Morphology Changes Upon Scaling a High-Efficiency, Solution-Processed Solar Cell From Spin-Coating to Roll-to-Roll Coating

Abstract

Solution processing via roll-to-roll (R2R) coating promises a low cost, low thermal budget, sustainable revolution for the production of solar cells. Poly[(5,6-difluoro-2,1,3-benzothiadiazol-4,7-diyl)-alt-(3,3‴-di(2-octyldodecyl)-2,2′;5′,2″;5″,2‴-quaterthiophen-5,5-diyl)], PffBT4T-2OD, has recently been shown to achieve high power conversion efficiency (>10%) paired with multiple acceptors when thick films are spun-coat from hot solutions. We present detailed morphology studies of PffBT4T-2OD based bulk heterojunction films deposited by the volume manufacturing compatible techniques of blade-coating and slot-die coating. Significant aspects of the film morphology, the average crystal domain orientation and the distribution of the characteristic phase separation length scales, are remarkably different when deposited by the scalable techniques vs spun-coat. Yet, we find that optimized blade-coated devices achieve PCE >9.5%, nearly the same as spun-coat. These results challenge some widely accepted propositions regarding what is an optimal BHJ morphology and suggest the hypothesis that diversity in the morphology that supports high performance may be a characteristic of manufacturable systems, those that maintain performance when coated thicker than ≈200 nm. In situ measurements reveal the key differences in the solidification routes for spin- and blade- coating leading to the distinct film structures.

1. Introduction

Intense interest has developed in sustainable, low thermal budget, high volume production of functional coatings and devices via solution-processed roll-to-roll (R2R) manufacturing. R2R manufacturing promises to facilitate markets as diverse as wearable electronics, large area displays, healthcare, and inventory management.¹ The specific advantages of low embodied energy and high volume production may enable R2R to make early impacts in energy production and storage technologies such as batteries, super capacitors, and photovoltaics.² Embodied energy is critical in evaluating photovoltaic technologies if they are to scale to a significant fraction of world power production³ and eventually replace conventional power sources.^4,5 This fact has led to considerable research in solution processed photovoltaic architectures, including but not limited to, dye sensitized solar cells, organic photovoltaic (OPV) devices and organic-inorganic perovskite based devices. Remarkably, almost all the research in these technologies has been on small devices fabricated via spin-coating. Spin-coating was established in the batch fabrication environment of semiconductor manufacturing and is a robust, simple technique for producing uniform thin films. However, it is material consumptive and scalability is limited, creating a significant challenge to progress from the laboratory to large scale manufacturing.⁶

Great advances have been made in the performance of spin-coated, small-area, single-junction OPV devices, with multiple material systems demonstrating > 10 % power conversion efficiency (PCE).^7–10 Yet a survey of small-area, single-junction devices fabricated by manufacturable deposition techniques such as blade-coating^11,12 or slot-die coating^13,14 finds PCE levels typically (6 to 7) % with a record, proprietary material reported at 8.5%.¹⁵ Similar issues are reflected in the performance of perovskite systems, where PCEs over 20 % have been reported for spin-coated devices,¹⁶ but blade-coating¹⁷ and slot-die coating¹⁸ demonstrations are currently at ≈ 10 %. For the OPV devices studied here, functionality is achieved via the ubiquitous bulk heterojunction (BHJ) motif¹⁹ where spontaneous phase separation generates a nanoscale network of donor and acceptor species. To achieve high PCE, BHJ layers must both efficiently separate charges (primarily determined by the length scale of donor acceptor phase separation and the structure of the donor-acceptor interface) and then extract the charges (correlated to the local mobility and bi-continuous structure). These complex constraints for BHJ layers raise the significant question as to whether the spin-coated film morphology and functionality can be produced via alternate deposition approaches. In addition to challenges related to variations in the process-structure-function relationships²⁰ of solution deposited inks upon scaling the deposition technique, significant device design and engineering challenges are associated with scaling from small- to large-area devices,²¹ and with elimination of non-earth abundant materials. In addition to a scalable deposition system that maintains function at high film quality, flexible electrodes that minimize series resistance must be developed,²² along with cell pattering approaches that minimize geometric losses,²³ and optical designs that maximize efficiency.²¹ With appropriate material sets, scaling of both single junction and tandem device structures has been demonstrated.^23,24

In this paper we focus on the material processing challenges associated with scaling OPV devices based on poly[(5,6-difluoro-2,1,3-benzothiadiazol-4,7-diyl)-alt-(3,3‴-di(2-octyldodecyl)-2,2′;5′,2″;5″,2‴-quaterthiophen-5,5-diyl)], PffBT4T-2OD. PffBT4T-2OD (Figure 2a), is one of three high-hole-mobility polymers recently shown to achieve > 10 % PCE in spin-coated devices when paired with a fullerene derivative in a BHJ active layer.⁹ A minor variant, PffBT4T-C₉C₁₃, has achieved >11% PCE when processed from halogen free solvents.²⁵ These polymers are particularly suited to large scale fabrication as they achieve high PCE in thick films (> 300 nm); it is generally accepted that active layers ≥ 250 nm are necessary to fabricate devices tolerant of the roughness and electrode flagging intrinsic to web substrates. The excellent performance of PffBT4T-2OD was attributed to a strong tendency to aggregate resulting in robust polymer:fullerene morphologies that are independent of choice of fullerene acceptor.⁹ Of particular note is the requirement that these polymers be spun-coat from a hot (>90 °C) solution onto a pre-heated substrate, foreshadowing complications upon transitioning to blade-coating or slot-die coating.

Figure 2.

a, Structure of PffBT4T-2OD. b, Temperature and thickness dependence of blade-coated conventional (Ca/Al cathode) device PCE. Optimal performance is achieved at ≈ 90 °C. c,d Temperature dependent UV-Vis absorbance spectra of the coating solution, indicating transition from solution to aggregate in the temperature range (55 to 65) °C: horizontal lines indicate positions of spectra shown in d.

We demonstrate that blade-coated small-area devices achieve PCE values (>9.5 %) comparable to the initial spun-coat demonstration. Importantly, the optimal blade-coated films have qualitatively different morphologies compared to spun-coated films,^9,26 demonstrating that multiple morphologies can result in high PCE devices. This result challenges widely accepted propositions regarding what is an ideal BHJ morphology and further suggests characteristics for robust BHJ material sets. We establish the origin of the morphological differences through in-situ measurements of both the spin- and blade- coating processes. Additionally, we demonstrate that the morphology differences are robust when scaling small piece blade-coating to slot-die coating on a continuous web.

2. Results & Discussion

2.1 Coating Processes

The solidification physics controlling film thickness and drying behavior of BHJ layers is a necessary background for discussion and highlights the challenges of up-scaling to R2R manufacturing. A summary of the coating techniques employed is presented in Figure 1. For commercial scale deposition, slot-die coating is generally the most precise for the thin films relevant to organic photovoltaics. However, to date, spin-coating (Figure 1a) is the most commonly employed thin-film deposition technique in the laboratory. Spin-coating nominally proceeds in two stages²⁷ and is self-metered, i.e. the final film thickness is independent of the amount dispensed. In the first stage, the dispensed fluid is ejected from the edge of the substrate due to centrifugal forces at a rate proportional to the cube of the film thickness. This leads to highly uniform films as they self-level. This also causes the technique to be extremely wasteful of material. When the film is sufficiently thin that the ejection rate is less than the evaporation rate, the wet film thickness is established and the film dries via evaporation. The centrifugal flow of ambient atmosphere over the sample face results in extremely rapid drying and can lead to instabilities.²⁸ However, the film meniscus never passes over the substrate, so spin-coating is relatively robust with regard to ink wetting characteristics, provided that initial wetting can be established.

Figure 1.

Schematic of a, spin coating, b, blade coating, and c, slot-die coating.

Blade-coating (Figure 1b) or knife edge coating is a prototyping tool for slot-die coating and used for piece coating. Like spin-coating, it is a self-metered technique. The wet film thickness is determined by the blade velocity and the film viscosity. In the horizontal dip-coating or Landau-Levich regime,²⁹ the wet film thickness varies ~v^2/3 where ν is the substrate velocity. Unlike spin-coating, the technique is material conservative as it is simple to dispense volumes nominally matched to the final desired film area. Coating stability is determined by the wetting characteristics of the solution on the substrate and the final film morphology is determined by the drying conditions (substrate temperature and ambient gas flow) after wet film deposition.

Slot-die coating (Figure 1c) is an example of a pre-metered technique: the wet film thickness is set by the ratio of the solution delivery rate to the moving substrate (web) speed. In this regard it is very easy to control thickness and the system is economical in terms of material consumption. However, the technique is prone to instabilities (for example chatter) that are set by the wetting characteristics of the ink on the leading edge of the web and the underside of the die. Thus there will be a range of web-speed and solution rates that can be used.³⁰ Like blade-coating, the final film morphology is set by the evolution of the wet film in the downstream dryer, often under moderate heating (< 100 °C). While an optimized slot-die coater conserves material, optimization of web coaters requires a significant amount of material because one must coat the full length of the dryer in order to determine the film characteristics. For a research scale system with a 2.5 cm slot and a 2 m drier, of order a gram of material is needed for empirical optimization. Typical research syntheses produce on the order of 100 mg per series, prohibiting extensive web demonstrations. Interestingly, it has recently been shown that slot-die heads, when operated as small piece coaters, are not pre-metered, but instead act as self-metered blade-coaters with film thickness increasing with coating speed.¹³

2.2 Blade-coated device performance

Shown in Figure 2b are results for conventional (Ca/Al cathode) blade-coated devices as a function of substrate temperature and film thickness (see SI for additional device characteristics). The BHJ solution was 1:1.2 by mass (PffBT4T-2OD:PCBM-71) in a mixed solvent, 1:1 chlorobenzene (CB):o‐dichlorobenzene (oDCB), containing 3 % volume fraction of the low vapor pressure additive 1,8‐diiodooctane (DIO), see Methods for details. Excellent performance is achieved, with champion devices exhibiting PCE > 9.3 %. As observed in earlier spin-coating results,⁹ the active layer supports nearly optimal performance at thicknesses exceeding 400 nm. Substrate temperature is revealed to be the critical process variable, optimal devices are coated at (90 to 100) °C. This can be easily understood in terms of the planarization of the polymer backbone in solution, which we will hereafter describe as “aggregation.” Aggregation is readily characterized by UV-Vis absorbance spectroscopy of the coating solution as a function of temperature, shown in Figure 2c. The significant red shift in the absorbance spectrum indicates aggregation of the polymer at temperatures below ≈ 58 °C. Devices blade-coated at 55 °C exhibit poorer PCE and much lower reproducibility owing to poor film formation characteristics; the films are hazy with significant roughness. However, the consistently low PCE for devices coated at 55 °C, regardless of film thicknesses, suggests that the non‐ideality is intrinsic to the film morphology, as increasing thickness should result in increasing robustness of films with respect to most coating artifacts.

Like many high performing OPV systems,^8,31,32 PffBT4T-2OD:PCBM-71 achieves consistently higher PCE in inverted (transparent ZnO cathode) devices relative to conventional devices, as shown in Table 1 and S1. The champion and mean PCE for our blade-coated inverted devices are quite similar to those obtained for devices produced by spin-coating. However, scattering from the nanostructured ZnO layer in inverted devices interferes with precise morphological characterization; we therefore restricted the following morphological studies to active layers coated onto PEDOT:PSS. Given the similarity in surface energy of PEDOT:PSS and the polyethylenimine ethoxylated (PEIE) used as an electron injection layer for the inverted blade-coated devices, we expect similar film morphologies.

Table 1.

Summary of OPV Device Characteristics

	Short-circuit current density, Jsc (mA cm−2)	Open-Circuit Voltage, Voc (V)	Fill Factor, FF	Power conversion efficiency, PCE (%)	Thickness (nm)
Spin coated [inverted]b	18.4	0.77	0.74	10.5 [10.2]a	300 [300]
Blade 90°C [inverted]	18.4	0.77	0.70	9.9 [9.6]	243 [218]
Blade 90°C [conventional]	16.8	0.77	0.71	9.3 [9.1]	250 [254]
Blade 55°C [conventional]	14.6	0.67	0.64	6.2 [5.6]	213 [219]

^aValues in [] are average over 4 devices from a single coating. Isolated values correspond to champion devices

^btaken from literature⁹

2.3 Blade-coated film morphology

We undertook a detailed morphology study of the blade-coated films as a function of coating temperature to understand the origin of the performance differences. Shown in Figure 3 are grazing incidence X-ray diffraction (GIXD) and resonant soft X-ray scattering (R-SoXS) results for films blade-coated at the optimal 90 °C, where polymer chains are well dissolved in the solvents, and from the aggregated solution formed at 55 °C. We see that significantly different morphologies are created. GIXD provides insight into the crystalline regions of the polymer. As observed for most “hairy rigid-rod” (flexible side chains attached to a rigid backbone) polymers, the crystal structure consists of π-packed conjugated cores separated into lamella by the solubilizing side chains. There is no evidence for crystalline PCBM domains in the 2-d GIXD patterns (Figure 3b), only an amorphous halo at q ≈ 1.35 Å ⁻¹. The GIXD data further indicate that at 90 °C, for the optimal films, the polymer crystallites exhibit a strong preferential edge-on orientation with respect to the substrate (Herman’s order parameter is 0.82, 1.0 is a perfectly edge on distribution) with up to 4 orders of the (h00) diffraction series due to the lamella separation along surface normal observed.

Figure 3.

a, UV-Vis spectra from devices coated at 90 °C and 55 °C, compared to the solution absorbance of aggregates formed at 45 °C. b, GIXD of films coated at (55 and 90) °C. c, (100) pole figures from GIXD of films coated at (55 and 90) °C, and relative degree of crystallinity (DOC) from the integrated (100) pole figure. χ=0° is the out-of-plane axis and χ=90° is the in-plane axis. d, R-SoXS profiles from films coated at 55 °C and 90 °C and relative scattering intensities normalized to the 55 °C sample. All error bars are standard deviation of the mean of multiple measurements.

In contrast, for films coated at 55 °C, a more face-on orientation (characterized by the prominent (010) π-packing feature along the surface normal) is adopted (Herman’s order parameter −0.09, −0.5 is perfectly face on). Detailed analysis of the GIXD patterns can be found in the SI. The differences in crystallite orientation of the polymer between the 90 °C and 55 °C samples is also evidenced by the variation of intensity at different scattering angles in the (100) pole figure plot shown in Figure 3c. The relative degree of crystallinity (DOC) assessed from the integrated (100) pole figure is shown in Figure 3c for BHJs and for the neat polymer. Coating the neat polymer at 55 °C results in an increase in the DOC relative to coating at 90 °C, consistent with the hypothesis that early solution aggregation should promote higher DOC. However, coating the BHJ at 55 °C results in no significant change in the DOC relative to the BHJ coated at 90 °C, and both BHJ films exhibit a DOC lower than the 55 °C neat polymer reference. The relative decrease in DOC of the BHJ at the lower casting temperature may be attributed to slower crystallization due to the presence of the high glass transition temperature (T_g ≈ 163 °C) fullerene.³³ A clear T_g has not been observed for the polymer, but a strong melting transition is observed at ~283 °C in DSC. Using typical scaling rules, T_g of the polymer should lie between (60 and 116) °C and the T_g of the 1:1.2 blend estimated from the Fox equation (110 to 140) °C is such that vitrification will more significantly impact the film kinetics at the 55 °C drying temperature than at 90 °C.

R-SoXS provides insight into the spatial distributions of the composition correlations (phases) along with the relative phase purities over length scales spanning ≈ (10 to 1000) nm. When compared to the GIXD results, similar significant differences are observed in the R-SoXS. Interestingly the optimal 90 °C structure exhibits a clear bimodal scattering pattern, with characteristic length scales or long period (=2π/q_peak) of 55 nm and 220 nm. In contrast, the 55 °C film exhibits a single, broad scattering feature with long period of 105 nm (Figure 3d). The average composition variation (and relative average domain purity) can also be revealed by R-SoXS via computation of the total scattering invariant, TSI≡∫Iq² dq. Completely mixed domains result in no scattering over the q-range probed, while a two-phase morphology with pure phases will yield maximum scattering. The optimal 90 °C film is found to have ≈30 % greater phase purity than films cast at 55 °C. In addition to size and composition variation, the molecular orientation of the polymer relative to the dominant, discrete polymer-fullerene interface can also be a critical structure parameter that impacts performance. This orientation may influence both exciton dissociation at and charge transport near an interface, with “face-to-sphere” orientation (face of the conjugated plane of the polymer oriented toward the fullerene-rich domain) being correlated to improved J_sc and FF.^34,35 The extent of molecular orientation in the blade-coated films was quantified using polarized R-SoXS (results shown in Figure S3 and Table S1) following previously established procedures.³⁶ The anisotropy order parameter is small and positive, with similar values for films coated at 90 °C and 55 °C. The positive values indicate mild preference for the beneficial “face-to-sphere”.

The three most significant differences between the 90 °C blade-coated films and the 55 °C blade-coated films lie in the crystal orientation (edge-on vs face-on), the phase segregation morphology (bimodal vs simple), and the phase purity. The very similar device PCE achieved for the blade-coated films, compared to the spin-coated devices, together with the similar casting temperatures of (90 to 100) °C would suggest that similar film morphologies are created. The morphology of spin coated PffBT4T-2OD:PCBM-71 devices has been studied in detail.^9,26 Compared to poor-performing high-spin-speed films, optimal low-spin-speed films were characterized by face-on crystal orientations, simple R-SoXS scattering distributions peaked at ≈80 nm, and high phase purity. Thus optimized spin-coated films have very distinct morphology compared to optimized blade-coated films. This indicates that PffBT4T-2OD can support high performance with a diversity of morphologies and challenges widely accepted propositions that specific orientations (face-on),³⁷ hierarchical phase distributions,³⁸ or a significant mixed amorphous phase are critical to high performance. Comparison of the morphologies of high performing (warm, bladed or slow, hot, spun-coat^9,26) PffBT4T-2OD with poorer performing (cold, bladed or fast, hot, spun-coat^9,26) processes allow us to identify relevant and irrelevant features of the morphology in this system. Achieving sufficient carrier mobility is considered critical to high FF and extraction efficiency.³⁹ For all high performing PffBT4T-2OD films, a high level of crystalline order is observed, presumably supporting high mobility, yet crystal orientation (edge-on vs face-on) is clearly not important, consistent with the emerging consensus that transport in most semicrystalline polymers is highly 1-dimensional along the backbone.^40,41 While all high performing films exhibit characteristic phase-separation length scales between (50 and 200) nm, the specifics of the distribution (hierarchical nature³⁸) do not appear to be critical. However, the overall phase purity appears to be quite important. While in some systems, such as P3HT, a significant mixed phase is considered important to support transport energy gradients,⁴² PffBT4T-2OD clearly optimizes at the highest phase purity supported by the processing scheme in agreement with recent studies that show that high average domain purity can be an important factor in reducing geminate as well as bimolecular recombination.⁴³ Quantitative comparison of the phase purity and DOC of high performing blade- and spun-coat films (see SI) indicates the two coating methods achieve the same high level (differ by only 5%) of relative phase purity and similar levels of DOC.

2.4 In situ experiments

Insight into the origin of the processing dependence of the morphology and performance of PffBT4T-2OD is obtained through in-situ studies of the film drying. As shown in Figure 2, the absorbance at ≈ 700 nm can be used as a diagnostic for polymer aggregation. Shown in Figure 4 is the film thickness (from white light interferometry) and film absorbance at 700 nm (from UV-Vis transmission) during spin-coating and blade-coating. Full spectra are in the SI. The data clearly demonstrate significant differences between the two deposition techniques. During blade-coating at 90 °C, the thickness evolves in two distinct stages: rapid drying of the primary binary solvent mixture, ending at ≈ 7 s, followed by slow drying of the lower vapor pressure additive (DIO), ending at ≈ 110 s. The polymer aggregates very late in the solvent drying stage, at a wet film thickness of ≈ 800 nm that corresponds to a solution concentration of 370 mg/mL. The aggregation is essentially complete upon removal of the primary solvent; there is very little evolution of the spectrum during the DIO drying phase. During spin-coating, aggregation is seen to occur much earlier in the solvent drying stage, at a wet film thickness of ≈ 8 μm. The spun-coat solution concentration at aggregation onset is ≈ 50 mg/mL, similar to the concentration of the solution studied in Figure 2a. This suggests that, at 8 s, the initially 110 °C substrate and solution have cooled by convection to below the aggregation point of ≈ 58 °C. There is a slight additional increase in the aggregation as the primary solvent leaves the film (up to ≈ 80 s). Assuming the cooling rate is linear in time, the initially 110 °C substrate will reach room temperature well before the primary solvent leaves, consistent with the significantly slower evaporation of the primary solvent compared to the 90 °C isothermal blade-coating. The subsequent drying of the DIO, near room temperature, is not complete after 1 hr (364 nm thickness). Dry film thickness (315 nm) was record after vacuum drying. The origin of the similarity between the morphology of the optimized spin-coated films and the 55 °C blade-coated films is now clear, as the spin-coated films are deposited from a cooled, aggregated solution as posited in the original report.⁹ However, it is also clear that the complex thermal history of the spun-coat aggregates result in a final structure more conducive to OPV performance than that created by isothermal deposition at 55 °C.

Figure 4.

Film thickness (black) and absorbance at 700 nm (red) from in-situ UV-Vis absorbance studies during spin-coating and blade-coating at optimized conditions [spin speed 800(2π) rad/min, i.e. 800 rpm, blade velocity 40 mm/s]. Increase in red-shifted absorbance (see Fig.1) is indicative of aggregation.

Having identified the critical difference between spin-coating (complex and uncontrolled thermal trajectory) and blade-coating (isothermal drying) we have further studied the kinetics of blade-coated film structure evolution. Shown in Figure 5 is the time dependence of the (100) pole figure from in situ GIXD for a film blade-coated at the optimal conditions (90 °C, 250 nm final thickness) and blade-coated at a temperature (55 °C) just below the aggregation threshold; also shown in the figure are the film thicknesses (from simultaneous optical reflectometry) and the thickness normalized pole figure integrals (proportional to crystallinity). The structure evolution of the film during 90 °C coating is very similar to that recently reported for BHJs of both the canonical crystalline polymer poly(3-hexyl-thiophene) (P3HT)⁴⁴ and that of a high PCE small molecule.⁴⁵ There is no evidence for polymer crystallization until very late in the primary solvent drying phase (suggesting extreme super-saturation and a high nucleation barrier). A small amount of crystal is formed as the last of the primary solvent evaporates; however, the majority of crystallization occurs late during the evaporation of the additive. This majority of crystallization is characterized by a slight broadening of the orientation distribution. The evolution of the film thickness at 55 °C is similar, but slower due to the reduced vapor pressure of the components. Unlike the 90 °C case, there is trace development of crystals early in drying, suggesting a higher nucleation rate, consistent with the lower temperature of the solution. However, as at 90 °C, significant diffraction again only occurs late in the removal of the DIO additive. Note that the evolution of the crystallinity (GIXD) is quite distinct from the evolution of the aggregation (UV-Vis, Figure 4). A lack of correlation between aggregation and crystallization has also been seen for the drying of P3HT.^44,46 The strong red shift in UV-Vis is due to planarization of the polymer back-bone and does not require the development of long-range crystalline order.

Figure 5.

Results from in-situ GIXD during blade-coating as a function of temperature. Top row: false-color plot of (100) pole figure vs time. Bottom row: film thickness (black) and integrated pole figure (red) vs time for films blade-coated at 90 °C (left panel) and 55 °C (right panel).

We noted earlier that a significant difference between films coated at 90 °C and 55 °C is the resultant crystal orientation distribution. For an alkyl side chain polymer in the conventional π-stacked lamella structure, the lowest free energy crystal faces are the alkyl dominated (100) faces while the π-π (010) packing direction should be of higher energy. At the air interface, edge-on packing, that minimizes the interfacial energy, would be expected. For example in the case of P3HT films, polymer orientation can be controlled by drying speed. At low spin speeds, closer to equilibrium, an edge-on structure is adopted. However, at high spin speeds, a face-on structure can be forced.²¹ We attribute the temperature dependence in orientation exhibited by the blade-coated films to comprehensive film drying nucleated at the air interface in the case of the film coated at 90 °C (note the highly oriented early time crystals) vs pre-formed solution aggregate deposition in the case of films coated at 55 °C. The higher free energy (010) face of the aggregate may favor face-on adsorption at the higher free energy PEDOT:PSS interface. The structurally similar polymer PBTffT4T-2OD⁹ exhibits a similar orientation dependence on substrate temperature (edge-on at high temperature, face-on at lower temperature, see SI).

As we discuss in the previous section, small-angle X-ray scattering techniques such as R-SoXS can provide valuable information about the length scale of composition fluctuations and the extent of phase separation and phase purity. Unfortunately, soft x-rays typically require a vacuum environment and thus RSoXS is not currently used for in situ studies. Instead, we can apply grazing-incidence small angle X-ray scattering (GISAXS) at hard X-ray energies to follow the evolution of these quantities during drying. Although GISAXS has poorer phase contrast than R-SoXS, it is straightforward to make GISAXS measurements in air at atmospheric pressures. Shown in Figure 6 are in situ GISAXS results under comparable conditions to the GIXD results of Figure 5. Quantification of GISAXS is challenging due to the complexity of electric field structure above the surface when the incident beam is above the critical angle for the film but below the critical angle for the substrate. See Methods for the approximations we used in the development of a pseudo total scattering invariant. In general, the GISAXS evolution is similar to that of the GIXD: appearance of minor structure late in the primary solvent drying stage and dominant feature growth late in the additive drying stage. It is interesting to note that, for 90 °C coating, the initial scattering pattern is distinctly different from the final film: a single feature is present at q_x ≈ 0.012 Å⁻¹ (2π/q_x ≈ 52 nm). The low q dominant peak only grows in during the DIO removal stage. This is suggestive of classic nucleation and growth of features, consistent with the initial well solvated polymer. In contrast, for 55 °C coating the GISAXS pattern is nominally time invariant, and only the intensity changes. This suggests that the characteristic length scale is pinned by the polymer characteristics (mol mass and dispersity) and the aggregate formation in solution. Thus the different final phase separation distributions can be attributed to the different kinetic paths at (90 and 55) °C: nucleation and growth in the confined, nearly dry film at 90 °C, and slow refinement of adsorbed aggregates at 55 °C.

Figure 6.

In situ GISAXS as a function of time and temperature for 90 °C (left panel) and 55 °C (right panel) blade-coated samples. Top row: false-color image of Iq² vs time, middle row: Kratky plots of Iq² at 10 s and when dry, bottom row: film thickness (black) and pseudo-TSI (red) vs time.

2.5 Slot-die coated films

Given the remarkable morphological differences between optimal blade-coated films (edge‐on crystallite orientation, bimodal characteristic length scale) and optimal spin-coated films (face‐on crystallite orientation, monomodal characteristic length scale), we performed detailed studies of active layers deposited by slot-die coating on a continuous web. The work was performed on a prototype scale system with a 2.5 cm wide slot-die head and a ≈ 2 m long, open, forced-air dryer (see SI for details). The system was designed for in-situ monitoring of film drying behavior and is not capable of fabricating completed devices without winding and unwinding multiple times resulting in degraded devices. Shown in Figure 7a is the coating region and a deposited film on a PEDOT-PSS coated PET web. Accurate X-ray characterization on PET is difficult due to roughness and scattering from the web itself.⁴⁷ Similarly, device measurements on large area films are very sensitive to the overall engineering of the cell and electrode structure² and thus are not good tests of the film quality. Therefore, for precise comparison to the blade-coated films, the slot-die films were transferred to Si substrates for R-SoXS and GIXD characterization.

Figure 7.

a) Image of coating region of prototype system. Visible are the CCD camera for slot-die web monitoring, fiber based UV-Vis system for wet film monitoring, and heater for transport to forced air drier. Arrow indicates die/web gap. b–d) Comparison of structural results on slot-die coated vs blade coated films at 90°C, with 3 % DIO. b) in-situ UV-vis c) ex-situ R-SoXS (Inset: relative scattering intensity) d) ex-situ GIXD.

We performed slot-die coating with 90 °C slot, support roller, transport heater and drier temperatures with an identical solution to that used for blade coating. Stable coating was achieved at web speeds between (1.0 and 25) mm/s. Shown in Figure 7b are UV-Vis spectra recorded as the web leaves the die roller, ≈ 7.2 s after film deposition. Also shown is in-situ UV-Vis during blade-coating at 90 °C, ≈ 7.2 s after blade passage. Due to the fast evaporation at 90 °C, all of the primary binary solvent is absent at this point and the web consists of the DIO swollen active layer (see Figure 5). The spectra are characteristic of a fully aggregated film as can be seen by comparison to Figure 2. Also shown in Figure 7b are comparisons of the dry slot-die coated and blade coated films; again, the details of the spectra are nominally identical. Comparisons between slot-die coated films, after drying and transfer to Si substrates, and blade-coated films are shown in R-SoXS (Figure 7c) and GIXD (Figure 7d). Note the similarity of both the phase segregation length scale (shape) and phase purity (intensity) based on the R-SoXS and orientation distribution based on the GIXD. It is clear that the final film morphology from the blade-coated and slot-die coated films are quite similar, confirming that self-metered small-piece blade-coating is an excellent prototyping tool for pre-metered slot-die coating on a continuous web.

3. Conclusion

We have demonstrated that with proper care, PffBT4T-2OD BHJ active layers can be deposited by manufacturable techniques, maintaining the very high PCEs initially obtained by spin-coating. Surprisingly, we find that the detailed morphology of the active layer in the optimized blade- and slot-die coated layers is distinctly different than in the spin-coated layers. In situ measurements clearly establish that the morphological differences arise due to the radically different cooling characteristics across the deposition techniques. We suggest that the ability of multiple morphologies to support high performance may be a characteristic of manufacturable systems that are thickness tolerant as homogeneous structures satisfying rigid morphological constraints become less likely as film thickness increases. This hypothesis is supported by a recent study of the polymer PBnDT-FTAZ where films of ≈250 nm thickness achieved comparable performance when processed to produce markedly different phase structure.⁴⁸ Similarly, the canonical polymer P3HT, that can support films up to 1 μm in thickness,⁴⁹ exhibits markedly different morphologies when optimally thermally annealed⁵⁰ vs solvent annealed.⁵¹ In spite of the morphological diversity displayed, high crystallinity (driven by the strong tendency to aggregate) and high phase purity (set by subtle details of the kinetic paths) appear to be key to high performance for PffBT4T-2OD.

The demonstration of manufacturable deposition of thick active-layers that support near 10 % performance in single-junction small-area devices bodes well for the development of light-weight, flexible, integrated energy generation products such as window coverings for buildings and automobiles and self-powered smart signage. A significant degree of morphological tolerance is expected to be a common feature of manufacturable systems, indicating that the search space for even higher performing systems is of encouraging size.

HR, JD, SE, SM, LJ contributed equally, performing the bulk of the experimental work. AH perform STEM measurements and analysis. DD, LR, HA, AA, and HY participated in initial experimental design, all authors participated in manuscript preparation and review.

Methods

1) Film preparation

a. blade-coating

The BHJ master solution was a 1:1.2 (by mass) mixture of PffBT4T-2OD (1-Material, Inc. M_w (1 to 1.5)×10⁵, PS Standard; polydispersity index (2.0 to 2.5))⁵² and PCBM-71 (Nano-C) well-dissolved in 1:1 chlorobenzene (CB):1,2 ortho-dichlorobenzene (oDCB) at 44 mg/mL (total solids) concentration at 100 °C. Additive solutions were prepared by addition of 3 % (by volume with respect to CB:oDCB) of 1,8-diiodooctane. Films were prepared using the blade-coating technique, as previously described.⁵³ Approximately 20 μL of the BHJ solution was dispensed from a warm syringe under the edge of glass blade fixed at 300 μm above the substrate surface. The substrate temperature was varied from (55 to 110) °C. The stage was immediately translated at (5 to 77) mm/s of velocity. The final BHJ film ellipsometric thickness was nominally (100 to 500) nm. Care was taken to orient the blade translation direction parallel to the sample stage of the instrument to avoid any possible thickness gradient. All films used for structure characterization were prepared on PEDOT:PSS treated substrates to closely mimic the device films (see below).

a. slot-die-coating

Slot-die coated films were produced with a prototype web coater custom designed by Innovative Technologies. The coater consists of a tensioned web drive, 25 (there was a wrong comma) mm wide slot-die head, and 2 m long, open, linear, air levitation drier. Both the slot-die head and support roller are heated internally by a common fluid bath, allowing operation up to ≈100 °C. The air levitation drier can operate at up to ≈130 °C. The substrate was 62 mm wide PET (0.125 mm thick, Melinex®, ST505 adhesion treated 2-sides, from Tekra). A PEDOT-PSS layer was coated at 1.5 m/min web speed from a 1:1 diluted PEDOT-PSS:IPA solution. The slot-die temperature and dryer temperature were 40 °C and 70 °C, respectively. The dry film was re-wound, and then coated with the BHJ solution as described above at a web speed of 1m/min. Coating was established at 127 μm die-web gap and then incrementally increased to establish stable coating of a 25 mm wide film. The slot-die temperature and dryer temperature were 90 °C.

2) OPV devices

Conventional blade coated devices: Indium-tin-oxide-(ITO)-patterned Eagle 2k glass substrates (150 nm thickness, sheet resistance < 20±5 Ohms/□, Thin Film Devices, Inc.) were carefully cleaned by sonication in acetone and isopropanol for 10 min, respectively, followed by exposure to ultraviolet-generated ozone (UVO) for 10 min. PEDOT:PSS (CleviosTM PVP Al 4083) was spin coated at 4000(2π) rad/min (i.e., 4000 rpm) for 60 sec, followed by thermal treatment at 130 °C for 20 min in ambient air, resulting in a nominally 35 nm film. Active layer BHJ films were deposited in a N₂ glove box by blade-coating. Active layers deposited in air exhibited ≈20 % lower PCE. Thermally evaporated calcium (20 nm) and aluminum (80 nm) layers were deposited through shadow masks in a high vacuum chamber. Current density-voltage curves were measured under AM 1.5G 100 mW cm⁻² illumination using an Agilent 4155C parameter analyzer in an Ar atmosphere. The OPV device performance was referenced to a KG5 filtered silicon photodiode calibrated by the National Renewable Energy Laboratory (NREL). Active area of .04 cm⁻² was defined by perpendicular overlap of 2 mm ITO and 2 mm metal electrodes.

Inverted blade coated devices: ITO-patterned glass substrates were carefully cleaned as above. A 30 nm ZnO film was spin-coated from solution precursor following Ref [⁵⁴] and converted in air at 200 °C for 60 min. A ~ 2 nm PEIE layer was spin coated and dried at 110 °C for 5 min. The BHJ solution was blade coated as above. Devices were finished by physical vapor deposition of 10 nm MoO_x and 80 nm Ag.

For conventional devices, optimal performance was achieved with films of (200 to 300) nm thickness, coated at (90 to 100) °C. For inverted devices, optimal performance was achieved with films of (200 to 300) nm thickness, coated at (70 to 90) °C.

For spin-coated devices: A 20 mg/mL total solids (1:1.2 polymer:fullerene) solution in 6:4 CB:oDCB with 2.5 % DIO, held at 110 °C was deposited onto a substrate, preheated to 110 °C. The spin speed was 800(2π) rad/min. Coating was performed in an inert air glove box. The inverted device structure was that of Ref [⁵⁵]

3) In-situ optical measurements

All in-situ measurements (optical, X-ray) were performed in laboratory ambient conditions. Minor variations in the drying kinetics are observed due to the specific circulation conditions in the distinct bays housing the lab-scale blade-coating optical measurements, the lab-scale spin-coating optical measurements, and the ALS sited blade-coating X-ray-scattering measurements.

In-situ transmission (absorbance) and reflection (thickness) data was acquired with custom built sample stages and commercial diode array based spectrometers optimized for the specific applications. For thin solution measurements, a drop of solution was placed between two 50 mm by 75 mm glass slides and the slide placed on a temperature controlled stage with a central hole for normal incidence transmission measurements. Evaporative loss from the edges was found to be negligible over the course of the measurement. For blade coated film drying measurements, an unpolarized beam was incident on the film at nominally 55°. The p-pol transmission (calcite polarizer) and s-pol reflection (broad band thin film polarizer) were defined in the collection path. This allows robust determination of film thickness from the reflectivity channel and acquisition of absorbance spectra nominally unperturbed by thin film interference. In-situ transmission measurements on the web were again performed with p-polarization at ≈ 51° angle of incidence from the web normal. While blade coating can produce anisotropic films, both polarized UV-vis and anisotropic ellipsometry measurements on both neat polymer films and BHJ films cast at both 90 °C and 50 °C detected no significant (less than 3%) variation in optical properties along vs across the blading direction. Thus isotropic optical models were used in the analysis of all data. Details of the determination of thickness can be found in Ref [⁵⁶]. For spun-coat film drying measurements, a combined spectroscopic ellipsometry (SE; M-2000XI, J. A. Woollam Co., Inc) and reflectometry (F20-UVX spectrometer, Filmetrics, Inc.) setup was used for thickness measurements. The SE measurements were obtained at an incidence angle of 70° from substrate normal. For reflectometry measurements, an unpolarized beam was incident on the film at an angle of ≈ 0° from substrate normal. The in- situ absorbance measurements were performed using a setup described in a previous work Ref [⁵⁶].

4) Resonant Soft X-ray Scattering (R-SoXS)

Ex-situ R-SoXS measurements were performed at beamline 11.0.1.2⁵⁷ at the Advanced Light Source (ALS), following previously used established methods and protocols,³⁶ by transferring sections of blend films onto silicon nitride windows (Norcada Inc.).

The R-SoXS technique utilizes the unique optical contrast between the donor molecule and fullerene near the carbon K-edge to achieve high sensitivity in thin films observed in transmission. The real dispersive part of the refractive index, 1−δ, and the imaginary absorptive part, β, for PBffT4T-2OD and PC₇₁BM are unique fingerprints of each material and provide scattering contrast that is proportional to Δn²=Δδ² + Δβ². Data were acquired below the absorption edge at 284.2 eV to optimize material contrast over the mass-thickness contrast and avoid damage⁵⁸ and fluorescence background. In transmission R-SoXS, the path length of the X-rays through the sample affects the scattering intensity following the Beer-Lambert law.⁴⁸ The Lorentz corrected⁵⁹ circular and sector averaged R-SoXS scattering profiles were subsequently normalized for absorption and thickness.

5) Ex-situ STEM

Specimens were prepared for scanning transmission electron microscopy (STEM) analysis by delaminating the film via dissolution of the PEDOT:PSS layer in deionized water. The floating films were then attached to a 200-mesh copper TEM grid and allowed to dry. Characterization was performed using an aberration-corrected FEI Titan 80–300 operated with a primary beam energy of 300 keV and a probe current of approximately 60 pA. Low-angle annular dark field (LAADF) images were collected using a Fischione model 3000 detector configured such that the inner angle of collection was approximately 17 mrads while the outer was 96 mrads. A small convergence angle (4 mrads) was used to limit contribution of the transmitted beam to the LAADF signal. The images presented in this study are shown with inverted contrast, so as to emphasize the donor material network.

6) Hard X-Ray Scattering

In-situ hard x-ray scattering measurements were performed at the Advanced Light Source⁶⁰ beam line 7-3-3, with a beam energy of 10 keV. For GIXD measurements a 2D image detector (Dectris Pilatus 2M) was located at a distance of ≈ 260 mm from the sample center. Slits were adjusted to produce a nominally 0.3 mm high beam which overfilled the nominally 2 cm wide substrate. In-situ GISAXS was performed at the same beam line with the 2D image detector-to-sample distance 3850 mm. An evacuated flight tube was used to minimize air scatter. The X-ray beam was attenuated to eliminate sample damage. The detector was calibrated with a silver behenate standard. Both GIXD and GISAXS data was reduced with the Nika software package.⁶¹ Simultaneous with both GIXD and GISAXS measurements, normal incidence spectral reflectometry was performed with a home-made, fiber spectrometer based system. Care was take to overlap the reflectometry probe beam (≈ 0.3 mm diameter) with the stripe illuminated by X-rays. The reflectometry was analyzed using a commercial ellipsometry code (JA Woollam WVASE32). X-ray data were recorded with a variable integration time and period. The initial 120 s were recorded at ≈0.1 s integration and period while the next (3 to 8) min were recorded with 1 s integration and 1.5 s period. The reflectometry was recorded at a constant ≈0.1 s integration and period. Ex-situ (static) scattering measurements were performed with 9 s integration.

For quantitation of GIXD, pole figures were constructed from the data after correcting for both the ‘missing wedge’ and detector solid angle effects.⁶² For quantitation of GISAXS, we make the approximation that the enhanced signal at q_z~0 (the Yoneda peak) is a good proxy for q_x and assume a 3D isotropic scattering pattern such that the total scattering invariant becomes ∫dq Iq². This approximation is well supported by the excellent agreement between Iq² from the GISAXS and Iq² from the R-SoXS (see SI for detailed comparison), nevertheless we will refer to pseudo TSI in the case of our in-situ measurements to indicate the grazing incidence approximation with respect to the results obtained via transmission R-SoXS. We note that the ≈ 4 m flight path of the GISAXS measurement limits the low q range and, as can be seen from comparison to the R-SoXS, a significant amount of low q scattering is missed in the in-situ experiment. This can lead to a bias in relative phase purity when interpreting the evolution of the pseudo-TSI.

Broader context.

Large area processing of solution deposited active layers can enable significant advances in sustainable manufacturing of diverse products (energy capture, energy storage, electronic displays, medical sensors) due to the intrinsic benefits of efficient material consumption and low energy processing. The epitome of large area processing is continuous roll-to-roll manufacturing (R2R). R2R processing has met with significant challenges when applied to highly functional products such as solar cells and organic light emitting diodes. Significant engineering challenges are faced in transitioning traditional rigid material sets, such as ITO based transparent electrodes, to flexible replacements. Additionally, significant processing challenges arise associated with maintaining active layer performance when transitioning from spin-coated demonstration platforms to continuous manufacturing, such as slot-die, deposition techniques. In this report we apply detailed in-situ and ex-situ film morphology measurements to characterize the transition from spin-coating to slot-die coating of a high-performance organic photovoltaic candidate system. The temperature dependent aggregation characteristics of the material result in a significant processing dependence to the morphology, yet distinctly different morphologies are found to provide similar ultimate performance. This allows significant insight into the essential characteristics of high performance and provides design cues for materials optimized for R2R processing.