of foreign species from the synthesis solution.
Finally, the electronic properties of the prepared materials were analyzed by DR UV-Vis spectroscopy, by calculating the band gap energy E g with the Tauc plot method. This shows that only the sample prepared at 6 hours reaction (ZnS_6h) without any post-treatment has an E g value sensibly higher with respect to bulk ZnS (4.2 vs 3.7 eV), as a result of nanostructuration. When the size of the primary crystallite moves to 3-4 nm as a result of washing with hot water, longer reaction time or employ of Na 2 S, the band Energy gap, E g , as a function of the particle size. The dashed red line is the value for bulk ZnS. In the insert band structure and projected density of states for bulk ZnS. B3PW calculations.
gap of the materials corresponds to the bulk one. This trend is rationalized by DFT calculations on ZnS NP, predicting the modification of the semiconducting electronic structure as a function of particle size.

ACKNOWLEDGMENTS
Italian Ministry for Research and University (MIUR) is gratefully acknowledged for funding project PRIN 2009BLNJC5 "Optimization of the photothermocatalytic sulfur-ammonia process for hydrogen production". A.M. F. acknowledges CINECA (Iscra project B HOME) for computational resources.