A. Reflectometry

One of the most utilized thin film metrology techniques during BHJ film formation is thin film reflectometry. The thickness of a drying film can be calculated by recording the interference fringe patterns (Fig. 2) in reflection mode with Eqs. (1) and (2) ^{Reference Harrick32} :

where θ is angle of incidence, n is refractive index, d is film thickness, λ is the wave length of light, and m is the order of the fringe. For reflection from a film on a substrate with higher refractive index, Eqs. (1) and (2) represent the locations of maxima and minima, respectively. If the refractive index is known, the film thickness can be calculated easily with this principle. The accuracy of determining the thickness depends on the accuracy of measuring fringe spacing ∆λ. Hence, it is important that the fringes are not distorted by any absorption bands, especially when absorbing films are measured. Additionally, using many fringes (large m) increases the accuracy, which is only possible if the refractive index does not change drastically.

FIG. 2.

Schematic for thin film interference on substrates with higher refractive indices (left). Exemplary reflectance spectrum of P3HT:PCBM drying from ODCB solution measured with white light reflectometry. Interference fringes are present after the absorption edge of P3HT. Reproduced from Ref. Reference Güldal, Kassar, Berlinghof, Ameri, Osvet, Pacios, Li Destri, Unruh and Brabec49 with permission from RSC.

The calculation of film thickness from white light reflectometry (i.e., spectral reflectometry) varies slightly, especially when the measured layer’s refractive index changes over a certain wave length range where fringes are present. The film thickness from such layers can be determined from two adjacent maxima (or minima) as in Eq. (3) ^{Reference Swanepoel33} :

Although not as accurate as ellipsometry, reflectometry is a powerful tool when the experiments and data processing are handled correctly. Caution must be exercised when reflectometry is utilized for drying of BHJ layers:

1. (i) The interference phenomenon stems from ‘coherence’. The interference fringes vanish when the path length difference between two reflected light waves is greater than the coherence length of the source. ^{Reference Ficek and Swain34} Hence, if a white light source is taken, which has finite coherence length, the wet layer thickness of a freshly coated layer should not exceed few tens of µm. This can be resolved by utilizing a narrow band filter or a monochromatic light source with longer coherence length, for example a laser. Nevertheless, the spectral information would be lost in the case of laser being the light source.

2. (ii) Despite the fact that spectral reflectometry with white light source has limited coherence length, the spectral information can be advantageous for thin film drying experiments. Although it cannot be translated directly, reflection spectrum of a drying BHJ layer on a higher refractive indexed substrate would show the absorption behavior within a certain wave length range depending on the band gap of the materials used. Within the absorption range, interference fringes are either damped or completely lost. Therefore, as mentioned previously, thickness calculations are generally conducted in the transparent range (between absorption edge and 1100 nm), where interference fringes are seen. However, changes in the absorption spectrum during drying affect the optical constants of the layer which translate into reflection spectrum. This allows the user to track the evolution of absorption behavior of the layer.

3. (iii) The calculation based on fringe counting, as explained above, is only possible if the optical constants do not change drastically during drying. This is however hardly the case with a crystallizing layer. Since most of the donor materials used for BHJs are semicrystalline materials, crystallization takes place during drying. Due to the increasing ordering and solvent evaporation, optical constants change massively, ^{Reference Campoy-Quiles, Schmidt, Nassyrov, Pena, Goni, Alonso and Garriga35} hence interference fringes will be damped or even lost. Another mechanism which might affect the interference patterns detrimentally is scattering. If for any reason (e.g., immiscibility) two phases with different optical densities form within a drying layer with sizes in the order of the incident wave length, the main optical process would be scattering. Due to its incoherent nature, ^{Reference Berne and Pecora36} scattering causes interference losses. This additionally causes decay of fringe intensity (lower fringe visibility) toward UV region, since scattering is stronger in UV region than in visible or IR region (Rayleigh scattering). ^{Reference Berne and Pecora36}

An alternative to reflection measurements is transmission-based optical measurements. In transmission geometry, the information of the film thickness would be lost. Nevertheless, the absorption behavior during the transition from solution state to solid state can be recorded. ^{Reference Zhao, Hu, Spada, Jagadamma, Yan, Abdelsamie, Yang, Yu, Munir, Li, Ndjawa and Amassian37,Reference Abdelsamie, Zhao, Niazi, Chou and Amassian38}

B. Ellipsometry

Ellipsometry in general is a very sensitive measurement technique based on polarized light for characterization of thin films and surfaces. It analyzes the phase and amplitude of reflected polarized light to extract information from the layer. Since the information contains ‘phase’ information, this measurement technique is more sensitive than reflectometry. ^{Reference Woollam39} Some of the common material properties which can be determined by ellipsometry are optical constants, thin film thickness, surface roughness, optical anisotropy, and compositional changes within the measured material/thin film. One of the most used ellipsometry type is variable angle spectroscopic ellipsometry (VASE). The fitting of the data can be conducted more easily if the measured data provide more details for the model parameters which are to be determined or fixed. Hence, spectral measurements and various incidence angles give the measurement technique more flexibility than single-angle and/or single-wavelength ellipsometers.

Spectroscopic ellipsometry (SE) measures the response of the material as a function of the angular frequency ω. The determined properties are the refractive index n(ω) and dielectric function ε(ω). In absorbing materials like BHJ layers, the dielectric function and the refractive index are complex [Eqs. (4) and (5)] ^{Reference Losurdo and Hingerl40} :

The imaginary part of the equations is proportional to energy dissipation (absorption of the light). The real and imaginary part of the dielectric function can be directly calculated from the refractive index (n) and extinction coefficient (k) [Eqs. (6) and (7)] ^{Reference Losurdo and Hingerl40} :

In the case of the reflection of a light beam from a planar interface between two optically isotropic media, the ratio of p- and s-polarized Fresnel reflection coefficients (i.e., complex reflection ratio) is the quantity measured by SE with the help of ellipsometric angles Ψ and Δ [Eq. (8)] ^{Reference Tompkins and Irene41} :

Fresnel coefficients provide a connection between the measured Ψ and Δ with intrinsic sample properties like dielectric function. Originally these coefficients stem from the boundary conditions defined by the Maxwell equations, which lead to Fresnel equations. ^{Reference Losurdo and Hingerl40} At the end, a dielectric function is derived by using the measured parameters and Fresnel coefficients [Eq. (9)] ^{Reference Losurdo and Hingerl40} :

Quantitative data for the measured samples can be derived by theoretical models to fit the real and imaginary parts of the dielectric function. Two of the models used for transparent films are Cauchy and Sellmeier models. One of the restrictions of Cauchy model is its Kramers–Kroening inconsistency. Hence, compared to Sellmeier which is Kramers–Kroening consistent, the Cauchy model can lead to unphysical optical dispersions. ^{Reference Tompkins and Irene41} Nevertheless, both models provide good approximations for transparent films or absorbing films in transparent range. Absorbing films in the absorption range, on the other hand, must account for real and imaginary parts, as explained previously. For this purpose, many dispersion relationships use oscillator theories. Some of the most common oscillators are Lorentz, Harmonic, and Gaussian oscillators. Since they are all Kramers–Kroening consistent, the modeled imaginary part is used to define the real component. More information on the theory and explanation of these oscillators and their more advanced versions (e.g., Tauc-Lorentz) can be found elsewhere. ^{Reference Tompkins and Irene41}

SE can be operated in in situ mode, which enables direct observation of BHJ drying. Including to the notes mentioned earlier for reflectometry measurements, we must add few more comments on the application of SE for drying BHJ layers:

1. (i) Theoretically, all optical models (especially Fresnel) for SE considers planar interface, as explained previously. However, a drying layer hardly provides such interface due to the dynamic motion within the film caused by solvent evaporation. Therefore, it is extremely challenging to model such ellipsometry (also reflectometry) data since all the provided models are insufficient. It is not unlikely that the intensity of the fringes is not matched with the model selected. Hence, it should always be kept in mind that the calculated SE data (or reflectometry data) for drying layers might have greater error range than usual.

2. (ii) Ellipsometry signal can actually provide much more information on the drying behavior of a BHJ thin film than just its intrinsic properties or thickness. One of the most common signal behaviors is the beating effect. It is a phenomenon which occurs when two or more signals with different frequencies are combined. ^{Reference Weik42} It is not unlikely that a drying layer would go through a vertical segregation. Such vertical phase separation creates another interface within the film, which reflects the light differently than the bulk due to different optical properties. Hence, two (or more) light waves with different frequencies induce a beat in the interference pattern, which can then be modeled accordingly. Although, this phenomenon has not been reported for drying BHJ layers so far, we believe that ellipsometry signal can be examined more thoroughly, and such effects can provide greater insight into the dynamics of BHJ drying.

3. (iii) An inhomogeneous medium composed of different materials in structured manner must take into account the collective behavior of the particles within a medium and their effect on each other. Hence, such media can be modeled with effective medium approximation (EMA). ^{Reference Hinrichs and Eichhorn43} A layer or an interface can be defined as an effective material if the particles within the medium are smaller than the incident wave length, and if the particles are randomly distributed. Widely used is Bruggeman theory, which is a good approximation for rough surface or interfaces and for randomly disturbed particles. ^{Reference Hinrichs and Eichhorn43} The idea behind EMA concept is to connect macroscopic permittivity of the medium as whole with the microscopic polarizabilities of the constituents. Hence, it is a relatively good approximation for drying BHJ films and swollen BHJ layers (e.g., during solvent annealing).

C. Photoluminescence

Although photoluminescence (PL) measurements are generally easy to conduct in in situ mode, the processing of in situ data can become very challenging. To be able to utilize PL measurements in in situ mode for BHJ film drying, luminescence theory of organic compounds and relevant quenching mechanisms should be understood well.

Luminescence in general is a phenomenon of light emission from any substance from excited electronic states. ^{Reference Lakowicz44} In quantum mechanics, electronic states with spin 0 and 1 are called singlet and triplet states, respectively. Two types of luminescence can be explained very briefly with the spin selection rules ^{Reference Mikhnenko45} : (i) fluorescence is the emission from excited singlet states, where the electron is paired with ground state in opposite spin. Hence, the return to ground state is spin allowed, and it usually occurs within few nanoseconds. ^{Reference Lakowicz44,Reference Mikhnenko45} (ii) Phosphorescence is the emission from excited triplet states, where the electron has the same spin direction as the ground state. Since the decay from an excited triplet state involves a spin flip, it is spin forbidden; hence the emission lifetimes can take up to seconds. ^{Reference Lakowicz44,Reference Köhler and Bässler46,Reference Yersin and Finkenzeller47}

In an organic compound (small molecule or polymer), these transitions between electronic states occur between highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO). Although there are several ways to excite an electron from HOMO to LUMO (such as electrical current, chemical reaction…), photoexcitation is one of the most common methods, hence luminescence upon photon absorption inherits the term PL. Phosphorescence is generally very weak in organic compounds due to the selection rules explained previously. Additionally, in order for the triplet excitons to be generated, photoexcitation alone is not sufficient. Thus, spin–orbit coupling is necessary for intersystem crossing from singlet to triplet state and radiative and nonradiative transitions from triplet to ground state. ^{Reference Mikhnenko45,Reference Yersin and Finkenzeller47} With all the lightweight atoms organic compounds have, they show only weak spin–orbit coupling. ^{Reference Lakowicz44} For more visible phosphorescence, organic compounds are combined with transition metals, which are triplet emitters, forming organo-transition metal complexes [e.g., Ir(ppy)₃]. ^{Reference Yersin and Finkenzeller47} It should be noted that, BHJs are organic molecules without any transition metals, hence the term ‘PL’ is used in this review to indicate fluorescence emission.

PL quenching is a phenomenon of reduction in PL intensity of a compound due to a certain mechanism. There are several PL quenching mechanisms for fluorescent conjugated systems: intra-chain defects, impurities/dopants, photochemical oxidation, concentration quenching, aggregation quenching leading to interchain species (e.g., excimers, exciplexes, polaron pairs), as well as an efficient photo-induced charge transfer to a quencher (e.g., fullerene or nonfullerene acceptors). The observed PL quenching in a material system is generally a result of few mechanisms occurring simultaneously. Since drying of BHJ systems is a transition from solution to solid state, some of these mechanisms would dominate the quenching process:

1. (i) Concentration quenching: in theory and practice, it was shown that PL decay kinetics is a function of concentration. ^{Reference Lakowicz44,Reference Nguyen, Martini, Liu and Schwartz48} At sufficiently diluted concentrations, the molecules (or polymer chains) do not ‘see’ each other, and the PL signal remains constant during the drying. ^{Reference Güldal, Kassar, Berlinghof, Ameri, Osvet, Pacios, Li Destri, Unruh and Brabec49} Additionally, other quenching mechanisms (impurities/dopants, existence of a second compound acting as a quencher and photochemical oxidation) are also concentration dependent. This concentration dependency can be explained by the Stern–Volmer theory: it is suggested that two of the most common quenching types are dynamic and static quenching. Both quenching types require molecular contact between the fluorophore and quencher. While dynamic quenching [Eq. (10)] ^{Reference Lakowicz44} requires the quencher to be diffused to the fluorophore within the excited state lifetime, static quenching [Eq. (11)] ^{Reference Lakowicz44} is a nonradiative relaxation of a ground state complex formed by fluorophore and quencher:

where I ₀ is the PL intensity without any quencher, I is the PL intensity with quencher, [Q] is the quencher concentration, k _q is the bimolecular quenching constant, τ₀ is the PL lifetime with the absence of quencher, and K _S is the association constant. The bimolecular quenching constant in Eq. (10) can be related to bimolecular rate constant, k ₀, and quenching efficiency, f _Q. The bimolecular rate constant, additionally is a function of diffusion coefficients, D, and sphere radii, R, of fluorophore and quencher:

The collision frequency Z of a fluorophore can then be defined as the function of bimolecular rate constant and the concentration of the quencher:

Concentration quenching in general is a mechanism which is observed at every quencher concentration. However, the relation between concentration [Q], diffusion D, collusion frequency Z, and bimolecular quenching constant k _q presented above is established for sufficiently dilute solutions where free diffusion of molecules is possible. ^{Reference Lakowicz44} This condition is generally not fulfilled when a random BHJ solution prepared specifically for solar cell processing is taken for drying experiments with PL due to high concentration. Therefore, it should be noted that the calculation of sphere radii (also called: sphere of action) becomes cumbersome at high concentrations since PL signal (I) would not be dominated only by one quencher type but by many quenchers, including the self-quenching mechanisms of the fluorophore.

1. (ii) Aggregation quenching: PL quenching occurs additionally by the effect of polymer chain aggregation which leads to high quantum yield of interchain species which tends to relax nonradiatively. ^{Reference Wang, Collison and Rothberg50} This is a well-known self-quenching mechanism of the fluorophore, which is a function of fluorophore concentration. Therefore, during BHJ drying, while the increase in the acceptor concentration induces concentration quenching, the increase in donor concentration would additionally trigger aggregation quenching.

If PL measurements are conducted correctly, aggregation and concentration quenching are the dominant quenching mechanisms for BHJ drying experiments. Care must be taken that the oxidation and laser-induced degradation of the organic molecules do not take place during the PL measurements. This ensures that the recorded PL signal is a direct translation of the molecular self-assembly of the donor and acceptor materials during drying. Although PL does not provide any quantitative result, it is an invaluable measurement method for in situ drying experiments for BHJs, ^{Reference Güldal, Kassar, Berlinghof, Ameri, Osvet, Pacios, Li Destri, Unruh and Brabec49,Reference Engmann, Bokel, Ro, Delongchamp and Richter51} since it is the only measurement method which is very sensitive to the thermodynamic properties of the solution and its components (solubility, miscibility…).

D. Light scattering

Light scattering is the direction and intensity alteration of a light beam upon striking an object due to the combined effects of reflection, refraction, and diffraction. ^{Reference Diebold52} More specifically, scattering occurs when the electron cloud of the material re-emits the energy of the light beam (photon by photon) without altering it otherwise. In other words, due to the polarizability of the molecule, the oscillating dipole emits the electromagnetic wave of the same wave length as the incident wave. Hence, it is called elastic scattering. ^{Reference Schaertl53}

The intensity of scattered light is a function of wave length, particle size, scattering angle, and the relative refractive index of the particles and the medium in which the particles exist. Light scattering is only observable when there is a difference in the refractive indices between the particles and the medium. ^{Reference Diebold52} If this condition is fulfilled, the scattering phenomenon is determined to a great extent by the size of the particle with respect to the wave length. For the particle sizes comparable or larger than the wave length, Mie theory has been developed. The general Mie theory applies to the scattering of monochromatic light by isotropic spherical particles. For particles smaller than λ/20, Rayleigh theory has been developed. ^{Reference Schaertl53} For visible light range, the particle size must be then between 20 and 40 nm, which is mostly violated by solutions containing polymers with good or theta solvents. For large particles, the light can scatter from different parts of the particle. Hence, this might lead to constructive and destructive interferences. These interferences are scattering angle dependent, and they must be corrected by so-called particle form factor. ^{Reference Teraoka54} This correction leads to Zimm equation:

where K is the contrast factor, R _θ is the absolute scattered intensity (also called Rayleigh ratio), M _W is the molecular weight, A ₂ is the second virial coefficient which is a quantitative measure for solvent–solute interactions (originating from van’t Hoff law), ^{Reference Teraoka54} c is the mass concentration, θ is the scattering angle, and 〈s ²〉 is the radius of gyration. When the Rayleigh theory is corrected for large particles, particle size must be smaller than λ/2, instead of λ/20. ^{Reference Schaertl53} Therefore, within the visible range polymer coils or any other scattering substance should be in the range of 200 and 400 nm.

Light scattering has been used widely for the drying experiments of BHJ layers. ^{Reference Güldal, Kassar, Berlinghof, Ameri, Osvet, Pacios, Li Destri, Unruh and Brabec49,Reference van Franeker, Turbiez, Li, Wienk and Janssen55} Due to the measurements being conducted at a single scattering angle, calculations for second virial coefficient and radius of gyration are not possible. However, since light scattering originates from refractive index differences, a type of a phase separation between donor and acceptor, which leads to structures with sizes in the order of the wave length, would induce scattering. Therefore, the light scattering signal during drying generally allows the operator to track the course of phase separation within the wet film.

E. GIWAXS/GISAXS

Grazing incidence small-angle and wide-angle X-ray scattering GISAXS/GIWAXS have become major techniques for the characterization of thin films and nanostructures. ^{Reference Renaud, Lazzari and Leroy56} They can be used for any kind of sample environment providing high depth sensitivity and statistical averaging as opposed to imaging techniques. GISAXS and GIWAXS have already provided an insight into complex multilength-scale structures in BHJ thin films. ^{Reference Chen, Nikiforov and Darling57,Reference Liu, Gu, Shen, Ferdous, Wang and Russell58} By the means of these well-established techniques, structure formation on the nano- and microscale of BHJ thin films can be investigated and analyzed dynamically.

The incident beam in grazing-incident scattering geometry hits the sample surface under very shallow angle, i.e., α_i < 0.5°. Figure 3 shows a schematic drawing of the scattering geometry for GISAXS/GIWAXS with the incident wavevector k _i in the xz plane, the incident and exit angles (α_i and α_f) and the in-plane angle φ (or 2θ_f) of the scattered wavevector k _f projected on the sample plane. The absence of scattering features results in pure specular scattering, which fulfills the law of optics α_i = α_f. ^{Reference Born and Wolf59} The presence of scatterers gives rise to diffuse scattering and diffraction peaks, which carry the desired information about the morphology of the probed film. The scattering intensity, I(q), is measured as a function of the wavevector transfer |q| = |k _f − k _i| = 4π/λ × sin(θ), with λ being the wave length of the incident X-ray beam, and 2θ being the angle between incoming and outgoing wavevectors. The angular coordinates are related to the wavevector transfer coordinates according to Eq. (16):

FIG. 3.

Schematic picture of the geometry of grazing incidence small/wide angle X-ray scattering (GISAXS/GIWAXS) measurements. The scattered signal is collected using a 2D detector.

The scattering signal obtained from thin films by standard diffraction techniques in transmission geometry is very poor due to the small scattering volume. For organic thin films, it is even worse because X-ray scattering cross-section of organic materials is small. However, thin films can be probed with sufficient statistics due to the footprint effect of the grazing incidence conditions. Although the footprint can be confined with microfocusing X-ray optics, ^{Reference Giri, Li, Smilgies, Li, Diao, Lenn, Chiu, Lin, Allen, Reinspach, Mannsfeld, Thoroddsen, Clancy, Bao and Amassian60,Reference Smilgies, Li, Giri, Chou, Diao, Bao and Amassian61} the sampled region is generally in a macroscopic size and thereby produces averaged information. Since the incidence angle is small, it is necessary to consider the effects of refraction and reflection at the interfaces. ^{Reference Sinha, Sirota, Garoff and Stanley62–Reference Yoneda64} The scattering depth is strongly affected by refraction when α_i is close to the critical angle for total external reflection of the surface ${\alpha _{\rm{c}}} = \sqrt {2 \times {\mathop{\rm Re}\nolimits} (1 - n)} \propto \sqrt \rho$ , with n being X-ray refractive index of the material and ρ being the electron density. When α_i ≪ α_c, the scattering depth is limited to few nanometers. It rapidly increases to hundreds of nanometers or even few micrometers when α_i is larger than α_c. ^{Reference Renaud, Lazzari and Leroy56} Therefore, grazing incidence is used to minimize the undesired background and bulk scattering and to enhance the near-surface scattering. The scattering volume depends on the footprint of the beam on the sample and on the penetration depth. An interference phenomenon between the incident and transmitted X-ray beam occurs when α_i,f = α_c, resulting in so-called ‘wave guiding enhancement’ of the scattering intensity, ^{Reference Jiang, Lee, Narayanan, Wang and Sinha65} and the maximum of the diffuse scattering called Yoneda peak. ^{Reference Yoneda64}

The scattering pattern is usually acquired using a 2D detector. Depending on the distance from the sample to the detector, different ranges of q ranges can be probed, and the related experiments can be typically referred as GIWAXS or GISAXS. GIWAXS, also called GIXD for grazing incidence X-ray diffraction, is usually used to probe crystalline structures on the atomic length scale, e.g., the orientation of crystallites, an estimation of crystallite sizes, and the degree of crystallinity. ^{Reference Wu, Jeng, Su, Wei, Su, Chiu, Chen, Su, Su and Su66–Reference Roehling, Baran, Sit, Kassar, Ameri, Unruh, Brabec and Moulé68} GISAXS, on the other hand, probes length scales up to several hundred nanometers. ^{Reference Hexemer and Müller Buschbaum69} In GIWAXS measurements, the incidence angle lies between the critical angles of the film, and a substrate is often selected to obtain information about the crystalline structures in the entire film as well as weak background from the substrate. Bragg peaks appear along the out-of-plane and in-plane directions in the case of a film with an orientation of the lattice planes (or crystals) parallel and perpendicular to the substrate surface, respectively. Usually an arc is observed when the film has an angular distribution of the orientation of crystals.

The positions of peaks for organic conjugated molecules characteristic to the lamellar stacking can be used to determine whether the crystallites are preferentially face-on or edge-on oriented with respect to the substrate. 1D line cut profiles in the out-of-plane direction and in-plane direction are used to obtain peak position and peak width. Peak positions are used to determine the lattice parameters. Full-width-half-maximum (FWHM) of the peak can be utilized to calculate the crystalline correlation length (CCL) via the Scherrer equation if the resolution of the apparatus is accurately accounted for. ^{Reference Smilgies70} The para-crystalline nature of the conjugated polymers contributes to the broadening of Bragg peaks; hence higher order Bragg peaks need to be used in the analysis. ^{Reference Rivnay, Mannsfeld, Miller, Salleo and Toney71} The relative degree of crystallinity (rDoC) can also be calculated. ^{Reference Boudouris, Ho, Jimison, Toney, Salleo and Segalman72,Reference Gomez, Barteau, Wang, Toney and Loo73} Due to the grazing incidence geometry and ill-defined X-ray absorption, it is still hard to obtain the absolute crystallinity from GIWAXS. ^{Reference Liu, Gu, Shen, Ferdous, Wang and Russell58}

GISAXS probes larger length scales, e.g., fullerene aggregates, and provides morphological information such as phase sizes, structural length scales, and intermixing ratios of the components. If the incident angle is above the critical angle of all involved materials, the specular and the Yoneda peak are well separated along the z-axis of the detector allowing for resolving large scale structures. ^{Reference Müller-Buschbaum, Gomez, Nogales, Garcia-Gutierrez and Ezquerra74} To enhance the GISAXS signal and reduce air scattering background, an evacuated flight path is used. The probed depth can be varied by tuning the incident angle; hence the surface-near structures can be separated from the structure in the bulk film. Horizontal line cuts of the 2D GISAXS data are performed at the film’s Yoneda peak to enhance the sensitivity to its structures. ^{Reference Müller Buschbaum75} They can be fitted by a model of polydisperse spheres with a Schultz size distribution to account for the polydispersity, a structure factor of hard spheres and Debye–Buche correlation function to account for the increase in intensity at very low q _y values. ^{Reference Wu, Jeng, Su, Wei, Su, Chiu, Chen, Su, Su and Su66} However, this analysis model is adapted from transmission scattering experiments using Born approximation (BA) which is only valid when α_i and α_f are larger than ∼3αc. The distorted wave born approximation (DWBA) is more suitable at lower angles as it accounts for the contribution of different single scattering processes arising from perturbations of the incident, reflected, and refracted wavefields by the nanostructures. ^{Reference Sinha, Sirota, Garoff and Stanley62,Reference Rauscher, Salditt and Spohn63} The form factor and the shape of nanoparticles can be extracted by modeling the full 2D GISAXS pattern in the framework of the DWBA ^{Reference Renaud, Lazzari and Leroy56} by several software packages: IsGISAXS, ^{Reference Lazzari76} FitGISAXS, ^{Reference Babonneau77} and BornAgain. ⁷⁸

It should be taken care that the incident X-rays do not damage the organic molecules, since they are vulnerable against high intensity irradiation (i.e., hard X-rays). Additionally, care must be taken when BHJ system contains chlorinated solvents. Due to X-ray absorption of the solvent, the probing depth reduces to few hundreds of nanometers. Nevertheless, kinetic investigations making use of time-resolved GISAXS/GIWAXS measurements help toward a better understanding of the processes taking place during the formation of BHJ thin films upon drying. ^{Reference Kassar, Güldal, Berlinghof, Ameri, Kratzer, Schroeder, Destri, Hirsch, Heeney, McCulloch, Brabec and Unruh79}

A. Development of in situ measurement methods with P3HT:PCBM

The very first in situ study on morphology formation during drying was introduced by Schmidt-Hansberg and coworkers in 2009. ^{Reference Schmidt-Hansberg, Klein, Peters, Buss, Pfeifer, Walheim, Colsmann, Lemmer, Scharfer and Schabel91} They have investigated P3HT:PCBM layers from ortho-dichlorobenzene (ODCB) solution, coated from doctor blade under different drying temperatures (15, 25, and 40 °C). Optical reflectometry was utilized with interference theory. The measured drying curves were additionally modeled by self-developed model based on vapor–liquid equilibria and mass transfer theories. ^{Reference Schmidt-Hansberg, Klein, Peters, Buss, Pfeifer, Walheim, Colsmann, Lemmer, Scharfer and Schabel91} Although it has been stated that the drying has shown a constant rate drying with a following falling rate period, it is a very limited information on the governing thermodynamic and kinetic processes during drying. To gain a full mechanistic understanding of microstructure formation, it is necessary to couple few measurement methods.

Wang and coworkers addressed this issue by utilizing ellipsometry and GIXD simultaneously in 2010. ^{Reference Wang, Dunbar, Staniec, Pearson, Hopkinson, MacDonald, Lilliu, Pizzey, Terrill, Donald, Ryan, Jones and Lidzey92} P3HT:PCBM films from chlorobenzene (CB) were doctor bladed at various temperatures (30, 40, and 50 °C). As shown in Figs. 8(a) and 8(b), three stages in film drying have been identified: (i) rapid solvent evaporation: the drying showed constant rate behavior, (ii) moderate solvent evaporation and crystallization: P3HT started to crystallize rapidly [Figs. 8(c) and 8(d)], hence the crystals hindered the diffusion of solvent toward the surface. The drying has become diffusion-dominated. (iii) Slow solvent evaporation and slow crystallization: P3HT crystallization finalized. It is additionally suggested that the nucleation dynamics of P3HT indicated heterogeneous nucleation behavior, as the crystallization growth mechanism was explored using the Avrami equation with Avrami exponent lower than 2, which is indicative for heterogeneous nucleation. ^{Reference Wang, Dunbar, Staniec, Pearson, Hopkinson, MacDonald, Lilliu, Pizzey, Terrill, Donald, Ryan, Jones and Lidzey92}

FIG. 8.

(a) The thickness of a P3HT:PCBM blend film as a function of time after casting determined using spectroscopic ellipsometry at different temperatures. (b) is the re-plot of the data for 30 °C after 10th second of drying. (c) is the out-of-plane cross sections X-ray intensity of P3HT [100] peak. (d) is the crystallinity data determined from (c). Reproduced from Ref. Reference Wang, Dunbar, Staniec, Pearson, Hopkinson, MacDonald, Lilliu, Pizzey, Terrill, Donald, Ryan, Jones and Lidzey92 with permission from RSC.

Schmidt-Hansberg and coworkers followed up with a very similar study on P3HT:PCBM with an in situ setup based on reflectometry and GIXD in 2011. ^{Reference Sanyal, Schmidt-Hansberg, Klein, Colsmann, Munuera, Vorobiev, Lemmer, Schabel, Dosch and Barrena93} P3HT:PCBM layers were doctor bladed from ODCB-based solution at various temperatures (10, 25, 40, and 80 °C). It has been shown that at all drying temperatures, P3HT (h00) diffraction peaks are formed along out-of-plane directions, which is indicative for edge-on orientation [Fig. 9(a)]. The drying at lower temperatures; however, led to more random distribution of P3HT crystallites (i.e., higher degree of mosaicity) but greater π–π stacking [Fig. 9(b)]. It was additionally proposed that the poor molecular kinetics at low drying temperature hindered the phase separation of P3HT and PCBM, which was confirmed by AFM measurements [Fig. 9(c)] indicating smaller domains for the sample processed at low temperature after washing PCBM away with octanedithiol (ODT). Therefore, it was advised that lowering substrate temperature can be a simple route to optimize device efficiency.

FIG. 9.

(a) 2D diffraction pattern of P3HT:PCBM dried at 10, 25, 40 and 80 °C. (b) Mosaicity of P3HT plotted with respect to drying time. (c) Atomic force microscopy topography images of P3HT:PCBM dried at 10 and 40 °C. The images on the left and right are the samples without and with ODT treatment, respectively. Reproduced from Ref. Reference Sanyal, Schmidt-Hansberg, Klein, Colsmann, Munuera, Vorobiev, Lemmer, Schabel, Dosch and Barrena93 with permission from Wiley VCH.

While the temperature was found to affect the orientation of P3HT crystallites, the effect of PCBM content on the P3HT orientation has been additionally studied by Schmidt-Hansberg and co-workers with in situ GIXD. ^{Reference Sanyal, Schmidt-Hansberg, Klein, Munuera, Vorobiev, Colsmann, Scharfer, Lemmer, Schabel, Dosch and Barrena94} P3HT:PCBM layers with 2:1, 1:0.8, and 1:2 weight ratios were doctor bladed from ODCB-based solutions. The crystallization process of P3HT was estimated by quantifying the integrated (100) Bragg peak of P3HT as solvent evaporates [Fig. 10(a)], and the orientation distribution of P3HT crystallites (i.e., mosaicity) was additionally plotted as a function of time [Fig. 10(b)]. It has been found that with increasing amount of PCBM, larger portion of edge-on P3HT crystallites were formed during drying. Furthermore, high PCBM content hindered the development of P3HT π–π stacking, as the characteristic peak (010) was absent for the sample with 1:2 blend ratio. This work has introduced the microscopic understanding of the film formation mechanism under the influence of blend ratio, which is vital for the further development of efficient OSCs.

FIG. 10.

(a) 2D X-ray diffraction images of P3HT:PCBM drying plotted at certain time segments for samples with 2:1, 1:0.8 and 1:2 ratios. (b) Evolution of total integrated intensity of P3HT (100) Bragg peak (top) and mosaicity of P3HT crystals (bottom) which was calculated from the azimuthal integration of the first-order Bragg peak shown in inset. Reproduced from Ref. Reference Sanyal, Schmidt-Hansberg, Klein, Munuera, Vorobiev, Colsmann, Scharfer, Lemmer, Schabel, Dosch and Barrena94 with permission from ACS.

The effect of PCBM content on the P3HT crystallization behavior has been additionally investigated by Chou and coworkers. ^{Reference Chou, Yan, Li, Li, Zhao, Anjum, Alvarez, Gassaway, Biocca, Thoroddsen, Hexemer and Amassian95} The solvent was selected as CB instead of ODCB. The spin coated samples were probed with GIXD/GISAXS, reflectometry and ellipsometry. The PCBM content has been varied in the range of 0–75 wt%. The drying was extremely rapid, and the crystallization and phase separation have occurred simultaneously and instantly. As the highest solar cell performance was observed for the blend with 37.5 wt% PCBM content, this blend was also showing the most phase separation. It was observed that the crystallization kinetics of P3HT has slowed down with an increasing content of PCBM, as PCBM hindered the aggregation process. Nevertheless, the mosaicity of the P3HT crystallites was lower under the presence of high PCBM content, which was also observed by Schmidt-Hansberg and coworkers, as explained previously. ^{Reference Sanyal, Schmidt-Hansberg, Klein, Munuera, Vorobiev, Colsmann, Scharfer, Lemmer, Schabel, Dosch and Barrena94,Reference Chou, Yan, Li, Li, Zhao, Anjum, Alvarez, Gassaway, Biocca, Thoroddsen, Hexemer and Amassian95}

To design better performing materials and to optimize fabrication methods, the desire to gain greater insight into the dominant mechanisms for self-assembly during film formation led the research toward establishing phase diagrams from in situ measurements. Although it was only presented on P3HT:PCBM, Schmidt-Hansberg and coworkers investigated the thermodynamics of molecular ordering during film drying and provided a method to establish ternary phase diagrams based on structural changes during crystallization. ^{Reference Schmidt-Hansberg, Sanyal, Klein, Pfaff, Schnabel, Jaiser, Vorobiev, Müller, Colsmann, Scharfer, Gerthsen, Lemmer, Barrena and Schabel96} For that, binary phase diagrams of P3HT and PCBM with the solvent (DCB) was measured with the help of a refractometer at different temperatures. It was found that the P3HT solution in DCB was stable for concentrations below 3 wt% at 25 °C [Fig. 11(a)], and the solubility limit was reached at a mass fraction of 5.1 wt%. While the solution has started to experience bimodal demixing, polymer chains have started to aggregate at the mass fraction of 10.2 wt%, leading to partially gel state. PCBM, on the other hand, experienced a single solubility limit without any gel state, and its solubility limit has increased stronger than P3HT with increasing temperature [Fig. 11(b)]. The crystallization behavior of P3HT:PCBM layer casted from DCB at 25 °C on PEDOT:PSS was observed in situ with GIXD and compared with the superposition of the binary phase diagrams. The solid mass fraction from the drying experiments were calculated from reflectometry measurements. The evolution of GIXD profiles during drying indicate that the P3HT (100) Bragg peak has appeared at around 156th second of the drying, which refers to a mass fraction of 4.6 wt%. On the other hand, it was rather impossible to recognize the time point for the initiation of PCBM diffraction due to the overlap of PEDOT:PSS diffraction with the PCBM. Hence, another 25 °C measurement was conducted on a Si wafer [Fig. 11(c)]. The P3HT (100) Bragg peak appeared at a mass fraction of 14 wt%, while the PCBM diffraction peak started to form at 46 wt%, which is much lower than the measured value presented on the binary phase diagram. As a conclusion, it was suggested that P3HT has a nucleation range between 4.6 and 14 wt%, which was in agreement with the binary phase diagram presented. However, the suppressed aggregation of PCBM has suggested that there are strong interactions between P3HT and PCBM [Fig. 11(d)], which prevail over fullerene–fullerene interactions, thus hindering PCBM aggregation. ^{Reference Schmidt-Hansberg, Sanyal, Klein, Pfaff, Schnabel, Jaiser, Vorobiev, Müller, Colsmann, Scharfer, Gerthsen, Lemmer, Barrena and Schabel96}

FIG. 11.

(a) Binary phase diagram of P3HT solution in ODCB. (b) Binary phase diagram of PCBM solution in ODCB. (c) GIXD profile plots at increasing drying times indicating the emergence of the diffraction peaks of PCBM, P3HT (h00) Bragg peaks. (d) Scheme shows the stages of molecular arrangement during solvent evaporation with nucleation at the substrate or air interface. Purple squares are crystalline P3HT, and dots are PCBM molecules. Reproduced from Ref. Reference Schmidt-Hansberg, Sanyal, Klein, Pfaff, Schnabel, Jaiser, Vorobiev, Müller, Colsmann, Scharfer, Gerthsen, Lemmer, Barrena and Schabel96 with permission from ACS.

B. Additive effects on microstructure formation investigated by in situ methods

To achieve the most optimized BHJ active layer morphology, several methods have been utilized, such as solution optimization, thermal annealing and solvent annealing. ^{Reference Diao, Shaw and Mannsfeld20,Reference Engmann, Ro, Herzing, Snyder, Richter, Geraghty and Jones97,Reference Huang, Kramer, Heeger and Bazan98} Within all these methods, employing low volatile solvents, i.e., additives, has become a widely accepted microstructure optimization method. ^{Reference Liao, Ho, Chang, Jao, Darling and Su17,Reference Perez, Chou, Love, Van Der Poll, Smilgies, Nguyen, Kramer, Amassian and Bazan25} In spite of the ex situ characterization data providing the information on how additives influence final BHJ structure, little was known regarding the dynamics of the morphology formation mechanism under the presence of an additive due to the lack of proper characterization methods. Hence, the development of in situ techniques has allowed the deeper investigation of additive effect on morphology formation.

One of the very first in situ studies regarding the additive effect on a polymer:fullerene blend system was introduced by Rogers and coworkers in 2012. ^{Reference Rogers, Schmidt, Toney, Bazan and Kramer99} They have investigated the crystallization evolution of the blend of poly{2,1,3-benzothiadiazole-4,7-diyl[4,4-bis(2-ethylhexyl)-4H-cyclopenta[2,1-b:3,4b′]dithio-phene-2,6-diyl]} (PCPDTBT) and PC₇₀BM, where 3 vol% ODT to the chlorobenzene (CB) solution was reported to increase the efficiency significantly. ^{Reference Peet, Kim, Coates, Ma, Moses, Heeger and Bazan100} It has been observed that the film coated from neat CB solution showed an amorphous, isotropic scattering pattern which has not changed during the drying. On the other hand, the film coated from CB + ODT solution has shown a (100) Bragg peak immediately after coating, and the scattering intensity has increased during the drying. The CCL was estimated as 9 nm, which stayed constant over the course of the drying. Hence, it was suggested that the ODT has introduced additional nucleation processes rather than further growth of the existing crystallites. All in all, this study was the first additive-related study, which suggested that the functioning mechanism of additives on BHJ blends might be related to the early aggregation of the donor material and prolonged reorganization time within the layer. ^{Reference Rogers, Schmidt, Toney, Bazan and Kramer99}

Additives, such as ODT or 1,8-diiodooctane (DIO), are accepted as selective solvents for fullerenes in the organic photovoltaic community. ^{Reference Liao, Ho, Chang, Jao, Darling and Su17} On the other hand, additives, such as 1-chloronapthalene (CN) or DIO, are also capable of dissolving both polymer and the fullerene. ^{Reference Machui, Maisch, Burgués-Ceballos, Langner, Krantz, Ameri and Brabec101,Reference Güldal, Berlinghof, Kassar, Du, Jiao, Meyer, Ameri, Osvet, Li, Li-Destri, Fink, Ade, Unruh and Brabec102} Since the kinetics of microstructure formation is determined by the thermodynamics of the coated solution during drying, the quality of the selected additive regarding the component solubility is an attractive aspect to investigate, especially with in situ characterization methods. A very detailed study was presented by Shin and coworkers on P3HT:PCBM blend with different amounts of ODT (as poor solvent for P3HT) and CN (as good solvent for P3HT). ^{Reference Shin, Richter, Herzing, Kline and DeLongchamp29} The polymer order and the changes in the thickness and morphology were tracked with spectroscopic ellipsometry. The drying curves for all the samples indicate the aggregation onset (i.e., solidification onset) and final film formation (i.e., solidification end) [Fig. 12(a)]. It has been observed that the rate of CB drying was similar for all samples, independent of the amount of additive used. However, the aggregation onset has shifted to earlier times with increasing amount of both additives, indicating that both additives have promoted early crystallization of P3HT, which was promoted by supersaturation. The mechanism behind this behavior was attributed to the additives lowering the host solvent quality in terms of solubility. Nevertheless, CN being a partially good solvent for P3HT, the ordering of the polymer has continued even after CB removal, leading to a prolonged reorganization in the system.

FIG. 12.

(a) The thickness versus time data plotted on a log–log scale for P3HT:PCBM films processed without additive and with different amounts of ODT and CN, as indicated accordingly. Graphical summary of film solidification for P3HT:PCBM films studied from solutions with (b) CB only, (c) CB + CN and (d) CB + ODT. Reproduced from Refs. Reference Shin, Richter, Herzing, Kline and DeLongchamp29 and Reference Richter, Delongchamp, Bokel, Engmann, Chou, Amassian, Schaible and Hexemer103 with permission from Wiley VCH.

A follow up work on doctor bladed P3HT:PCBM in CB with ODT or CN was presented by Richter and coworkers with in situ GIXD/GISAXS and reflectometry. ^{Reference Richter, Delongchamp, Bokel, Engmann, Chou, Amassian, Schaible and Hexemer103} As explained earlier, the layer with neat CB has dried quickly with a very rapid crystallization of P3HT and the formation of a mixed phase of amorphous P3HT and PCBM. Under the presence of CN [Fig. 12(b)], early aggregation of polymer has taken place, and the crystallization has followed a similar pattern to that of additive-free layer. It was suggested that the layer has adapted into a two phase structure after the drying of host solvent: P3HT crystal and a CN mixture of P3HT and PCBM. Hence, CN regulated the rest of the crystallization and ordering in the layer during its drying. With ODT, however, most of the P3HT crystallization has happened during CB drying, triggered by early aggregation. In this case, PCBM has been removed from the P3HT-rich phase by ODT, which has improved the phase separation compared to additive-free layer. ^{Reference Richter, Delongchamp, Bokel, Engmann, Chou, Amassian, Schaible and Hexemer103}

The dynamics and interplay of phase separation and polymer aggregation under the presence of additives were additionally investigated by van Franeker and coworkers with a different blend system than P3HT:PCBM. ^{Reference van Franeker, Turbiez, Li, Wienk and Janssen55} They have taken diketopyrrolopyrrole-quinquethiophene copolymer (DPP5T) with PC₇₀BM in chloroform (CF) solution and added different amounts of ODCB or DIO as additives. The evolution of microstructure formation has been measured by absorption, reflectometry and light scattering. It has been observed that DPP5T:PC₇₀BM in CF has formed big fullerene droplets [Fig. 13(b)], which has introduced a high light scattering signal before any aggregation was observed at the absorption plot [Fig. 13(a)]. Hence, the nature of the phase separation was labeled as liquid–liquid phase separation i.e., spinodal demixing, which preceded polymer aggregation and influenced the cell performance detrimentally. However, if an additive such as ODCB or DIO were used as bad solvents for the polymer, DPP5T was then forced into aggregation under supersaturation conditions, which hindered the liquid–liquid phase separation by either forming gel aggregates or precipitates.

FIG. 13.

(a) Detecting liquid–liquid phase separation with time. Film thickness, normalized interference and light scattering of DPP5T:PC₇₀BM in solvent mixtures is monitored during spin coating. The solvent is chloroform (blue squares) and chloroform with 5 vol% ODCB (red circles). Green arrows schematically indicate the thickness reconstruction from the measured interference signal. The dashed lines indicate the onset of scattering. (b) Bright-field TEM images of DPP5T:PC₇₀BM blends spin coated from chloroform and chloroform + 3 vol% ODCB solutions. Reproduced from Ref. Reference van Franeker, Turbiez, Li, Wienk and Janssen55 with permission from Nature Publishing Group.

Especially with the help of in situ characterization studies, it has been widely accepted that there are two guidelines followed when selecting an effective additive: (i) the additives should be non-solvents or worse solvents than host solvents for one component (generally the donor material since the solubility of fullerenes are generally very high in such solvents), and (ii) the additive should have lower volatility than the host solvent, ensuring a more controlled microstructure formation. Although CN is an exception among all additives since it provides good solubility for both components, the presented in situ studies so far indicate that the working principle of all additives follows the same manner: additives reduce the quality of the host solvents by accelerating the polymer aggregation/crystallization, hence they additionally prolong the crystallization procedure, and the total drying time would be increased due to the low vapor pressure. However, the interplay between component solubility, solid solutions and microstructure formation kinetics in the presence of additives was not well understood.

To investigate the effect of component solubility in additives on microstructure formation, Güldal and coworkers presented an in situ study on a blend system with poly-thieno[3,2b]thiophene-diketopyrrolopyrrole-co-thiophene (DPP-TT-T) and PCBM. ^{Reference Güldal, Berlinghof, Kassar, Du, Jiao, Meyer, Ameri, Osvet, Li, Li-Destri, Fink, Ade, Unruh and Brabec102} In this study DPP-TT-T:PCBM system was utilized with the additive DIO as a better solvent for both components than the host solvent, o-xylene:mesitylene mixture, and an alternative morphology formation mechanism has been introduced. The drying kinetics and solution thermodynamics have been observed by reflectometry, light scattering and PL simultaneously on doctor bladed solutions with different amounts of DIO, where PL has been specifically used for tracking molecular self-assembly. It has been observed that without any DIO, the blend has gone through a polymer gelation due to o-xylene and then aggregation/crystallization. ^{Reference Güldal, Kassar, Berlinghof, Ameri, Osvet, Pacios, Li Destri, Unruh and Brabec49} However, when DIO was added (Fig. 14), it has led the evolution of morphology formation toward thermodynamic equilibrium (i.e., all kinetic mechanisms were suppressed), and a solid solution was formed. It was observed that the kinetic mechanisms (e.g., crystallization) have taken place during DIO drying only. It was confirmed that host solvents do not play a major role on morphology evolution, and the component solubility is an important factor playing role on the formation mechanism of solid solutions and microstructure.

FIG. 14.

Time evolution of DPP-TT-T:PCBM drying was measured by WLR (a), LS (b) & (d), PL (c) & (e). Figures (b) and (c) are additionally ‘zoomed out’ (i.e., plotted over longer time scale), which are presented at figures (d) and (e), respectively. Increasing amount of DIO hinders PCBM aggregation (i.e., lower LS signals) and the self-quenching mechanisms of DPP-TT-T (i.e., higher PL signals right after host solvent drying). Reproduced from Ref. Reference Güldal, Berlinghof, Kassar, Du, Jiao, Meyer, Ameri, Osvet, Li, Li-Destri, Fink, Ade, Unruh and Brabec102 with permission from RSC.

C. Up-scaled BHJ layers investigated by in situ characterization methods

No matter how extensive studies are being conducted on lab scale BHJ layers with in situ and ex situ methods, the transfer of the coating and drying information from lab scale processing methods, such as spin coating and doctor blading, to industrial scale processing methods, such as slot-die coating and roll-to-roll coating (R2R), is not straightforward. Although some processes are similar to each other, industrial scale printing requires higher amount of solutions to be printed on wider and longer substrates under very different coating and drying conditions. Hence, the need of dynamic microstructure formation analysis has emerged not only for lab scale BHJ layers but also for upscaled BHJ layers.

In 2014, Rossander and coworkers were the first group installing a miniature R2R coater on a SAXS unit (Fig. 15) and observing the crystal formation of P3HT:PCBM layers in situ. ^{Reference Rossander, Zawacka, Dam, Krebs and Andreasen104} P3HT:PCBM solution in CB was injected in a coating head and printed on a PET foil at 70 °C. SAXS measurements were conducted at two different distances between coating head and illumination point, leading to two different drying times (14 and 23 s). It was observed that the P3HT (100) Bragg peak appeared at out-of-plane direction, indicating edge-on oriented crystals [Fig. 15(a)]. When the in situ SAXS results have been compared with an ex situ measurement of the dried film after 1 h, the intensity of the randomly distributed crystals (i.e., the ring) has increased, indicating a residual solvent being trapped in the film causing further ordering [Fig. 15(c)]. However, the ring formation has occurred very slowly and much after than the formation of edge-on crystals [Fig. 15(b)]. Hence, it was concluded that edge-on oriented crystals started to build up at interfaces (i.e., heterogeneous nucleation) while randomly distributed crystal formation was controlled by supersaturation, which contradicts with the results presented for lab-scale P3HT:PCBM formation, indicating that the governing nucleation and crystal formation mechanism for oriented and randomly distributed crystals is different for R2R coating compared to lab scale processing methods.

FIG. 15.

Schematic of the SAXS roll coater (left), with 3 examples of measurements on the right with different distances from the coating head, corresponding to 14 s (a), 23 s (b) and 1 h (c) drying time, respectively. The 1 h experiment was recorded ex situ by rewinding back the coated film after 1 h. Reproduced from Ref. Reference Rossander, Zawacka, Dam, Krebs and Andreasen104 with permission from AIP Publishing.

Liu and coworkers, on the other hand, presented a mini version of a slot-die coater installed on a GIXD/GISAXS stage by observing the solvent mixture effect on the crystallization and domain formation of a blend mixture composed of low band gap conjugated copolymer of diketopyrrolopyrrole and quaterthiophene (DPPBT) and PC₇₁BM. ^{Reference Liu, Ferdous, Schaible, Hexemer, Church, Ding, Wang and Russell105} The crystallization was controlled with a solvent mixture of CF and ODCB, where ODCB was taken as a non-solvent for DPPBT. In situ GIXD has revealed that with the increasing amount of non-solvent in the solution, the ordering of DPPBT has occurred more and more in solution phase, and faster aggregate formation was observed. In situ GISAXS has indicated that for low non-solvent concentrations, the phase separation was seen before crystallization. With more ODCB in the solution, the onset of the crystallization and phase separation initiation were in sync. It was, however, expected that the phase separation has occurred at earlier time scales due to the high amount of non-solvent separating PC₇₁BM from DPPBT more readily.

D. Morphology formation of BHJs based on polymers with aggregation extremities

Although most conjugated polymers for OPV are semi crystalline in nature, some polymers are either amorphous or highly crystalline. Such polymers are of interest for improving the performance of OPV as well. Therefore, within the last years such polymers were also subjects of in situ investigations.

One of the most well-known and widely used amorphous polymers in organic solar cells is poly[N-9′-heptadecanyl-2,7-carbazole-alt-5,5-(4′,7′-di-2-thienyl-2′,1′,3′-benzothiadiazole)] (PCDTBT), which was once exhibiting the highest efficiency with 6.1% and internal quantum efficiency up to 100%. ^{Reference Park, Roy, Beaupre, Cho, Coates, Moon, Moses, Leclerc, Lee and Heeger106} It was additionally reported that amorphous polymers generally need high loading of acceptor material due to the miscibility, which is not the case for the semi crystalline polymers. Hence, PCDTBT was especially interesting material to investigate in situ in combination with PC₇₁BM to distinguish the differences in the microstructure formation mechanism in an amorphous polymer:fullerene blend. Nevertheless, it is challenging to measure a blend based on an amorphous polymer, given the fact that most of the in situ studies utilize GIXD. Pearson and coworkers, on the other hand, have utilized GIXD and ellipsometry, with which they observed the microstructure formation of doctor-bladed PCDTBT:PC₇₀BM films. ^{Reference Pearson, Wang, Dunbar, Yi, Watters, Coles, Staniec, Iraqi, Jones and Lidzey107} With GIXD, it has been observed that PCDTBT underwent a very weak lamella ordering at the very end of the drying while fullerene has aggregated right after its solubility limit has been reached. However, unlike what has been reported so far on semi crystalline systems, fullerene aggregation has preceded the weak polymer arrangement, and the polymer chains had to arrange themselves around the fullerene aggregates.

In contrast to PCDTBT, one of the most efficient polymers reported to date is a highly aggregating polymer, which shows a strong temperature-dependent aggregation ^{Reference Liu, Zhao, Li, Mu, Ma, Hu, Jiang, Lin, Ade and Yan108} : poly[(5,6-difluoro-2,1,3-benzothiadiazol-4,7-diyl)-alt-(3,3‴-di(2-octyldodecyl)-2,2′;5′,2″;5″,2‴-quarter thiophen-5,5-diyl)] (PffBT4T-2OD). Although PffBT4T-2OD and similar variants’ advantage is the strong tendency to aggregate resulting in robust morphologies which lead to high performances, temperature dependency pushes the processing conditions toward high temperatures (>90 °C). Hence, the adaptation of the process from spin coating to doctor blading (or to slot-die coating) becomes challenging. Nevertheless, this transition was investigated in situ by Wook Ro and coworkers very recently. ^{Reference Ro, Downing, Engmann, Herzing, DeLongchamp, Richter, Mukherjee, Ade, Abdelsamie, Amassian, Jagadamma, Yan and Liu23} PffBT4T-2OD:PC₇₁BM solutions were doctor bladed and spin coated at 90–100 °C from CB:ODCB-based solutions with DIO as additive. The in situ measurements were conducted with reflectometry and UV-Vis absorbance at 700 nm which is indicative of aggregation. The thickness has evolved in two stages: primary solvent drying and DIO drying (Fig. 16). The aggregation has occurred very late in the host solvent drying stage and was completed until DIO drying phase. On the other hand, spun-coat film has shown a much earlier aggregation onset (Fig. 16), which was explained with the cooling effects on the sample by convection reducing the aggregation onset. It was suggested that the PffBT4T-2OD has the ability to have multiple morphologies leading to similar performances, independent from the coating method.

FIG. 16.

Film thickness (black) and absorbance at 700 nm (red) from in situ UV-Vis absorbance studies during spin-coating and blade-coating. Increase in red-shifted absorbance at 700 nm is indicative of aggregation. Reproduced from Ref. Reference Ro, Downing, Engmann, Herzing, DeLongchamp, Richter, Mukherjee, Ade, Abdelsamie, Amassian, Jagadamma, Yan and Liu23 with permission from RSC.