Polymer Photovoltaic Cells: Enhanced Efficiencies via a Network of Internal Donor-Acceptor Heterojunctions

G. Yu,* J. Gao, J. C. Hummelen, F. Wudl, A. J. Heeger

The carrier collection efficiency (c) and energy conversion efficiency (e) of polymer photovoltaic cells were improved by blending of the semiconducting polymer with C60 or its functionalized derivatives. Composite films of poly(2-methoxy-5-(2-ethyl-hexyloxy)1,4-phenylene vinylene) (MEH-PPV) and fullerenes exhibit c of about 29 percent of electrons per photon and e of about 2.9 percent, efficiencies that are better by more than two orders of magnitude than those that have been achieved with devices made with pure MEH-PPV. The efficient charge separation results from photoinduced electron transfer from the MEH-PPV (as donor) to C60 (as acceptor); the high collection efficiency results from a bicontinuous network of internal donor-acceptor heterojunctions.

The need to develop inexpensive renewable

energy sources continues to stimulate new approaches to production of efficient, lowcost photovoltaic devices. Although inorganic semiconductors (silicon, amorphous silicon, gallium arsenide, and sulfide salts) have been the primary focus, the photosensitivity and the photovoltaic effects in devices made with organic materials have also been explored, including conjugated polymers (1), organic molecules (2), stacked discostic liquid crystals (3), and selfassembling organic semiconductors (4). Because of the advantages that would be realized with polymer-based photovoltaics (such as low-cost fabrication in large sizes and in desired shapes), efficient plastic solar cells would have a major impact. Energy conversion efficiencies of photovoltaic cells made with pure conjugated polymers were typically 10-3 to 10-2%, (1), too low to be used in applications. The discovery of photoinduced electron transfer in composites of conducting polymers as donors and buckminsterfullerene (C60) and its derivatives as acceptors (5) provided a molecular approach to high-efficiency photovoltaic conversion (6, 7). Because the time scale for photoinduced charge transfer is subpicosecond, more than 103 times faster than the radiative or nonradiative decay of photoexcitations (5), the quantum efficiency of charge separation from donor to acceptor is close to unity. Thus, photoinduced charge transfer across a donor-acceptor (D-A) interface provides an effective way to overcome early time-carrier recombination in organic systems and thus to enhance their optoelectronic response. For example, with the addition of only 1% C60, the photoconductivity increases by an order of magnitude over that of pure MEH-PPV (6, 7). Although the quantum efficiency for

Institute for Polymers and Organic Solids, University of California at Santa Barbara, Santa Barbara, CA 931065090, USA. *Present address: UNIAX Corporation, 6780 Cortona Drive, Santa Barbara, CA 93117-3022, USA.

photoinduced charge separation is near unity for a D-A pair, the conversion efficiency in a bilayer heterojunction device is limited (8). Efficient charge separation occurs only at the D-A interface; thus, photoexcitations created far from the D-A junction recombine before diffusing to the heterojunction. Even if charges are separated at the D-A interface, the conversion efficiency is limited by the carrier collection efficiency. Consequently, interpenetrating phaseseparated D-A network composites (see Fig. 1) would appear to be ideal photovoltaic materials (9-11). Through control of the morphology of the phase separation into an interpenetrating network, one can achieve a large interfacial area within a bulk material. Because any point in the composite is within a few nanometers of a D-A interface, such a composite is a bulk D-A heterojunction material. Because of the interfacial potential barrier, as demonstrated by the built-in potential in the bilayer D-A heterojunction diode (8), ultrafast photoinduced charge transfer and charge separation will occur with quantum efficiency approaching unity, leaving holes in the donor phase and electrons in the acceptor phase. This process is illustrated in the upper portion of Fig. 1. If the network is bicontinuous, the collection efficiency can, in principle, be equally efficient. Such a bicontinuous D-A network material is promising for use in thin-film solar cells. In addition to the high quantum efficiency of charge separation, the electronic structure of such a bicontinuous D-A network allows one to choose contact electrodes with work functions that optimize the carrier collection efficiencies of holes from the donor phase and electrons from the acceptor phase. Thus, thin-film sandwich devices with bicontinuous D-A composites as the active material promise to function as efficient solar cells with high c and e. Here, we report efficient photovoltaic cells made with such MEH-PPV:C60 composites (see Fig. 1).
SCIENCE VOL. 270 15 DECEMBER 1995

The device structure consisted of a metal (Ca or Al) contact on the surface of a blend film on a glass (or mylar) substrate, coated with transparent indium-tin-oxide (ITO), as shown in the bottom portion of Fig. 1. The active area of the test devices was 0.1 cm2 (12). Large photocells (>15 cm2) with similar efficiencies have also been fabricated. MEH-PPV:C60 films were spin-cast from 0.3 to 0.5 weight % xylene solutions prepared by mixing of two master solutions in proper ratios. The MEH-PPV (13) was obtained from UNIAX Corporation (Santa Barbara, California) and the C60 from MER Corporation (Tucson, Arizona). Typical film thicknesses were 1000 to 2000 . The Al or Ca electrode was vacuum evaporated to a thickness between 1000 and 5000 . The limited solubility of pure C60 in organic solvents and its tendency to crystallize during film formation limit its use in highconcentration blends. To overcome these problems, a series of soluble C60 derivatives has been developed (14). The molecular structures of two C60 derivatives, denoted as [6,6]PCBM and [5,6]PCBM, are shown in the top portion of Fig. 1. With these soluble derivatives we have been able to make homogeneous stable blends containing more than 80 weight % fullerene molecules. Figure 2 compares the current-voltage (I-V) characteristics of a Ca/MEH-PPV:

Fig. 1. Schematic diagram of the photoinduced charge transfer process in MEH-PPV:C60 D-A blends. The structures of the two soluble C60 derivatives used in this study (denoted as [6,6]PCBM and [5,6]PCBM) are included. When cast as a film, the D and A species phase-separate into a bicontinuous network (bulk heterojunction material), as shown schematically. The structure of the photovoltaic cell fabricated with this bulk heterojunction material is sketched at the bottom.

1789

[6,6]PCBM/ITO device with 1:1 weight ratio (Fig. 2A) and a Ca/MEH-PPV/ITO device (Fig. 2B) in the dark and then illuminated with 20 mW/cm2 at 430 nm. For pure MEH-PPV devices, the sharp current turnon and the open circuit voltage Voc (the minimum in the I-V under photoexcitation) at 1.6 V are consistent with the workfunction difference between Ca and ITO. The similar I-V characteristics of the devices shown in Fig. 2 imply similar device physics. In the Ca/MEH-PPV:[6,6]PCBM/ ITO device, the exponential current turn-on shifts to lower voltage by ~0.8 V. Thus, the lowest unoccupied molecular orbital of [6,6]PCBM is ~0.8 eV below the Fermi energy of Ca. Capacitance measurements of the blend devices yield values similar to those of pure MEH-PPV devices (independent of bias), implying that the energy gap remains clean and trap-free. No indication of ground-state charge transfer was observed. The forward I-V characteristics of the MEH-PPV devices (Fig. 2B) exhibit three regions: a small shunt current (<10-10 A/cm2) for V < 1.2 V, an exponential increase by more than four orders of magnitude between 1.3 and 1.8 V, and current saturation at higher voltages (the current increases, but at a less than exponential rate). In reverse bias, the current saturates at approximately 10-10 mA/cm2 for |V| < 3 V. The I-V characteristics at high fields have been interpreted in terms of tunneling (15) into a thin-film semiconductor deplet-

ed of carriers. Capacitance-voltage and alternating current conductivity experiments provide evidence in support of this model (16). The exponential current turn-on and the Voc under strong illumination correspond to the flat band condition. Because photoinduced electrons in MEH-PPV will lower their energy by transferring to the C60, and photoinduced holes in C60 will lower their energy by transferring to the MEH-PPV, the photosensitivity (PS) is substantially enhanced in the phaseseparated composite (as shown in Fig. 2). The short-circuit current (Isc) in the MEHPPV:[6,6]PCBM device is Isc = 0.5 mA/cm2; under 20 mW/cm2, corresponding to PS = 25 mA/W and c = 7.4% electrons per photon (e/ph), both of which are approximately two orders of magnitude higher than those of MEH-PPV devices (Fig. 2B). The electroluminescence quantum efficiency of the blend device was ~3 10-7% photons per electron, 103 to 104 times less than in pure MEH-PPV devices, which is consistent with ultrafast photoinduced charge separation (5). The carrier collection and energy conversion efficiencies of the bicontinuous DA network material are critically dependent on the networks morphology and chemical composition. By changing the solvent from xylene to 1,2-dichlorobenzene we have been able to cast high-quality MEHPPV:[6,6]PCBM films with methanofullerene compositions up to 1:4 weight

Fig. 2. (A) I-V characteristics of a Ca/MEH-PPV: [6,6]PCBM/ITO device in the dark (open circles) and under 20 mW/cm2 of illumination at 430 nm (solid circles). (B) The corresponding data from a Ca/MEH-PPV/ITO device. Currents were plotted as absolute values.

Fig. 3. c (A) and e (B) of Ca/MEH-PPV: [6,6]PCBM(1:4)/ITO (solid squares); Ca/MEHPPV:[6,6]PCBM(1:1)/ITO (open squares); Al/ MEH-PPV:[6,6]PCBM(1:1)/ITO (diamonds); Ca/ MEH-PPV:[5,6]PCBM(1:1)/ITO (open circles); Ca/ MEH-PPV:C60(3:1)/ITO (triangles); and Ca/MEHPPV/ITO (solid circles).

ratio (approximately one acceptor for every polymer repeat unit). For devices made from this blend, Isc = 2 mA/cm2, PS = 100 mA/W, and c 29% e/ph under 20 mW/ cm2 at 430 nm. Figure 3 shows c (Fig. 3A) and e (Fig. 3B) as a function of illumination intensity for several blend devices and for pure MEHPPV devices. The blend weight ratios are 3:1 for the MEH-PPV:C60 device, and 1:1 and 1:4 for the devices using the functionalized fullerenes. For Ca/MEH-PPV:[6,6]PCBM (1:4)/ITO devices, the photosensitivity is slightly sublinear with light intensity (I 0.94): c = 29%, e/ph and e = 2.9% at 20 mW/cm2; c = 45% e/ph and e = 3.2% at 10 W/cm2 (17). In reverse bias, these devices are excellent photodetectors; for example, at -2 V, PS > 200 mA/W and c > 60% e/ph at 20 mW/cm2, a substantial improvement over earlier results and better than ultraviolet-enhanced silicon photodiodes at 430 nm (7). The spectral response of these photovoltaic devices is similar to that obtained from photoconductive cells made from MEHPPV:C60 blends (6). The photovoltaic response turns on at approximately 1.5 eV and is relatively flat for photon energy >2.5 eV. The PS and the c are nearly the same when Al is substituted for Ca as the cathode, although Voc decreases slightly from 0.82 V to 0.68 V at 20 mW/cm2. Typical data from Al/MEH-PPV:[6,6]PCBM/ITO devices are also included in Fig. 3. Thus, stable metals with work functions as high as 4.3 eV can be used as the cathode electrode to collect electrons effectively from the acceptor phase. Ideally, the work function of the anode metal should be close to the donor highest occupied molecular orbital (HOMO) and the work function of the cathode should be close to the acceptor lowest unoccupied molecular orbital (LUMO). An ohmic contact would then form selectively at each interface, being ohmic for holes at the interface between the donor and the highwork-function metal contact but blocking for electrons at the same interface, and vice versa. The built-in potential would be the difference between the acceptor LUMO and the donor HOMO. The bulk D-A heterojunction material provides an automatic selection of electrons from the acceptor component and holes from the donor component. The use of Ca (or Al) as the cathode and ITO as the anode creates a large internal field (for a given film thickncss, the internal field is larger with Ca). These two contact materials automatically extract electrons from C60 and holes from MEH-PPV. Even though the lowwork-function metal is in direct contact with the donor, holes will not be extracted at this interface, because the internal field forces the holes toward the

1790

SCIENCE

VOL. 270 15 DECEMBER 1995

high-work-function contact. For the same reason, electrons are extracted from C60 at the Ca/MEH-PPV:C60 interface. The result, then, is that separated carriers are not wasted; they are automatically collected by the proper electrode so that external work can be done. The substantial enhancement in c achieved with the bicontinuous D-A network material results from the large increase in the interfacial area over that in a D-A bilayer and from the relatively short distance from any point in the polymer to a chargeseparating interface. Moreover, the internal D-A junctions inhibit carrier recombination and thereby improve the lifetime of the photoinduced carriers (6), so that the separated charge carriers can be efficiently collected by the built-in field from the asymmetric electrodes. Similar effects have been observed in MEH-PPV:Cyano-PPV polymer blends (10, 11). The device efficiencies are not yet optimized. Because only ~60% of the incident power was absorbed at 430 nm in the thinfilm devices used for obtaining the data in Fig. 3, the internal carrier collection efficiency and energy conversion efficiency are approximately 1.7 times larger; that is, c 90% e/ph and e 5.5% at 10 W/cm2. Although nearly 100% absorption can be achieved by using thicker films, c is currently limited in thick-film devices by internal resistive losses. Further improvements in device efficiencies are expected when the blend composition and the network morphology are optimized.
REFERENCES AND NOTES
1. For a review of photovoltaic devices made with polyacetylene, see J. Kanicki in Handbook of Conducting Polymers, T. A. Skotheim, Ed. (Dekker, New York, 1986), pp. 543-660; S. Glenis, G. Tourillon, F. Garnier, Thin Solid Films 111, 93 (1984); S. Karg, W. Riess, V. Dyakonov, M. Schwoerer, Synth. Metals 54, 427 (1993); H. Antoniadis, B. R. Hsieh, M. A. Abkowitz, S. A. Jenekhe, M. Stolka, ibid. 64, 265 (1994); R. N. Marks, J. J. M. Halls, D. D. D. C. Bradley, R. H. Frield, A. B. Holmes, J. Phys. Condens. Matter 6, 1379 (1994); G. Yu, C. Zhang, A. J. Heeger, Appl. Phys. Lett. 64, 1540 (1994). 2. For an early review, see G. A. Chamberlain, Solar Cells 8, 47 (1983); C. W. Tang, Appl. Phys. Lett. 48, 183 (1986); B. Miller et al., J. Am. Chem. Soc. 113, 6291 (1991); C. H. Lee, G. Yu, D. Moses, A. J. Heeger, Appl. Phys. Lett. 65, 664 (1994). 3. D. Adam et al., Nature 371, 141 (1994). 4. For examples, see B. A. Gregg, M. A. Fox, A. J. Bard, J. Phys. Chem. 94, 1586 (1990); C. Y. Liu, H. L. Pan, H. Tang, M. A. Fox, A. J. Bard, ibid. 99, 7632 (1995). 5. N. S. Sariciftci, L. Smilowitz, A. J. Heeger, F. Wudl, Science 258, 1474 (1992); N. S. Sariciftci and A. J. Heeger, Intern. J. Mod. Phys. B 8, 237 (1994). 6. C. H. Lee et al., Phys. Rev. B. 48, 15425 (1993). 7. G. Yu, K. Pakbaz, A. J. Heeger, Appl. Phys. Lett. 64, 3422 (1994). 8. N. S. Sariciftci et al., ibid. 62, 585 (1993). 9. N. S. Sariciftci and A. J. Heeger, U.S. Patent 5,331,183 (1994); U.S. Patent 5,454,880 (1995). 10. G. Yu and A. J. Heeger, J. Appl. Phys. 78, 4510 (1995). 11. J. J. M. Halls et al., Nature 376, 498 (1995). 12. Electrical measurements were performed with a

13.

14. 15. 16. 17.

Keithley 236 Source-Measure Unit. The excitation source was a tungsten-halogen lamp with a bandpass filter (centered at 430 nm, with a bandwidth of 100 nm). The maximum optical power at the sample was ~20 mW/cm2. F. Wudl, P.-M. Allemand, G. Srdanov, Z. Ni, D. McBranch, in Materials for Nonlinear Optics Chemical Perspectives, S. R. Marder, J. E. Sohn, G. D. Stucky, Eds. (American Chemical Society, Washington, DC, 1991), pp. 683-686. J. C. Hummelen, B. W. Knight, F. Lepec, F. Wudl, J. Org. Chem. 60, 532 (1995). I. D. Parker, J. Appl. Phys. 75, 1656 (1994). I. H. Campbell, D. L. Smith, J. P. Ferraris, Appl. Phys. Lett. 66, 3030 (1995). e was calculated from the relation e = FFIscVoc/Pin using the following definition of filling factor:

[see (8)]. In engineering applications, another definition of FF is sometimes used FF = ImVm/lscVoc where Im and V m are current and voltage for maximum power output. The FF and the e following this definition are approximately half of the values shown in the text. 18. The authors are grateful to N. S. Sariciftci and C. H. Lee for many valuable discussions. Supported by the Department of Energy under a grant from the Advanced Energy Projects Program (DOE-FG0393ER12138). 20 June 1995; accepted 25 October 1995

SCIENCE

VOL. 270

15 DECEMBER 1995

1791