Unravelling donor-acceptor film morphology formation for environmentally-friendly OPV ink formulations

The challenge of coating organic photovoltaics (OPV) from green solvents is to achieve the required nanostructured interpenetrating network of donor and acceptor domains based on a rational choice of solvent approach as opposed to the usual trial - and - error methods. We demonstrate here that we can achieve a bicontinuous interpenetrating network with nanoscale phase separation for the chosen donor – acceptor material system poly[2,3-bis-(3-octyloxyphenyl)quinoxaline-5,8-diyl- alt -thiophene-2,5-diyl]: phenyl-C 61 butyric acid methyl ester (TQ1:PC 61 BM) when processing from green solvent ink formulations. This structure is achieved by ﬁ rst calculating the Hansen solubility parameters (HSP) of the donor and acceptor materials, followed by careful choice of solvents with selective relative solubilities for the two materials based on the desired order of precipitation necessary for forming a nanostructured interdigitated network morphology. We found that the relative distances in Hansen space ( R a ) between TQ1 and the primary solvent, on the one hand, and PC 61 BM and the primary solvent, on the other hand, could be correlated to the donor – acceptor morphology for the formulations based on the solvents d-limonene, anisole, and 2-methyl anisole, as well as the halogenated reference solvent o -dichlorobenzene. This nanostructured blend ﬁ lm morphology was characterised with scanning transmission X-ray microscopy (STXM) and transmission electron microscopy (TEM), and the ﬁ lm surface composition was analysed by near edge X-ray absorption ﬁ ne structure (NEXAFS) spectroscopy. Hansen solubility theory, based on solution thermodynamics, has been used and we propose an HSP-based method that is a general plat-form for the rational design of ink formulations for solution-based organic electronics, in particular facili-tating the green solvent transition of organic photovoltaics. Our results show that the bulk heterojunction morphology for a donor – acceptor system processed from customised solvent mixtures can be predicted by the HSP-based method with good reliability.


Introduction
Printed organic electronics is an expanding field with multiple benefits: solution processability of the organic semiconductors; flexibility, light-weight and semi-transparency of the printed devices; and the ability to utilise conventional printing techniques to manufacture devices without the need for expensive infrastructure, making the technology low-cost and globally achievable (Fig. 1). 1 High throughput manufacturing involving high coating speeds (>1 m min −1 ) is critical for an economically feasible industrial scale-up of printed organic electronics, since cost is an important parameter to consider for industrial production. 1,2 The commercialisation of polymer solar cell technology will depend on three key factors: device efficiency, lifetime and cost. 3 It has been demonstrated that photoactive layers can be routinely slot-die roll-to-roll (R2R) coated at a speed of 1.5 m min −1 and Ag grid electrodes are flexographically R2R printed at 20 m min −1 . 4 Transitioning to eco-friendly printing inks for the largescale manufacture of printed organic electronics is a necessary step, as the use of halogenated solvents is environmentally hazardous, harmful to the health of workers, and will increase the cost of production. 5 This transition will require significant effort considering the highest performing organic photovoltaic (OPV) devices are at present processed using halogenated solvents. [6][7][8] Green solvents are environmentally friendly solvents which pose low environmental impact in both production and use, and can include bio-solvents. 9 The transition to green solvent processing and green synthetic techniques has been reported in various fields, including catalyst development, 10,11 synthesis of inorganic semiconductors for solar cell applications, 12 medicinal chemistry, 13,14 and analytical chemistry. 15 Shifting from halogenated solvents (such as chloroform, chlorobenzene and o-dichlorobenzene) to green solvents requires substantial film microstructure optimisation, 16,17 due to the vast effect that the casting solvent has on self-assembly of the complex donor-acceptor film morphology for these systems. 18,19 Factors determining donoracceptor film morphology formation include the vapour pressure of the solvent and additive, relative solubilities of the donor and acceptor molecules, 20,21 and miscibility of the donor and acceptor material. 22,23 To achieve efficient OPV device performance, it is necessary to control the degree of phase separation between the donor and acceptor components. 21,24 The size scale of donor and acceptor domains and the degree of donor-acceptor phase intermixing ( phase purity) have a direct impact on the exciton dissociation efficiency (η ED ), 25 charge carrier mobility, and recombination rates in these systems, 26 all of which directly influence the maximum achievable power conversion efficiency (PCE) of solar cells. It is common to utilise small fractions of co-solvents (additives) to avoid the formation of undesired large circular fullerene domains, during solutionprocessing of OPV, that reduce device performance. 21,27,28 Hansen solubility parameters (HSP) have emerged as a practical semi-empirical method of classifying a range of molecule types ( polymers, macromolecules, solvents, additives) and predicting the solubility of macromolecules. [29][30][31][32][33][34] HSP has also recently been shown to be a useful tool in identifying molecular phase stabilisers for overcoming microstructure instabilities in high performance PCE11:PC 71 BM OPV. 35 Studies of OPVs fabricated from green solvents have slowly emerged in the past decade, [36][37][38] but considerable research effort is still required to shift from halogenated solvent to green solvent printing and, in particular, through a rational choice rather than the more usual trial-and-error approach of solvent selection. The application of the green solvent system anisole:diphenyl ether to PBQ-4:PC 71 BM bulk heterojunction (BHJ) OPV achieved an impressive PCE of 8.37% in 2015. 17 Zhang et al. then reported PBDT-TS1:PC 71 BM-based OPV devices processed from green solvent 2-methylanisole, achieving a PCE of 9.67%. 39 PCEs exceeding 10% were achieved for donor-acceptor systems PffBT4T-C 9 C 13 :PC 71 BM 19 and FTAZ: IT-M 40 by coating from non-halogenated solvents 1,2,4-trimethylbenzene:1-phenylnaphthalene and toluene, respectively. Duong et al. explored the possibility of correlating HSP and BHJ film morphology; however, concluded that for their dataset the HSP values alone did not play a critical role in determining the morphology of the organic semiconductor BHJ films. 32 Machui et al. determined the HSP of poly(3-hexyl thiophene) (P3HT) and PC 61 BM and varied the composition of solvent blends in order to effectively demonstrate the processing limits of solvent systems for OPV fabrication. 30 An HSP study of the small molecule donor-acceptor system D-π-A tris{4-[5″-(1,1-dicyanobut-1-en-2-yl)-2,2′-bithiophen-5-yl]phenyl} amine N(Ph-2T-DCN-Et) 3 and PC 71 BM yielded promising green solvent alternatives for the fabrication of OPV devices, with the OPV devices yielding comparable efficiencies to those obtained with halogenated solvents. 31 In this study we demonstrate that the relative solubilities of the donor and acceptor material in the primary solvent are of prime importance for the TQ1: PC 61 BM BHJ film morphology. We utilise STXM and TEM to assess lateral film morphology, as well as NEXAFS to investigate the vertical film composition in spin coated BHJ films for the series of selected green solvent ink formulations. We demonstrate the formation of a finely mixed donor-acceptor blend film morphology that resembles the morphology of films coated from the halogenated solvent o-dichlorobenzene, a solvent that gives the highest PCE for this material system. Our study effectively presents an approach which uses both thermodynamics and kinetics to predict and quantify morphology.

HSP calculations
HSP calculations for TQ1 and PC 61 BM were performed using the software package Hansen Solubility Parameters in Practice (HSPiP). 44 A set of 36 solvents was used for both TQ1 and PC 61 BM, listed in ESI (Table S1 †). The concentration 1 mg ml −1 was chosen as the solution concentration threshold when determining the solubility of TQ1 and PC 61 BM, in order to generate input data to the HSP software. Solubility was assigned following 24 h dissolution at 50°C and ambient pressure. Visual inspection of the solutions led to the assignment in the software of a score of '1' when the solvent dissolved the material and a score of '0' when the solvent did not dissolve the material.

Ink formulation
The D : A ratio (TQ1:PC 61 Table 1 for each ink formulation. These inks were utilised to spin coat BHJ films for STXM, TEM and NEXAFS analysis, with spin coating and film drying conditions provided in ESI (Table S2 †). We note that mild thermal drying conditions were chosen carefully in order to avoid exposing the films to conditions that would lead to morphological changes attributed to thermal annealing. The purpose of the thermal drying conditions was solely to accelerate the removal of remaining solvent and additive from the films following spin coating, because if unremoved the remaining solvents and additives can lead to a continually evolving donor-acceptor morphology while the film is in a 'wet' semi-solvated state.

NEXAFS spectroscopy
Near-edge X-ray absorption fine-structure (NEXAFS) spectra were collected at the Australian Synchrotron on the SXR beamline in partial electron yield (PEY), total electron yield (TEY) and fluorescence yield (FY) modes. 45 All NEXAFS spectra were collected at the C K-edge in the photon energy range 280-320 eV at a 55°incidence angle with respect to the sample surface. For TEY, the sample drain current was measured, PEY was measured with a channeltron detector, and FY was measured with a multichannel plate detector. Sample spectra were normalised to reference spectra collected on an Au mesh sample. 46 Energy calibration at the C K-edge was performed by measuring the spectrum of highly ordered pyrolytic graphite (HOPG) and using the position of the exciton resonance at 291.65 eV. 47 NEXAFS spectra were analysed using QANT software developed by Gann et al. 48 C K-edge NEXAFS spectra collected in TEY mode are presented in Fig. 5 and 6. The mass fractions of TQ1 and PC 61 BM were determined by fitting the blend film spectra to those of the pristine components, and are listed in Table 5. TEY mode collects signal from the surface 5-10 nm of the film, giving a good representation of sub-surface composition. PEY mode probes the surface 2-3 nm of the film, the area of the film most prone to degradation effects due to air and light exposure, 49 and is hence of less relevance to the current study. FY mode probes to a depth of ∼250 nm, a thickness greater that the BHJ film thickness, hence collecting signal from the underlying hole transport layer (HTL) and substrate and thus of less relevance to the current study. The top surface of TQ1:PC 61 BM BHJ films were measured in order to study the film composition at the BHJ/ETL interface (electron extraction side). For the film top surface measurements, TQ1:PC 61 BM BHJ films, spin coated from green ink formulations directly The samples on Cu grids were mounted in the STXM sample chamber and rastered with respect to the X-ray beam, the STXM sample chamber was backfilled with helium (0.33 atm). The STXM Fresnel zone plate had an outer most zone width of 25 nm, setting the spatial resolution limit of the measurement.
The transmitted X-ray beam is detected by a scintillator and a photomultiplier tube. 51 The energy of the X-ray beam was varied between 270 and 340 eV, spanning the C K-edge region with an energy resolution of 0.1 eV. Energy calibration at the C K-edge was first performed on CO 2 gas. Second-and thirdorder light was removed by an order sorting aperture and also N 2 gas filter as detailed in Kilcoyne et al. 51 Orthogonal energies were determined from NEXAFS spectra of pristine films of TQ1 (285.2 eV) and PC 61 BM (284.4 eV). Singular value decomposition was used to fit a sum of the pristine TQ1 and PC 61 BM NEXAFS spectra to the measured blend spectrumat each pixelin the STXM images. Image analysis was performed with the aXis2000 package. 52 The average domain sizes presented in Table 4 were calculated in the aXis software by sampling up to 25 domains in the scanned area. Average domain size with standard deviation are presented.

Electron microscopy
TEM was performed on a JEOL 1200 EXII at an accelerating voltage of 80 kV at varying magnification ranges. BHJ films on 300 mesh Cu grids (20 μm bar, 63 μm hole, 3 mm diameter, purchased from ProSciTech Pty Ltd) were prepared as detailed in section 2.5.

HSP calculation of organic semiconductors
Hansen solubility parameters are a valuable tool for predicting the solubility of materials such as polymers and organic molecules, and for formulating inks. The organic solute and solvent materials are characterised using three parameters: δ D for dispersion (related to van der Waals interaction), δ P for polarity (related to dipole moment) and δ H for hydrogen bonding. The solubility parameters of TQ1 and PC 61 BM were determined by the conventional Hansen method, that is, performing solubility tests in a broad range of solvents (spanning the three dimensions of Hansen space), with solvents listed in Table S1 (ESI †). The HSP sphere for TQ1 is plotted in Fig. 2a and the HSP sphere for PC 61 BM in Fig. 2b. The blue dots in Fig. 2 represent good solvents for the material, lying within the solubility sphere; the red dots represent non-solvents for the material, lying outside of the solubility sphere (see designation of good solvents and non-solvents in Table S1 †). The HSP of each material are the δ D , δ P , δ H coordinates at the centre of the sphere. The HSP for TQ1 and PC 61 BM are tabulated in Table 2, including the δ D , δ P , δ H coordinates and R 0 , where R 0 is the radius of the solubility sphere in Hansen space. The model assumes that solubilities are spherical in the Hansen space and the calculated sphere is the minimal volume only including "good" solvents. Changing process parameters such as the concentration threshold will change R 0 and sometimes even affect the resulting HSP values. Hence, we have kept the process parameters constant when testing the solubility of the two semiconductors (TQ1 and PC 61 BM), that is, dissolution at 50°C and ambient pressure, dissolution time of 24 h, solution concentration of 1 mg ml −1 . We note that the HSP and R 0 values determined for TQ1 and PC 61 BM in this work are only comparable to other studies if the process parameters are consistently maintained.

Identification of green solvent alternatives
The calculated HSPs of TQ1 and PC 61 BM enabled a range of non-halogenated solvents to be identified for these organic semiconductors. We chose as primary solvents d-limonene, 2-methyl anisole, anisole, and the additives diphenyl ether, MO and MY as suitable candidates with similar HSP to the organic semiconductors (Table 3). d-Limonene is derived from citrus fruits and is often used in household cleaning products, while anisole is a parent molecule to multiple ethers naturally present in essential oils. 2-Methylanisole is a non-halogenated, biodegradable chemical with low toxicity, and is significantly safer than chlorinated solvents for mass production. 39 Further toxicity and biodegradability information, including hazard statements for each solvent and additive, is provided in ESI (Table S3 †). These solvents and additives were then combined to form seven ink formulations (detailed in Table 1).

Thin film morphologylateral and vertical
TQ1:PC 61 BM BHJ thin films were spin coated from each of the eight ink formulations and the lateral film morphology was imaged with STXM ( Fig. 3) and TEM (Fig. 4). Whilst TEM can resolve nanostructured domains at high spatial resolution, it is the chemical sensitivity of NEXAFS at the carbon K-edge and STXM that enable the determination of donor-rich and acceptor-rich domain compositions. 53 The STXM and TEM analysis show a wide variety in film morphologies. Some of the green solvent ink formulations yielded a fine donor-acceptor film morphology, whilst others show a significant degree of lateral phase separation. The film morphology types observed for the TQ1:PC 61 BM BHJ film series were divided into three categories, based on the size scale of the film microstructure. Category 1: micron sized fullerene aggregates; Category 2: ∼ 200 nm fullerene aggregates; Category 3: <100 nm TQ1-rich and PC 61 BM-rich domains forming an interdigitated network. This morphology classification is represented in conceptual schematic form in Fig. 7. The ink formulations Lim and Lim:2-MA fall in Category 1, and the ink formulations Lim:MO and Lim:MY fall in Category 2. The remaining four ink formulations fall in Category 3. All categories will be discussed in further detail here.
Macro-scale (>150 nm) PC 61 BM-rich domains, resembling PC 61 BM aggregates, are observed in TQ1:PC 61 BM BHJ films spin coated from ink formulations Lim (Fig. 3a, b and 4a) Lim:2-MA (Fig. 3c, d and 4b), Lim:MO (Fig. 3e, f and 4c) and Lim:MY (Fig. 3g, h and 4d). These macro-scale domains are confirmed to be PC 61 BM-rich from STXM analysis, and the domain compositions measured from STXM maps are provided in Table 4. In addition, irregularly shaped structures, indicative of residual undissolved polymer, are observed in the   STXM maps of TQ1:PC 61 BM BHJ films spin coated from ink formulations Lim ( Fig. 3a and b) and Lim:MY ( Fig. 3g and h), and in the TEM images of films prepared from the four ink formulations Lim, Lim:2-MA, Lim:MO and Lim:MY (Fig. 4a-d).
For the ink formulation Lim we observe micron-sized fullerene aggregates of average diameter 860 ± 210 nm and secondary fullerene aggregates of diameter 250 ± 80 nm (where ±σ is the standard deviation, representing the variance in domain size) (Table 4). Similarly, for the ink formulation Lim:2-MA we observe micron-sized fullerene aggregates of diameter 870 ± 140 nm and     (7) 70 (10) 60 (20) a Since STXM is a transmission technique, the actual compositions of the fullerene domains (especially the island domains) cannot be measured when there are polymer-rich wetting and capping layers to the film. Hence the measured compositions listed in Table 4 are lower than the actual domain compositions. meter) formation in films of PBDT-TS1:PC 71 BM spin coated from xylene, a poor solvent in the ink formulation series studied. 39 Similarly, Hansson et al. reported 500-800 nm diameter fullerene-rich islands in films of TQ1:PC 71 BM spin coated from chloroform, a poor solvent for PC 71 BM. 20 The morphology of the OPV active layer is key to device performance. The domain size should be similar to the exciton diffusion length (∼10 nm in organic semiconductors 54 ) to maximise exciton dissociation, but maintain bicontinuous percolation pathways for efficient charge extraction. 55 The reference TQ1:PC 61 BM BHJ film spin coated from halogenated solvent o-DCB achieves a nanostructured film (Fig. 3o, p and  4h), with domain sizes <25 nm, below the spatial resolution limit of the STXM technique. We have very closely approached this small domain size with three of the green ink formulations, classified as Category 3, Ani:DE, Ani:DIO and 2-MA: Lim (noting that ink formulation Ani:DIO still contains 1 v/v% of halogenated additive DIO and hence is not completely halogen-free). The STXM and TEM analysis of the TQ1:PC 61 BM BHJ film spin coated from ink formulation Ani:DE shows a finely structured morphology comprising an interdigitated donor-acceptor domain network (Fig. 3i, j and 4e) with an average domain size of 100 ± 20 nm and a donor-rich domain composition of 72 ± 5% TQ1 and acceptor-rich domain composition of 67 ± 4% PC 61 BM (where ±σ is the standard deviation, representing the variance in domain composition). For the ink formulation Ani:DIO (Fig. 3k, l and 4f ) we observe a further reduction in domain size to 80 ± 20 nm, and measured via STXM a donor-rich domain composition of 66 ± 7% TQ1 and acceptor-rich domain composition of 68 ± 5% PC 61 BM. We then achieve the finest morphology within the green ink formulation series for the 2-MA:Lim BHJ film (Fig. 3m, n and 4g), with a feature size of 60 ± 20 nm, and measured donor-rich domain composition of 68 ± 7% TQ1 and acceptor-rich domain composition of 70 ± 10% PC 61 BM.

Green Chemistry Paper
Large domains are detrimental for organic solar cell performance. For Category 1 and Category 2 morphologies, the donor and acceptor domains are too large, and OPV devices would suffer from a low exciton dissociation efficiency. Category 1 and Category 2 morphologies are absent of an interconnected fullerene network required for efficient electron charge extraction due to the fullerene aggregate formation which manifests as circular island domains within the film. While the TQ1-rich domains are interconnected for all three categories of morphology, it is essential to have bicontinuous percolation pathways to enable both efficient hole extraction through the donor material and efficient electron extraction through the acceptor material. For Category 1 and Category 2 morphologies, the PC 61 BM content of the TQ1-rich phase is less than 20%, which is below the percolation threshold for effective charge transport of electrons (the minimum fraction of PC 61 BM required to form continuous PC 61 BM percolation pathways in a blend film). 56,57 We have effectively achieved these bicontinuous pathways with the Category 3 ink formulations, where an interdigitated network exists with domain size approaching the exciton diffusion length in these systems. Importantly, this thin film structure has been achieved with green solvents for the TQ1:PC 61 BM material system. Surface C K-edge NEXAFS spectra collected in TEY mode are presented in Fig. 5 for pristine films of TQ1 and PC 61 BM, and Fig. 6 for the eight TQ1:PC 61 BM BHJ films. All BHJ films possessed a TQ1-enriched surface, with 2-MA:Lim and o-DCB films containing the highest degree of blending, with a measured TQ1 composition of 74 ± 1% and 76 ± 1%, respectively ( Table 5). Enrichment of the donor material at the blend film surface will limit electron extraction in an OPV device of conventional architecture and facilitate electron extraction in an inverted architecture OPV device. The surface energy of the TQ1 and PC 61 BM drives the vertical structure formation observed in the blend films. TQ1, with a lower surface energy (29 mJ m −2 ) 20 than PC 61 BM (38 mJ m −2 ), 58 is driven to the film surface upon solvent evaporation, to minimise total energy.

Correlation of HSP to morphology
There is a rich complexity to phase evolution during solvent evaporation. 59,60 To develop tools to predict this phase evolution, and in turn control morphology formation, is the ultimate goal for organic photovoltaic active layers. 61 We here calculated the distance in HSP space (R a ) between both the donor and acceptor material to the primary solvent in order to determine a correlation between HSP and donor-acceptor film morphology. The calculation of R a uses the HSP for component 1 (δ D1 , δ P1 , δ H1 ) and component 2 (δ D2 , δ P2 , δ H2 ) in eqn (1).
The calculated values of R a are listed in Table 6 for the materials TQ1 (R a i ) and PC 61 BM (R a j ) to the four primary solvents Lim, Ani, 2-MA and o-DCB. A difference parameter (Δ ij ) can then be established which factors in the relative distance of a given solvent to the two solubility sphere centres (eqn (2)), using the solubility sphere radius for TQ1 (R 0 i ) and PC 61 BM (R 0 j ).
If Δ > 0 then i is more soluble than j; if Δ < 0 then j is more soluble than i. The closer R a i is to R 0 i , the system tends to insolu- bility. Hence, the solvent that is closer to the origin of the HSP sphere when comparing a binary material system is a superior solvent for that given material, under the parameters set in the HSP analysis. Calculated Δ ij values are presented in Table 6 for the four primary solvents studied. Now, there is a correlation observed between Δ ij and the degree of phase separation in the  TQ1:PC 61 BM films, as quantified by the average domain size. As Δ ij increases (that is, PC 61 BM becomes increasingly less soluble compared to TQ1) then the phase segregation (average domain size) increases. When utilising the principle of differential solubilities, the goal is to avoid the fullerene acceptor precipitating first (positive Δ), and forming large (>150 nm) aggregates in the film. We hypothesise that it is preferable for the donor polymer to precipitate first during film casting forming a web-type structure; if the fullerene acceptor is more soluble it will remain dissolved for longer and precipitate second (negative Δ), filling the interstices in the film. This order is effectively what we achieve as Δ ij decreases and we move through the solvent series: However, HSP theory only considers equilibrium thermodynamics, not kinetics. The time that the system takes to come out of solution is governed by the vapour pressure of the solventthis is the role of the additives. Considering the d-limonene series, we observe an increase in phase segregation when moving from Lim:MO → Lim:MY → Lim → Lim:2-MA and this trend correlates with an increase is vapour pressure of the system 0.002 → 0.009 → 0.15 → 0.25 kPa. When considering the anisole series, we observe an increase in phase segregation when moving from Ani:DIO → Ani:DE and this trend correlates to an increase in vapour pressure of the system (and hence decrease in drying time) from 0.00004 → 0.003 kPa. Note the vapour pressure will decrease significantly when the system reaches low amounts of remaining solvents due to the interaction between solvent and solutes. The physical parameters used here reflect bulk properties only; differences will exist between bulk properties and thin film properties.
Controlling morphology formation in these donor-acceptor systems involves managing the crystallisation kinetics and molecular diffusion kinetics. The thermodynamically stable state for a TQ1:PC 61 BM blend film is phase segregated, hence to avoid this structure formation the kinetics must be controlled, or limited. As we have observed in this study, green solvents typically have high boiling points and low vapour pressures, which means that upon spin coating or printing the solvent evaporation rate is low and donor and acceptor materials have more available time to migrate and reorganise into a thermodynamically favourable state. A similar principle to solvent annealing, which is often used to "slow down" the drying speed of the deposited wet donor-acceptor film. 62 Unfortunately, the thermodynamically favourable state of the donor-acceptor BHJ film is often not the optimal microstructure for an efficient OPV, since alike molecules are attracted to each other this means the film morphology tends towards a state comprised of large phase segregated domains. We have effectively shown in this study that HSP enables the prediction of film morphology formation in thin films of donor-acceptor systems, as the relative solubilities of the two materials can be gauged prior to experimental studies, and a nanostructured interdigitated network morphology can be achieved.

Outlook for the application of HSP to OPV printing
In this study we have successfully applied the HSP-based method to a well-known polymer donorfullerene acceptor system. Recently the field of OPV has been moving to non-fullerene acceptors (NFA), 63 due to their advantageous properties of tuneable HOMO and LUMO levels leading to higher V OC devices, broad and strong light absorption providing for high photocurrents, and superior morphological stability in thin film devices. 8,64,65 We have effectively demonstrated a strategy for the application of the HSP-based method to the screening of OPV green solvent ink formulations, which has established a clear pathway forward which can now be utilised for screening green solvents for the large suite of NFA systems available for cuttingedge OPV research and module printing. Indeed this process has already begun with reports from Xu et al. for the PBDTS-TDZ: ITIC system 66 and Ye et al. for the FTAZ:IT-M system. 40

Conclusions
Hansen solubility parameters have the potential to predict solubility in ternary (two solutes, one solvent) and quaternary (two solutes, two solvents) blends. A series of TQ1:PC 61 BM BHJ films were spin coated from green solvent ink formulations and X-ray spectromicroscopy was utilised to measure donoracceptor domain size and domain composition. We have shown that we can now engineer a nanostructured donoracceptor thin film from green solvents, with donor and acceptor domain sizes closely matched to exciton diffusion length in organic semiconductor systems. The HSP approach is versatile for different organic semiconducting material systems, and we have detailed the pathway for using the HSP semiempirical tool to predict morphologies, screen suitable ink formulations, and most importantly design nanostructured donor-acceptor thin film morphologies for efficient OPV device production. We demonstrate that by carefully considering HSP and choosing solvent blends with relative solubilities for the two materials, donor and acceptor can be driven to precipitate in the desired order necessary for forming a nanostructured interdigitated network morphology.

Conflicts of interest
There are no conflicts of interest to declare.