Supplementary Information  S1 – Sample fabrication    Fabrication  of  nanoporous  aluminum  oxide  templates:

Supplementary Information S1 – Sample fabrication Fabrication of nanoporous aluminum oxide templates: A two‐step anodization processes using high purity (99.999%) aluminum sheets (ESPI Metals) yielded highly ordered nanoporous aluminum oxide films. All anodization was performed at 40 V, at room temperature in 0.4 M oxalic acid. After anodizing for one‐hour, the resulting oxide layer (~12 μm thick) was stripped using a 0.4 M phosphoric acid/0.2 M chromic acid solution heated at 85 °C for 45 minutes. The substrates were anodized a second time for 42‐90 s, depending on the desired template thickness (100‐300 nm). All templates were then etched in 5 wt% phosphoric acid for 30 minutes at room temperature in order to widen the pore diameters to ~50‐60 nm. Nanoporous aluminum oxide pattern transfer procedure: The completed, hexagonally ordered aluminum oxide templates were transferred to device substrates using the following procedure. A protective PMMA layer (Microchem 950PMMA C4) was spin cast on top of the template at 1000 rpm and subsequently baked on a hotplate at 180 °C for 90 s. The aluminum substrate was removed using a solution of 20 ml H2O, 20 ml HCl, 1 g CuCl2, leaving only the aluminum oxide template and the protective PMMA layer. After rinsing in water, the structure was then etched in 5 wt% phosphoric acid for 2 hours to expose the pore bottoms. During this step the PMMA layer protected the pore interiors until the pore bottoms were opened. There is additional pore widening after the etchant enters the template and begins to etch the pore interiors. This typically results in ~10 nm of additional pore widening. Finally, the templates are transferred to an acetone bath where the PMMA is dissolved, leaving only the template behind. The templates are physically transferred to the desired device substrate and blown dry using nitrogen. Solar cell fabrication: All devices are fabricated on commercially purchased ITO‐coated (150 nm) glass substrates. V2O5 hole contact layers (15 nm) were deposited via thermal evaporation from a powder source material. In templated devices, the nanoporous template was transferred onto the V2O5 layer using the procedure described above. Prior to depositing the active material, samples were coated with hexamethyldisilazane (HMDS) by spin coating at 5000 rpm and baking on a hotplate at 115 °C for 90 s in order to aid wetting and pore filling of the P3HT:PCBM active layer. A 1:1 mixture of P3HT:PCBM (2 wt% in chlorobenzene) was then deposited by spin coating in ambient air. For templated devices, the spin speed was 1000 rpm, while for control planar devices the spin speed was varied to achieve the desired thickness (400‐1000 rpm for 200‐80 nm, respectively). Finally, aluminum top contacts were deposited via thermal evaporation, with the device contact size physically defined by a shadow mask (500 μm diameter circles). 1 Supplementary Information S2 – Effect of HMDS interfacial layer on device performance Figure S2_1. J‐V characteristics for control devices fabricated with (red) and without (blue) HMDS. The dashed curves are the same devices after a 10‐minute, 150 °C vacuum anneal. Before annealing, the HMDS greatly increases JSC of the device and reduces the overall series resistance. After annealing the HMDS device improves only slightly and the series resistance remains unchanged. Annealing the control device results in a low series resistance and comparable performance to the HMDS device. The blend film quality and thickness were similar for devices with and without HMDS. 2 Supplementary Information S3 – Confirmation of pore filling and analysis of the residual material Figure S3_1. Porous aluminum oxide template pore filling. (a) Cross‐sectional SEM image of a P3HT:PCBM blend in template pores without HMDS pretreating. (b) P3HT:PCBM blend in template pores with HMDS pretreating. HMDS treatement improves uniformity and generates smooth polymer rods. (c) Bright‐field TEM image of a filled template suspended on lacey carbon, showing uniform pore filling by the organic. 3 Figure S3_2. Sample preparation parameters were selected to minimize excess material on the template tops, irrespective of template height: (a) 210 nm, (b) 150 nm and (c) 120 nm. 4 Supplementary Information S4 – GIXRD analysis Grazing‐incidence x‐ray diffraction measurements were performed at the X9 endstation at the National Synchrotron Light Source (NSLS), Brookhaven National Laboratory. Two‐dimensional scattering images were acquired using an area detector, positioned ≈ 200 mm from the sample, and using an x‐ray wavelength of 0.0780 nm (photon energy of 15.90 keV). Data conversion to q‐space was accomplished by measuring a standard sample with known scattering features (Silver Behenate), and accounting for detector position and tilt angle. The incident beam was collimated using slits, and focused onto the sample position using a KB mirror system; the beam size at the sample position was approximately 100 μm horizontal width, and 80 μm vertical width. Measurements were performed at a variety of incident angles (0.10º, 0.15º, 0.20º, 0.30º, 0.40º, 0.50º); all measurements above the critical angle had similar features. The results presented in the manuscript used the data at 0.20º, which is well above the critical angle for the filled template (calculated to be 0.13º at 15.90 keV), and is thus representative of the entire film. Absorption effects are negligible (the absorption length of 15.90 keV x‐rays at 0.20º into Al2O3 is 1.6 μm, much larger than the film thickness). In order to quantify the orientation distribution of ordered P3HT, we integrated the 100 lamellar peak along the arc at |q| = 0.38 Å–1 at each angle (ω) with respect to the qz axis. The background at each ω was subtracted by fitting the peak to a Gaussian with a linear background; representative line‐cuts are shown in Fig. S4_1. This method removes the higher diffuse scattering observed in the templated samples. The full peak width was integrated at each ω, in order to account for variation in peak width between samples and with angle for a given sample. The ω scale was corrected to account for the intersection of the Ewald sphere with reciprocal‐space.1 Finally, to convert from detector intensity to the amount of material oriented at each angle, the integrated peak intensity was multiplied by sin(ω). The detector does not measure the full scattering in reciprocal space, therefore the sin(ω) geometric factor is required to correctly estimate the orientational distribution. For instance, all grains oriented out‐of‐plane satisfy the Bragg condition and contribute to the peak near the qz axis, whereas only a subset of the grains oriented in‐plane contribute to the intensity seen along the qr axis. 1
J.L. Baker, L.H. Jimison, S. Mannsfeld, S. Volkman, S. Yin, V. Subramanian, A. Salleo, A.P. Alivisatos, and M.F. Toney, Langmuir 26, 9146‐9151 (2010).
5 Figure S4_1. Radial reciprocal‐space line profiles taken at various angles with respect to the qz axis, comparing the scattering intensity for the first‐order lamellar packing peak in blends on planar control substrates (solid line), and confined in porous aluminum oxide pores (dashed line). The peaks are fit (red lines) in order to perform a Scherrer grain‐size analysis. The decrease in scattering intensity for the confined sample is not uniform with respect to angle. In particular, there is a large suppression of the population at 0º. This change in orientation distribution suggests that the confined material exhibits worse ordering, with a greater relative contribution of in‐plane lamellar packing compared to materials on a planar substrate. 6 Supplementary Information S5 – PCBM devices Devices for measuring PCBM properties consisted of PCBM films layered between V2O5 and Al electrical contacts (Figure S5_1b). For confined samples, successful filling of the porous template was confirmed by cross‐sectional SEM imaging. These structures display Schottky diode behavior due to an energy barrier for electron injection from V2O5 to PCBM (Al forms a low‐resistance contact). At high forward bias, the unconfined device conductivity is limited by space‐charge current through the PCBM, allowing us to determine the average electron mobility, 5×10‐4 cm2V‐1s‐1. This value is consistent with previous reports and ~6 times higher than our measured unconfined P3HT hole mobility. The conductance of confined PCBM devices is not well‐described by a SCLC model at high bias, thus preventing determination of the confined PCBM mobility by this method. However, the current density is reduced by a factor of 50 compared to an equivalent volume of unconfined material (Figure S5_1b), signifying a corresponding reduction in either electron mobility or carrier concentration. Figure S5_1c shows a summary of the extracted mobility values for both P3HT and PCBM in this study. While the unconfined PCBM mobility is higher than either than the unconfined P3HT, confining P3HT increases the mobility well above that of PCBM. This implies that PCBM electron mobility may be a limiting factor in the confined BHJ devices. Figure S5_1. Summary of SCLC mobility measurements for unconfined (blue) and confined (red) material. (a) J‐V data for P3HT‐only devices. Inset: Device band structure schematic. (b) J‐V data for PCBM‐only devices. Inset: Device band structure schematic. (c) Graph showing all extracted SCLC mobility values from this study. Error bars represent the standard deviation across at least four devices. 7 Supplementary Information S6 – Dark I‐V Figure S6_1. Dark I‐V data of P3HT:PCBM solar cells comparing unconfined (blue) and confined (red) active material. Under forward bias, the confined material show enhanced conductivity despite having a significantly smaller cross sectional area. Inset: Illuminated I‐V data for the same devices. This data has not been scaled and shows that performance for the two devices is similar despite having less active material in the confined sample. 8