Organic photovoltaics (OPVs) offer the possibility for creating low-cost, lightweight, and flexible renewable energy sources [1]; however, further improvements are required to reach commercial viability. One limitation of OPVs is their low open-circuit voltage (V_oc), which is typically three to four times lower than the optical energy gap of the materials employed [2]. Low short-circuit current (J_sc) is also typically observed due to the tradeoff between the relatively long optical absorption length and the short exciton diffusion length [3]. One means to improve solar cell performance is to control crystalline ordering [4,5]. In past work, for example, we have shown that the excition diffusion length is significantly increased with order [6]. Furthermore, anisotropies native to the structure of many organic crystals can result in control over both the optical absorption and charge transport properties of the resulting film. Hence, considerable work has been focused on controlling crystal structure used in the active region of organic solar cells to result in an improvement of its several operating parameters.

Following the approach of controlling crystalline order in the device active region, OPVs using a thin 3,4,9,10-perylenetetracarboxlic dianhydride (PTCDA) structural templating layer led to a change in orientation of the subsequently deposited polycrystalline copper phthalocyanine (CuPc) donor layer. Here, PTCDA is notable for its tendency to lie flat when deposited on amorphous substrates such as SiO₂, or rough surfaces such as indium tin oxide (ITO) [7,8]. While polycrystalline CuPc grown on ITO results in the upright (100)-α-phase molecular configuration, the presence of PTCDA orients CuPc into a nearly flat-lying configuration that leads to improved π-orbital overlap between molecules, and hence enhanced exciton diffusion and charge transport [3,6]. This advantageous orientation has been known to lead to an increased J_sc [9–11] compared to films deposited in the absence of the template. In one case this led to a decrease in V_oc, leading to an 11% improvement in power efficiency [10], while in other cases an increase in V_oc was observed [11]. In this work, PTCDA is used as a self-organizing template [3,9,12–14] for the growth of subsequent layers. The addition of a diindenoperylene (DIP) layer on the PTCDA serves three purposes: propagating the templating effect of PTCDA, acting as an exciton blocking layer (EBL) [15], and influencing the surface morphology of subsequently deposited films. As a result, we observe an increase in both J_sc and V_oc due to control of molecular crystalline orientation, leading to a concomitant increase in η_p.

Organic thin films were grown on 150 nm thick layers of indium tin oxide (ITO) pre-coated onto glass substrates. Prior to deposition, the ITO/glass substrates were cleaned in surfactant and a series of solvents as previously [16], and then exposed to ultraviolet-ozone for 10 min before loading into a high vacuum chamber (base pressure < 10⁻⁶ Torr). First purified by thermal gradient sublimation in vacuum [17], PTCDA, DIP, CuPc, C₆₀, and bathocuproine (BCP) were then thermally evaporated at 0.2, 0.05, 0.1, 0.15, and 0.1 nm/s, respectively, followed by a 100nm thick Al cathode deposited through a shadow mask with an array of 1 mm diameter openings. For each experiment, CuPc, C₆₀, BCP, and Al were simultaneously grown with and without structural templating layers, the latter for control purposes.

Ultraviolet photoelectron spectroscopy (UPS) was used to measure the film ionization energies relative to vacuum For UPS, the samples were transferred in nitrogen from the growth chamber to an ultrahigh vacuum system (base pressure < 5x10⁻⁹ Torr) where they were illuminated with the He I source. X-ray diffraction (XRD) was performed using a rotating anode Rigaku Cu-Kα diffractometer in the Bragg-Brentano configuration, and atomic force microscope (AFM) images were obtained in the tapping mode. Active region absorption was calculated from measurement of the device reflectivity (R) obtained at 6° (near-normal) incidence after subtracting the loss measured for an ITO/BCP/Al reference. The active layer absorption is then equal to (1 – R). Internal quantum efficiency (IQE) was obtained from the ratio of the external quantum efficiency (EQE) to the absorption. Current density versus voltage (J-V) characteristics were measured in the dark and under simulated AM1.5G solar illumination. The illumination intensity and quantum efficiency measurements were referenced using an NREL-calibrated Si detector [18]. Errors quoted correspond to the standard deviation in values determined by measuring three positions on the same substrate.

Figure 1 shows the XRD patterns for films grown on oxidized Si substrates [19]. A weak diffraction peak at 2θ = 27.5° is observed for a 1.5 nm thick layer of PTCDA, indicating the existence of the flat-lying α-phase (102) orientation. For a 25 nm thick layer of CuPc, the “standing-up” (molecular normal parallel to the substrate plane, as seen in Fig. 1(b)) α-phase (200) orientation is inferred from the peak at 2θ = 6.8°. When a 25 nm thick layer of CuPc is grown on a 1.5 nm thick DIP layer, the CuPc orientation is largely unchanged, whereas, when grown on 1.5 nm thick PTCDA, there is a nearly complete disappearance of the (200) orientation along with the appearance of peaks at 2θ = 26.7° and 27.7°, corresponding to the flat-lying CuPc (312) and ($\overline{3}$13) orientations as shown in Fig. 1(c). When a 25 nm thick CuPc layer is grown on a bilayer of 1.5 nm thick DIP on 1.5 nm PTCDA, we see similar changes in CuPc orientation to that grown directly on PTCDA. These data suggest that by using PTCDA as a templating layer, the orientation of DIP changes from (001) β-phase on glass, to the (020) β-phase on PTCDA [3,20], which in turn orients the CuPc molecules for maximum out-of-substrate-plane conductivity.

 

Fig. 1 (a) X-ray diffraction patterns of PTCDA, CuPc, DIP, and combinations of these layers on Si. The standing-up CuPc (200) orientation (b) disappears when CuPc is grown on a pre-deposited PTCDA template layer. This orientation is then replaced by the (c) flat-lying orientations as evidenced by the appearance of the (312) and ($\overline{3}$13) diffraction peaks.

Download Full Size | PDF

It has also been found that the energies of the frontier orbitals of organic materials are influenced by their crystalline structure. For example, an increase in highest occupied molecular orbital (HOMO) energy was previously reported for CuPc lying flat on highly-oriented pyrolytic graphite [21]. Figure 2a shows the UPS data for PTCDA, CuPc, templated CuPc, and templated DIP, where dashed lines indicate the intercepts. Comparing CuPc (black/circle) to templated CuPc (red/diamond), we measure a shift in the highest energy cutoff from −0.93 ± 0.01 eV to −0.70 ± 0.01 eV (a difference of −0.23 eV) below the Fermi level upon templating. We also see a vacuum level shift of 0.15 ± 0.01 eV, as indicated by the change in low-energy cutoff. Adding these values, we infer that the HOMO energy of CuPc (measured for 5.0 nm thick films deposited on ITO) is increased by −0.08 ± 0.02 eV when a 1.5 nm thick layer of PTCDA is used for templating the CuPc. The relative positions of the HOMO levels for PTCDA, DIP, and CuPc taken from UPS measurements are shown schematically in Fig. 2(b) assuming vacuum level alignment. It is apparent that the PTCDA/CuPc interface acts as a type-II (staggered) heterojunction [22], and DIP can function as an EBL in a type-I (nested) heterojunction with CuPc.

 

Fig. 2 (a) Ultraviolet photoelectron spectroscopy data for 1.5 nm thick PTCDA, 5.0 nm thick CuPc, and 5.0 nm thick templated films of DIP and CuPc on indium tin oxide (ITO). The high energy cutoff of CuPc shifts ~0.2 eV when templated on PTCDA compared to films on ITO. Dashed lines show extrapolations of the data to the energy axis. (b) Energy level diagrams inferred from the measured highest occupied molecular orbital energies CuPc and PTCDA (units of eV). Symbols and colors in (a) correspond to those in (b).

Download Full Size | PDF

Finally, the surface morphology of the CuPc layer changes from a root mean square (RMS) roughness of 1.8 nm when grown directly on ITO (Fig. 3(a) ), to a roughness of 3.9 nm when grown on either PTCDA or DIP (Figs. 3(b) and 3(c)). In the latter two cases, the underlying grain structure of ITO becomes apparent. When crystalline DIP is grown on top of PTCDA, a CuPc roughness of 6.8 nm and an island size of ~100 nm results, as shown in Fig. 3(d). The surface area ratio (compared to a perfectly planar junction) of these morphologies is 1.01, 1.05, 1.03, and 1.12, respectively.

 

Fig. 3 Atomic force microscope images of (a) 25 nm thick CuPc, (b) 1.5 nm thick PTCDA/25 nm thick CuPc, (c) 1.5 nm thick DIP/25 thick nm CuPc, and (d) 1.5 nm thick PTCDA/1.5 nm thick DIP/25 nm CuPc. Lateral spans of each image are 5 μm. The cluster-like morphology of 3(d) suggests a bulk heterojunction interface between CuPc and C₆₀.

Download Full Size | PDF

Device performances under one sun, AM1.5G illumination the structures glass/ITO/(0, 1.5nm) PTCDA/(0, 1.5 nm) DIP/25 nm CuPc/40 nm C₆₀/10 nm BCP/Al are summarized in Table 1 . In this case, the templating layers consist of Device (I) none (control), (II) 1.5 nm DIP, (III) 1.5 nm PTCDA/0 nm DIP, and (IV) 1.5 nm PTCDA/1.5 nm DIP EBL. The efficiency of the control (1.42 ± 0.04%) is similar to recently published data [16,23], although it is below the highest efficiencies reported for this material combination [24]. Device II performs similarly to the un-templated structure, while for Device III, structural templating leads to an increase of 0.06 V in V_oc and a small increase in J_sc, resulting in η_p = 1.76 ± 0.04%. This increase in V_oc is due to the increase in the HOMO energy of CuPc as shown in Fig. 2(b). This is consistent with previous work suggesting that V_oc is proportional to the interface energy gap, which is defined as the difference between the donor HOMO and acceptor lowest unoccupied molecular orbital energy [2,25]. Incorporating both PTCDA and DIP in Device IV shows the same V_oc as for Device III, while J_sc is substantially increased, leading to η_p = 2.19 ± 0.05%. The FF for all devices is ≥0.60, indicating that they have similar diode characteristics and shunt resistances under illumination [26].

Table 1. OPV performance for planar heterojunction (PHJ) and planar-mixed heterojunction (PMHJ) devices under simulated 1 sun, AM1.5G illumination

View Table

Devices incorporating a planar-mixed heterojunction (PMHJ) [27] were also fabricated with the structure glass/ITO/(0, 1.5nm) PTCDA/(0, 1.5 nm) DIP/15 nm CuPc/10 nm CuPc:C₆₀ (1:1)/35 nm C₆₀/10 nm BCP/Al. As shown in Table 1, there is a similar increase in J_sc from 6.2 to 8.1 mA/cm² when incorporating both the PTCDA and DIP layers due to the increase in absorption coefficient due to a more advantageous orientation of the initial CuPc donor region. However, this is accompanied by a decrease in V_oc from 0.50 to 0.48 V. This is due to a previously reported frustration of crystallinity in co-evaporated CuPc:C₆₀ films [16,27]. The resulting amorphous film does not have the preferred stacking, which results in the deeper CuPc HOMO in Fig. 2(b). Nevertheless, the combination of a templating layer and an EBL, the efficiency increases to 2.49 ± 0.03%, or nearly double that of the planar control.

The mechanisms for OPV efficiency enhancement are further understood by comparing the internal (IQE) and external quantum efficiencies (EQE) of the cells. Figure 4(a) shows EQE (symbols) and absorption (lines) for the approximately planar heterojunction (PHJ) devices. For Device III, which employs a PTCDA template, the absorption (corresponding to 1-R) of CuPc at λ = 690 nm is increased from 0.50 to 0.58. This leads to an increase in EQE from 14 to 16% in the same spectral region, accompanied by a decrease in EQE at shorter wavelengths. This decrease could arise from a decrease in the exciton diffusion length due to morphology changes. Integrating across the solar spectrum, Devices I, II, and III have comparable photocurrent, while Device IV is 25% higher. The ratio of the IQE of Device IV to Device III is >1 across the spectrum, shown in Fig. 4(b). The 10% increase in surface area accounts for an increase in IQE across the spectrum (c.f. Figure 3). The additional increase in the spectral region from λ = 550 nm to 750 nm (where CuPc absorbs) is attributed to exciton blocking by DIP. Whereas in Device III, excitons generated in CuPc can quench at the PTCDA/CuPc interface, by incorporating DIP in Device IV the contribution of CuPc to the IQE is increased ~20% according to optical models utilizing the transfer matrix approach [28].

 

Fig. 4 (a) External quantum efficiency (EQE) and absorption measured for Devices I - IV. (b) Ratio of the internal quantum efficiencies (IQE) of Device IV to Device III.

Download Full Size | PDF

In summary, we have demonstrated improved OPV performance resulting from a change in crystalline orientation achieved via structural templating of subsequently deposited layers of DIP and CuPc. Using PTCDA as a crystalline template, the DIP and CuPc molecular stacking were modified from a standing-up to a flat-lying orientation relative to the substrate plane. For CuPc, this leads to improvement in orbital overlap between adjacent molecules, and hence changes in frontier energy levels and absorption coefficient that combine to substantially increase the power conversion efficiency. In addition, DIP propagates the structural templating, changes CuPc film morphology, and serves as an EBL between PTCDA and CuPc. The OPV efficiency increases from 1.42 ± 0.04% to 2.19 ± 0.05% for a PHJ, and from 1.89 ± 0.05% to 2.49 ± 0.03% for a PMHJ by the improved stacking arrangements of CuPc in a CuPc/C₆₀ OPV cell. Our results show the impact of controlling the crystalline morphology and orientation on organic optoelectronic properties.