Environmentally friendly preparation of nanoparticles for organic photovoltaics

Aqueous nanoparticle dispersions were prepared from a conjugated polymer poly(2,5-thiophene-alt-4,9-bis(2-hexyldecyl)-4,9-dihydrodithieno[3,2-c:3',2'-h][1,5]naphthyridine- 5,10-dione) (PTNT) and fullerene blend utilizing chloroform as well as a non-chlorinated and environmentally benign solvent, o -xylene, as the miniemulsion dispersed phase solvent. The nanoparticles (NPs) in the solid-state film were found to coalesce and offered a smooth surface topography upon thermal annealing. Organic photovoltaics (OPVs) with photoactive layer processed from the nanoparticle dispersions (PCE) of 1.04%, which increased to 1.65% for devices utilizing NPs prepared from o -xylene. Physical, thermal and optical properties of NPs prepared using both chloroform and o -xylene were systematically studied using dynamic mechanical thermal analysis (DMTA) and photoluminescence (PL) spectroscopy and correlated to their photovoltaic properties. The PL results indicate different morphology of NPs in the solid state were achieved by varying miniemulsion dispersed phase solvent.


M A N U S C R I P T
A C C E P T E D ACCEPTED MANUSCRIPT 3 In recent years a number of publications have focussed on developing alternative OPV fabrication methods, which are scalable at low cost such as roll-to-roll printing. 8,[12][13][14] Although remarkable success has been achieved in printing OPVs on flexible substrates, most of the best performing materials used for printing are still processed from chlorinated solvents such as ortho-dichlorobenzene (o-DCB). 15 The large-scale use of chlorinated solvents is harmful to human health and has a detrimental impact on the environment. [16][17] Moreover, the usage of chlorinated solvents increases the cost of large-scale fabrication of OPVs, which results from expensive halogenated solvent recovery systems. The harmfulness and high cost of chlorinated solvents used for processing photoactive materials in OPV fabrication is one of the main hurdles to be overcome before the knowhow of fabricating high performing OPVs can be transferred from a lab-scale to an industrial scale fabrication. Thus, it is of utmost importance to develop green deposition methods by utilizing benign and nonchlorinated solvents. [18][19][20] OPVs with active layers processed from a water or alcohol based nanoparticle dispersion has been reported in recent years. [21][22][23][24] These nanoparticles are processed from donor-acceptor blends either through a miniemulsion process with the presence of surfactant [25][26][27][28] or a precipitation method [21][22][23]29 . Furthermore, conjugated polymer nanoparticles were also reported to be synthesized by direct Suzuki-Miyaura dispersion polymerization. [30][31][32] In the miniemulsion method, the organic solvent utilized to dissolve the active materials should ideally have high vapour pressure as well as be immiscible with water. Most procedures reported in the literature to date use harmful chlorinated solvents, such as chloroform (CHCl 3 ) 21, 33 , chlorobenzene 34 or o-DCB 26 . Compared to conventional bulk heterojunction (BHJ) OPV fabrication, the NP method is still environmentally superior considering (a) the volume of chlorinated solvents utilized is comparably less; (b) the roll-to-roll printing of solar cells is free from chlorinated solvents as they can be removed in a closed loop system prior to M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 4 printing. Considering that the upscaling of OPVs with NP active layers will lead to an increase in the consumption of solvents used to prepare NPs, it is timely to consider the use of industrially relevant solvents during the preparation of the NPs of the photoactive materials in addition to the subsequent deposition.
In this paper, we report for the first time the preparation of water-dispersed nanoparticles using a relatively benign and industrially relevant solvent, o-xylene, as the miniemulsion dispersed phase solvent, and successfully demonstrate the fabrication of solar cells with comparable device parameters to BHJ OPVs. 19 The nanoparticles were prepared from a wide bandgap semicrystalline conjugated polymer namely poly(2,5-thiophene-alt-5,10bis(octyloxy)dithieno[3,2-c:3',2'-h] [1,5]naphthyridine-5,10-dione) (PTNT) 35 and PC 71 BM (phenyl C 71 butyric acid methyl ester). PTNT polymer was chosen in this study as it was demonstrated to perform well in an active layer thickness of up to 400 nm 35 , which makes it a relevant polymer for devices fabricated via printing. To gain a better understanding of the influence of the miniemulsion dispersed phase solvent on the nanoparticle properties,

Nanoparticle preparation
PC 71 BM was purchased from Solenne BV. PTNT:PC 71 BM nanoparticles were prepared with the weight ratio of 1:2 through the miniemulsion method. 33,37 The weight ratio was chosen based on the best performance of BHJ devices from a PTNT:PC 71 BM blend without solvent additive. 35 10 mg of PTNT (10 mg) and 20 mg of PC 71 BM (20 mg) were dissolved in 540 µL of organic solvent (chloroform or o-xylene) at 35 °C with stirring at 500 rpm for 2 hours.
Meanwhile the aqueous phase was prepared by dissolving 33 mg of sodium dodecyl sulphate (SDS) (33 mg) in 2.8 mL of MilliQ water. After ensuring complete dissolution of PTNT and PC 71 BM, the aqueous phase was combined with the organic phase under stirring at 1200 rpm.
A macroemulsion was then formed by stirring the mixture at 1200 rpm at 30 °C for approximately 1 hr. To generate the miniemulsion, Vibra-Cell ultrasonic processor VCX 750 with 1/8" stepped probe was introduced to ultra-sonicate the macroemulsion at 30% amplitude for 3 min. Then the miniemulsion was transferred immediately to a heating block and stirred at 1200 rpm to form a stable water dispersion of NPs after complete removal of the residual organic solvent. For nanoparticles prepared using chloroform, the miniemulsion was heated at 60 °C for 3 hr to ensure the complete removal of chloroform whereas in the case of using o-xylene, a NPs dispersion was achieved by heating at 75 °C for 6 hr.
Additional water was added every hour to compensate for the water loss, which could otherwise results in aggregation of material and precipitation on the wall of the vials. To minimise unwanted SDS from negatively impacting the device performance 27 , centrifugal dialysis was introduced to remove excess free surfactant in the dispersion as well as concentrate the active materials in the water dispersion, giving the final dispersion a solids content of 6 wt% in 0.5 mL water.

Nanoparticle characterization
Scanning electron microscopy (SEM) was performed using an ultra-high resolution fieldemission gun scanning electron microscope (Zeiss Merlin) at an accelerating voltage of 2 kV with magnification ranges of 50,000-150,000 X. All SEM samples were spin-coated (3000 rpm for 1 minute) from diluted nanoparticle water dispersion with 1wt% of solids content on conductive silicon substrate. The size distribution of PTNT:PC 71 BM (1:2 weight ratio) NPs prepared from different organic solvents were characterized from SEM images with a circular Hough transform algorithm. [37][38] In the varied annealing temperature study, all films were predried at 90 °C for 4 min immediately after spin-coating for consistency with device fabrication.
The ultraviolet-visible (UV-vis) study was performed on a Perkin Elmer UV-vis-NIR Lambda950 spectrophotometer. The photoluminescence (PL) measurements were performed on a Varian Cary Eclipse fluorescence spectrophotometer at the excitation wavelength of 450 nm. The PL measurements of NP water dispersions were performed on diluted NP-xylene and NP-chloroform dispersion with the same concentration. Measurements of the solid state were performed on spin coated NP films from original NP dispersion. It should be noted that the study of absorbance change of NP film with and without annealing was measured on the same area on the same film for comparison.
Dynamic mechanical thermal analysis (DMTA) samples were prepared by repeatedly dropcasting the PTNT:PC 71 BM NP water dispersion onto a 20-30 mm long, approximately 5 mm wide piece of glass fiber mesh, as described elsewhere 39 , followed by drying under ambient condition until a uniformly fully covered film on the glass mesh was obtained. The samples were stored in a desiccator overnight to remove most of the water before performing DMTA measurements. DMTA measurements were carried out on a TA Q800 DMA in strain- controlled mode at a frequency of 1 Hz and an amplitude of 5 µm. The samples were measured under a continuous flow of nitrogen gas with a heating rate of 3 °C per minute. The first run was performed from room temperature up to 80 °C for further drying the sample, followed by a second run from -110 °C to 200 °C. The data from the second run was utilized to study thermomechanical properties of materials in this study.

Device fabrication
Inverted solar cells with the structure ITO/ZnO/NPs/MoO 3 /Ag were fabricated using water dispersed NPs to coat the active layer. Patterned ITO-coated glass substrates (10 Ω/sq, purchased from Xin Yan Technology Ltd) were cleaned using the procedure described elsewhere. 36 ITO-coated glass substrates were first cleaned by soaking in a 5% detergent solution (pyroneg from Johnson Diversey) at 90 °C for 20 minute and then rinsing in deionized (DI) water, before sonicating in DI water, acetone and isopropanol for 10 min each.
Substrates were then cleaned in UV-ozone for 20 minute immediately before spin coating the ZnO layer. ZnO sol-gel 40

Property of nanoparticles
PTNT:PC 71 BM (1:2 weight ratio) NPs were prepared using the miniemulsion method 24 with chloroform or o-xylene as a miniemulsion dispersed phase solvent. To achieve stable aqueous NP dispersions, chloroform (utilized to form the miniemulsion) was completely removed after heating at 60 °C for 3 hr and o-xylene was evaporated by heating at 75 °C for 6 hr, respectively. To evaluate the size of prepared NPs, SEM measurement was performed on NPs in the solid state. SEM images presented in Fig. 2a and b depict the PTNT:PC 71 BM nanoparticulate films for NPs spun prepared from chloroform and o-xylene, respectively. NPs were as spun (no thermal treatment) from water-based colloidal dispersions without further treatment. The particulate shape of PTNT:PC 71 BM NPs was found not to be completely spherical, which is typical for NPs made of semicrystalline polymer. 25,27,[41][42][43] The size distribution of NPs (Fig. 2c) was measured by applying a circular Hough transform algorithm to the SEM images of NP films. Since the circular Hough transform algorithm is based on circle calculation, the model resulted in several mismatches (see Fig. S2). Nevertheless the mean diameter of PTNT:PC 71 BM NPs prepared using chloroform (NP-chloroform) was calculated to be 32 ± 12 nm whereas the nanoparticles from the o-xylene batch (NP-xylene) had a mean diameter of 27 ± 11 nm in the solid state. Compared to NP-chloroform, NPxylene resulted in a slightly narrower size distribution with the presence of fewer large sized NPs on the surface. This observation indicates that the higher temperature and the longer time required to convert the NP-xylene miniemulsion to an aqueous dispersion does not result in material aggregation or further increase in the particle size. To achieve well performing solar devices using water dispersed NPs, it is imperative to achieve good morphology of the active layer. [44][45] Due to the particulate shape and the coreshell structure of NPs prepared through miniemulsion method 25,37 , the coalescing of NPs could ideally tune the morphology of the active layer and is thus desirable for better charge transport and extraction. 21,46 To coalesce NPs without inducing defects in the film, thermalannealing was introduced, which is widely applied to both NP and BHJ solar devices to improve the morphology of the active layer, therefore enhancing the performance of devices. [46][47][48] The schematic shown in Fig. 3 depicts coalescence of NPs upon annealing in the solid state. SEM and atomic force microscopy (AFM) (see SI varied annealing temperature study) was thus used to systematically study the changes induced in the surface topography of the NP films as a function of annealing temperatures, and to find the ideal annealing temperature which resulted in best performing devices. The as spun film prepared from NP-chloroform (Fig. 2a) shows that the separate NPs are clearly distinguishable. The near-edge X-ray absorption fine structure (NEXAFS) spectroscopy measurements (Fig. S3) and scanning transmission X-ray microscopy (STXM) results (Fig. S4) reveal the core-shell structure of PTNT:PC 71 BM NPs without thermal annealing. Thermal annealing of NP-chloroform films at 100 °C (see Fig. S5a) or 120 °C (Fig.   4a) was not found to make any significant changes on the surface features of the films, as a large degree of nanoparticulate structure was still observed by SEM. The particles were found to interconnect with each other (sinter) after the NP-chloroform film was annealed at 140 °C M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 12 (Fig.4b), with some residual nanoparticles presented on the surface. When thermally annealed at 160 °C (Fig. 4c), NPs were sintered and the relatively homogenous film was obtained, which could improve the charge transport 43 and reduce the possibility of charge recombination in the coalesced NP active layer in OPVs. 21 Upon annealing at 180 °C, large aggregates in the NP-chloroform film were observed (Fig. 4d), which was attributed to the crystallization of PC 71 BM. 49 The gross phase separation deteriorated the NP film morphology, which is detrimental for the device performance of NP solar devices. 33,43,50 Thus, precise control over the thermal annealing temperature is crucial to achieve a homogenous active layer in OPVs as well as removing the moisture and avoiding large phase separation. 37 Similar behaviour was also found in the case of NP-xylene films, when annealed at different temperatures ( Fig. 4e-  To further probe the morphological changes and thermomechanical behaviour of PTNT:PC 71 BM NP films, DMTA measurements 39 were performed on thin films cast from NP-xylene (Fig. 5) and NP-chloroform (see Fig. S6a) as well as pure PTNT (Fig. S6b) and NP-chloroform as well as pure PTNT (Fig. 5). Since DMTA measures the thermal properties of materials with high sensitivity 39, 52-53 , it further compliments the SEM study which only probes the surface topography. Fig. 5a shows the storage modulus (E') in the DMTA scan of the NP-xylene sample. The storage modulus (E') was found to drop around 100 °C with a significant loss at 160 ˚C, indicating the softening of the amorphous part of the PTNT-rich phase. 54 Based on the STXM results (Fig. S4), it is known that the nanoparticles have a coreshell structure with a PTNT-rich NP shell and a PC 71 BM-rich NP core. As such, the nanoparticles should start to coalesce only when PTNT in the polymer-rich shell reaches the rubbery state 55 similarly to P3HT:PCBM NPs 25 . Thus, the significant drop in the E' observed using DMTA agrees well with the coalescence of the nanoparticles in thin films observed using SEM. Beyond 160 ˚C, the stiffness (E') of the DMTA sample was again found to increase, which is attributed to the crystallization of PC 71 BM in the PC 71 BM-rich phase after the coalescence of the NPs, revealing the maximum temperature before detrimental crystallization occurs. This behaviour was also observed in SEM as drastic phase separation and coarsening of the film (Fig. 4h). The DMTA temperature scan of the NPs prepared using chloroform can be found in the Supplementary Information Fig. S6a and they show a similar thermomechanical behaviour, whereas the PTNT DMTA scan (Fig. S6b) doesn't show the increase of E' above 160 °C. The DMTA temperature scan of the NPs prepared using chloroform (Fig. 5b) shows a similar thermomechanical behaviour, whereas the neat PTNT DMTA scan (Fig. 5c)

3.1 Optical properties of nanoparticles
To probe the optical properties of NPs prepared using different solvents, UV-vis and photoluminescence (PL) measurements were carried out on NP films as well as NP dispersions. Fig. 6 shows the UV-vis absorption spectra of solid state films and PL spectra backbones. 35 The absorption peak below 400 nm is primarily attributed to the absorbance of PC 71 BM, while the vibronic peaks at 522 nm in the absorption spectra are attributed to the ππ stacking of polymer backbone 35 (Fig. S6a). It can be observed that after thermal annealing the absorbance of PTNT:PC 71 BM NP film was increased (Fig. 6a,b). The increased absorption upon annealing could be due to the increased crystallinity and enhanced ordering of PTNT in the annealed films 22,57 , which was also observed in annealed PTNT:PC 71 BM BHJ films (presented in Fig. S6b and c). The thermal annealing at 160 °C for 4 min did not lead to blue-shifted onset or decrease of the light absorbance of the PTNT:PC 71 BM NP films, revealing no thermal degradation of materials in the NP film during short-term thermal annealing.
The PL spectra shown in Fig. 6c compare the quenching of PL signal in NP-chloroform and NP-xylene water dispersion. A higher PL intensity is observed for the NP-xylene dispersion ( Fig. 6c), which was attributed to a higher degree of donor-acceptor material phase separation

3.2 Photovoltaic characterizations
In order to test the photovoltaic properties of NPs, solar cells were fabricated and tested with the active layer spin-coated from NP dispersion (presented in Fig. S7) Table 1. NP-chloroform films after thermal annealing at 160 °C were found to result in solar cells with a maximum PCE of 0.76%, a J SC of 2.60 mA/cm 2 and fill factor (FF) of 32%. The PCE of these devices was further increased up to 1.04% accompanied by slight increase in J SC to 2.84 mA/cm 2 and a significant increase in FF to 43% upon post-annealing the complete devices at 160 °C. Thermal treatment after the deposition of electrode is known to enhance the performance of NP based solar cells. 37,61 Although post-annealing at 160 °C resulted in a slightly higher PCE of 1.04% as compared to the PCE of 0.98% when post-annealed at 140 °C, the standard deviation in the device characteristics in the case of devices post-annealed at 160 °C was higher than that at 140 °C.
The increase in the standard deviation could be a result of the uneven distribution of gross phase separation happening during post-annealing at high temperature. The representative J-V curves of NP-chloroform based devices with and without post-annealing are shown in Fig.   7a. It must be noted that the open circuit voltage (V OC ) of these devices was found to be as high as 0.90 V, which is comparable to their BHJ counterparts processed from organic solvents (see Table S3). [35][36] This suggests that the performance of PTNT:PC 71 BM NP solar cells is mainly limited by the lower current and lower FF, both of which can be highly dependent on the NP film morphology and the nature of the donor/acceptor interface. 27 Furthermore, compared to solution-processed BHJ solar cells 35 , the lower performance of NP devices could be explained by the relatively isolated polymer and PC 71 BM domains with insufficient connection of fullerene-rich core, which impedes the charge separation 62 (Fig. 7c). This could possibly be explained due to better exciton dissociation or reduced charge recombination resulted from the phase separation between polymer and PC 71 BM. 64 When a higher post-annealing temperature (i.e. 160 °C) was applied to the device, a decline in the mean value of J SC was observed as well as V OC compared to the device performance without post-annealing.
However, an improvement in FF compensated the decrease in J SC and V OC , and resulted in slightly improved efficiency, which is 1.33% on average, compared to the values in device without post-annealing. The EQE of NP-xylene based device post-annealed at 160 °C showed the contribution of photon harvesting property in PC 71 BM region (below 400 nm) was increased compared to the contribution from PTNT region (above 400 nm), which can be revealed from the decline of EQE between 450 nm to 600 nm photon wavelength (Fig. 7c).
Considering 160 °C is approaching the crystallisation temperature of PC 71 BM 49 , the increase in EQE in the wavelength below 400 nm could be attributed to the crystallization appearing in the pure PC 71 BM domains, which also agrees with the storage modulus increasing beyond M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 19 Either with or without post-annealing, the NP solar devices fabricated from o-xylene batch PTNT:PC 71 BM (1:2 weight ratio) NPs achieved higher J SC and efficiency compared to the NP devices produced from NP-chloroform. However, there is no significant difference in FF and V OC between different batches of NPs as long as they were thermally treated in the same way. The increase in J SC is attributed to a finer intermixed nanomorphology in annealed NPxylene film, which is also suggested by the PL of annealed NP films. The difference in the morphology of NP active layer indicates that post-preparation of NPs, the internal morphology of NPs was changed by altering the miniemulsion dispersed phase solvent from chloroform to o-xylene. The difference of the internal morphology between NP-xylene and NP-chloroform was revealed from the varied PL intensity observed in different batches of NP dispersions.
BHJ solar cells were also fabricated from PTNT:PC 71 BM (1:2 weight ratio) blends using chloroform and o-xylene as the processing solvents, with the same device geometry. The BHJ devices gave highest PCE of 1.20% with chloroform and 0.84% using o-xylene (see Table   S3), due to the unfavourable film quality induced by the processing solvents. The NP-OPVs based on aqueous NP-xylene achieved comparable photovoltaic performance without using any harmful solvent during the device fabrication, though the efficiency of PTNT:PC 71 BM BHJ devices could be improved by altering the processing solvent and using additives [35][36] .

Conclusions
Water Using o-xylene leads to higher photovoltaic performance compared to using chloroform.
DMTA probes morphological change of NP film as a function of annealing temperature.
Thermal annealing to the complete device improves the device performance.