Design of solar cell materials via soft X-ray spectroscopy

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/authorsrights

Author's personal copy Journal of Electron Spectroscopy and Related Phenomena 190 (2013) 2–11

Contents lists available at ScienceDirect

Journal of Electron Spectroscopy and Related Phenomena journal homepage: www.elsevier.com/locate/elspec

Design of solar cell materials via soft X-ray spectroscopy F.J. Himpsel a,∗ , P.L. Cook b , G. de la Torre c , J.M. Garcia-Lastra d,e , R. Gonzalez-Moreno d , J.-H. Guo f , R.J. Hamers g , C.X. Kronawitter h , P.S. Johnson a , J.E. Ortega d , D. Pickup d , M.-E. Ragoussi c , C. Rogero d , A. Rubio d , R.E. Ruther g , L. Vayssieres i , W. Yang f , I. Zegkinoglou a,f a

Department of Physics, University of Wisconsin Madison, Madison, WI 53706, United States Natural Sciences Department, University of Wisconsin Superior, Superior, WI 54880, United States Departamento de Química Orgánica, Facultad de Ciencias, Universidad Autónoma de Madrid, Campus de Cantoblanco, 28049 Madrid, Spain d Material Physics Center (MPC), Centro de Física de Materiales (CSIC-UPV/EHU), Donostia International Physics Center (DIPC), Departamento de Fisica Aplicada I, Universidad del Pais Vasco, 20018 San Sebastian, Spain e Department of Physics, Center for Atomic-scale Materials Design, Technical University of Denmark, DK-2800 Kgs. Lyngby, Denmark f Advanced Light Source, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, United States g Department of Chemistry, University of Wisconsin Madison, Madison, WI 53706, United States h Environmental Energy Technologies Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, United States i International Research Center for Renewable Energy, State Key Laboratory of Multiphase Flow in Power Engineering, School of Energy & Power Engineering, Xian Jiaotong University, Xian 710049, PR China b c

a r t i c l e

i n f o

Article history: Available online 22 October 2012 Keywords: Solar cell Photovoltaics X-ray absorption spectroscopy Band offset Dye molecule Diamond

a b s t r a c t This overview illustrates how spectroscopy with soft X-rays can assist the development of new materials and new designs for solar cells. The starting point is the general layout of a solar cell, which consists of a light absorber sandwiched between an electron donor and an electron acceptor. There are four relevant energy levels that can be measured with a combination of X-ray absorption spectroscopy and photoelectron spectroscopy, as illustrated for an organic dye as absorber attached to a p-doped diamond film as donor. Systematic measurements of organometallic dyes (phthalocyanines and porphyrins) as a function of the metal atom are presented for the metal 2p and N 1s absorption edges. In combination with density functional theory one can discern trends that are useful for tailoring absorber molecules. A customized porphyrin molecule is investigated that combines an absorber with a donor and a linker to an oxide acceptor. The bridge to device fabrication is crossed by correlating spectroscopic features with the photocurrent in hematite photoanodes for water splitting. For speeding up the development of new materials and designs of solar cells a feedback loop between spectroscopy, theory, synthesis and device fabrication is envisioned. © 2012 Elsevier B.V. All rights reserved.

1. Introduction The search for clean energy is rapidly becoming one of the most pressing technological challenges. It is natural to tap into solar energy, the origin of most types of energy being used today, such as fossil, wind, and hydroelectric. Ideally, one would want to convert solar energy directly into electricity rather than using thermal energy as intermediate, such as in power plants and combustion engines. Electricity can be converted nearly lossless into any other type of energy, while conversion of thermal energy is limited by the Carnot efficiency. Such reasoning suggests solar electricity as the ultimate energy source. Direct conversion of solar energy to electricity by photovoltaics is not yet financially competitive with traditional (polluting) energy

∗ Corresponding author. E-mail address: [email protected] (F.J. Himpsel). 0368-2048/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.elspec.2012.10.002

sources, such as coal, oil, and gas, although photovoltaics has already obtained grid parity in specific locations [1]. Progress in photovoltaics has been achieved mainly by the Edisonian method of educated trial and error. Most of the R&D in industry has been restricted to the optimization of a specific design, such as silicon, CdTe, or CIGS (copper indium gallium selenide). Advances in cost per watt have been small compared to the rapid improvements in microelectronics exemplified by Moore’s law, because solar cells cannot take advantage of miniaturization. Therefore, photovoltaics has not been a major contender for solving the energy problem, even though it is one of the cleanest possible sources of energy and has experienced enormous growth rates in favorable times, when fossil fuels became expensive. This situation calls for a new strategy, which accelerates the slow evolution of traditional concepts by considering a large variety of new materials and devices. They need to be winnowed down quickly to the most promising technologies. This is where spectroscopy comes into the picture. It makes it possible to measure the energy levels that determine whether or

Author's personal copy F.J. Himpsel et al. / Journal of Electron Spectroscopy and Related Phenomena 190 (2013) 2–11

3

Fig. 1. Generic energy level diagram of a solar cell, consisting of a light absorber sandwiched between an electron donor and an electron acceptor. An example is the dye-sensitized solar cell. A traditional semiconductor solar cell combines all three functions in one material. Choosing three different semiconductors provides extra flexibility. A large energy drop between adjacent levels increases the photocurrent by speeding up carrier separation, but reduces the (open circuit) photovoltage VOC .

not a solar cell design has a chance to separate electrons and holes and thereby provide a photovoltage. Consider a generic solar cell, as shown in Fig. 1. It contains three essential parts, a light absorber, an electron acceptor, and an electron donor. Four energy levels need to be arranged correctly to pull the excited electron toward the acceptor and refill the hole with a donor electron. Using semiconductors, the valence band maximum (VBM) acts as donor, and the conduction band minimum (CBM) as acceptor. In a molecule the VBM corresponds to the highest occupied molecular orbital (HOMO) and the CBM to the lowest unoccupied molecular orbital (LUMO). Both semiconductors and molecules will be considered here. In order to separate the electron–hole pairs excited by the photons, a small energy drop is needed between the CBM of the absorber and the dye, as well as between the VBM of the donor and the absorber. The open circuit output voltage VOC is maximized by minimizing this energy drop. But for maximizing the current one needs a large energy drop, which separates electrons and holes rapidly and thereby prevents recombination. Optimizing the voltage–current product requires a delicate trade-off in the spacing of adjacent energy levels. A desirable range of band offsets is indicated by blue double-arrows in Fig. 2. In the generic solar cell of Fig. 1, each of the three parts can be optimized separately, but in a standard silicon solar cell these functions are all combined into one elemental material. That provides little room for innovations. Dye-sensitized solar cells (or Grätzel cells) are more flexible by separating the three roles [2–5]. Typically, TiO2 nanoparticles act as acceptor, a redox electrolyte as donor, and organic dye molecules absorb sunlight. Those can be tailored in many ways by organic chemistry. Interestingly, the trade-off between maximum voltage and maximum current is handled asymmetrically in a typical Grätzel cell (see the energy diagram in Fig. 3 of [5]). The voltage drop between dye and acceptor is very small, but about half of the cell voltage is lost between the donor and the dye. The small voltage drop on the acceptor side keeps the excited electrons lingering near their holes. A large voltage drop is needed on the donor side to fill the holes before recombination takes place. It remains to be investigated whether a symmetric or asymmetric voltage drop is best for optimizing the power product. Here we envisage an even more general design, where any of the three components can be molecular, liquid, or solid state. Although it is riskier to stray from proven avenues, such as semiconductor pn-junctions, this strategy opens up a much larger playing field for finding truly innovative combinations. It will take extra work, though, to refine a new design to the same level as traditional

Fig. 2. Example illustrating the design of a solar cell based on spectroscopic input about the energy levels. H-terminated p-type diamond is selected as donor and ntype ZnO or TiO2 as acceptor. The desirable range for the band edges of the absorber is indicated by double-arrows. The bands are referenced to the vacuum level of the isolated materials (adapted from Fig. 2 of [5], with H-terminated diamond data from [6–8]). When interfaces react, this simple picture breaks down. The actual band offsets have to be determined spectroscopically, because it is very difficult to predict the structure of a reacted interface theoretically.

pn-junctions. For example, dye-sensitized solar cells have not yet reached the efficiency of Si solar cells, despite their more flexible design. But they use much less energy during their fabrication and thus have an intrinsic cost advantage over silicon. To be specific, we discuss in Fig. 2 the design of an all solidstate solar cell consisting of three semiconductors. It uses a p-type diamond film as inert donor, GaAs as absorber, and n-type ZnO or TiO2 as acceptor. The conduction band minima are shown in red and the valence band maxima in yellow. The simplest model for obtaining the positions of the band edges uses their positions relative to the vacuum level in an isolated solid (“electron affinity model”). For ZnO, TiO2 , and GaAs these are taken from Fig. 2 of [5], and for diamond from [6–8]. This is equivalent to the Schottky Model for semiconductor–metal junctions. For Schottky barriers several more sophisticated models have been developed that take the specific nature of the interface states into account. These have counterparts for explaining band offsets at heterojunctions. The main difficulty in predicting accurate Schottky barriers and band offsets is the unknown arrangement of the atoms near the interface. Growth and annealing conditions determine the interface bonds and cause interdiffusion. Interface dipoles are generated which create a potential step across the interface. Various spectroscopic methods make it possible to determine Schottky barriers and band offsets experimentally. The valence band maximum is obtained relative to the Fermi level by photoelectron spectroscopy. Likewise, the conduction band minimum can be determined relative to the Fermi level by inverse photoemission or two-photon photoemission. In addition, one can probe the band edges relative to a core level. The valence band maximum is probed by X-ray emission spectroscopy and the conduction band minimum by Xray absorption spectroscopy. These results can then be combined with X-ray photoelectron spectroscopy of the core level to relate the band edges to the Fermi level.

Author's personal copy 4

F.J. Himpsel et al. / Journal of Electron Spectroscopy and Related Phenomena 190 (2013) 2–11

While interface dipoles lead to potential steps, bulk charges in the depletion zone lead to band bending (both omitted in Figs. 1 and 2 for clarity). Consequently the Fermi level position in the gap differs between bulk and surface, an effect that has been studied extensively for semiconductor surfaces, both inorganic and organic [9–11]. Spectroscopy can also get a handle on band bending by varying the probing depth, for example by changing the photon energy in photoelectron spectroscopy. In special cases it is possible to detect the acceptor level and thus the bulk Fermi level by bulk-sensitive X-ray absorption spectroscopy with fluorescence detection, as shown for diamond below [8,12,13]. The versatile design in Fig. 1 allows taking advantage of the enormous progress that has taken place in tailoring materials by nano-structuring, supra-molecular assembly, and atomic layer epitaxy. Such highly controllable materials might become the key for producing radically different solar cells with much higher efficiency. With a large variety of materials under consideration it becomes crucial to speed up the pre-selection of the most promising combinations. Spectroscopy plays again an important role by correlating the performance of the acceptor, donor, and absorber with their energy levels. But a much wider feedback loop is required, which covers not only the analysis of energy levels, but also the synthesis of new materials, device fabrication, and electrical testing. In addition, it is important to have theory in the loop, which allows educated predictions of promising materials for the synthetic chemist and crystal grower. A team of experts covering many areas needs to work together in order to close the feedback loop. This article, for example, incorporates coauthors with expertise in spectroscopy, density functional theory, synthesis of dye molecules, covalent attachment of dye molecules to semiconductors, and the fabrication of nanostructured materials for donor and acceptor electrodes. In the following, we start with a brief description of common spectroscopic methods and their capabilities, with an emphasis on X-ray absorption spectroscopy. Systematic results on the electronic structure of dye molecules follow, which illustrate how a library of energy levels reveals trends that can be used for tailoring organometallic molecules. These measurements are closely coupled with theory, in order to understand the origin of various energy shifts and the orbital character of the observed states. As example for tailored materials we present results from a custom synthesized dye molecule that includes a donor and a link to a solid-state acceptor. Eventually we will make a connection between spectroscopy and the electrical performance of a device as a function of processing parameters.

2. Spectroscopic methods and their capabilities The spectroscopy used most widely in photovoltaics is ultraviolet/visible (UV/vis) absorption spectroscopy. It is indispensable for finding out whether the absorber is matched to the solar spectrum. However, the information about the energy levels is scrambled in such a spectrum, as shown schematically in Fig. 3. At a given photon energy one can have many simultaneous transitions from occupied to empty valence states, as indicated by vertical arrows of equal length. Thereby the densities of the occupied and empty states become convoluted with each other, making it difficult to discern spectral features originating from occupied and empty states. Using a core level instead of the valence band greatly simplifies the situation. The initial state is now a single, sharp energy level. Each observed transition corresponds to a unique final state. Since core levels are element specific, it becomes possible to probe the local density of states at selected atoms by using the absorption edges of different elements.

Fig. 3. Schematic comparison between UV/vis absorption spectroscopy and X-ray absorption spectroscopy, with the spectra plotted vertically to match the energy level schemes of Figs. 1 and 2. UV/vis is essential for determining how much solar light is absorbed, but it scrambles transitions between different combinations of occupied and empty valence orbitals, as indicated by three transitions occurring at the same photon energy. X-ray absorption spectroscopy starts with a well-defined core level and thereby allows unique identification of the empty orbitals. Choosing core levels of different elements (such as C 1s, N 1s, or first row transition metal 2p) makes it possible to select specific parts of a dye molecule.

A drawback of this type of spectroscopy is the need for a tunable light source in the soft X-ray region, which essentially requires going to a synchrotron light source rather than having a setup readily accessible in the lab. Facilitating rapid access and rapid return remains a challenge at crowded synchrotron radiation facilities, but progress has been made via robotics in high throughput areas, such as protein crystallography. The sharpest core levels of common elements lie in the photon energy region between 50 and 1000 eV (see Fig. 2 in [12]). These core levels typically have an intrinsic lifetime broadening of the order 0.1 eV. Starting from low atomic number, one first encounters the 1s levels of C, N, O, F, the key elements of organic dye molecules. Then come the 2p levels of Mg, Al, Si, P, S, which cover important semiconductors. The transition metal 2p levels provide strong transitions into the 3d valence states, which will be discussed later in the context of organometallic dye molecules. Heavy metals and rare earths require higher energies up to several keV, since the cross section of their sharp 4f levels is highly suppressed at the absorption edge. One needs to resort to deeper, but broader core levels. Synchrotron radiation has the additional advantage of being polarized. The polarization dependence of optical dipole transitions provides straightforward information about the orientation of unoccupied orbitals in both molecules and solids, as shown in Fig. 4 (from [14]). For transitions from the 1s core level into unoccupied 2p orbitals the dipole selection rules are quite simple. The transition probability is maximized when the photon polarization is parallel to the axis of the 2p orbital and zero when perpendicular. The angular distribution follows a cos2 () law. In Fig. 4, the transitions into various ␲* orbitals of a planar phthalocyanine molecule show maximum intensity at normal incidence ( = 0◦ ), where the polarization vector is parallel to the surface. Therefore, the ␲* orbitals are nearly parallel to the surface, and the molecular plane is perpendicular to it. This is confirmed by the opposite behavior of the higher-lying ␴* orbitals, which are oriented perpendicular to the ␲* orbitals. The inset shows the normalized intensity of the LUMO as a function of cos2 (). The line does not reach zero for cos2 () = 0, because the molecules are either disordered or somewhat tilted. The slope of the cos2 () plot

Author's personal copy F.J. Himpsel et al. / Journal of Electron Spectroscopy and Related Phenomena 190 (2013) 2–11

5

Fig. 4. Detecting the orientation of Fe-phthalocyanine dye molecules in a thin film via the polarization dependence of the C 1s absorption spectrum (from [14]). The ␲* peaks show maximum intensity at normal incidence ( = 0◦ ), demonstrating preferential orientation of the molecular plane perpendicular to the surface. The inset shows the normalized intensity of the LUMO as a function of cos2 () using p polarization. The slope of such plots is related to the average tilt angle of the molecules.

provides a quantitative measure of either the order or the tilt angle via modeling. In organic electronics both the order and the tilt angle of ␲ systems are important parameters in determining the overlap between the ␲-sytems of adjacent molecules. The polarization selection rules become more complicated when going from a 1s to a 2p core level. In that case there are two channels, 2p→d and 2p→s. The selection rules have been worked out for the S2p transitions of thiophene [15], a molecule with two orthogonal mirror planes. Polythiophene is an important donor in polymer-based solar cells. Being able to vary the probing depth is important for identifying band bending and interdiffusion in solar cells. In X-ray absorption spectroscopy this can be achieved by detecting secondary electrons or photons, which are both produced by the decay of the core hole that is created in the absorption process. Total electron yield (TEY) detection is surface sensitive (typically 5 nm), while total fluorescence yield (TFY) detection is bulk sensitive (typically 100 nm). A dramatic example of the difference between the two detection methods is given in Fig. 5 (from [8]). A TEY and a TFY spectrum were simultaneously acquired from a highly p-doped diamond film at the C 1s edge. The only common feature is the onset of transitions into the conduction band of diamond at about 289 eV. The bulk-sensitive TFY spectrum shows a pronounced exciton peak, which is characteristic of diamond [16], while the surface-sensitive TEY spectrum exhibits only a shoulder at that energy. Transitions into unoccupied gap states below the exciton are completely different. Peaks A and B in the bulk-sensitive spectrum can be assigned to transitions from different C atoms into the acceptor state [8,17], while peaks C and D in the surface-sensitive spectrum are caused by ␲-bonded carbon defects and hydrogen-terminated carbon [8,18,19]. Fig. 6 summarizes a combination of experiments aimed at measuring the energy levels that are relevant for a dye-sensitized solar cell (from [8]). A highly p-doped diamond film is used as donor. The dye molecule is attached covalently to the diamond surface, as shown by an inset [20]. According to Fig. 1, the relevant energy levels are the VBM of the diamond donor and the HOMO and LUMO of the dye molecule. The VBM is determined by subtracting the band gap of diamond (5.47 eV) from the CBM, which in turn is located 0.2 eV above the bulk core exciton peak at 289.1 eV [16]. The HOMO of the dye molecule is determined by photoelectron spectroscopy (shown in the inset for a thin dye film; data on diamond are given in [8]). The LUMO is determined by X-ray

Fig. 5. Comparison of bulk- and surface-sensitive X-ray absorption spectra (from [8]), obtained by simultaneously detecting the total fluorescence yield (TFY) and total electron yield (TEY). For a boron-doped, hydrogen-terminated diamond film the two spectra are completely different, with the TFY spectrum showing empty gap states induced by boron doping (A and B) and the TEY spectrum showing states induced by ␲-bonded carbon defects (C) and the C H monohydride termination (D).

absorption spectroscopy at the C 1s edge, where it can be referenced directly to the CBM of diamond. Since the dye coverage is well below a monolayer, the signal from the LUMO is rather weak. Nevertheless, it can be brought out clearly by taking the difference between surfaces with and without dye. It also helps to excite the dye levels at the C 1s edge, where they compete only with the low density of states in the band gap of diamond. At the N 1s edge, one has to cope with a large background from C 1s transitions into high-lying continuum states (not shown). The energy scale of the absorption spectra in Fig. 6 is tied to the VBM and CBM of diamond (vertical dashed lines). Due to band bending, the position of the Fermi level EF in the gap will depend on the probing depth. EF at the surface is obtained by measuring the energy of the C 1s level relative to EF via X-ray photoelectron spectroscopy (285.0 eV in Fig. 6). EF in the bulk is obtained directly from the acceptor level, which is seen as peak B in the bulk-sensitive absorption spectrum in Fig. 5. It lies 0.2 eV above the VBM, as one would expect for highly p-doped diamond. This method leads to a complete energy level scheme, which includes the HOMO, LUMO, VBM, CBM, and EF (both surface and bulk). The result for the LUMO represents a lower limit, since the Coulomb attraction between the excited electron and the core hole reduces the energy of the optical transition. The HOMO of the adsorbed molecules lies below the VBM of the diamond donor, as it should according to Fig. 1. But the energy difference is too large, leading to a substantial loss in the voltage output of a potential solar cell. This calls for a modification of the dye molecules. Alternatively, one could try introducing an appropriate dipole via the attachment chemistry. Thus we have a proof of principle for determining all the relevant energy levels by a combination of spectroscopies. The next step will be to systematically search for optimized combinations of donors, acceptors, and absorbers.

Author's personal copy 6

F.J. Himpsel et al. / Journal of Electron Spectroscopy and Related Phenomena 190 (2013) 2–11

Fig. 7. Detecting the oxidation state of the iron atom at the center of a Fephthalocyanine dye molecule by the multiplet structure of the X-ray absorption spectrum (from [14]). A vacuum-deposited thin film exhibits the native Fe2+ oxidation state (top). As the sample is exposed to air, the spectrum becomes a mixture of Fe2+ and Fe3+ multiplets. A dried ethanol solution of commercial Fe-phthalocyanine powder consists almost completely of Fe3+ within the TEY probing depth of about 5 nm (bottom). Such reactions with ambient oxidants have been a key factor in the degradation of organic LEDs.

Fig. 6. Energy levels of Ru(tpy)2 dye molecules attached covalently to a boron doped, H-terminated diamond thin film (from [8]). X-ray absorption spectroscopy is used for unoccupied states and PES (photoelectron spectroscopy) for occupied states. The position of the Fermi level in the C 1s spectra is obtained from PES of the C 1s level (not shown).

3. Systematics of dye molecules A systematic investigation of the electronic structure of potential absorbers, acceptors, or donors is highly desirable for designing solar cells from scratch. Organometallic dye molecules, such as phthalocyanines and porphyrins, lend themselves to this task, since their central metal atom can be varied across a significant part of the periodic table. For the first row of transition and noble metals the 2p–3d transitions are particularly informative. They exhibit a rich multiplet structure that originates from the interaction of a partially filled 3d shell with itself and with the 2p core hole, as shown in Figs. 7 and 8 (from [14,21], see also [22,23] for early transition metals). These spectra can be parametrized by a small number of parameters [24], which then provide information about the oxidation state and the crystal field splitting. The metal d states are quite relevant to dye-sensitized solar cells, which typically use organometallic molecules with a Ru atom at the center. The HOMO–LUMO transition has charge transfer character, with the HOMO typically located at the transition metal and the LUMO at the nitrogen atoms that surround it. This is shown in Fig. 9 for porphyrins (from density functional theory [25]). Having charge

separation already in the initial photoabsorption process facilitates the further separation of electrons and holes by the acceptor and donor. Since Ru is rare and expensive, it is worth considering the element above it in the periodic table, which is Fe. It occurs in biomolecules involved with charge transfer reactions, such as hemoglobin and cytochrome c [26,27]. The question arises why Fe can perform such tasks in biochemistry, while only Ru seems to work for photovoltaics. Investigations are under way to search spectroscopically for the characteristic differences between Fe and Ru that cause this disparity [28,29]. Similar questions are being asked in catalysis [30]. One of the potential handicaps of Fe in photovoltaics is its higher chemical reactivity. This is demonstrated in Fig. 7, where the Fe2+ phthalocyanine dye is shown to be unstable against oxidation to Fe3+ in air. Reference spectra for Fe2+ and Fe3+ are obtained by evaporating Fe2+ phthalocyanine in ultrahigh vacuum and drop-casting Fe3+ Cl-phthalocyanine. The equivalent Ru2+ COphthalocyanine does not show this tendency to oxidize [29]. Fe-containing biomolecules are protected by a three-dimensional protein cage, which enables them to reversibly switch between Fe2+ and Fe3+ . This suggests going from planar phthalocyanine and porphyrin dyes to organometallics with three-dimensional cages. Examples are dyes with approximately octahedral symmetry around the Fe atom, such as the Fe-tris(bipyridyl) group [29]. In the search for new materials for solar cells it is valuable to establish a database of spectroscopic properties. Fig. 8 shows systematic 2p→3d absorption spectra of the central metal atom in a phthalocyanine (Pc) or octaethylporphyrin (OEP) along the row of 3d metals from Mn to Zn (compiled from [14,21]). Analogous data are available for early 3d transition metals [22,23]. These transitions

Author's personal copy F.J. Himpsel et al. / Journal of Electron Spectroscopy and Related Phenomena 190 (2013) 2–11

Fig. 8. The 2p edges of organometallic dye molecules containing 3d metals (from [14,21]).

7

Author's personal copy 8

F.J. Himpsel et al. / Journal of Electron Spectroscopy and Related Phenomena 190 (2013) 2–11

Fig. 9. Character of the frontier orbitals of porphyrin dyes, obtained from firstprinciples calculations (from [25]). The LUMO has mainly N 2p character, while the orbitals close to the HOMO are dominated by metal 3d wave functions. The resulting HOMO-to-LUMO transition has charge transfer character and thus facilitates charge separation in photovoltaics.

form rich multiplets that provide information about the oxidation state and the ligand field. The effect of the oxidation state is demonstrated for Mn2+ - vs. Mn3+ -OEP and for Fe2+ - vs. Fe3+ -Pc (compare also Fig. 7). These were the only molecules that could be prepared in two oxidation states. Facile switching between oxidation states is an important factor for charge separation and charge transfer. While the multiplet structure varies strongly with the number of 3d electrons, multiplets of isoelectronic configurations show some resemblance, such as Mn2+ -OEP and Fe3+ Cl-Pc. The more subtle effect of the crystal field can be seen by comparing Pcs and OEPs containing the same metal in the same oxidation state. As illustrated at the top of Fig. 8, the metal atom is surrounded by two rings of N atoms in a Pc, compared to single ring in an OEP. The strength and symmetry of the crystal field can be characterized by a few parameters, which are obtained by modeling the data with atomic multiplets [24,26]. As the 3d shell becomes filled toward the end of the series in Fig. 8, transitions into metal 4s and nitrogen 2p states appear above the 3d multiplet (magnified 10× for Cu and Zn) [21]. This is particularly pronounced for Zn where the 3d shell in nominally filled in the 2+ oxidation state. Zn-based porphyrins and phthalocyanines play an important role in dye-sensitized solar cells [2] and organic LEDs. Instead of analyzing the transition metal 2p absorption edge, one can also investigate the 1s absorption edge of the nitrogen atoms surrounding the metal. These usually form the LUMO, as shown in Fig. 9 (from [25]). In this case a first-principles calculation has been performed for analyzing the contributions to the observed energy of the optical transition. These are the energy of the N 1s core level, the energy of the LUMO, and the Coulomb interaction between the electron in the LUMO and the N 1s core hole. The N 1s core level energy can be measured by X-ray photoelectron spectroscopy. But the electron–hole interaction is difficult to track down experimentally. In Fig. 10, a systematic shift of the N1s→LUMO transition is observed when varying the central metal atom. The calculation reproduces that shift, even though the absolute energy is off (lower panel). Therefore it is possible to analyze the calculated shift and find out which of the three energies causes it (upper panel). It turns out to be the N 1s core level binding energy.

Fig. 10. First-principles calculations disentangle various contributions to the energy shift observed in N 1s absorption spectra for a series of 3d metal porphyrins (from [25]). These are the energy of the N 1s core level, the LUMO, and the electron–hole interaction (top). Their sum matches the observed trend (bottom). The main contribution is a chemical shift of the N 1s level induced by charge transfer from the metal atom.

One can go one step further and explain the energy shift of the N 1s core level by variations in the charge transfer from the metal atom to the surrounding N atoms (see Fig. 3b in [25]). Electronegative metals donate less electronic charge to the nitrogen, which makes the electrostatic potential more positive at the N and thereby lowers the energy of the N 1s core level. Thus, a higher photon energy is required to excite a N 1s electron into the same ␲* level. This charge transfer is related to the electronegativity of the various 3d metals. For example, the maximum excitation energy between Ni and Co in the lower panel of Fig. 10 matches a maximum in the electronegativity along the row of 3d metals. A similar effect has been found in the comparison of Fe- and Ru-based dye molecules [29]. 4. Customized dye molecules Recently, the concept of donor-␲-acceptor molecules has been gaining interest [2], where ␲ stands for the ␲-electron system of organic dye molecules. The functions of the absorber, donor, and acceptor are all combined in one molecule. This allows atomic precision and control of the architecture in a solar cell. Fig. 11 shows N 1s absorption spectra from such a molecule (labeled GT15) and two reference molecules (GT14 and tris(DPA)), synthesized in Madrid and drop-cast from a solution in THF (from [31]). GT15 combines a Zn-porphyrin dye with three diphenylamine (DPA) donor groups and a carboxyl (COOH) group which is used to link the molecule with a standard TiO2 nanoparticle acceptor. GT14 lacks the DPA groups and tris(DPA) consists exclusively of DPA groups. The spectra in Fig. 11 allow a clear identification of the orbitals derived from the three active groups. The peak and shoulder below

Author's personal copy F.J. Himpsel et al. / Journal of Electron Spectroscopy and Related Phenomena 190 (2013) 2–11

9

Fig. 11. Using X-ray absorption spectroscopy to identify individual functional groups in a custom-synthesized molecule that combines a Zn-porphyrin dye with three diphenylamine (DPA) donor groups and a carboxyl linker for attachment to metal oxide acceptors (from [31]). The N 1s spectrum of this molecule is compared to a similar dye without the donor groups and a molecule containing only donor groups. Combining the functions of donor, dye, and acceptor in a single molecule allows ultimate control.

400 eV and the doublet near 402 eV originate from the ␲-system of the Zn-porphyrin dye. The peak near 403 eV is characteristic of the DPA donor group. And the signature ␲* orbitals of the carboxyl linker can be observed at the C 1s and O 1s edges (not shown, see [31]). This once more demonstrates the selectivity of X-ray absorption spectroscopy by identifying not only specific atoms (such as the central metal atom), but also specific functional groups (such as the DPA donor). We have investigated additional modifications of these molecules where the number of carboxyl linkers has been increased to strengthen the attachment to oxide acceptors [32]. Indeed, a substantial increase of the molecular coverage was observed from the signal/background ratio of the LUMO peak. Insufficient coverage can be a major problem with acceptors such as single crystalline ZnO nanorods [4,33]. Such high quality acceptor materials would be ideal for the transport of electrons to the negative electrode. But the side faces of the hexagonal nanorods are non-polar and have such a low surface energy that the standard carboxyl linker fails to establish a viable bond. 5. Spectroscopy of real devices The discussion about efficient attachment of dye molecules already led toward the use of spectroscopy as a diagnostic tool in device fabrication. Such investigations have begun, for example by comparing J–V curves with band diagrams similar to Fig. 1 for differently prepared interfaces between an AlCl-Pc donor and a C60 acceptor in [34]. Fig. 12 shows another example where a connection between the electrical properties and a spectroscopic signal has been established (from [35]). This is from an effort to create fuel from sunlight by a photoelectrochemical method (for reviews see [5,36]). Nanostructured hematite (␣-Fe2 O3 ) has been used as photoanode for solar water splitting [37]. It can be combined with

Fig. 12. Connection between spectroscopy and device properties, exemplified by an oxide photoanode for solar water oxidation (from [35]). The photocurrent increases by more than two orders of magnitude after annealing the device. X-ray absorption spectroscopy reveals diffusion of Sn into the Fe2 O3 photoanode from an underlying transparent electrode consisting of fluorine-doped tin dioxide. Such interface reactions affect the interface dipole that determines the electric field for charge separation at interfaces.

solar cells in a tandem configuration [38]. Resonant inelastic X-ray scattering studies of hematite quantum rods have revealed a 1D confinement effect [39], where the bottom of the conduction band shifts upward by 0.3 eV matching the level of H2 /H+ and increasing the band gap to an ideal 2.5 eV for solar water splitting at low bias. These quantum rods show good quantum efficiency (as high as 56% at 350 nm), due to a very good match between minority carrier diffusion length and rod diameter [40]. The remaining obstacle is the slow extraction of photoexcited electrons to an external circuit, which increases the probability of carrier loss by recombination [41]. When Fe2 O3 is deposited onto a conductive fluorine-doped tin dioxide (SnO2 :F) substrate, high-temperature annealing dramatically increases conversion efficiencies for the solar-driven oxidation of water, as shown in the lower panel of Fig. 12. The upper

Author's personal copy 10

F.J. Himpsel et al. / Journal of Electron Spectroscopy and Related Phenomena 190 (2013) 2–11

the DOE under the contracts DE-SC0006931, DE-AC02-05CH11231 (ALS), and DE-FG02-01ER45917 (end station), by the Spanish Ministerio de Economia y Competitividad (MAT2010-21156-C03-01, C03-03, and PIB2010US-00652), and by the Basque Government (IT-257-07). RER and RJH acknowledge support from the NSF with grants CHE-0613010 and CHE-0911543.

References

Fig. 13. The concept of a feedback loop between spectroscopy, theory, materials synthesis, and device fabrication/testing. Having close connections between these four pillars will speed up the rate of innovation and allow a much larger variety of solar cell designs to be tried out.

panel shows O 1s absorption spectra, which exhibit distinct peaks for electrons excited into Fe 3d and Sn 5s states of the respective oxides (hybridized with O 2p states). Depending on the thickness of the Fe2 O3 and the annealing temperature, one can observe clear changes in the intensity of the Sn 5s peak. It appears for small film thickness or high annealing temperature, indicating thermal diffusion of Sn into the Fe2 O3 film. The resulting mixed Sn–Fe oxide has a different electronic structure, which can be detected by subtle changes in the line shape of the Fe 2p absorption edge (not shown). Such studies can provide direct feedback between spectroscopy and device fabrication. In order to handle real devices it is necessary to adapt traditional soft X-ray spectroscopy with synchrotron radiation from ultrahigh vacuum to ambient environments, such as air, liquids, and electrolytes. This can be achieved by photon-in photon-out techniques via thin windows separating the sample region from vacuum [12,42,43]. Even photoelectron spectroscopy can be adapted to these conditions by a sophisticated differential pumping design [44,45]. 6. Conclusions and outlook This article shows how various types of spectroscopy come together for obtaining a complete picture of the energy levels that underlie the most general solar cell design. Such spectroscopic measurements are part of a bigger picture. The ultimate goal is to speed up the improvement of solar cells by obtaining better efficiency at equal or lower cost. For this purpose we are in the process of setting up a feedback loop, which is sketched in Fig. 13. Spectroscopy of molecules and semiconductors provides input for theory, which helps explaining the observed energy levels and their wave functions. First-principles calculations even have the capability to predict new materials of potential interest to photovoltaics, such as acceptors [46], anchoring groups [47], or all three components of a solar cell [48]. These blueprints are passed on to synthetic chemists and crystal growers. The newly synthesized materials are then incorporated into device structures, which allow measurements of the electric performance. Since device processing may change the energy levels via interface reactions or creation of bulk defects, the ball comes back to spectroscopy. If such a feedback loop can be executed efficiently, the time from discovery to practical use in solar cells will be shortened considerably. It will be possible to consider a much larger variety of materials by screening them spectroscopically. Many new concepts can be tested, including far-flung ideas that otherwise might never get a chance. Acknowledgments This work was supported by the NSF with the awards CHE1026245, DMR-1121288 (MRSEC), and DMR-0537588 (SRC), by

[1] K. Branker, et al., Renew. Sustain. Energy Rev. 15 (2011) 4470. [2] A. Yella, H.-W. Lee, H.N. Tsao, C. Yi, A.K. Chandiran, Md K. Nazeeruddin, E.W.-G. Diau, C.-Y. Yeh, S.M. Zakeeruddin, M. Grätzel, Science 334 (2011) 629. [3] B.E. Hardin, H.J. Snaith, M.D. McGehee, Nat. Photonics 6 (2012) 162. [4] J.A. Anta, E. Guilleˇın, R. Tena-Zaera, J. Phys. Chem. C 116 (2012) 11413. [5] Michael Grätzel, Nature 414 (2001) 338. [6] F.J. Himpsel, J.A. Knapp, J.A. Van Vechten, D.E. Eastman, Phys. Rev. B 20 (1979) 624. [7] C. Bandis, B.B. Pate, Phys. Rev. B 52 (1995) 12056. [8] I. Zegkinoglou, P.L. Cook, P.S. Johnson, W. Yang, J. Guo, D. Pickup, R. GonzálezMoreno, C. Rogero, R.E. Ruther, M.L. Rigsby, J.E. Ortega, R.J. Hamers, F.J. Himpsel, J. Phys. Chem. C 116 (2012) 13877. [9] M. Gorgoi, D.R.T. Zahn, Org. Electron. 6 (2005) 168. [10] B.P. Rand, D.P. Burk, S.R. Forrest, Phys. Rev. B 75 (2007) 115327. [11] A. Kahn, N. Koch, W. Gao, J. Polym. Sci. B: Polym. Phys. 41 (2003) 2529. [12] F.J. Himpsel, Phys. Status Solidi B 248 (2011) 292. [13] M. Bär, B.-A. Schubert, B. Marsen, R.G. Wilks, M. Blum, S. Krause, S. Pookpanratana, Y. Zhang, T. Unold, W. Yang, L. Weinhardt, C. Heske, H.-W. Schock, J. Mater. Res. 27 (2012) 1097. [14] P.L. Cook, X. Liu, W. Yang, F.J. Himpsel, J. Chem. Phys. 131 (2009) 194701. [15] E.M. Mannebach, J.W. Spalenka, P.S. Johnson, Z. Cai, F.J. Himpsel, P.G. Evans, Adv. Funct. Mater. (2012); http://dx.doi.org/10.1002/adfm.201201548 [16] J.F. Morar, F.J. Himpsel, G. Hollinger, G. Hughes, J.L. Jordan, Phys. Rev. Lett. 54 (1985) 1960. [17] P.-A. Glans, T. Learmonth, K.E. Smith, S. Ferro, A. De Battisti, M. Mattesini, R. Ahuja, J.-H. Guo, Appl. Phys. Lett., submitted for publication. [18] J.F. Morar, F.J. Himpsel, G. Hollinger, J.L. Jordan, G. Hughes, F.R. McFeely, Phys. Rev. B 33 (1986) 1346. [19] K. Bobrov, G. Comtet, G. Dujardin, L. Hellner, P. Bergonzo, C. Mer, Phys. Rev. B 63 (2001) 165421. [20] R.E. Ruther, M.L. Rigsby, J.B. Gerken, S.R. Hogendoorn, E.C. Landis, S.S. Stahl, R.J. Hamers, JACS 133 (2011) 5692. [21] P.L. Cook, W. Yang, X. Liu, J. María, G. Lastra, A. Rubio, F.J. Himpsel, J. Chem. Phys. 134 (2011) 204707. [22] Y. Zhang, S. Wang, T. Learmonth, L. Plucinski, A.Y. Matsuura, S. Bernardis, C. O’Donnell, J.E. Downes, K.E. Smith, Chem. Phys. Lett. 413 (2005) 95. [23] Y. Alfredsson, H. Rensmo, A. Sandell, H. Siegbahn, J. Electron Spectrosc. Relat. Phenom. 174 (2009) 50. [24] F. de Groot, Coord. Chem. Rev. 249 (2005) 31. [25] J.M. García-Lastra, P.L. Cook, F.J. Himpsel, A. Rubio, J. Chem. Phys. 133 (2010) 151103. [26] R.K. Hocking, E.C. Wasinger, Y.-L. Yan, F.M.F. deGroot, F. Ann Walker, K.O. Hodgson, B. Hedman, E.I. Solomon, JACS 129 (2007) 113. [27] P.L. Cook, P.S. Johnson, X. Liu, A.-L. Chin, F.J. Himpsel, J. Chem. Phys. 131 (2009) 214702. [28] E.M.J. Johansson, M. Odelius, S. Plogmaker, M. Gorgoi, S. Svensson, H. Siegbahn, H. Rensmo, J. Phys. Chem. C 114 (2010) 10314. [29] P.S. Johnson, P.L. Cook, I. Zegkinoglou, J.M. Garcia-Lastra, A. Rubio, R.E. Ruther, R.J. Hamers, F.J. Himpsel, J. Chem. Phys., submitted for publication. [30] G. Bauer, K.A. Kirchner, Angew. Chem. Int. Ed. 50 (2011) 5798. [31] I. Zegkinoglou, M.-E. Ragoussi, P.S. Johnson, D. Pickup, J. Enrique Ortega, G. de la Torre, F.J. Himpsel, submitted for publication. [32] I. Zegkinoglou et al., submitted for publication. [33] R. González-Moreno, P.L. Cook, I. Zegkinoglou, X. Liu, P.S. Johnson, W. Yang, R.E. Ruther, R.J. Hamers, R. Tena-Zaera, F.J. Himpsel, J.E. Ortega, C. Rogero, J. Phys. Chem. C 115 (2011) 18195. [34] V. Chauhan, R. Hatton, P. Sullivan, T. Jones, S.W. Cho, L. Piper, A. deMasib, K. Smith, J. Mater. Chem. 20 (2010) 1173. [35] C.X. Kronawitter, I. Zegkinoglou, C. Rogero, J.-H. Guo, S.S. Mao, F.J. Himpsel, L. Vayssieres, J. Phys. Chem. C (2012); http://dx.doi.org/10.1021/jp308918e [36] Y. Tachibana, L. Vayssieres, J.R. Durrant, Nat. Photonics 6 (2012) 511. [37] T. Lindgren, H. Wang, N. Beermann, L. Vayssieres, A. Hagfeldt, S.E. Lindquist, Sol. Energy Mater. Sol. Cells 71 (2002) 231. [38] H. Wang, J. Turner, ECS Trans. 25 (2010) 49. [39] L. Vayssieres, C. Sathe, S.M. Butorin, D.K. Shuh, J. Nordgren, J.-H. Guo, Adv. Mater. 17 (2005) 2320. [40] N. Beermann, L. Vayssieres, S.-E. Lindquist, A. Hagfeldt, J. Electrochem. Soc. 147 (2000) 2456.

Author's personal copy F.J. Himpsel et al. / Journal of Electron Spectroscopy and Related Phenomena 190 (2013) 2–11 [41] C.X. Kronawitter, L. Vayssieres, S. Shen, L. Guo, D. Wheeler, J.Z. Zhang, B.R. Antoun, S.S. Mao, Energy Environ. Sci. 4 (2011) 3889. [42] E.F. Aziz, S. Eisebitt, F. de Groot, J.W. Chiou, Ch. Dong, J. Guo, W. Eberhardt, J. Phys. Chem. B 111 (2007) 4440. [43] L. Weinhardt, M. Blum, O. Fuchs, A. Benkert, F. Meyer, M. Bär, J.D. Denlinger, W. Yang, F. Reinert, C. Heske, J. Electron. Spectrosc. Relat. Phenom. (2012); http://dx.doi.org/10.1016/j.elspec.2012.10.006, in press. [44] D.F. Ogletree, H. Bluhm, G. Lebedev, C.S. Fadley, Z. Hussain, M. Salmeron, Rev. Sci. Instrum. 73 (2002) 3872.

11

[45] C. Zhang, M.E. Grass, A.H. McDaniel, S.C. DeCaluwe, F. El Gabaly, Z. Liu, K.F. McCarty, R.L. Farrow, M.A. Linne, Z. Hussain, G.S. Jackson, H. Bluhm, B.W. Eichhorn, Nat. Mater. 9 (2010) 944. [46] S. Meng, E. Kaxiras, Md.K. Nazeeruddin, M. Grätzel, J. Phys. Chem. C 115 (2011) 9275. [47] F. Ambrosio, N. Martsinovich, A. Troisi, J. Phys. Chem. Lett. 3 (2012) 1531. [48] F. Labat, T. Le Bahers, I. Ciofini, C. Adamo, Acc. Chem. Res. 45 (2012) 1268.