2.1. Development of the in situ GIWAXS technique

The small-angle X-ray scattering (SAXS) beamline (BL16B1) at the SSRF [15], which mainly engages in the application of X-ray scattering [16], provides perfect conditions for the development of the GIWAXS method. Figure 1 shows the developed GIWAXS setup for BL16B1. PILATUS3 X 2M was employed as a 2D area detector, and the pixel size was 172 μm × 172 μm, the number of pixels was 1475 × 1679, and the maximum frame rate was 250 Hz. The high frame rate guaranteed the ability to collect data during time-resolved measurements. The incident angle could be adjusted by the sample stage, which comprised four Kohzu stages, from 0° to 2°. The four Kohzu stages control the height and shift in the YZ plane, rotation on the y-axis, and rotation on the x-axis. The glove box, as shown in Fig. 1, provided a gaseous atmosphere for the samples and reduced background scattering. The sample temperature was controlled with a homemade temperature controller using a thermocouple that was thermally connected to the top of the sample stage. The temperature range was 25–200°C.

Fig. 1Full-screen View

(Color online) The grazing incidence wide-angle X-ray scattering setup at BL16B1

2.2. GIWAXS and GISAXS measurements

For GIWAXS and GISAXS measurements, samples were fabricated by spin-coating the respective solutions of active layers onto 1.5 × 1.5 cm² silicon substrates with pre-coated PEDOT: PSS (30 nm). Silicon substrates were cleaned in sequence with detergent; deionized water; acetone; and isopropyl alcohol in an ultrasonic bath for 15 min, dried by ultra-pure N₂, and then treated with O₂ plasma for 8 min. The GIWAXS and GISAXS measurements were carried out using beamline BL16B1 at the SSRF. The beam size was 0.4 × 0.5 mm², and the photon flux at the sample was 1 × 10¹¹ phs/s. The X-ray wavelength was 0.124 nm (E = 10 keV), and the incident angle was set at 0.15°, which was close to the critical angle, to maximize the intensity. The sample-to-detector distance was 266.2 and 1997.6 mm for GIWAXS and GISAXS measurements, respectively, which was calibrated with the standard sample of silver behenate. For the static measurements, the exposure time of individual measurements was 60 s. For in situ measurements, samples were heated with a hotplate under an air atmosphere. Single images were collected every 60 s to obtain a high signal-to-noise ratio and high sampling rate. For each set of experiments, the pixel coordinates of the images were converted into q values after correction of the beam center by standard samples. One-dimensional profiles in the out-of-plane direction were obtained by integrating at the azimuthal angle around 90°, while the profiles in the in-plane direction were extracted with an azimuthal angle of approximately 0°.

2.3. Materials

All reagents and solvents were obtained from commercial sources and used directly without further purification. N, N-Dioxide of bis(N,N’-dimethylaminopropyl)perylene diimide (PDINO) was purchased from Solarmer Materials Inc. (Beijing, China). Poly(3,4 ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) (PH1000) was obtained from Heraeus Inc. PTB7-Th was purchased from One-Material Corp. The PC₇₁BM, chlorobenzene, and 1,8-diiodooctane (DIO) used in the device fabrication processes were obtained from Sigma Aldrich.

2.4. Device fabrication

The OSC devices were fabricated with the configuration ITO/PEDOT:PSS/active layer/PDINO/Al. ITO substrates were cleaned with detergent, deionized water, acetone, and isopropyl alcohol in an ultrasonic bath sequentially for 15 min, then dried by ultra-pure N₂, and finally treated with O₂ plasma for 8 min. Then, a 30 nm PEDOT:PSS layer was spin-coated onto the ITO-coated glass at 4,000 rpm and baked at 150°C for 20 min in an oven. The PTB7-Th/PC₇₁BM solution was prepared by dissolving blends with a weight ratio of 1:1.5 in chlorobenzene (total concentration 20 mg/mL, with 0, 1, 3, and 8 vol% DIO additive). When the substrates were transferred into a glove box, the solutions were stirred overnight and spin-coated on the PEDOT:PSS layer to form active layers (the optimal thickness was approximately 100 nm). Subsequently, a PDINO solution (in CH₃OH, 1 mg/mL) was spin-coated at 3,000 rpm for 15 s to form the electron transfer layer. Finally, the Al (100 nm) metal electrode was thermally evaporated under approximately 4 × 10⁻⁴ Pa, and the device area was 0.1 cm², defined by a shadow mask.

2.5. Annealing process

To perform in situ investigations on device PCE and active layer morphology, an annealing strategy that involved placing devices on the hot plate at room temperature (25°C) and then heating them to the annealing temperature (100°C) at a heating rate of 20°C/min was employed to simulate the conventional thermal annealing strategy that involved directly placing devices on a hot plate that had been heated to a specified temperature. Hence, it would remain for 3.75 min in the heating stage process from 25°C to 100°C, 10 min in the temperature holding process at 100°C, and 3.75 min in the cooling process from 100°C to 25°C.

3.1. DIO addition effects on device PCE and thermal stability

To demonstrate the effects of using DIO on the PTB7-Th/PC₇₁BM device, devices with various ratios of DIO ranging from 0 to 8 vol% were fabricated with a conventional device structure of ITO/PEDOT:PSS/active layer/PDINO/Al, as shown in Fig. 2. The J-V curves are shown in Fig. 3(a), and the detailed photovoltaic parameters for devices with various ratios of added DIO are shown in Fig. 3(b) and Table 1. As shown in Fig. 3(a), a drastic change in the J-V curve of the PTB7-Th/PC₇₁BM device was induced by the addition of DIO. The J-V curve for the device without DIO showed an unfavorable shape with a low short-circuit current density (J_SC) and fill factor (FF) compared with the values for devices with added DIO. While the values of open circuit voltage (V_OC) for the device without the addition of DIO were basically the same as the value for devices with different ratios of DIO, a trend of decreasing with the increase in DIO was observed. As shown in Fig. 3(b), the PCE was only 5.49% when no DIO was added. As the ratio of DIO increased, the value of PCE first increased to 8.48% for the device with 1 vol% DIO, and a maximum value of 9.17% was achieved for devices with 3 vol% DIO. However, the PCE value decreased to 8.36% when the DIO ratio increased to 8 v%. These results demonstrate that the appropriate addition of DIO can improve the PCE of PTB7-Th/PC₇₁BM devices, while the excessive addition of DIO could harm the device performance. A similar phenomenon in PTB7/PC₇₁BM was reported by Kim et al. [17].

Fig. 2Full-screen View

(Color online) Device configuration and the molecular structures of PTB7-Th and PC₇₁BM

Fig. 3Full-screen View

(Color online) (a) J-V curves and (b) photovoltaic parameters of organic solar cell devices based on PTB7-Th/PC₇₁BM devices with various ratios of 1,8-diiodoctane ranging from 0 to 8 vol %. The average parameters were calculated from 20 independent devices. DIO, 1,8-diiodoctane; V_OC, open circuit voltage; J_SC, low short-circuit current density; FF, fill factor; PCE, power conversion efficiency.

As shown in Fig. 3(b), for devices with no added DIO, the values of V_OC, J_SC, and FF were 0.83 V, 15.37 mA/cm², and 43.37%, respectively. After adding DIO, the value of V_OC showed a trend of decreasing with an increase in DIO proportion, where the median value of 0.82 V was obtained from the device with 3 vol % DIO added. For J_SC and FF, the values significantly increased after the addition of DIO, and maximum values of 17.77 mA/cm² and 63.33% were achieved, respectively, when the ratio of DIO was 3 vol %, leading to the optimum PCE. However, when the ratio of DIO was increased to 8 vol %, both the FF and J_SC values decreased, resulting in a decrease in PCE. The increase in J_SC and FF appeared to be the main reason for the improvement in PCE induced by the addition of DIO. For OSCs, V_OC is generally influenced by the morphology of the active layer, especially large-scale phase separation [18]. J_SC and FF are usually dominated by the generation, diffusion, and dissociation of photogenerated excitons and the transport and collection of charge, which are strongly dependent on the donor/acceptor interface and charge transport channel in the active layer [19]. Therefore, the photovoltaic parameters changed with various ratios of DIO, suggesting that there was a morphological evolution in the active layer.

Typical PTB7-Th/PC₇₁BM devices suffer severe PCE degradation even after short-term thermal annealing at temperatures ranging from 70°C to 150°C [20]. The addition of DIO can further harm the thermal stability of OSC devices [21]. However, the mechanism by which DIO can harm the thermal stability of PTB7-Th/PC₇₁BM devices is not yet completely understood. Figures 4(a) and 4(b) show the J-V curves and normalized photovoltaic parameters obtained after 10 min of thermal annealing at 100°C for PTB7-Th/PC₇₁BM blends with various ratios of added DIO. As shown in Fig. 4(a), the J-V curve after thermal annealing for the device without DIO showed an insignificant change compared to that of the J-V curve before annealing (Fig. 3(a)). However, drastic changes were observed for devices with various proportions of DIO, where the changes were primarily induced by the decrease in J_SC and FF. As shown in Fig. 4(b), after thermal annealing, the residual J_SC, FF, and PCE values were 99.3%, 92.4%, and 90.4%, respectively, of their initial values for devices without DIO. However, only 95.1%, 82.9%, and 77.6% of J_SC, FF, and PCE, respectively, were retained for the devices with 1 vol% DIO. For devices with 3 vol% DIO, the J_SC, FF, and PCE values were 91.4%, 83.2%, and 75.8%, respectively. When the proportion of DIO was further increased to 8 vol%, the residual J_SC, FF, and PCE values decreased to 88.7%, 82.1%, and 71.4%, respectively. The degradation in J_SC, FF, and PCE increased after the addition of DIO. In addition, these three parameters showed a decreasing trend as the proportion of DIO increased. However, the value of residual V_OC was approximately unchanged with the increase in DIO, which demonstrated that the effect of DIO on V_OC could be negligible and no large-scale phase separation was induced during thermal annealing. The photovoltaic parameters after thermal annealing are summarized in Table S1. These results demonstrate the negative effect of DIO addition on the thermal stability of the PTB7-Th/PC₇₁BM devices.

Fig.4Full-screen View

(Color online) (a) J-V curves and (b) normalized photovoltaic parameters of organic solar cell devices based on PTB7-Th/PC₇₁BM devices with various ratios of 1,8-diiodoctane ranging from 0 to 8 vol % after 10 min of thermal annealing at 100°C. The average parameters were calculated from 20 independent devices. DIO, 1,8-diiodoctane; V_OC, open circuit voltage; J_SC, low short-circuit current density; FF, fill factor; PCE, power conversion efficiency.

3.2. Effects of Addition of DIO on the morphology of the active layer to improve device performance

To further explore the mechanism by which addition of various ratios of DIO improves device performance, GIWAXS measurements were carried out to monitor the evolution of the active layer morphology after adding various ratios of DIO. The 2D GIWAXS patterns of the blends with various ratios of DIO are shown in Fig. S1. Figure 5 (a) and (b) show the 1D profiles integrated from the 2D patterns in the out-of-plane and in-plane directions, respectively. For the typical PTB7-Th/PC₇₁BM blends with 3 vol% DIO, the polymer alkyl stacking peaks showed an almost ring-like scattering pattern at q ≈ 2.90 nm^-1 with an associated lattice spacing of d ≈ 2.17 nm, which corresponds to the (100) Bragg reflection and is consistent with that of previous results [22]. The almost ring-like scattering pattern of PC₇₁BM was observed at q ≈ 13.60 nm^-1, which demonstrated that the PC₇₁BM molecules were in an amorphous state. A weak π-π stacking peak, which could hardly be distinguished from the ring of PC₇₁BM, occurred in the out-of-plane direction, implying that the PTB7-Th lamellae had a preference for face-on orientation. The lamellar structure was likely to form nanofibrils under common processing conditions, which could be inclined relative to the substrate, causing the peaks to have an angle spread in the scattering patterns.

Fig. 5Full-screen View

(Color online) 1D profiles integrated from 2D patterns in the (a) out-of-plane and (b) in-plane directions. Crystal coherence length and d-spacing values of (c) (100) peaks and (d) [6,6]-phenyl C₇₁-butyric acid methyl ester rings of blends with various ratios of 1,8-diiodoctane. CCL, crystal coherence length; DIO, 1,8-diiodoctane.

For the (100) peak of PTB7-Th shown in Fig. 5(a) and (b), a weak peak occurred in the blend film without DIO owing to the poor crystallinity of PTB7-Th [23]. The peak intensity increased with decreasing peak width in both the out-of-plane and in-plane directions as the ratio of DIO increased. These results suggested that the crystallization of PTB7-Th could be enhanced by increasing the proportion of DIO. The changes in the peak position and full width at half maximum (FWHM) were also observed after adding various proportions of DIO, indicating that DIO could influence the lattice space (d-spacing) and crystal coherence length (CCL) of PTB7-Th. For the peak of PC₇₁BM, slight changes in FWHM were also observed for blends with various ratios of DIO, while the peak position was approximately constant.

The d-spacing and CCL values of the (100) peaks are shown in Figure 5(c). The detailed values integrated from the out-of-plane and in-plane directions are summarized in Table 2 and S2, respectively. The CCL values were determined using the Scherre equation, as shown in Eq.(1) [24]:

where B is the FWHM of the peaks, K is a constant with a value of 0.89, γ is the X-ray wavelength, and θ is the Bragg diffraction angle. For the data processing process, the software Fit2D [25] and SG-Tool [26] were used to subtract background scattering and convert 2D patterns into 1D profiles, and the software Peakfit [27] was employed to fit the peak position and FWHM from 1D profiles. The detailed data processing process is shown in the Supporting Information (Sect. 2). The original experimental profiles with background and (100) profiles from Fig. 5(a) are shown in Fig. S2 and S3, respectively.

As shown in Fig. 5(c), the d-spacing increased as the proportion of DIO increased, which suggested that the packing distance between the lamellae was enlarged. The CCL also increased with increasing ratio of DIO, indicating that a larger crystal size was obtained when adding more DIO. When the ratio of DIO was 0 vol%, the CCL value of PTB7-Th was approximately 3.98 nm. The values of CCL for blends with various ratios of DIO were significantly larger (5.10, 5.71, and 6.15 nm when adding 1, 3, and 8 vol% DIO, respectively). The values of CCL obtained from the out-of-plane and in-plane directions were approximately the same regardless of the amount of DIO added, indicating that the orientation of the PTB7-Th crystal was not affected by DIO.

The d-spacing and CCL values for PC₇₁BM are shown in Fig. 5(d). The value of d-spacing for PC₇₁BM remained approximately unchanged at 0.46 nm with various ratios of DIO. The CCL value first decreased and then increased as the proportion of DIO increased, where the CCL values were 2.89 nm, 2.84 nm, 2.77 nm and 2.83 nm. This phenomenon indicated that adding an appropriate proportion of DIO decreases the size of the PC₇₁BM phase, while excessive addition of DIO increases it.

To explore the effect of adding DIO on the aggregation of the amorphous phase in the active layer, GISAXS analyses were performed for blends with 0 and 3 vol% DIO to monitor the large-scale morphological information in the active layer. Figure 6(a) and 6(b) show the 2D GISAXS patterns for blends with 0 and 3 vol% DIO. The 1D GISAXS profiles are shown in Fig. 6(c), which were integrated from the 2D patterns in the in-plane direction at the value of q_z near the reflected position to obtain enhanced signals. The fitting parameters obtained from GISAXS data were fitted with the Debye-Anderson-Brumberger term [28] in the low q region (0.025–0.5 nm⁻¹) as shown in Table 3. The fitting equation is shown in Eq. (2):

Fig. 6Full-screen View

(Color online) 2D Grazing incidence small angle X-ray scattering patterns for blends with (a) 0 and (b) 3 vol % 1,8-diiodoctane. (c) The 1D profiles integrated from 2D patterns and the fitting curves. DIO, 1,8-diiodoctane.

In polymer fullerene systems, the first term is used to model the scattering of dispersive PC₇₁BM in the amorphous domain, and the second term represents the contribution of fractal-like aggregations of PC₇₁BM, where q is the scattering wave vector, A1 and A2 are independent fitting parameters, and ξ is traditionally used to estimate the size of the domain where the acceptor is dispersed in polymer, which can be easily identified from the “shoulder” occurring in the 1D GISAXS profile. In the second term, P(q, R) is the form factor modeled as spheres, and S(q, R, η, D) is the structure factor of the cluster domain modeled as a fractal-like network expressed in Eq. (3):

where R is the mean spherical radius of the primary particles, η is the correlation length of the fractal-like network, D is the fractal dimension of the structure, and η is the correlation length of the fractal-like network. 2R_g-polymer and 2R_g-PCBM are the Guinier radii employed to approximately characterize the average domain sizes of the aggregated cluster phases of PTB7-Th and PC₇₁BM, which are calculated according to Eq. (4) and (5):

As shown in Table 3, the values of ξ were approximately constant for blends with 0 and 3 vol% DIO, while the CCL values of blends without DIO were apparently smaller than those of blends with 3 vol% DIO, as shown in Fig. 5(c). This indicated that the addition of DIO had no effect on the size of the PTB7-Th cluster, but both d spacing and crystal size were enhanced with the addition of DIO. The stronger aggregation of PC₇₁BM molecules was obtained in blends without DIO, which was demonstrated by the significantly larger η and 2R_g-PCBM values shown in Table 3. However, there were more discrete PC₇₁BM molecules in the active layer with 3 vol% DIO, in which a smaller 2R_g-polymer value of the aggregation size of PTB7-Th was obtained (Table 3). Therefore, the addition of DIO weakened the phase separation in the active layer by weakening the aggregation of PC₇₁BM molecules. A homogeneous phase separation scale was obtained, accompanied by smaller clusters of both PC₇₁BM and PTB7-Th. Schematic diagrams of the 3D nanostructures for blends with 0 and 3 vol % DIO were drawn according to the results of GIWAXS and GISAXS measurements, as shown in Fig. 7.

Fig. 7Full-screen View

(Color online). Schematic diagrams of the 3D nanostructures drawn according to the results of grazing incidence wide-angle X-ray scattering and GISAXS measurements for blend films with (a) 0 and (b) 3 vol % 1,8-diiodoctane. DIO, 1,8-diiodoctane; PTB7-Th, poly[(ethylhexyl-thiophenyl)-benzodithiophene-(ethylhexyl)-thienothiophene]; PC₇₁BM, 6,6-Phenyl C₇₁ butyric acid methyl ester.

For the blend without DIO shown in Fig. 7(a), the CCL value of PTB7-Th was significantly lower than that of blends with 3 vol% DIO, according to the GIWAXS results (Fig. 5(c)), resulting in poor charge transport efficiency due to the lack of charge transport channels. The larger D value (2.97) of the blends without DIO suggested that molecules tended to form spherical clusters, which was also harmful to the charge transport. In addition, the maximum values of CCL and 2R_g-PCBM for PC₇₁BM were obtained, indicating that stronger PC₇₁BM aggregation was formed in this case, which resulted in a relatively large phase separation in the active layer. Moreover, a stronger phase separation was also demonstrated by the larger 2R_g-polymer value. Thus, exciton dissociation was harmed because of the lack of a donor/acceptor interface caused by the strong phase separation. Owing to these two factors, devices without the addition of DIO showed significantly lower J_SC and FF values than devices with added DIO, leading to an apparently lower PCE of 5.49% (Fig. 3). After the addition of DIO (Fig. 7(b)), the value of CCL for PTB7-Th increased to 5.70 nm. The value of D decreased to 2.04, which implied that molecules preferred to form planar aggregations accompanied by a large crystal size, which was beneficial to the improvement of charge transport efficiency, hence, increasing J_SC and FF. Simultaneously, the values of CCL for PC₇₁BM, 2R_g-PCBM, and 2R_g-polymer decreased with the addition of DIO, implying the formation of weaker phase separation with smaller aggregation clusters, which was also favorable for the enhancement of exciton dissociation and the improvement of J_SC and FF by forming more donor/acceptor interfaces. Therefore, the PCE of the device can be improved (Fig.3).

3.3. In situ investigation of morphology evolution and correlations with the degradation of thermal stability

In situ GISAXS measurements were performed for PTB7-Th/PC₇₁BM blend films with 0 and 3 vol % DIO to explore the large-scale morphological evolution during the thermal annealing process, as shown in Figs. S4 and S5. No significant change in the 1D curves was observed during thermal annealing of the blend films, regardless of the proportion of DIO added. This indicates that the effect of short-term thermal annealing on the aggregation of amorphous molecules in the active layer was insignificant. The size of the aggregation, size of the fractal network, and fractal dimension were almost unchanged. No large-scale phase separation was induced after thermal annealing, which is consistent with the results in Fig. 4(b), where the values of V_OC remained approximately constant for devices after thermal annealing, regardless of the addition proportion of DIO. Therefore, the degradation of the device PCE was attributed to the small-scale morphological evolution, which was closely related to the J_SC and FF of the devices.

To explore the mechanism of increasing DIO harming device thermal stability, in situ GIWAXS measurements were performed during the thermal annealing process for PTB7-Th/PC₇₁BM blends with various ratios of DIO. To simulate the conventional thermal annealing strategy that directly places devices on a hot plate that has been heated to a specified temperature, an annealing strategy that placed devices on the hot plate at room temperature (25°C), which was then heated to the annealing temperature (100°C) at a heating rate of 20°C/min, was employed. Figure 8 shows the 1D profiles of GIWAXS during the thermal annealing process obtained in the out-of-plane direction for blend films with different ratios of DIO. A similar transition in the (100) peak was observed for all blend films, regardless of the proportion of DIO. Taking the blend film with 3 vol% DIO as an example, as the annealing time increased, the FWHM of the original (100) peak in the out-of-plane direction gradually increased from 0.99 to 1.23 nm⁻¹, and the peak intensity gradually decreased, which led to the gradual disappearance of the original (100) peak. Simultaneously, a newly formed (100) peak occurred at the position with a higher q value (q = 3.04 nm⁻¹) accompanied by a gradual decrease in the FWHM to 0.78 nm⁻¹, and peak intensity gradually increased. The detailed values of the peak position and FWHM are summarized in Table S3–S6. This indicates that PTB7-Th crystallization could be refactored during the temperature holding process, where the original crystallization was destroyed and new crystallization was formed, accompanied by molecular nucleation and crystal growth processes. It can be concluded from Fig. 8 that the transition of the (100) peak became more drastic as the proportion of DIO increased. In addition, as shown in Table S7–S10, it remained approximately unchanged for both the peak width and peak position of PC₇₁BM during thermal annealing, which implies that the effect of short-term thermal annealing on the PC₇₁BM crystal could be ignored.

Fig. 8Full-screen View

(Color online) The 1D profiles of grazing incidence wide-angle X-ray scattering obtained in the out-of-plane direction for blend films with (a) 0, (b) 1, (c) 3, and (d) 8 vol% 1,8-diiodoctane during thermal annealing at 100°C for 10 min, where 0 represents the temperature before annealing (25°C). DIO, 1,8-diiodoctane.

The evolution of the normalized d-spacing integrated from the out-of-plane (100) peak for blend films with various ratios of DIO during thermal annealing at 100°C is shown in Fig. 9(a). The detailed values of d-spacing and CCL after thermal annealing are summarized in Table 4. As shown in Figure 9(a), the same trends occurred, wherein the values of d-spacing increased initially and then decreased for all blend films with various ratios of DIO. The difference was that both the increasing and decreasing rates of d-space increased as the proportion of DIO increased. The maximum values of d-spacing were 102.8%, 102.9%, 103.2%, and 103.4% of their initial values, and the final values were 99.3%, 98.0%, 97.3%, and 96.5% of their initial values for blends with 0, 1, 3, and 8 vol% DIO, respectively. It was indicated that the blend film with more DIO added would suffer more drastic evolution in the d-spacing. Moreover, the time at which d-spacing started to decrease and CCL started to increase was advanced as the proportion of DIO increased, where the time was approximately 2 min earlier for blends with 8 vol% DIO (about 8 min after thermal annealing started) than for the blends without DIO (approximately 10 min after thermal annealing started), which also suggested that the addition of DIO could facilitate the morphological evolution during the annealing process.

Fig. 9Full-screen View

(Color online) Normalized (a) lattice space and (b) crystal coherence length obtained from the (100) peak in the out-of-plane direction for blends with various ratios of 1,8-diiodoctane during thermal annealing at 100°C. “0” represents the value obtained at 25°C. The heating and cooling rate was 20°C/min. d-spacing, lattice space; CCL, crystal coherence length; DIO, 1,8-diiodoctane.

The evolution of the normalized value of CCL integrated from the out-of-plane (100) peak for blend films with various ratios of DIO during thermal annealing at 100°C is shown in Fig. 9(b). The same trends wherein CCL first decreased and then increased were obtained from blend films, regardless of the amount of DIO. Similarly, larger increasing and decreasing rates of CCL were observed as the proportion of DIO increased. The minimum values of CCL were 82.1%, 81.4%, 81.0%, and 79.5%, and the final values were 111.4%, 118.9%, 126.6%, and 137.8% of their initial values for blends with 0, 1, 3, and 8 vol% DIO, respectively. These results indicate that the morphological stability of the blends decreased with an increase in the proportion of DIO, which behaved as a faster morphological evolution by a large margin, as shown in Figure 8 and 9, resulting in more severe degradation of device thermal stability as more DIO was added (Fig. 3).

In the heating stage, as shown in Figure 9, the CCL first decreased, which resulted in the reduction of charge transport channels. Simultaneously, a looser packing with a larger d-spacing value was formed due to the destruction of the original crystal, which allowed the PC₇₁BM molecules to diffuse into the polymer domains to form more interfaces between the donor and acceptor, thereby enhancing exciton dissociation. In the temperature holding stage, the CCL continued to decrease at a relatively slower rate compared to the decreasing rate in the heating stage. Meanwhile, the d-spacing increased at a slower rate. These results demonstrated that the destruction of polymer crystalline phase was still in progress, while the slower evolution slowed the decay rate of charge transport, which may have balanced the effects of weakened charge transport and improved the donor/acceptor interface. Subsequently, as the annealing time increased, the value of CCL increased, accompanied by newly formed charge transport channels, which facilitated the electron mobility and hole mobility in the active layer, thereby enhancing the device PCE and decreasing the recombination loss. However, the interface between the donor and acceptor was weakened owing to the enlarged phase separation due to crystal growth. In the cooling stage, the CCL continued to increase, while the growth rate gradually slowed. However, the final crystal size became too large, resulting in unfavorable phase separation that harmed the photogenerated exciton dissociation, thereby decreasing J_SC and FF.

When the devices were thermally annealed, the varying rates of CCL and d-spacing in all stages increased as the proportion of DIO increased, which suggested that the addition of DIO causes the active layer to become unstable and facilitate the rate of morphological evolution. The blends with more DIO suffered a more drastic morphological evolution, which was apparently unfavorable for the thermal stability of the device. Therefore, the residual PCE gradually decreased as the proportion of DIO increased (Fig. 4).

3.4. Mechanism of analysis of the effect of DIO on the morphology–performance relationship

Figure 10 shows a schematic diagram of the morphological evolution of blends with 0 and 3 vol% DIO during the thermal annealing processes. DIO is a selective solvent additive that can selectively dissolve PC₇₁BM [29]. In addition, the boiling point of DIO (332°C) [30] is significantly higher than that of the solvent chlorobenzene (130.9°C) [31]. As shown in Fig.10(a), when no DIO was added, large-scale sphere PC₇₁BM clusters were formed, leading to a strong phase separation between PTB7-Th and PC₇₁BM, which was demonstrated by the larger values of 2R_g-PCBM and 2R_g-polymer listed in Table 3. When DIO was added (Fig. 10(b)), part of the clustered PC₇₁BM was dissolved into discrete molecules, and the shape of the cluster became planar, accompanied by decreased values of CCL for PC₇₁BM and 2R_g-PCBM (Fig. 5(d) and Table 3). Discrete PC₇₁BM molecules could diffuse into the polymer-rich domains and even the crystal domains of PTB7-Th during the film-forming process, owing to the slower volatilization of DIO, leading to an increase in the d-spacing of PTB7-Th as the ratio of DIO increased (Fig.5(c)). Subsequently, a heterogeneous nucleation and crystal growth process was observed, where the discrete PC₇₁BM molecules had an effect similar to that of the nucleating agent for PTB7-Th. The d-spacing, and crystal size of PTB7-Th increased as the proportion of DIO increased (Fig. 5(c)). When the DIO ratio was increased to 8 vol%, many PC₇₁BM molecules were dissolved in DIO. Owing to the higher boiling point of DIO, chlorobenzene volatilized before DIO during the film-forming process. Afterward, the excessive PC₇₁BM molecules dissolved in DIO formed a new large-scale PC₇₁BM cluster during the volatilization of DIO, leading to an enlarged CCL of PC₇₁BM, as shown in Fig. 5(d).

Fig. 10Full-screen View

(Color online) Schematic diagram of the morphological evolution of blends with 0 and 3 vol% 1,8-diiodoctane during thermal annealing. DIO, 1,8-diiodoctane; PC₇₁BM, [6,6]-phenyl C₇₁-butyric acid methyl ester.

For the morphological evolution during thermal annealing, there were two stages; the crystal destruction process and recrystallization process, which corresponded to the CCL decreasing and increasing stages, respectively, as shown in Fig. 9(b). Organic photovoltaic molecules have long flexible side chains, allowing molecular chains to relax at temperatures below the glass transition temperature (T_g) [32], which is approximately 125°C for PTB7-Th [33]. When the annealing process started, the original crystal was destroyed first because of the strong relaxation of the molecules at this temperature. The addition of DIO can decrease the T_g of PTBT-Th [33]; therefore, a stronger molecular relaxation occurred at the annealing temperature for blends with added DIO, which led to a more significant decrease in CCL in the first stage of thermal annealing (Fig. 9). In the second stage, molecular interaction and movement caused amorphous molecules to nucleate and crystallize again to form new crystals with gradually decreased d-spacing and increased CCL. For blends with DIO, there were many discrete PC₇₁BM molecules distributed in the polymer domain, which could facilitate the nucleation and crystallization of PTB7-Th (Fig. 10(b)). Hence, the crystal growth rate and growth starting time both increased as the proportion of DIO increased (Fig. 9). Therefore, more drastic morphological evolutions induced when the proportion of DIO was increased resulted in morphological instability and subsequently, led to a relatively poor thermal stability compared to that of devices without DIO.