Band structure engineering in organic semiconductors

See allHide authors and affiliations

Science  17 Jun 2016:
Vol. 352, Issue 6292, pp. 1446-1449
DOI: 10.1126/science.aaf0590

Organic solar cells tuned by blending

Electrical engineers can finetune the energetics of rigid photovoltaics and transistors by blending different semiconducting materials. However, it's hard to apply this tuning protocol to the flexible class of carbon-based semiconductors. Schwarze et al. now show that continuous band energy tuning is indeed possible by varying the blend ratios of certain organic phthalocyanines and their fluorinated or chlorinated derivatives (see the Perspective by Ueno). They demonstrated the effect, which they attribute to quadrupolar interactions, in model solar cells.

Science, this issue p. 1446; see also p. 1395


A key breakthrough in modern electronics was the introduction of band structure engineering, the design of almost arbitrary electronic potential structures by alloying different semiconductors to continuously tune the band gap and band-edge energies. Implementation of this approach in organic semiconductors has been hindered by strong localization of the electronic states in these materials. We show that the influence of so far largely ignored long-range Coulomb interactions provides a workaround. Photoelectron spectroscopy confirms that the ionization energies of crystalline organic semiconductors can be continuously tuned over a wide range by blending them with their halogenated derivatives. Correspondingly, the photovoltaic gap and open-circuit voltage of organic solar cells can be continuously tuned by the blending ratio of these donors.

The working principles of semiconductor devices crucially depend on the electronic energy levels of conduction and valence states. This also holds for organic semiconductors, which have recently gained much attention for their application in flexible large-area devices not realizable with traditional inorganic semiconductors. In organic semiconductor devices such as organic solar cells (OSCs), light-emitting diodes (OLEDs), and field-effect transistors (OFETs), an accurate energy-level tuning of electron- and hole-transporting electronic states—represented by the frontier orbitals, the highest occupied (HOMO) and lowest unoccupied molecular orbital (LUMO)—is necessary to optimize device efficiency (14). For instance, the relative energy position of the HOMO of the donor and the LUMO of the acceptor material is paramount in OSCs, as it controls the open-circuit voltage (1, 5). Also, the tuning of the energy level alignment at the electrode/semiconductor interface is crucial for efficient charge injection in OLEDs and OFETs, and extraction in OSCs (4, 68).

In inorganic semiconductors, band-gap engineering (or more generally, band structure engineering) has opened up a new dimension in device design: By blending materials with different energy levels, it became possible to continuously tune band energies by simply varying the composition of binary, ternary, or quaternary alloys (911).

So far, this design principle has not been available in organic semiconductors. Here, the established means of tuning the electronic levels is by varying the molecular design—for example, by halogenation of organic compounds (1215). However, such changes in molecular design require a new synthesis and do not provide the same degree of fine-scale tunability obtained by the blending of inorganic semiconductors. Blending of organic molecules has been applied to the active layer in organic solar cells (16, 17) and for interface layers in OLEDs (18, 19). However, in those experiments, only the electronic states of individual molecules were relevant. Accordingly, blend properties were not used to achieve continuous tuning.

In this study, we show that band structure engineering is possible in organic semiconductors. Despite the strongly localized electronic states of the individual molecules, long-range Coulomb interactions enable continuous tuning in blends (20, 21). As a model material system, we chose zinc phthalocyanine and boron subphthalocyanine chloride (ZnPc and SubPc; Fig. 1A and fig. S3) and their halogenated derivatives (F4ZnPc, F8ZnPc, F16ZnPc, and Cl6SubPc; structural formulas in figs. S1 and S3) (22). The energetic, optical, and structural properties of phthalocyanines have already been well investigated in the literature, as also motivated by their good device performance (23).

Fig. 1 Molecular structural properties of (halogenated) ZnPc and SubPc.

Isopotential surfaces at –0.3 V (blue) and +0.3 V (red) for (A) ZnPc, (B) F16ZnPc, (C) SubPc, and (D) Cl6SubPc calculated via the DFT method B3LYP/6-311+g(d,p).

We first performed ultraviolet photoelectron spectroscopy (UPS) measurements on ZnPc, F4ZnPc, F8ZnPc, and F16ZnPc thin films to obtain measurements of the ionization energy (IE) of neat layers (fig. S1). The IE increases with higher degrees of fluorination from 5.1 eV for ZnPc to 6.75 eV for F16ZnPc, whereas the shapes of the spectra remain very similar, as reported previously (24). This increase of IE in fluorinated compounds has been explained by a stabilization of the HOMO level in compounds with electron-withdrawing ligands (13). In ordered thin films, large orientation dependencies of the IE were seen for molecules containing polar bonds (25). For highly ordered thin films of pentacene and perfluoropentacene, this was explained by an orientation dependence of the polarization energy originating from different charge-quadrupole interactions (26, 27). By comparing literature values for the IE of ZnPc (6.3 eV) and F16ZnPc (7.2 eV) molecules measured by gas-phase UPS with the IE obtained in this study, we can obtain a first estimation of the relaxation energy contribution to the IE of our solid thin-film samples (14). The large difference of ~0.75 eV is rationalized by their opposing quadrupole moments (Fig. 1 and table S3).

Salzmann et al. (28) showed large shifts in the IE by intermixing of pentacene and perfluoropentacene. However, the IE did not shift linearly with mixing ratio, which was explained by an undesired phase separation in the blend films. In contrast, good intermixing on a molecular scale was shown by x-ray diffraction (XRD) and absorption measurements for a blend of CuPc:F16CuPc (29). As shown by comparison of our XRD results on a ZnPc, an F4ZnPc, and a ZnPc:F4ZnPc (1:1) blend layer (fig. S2) to previous results, the α-phase crystal structure with an edge-on orientation and fine molecular intermixing in blend layers can be expected (24, 29, 30). Crystallinity of mixed systems and large differences in quadrupole moments of single molecules are key preconditions for tuning energy levels over a wide range (see below).

UPS results on mixed ZnPc:FnZnPc layers show a monotonous shift of the HOMO distribution to lower binding energies with increasing ZnPc content, corresponding to a decrease of the IE (Fig. 2, A to C). Whereas the HOMO distributions of neat layers can be fitted by single Gaussian functions, the broader HOMO distributions of the blend layers can only be described by a superposition of two Gaussian peaks (see supplementary materials). The molar ratio of ZnPc and FnZnPc in the blends, obtained from the intensity ratio of fluorine and zinc emission in x-ray photoelectron spectra, was used as the intensity ratio of the peaks.

Fig. 2 Thin-film ionization energies at different mixing ratios.

UPS spectra (gray circles) and fits (solid lines) for mixed blends of (A) ZnPc:F4ZnPc, (B) ZnPc:F8ZnPc, and (C) ZnPc:F16ZnPc. In blends, parts of spectra belonging to ZnPc (blue) and FnZnPc (red) are obtained from Gaussian fits with Shirley backgrounds (dashed lines). Mixing ratios of the two compounds in the blends are given by the mass ratios obtained from quartz crystals. Fermi level positions are depicted as short gray solid lines. (D to F) Maximum positions of HOMO distributions of ZnPc (blue squares) and FnZnPc (red circles) over ZnPc content are shown together with simulation results (dashed lines) of the IE of ZnPc (light blue squares) and FnZnPc (light red circles). Simulated onset values are shifted by 500 meV, which is the typical difference between the maximum and the onset of the HOMO peak in the spectra.

A detailed analysis of the HOMO peaks in the blends yields several remarkable results. First, the difference of the IE of ZnPc and FnZnPc in the blends is considerably smaller than the difference in the neat films. Second, this difference is constant over a broad range of mixing ratios: The IE of ZnPc and FnZnPc linearly shift to higher (lower) binding energies with increasing content of FnZnPc (ZnPc) in the blends. Hence, the IE in blends of ZnPc with F4ZnPc, F8ZnPc, and F16ZnPc can be continuously shifted over a range of 0.5 eV, 0.86 eV, and 1.17 eV, respectively (Fig. 2, D to F, maximum position of Gaussian fits is used), by simply blending these materials. As the linear IE change with mixing ratio is shared by all systems, with larger degrees of fluorination enhancing the slope, a common physical mechanism can already be anticipated.

Density functional theory (DFT) allows estimation of intermolecular interactions along the stacking direction of molecular crystals (3133). Here, we applied DFT calculations to stacks of FnZnPc molecules with the geometry of the β phase. For model geometries consisting of five stacked molecules, interactions along the stacking direction produce substantial shifts of the HOMO energies. In ZnPc stacks, the inner sites suffer a HOMO destabilization of 0.24 eV on average, whereas in F8ZnPc stacks the respective states are stabilized by –0.22 eV, and the HOMO energies in F4ZnPc stacks remain nearly unaffected relative to the gas phase. HOMO shifts on surface molecules of the finite stacks are about half as large.

DFT calculations reproduce an essential finding of the UPS measurements (Fig. 3): In mixed stacks of ZnPc and FnZnPc, intermolecular interactions result in two distinct groups of HOMO and LUMO energies, one arising from the fluorinated and one from the nonfluorinated sites, but the splitting between these two groups of orbitals is significantly reduced with respect to the pure phases. The energy difference between the two subgroups reaches a value of 0.20 eV for ZnPc/F4ZnPc and of 0.34 eV for ZnPc/F8ZnPc, in both cases in excellent agreement with the splitting found in UPS.

Fig. 3 Calculated DOS of the HOMO and LUMO levels.

Mixed stacks of (A) ZnPc:F4ZnPc and (B) ZnPc:F8ZnPc consisting of five molecules according to the geometry of β-ZnPc. The results are averaged over all possible orderings of the stacks, convoluted with a Gaussian broadening of 100 meV (thin solid lines) and 500 meV (dashed lines) for ZnPc (blue) and FnZnPc (red). Thick solid lines depict the sum of ZnPc and FnZnPc levels with a broadening of 500 meV.

The reduced splitting between the orbital energies of fluorinated and nonfluorinated phthalocyanines seems counterintuitive, because any off-diagonal interaction between nondegenerate levels should increase the splitting. Instead, DFT calculations on mixed stacks demonstrate that electrostatic interactions of a molecular orbital with neighboring molecules move the on-site energies of the HOMOs of fluorinated and nonfluorinated ZnPc toward each other. This effect can be rationalized by accounting for electrostatic interactions of a net charge with quadrupole moments of surrounding molecules.

A deficit of the DFT calculations concerns the overall shift of the HOMO energy arising from fluorination: The UPS measurements reveal a shift of 0.7 eV from ZnPc to F4ZnPc, whereas DFT predicts only 0.46 eV. For a more quantitative description of the ionization energies, an explicit account of the excess charge (hole) is required, achieved here via a perturbative embedding procedure applied to large atomistic models (fig. S8; see supplementary materials for simulation details). Such a coarse-grained approach allows the investigation of much larger molecular assemblies.

Our simulations reveal that the microscopic mechanism responsible for the linear behavior is the superposition of quadrupolar molecular fields of two compounds that act on the excess hole. Indeed, in the absence of intermolecular interactions, the density of states (DOS) of the binary mixture would consist only of two peaks corresponding to the ionization energies of ZnPc and FnZnPc in the gas phase. Changes in composition would then only shift the relative height of these peaks, but not their positions. The linear dependence on mixing ratio is thus conditioned solely by the external, solid-state contribution determined by the molecular packing, degree of molecular ordering, and molecular architecture. In fact, ZnPc carries a net-negative out-of-π-plane quadrupole component, as opposed to the positive out-of-π-plane quadrupole component that characterizes the FnZnPc family (see again the isopotential maps of ZnPc and F16ZnPc in Fig. 1, A and B, and table S3). The long-range character of the charge-quadrupole interaction effectively results in a concentration-weighted average over both contributions, which thus serves as the mechanism for level engineering. Note that the long-range electrostatic contribution is virtually homogeneous across the film. Polarization additionally reduces the level offset between ZnPc and FnZnPc, which accounts for the anomalously small IE offset in the thin film relative to the gas phase. This microscopic view is fostered by the very good agreement between experimentally measured and simulated, composition-dependent IEs (see Fig. 2, D to F, and fig. S9).

To prove the generality of the observations, we performed UPS experiments on blends of SubPc and Cl6SubPc, a material system with a different symmetry (fig. S3). This system also shows a tuning effect of the ionization energies with mixing ratio and the difference of the IE in the blend is smaller than in neat films. Notably, the IE difference of around 300 meV is similar to the value in ZnPc:F8ZnPc blends, where the degree of halogenation (i.e., the amount of electron-withdrawing ligands) is comparable. However, the shift of the ionization energies is much smaller than the shift in blends of ZnPc:F8ZnPc, which already is partly explained by the smaller difference of quadrupole moments of the single molecules (table S3). An orientation deviating from edge-on stacks would decrease the tuning effect even further.

We next address the transport properties of such tunable systems. Previously, a substantial decrease of conductivity in p-doped CBP was seen by admixing molecules with lower ionization energies—that is, by the introduction of traps in the energy gap of CPB (34). Because of the energy distance of the HOMO and LUMO levels between the single molecules in the blends, a negative impact on the charge transport properties might be expected. We obtained mobility data for undoped and temperature-dependent conductivity data for n-doped blends (see supplementary materials). The mobilities slightly decrease for the mixed systems but still allow device application (fig. S4). The room-temperature conductivity increases with mixing ratio from the value of n-doped ZnPc to n-doped F4ZnPc (figs. S5 and S6). Accordingly, the activation energy does not show an increase in the blend layers. Obviously, the transport level shifts with the mixing ratio without significantly impairing the transport properties.

To prove that the continuous shift of energy levels can be exploited in devices, we applied the approach in organic solar cells. In previous studies, a correlation of the open-circuit voltage (Voc) and the effective gap was measured for several donor-acceptor systems. Here, the effective gap is defined as the difference between the IE of the donor and the electron affinity (EA) of the acceptor (1, 5). As tuning of the IE of the donor by intermixing of ZnPc and F4ZnPc is possible, this should equally apply to the effective gap. We chose to study a ternary bulk heterojunction of the two donors with C60 as acceptor. The heterojunction was embedded in a p-i-n–type solar cell (see supplementary materials). The mixing ratio of F4ZnPc and ZnPc was varied from pure F4ZnPc to pure ZnPc while the C60 content was fixed at 60 weight percent.

Tuning the IE of the donor indeed changes the current-voltage characteristics (Fig. 4A and table S1). The Voc shifts almost linearly to higher values with increasing F4ZnPc content, as observed in earlier studies on other ternary blends (35, 36). The 300-meV shift of Voc in our solar cells is, however, smaller than expected from UPS measurements on binary ZnPc:F4ZnPc blends. For that reason, we also performed UPS measurements on ternary blends with varying donor:donor ratio (fig. S9). From pure ZnPc to pure F4ZnPc, the IE of the donor shifts to higher binding energies by 700 meV (i.e., the same magnitude as in binary blends). The IE of the acceptor C60 also shifts to higher binding energies by 300 meV, mirroring the increase in IE that is driven by the change in long-range quadrupolar fields. Assuming a constant distance between the IE and the EA of 2.4 eV for C60 in all ternary blends, the effective band gap for all samples can be calculated (37). The relation of Voc and the effective gap deviates from a linear correlation only at higher F4ZnPc contents (Fig. 4B and table S1).

Fig. 4 Solar cell performance of ternary blends.

(A) Current density–voltage curves of ternary blends with varied donor composition. (B) Open-circuit voltage (Voc, blue circles) and effective gap (red squares) obtained by UPS as a function of F4ZnPc content.

Tuning of energy levels by superimposing quadrupole fields is expected to work in a variety of semiconducting small molecules and polymers, but a few preconditions appear necessary. Besides the large difference in the magnitude of the quadrupole moments along the thin-film normal, the superposition of their fields must be coherent facilitated by a systematic orientation of both species in the films. A tuning effect can even be realized with constituents of the same polarity, but different orientation in the film, which thus effect a different out-of-plane quadrupole moment. Although the spatial range of this tuning effect and the required degree of molecular intermixing need further investigation, not only bulk tuning but also tuning the energy levels spatially by gradients of the mixing ratios is possible. This may motivate entirely new designs of device architectures for organic semiconductors.


Materials and Methods

Figs. S1 to S11

Tables S1 to S3

References (3849)


  1. Acknowledgments: We thank O. Inganäs for his support and insightful discussions, D. Wöhrle for supplying F8ZnPc, O. Kaveh for conductivity measurements, D. Schütze for building OFETs, L. Wilde for performing GIXRD measurements, and F. Holzmüller for evaluating and discussing GIXRD measurements. The research in Dresden was funded by the DFG project MatWorldNet LE-747/44-1, as well as the European Community’s Seventh Framework Programme under grant agreement FP7-267995 (NUDEV). The research at Linköping was supported by the Knut and Alice Wallenberg Foundation through a Wallenberg Scholar grant to O. Inganäs, the Swedish Research Council (VR, 330-2014-6433), and the European Commission Marie Skłodowska-Curie Actions (INCA 600398). Also supported by Bundesministerium für Bildung und Forschung project MEDOS grant FKZ 03EK3503B (C.P. and D.A.) and by the Dr. Isolde-Dietrich-Stiftung (A.A.G.). K.L. is a fellow of the Canadian Institute for Advanced Research (CIFAR). Author contributions: M.S. and K.O. acquired UPS data; W.T., B.B., and F.G. acquired and evaluated solar cell data; M.S. evaluated UPS and conductivity data; R.S. performed DFT calculations; C.P. performed molecular simulations; A.A.G. obtained mobility values; M.S., K.L., C.P., R.S., and D.A. wrote the manuscript; all authors contributed to discussions and finalizing the manuscript. There are no competing financial interests.
View Abstract

Navigate This Article