An Investigation into the Selenization Mechanisms of Wurtzite CZTS NanorodsGerard Bree, Claudia Coughlan, Hugh Geaney, and Kevin M. Ryan
ACS Appl. Mater. Interfaces, Just Accepted Manuscript • DOI: 10.1021/acsami.7b18711 • Publication Date (Web): 02 Feb 2018
Downloaded from http://pubs.acs.org on February 6, 2018
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are postedonline prior to technical editing, formatting for publication and author proofing. The American ChemicalSociety provides “Just Accepted” as a service to the research community to expedite the disseminationof scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear infull in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fullypeer reviewed, but should not be considered the official version of record. They are citable by theDigital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,the “Just Accepted” Web site may not include all articles that will be published in the journal. Aftera manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Website and published as an ASAP article. Note that technical editing may introduce minor changesto the manuscript text and/or graphics which could affect content, and all legal disclaimers andethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors orconsequences arising from the use of information contained in these “Just Accepted” manuscripts.
1
An Investigation into the Selenization Mechanisms of Wurtzite CZTS Nanorods
Gerard Bree, Claudia Coughlan, Hugh Geaney, and Kevin M. Ryan*
Bernal Institute, University of Limerick, Limerick, Ireland
ABSTRACT: Here we report the first detailed investigation into the selenization mechanism of thin films of wurtzite cop-per zinc tin sulfide (CZTS) nanorods (NRs), giving particular emphasis to the role of the long-chain organic ligands sur-rounding each NR. During selenization, the NRs undergo a selenium-mediated phase change from wurtzite to kesterite, concurrent with the replacement of sulfur with selenium in the lattice and in-situ grain growth, along with the recrystalli-zation of larger copper zinc tin selenide (CZTSe) kesterite grains on top of the existing film. By utilizing a facile ligand removal technique, we demonstrate that the formation of a large grain overlayer is achievable without the presence of ligands. In addition, we demonstrate an elegant ligand-exchange based method for controlling the thickness of the fine grain layer. This report emphasizes the key role played by ligands in determining the structural evolution of CZTS nano-crystal films during selenization, necessitating the identification of optimal ligand chemistries and processing conditions for desirable grain growth.
Keywords: Cu2ZnSn(S,Se)4, wurtzite, nanoparticle, selenization, thin film solar cells, ligand, crystal phase evolution
INTRODUCTION
CZTSSe (Cu2ZnSnSxSe4-x) has been widely studied as a high performance and low cost absorber material for solar cell applications1,2,3,4,5 with efficiencies above 12% for the best devices.6 One particularly attractive route to fabricat-ing this material involves the use of solution processed nanocrystals (NCs), exploiting wet chemical approaches to form low-cost building blocks via facile synthetic methods.7,8,9 With this approach, CZTSxSe1-x NCs (most often in the kesterite crystal phase) are synthesized as a first step, after which they are deposited to form a thin film. CZTS NCs can also be synthesized in the wurtzite phase,10,11,12,13 which is of particular interest due to the rela-tive ease of NC shape and stoichiometric control.14,15,16 This in turn allows engineering of the Fermi level and bandgap, enabling novel optical properties.17 Further-more, a monodisperse nanorod (NR) morphology is uniquely stable in the wurtzite phase,13,18 enabling dense, highly ordered film formation.16,19 Thin films of CZTS NCs are usually formed by deposition of the NC’s onto molyb-denum (Mo) substrates followed by annealing at high temperatures (>500 °C). Annealing can be carried out in the presence of added sulfur10 or selenium20 (processes known as “sulfurization” and “selenization” respectively), in order to increase the crystal grain size in the films. This in turn reduces defect and grain boundary concentration, enhancing the charge transport characteristics of the film and improving device performance.21,22 In nanoparticle films, selenization is the most widely used process and usually results in the formation of a bi-layer structure with bi-modal grain size, consisting of a large-grain upper
layer (LGL) formed by recrystallization of “abnormal” grains, and a fine-grain underlayer (FGL), formed by in-situ growth of “normal” grains.23,24 The LGL (consisting of grains typically ~ 1 µm in size) is responsible for photon absorption and charge separation; however the role of the FGL in the completed cell is unclear. FGLs are often rich in graphitic carbon,25,26 meaning that they allow charge carriers to pass through the layer, however they may not be photovoltaically active27 and can significantly add to the series resistance of the cell.22 Control over (including full elimination of) the carbon-rich FGL has been demon-strated previously, involving the use of formamide,28 sul-fide anion29 and dodecylamine25 capped NCs, by ligand removal,9,30 and by varying selenium pressure during selenization.24 However, despite the development of these control techniques, the best performing CZTSSe NC-based solar cells fabricated to date have demonstrated this bi-layer morphology,8,9 in contrast with record cells fabricated by other methods, which exhibit a single layer of large grains.6,31 Therefore a deeper understanding of the mechanism of bi-layer formation, and morphological evo-lution associated with CZTS NCs in general, is desirable.
Uniquely in the case of wurtzite NCs, high temperature treatments can induce phase-transition driven grain growth due to the metastable nature of the phase,13 and have yielded fully sintered films10,11 with grains that span the entire depth of the layer.12 Although solar cell perfor-mances to date have typically been lower for these cells (record is 4.83%10), this phase offers the potential to form films morphologically similar to the highest performing CZTSSe cells,6,31 and the existence of a morphological evo-
lution distinct from kesterite NC-based cells warrants further investigation.
The presence of long-chain organic ligands adsorbed on the NC surface can play a significant role in morphologi-cal changes during the selenization process. Previously, ligands have been thought to slow or prevent annealing of NCs due to the physical separation between adjacent NCs.13 More recently it has been suggested that the physi-cal presence of ligands (or rather the nanoporosity and resultant high surface area that they provide) is a key fa-cilitator of the growth of abnormal grains.9 The chemical nature of the ligand can also play a role in determining the morphology and composition of the post selenization film,25,28,24,30 as its decomposition characteristics can alter the kinetics of the selenium-NC interactions.32 In addi-tion, the oleylamine ligands most commonly used in CZTS synthesis have been found to remain in the post-selenization film in the form of graphite crystallites25 and amorphous carbon.27 This transformation has been shown to be dependent upon ligand reactivity, with more reac-tive oleylamine resulting in larger graphite crystallites and greater device performance.25 Given these, the relative lack of studies into the role played by ligands in the selenization process, in particular the alkanethiol ligands crucial for synthesis of wurtzite CZTS NCs,33 indicates that this area warrants further investigation.
Herein we report an ex-situ examination of the seleniza-tion process of wurtzite CZTS NR films, observing a phase change to kesterite concurrent with both the beginning of normal and abnormal grain growth, and the replacement of the majority of sulfur with selenium in the CZTSSe lat-tice. By comparing NR films annealed with and without the presence of selenium, both the abnormal recrystalli-zation of large CZTSSe grains on top of the existing film and normal grain growth within the layer itself are found to be selenium mediated. The role of the organic ligands in the film during selenization is particularly emphasized, in that a transformation to disordered graphite crystallites occurs, which subsequently grow in size as the process continues. Furthermore, by utilizing a ligand removal treatment, we demonstrate that large grain growth can be achieved with carbon-free films. Finally, we show that the fine grain layer thickness can be controlled by modulating the ligand chain length, thereby demonstrating a facile chemical method of morphological control. This will ena-ble the development of nanocrystal synthesis procedures specifically designed to tune absorber film properties.
EXPERIMENTAL METHODS
Chemicals. All chemicals were purchased from Sigma Aldrich except octylphosphonic acid & octade-cylphosphonic acid (PCI Synthesis).
CZTS NR Synthesis. CZTS NRs were synthesized accord-ing to a modified method previously published.18,13 Briefly, in a typical synthesis, 236 mg copper (II) acetylacetonate,
106 mg zinc acetate, 178 mg tin (IV) acetate and 1353 mg tri-octylphosphine oxide (TOPO) were mixed in 10 mL 1-octadecene in a 25 ml 3-neck flask. The flask was then evacuated for 30 mins using a Schlenk line, after which they were exposed to an argon atmosphere. This follows the “Cu-poor, Zn-rich” (Cu/[Zn+Sn] ≈ 0.8 & Zn/Sn ≈ 1.2) formulation, which has been found to result in the high-est performing CZTSSe solar cells.34 The temperature was then ramped up to 270 °C, during which a mixture of 1.75 ml tert-dodecylmercaptan and 0.25 ml 1-dodecanethiol was injected into the flask at 155 °C. The solution immedi-ately turned from dark green to clear yellow after the in-jection. The reaction was allowed to proceed for 30 mins after injection, before being naturally cooled to 100 °C and dispensed into a vial.
Film Deposition. The as-synthesized NRs were typically washed twice to remove excess ligands and unreacted materials. For the first wash, isopropanol was added to the product at a ratio of 1:2, followed by vortexing and sonicating (1 min each), before the NRs were isolated by centrifugation at 4,000 rpm for 5 min. After redispersing in toluene, a 2nd wash, using 2:1 isopropanol to toluene, was performed. The centrifuged NRs were then dried un-der a flow of argon for 10 mins, before being dispersed in an ink (200 mg/ml in toluene or hexanethiol). Layers were deposited by doctor blading using a strip of scotch tape (~50 µm) as a spacer. The film was allowed to dry naturally at room temperature for ~5 minutes, before be-ing placed on a hotplate at 300 °C (for hexanethiol ink) or 100 oC (for toluene ink) to ensure full solvent evaporation. Typically, two coats were deposited to obtain a layer 1-1.5 µm thick.
Ligand Exchange. After the first wash and redispersing in toluene, an 8 ml extract was taken and 50 mg/ml of the required ligand (e.g. an alkylphosphonic acid) was added, at which point the dispersion was sonicated for 30 mins. The 2nd wash was then performed as before.
Ligand Removal. Ligand removal was performed at one of two points in the process, “post-deposition” (denoted “Type-2” films) or “pre-deposition” (denoted “Type-3” films). “Type-1” films did not undergo ligand removal. To form Type-2 films, we utilized a method previously pub-lished,35 in which the dried films were immersed in a 20 mM solution of ammonium sulfide in methanol for 30 s, followed by rinsing in pure methanol, after each coat. To form Type-3 films, the NRs were washed once using a 1:2 ratio of product to isopropanol, and after redispersal a 1 ml solution of 70 µL ammonium sulfide (40%) in metha-nol was added to the NR dispersion. The dispersion was sonicated for 30 mins, after which more IPA was added. The NRs were then isolated by centrifugation, and pro-cessed into an ink as normal.
Selenization. In order to avoid oxidation, all films were stored in a glovebox prior to further treatment. The sub-
strates were placed onto a graphite plate in a quartz box (~100 x 50 x 20 mm external volume) along with 2 seleni-um pellets (50 +/- 5 mg each). This was then placed in the cold zone of a tube furnace and evacuated for 30 mins before being filled with 100 mbar argon. The box was then moved into the preheated hot zone of the furnace and left for the desired time. During selenization, a yellow/orange color was visible in the box, indicating the presence of selenium vapor. Typically the thin films of black/brown NRs turned grey during the process, indicating that CZTSe has formed. In addition, a matte surface is ob-served if large grains are present.
Annealing without Selenium. In order to investigate the effect of selenium on the evolution of film morpholo-gy during heat treatment, an identical process to that of selenization was performed, however no Se pellets were added, i.e. films of NRs on Mo were placed into a quartz box, which was inserted into the preheated hot zone of tube furnace under a 100 mbar argon atmosphere. To achieve a higher ramp rate, substrates were placed on a hotplate preheated to 450 °C in an Argon-filled glovebox for 15 mins.
Characterization. Transmission electron microscopy (TEM) was conducted by using a 200 kV JEOL JEM-2100F field emission microscope, equipped with a Gatan Ultras-can CCD camera. Scanning electron microscopy (SEM) was performed with a Hitachi SU-70 system equipped with an Oxford Instruments EDS detector, X-Ray diffrac-tion (XRD) with a PANalytical X’Pert PRO MPD instru-ment with a Cu Kα radiation source (λ = 1.5418 A) with a 1-D X’celerator strip detector. Raman spectroscopy was carried out using a Horiba Labram 300 spectrometer sys-tem equipped with a 633nm laser, Fourier Transform In-frared (FTIR) spectroscopy with a Perkin Elmer Spectrum 100 spectrometer and X-ray Photoelectron Spectroscopy (XPS) with a Kratos Axis 165 spectrometer. Thermogravi-metric Analysis was performed using a TA TGA-Q50.
RESULTS AND DISCUSSION
The CZTS NRs obtained in this synthesis were highly monodisperse (Figure 1a and b) with dimensions of ap-prox. 50 x 10 nm. The thin films formed by selenization (at different times 5, 10, 20 and 30 mins) of the deposited rods are shown in the top-down SEM images in Figure 1 (c-f). Initially, the as-deposited film consisted of a uni-form layer of randomly orientated NRs approx. 1.5 µm thick (Figure 1c and 1i). 5 mins after insertion, the tem-perature inside the box had reached 430 oC (as measured by a thermocouple placed inside the box in a separate calibration test, see Figure S10 for temperature profile), indicating a selenium vapor pressure of 13 mbar; however the morphology of the film had not yet changed (Figure 1d). Typically, the yellow/orange selenium vapor was first visible to the naked eye at ~ 7 mins (at a temperature of 490 oC, Se vapor pressure of 40 mbar). Abnormal grain growth was first observed at 10 mins as shown in Figure
Figure 1: TEM images of (a) NR ink, and (b) a single NR. Top-Down SEM images showing the CZTS films selenized at a setpoint of 550 oC for (c) 0 mins, (d) 5 mins, (e) 10 mins, (f) 20 mins, and (g) 30 mins. (h) shows the changes in the atom-ic ratio of chalcogen to total metal (Cu+Zn+Sn) content as measured by EDS. All EDS data was obtained from the FGL only, except in the sample selenized for 30mins, where con-tributions from both FGL and LGL were unavoidable (see SI for more information). Cross sectional SEM images of films selenized for 0 mins and 30 mins are shown in (i) and (j)
1e, at which time the grains were relatively small (100-500 nm). From 10 to 20 mins, the abnormal grains grew in size up to 1-2 µm (Figure 1f); however the coverage was still relatively low (i.e. the fine grain underlayer was still visi-ble in most areas). After 30 mins, the individual grains had not grown in size any further; however a dramatic increase in layer coverage was observed, almost complete-ly obscuring the FGL (Figure 1g). The formation of this bi-
layer structure is clearly observed in the cross-sectional SEM image in Figure 1j. Energy-dispersive X-ray spectros-copy (EDS) was used to ascertain the changes in the chal-cogenide/metal ratio (atomic percentages) during the selenization process (Figure 1h). The EDS data indicated that a large amount of Se was added to the film in the first 5 minutes, which includes contributions from both atoms incorporated into the lattice and elemental Se that had condensed during cooling. Interestingly, little S had been lost from the film at this point (the sulfur/metal atomic ratio falling from 0.88 to 0.84), indicating that the majori-ty of Se was in elemental form. A significant decrease in S content (to a ratio of 0.23) was observed from 5 to 10 mins, which suggested that the majority of S was lost dur-ing this time. The replacement of S with Se had effectively terminated at 20 mins, falling only slightly from 0.07 to 0.06 at 30 minutes. At this point, the selenium to metal atomic ratio was 0.87 (approx. the value for S pre-selenization), indicating that all Se present in the film had been incorporated into the lattice.
The changes in the morphology of the FGL associated with the selenization process are shown in top-down SEM images in Figure 2. Again, the major morphological changes occurred between 5 and 10 mins after insertion, in which the NRs were transformed into small grains of 50-100 nm in size. An evolution of the NC shape associat-ed with a crystal phase change from wurtzite to kesterite is predicted based on the number of bonds of surface
Figure 2: Top-Down SEM images showing the “fine-grain layer” of the CZTS films selenized at a setpoint of 550 oC for (a) 0 mins, (b) 5 mins, (c) 10 mins, (d) 20 mins, and (e) 30 mins. (f) shows an SEM image of a NR film heated in the same manner as the film in (e), but without the addition of selenium
the surface atoms on the sidewalls, consisting of (100) planes, have 3 bonds. However, upon phase change to kesterite, the reorganization causes this to reduce to 2 bonds, and a change in shape to form surfaces of (112) planes is preferred.36 The grains then increased in size to 100-200 nm, after which they did not undergo any further growth. It is important to note that this growth was Se-mediated, as a control film subjected to the same 30 min heating process without the addition of Se pellets (Figure 2f) remained in NR form.
X-ray Diffraction (XRD) analysis was used to further in-vestigate the structural changes in the film during the selenization process (Figure 3). The as-synthesized NRs (Figure 3a - 0 mins) were in the wurtzite phase, with peaks corresponding to the (100), (002), (101), (110), (103) and (112) planes evident in the XRD pattern. Peak broad-ening (Scherrer) due to the nanoscale lengths of the NRs was observed for all reflections except the (002), confirm-ing an elongation in the z-direction (see Figure S1). This elongation was maintained after 5 mins selenization (Fig-ure 3a – 5 mins), indicating that the NR morphology was retained. A phase change from wurtzite to kesterite oc-curred from 5 to 10 mins, as observed visually in the XRD patterns in Figure 3a, which was also concurrent with the appearance of Se vapor in the quartz box and the for-mation of normal and abnormal grains.
In order to allow further examination of the phase and morphological changes during this key stage (5 to 10 mins) in the process, an identical film was exposed to a slower (~ 0.3 K s-1) ramp to 450 oC in the presence of Se, after which the sample was removed. XRD and SEM anal-ysis (Figure S2) of the film confirmed the simultaneous occurrence of both the wurtzite and kesterite phases, as well as the presence of both NRs in the fine grain layer and small abnormal grains on the surface. This confirms that the initial recrystallization of abnormal kesterite grains can in fact begin prior to phase and morphological changes in the FGL. After 5 mins, the shift of the (100) wurtzite reflection (Δ 2θ = 0.2o, corresponding to a d-spacing change of ~0.03 Å) when compared with its initial position is consistent with the incorporation of the larger selenium atom.37,38 This indicates that some selenium had vaporized and reacted with the film at this point (at 400 °C the vapor pressure of selenium was ~4 mbar). This ef-fect was also observed in the change of the kesterite (112) peak position from 10 to 20 mins, whereby a similar shift of 0.2o was evident, which indicated further selenium in-corporation. Scherrer analysis was conducted to quantify the change in the crystallite size after selenization. No change of the crystallite size was observed between 0 and 5 mins for the wurtzite (100) peak, whereas a significant increase in the crystallite size was observed at 10, 20 and 30 mins respectively for the kesterite (112) peak, as shown in Figure 3b.
Figure 3: (a) XRD spectra obtained from the 5 films which underwent selenization for varying lengths of time, the intense Mo peaks at ~41o are omitted for clarity. The reference spectra for wurtzite ZnS and kesterite CZTSe are also included for compari-son. (b) The same spectra in the Mo and MoSe2 peak range, demonstrating the selenization of the Mo back contact. (c) Full width at half maximum (FWHM) and the crystallite size calculated by peak fitting and Scherrer analysis (D= Kλ/[Bcosθ], where D is the crystallite size, K is the shape factor, λ is the incident X-ray wavelength, B is the FWHM of the peak in question and θ is the Bragg angle). The peaks analyzed were the wurtzite (100) (for 0 and 5 mins) and the kesterite (112) for the remain-der
This trend in both the XRD and Scherrer analysis data is consistent with the visual changes in the film from the SEM analysis (shown previously in Figure 2).
The XRD study also demonstrated the selenization of the Mo layer.39 Initially only the Mo (011) peak was visible in the XRD spectra in Figure 3c, however after 20 mins selenization it had almost disappeared, and was replaced by the (101), (012) & (110) reflections of MoSe2.
40 The delay time between the selenization of the NR film, and that of the Mo, is likely associated with the time taken for con-densed selenium to percolate fully through the CZTSSe layer.41 The formation of this MoSe2 layer separating the absorber layer and the elemental Mo is a necessary step in forming an ohmic back contact, however the material itself is a poor conductor and hence a thin layer is favora-ble.40,42 It has been demonstrated previously that its thickness can be effectively controlled by variation in selenization temperature.40
Given the challenges in distinguishing secondary phases from XRD analysis alone,43 Raman spectra for the various films were also obtained (Figure 4). Both the expected chalcogenide wavenumber range (150-500 cm-1), as well as a larger wavenumber region which enabled the investiga-tion of the graphitization of the organic material,25 were
examined. The spectra confirmed the transformation from CZTS to CZTSe (Figure 4a), including a redshift of the CZTS A1 Raman mode peak from 332 cm-1 at 0 mins to 327 cm-1 at 5 mins, again associated with replacement of sulfur with the heavier selenium atom.38 Similarly, after 5 mins a broad feature centered around 209 cm-1 was ob-served, and is likely a blueshifted A1 mode CZTSSe peak (typically the A1 mode of CZTSe is observed at 196 cm-
1,9,38). Small features at 252 cm-1 and 258 cm-1 after 10 mins indicated the presence of secondary phases ZnSe and Cu2Se phases, however by this time the majority of sele-nium was in in the form of CZTSe. The presence of these small impurities is supportive of a proposed mechanism in which liquid Se allows transport of cations to the sur-face, in the form of binary selenides, leading to the ob-served growth of large CZTSe grains on top of the layer.9,44 These peaks disappeared after 20 mins, indicating the incorporation of all cations into the CZTSe grains. Analy-sis of the larger wavenumber region confirmed the for-mation and growth of disordered graphite crystallites (Figure 4b), originating from the organic content in the ligand shell surrounding each NR. Raman peaks for disor-dered graphite typically appear at 1,580 cm-1 (graphene, “G” peak) and 1,370 cm-1 (disorder induced mode, “D” peak).45
Figure 4: Raman spectra obtained for the films pre-selenization and selenized for 5, 10, 20 and 30 mins. Spectra in the short (a) and long (b) wavenumber regions are shown. Dotted lines indicate the positions of peaks associated with relevant compounds.
The pre-selenization film showed no significant peaks in this region, however after 5 mins a peak had appeared at 1,423 cm-1 along with a small feature at ~1,550 cm-1. A pre-vious study has attributed these features to the D and G peaks of graphite,25 with the shift in the position of the D peak relative to its expected position caused by interac-tions with the CZTSSe NCs. The ratio of the peak intensi-ties (ID > IG) is indicative of the level of disorder present, and of relatively small graphite crystallites.46 Interesting-ly, the position of the D peak shifted to its natural posi-tion after 20 mins, and also reduced in intensity relative to the G peak, suggesting that the crystallites grew during the latter stages of the selenization process.
It has previously been observed9,25,28,30 that the nature of organic material present in the CZTS NC film can play an extremely significant role in structural changes observed during the selenization process. However, the relative lack of systematic studies into the role of the organic lig-ands in achieving effective grain growth has thus far con-strained the design of optimal ligand chemistries, sug-gesting that a deeper understanding of this mechanism would be beneficial. This is particularly important given that the organic matter remains part of the final selenized
film in the form of graphite. In order to investigate this, a chemical method of stripping the alkylthiol and phos-phine molecules from the NR surface (utilizing ammoni-um sulfide) was employed. This method has been shown to replace the long-chain ligands by short metal-sulfur bonds and reduce the inter-particle distance significant-ly.35 The treatment was performed on CZTS NRs dispersed in a solvent, and on NR films post-deposition (see Exper-imental Methods for details). FTIR, TGA and EDS data indicated the successful removal of organic content from the NR surface (see Figure S4 & S5). Three types of films were deposited; one in which ligands were present (“Type-1”), one in which they were stripped after deposi-tion (“Type-2”) and one in which they were stripped prior to deposition (“Type-3”).
Pre-selenization SEM analysis of the NR films revealed that the Type-1 and Type-2 films had identical structure and morphology, as shown in Figure 5a and 5b, respec-tively, in which the NRs were randomly orientated and the film was uniform and nanoporous. However, in the Type-3 film (Figure 5c) larger aggregates of NRs (~100-200 nm in diameter) were formed immediately upon addition of ammonium sulfide, resulting in a layer which was far
less uniform and contained pores on the order of hun-dreds of nm. The three film types were then selenized and analyzed. The design of these films allowed decoupling of the effects of (a) layer morphology and porosity caused by ligands, and (b) the presence of the ligands, on the seleni-zation process. This was possible because the Type-2 film had low ligand content but maintained high nanoporosi-ty/surface area, whereas the Type-3 film had both low ligand content and low nanoporosity/surface area.
The SEM images (Figures 5 d,e,f,g,h & i) show that the morphology of the selenized films was significantly al-tered by the ligand environment. In all cases, complete (or almost complete) selenization (replacement of sulfur with selenium) occurred, as observed in the EDS in Fig-ures 5 j, k & l. The Type-1 film formed a bi-layer structure, consisting of an upper LGL and a lower carbon-rich FGL. On the contrary, the Type-3 film formed few large grains and exhibited an extremely porous layer morphology (Figure 5 f & i). Interestingly, selenization of the Type-2 film resulted in an intermediate situation (Figure 5 e & h), in which a bi-modal grain size distribution was observed, whereby large grains did form but were isolated from one
another for the most part. The ligand removal treatment allowed for the elimination of organic content from both the Type-2 and Type-3 selenized films (note the lack of any carbon in the EDS data, Figures 5k & l, compared with its presence in 5j). However, by performing the treatment after deposition (Type-2), a degree of nanopo-rosity sufficient for widespread large abnormal grain growth was maintained. In addition, the poor pre-selenization layer morphology observed in the Type-3 film (i.e. large pores) means that liquid selenium/copper sele-nide likely does not pool on top of the NRs, meaning the reservoir for abnormal grain growth was not present. This further supports the hypotheses9 that the formation of larger grains on top of the layer can only proceed in films with a high surface area along with the presence of liquid chalcogen acting as a cation transport medium. It is par-ticularly interesting to note that the formation of the bi-layer structure occurred with and without the presence of organic material. The EDS demonstrates that, in general, the large grains were close to stoichiometric CZTSe com-position, whereas the fine grain layer was extremely Cu- poor and Zn-rich.
Figure 5: SEM images and EDS data for layers of CZTS NRs selenized at a setpoint of 550 °C for 40 mins. Top-down (a, b, c & d, e, f) and cross-sectional (g, h & i) images comparing films before and after selenization, in cases where ligands were present and removed before and after film deposition. Charts (j), (k) and (l) show the elemental composition of the fine and large grain por-tions of the layers as determined by EDS (note that the large grain layer measurements will include some signal from the fine grain layer). In order to perform analysis of the fine grain layer in (d), a location free from large grains was examined.
Figure 6: Cross sectional SEM images of CZTS NR films annealed at a setpoint of 550 °C for 40 mins under 0.1 bar Argon in which the native thiol ligands were (a) present, (b) removed after deposition, and (c) removed prior to deposition. Images (d), (e) and (f) are of similar films annealed on a hotplate preheated to 450 °C for 15 mins in 1 bar Argon atmosphere.
While the nucleation and growth of normal crystal grains in the FGL is also selenium mediated, it appears to be restricted by organic content acting as a barrier/separator. The crystallites in the FGL were largest (100 - 400 nm, Figure 5i) in the Type 3 film, compared with crystallites of 50 - 100 nm (Figure 5 g&h) in the other two films. The NRs in the Type 2 film had aggregated prior to deposition, and indeed orientated assembly can be observed in places (see Figure S6a). This effectively reduced the interparticle distance to zero, and enabled a faster in-situ crystal growth rate (while maintaining a similar nucleation rate). In fact, isolated grains with diameters similar to those found in the large grain layer of the other films were ob-served evenly spread throughout the depth of the film (see Figure S6b).
In order to further assess the interaction between chalco-genide vapor and ligands during heat treatment, the 3 types of film were also subjected to a similar heating pro-cess, but without the presence of selenium, and examined by SEM (Figure 6 a-c respectively). In all cases the NR morphology did not change and an XRD measurement upon cooling confirmed the sole presence of the wurtzite phase (see Figure S7), despite the relatively high tempera-ture heating (550 oC).
This need for selenium to facilitate normal grain growth and the phase change in the fine grain layer can be over-come if the heating ramp rate is sufficient, as demonstrat-ed previously.13 Placing the 3 NR films onto a hotplate preheated to 450 °C for 15 mins in an argon atmosphere resulted in a higher rate of heat transfer to the films (con-ductive heat transfer as opposed to solely radiative). With this process, kesterite grains were formed in each of the three cases, however they were notably larger when lig-
ands were removed (100-500 nm) as opposed to when they were present (<100 nm) (Figure 6 d-f). The Type 2 film exhibited properties particularly suited to high ramp-rate induced grain growth (Figure 6e), namely a high par-ticle density and low organic content.
The inter-particle distance in the film is determined by the length of the ligand molecules in the organic shell surrounding each NR (~ 1 nm for dodecanethiol ligands). This distance can be reduced to zero by stripping of the ligand entirely (as observed in the Type-3 film in Figure 5), however by varying it in more discrete steps, a deeper understanding of the mechanism, as well as the identifi-cation of ligands for optimum film structure, are enabled. Previous studies with similar systems16,35 have shown that ligand molecules can be effectively exchanged for others, altering the chemical and physical properties of the or-ganic shell, but not the NC itself.
In order to investigate the effect of ligand length on selenization, an as synthesized batch of NRs was divided into 3, and each underwent a ligand exchange from the native thiol to a) methylphosphonic acid (MPA), b) oc-tylphosphonic acid (OPA) and c) octadecylphosphonic acid (ODPA) (see schematic in Figure 7a). These mole-cules have theoretical lengths of 0.3 nm, 1.1 nm and 2.5 nm respectively (while maintaining the same functional groups). This particular ligand family was chosen as they have previously been shown to effectively replace thiol adsorbed on CZTS NRs.16 The ligand exchange process was successful (as determined by FTIR, TGA & EDS, see Figure S8), with only small residues of thiol present.
The NRs were dispersed in toluene (200 mg/ml), forming inks for doctor blading (here toluene was used as the
Figure 7: (a) Schematic demonstrating the process of ligand exchange and selenization, for three films with phosphonic acid ligands of varying length. (b–d) Cross sectional SEM images of the selenized films of CZTS NRs with methylphosphonic acid, octylphosphonic acid and octadecylphosphonic acid ligands. The thickness of the FGL is indicated.
solvent instead of the more common hexanethiol in order to avoid any interactions/exchanges that may arise from exposing the NRs to another potential thiol ligand). Lay-ers were then doctor bladed onto Mo substrates as before. NRs with OPA and ODPA ligands formed uniform films similar to those with thiol, whereas for MPA-passivated NRs, some aggregation was observed, resulting in some-what porous and non-uniform films (see Figure S9).
Figure 7 b-d shows cross sectional SEM images of the re-sultant selenized films. Films with phosphonic acid lig-ands formed large abnormal grains of 600-1,000 nm inde-pendent of ligand length, indicating that percolation of liquid selenium was successful in all cases. However the thickness of the FGL increased with ligand length (200 nm for MPA, 450 nm for OPA and 800 nm for ODPA). The LGL to FGL thickness ratio increased from approx. 1:1 in the case of ODPA to 3:1 in the case of MPA. It is im-portant to note that this was likely not simply related to the total organic content of the precursor layer. This is because, as we have shown above (Figure 5h), the Type-2 film with zero carbon content in the precursor layer ex-
hibited a bi-layer structure with a relatively thick carbon-free FGL (up to ~1 µm). Furthermore, the elimination of the FGL has been demonstrated on carbon-rich precursor films by modulation of selenization parameters,24 suggest-ing that alternate mechanisms can dominate. We propose that, while all ligands allowed for selenium percolation through the layer, a small ligand maximized its physical contact with the NR surface. This enabled increased cati-on mobility and hence an increased rate of abnormal grain growth. In turn, this increased rate meant that a larger portion of the cation content was transported out of the FGL prior to the termination of the process, result-ing in a thinner FGL. Thus, ligand exchange in dispersion was an elegant method for controlling the fine grain layer of the nanoparticle processed layer.
CONCLUSIONS
An ex-situ study into the process of selenization of wurtz-ite CZTS NR films revealed a selenium-mediated phase change to kesterite concurrent with the replacement of sulfur with selenium, and the growth of both normal and abnormal grains. More specifically, the cation mobility
provided by the presence of liquid selenium in the layer allowed both; (a) small kesterite CZTSe grains to grow in-situ (i.e. due to the fusing of adjacent NRs) and (b) the recrystallization of larger kesterite grains on top. Grain growth terminated at 100 – 200 nm in the FGL and 1 – 2 µm in the LGL.
We have demonstrated several facile methods of control-ling the structure and morphology of the post-selenization film by altering the ligand environment. Lig-ands can be removed chemically after deposition, result-ing in a carbon-free film with desired large grains. Fur-thermore, modulating ligand molecule length enabled control of the final thickness of the fine grain layer. This has implications for solar cell applications given that this layer can contribute to the parasitic series resistance of the cell. In particular, chemically exchanging to a short ligand molecule, methylphosphonic acid, enabled mini-mization of the inactive carbon-based portion of the ab-sorber film, while maintaining the large grain growth that would be necessary for high efficiency solar cells.
ASSOCIATED CONTENT
Supporting Information. Table quantifying FWHM of XRD peaks at 0 and 5 mins selenization, XRD spectrum and SEM images of NR film selenized under a slow heating rate to 450 oC, XPS spectra of CZTS film pre and post selenization, Ra-man spectra of ex-situ samples, FTIR, TGA & EDS data of NRs pre and post ligand removal, further SEM images of films with ligands removed pre-deposition, XRD spectra of NRs annealed at 550 oC without selenium, FTIR, TGA and SEM analysis of NRs after ligand exchange to phosphonic acids. Measured temperature profile inside quartz box during selenization. SEM image and XRD spectrum of a sulfurized NR film. This material is available free of charge via the In-ternet at http://pubs.acs.org
The manuscript was written through contributions of all authors. / All authors have given approval to the final version of the manuscript.
Notes
The authors declare no competing financial interest
Funding Sources
The authors would like to acknowledge Science Foundation Ireland (SFI) Grant No. 11-PI-1148 for funding this research.
REFERENCES
(1) Jiang, M.; Yan, X. Cu 2 ZnSnS 4 Thin Film Solar Cells : Present Status and Future Prospects. Sol. cells - Res. Appl. Perspect. 2013.
(2) Liu, X.; Feng, Y.; Cui, H.; Liu, F.; Hao, X.;
Conibeer, G.; Mitzi, D. B.; Green, M. The Current Status and Future Prospects of Kesterite Solar Cells : A Brief Review. 2016, No. January, 879–898.
(3) Wu, S.; Chang, C.; Chen, H.; Shih, C.; Wang, Y.; Li, C.; Chan, S. High-Efficiency Cu2ZnSn(S,Se)4 Solar Cells Fabricated through a Low-Cost Solution Process and a Two-Step Heat Treatment. 2017, No. September 2016, 58–66.
(4) Larramona, G.; Levcenko, S.; Bourdais, S.; Jacob, A.; Choné, C.; Delatouche, B.; Moisan, C.; Just, J.; Unold, T.; Dennler, G. Fine-Tuning the Sn Content in CZTSSe Thin Films to Achieve 10.8% Solar Cell Efficiency from Spray-Deposited Water – Ethanol-Based Colloidal Inks. Adv. Energy Mater. 2015, 5 (24), 1–10.
(5) Coughlan, C.; Iba, M.; Dobrozhan, O.; Singh, A.; Cabot, A.; Ryan, K. M. Compound Copper Chalcogenide Nanocrystals. Chem. Rev. 2017.
(6) Wang, W.; Winkler, M. T.; Gunawan, O.; Gokmen, T.; Todorov, T. K.; Zhu, Y.; Mitzi, D. B. Device Characteristics of CZTSSe Thin-Film Solar Cells with 12.6% Efficiency. Adv. Energy
Mater. 2014, 4 (7), 1–5.
(7) Guo, Q.; Ford, G. M.; Wei-Yang, C.; Walker, B. C.; Stach, E. A.; Hillhouse, H. W.; Agrawal, R. Fabrication of 7.2% Efficient CZTSSe Solar Cells Using CZTS Nanocrystals. J. Am. Chem. Soc 2010, 132, 17384–17386.
(8) Miskin, C. K.; Yang, W. C.; Hages, C. J.; Carter, N. J.; Joglekar, C. S.; Stach, E. A.; Agrawal, R. 9.0% Efficient Cu2ZnSn(S,Se)4 Solar Cells from Selenized Nanoparticle Inks. Prog.
Photovoltaics Res. Appl. 2015, 23 (5), 654–659.
(9) Hages, C. J.; Koeper, M. J.; Miskin, C. K.; Brew, K. W.; Agrawal, R. Controlled Grain Growth for High Performance Nanoparticle-Based Kesterite Solar Cells. Chem. Mater. 2016, 28 (21), 7703–7714.
(10) Liu, X.; Zhou, F.; Song, N.; Huang, J.; Yan, C.; Liu, F.; Sun, K.; Stride, J. A.; Hao, X.; Green, M. Exploring the Application of Quaternary Metastable Wurtzite Nanocrystals in Pure-Sulfide Cu2ZnSnS4 Solar Cells by Forming Nearly Micron-Sized Large Grains. J. Mater.
Chem. A 2015, 3, 23185–23193.
(11) Ghorpade, U. V; Suryawanshi, M. P.; Shin, S. W.; Hong, C. W.; Kim, I.; Moon, J. H.; Yun, J.
H.; Kim, J. H.; Kolekar, S. S. Wurtzite CZTS Nanocrystals and Phase Evolution to Kesterite Thin Film for Solar Energy Harvesting. Phys.
Chem. Chem. Phys. 2015, 17 (30), 19777–19788.
(12) Yang, W.-C.; Miskin, C. K.; Hages, C. J.; Hanley, E. C.; Handwerker, C.; Stach, E. A.; Agrawal, R. Kesterite Cu2ZnSn(S,Se)4 Absorbers Converted from Metastable, Wurtzite-Derived Cu2ZnSnS4 Nanoparticles. Chem. Mater. 2014, 26 (11), 3530–3534.
(13) Mainz, R.; Singh, A.; Levcenko, S.; Klaus, M.; Genzel, C.; Ryan, K. M.; Unold, T. Phase-Transition-Driven Growth of Compound Semiconductor Crystals from Ordered Metastable Nanorods. Nat. Commun. 2014, 5, 3133.
(14) Coughlan, C.; Ryan, K. M. Complete Study of the Composition and Shape Evolution in the Synthesis of Cu2ZnSnS4 (CZTS) Semiconductor Nanocrystals. CrystEngComm 2015, 17 (36), 6914–6922.
(16) Liu, P.; Singh, S.; Bree, G.; Ryan, K. M. Complete Assembly of Cu2ZnSnS4 (CZTS) Nanorods at Substrate Interfaces Using a Combination of Self and Directed Organisation. Chem. Commun. 2016, 1 (i), 2–5.
(17) Yunxia, Q.; Qiangchun, L.; Kaibin, T.; Zhenghua, L.; Zhibiao, R.; Xianming, L. Synthesis and Characterization of Nanostructured Wurtzite CuInS 2: A New Cation Disordered Polymorph of CuInS 2. J.
Phys. Chem. C 2009, 113 (10), 3939–3944.
(18) Singh, A.; Geaney, H.; Laffir, F.; Ryan, K. M. Colloidal Synthesis of Wurtzite Cu2ZnSnS4 Nanorods and Their Perpendicular Assembly. J. Am. Chem. Soc. 2012, 134 (6), 2910–2913.
(19) Singh, A.; Coughlan, C.; Laffir, F.; Ryan, K. M. Assembly of CuIn1‑xGaxS2 Nanorods into Highly Ordered 2D and 3D Superstructures. ACS Nano 2012, 6 (8), 6977–6983.
(20) Qu, Y.; Zoppi, G.; Beattie, N. S. Selenization Kinetics in Cu2ZnSn(S,Se)4 Solar Cells Prepared from Nanoparticle Inks. Sol. Energy
Mater. Sol. Cells 2015, 1–8.
(21) Collord, A. D.; Hillhouse, H. W. The Effect of Nanocrystal Reaction Time on Cu2ZnSn(S,Se)4 Solar Cells from Nanocrystal Inks. Sol. Energy
Mater. Sol. Cells 2015, 141, 383–390.
(22) Zhou, H.; Hsu, W.-C.; Duan, H.-S.; Bob, B.; Yang, W.; Song, T.-B.; Hsu, C.-J.; Yang, Y. CZTS Nanocrystals: A Promising Approach for next Generation Thin Film Photovoltaics. Energy
Environ. Sci. 2013, 6 (10), 2822–2838.
(23) Chernomordik, B. D.; Béland, A. E.; Deng, D. D.; Francis, L. F.; Aydil, E. S. Microstructure Evolution and Crystal Growth in Cu2ZnSnS 4 Thin Films Formed by Annealing Colloidal Nanocrystal Coatings. Chem. Mater. 2014, 26 (10), 3191–3201.
(24) Chernomordik, B. D.; Ketkar, P. M.; Hunter, A. K.; Béland, A. E.; Deng, D. D.; Aydil, E. S. Microstructure Evolution during Selenization of Cu2ZnSnS4 Colloidal Nanocrystal Coatings. Chem. Mater. 2016, 28 (5), 1266–1276.
(25) Martin, T. R.; Katahara, J. K.; Bucherl, C. N.; Krueger, B. W.; Hillhouse, H. W.; Luscombe, C. K. Nanoparticle Ligands and Pyrolized Graphitic Carbon in CZTSSe Photovoltaic Devices. Chem. Mater. 2016, 28 (1), 135–145.
(26) Clark, J. A.; Uhl, A. R.; Martin, T. R.; Hillhouse, H. W.; Clark, J. A.; Uhl, A. R.; Martin, T. R.; Hillhouse, H. W. Evolution of Morphology and Composition during Annealing Evolution of Morphology and Composition during Annealing and Selenization in Solution-Processed Cu2ZnSn(S,Se)4. Chem. Mater. 2017.
(27) Wu, W.; Cao, Y.; Caspar, J. V; Guo, Q.; Johnson, L. K.; Malajovich, I.; Rosenfeld, H. D.; Choudhury, K. R. Studies of the Fine-Grain Sub-Layer in the Printed CZTSSe Photovoltaic Devices. J. Mater. Chem. C 2014, 2 (19), 3777–3781.
(28) Huang, T. J.; Yin, X.; Tang, C.; Qi, G.; Gong, H. Influence of Ligands on the Formation of Kesterite Thin Films for Solar Cells: A Comparative Study. ChemSusChem 2016, 9 (9), 1032–1041.
(29) Huang, T. J.; Yin, X.; Tang, C.; Qi, G.; Gong, H. A Low-Cost, Ligand Exchange-Free Strategy to Synthesize Large-Grained Cu2ZnSnS4 Thin-Films without a Fine-Grain Underlayer from Nanocrystals. J. Mater. Chem. A 2015, 3 (34),
(30) Williams, B. A.; Smeaton, M. A.; Trejo, N. D.; Francis, L. F.; Aydil, E. S. Effect of Nanocrystal Size and Carbon on Grain Growth During Annealing of Copper Zinc Tin Sulfide Nanocrystal Coatings. Chem. Mater. 2017, 29 (4), 1676–1683.
(31) Lee, Y. S.; Gershon, T.; Gunawan, O.; Todorov, T. K.; Gokmen, T.; Virgus, Y.; Guha, S. Cu2ZnSnSe4 Thin-Film Solar Cells by Thermal Co-Evaporation with 11.6% Efficiency and Improved Minority Carrier Diffusion Length. Adv. Energy Mater. 2015, 5 (7), 2–5.
(32) Kim, Y.; Woo, K.; Kim, I.; Cho, Y. S.; Jeong, S.; Moon, J. Highly Concentrated Synthesis of Copper-Zinc-Tin-Sulfide Nanocrystals with Easily Decomposable Capping Molecules for Printed Photovoltaic Applications. Nanoscale 2013, 5 (21), 10183–10188.
(34) Katagiri, H.; Jimbo, K.; Maw, W. S.; Oishi, K.; Yamazaki, M.; Araki, H.; Takeuchi, A. Development of CZTS-Based Thin Film Solar Cells. Thin Solid Films 2009, 517 (7), 2455–2460.
(35) Zhang, H.; Hu, B.; Sun, L.; Hovden, R.; Wise, F. W.; Muller, D. A.; Robinson, R. D. Surfactant Ligand Removal and Rational Fabrication of Inorganically Connected Quantum Dots. Nano
Lett. 2011, 11 (12), 5356–5361.
(36) Xu, P.; Chen, S.; Huang, B.; Xiang, H. J.; Gong, X.; Wei, S. Stability and Electronic Structure of Cu2ZnSnS4 Surfaces: First-Principles Study. 2013, 45427, 1–8.
(37) Brandl, M.; Sayed, M. H.; Chory, C.; Hammer-Riedel, I.; Parisi, J.; Hock, R.; Gütay, L. In-Situ XRD Investigation of Selenization of CZTS Nanoparticles. J. Alloys Compd. 2017, 714, 35–38.
(38) Shin, S. W.; Han, H.; Park, C.; Agawane, G. L.; Jeong, H.; Yun, H.; Moholkar, A. V; Lee, Y.; Hyeok, J. A Facile and Low-Cost Synthesis of Promising Absorber Materials on Cu2ZnSn(Sx,Se1- x)4 Nanocrystals Consisting of Earth Abundant Elements with Tunable Band Gap Characteristics. J. Mater. 2012, 22, 21727–21732.
(39) S, C. Z.; Cells, S. T. S.; Scragg, J. J.; Wa, J. T.; Edo, M.; Ericson, T.; Kubart, T.; Platzer-bjo, C. A Detrimental Reaction at the Molybdenum Back Contact in Cu2ZnSn(S,Se)4 Thin-Film Solar Cells. J. Am. Chem. Soc 2012, 134, 19330−19333.
(40) Xiao, Z.-Y.; Yao, B.; Li, Y.-F.; Ding, Z.-H.; Gao, Z.-M.; Zhao, H.-F.; Zhang, L.-G.; Zhang, Z.-Z.; Sui, Y.-R.; Wang, G. Influencing Mechanism of the Selenization Temperature and Time on the Power Conversion Efficiency of Cu2ZnSn(S,Se)4 -Based Solar Cells. ACS Appl. Mater. Interfaces 2016, 8 (27), 17334–17342.
(41) Shin, B.; Bojarczuk, N. A.; Guha, S. On the Kinetics of MoSe 2 Interfacial Layer Formation in Chalcogen-Based Thin Film Solar Cells with a Molybdenum Back Contact. Appl. Phys. Lett. 2013, 102, 91907.
(42) Shin, B.; Zhu, Y.; Bojarczuk, N. A.; Chey, S. J.; Guha, S.; Shin, B.; Zhu, Y.; Bojarczuk, N. A.; Chey, S. J.; Guha, S. Control of an Interfacial MoSe2 Layer in Cu2ZnSnSe4 Thin Film Solar Cells: 8.9% Power Conversion Efficiency with a TiN Diffusion Barrier. Appl. Phys. Lett. 2012, 53903 (2012).
(43) Dale, P. J.; Hoenes, K.; Scragg, J.; Siebentritt, S. A Review of the Challenges Facing Kesterite Based Thin Film Solar Cells. 2009 34th IEEE
Photovolt. Spec. Conf. 2009, 2080–2085.
(44) Mainz, R.; Walker, B. C.; Schmidt, S. S.; Zander, O.; Weber, A.; Rodriguez-Alvarez, H.; Just, J.; Klaus, M.; Agrawal, R.; Unold, T. Real-Time Observation of Cu2ZnSn(S,Se)4 Solar Cell Absorber Layer Formation from Nanoparticle Precursors. Phys. Chem. Chem. Phys. 2013, 15 (41), 18281–18289.
(45) Street, T. Raman Spectroscopy of Graphite. Phil. Trans. R. Soc. Lond. 2004, 362 (1824), 2271–2288.
(46) Ferrari, A. C.; Robertson, J. Interpretation of Raman Spectra of Disordered and Amorphous Carbon. Phys. Beview B 2000, 61 (20), 95–107.
