Organometal halide perovskites have tremendous potential as light absorbers for photovoltaic applications. In this work we demonstrate hybrid solar cells based on the mixed perovskite CH3NH3PbI2Cl in a thin film sandwich structure, with unprecedented reproducibility and generating efficiencies up to 10.8%. The successfulness of our approach is corroborated by the experimental electronic structure determination of this perovskite. Hybrid solar cells based on organometal halide perovskites (e.g., CH3NH3PbI2Cl) as light absorbers are upcoming new players in the field of third-generation photovoltaics.1 Remarkably, this new type of solar cells has surpassed dye-sensitized and organic solar cells in terms of efficiency on a very short timescale, ranging from 3% to typically 8% on average, with a maximum reported value of 15%.2-11 Regarding their architecture, the active layers of perovskite photovoltaic devices are often similar to that of a classical hybrid solar cell: a mesoporous (‘meso-superstructured')6 metal-oxide framework is infiltrated with the perovskite aiming to maximize the interfacial area between the two materials.12 Surprisingly, it could be demonstrated very recently that efficient solar cells can be realized in a planar junction configuration as well, where the perovskite film is sandwiched between selective electrodes, obsoleting the need for a nanostructured acceptor.11, 13 Despite the resulting simplification of the fabrication process, where homogeneous infiltration of the perovskite precursor is no longer an issue, still very large efficiency fluctuations are observed between nominally identical samples with standard deviations approaching even 50% of the average value.13 Besides this poor reproducibility, issues also arise from a rather limited knowledge of even basic properties of the perovskite, which especially holds true for its electronic structure. The latter, however, is one of the key aspects controlling the proper functioning of any solar cell. Therefore, the aim of this contribution is twofold: Firstly, we demonstrate highly reproducible and efficient organometal halide perovskite solar cells in a planar sandwich configuration exceeding 10% efficiency, by optimizing the processing of the perovskite precursor. Secondly, the valence band structure of the absorber material is studied by combining different photoelectron spectroscopic methods. This allows to critically compare with recent theoretical work14 and, consequently, helps to derive a better picture about the distribution of electronic states in this new class of absorber materials. To date, the most efficient perovskite-based absorber materials are the mixed halides CH3NH3PbI2X with X representing I, Br, or Cl.14 In any solar cell, the energy gap of the absorber material determines the amount of absorbed sunlight and, thus, its efficiency to convert light into electricity. CH3NH3PbI2Cl was chosen as absorber material since it offers the smallest gap in the electronic density of states in this material family.14 As CH3NH3PbI2Cl simultaneously offers excellent hole and electron conduction,6 that could give rise to leakage currents, a solar cell based on this material requires hole and electron blocking layers at adjacent interfaces. This can be realized by sandwiching a perovskite film between TiO2 and poly-3-hexylthiophene (P3HT) layers serving as electron (TiO2) as well as hole (P3HT) conducting films, respectively (see Figure 1). P3HT is a particularly appropriate choice as hole collecting layer because it allows the full device fabrication to be done in nitrogen atmosphere, thereby protecting the humidity-sensitive perovskite. This cannot be done with the commonly used Spiro-OMeTAD, which requires exposure to ambient atmosphere for proper functioning, thus at the same time risking to degrade the perovskite.15 It has to be pointed out that, by combining perovskite with mesoporous TiO2 and P3HT, efficiencies of 3.8%,8 4.5%,16 and 6.7%12 could be achieved very recently. Using our thin-film concept, this value can be pushed to 10.8% as will be demonstrated in the following. A recent report announced the achievement of planar heterojunctions of CH3NH3PbI3 and fullerene derivatives with a maximum efficiency of 3.9%.17 This limitation mainly resulted from severe constraints in film thickness, where thicker films were found to be too course for successful solar cell operation. Another very recent report demonstrated more efficient planar cells based on CH3NH3PbxCl3-x (11.4% for the best device) by optimizing perovskite coverage, however standard deviations are observed of up to almost 50% of the average value.13 Our strategy yields reasonably smooth films with up to >95% coverage and thicknesses exceeding 200 nm, thus enabling light absorption close to 100%, and resulting in very reproducible efficiency values. This is achieved by using highly concentrated precursors in combination with optimized drying conditions (through spincoating). To this end, dimethyl sulfoxide (DMSO) is exploited as a solvent which, as compared to the commonly used solvents dimethyl formamide (DMF) and γ-butyrolactone, enables to prepare precursor solutions with considerably enhanced concentrations of the perovskite materials, up to 60 wt%.18 Slightly higher concentrations are still possible but were found too viscous for reliable spincoating. The crystalline quality of the resulting perovskite films is demonstrated in Figure 1c which displays the X-ray Diffraction (XRD) spectrum corresponding to a precursor concentration of 60 wt%. Four characteristic lines can be identified (14.10°, 28.47°, 43.27° and 58.88°) and assigned to the 110, 220, 330 and 440 diffraction peaks of a mixed-halide perovskite with an orthorombic crystal structure6 thus evidencing the presence of a phase-pure material with no indications for the admixture of other phases. Full experimental details on synthesis, device fabrication and characterization are to be found in the Supporting Information. Figure 2a provides JV-curves of solar cells derived from different precursor concentrations and spincoating speeds. Table 1 summarizes relevant photovoltaic parameters as well as the nominal thickness of the corresponding perovskite layers. For precursor concentrations ranging between 22.5 and 45 wt% (spincoated at 1500 rpm), the optimum is reached at 37.5 wt% yielding an efficiency of 8.5%. Interestingly, a value of 10.8% (Jsc = 21.3 mA·cm−2, Voc = 0.932 V, fill factor (FF) = 0.544) can be reached by significantly increasing the precursor concentration to 60 wt% which, however, requires an enhanced spin speed of 3000 rpm to limit the active layer thickness (solid line in Figure 2a). The latter approach allows to produce films of roughly the same nominal thickness as obtained from a 45 wt% solution spincoated at 1500 rpm, hence still warranting effective charge transport, but with higher ‘optical' density. The concentration-dependent trend in performance can be explained by taking a closer look at the optical and morphological properties of the perovskite films: Increasing the concentration from 22.5 to 30 wt% simply results in a thicker film and, thus, an augmented magnitude of the absorption (see Figure 2b) as well as of the short-circuit current Jsc, while the morphological features look quite similar (Figure 2c). Further increase in concentration to 37.5 wt% still induces a higher total absorption (again corroborated by a higher Jsc), albeit with a slight change in spectral shape. This latter observation clearly correlates with a transition in film morphology to a new regime that is now dominated by the formation of overlaying needle-like crystallites. Finally, when increasing the concentration to 45 wt%, absorption is partly lost below 600 nm, making the Jsc stagnant. The morphology in this case exhibits both needles and smaller crystallites, forming a more rough film, which could explain the slightly reduced fill factor. Notice that the shape of the absorption spectra of the 22.5 wt%, 30 wt% and 60 wt% perovskite films are identical, while the 37.5 wt% and the 45 wt% lack the features at lower wavelength, thereby ultimately limiting the absorption capability. This could be the consequence of a rather complicated scattering behavior, resulting from widths of needle-like perovskite structures (in the 37.5 and 45 wt% case) that partly match with the wavelengths of absorption. The observed complex absorption-morphology relation, together with the possibility of a more efficient cell using a 60 wt% precursor solution that is spincoated faster, strongly suggests that film formation during solvent evaporation is the key parameter that has to be tuned to create a favorable morphology for high efficiency devices.13, 17 Moreover, it is important to point out that, while large fluctuations in efficiency are typically observed in perovskite solar cells,6, 13 our fabrication method results in very little performance variation between different cells and batches (see Table 1). The open circuit voltage Voc of our optimized devices exceeds 0.9 V (Table 1) which is the highest value reported for solar cells based on organometal halide perovskite as light absorber using TiO2 and P3HT as hole and electron blocking layers, respectively. The average Jsc values found here are comparable to the highest values reported in literature, and the Jsc of our best device (21.3 mA·cm−2) nearly matches the highest ever reported Jsc for vapor deposited perovskite-based solar cells.11 The exciton binding energy for the comparable CH3NH3PbI3 was found earlier to be 37 meV19 which, as compared to typical values for organic solar cells of ∼0.5 eV,20 is not too far away from the case of silicon (∼15 meV),21 thus pointing to a considerable amount of free electrons within the perovskite due to thermal dissociation of excitons at ambient temperature. In addition, we also observe that excitons are efficiently dissociated at both the TiO2 and the P3HT interface, as shown by the photoluminescence spectra in Figure 2d, which is in accordance with their weakly bound Wannier-type nature:19 The characteristic photoluminescence of the pure perovskite is considerably lowered in intensity when in contact with a layer of TiO2, and completely absent when in contact with P3HT. This efficient quenching at either interface likely assists in generating the observed high photocurrents.8 A small additional contribution to the photocurrent is probably also coming from the absorption of the P3HT, as inferred from the shape of the external quantum efficiency (EQE) spectrum of the best device (inset of Figure 2a).22 The parameter that still limits the efficiency of perovskite photovoltaics realized in planar design is the fill factor, which is lower than that found for perovskite-infiltrated mesoporous metal-oxide based cells. As argued before, the most likely explanation for this is the fact that the perovskite films are not pinhole-free (as shown by Figure 2c), inevitably leading to parasitic loss currents and thereby decreasing the parallel resistance of the devices.13 Secondly, the selective contacts formed by TiO2 and P3HT are not yet optimized for minimal resistance and, therefore, present another source of loss in fill factor. Inferior P3HT purity is sometimes suggested to affect charge transport, however an X-ray Photoelectron Spectroscopy (XPS) survey scan shows that the material used here is free of impurities (Figure S1, Supporting Information).23 Turning our attention now towards a more detailed chemical analysis by XPS of the best performing devices (60 wt% precursor concentration), a composition of C:N:Pb:I:Cl = 1.04:1:1.05:2.02:0.99 can be derived by integrating appropriate core level spectra representing the different atomic species (with the exception of hydrogen). This experimental result agrees very well with the expected nominal stoichiometry (1:1:1:2:1), thus indicating the presence of a high quality material that can be used to further explore its electronic structure. Here, Ultraviolet Photoelectron Spectroscopy (UPS) has recently been exploited to derive fundamental photovoltaic parameters like, e.g., the position of the valence band maximum or the work function/electron affinity of the perovskite.4, 5 However, a more detailed evaluation of its electronic density of states (DOS) is still completely missing, despite its relevance to better understand all of the perovskite's physical properties. In order to unravel the electronic structure of the perovskite layer, UPS-based as well as XPS-based valence band measurements were carried out, the latter allowing to be less surface sensitive (information depth 2–5 nm) as compared to the former technique (information depth 0.5–1 nm). By using UV- or X-rays to excite photoelectrons in a non-single crystalline sample, information about the DOS of the material can be gained since the momentum selection rules are averaged out. Thus, UPS/XPS spectra taken from such samples represent the sum of angular-momentum projected partial densities of states, weighted by the corresponding photoionization cross sections. As these cross sections also strongly depend on the photon energy used in the experiment,24 XPS is generally believed to better represent the total DOS of a sample due to less differences in photoexcitation probability between electron wave functions of different symmetry. Furthermore, even in case of single-crystalline samples, XPS-derived valence band spectra can safely be interpreted in terms of the electronic DOS since indirect transitions are dominating the photoemission process. Figure 3a summarizes the photoelectron distribution curves acquired on a CH3NH3PbI2Cl perovskite film (spincoated from a 60 wt% precursor at 3000 rpm, as in the best devices) using UV- (hν = 21.2 eV) as well as X-rays (hν = 1486.6 eV) for photoexcitation, in comparison with recent theoretical work.14 Starting with the XPS valence band structure (bottom curve), a total band width of about 4.5 eV is detected, accompanied by 2 dominating spectral features at binding energies of 4.7 and 3.2 eV and a weak shoulder at 2.3 eV, respectively. A dashed line has been added to identify the secondary electron contribution to the spectrum (Shirley type of background).25 In case of UPS (middle curve), similar overall spectral features can be observed, superimposed by a significantly increased background of inelastically scattered electrons. Clearly, both photoemission techniques reveal basically the same structure of the valence band consisting of mainly two sub-bands. This similarity furthermore indicates, due to the different information depths of both spectroscopies, that the perovskite is rather homogeneous in the distribution of different atomic species resulting in depth-independent electronic properties. UPS valence band spectra acquired on solar cells are often utilized to derive key electronic parameters like the energetic position of the valence band maximum (VBM). This position is usually extracted by extrapolating the electron distribution curves decaying towards the band edge. As UPS spectra often suffer from satellite line contributions superimposing the photoelectron intensity close to the band edge, to precisely determine the VBM is a difficult task. Therefore, the XPS spectrum acquired using monochromatic X-rays is preferred for this purpose (Figure 3b). By extrapolating the decaying intensity towards zero (represented by the horizontal base line), a VBM position of 1.62 eV is extracted with respect to the Fermi energy, which is marginally smaller than the energy gap (1.64 eV) in the electronic density of states predicted by density functional theory (DFT) for the mixed halide perovskite CH3NH3PbI2Cl.14 Consequently, this material should be best described as n-type semiconductor with a very shallow donor level, the latter pinning the Fermi energy close to the conduction band edge (see below). This would support its high conductivity expressed by its extraordinarily high JSC value found in this work. Based on the extracted VBM position, a direct comparison between experimental and theoretical DOS is now possible, as it allows to adjust the binding energy scales to each other. As can be observed in Figure 3a, a rather good agreement is obtained between experimental and calculated DOS regarding the overall band shape and the band width, the latter being slightly underestimated by theory. It is important to note that, during photoemission, the finite lifetime of the photoionized states as well as the finite energy resolution of the electron spectrometer both lead to a smearing out of characteristic structures in the experimental DOS resulting in broadened spectral features as compared to the theoretical DOS. Besides, the good overall agreement between the experimental and calculated band structure permits to identify the two major sub-bands as being dominated by the 2 halides (sub-band centered at 3.2 eV) with a strong contribution by lead (sub-band centered at 4.7 eV). To this end, there still exists the important question as to which sub-system of the perovskite finally induces the pinning of the Fermi level close to the conduction band edge. To explore this, high-resolution core-level spectroscopy was performed to analyze the chemical neighborhood of the different atomic species. Figure 4 presents the Pb-4f7/2 core level spectrum acquired on a perovskite film deposited from a 60 wt% precursor solution. Beside the main line centered at 138.7 eV (reflecting Pb(+3)), a weak spectral feature is detected at a binding energy of 136.9 eV thereby evidencing the presence of elemental Pb(0) in the sample. Elemental Pb could in fact easily act as an electron donor in the semiconducting perovskite. In order to derive a more quantitative picture, the two spectral contributions to the Pb-4f7/2 core level spectrum have been integrated. That way, and by taking into account the number of atoms per unit cell of the perovskite, a concentration of 230 ppm of elemental Pb is obtained which would be consistent with a highly doped, n-type semiconductor. However, further studies with more sensitive probes (like Secondary Ion Mass Spectrometry (SIMS)) are required to get a more complete picture about the distribution of elements in the perovskite which are beyond the detection limit of XPS. In addition, the role of hydrogen as possible n-type dopant is also unclear as its concentration cannot be accessed by XPS. In conclusion, we report the fabrication and electron spectroscopic analysis of efficient hybrid solar cells based on the mixed metal halide perovskite absorber CH3NH3PbI2Cl, sandwiched between selective contacts of TiO2 and P3HT. Highly concentrated precursors allow for the deposition of optically dense perovskite films of >200 nm thickness, and excitons therein can be dissociated either thermally or when scattered at both the TiO2 and the P3HT interfaces. The resulting efficiencies of up to 10.8% are on par with very recent work on thin-film sandwich perovskite solar cells but can, in contrast, be obtained with little fluctuations. Finally, we observe a very high short circuit current Jsc, (>21 mA cm-2) which might also be related to an n-type semiconducting behavior combined with a shallow donor level of the absorber material as suggested by our photoemission results. This work was financially supported by BOF (Hasselt University), the Research Foundation Flanders (FWO) within the Odysseus program, and the Interreg project Organext. B.C. and C.D. are postdoctoral research fellows of the FWO. The authors thank dr. Henry Snaith (University of Oxford) and dr. Matthew Carnie (Swansea University) for fruitful discussions. As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.
