The impact of ultra-thin titania interlayers on open circuit voltage and carrier lifetime in thin film solar cells

Recent analyses have indicated that organic photovoltaics (OPVs) could realize the lowest overall environmental impact and fastest energy payback time among both existing and emerging solar cell technologies.¹ Record efficiencies for organic solar cells now reach power conversion efficiencies (PCEs) as high as 11.7% and 11.5% for single² and triple³ junction devices, respectively. These record devices have been achieved in inverted bulk heterojunction (BHJ) configurations where the electrons are extracted through a bottom electrode made of modified transparent conducting oxide (TCO), and the holes through a top electrode made of a high work function metal (gold or silver). These inverted device structures are also more stable than standard devices for several reasons, including elimination of both acidic poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS), which can slowly degrade the TCO and the top electrodes made of a low work function metal (such as aluminum) that can easily oxidize.^4–7 The issue of long-term stability remains an important challenge in the field of emerging thin-film PV materials to compete with conventional inorganic-based technologies in long-term applications.⁸ 

Inverted OPV device structures commonly adopt interlayers to modify the transparent conducting oxide in order to alter the work function of the TCO and improve the selectivity of charge extraction (CE). In this regard, zinc oxide (ZnO) and titania (TiO₂) layers are among the most promising interfacial layers to lower the work function of the TCO, such as indium tin oxide (ITO).⁷ However, the properties and performance of such interlayers can be sensitive to film thickness and growth methods.

Plasma-enhanced atomic layer deposition (PE-ALD) is an interesting method for growing conformal, pinhole-free interlayer films of tailored composition and carefully controlled thickness.^9–15 Here, we investigate the device physics of ITO electrodes modified with ultra-thin TiO₂ films deposited by PE-ALD. We show that the performance of BHJ solar cells based on blends of the polymer poly[N-9′-heptadecanyl-2,7-carbazole-alt-5,5–(4′,7′-di-2-thienyl-2′,1′,3′-benzothiadiazole)] (PCDTBT) and the fullerene derivative [6,6]-phenyl C71-butyric acid methyl ester (PC₇₁BM) depend on the TiO₂ layer thickness. We chose PCDTBT as a model p-type semiconducting polymer, because it is often used as a model of newer donor/acceptor polymers with high open-circuit voltages^16–19 and has achieved reasonable efficiencies (from 3.6% to 7.5% after device optimization), due to high open-circuit voltages (V_OC) and good device stability.^20,21

We grew TiO₂ layers on cleaned glass/ITO substrates by using PE-ALD to obtain varying thickness (1, 3, 5, 7, and 10 nm) using protocols previously described by Kim et al.²² We completed the OPV devices by spin-coating the active layer (45 μl/cm² at 5000 rpm for 120 s) from a PCDTBT:PC₇₁BM solution (1:4 weight ratio) and by drying the films at 70 °C for 20 min leading to an 80-nm-thick active layer (see Ref. 23 for details). Finally, we deposited the top (hole-collecting) contacts of molybdenum-(VI) oxide (10 nm) and silver (100 nm) by thermal evaporation. We performed the entire fabrication process in a nitrogen filled glovebox (O₂ and H₂O levels < 10 ppm).

Fig. 1 presents the performance of solar cells fabricated on bare ITO and ITO modified with different thicknesses of TiO₂ layer. Using only 1 nm of TiO₂ increases the V_OC to 0.83 V compared to 0.66 V for devices made with bare ITO. The V_OC then plateaus around 0.88–0.89 V between 3 and 7 nm of TiO₂. We note these values compare with the highest V_OC's reported in the literature for these carbazole-based devices.^20,21 However, when we increase the titania interlayer thickness further, the V_OC then dramatically drops to 0.39 V for devices with 10-nm-thick TiO₂ interlayers. We observe a similar trend for the fill-factor (FF), obtaining a maximum of 58% for 5-nm-thick TiO₂. The short-circuit current densities (J_SC) follow the same trend, but the drop is less pronounced for 10-nm-thick TiO₂. We thus measured the highest PCEs for 3- and 5-nm-thick TiO₂ while we measured the lowest PCE for the 10-nm-thick TiO₂ device, with the main loss in PCE due to the substantial decrease in V_OC. We note that the J_SC values of our devices fabricated on ultra-thin titania are around 20% lower compared to champion PCDTBT-based devices; however,^21,24,25 since our values were within the range of published values,^7,20,26,27 and since the purpose of this study was to compare relative changes in device performance with changes in interlayer properties, we did not pursue other optimizations such as solvent additives,²⁸ solvent mixtures,²⁹ or use of tailored optical interference.³⁰ 

To better understand these variations in device performance with varying titania interlayer thickness, we characterized bare ITO/TiO₂ electrodes using scanning Kelvin probe microscopy (SKPM) (see Ref. 31 for details). This high-resolution technique simultaneously measures the topography and work function difference (ΔΦ) between a metal-coated AFM tip and the sample,³² also called the contact potential difference (CPD = ΔΦ = Φ_{AFM tip} − Φ_sample). Fig. 2 (top row) exhibits uniform and smooth topography for each ITO sample modified with 1, 3, and 10 nm thick TiO₂, with root-mean-square (RMS) roughness of 379, 328, and 315 pm, respectively. The corresponding CPD images are shown in Fig. 2 (bottom row). We find an overall increase in CPD with increasing TiO₂ thickness progressing from 236 mV for a 3-nm-thick titania layer to 304 mV for a 10-nm-thick titania layer, indicating a shallower Φ for thicker TiO₂ layers.

Fig. 3 shows the CPD distributions from the CPD images in Fig. 2. We fit the distributions with Gaussian profiles to obtain the mean CPD and half-width values. Since SKPM only provides a relative work function, we use the work function value from recent ultraviolet photoemission spectroscopy (UPS) measurements²² on samples prepared under identical conditions using the same equipment, to assign a reference value of Φ = 3.7 eV to the ITO coated with 3 nm of TiO₂. By using the mean values from Fig. 3, we can determine the Φ values for 3-, 5-, and 10-nm-thick TiO₂ to be ∼3.70,²² ∼3.66, and ∼3.63 eV, respectively. The half widths of each were 4.7, 4.5, and 8.5 mV. Notably, the 10 nm thick TiO₂ electrode also exhibits the broadest distribution of Φ values, which is directly associated with increased spatial heterogeneity in the CPD. It has been previously shown that ALD deposited TiO₂ films remain amorphous until 5 nm thick, after which a transition occurs between 7 and 8 nm of thickness, and become predominantly anatase at 10 nm.^22,33 High-resolution conductive AFM measurements also reveal local high conductivity spots in such crystalline films.²² Our results indicate that these changes also lead to a larger variance in CPD across the surface of the 10-nm-thick TiO₂ films. Since the work function decreased monotonically with increasing TiO₂ thickness, the work function alone cannot explain this large drop in device performance we observed as the film thickness is increased past 5–7 nm. We suspected that this change was associated with changes in interfacial recombination kinetics upon the transition from amorphous to anatase TiO₂.

To explore this question further, we performed SKPM measurements on ITO/TiO₂ (5 and 10 nm) coated with active layers to study the local surface photovoltage (SPV) and the local recombination rate (or local carrier lifetime τ). We previously showed that the properties of the charge extracting contact significantly affects the V_OC and both the bulk and local recombination rates that can be determined by SKPM experiments.^23,34 We measured SPV by taking the difference in surface potential under constant illumination and in the dark,^35–38 and we used intensity modulated SKPM (IM-SKPM) to measure the local recombination rate. IM-SKPM was previously used by Takihara et al. to measure local sub-ms carrier lifetimes on inorganic photovoltaic microstructures.³⁹ This experiment is conducted in the frequency domain by measuring the time averaged surface potential, while the intensity and frequency of the probing source are modulated. We recently showed that IM-SKPM is also suitable to measure local recombination rates in OPVs, in some cases producing quantitative agreement with macroscopic device measurements.²³ 

Fig. 4 shows the local surface photovoltage (a) and carrier lifetime distributions (b) obtained by frequency-modulated SKPM (cf. Ref. 23 for specific experimental details). Table I summarizes the means and the widths of the distributions as shown in Fig. 4. First, the 5-nm-thick TiO₂ device has a SPV that is 6 ± 0.5 mV higher than the 10 nm thick titania layer. This result is in qualitative agreement with the device measurements (i.e., higher V_OC for the 5 nm thick TiO₂ device). The local carrier lifetime measured with IM-SKPM is also longer for the 5-nm-thick TiO₂ device. While qualitatively consistent with the better performance of the 5-nm-thick TiO₂ device, both of these changes are fairly small, however. Since SPV contrast can arise both from charge transfer at organic/TiO₂ interface^23,35 as well as vertical phase separation in the polymer:fullerene layer,^35,40–42 we conclude that these effects are masking much of the change in buried interface properties in these inverted devices.

To better understand how the changes in TiO₂ layer thickness affect performance of the completed devices, we performed transient photovoltage (TPV) and CE measurements to investigate the recombination dynamics of the PCDTBT:PC₇₁BM devices with different TiO₂ layer thicknesses under open-circuit conditions. The experimental methods are described elsewhere (Refs. 43–45 and specific details of our setup are reported in Ref. 34). Briefly, TPV uses a weak light pulse to induce a small excess of charge carriers in a device held at V_OC and measures the subsequent photovoltage transient to determine the pseudo-first-order recombination lifetime (τ_Δn). We performed TPV measurements at fixed background carrier densities (n), generated by constant illumination from a white-light bias at varying light intensities. We performed CE measurements to determine the carrier densities corresponding to each light intensity from the TPV measurements. Taken together, these techniques allow us to find the charge carrier lifetimes as a function of carrier density for each device.

Fig. 5 shows τ_Δn as a function of charge carrier density within the devices. The 3-, 5-, and 7-nm-thick TiO₂ devices show longer recombination lifetimes (lower recombination loss) and higher carrier densities compared to the device with 10 nm thick TiO₂, consistent with their higher V_OC and FF. The device with 5 nm thick TiO₂ exhibits longer carrier lifetimes compared to the 3 and 7 nm thick TiO₂ devices, in agreement with the highest FF for the 5 nm thick TiO₂. The SKPM data of Figs. 2 and 3 show a reduction in work function as the TiO₂ thickness increases, therefore lowering the energy barrier between the conduction band of the TiO₂ and the LUMO of PC₇₁BM. This lowering should also reduce the hole concentration near the electron-extracting contact by increasing the built-in electric field within the active layer to drive holes more efficiently toward the top, hole-extracting, contact. Therefore, the increase in lifetime from 3- to 5-nm-thick TiO₂ is most likely a result from a reduced recombination in the BHJ near the TiO₂ interface due to more favorable band bending.

The TPV and CE data indicate a rapid decrease in carrier lifetime associated with the drop in V_OC and FF occurring for the 10-nm-thick TiO₂ films. As is typical for organic solar cells, the pseudo-first order lifetime τ_Δn is strongly dependent on carrier density, with an apparent reaction order higher than 2.^27,46,47 Nevertheless, these results suggest that, somehow, the carrier recombination kinetics at the organic/TiO₂ layer become much less favorable as the TiO₂ layer grows from ∼7 to 10 nm thick. Interestingly, previous work has shown that over this same range, the ALD-grown TiO₂ films convert from conformal amorphous coatings to polycrystalline anatase TiO₂.^22,33 Indeed, a change from amorphous to anatase TiO₂ is consistent with our own scanning probe data from Figs. 2 and 3, showing an increase in heterogeneity in the work function, smaller photovoltage, and shorter carrier lifetime for the 10-nm-thick TiO₂ film. We interpret these results as a strong support for the hypothesis that the transition from amorphous to anatase TiO_2, known to occur over this thickness range^22,33 has an adverse effect on interfacial recombination kinetics. We conclude that the decrease in performance for the devices with 10-nm-thick TiO₂ interlayer occurs, because the anatase TiO₂ films exhibit less favorable recombination kinetics than the amorphous TiO₂ interlayers. Although further studies will be required to understand this unfavorable change in recombination kinetics upon conversion of the TiO₂ layer from amorphous to anatase, we speculate that the newly formed grain boundaries in the crystalline film could contribute to increased recombination current, consistent with recent study by Dang et al.⁴⁸ who have reported that the grain boundaries of anatase films can contribute to increased conductivity or leakage current of TiO₂.

In conclusion, we have studied the effects of PE-ALD-grown titania layers of different thickness on the performance parameters of inverted OPV devices. We find that there is an optimal titania thickness while the film is still amorphous and that the performance decreases once the film becomes thick enough to convert to anatase. The TiO₂ thickness influences OPV device performances primarily through V_OC and FF. We achieved the highest PCEs for devices with 3 and 5 nm thick TiO₂, when the crystal phase is amorphous and the work function in SKPM is homogeneous. As the TiO₂ thickness increases to 10 nm and develops into the anatase phase, we observe that the work function becomes heterogeneous and there is a decrease in device PCE mainly through a drop in V_OC and FF. We conclude that the recombination kinetics are not only affected by the interlayer work function, but that the phase of the interlayer plays a critical role, with more uniform amorphous TiO₂ interlayers having superior properties as electron extracting/hole blocking contacts compared to the more heterogeneous and polycrystalline anatase films that emerge for thicker PE-ALD-grown films. There are substantial literatures on the effect of titania crystal structure on recombination kinetics in dye-sensitized solar cells,^49–51 but little if any work on conformal amorphous films. This work suggests that such amorphous layers have desirable properties as interlayers for solid state devices.

This manuscript is based on work supported in part by the Office of Naval Research (N00014-14-1-0170) (D.S.G. and D.M., device measurements, microscopy, and TPV and CE measurements) and the Department of Energy through Bay Area Photovoltaics Consortium (DE-EE0004946) (PE-ALD TiO₂ growth by S.G. and H.K.). D.M. is grateful to B.A.E.F. for a postdoctoral fellowship. The authors thank Dr. Micah S. Glaz, and Jeffrey S. Harrison for helpful discussions regarding the SKPM measurements.