NP and film properties

The metal oxide NP suspensions were synthesized by controlled oxidative dissolution of micron-sized metal powders in the presence of H₂O₂ and HCl, using a modification of published procedure.^[Reference Redel, Huai, Dag, Petrov, O'Brien, Helander, Mlynarski and Ozin^21] We found that by replacing the original solvent of water with n-butanol, and by reducing the metal concentration to 0.1 M, the diameter of the dispersed NPs measured using dynamic light scattering (DLS) was reduced from ~5 nm to 0.7 ± 0.1, 1.6 ± 0.4, 1.0 ± 0.2, and 0.9 ± 0.4 nm for MoO₃, WO₃, NiO_x, and CoO_x, respectively (Table I). The surfactant-free NP suspensions exhibit different colors arising from band gap and/or intraband d–d transitions of the metal oxides [Fig. 1(b) inset], and remain dispersed without change in DLS size for at least 2 weeks. Figure 1(a) depicts a typical transmission electron microscopy (TEM) image of NiO_x NPs, showing spherical particles with diameter of 1.3 ± 0.7 nm, consistent with the DLS result. High magnification image of a single NiO_x NP gives a characteristic spacing of 0.24 nm [Fig. 1(a), inset], which matches the spacing from (111) planes of NiO. Thus, oxidation of metal powder by H₂O₂ in n-butanol produces stable suspensions of ultrasmall metal oxide NPs.

Figure 1.

(a) TEM images of NiO_x NPs. Inset: high magnification image of single NP, showing lattice spacing of 0.24 nm. (b) Φ values for ITO and various films on ITO. The error bar for Φ measurement is 0.03 eV. Inset: photograph of metal oxide NP suspensions.

Table I.

Properties of metal oxide NP suspensions and films.

^aMeasured in suspension by DLS.

^bMeasured by UPS.

^cMeasured by PESA.

Metal oxide NP films spin coated on indium tin oxide (ITO) were characterized in air unless otherwise noted to examine their electronic properties. The Φ was measured using a scanning Kelvin probe, and was found to be 5.20 eV for MoO₃, 5.31 eV for WO₃, 5.11 eV for NiO_x, and 5.16 eV for CoO_x [Table I and Fig. 1(b)]. These values are significantly higher than the bare ITO (4.85 eV) and PEDOT:PSS (4.95 eV), and is consistent with the film thickness of ~ 5 nm measured by spectroscopic ellipsometry (see Supplemental Material). The Φ values are similar to published data for the solution-processed NP films of MoO₃,^[Reference Meyer, Khalandovsky, Goerrn and Kahn^22] WO₃,^[Reference Stubhan, Li, Luechinger, Halim, Matt and Brabec^16] and NiO_x^[Reference Steirer, Chesin, Widjonarko, Berry, Miedaner, Ginley and Olson^13]; Φ for CoO_x NP film has not been reported. The ~1 eV decrease in Φ compared with the vacuum-deposited films^[Reference Greiner, Chai, Helander, Tang and Lu^23] is probably caused by the adsorption of hydrocarbon species^[Reference Meyer, Shu, Kroeger and Kahn^24] due to solution processing. The Φ and ellipsometry data suggest that spin coating results in significant coverage of the ITO surface with metal oxide NPs.

IE of MoO₃ and WO₃ NP films were measured by ultraviolet photoelectron spectroscopy (UPS) (Fig. S2) yielding values of 8.6 and 8.2 eV, respectively (Table I). These two oxides' IE cannot be measured by photoelectron spectroscopy in air (PESA), due to the upper limit of the deuterium source (6.2 eV). MoO₃ and WO₃ NPs exhibit band gap 2.95 and 3.45 eV, respectively, as measured by UV–vis absorbance spectroscopy [Fig. S3(a)], in good agreement with literature.^[Reference Greiner and Lu^1] In other words, the Φ is similar to electron affinity EA (EA = IE-band gap), and these oxides are n-type. For NiO_x and CoO_x NP films, IE measured by the PESA (Fig. S4) are 5.55 and 5.70 eV, respectively (Table I). Given band gaps from literature for NiO_x (3.7 eV)^[Reference Ratcliff, Meyer, Steirer, Armstrong, Olson and Kahn^25] and CoO (2.4 eV),^[Reference Greiner and Lu^1] we see that Φ is similar to IE, suggesting that they are p-type. The IE values are consistent with published results on MoO₃^[Reference Meyer, Khalandovsky, Goerrn and Kahn^22] and NiO_x^[Reference Ratcliff, Meyer, Steirer, Armstrong, Olson and Kahn^25] NP films, whereas IE values for solution-processed WO₃ and CoO_x NP films have not been reported. In summary, the electronic levels of our solution-processed metal oxide NP films are comparable with solution-processed and vacuum-deposited films of both n- and p-type semiconductors, without the need for post-deposition treatment.

The stoichiometry of metal oxide NP films on ITO was characterized by x-ray photoelectron spectroscopy (XPS). The Mo 3d spectra can be fitted to doublet peaks at 232.4 and 235.6 eV with the same full width at half maximum (FWHM) and 3:2 areal ratio [Fig. 2(a)], corresponding to Mo 3d _5/2 and Mo 3d _3/2 peaks of pure MoO₃. Similarly, the W 4f spectra can be fitted to doublet peaks at 35.7 and 37.8 eV with the same FWHM and 4:3 areal ratio [Fig. 2(b)], corresponding to W 4f _7/2 and 4f _5/2 peaks of pure WO₃. The Ni 2p _3/2 spectra contains peaks at 855.4, 860.9, and 864.1 eV [Fig. 2(c)], corresponding to either NiO or Ni(OH)₂.^[Reference Biesinger, Payne, Grosvenor, Lau, Gerson and Smart^26, Reference Ratcliff, Meyer, Steirer, Garcia, Berry, Ginley, Olson, Kahn and Armstrong^27] In addition, the peak at 857.3 eV matches NiOOH previously reported for post-processed NiO_x NP film.^[Reference Ratcliff, Meyer, Steirer, Garcia, Berry, Ginley, Olson, Kahn and Armstrong^27] The O 1s spectra shows peaks at 529.2, 530.5, and 531.7 eV, confirming the presence of NiO, Ni(OH)₂, and NiOOH, respectively [Figure S5(a)].^[Reference Ratcliff, Meyer, Steirer, Garcia, Berry, Ginley, Olson, Kahn and Armstrong^27] The NiOOH was previously shown to establish a surface dipole which increases Φ and IE of post-processed NiO_x NP films,^[Reference Ratcliff, Meyer, Steirer, Garcia, Berry, Ginley, Olson, Kahn and Armstrong^27] and may play a similar role in our films. The Co 2p _3/2 spectra contains peaks at 780.4, 782.2, and 786.4 eV [Fig. 2(d)], corresponding to CoO or Co(OH)₂.^[Reference Biesinger, Payne, Grosvenor, Lau, Gerson and Smart^26] In addition, a weak peak at 785.0 eV may correspond to Co₃O₄, although unambiguous assignment is difficult due to the complex surface chemistries of CoO_x and the resulting overlap in XPS peaks.^[Reference Biesinger, Payne, Grosvenor, Lau, Gerson and Smart^26] The O 1s spectra shows peaks at 529.6, 531.0, and 532.7 eV, corresponding to CoO, Co(OH)₂, and adsorbed H₂O, respectively [Figure S5(b)].^[Reference Biesinger, Payne, Grosvenor, Lau, Gerson and Smart^26] Thus, as deposited metal oxide NP films show stoichiometric MoO₃ and WO₃, and mixed NiO_x and CoO_x. The high oxidation states of the metal cations are consistent with the high Φ of the NP films.^[Reference Greiner, Chai, Helander, Tang and Lu^23]

Figure 2.

XPS spectra of metal oxide NP films. (a) Mo 3d spectra for MoO₃. (b) W 4f spectra for WO₃. (c) Ni 2p _3/2 spectra for NiO_x. NiOOH peak is labeled, and other peaks correspond to NiO or Ni(OH)₂. (d) Co 2p _3/2 spectra for CoO_x. Co₃O₄ peak is labeled, and other peaks correspond to CoO or Co(OH)₂. Crosses represent the experimental data and the lines represent fit to the data.

OPV performance

We examined the performance of metal oxide NP films as HCLs using conventional PCDTBT:PC₇₁BM OPVs [Fig. 3(a), inset]. Figure 3(a) shows that all devices with metal oxide NP HCLs perform similarly to references devices with PEDOT:PSS HCL under AM1.5G 100 mW/cm² illumination. Compared with the device without HCL [Fig. 3(a), purple crosses], the devices with HCL exhibit significantly higher V _oc (Table II) due to the larger drift field established between the high Φ HCL and Ca. The devices with HCL also exhibit much smaller SD in V _oc and power conversion efficiency PCE (Table II), highlighting the importance of the HCL in creating a uniform electrical contact between the BHJ and the anode. Examining the data in detail, devices with n-type MoO₃ NP [Fig. 3(a), red squares] and WO₃ NP [Fig. 3(a), orange triangles] HCLs exhibit slightly lower short-circuit current density J _sc compared with reference devices [Fig. 3(a), black circles], leading to small reductions in PCE (Table II). In contrast, devices with p-type NiO_x NP [Fig. 3(a), green inverted triangles] and CoO_x NP [Fig. 3(a), blue diamonds] HCLs exhibit similar J _sc and PCE as the reference devices. We also compared poly(3-hexylthiophene):phenyl-C₆₁-butyric acid methyl ester (P3HT:PC₆₁BM) devices with different HCLs and found that performance identical to reference devices can be achieved using MoO₃, WO₃, and NiO_x NP HCLs [Fig. S6(a) and Table S1). We note that to optimize OPV performance, the thickness of WO₃, NiO_x, and CoO_x NP HCLs must be decreased by diluting the NP suspensions to one-fourth of the concentration of the MoO₃ NP suspension. This effect can be explained by comparing current density–voltage (J–V) curves of P3HT:PC₆₁BM devices with MoO₃ and NiO_x NP HCLs of different thicknesses. When MoO₃ NP HCL thickness was varied, we found the minimal effect on the J–V curves [Figure S6(b)]. In contrast, when NiO_x NP HCL thickness was increased, the J–V curve developed a distinct S-shape, significantly decreasing device FF [Fig. S6(c)]. S-shaped J–V curves have been attributed to inefficient collection of one type of carriers that can be caused by low contact layer conductivity, changes in recombination, or variable electric field distributions, which depends on the specific metal oxide HCL.^[Reference Wagenpfahl, Rauh, Binder, Deibel and Dyakonov^28] Thus, by optimizing the thicknesses of metal oxide NP HCLs, performance matching PEDOT:PSS is achieved for conventional PCDTBT:PC₇₁BM and P3HT:PC₆₁BM OPVs. Decreasing HCL thickness from the optimal values lowered device yield.

Figure 3.

J–V response under AM 1.5 100 mW/cm² through a 2.5 mm diameter aperture for PCDTBT:PC₇₁BM OPVs with different HCLs. (a) Conventional devices with PEDOT:PSS (black circles), MoO₃ NP (red squares), WO₃ NP (orange triangles), NiO_x NP (green inverted triangles), CoO_x NP (blue diamonds), and no HCL (purple crosses). Device parameters are given in Table II. Inset: conventional device architecture. (b) Inverted devices with evaporated MoO₃ (black circles), as synthesized MoO₃ NP (blue diamonds), and butylphosphonic acid modified MoO₃ NP (red squares). Device parameters are given in Table III. Inset: inverted device architecture.

Table II.

Performance of conventional PCDTBT:PC₇₁BM OPVs with different HCLs under AM1.5G 100 mW/cm² illumination through a 2.5 mm diameter aperture.

To challenge the processability of our metal oxide NP suspensions, we fabricated inverted PCDTBT:PC₇₁BM OPVs with MoO₃ NP film as HCL on top of the active layer [Fig. 3(b), inset]. The PCDTBT:PC₇₁BM BHJ places multiple restrictions on solution-processed overlayers. First, the solvent must wet the organic BHJ and form a uniform film. Second, because the performance of PCDTBT:PC₇₁BM device decreases if heated to >100 °C,^[Reference Park, Roy, Beaupre, Cho, Coates, Moon, Moses, Leclerc, Lee and Heeger^29] the overlayer should exhibit high Φ with minimal processing. J–V curves under AM1.5G 100 mW/cm² illumination shows that PCDTBT:PC₇₁BM devices with HCLs formed from as-synthesized MoO₃ NPs were shorted [Fig. 3(b), blue diamonds], suggesting poor coverage of the BHJ by the MoO₃ NPs. This may be caused by the fact that our NPs are not functionalized with surfactants, and aggregate during spin coating instead of uniformly coating the BHJ. To test this theory, we reacted the MoO₃ NP suspension with an equal concentration of butylphosphonic acid, which has been shown to form a self-assembled monolayer on metal oxide surfaces via metal-phosphorus bonds.^[Reference Gao, Dickinson, Grozinger, Morin and Reven^30] Using HCL of butylphosphonic acid-modified MoO₃ NPs, the PCDTBT:PC₇₁BM devices worked well, with similar J _sc and slightly lower V _oc and FF compared with reference devices with evaporated MoO₃ as HCL, resulting in PCE of 3.94% versus 5.07% (Table III). The relative surface coverage of MoO₃ NP with and without modifier on PCDTBT:PC₇₁BM was reflected in Φ values from the Kelvin probe. Because the Kelvin probe tip size (mm²) is comparable with the device area, Φ values are a good indication of the presence and coverage of a contact layer on this length scale. Φ of PCDTBT:PC₇₁BM film was 4.23 eV, which increased to 4.97 eV with an evaporated MoO₃ film on BHJ. Φ of the as-synthesized MoO₃ NP film on BHJ was 4.50 eV, and Φ of the butylphosphonic acid modified MoO₃ NP film on BHJ was 4.72 eV. Thus, butylphosphonic acid modification improved the coverage of MoO₃ NPs on PCDTBT:PC₇₁BM and increased the collection efficiency for holes, which explains the improvement in FF. We anticipate that further PCE improvements can be achieved by optimizing the functionalization of the NP surface to achieve full coverage on top of BHJ, which is subject to further investigation.

Table III.

Performance of inverted PCDTBT:PC₇₁BM OPVs with different HCLs under AM1.5G 100 mW/cm² illumination through a 2.5 mm diameter aperture.

Metal oxide NP synthesis and characterization

Metal oxide NP suspensions in n-butanol were synthesized using a modified version of a published recipe.^[Reference Redel, Huai, Dag, Petrov, O'Brien, Helander, Mlynarski and Ozin^21] Mo (average particle size 2–4 µm), W (1–5 µm), Ni (2.2–3.0 µm), and Co (1.6 µm) powders were purchased from Alfar Aesar and were used as received. In a typical reaction, 0.3 mmol metal powder was added to 525 µL n-butanol (Fisher) and 75 µL HCl (37% in H₂O, Fisher) and stirred at 25 °C for 20 min. 2220 µL of n-butanol and 180 µL of H₂O₂ (30% in H₂O, Fisher) were then added, and the solution was stirred at 25 °C for 16 h, yielding 0.1 M suspension of metal oxide NPs. The product was passed through a 0.2 µm PTFE syringe filter (Pall), and was used without further purification. DLS was performed using a Malvern ZetaSizer Nano ZS. For TEM, NiO_x NP suspension was diluted to 0.005 M with n-butanol, drop casted on a TEM grid (01824, Ted Pella), and imaged at 200 kV with a JEOL JEM2100. NP diameter was analyzed using ImageJ and represents an average of 80 NPs. Metal oxide NP films of different thicknesses were formed by diluting the stock NP suspensions with n-butanol, and spin coating 2–4 layers. For WF and XPS measurements, 0.05 M metal oxide NP suspensions were spin coated twice at 8000 rpm on unpatterned ITO on glass (20 Ω/□, Thin Film Devices) that had been cleaned by rinsing with IPA and H₂O, follow by UV-O₃ cleaning for 20 min (BioForce Nanosciences). Φ was measured in air using a scanning Kelvin probe (SKP 5050, KP Technology), with Au as the reference material (Φ = 5.15 eV). XPS was measured using a Ulvac-PHI VersaProbe 2 with monochromated Al K_α source (1486.8 eV). All spectra were collected at an angle of 45° to the sample surface with energy step of 0.1 eV and pass energy of 23.5 eV, and were fitted using commercial software (MultiPak, Ulvac-PHI). UPS was measured on MoO₃ and WO₃ films using the same instrument with He I plasma source (21.2 eV) with energy step of 0.1 eV. PESA was performed on NiO_x and CoO_x NP films using a RKI Instruments Model AC-2 system with 100 nW deuterium lamp power and energy scanning between 4.8 and 6.2 eV with a step of 0.05 eV. The SD for Φ measurement was found to be 0.03 eV. The uncertainty of the UPS and PESA measurements were found to be 0.1 and 0.05 eV, respectively.

OPVs fabrication and characterization

Conventional and inverted PCDTBT:PC₇₁BM OPVs were fabricated following published procedure^[Reference Park, Roy, Beaupre, Cho, Coates, Moon, Moses, Leclerc, Lee and Heeger^29] with modifications. For conventional devices, metal oxide NP suspensions (0.025 M for MoO₃, 0.00625 M for others) were spin coated four times at 8000 rpm on patterned ITO on glass (20 Ω/□, Thin Film Devices) to form the HCL. For reference devices, PEDOT:PSS (Clevios P VP AI 4083, Heraeus) was mixed with isopropanol in a 7:3 volume ratio, spin coated at 4000 rpm on patterned ITO, and annealed in N₂ at 170 °C for 5 min to form the HCL. PCDTBT (1-Material) and PC₇₁BM (Solenne BV) were dissolved at 7:28 mg/mL in ortho-dichlorobenzene (Aldrich), spin coated at 2000 rpm, and annealed in N₂ at 70 °C for 15 min to form 50 nm thick BHJ. Finally, Ca (7 nm) and Al (150 nm) were thermally evaporated (Angstrom Engineering) to complete the devices. For inverted devices, ZnO sol–gel film ECL was spin coated on patterned ITO at 5000 rpm from a solution of 0.5 M zinc acetate dihydrate and 0.5 M ethanolamine in 2-methoxyethanol and annealed in air at 170 °C for 10 min, followed by rinsing with H₂O, IPA, and acetone. PCDTBT:PC₇₁BM was then spin coated and annealed. For butylphosphonic acid modification of MoO₃ NP, equal volumes of 0.1 M MoO₃ NP suspension and 0.1 M butylphosphonic acid (Alfar Aesar) were mixed and stirred at 25 °C for 16 h. For HCL, 0.025 M suspension of as synthesized MoO₃ NP or butylphosphonic acid modified MoO₃ NP was spin coated four times at 8000 rpm on the BHJ. For reference devices, 10 nm MoO₃ was thermally evaporated as HCL. Finally, 100 nm Ag was thermally evaporated. The class AAA solar simulator (Sun 3000, Abet Technologies) contained AM1.5G filter, with the intensity adjusted to 100 mW/cm² using a calibrated Si photodiode with KG5 filter (RR_227 KG5, Abet Technologies). The J–V measurement was performed in N₂ using a low-noise sourcemeter (2635A, Keithley) controlled by LabView (National Instruments). A 2.5 mm diameter aperture was placed in front of each device to rigorously define the illumination area to 0.049 cm². All data shown represent the average of at least five devices.