Sabastine
Ezugwu
a,
Hanyang
Ye
a and
Giovanni
Fanchini
*ab
aDepartment of Physics and Astronomy, University of Western Ontario, London, Ontario N6A 3K7, Canada. E-mail: gfanchin@uwo.ca; Fax: +1-519-661-2033; Tel: +1-519-661-2111 ext. 86238
bDepartment of Chemistry, University of Western Ontario, 1151 Richmond St., London, Ontario N6A 5B7, Canada
First published on 13th October 2014
In order to investigate the suitability of random arrays of nanoparticles for plasmonic enhancement in the visible-near infrared range, we introduced three-dimensional scanning near-field optical microscopy (3D-SNOM) imaging as a useful technique to probe the intensity of near-field radiation scattered by random systems of nanoparticles at heights up to several hundred nm from their surface. We demonstrated our technique using random arrays of copper nanoparticles (Cu-NPs) at different particle diameter and concentration. Bright regions in the 3D-SNOM images, corresponding to constructive interference of forward-scattered plasmonic waves, were obtained at heights Δz ≥ 220 nm from the surface for random arrays of Cu-NPs of ∼60–100 nm in diameter. These heights are too large to use Cu-NPs in contact of the active layer for light harvesting in thin organic solar cells, which are typically no thicker than 200 nm. Using a 200 nm transparent spacer between the system of Cu-NPs and the solar cell active layer, we demonstrate that forward-scattered light can be conveyed in 200 nm thin film solar cells. This architecture increases the solar cell photoconversion efficiency by a factor of 3. Our 3D-SNOM technique is general enough to be suitable for a large number of other applications in nanoplasmonics.
In relation to the dielectric function of a metal, it is shown that the electric field localized in the proximity of metallic nanoparticles induces charge polarization of free electrons, leading to their coherent excitation.5 In the far field, such coherent oscillations are highly screened if their frequency is lower than the plasmon frequency and the metal is highly reflective.12 The plasmon frequency thus provides a framework for describing the effectiveness of a plasmonic enhancement layer. Due to its relatively low plasmon frequency, copper has been considered for plasmonic photovoltaic enhancement layers at infrared frequencies, for which it is highly reflective in the far field, but seldom for visible light, because it exhibits strong optical absorption and poor reflectivity in the far field in that range.13 However, strong Mie scattering and forward-scattering in the near field at frequencies above the plasmon resonance of copper nanoparticles can also be useful to enhance light collection in thin film devices, if light is appropriately conveyed in their active region.6 Plasmon resonance effects have equally triggered considerable research interest for their applications in other areas. For instance, an increase in electroluminescence of about 12 orders of magnitude in light emitting diodes (LEDs) has been reported,14,15 in which metal nanoparticles were used to increase the external quantum efficiency of thin silicon-on-insulator LEDs. More recently, it has been shown that magneto-optical effects can be remarkably increased if implemented with nanostructured plasmonic crystals.16–18 In these reports, multilayered films covered with silver-capped nanospheres enhanced light transmission due to the enhanced interaction of light with surface plasmons. However, for these and other applications, copper is preferable over other metals commonly used for plasmonic enhancement in the visible region, including silver and gold, because of its cost-effectiveness.
In the specific case of copper,19 that can be easily oxidized,20,21 it is extremely important to grow the nanoparticles using physical deposition techniques that present significant advantages over wet chemistry methods that predispose the fabricated NPs to contamination and oxidation by the precursor residual materials, their by-products and the immediate environment. In physical deposition techniques, particle growth takes place in ultra-high vacuum (UHV), thereby minimizing the problems of contamination and oxidation. These techniques, including sputtering, thermal and ion-beam evaporation and pulsed laser deposition, can be used to deposit a variety of thin films and nanoparticles and are easily adaptable to large scale and industrial processes.22
In order to explore the suitability of Cu-NPs grown by direct-current biased, radio-frequency (RF) magnetron sputtering for plasmonic applications in thin film solar cells, we are here introducing three-dimensional (3D) scanning near-field optical microscopy (3D-SNOM) at 532 nm wavelength as a useful tool to directly image the amount of light that is forward-scattered in the near field from layers of metallic nanoparticles. 532 nm wavelength has been specifically chosen because it is situated in the proximity of the maximum of the AM 1.5 solar spectrum and is customarily used to design photovoltaics and other solar devices.23 SNOM is a nano-optical imaging technique that exploits light–matter interaction for the purpose of extracting relevant information about nanoscale objects.24 Aperture-type SNOM involves the electromagnetic interaction of two distinct nano-objects for which the principle of linear superposition can be invoked: a ‘probe’ nano-hole and a ‘sample’ nano-object (that in our case will be a Cu-NP in our sample). In the past, ‘nano-hole’ probes were typically tapered optical fibres. In our state-of-the art equipment, the nano-hole is a nanometric aperture machined at the end of an atomic force microscope cantilever. The 3D-SNOM method that we have developed for the present study is depicted in panel (a) of Fig. 1. With this method, the sample is illuminated with a 532 nm laser using an inverted microscope underneath the sample and an aperture-type SNOM probe attached to an atomic force microscope is utilized to collect light in the near field at variable distances z over the sample surface, controlled using a piezoelectric scanner, while maintaining the usual raster scan of the optical probe constant along the x- or y-axis (see panel (b)). At each point (x, y, z), the amount of light collected from the aperture of the SNOM probe is proportional to the light forward-scattered by the sample and is received by an upright confocal optical microscope focused on the tip hole and the intensity of radiation scattered in the near field is measured using a photomultiplier. This configuration allows us to map the forward-scattered waves of electric field arising from Mie scattering from various centers in our random arrays of Cu-NPs. The usefulness of our method for designing Cu-NP layers for plasmonic photovoltaic applications will be demonstrated.
The ratio of atomic contents of O and Cu (Fig. 2, panel d) shows that annealing in air led to severe oxidation of the Cu-NPs, while the O level was below the detection threshold (<1 at%) for the sample that was entirely annealed and treated inside the glove box. Fig. 2, panel (e), shows the optical transmission spectra of these samples. A dip centered at λp = 552 ± 5 nm, that can be assigned to resonant plasmon excitation,25,26 is evident in the transmission spectrum of the sample annealed and treated in the glove box, while it is undetectable in the air-annealed sample as a result of surface oxidation.27 On the other hand, the transmittance of the reference as-deposited Cu-NPs shows an enhancement (i.e. a peak in transmittance)28which is opposite to plasmonic enhancement (i.e. a dip in transmittance, corresponding to a peak in reflectance). The transmittance enhancement corresponds to metallic behavior of this sample; the morphology of the Cu-NPs presented in the following paragraph shows that the nanoparticles are highly inter-connected.
Fig. 3 shows the atomic force microscopy (AFM) images of the samples previously discussed in Fig. 2. Panel (a) shows the morphology of the sample prior to annealing, with a fine distribution of interconnected Cu-NPs that gave rise to a transmittance peak, as shown in Fig. 2e. Annealing in air (Fig. 3b) did not significantly alter the morphology of the sample. Although the Cu-NP diameter grows from 54 ± 5 nm to 61 ± 5 nm due to nucleation effects, particles are still interconnected and no plasmon-related feature was noticed in the optical transmission spectrum in Fig. 2e due to significant oxidation. Conversely, a more significant nucleation occurs upon annealing the sample in the glove box (Fig. 3c), with well separated and oxygen-free nanoparticles of 73 ± 2 nm in diameter. This demonstrates that, by growing our samples in an UHV deposition system attached to a glove box, we are able to completely overcome the effects of oxidation and to produce random arrays of Cu-NPs of suitable quality for nanoplasmonics. More experimental evidence on the nucleation of these particles, including the scanning electron microscopy images and particle distributions under different experimental conditions, can be found in the ESI.†
Fig. 3 AFM topography images of (a) as-deposited, (b) air-annealed and (c) glove box annealed Cu-NPs. Thermal annealing was at 300 °C for 1 hour. |
In addition to thermal annealing of interconnected Cu-NPs, it was also possible to initiate further nucleation by island formation during sputtering, also known as Volmer–Weber growth29 without any post-annealing process. In order to identify the optimal growth conditions for random arrays of Cu-NPs for plasmonic photovoltaic applications, we varied the Ar partial pressure in the chamber from 9 Pa to 23 Pa and sputtered a set of samples at −400 V dc negative bias voltage. External negative bias of the sample is essential to decrease the deposition rate and favor the process of nucleation during the sputtering deposition, thus promoting Volmer–Weber growth of individual Cu-NPs rather than interconnected particles.29 The average particle diameter increased from 58 ± 10 nm to 95 ± 10 nm with increasing chamber pressure. A complete characterization of this set of samples including AFM images, optical transmission spectra, collection-mode SNOM images and 3D-SNOM profiles of forward-scattered light is reported in the ESI.†
According to Mie theory, the multipolar nature of the scattered field is expected to manifest itself with the fact that the field intensity undergoes a number of oscillations, with maxima and minima at specific distances z, from the surface of the particles.30–33 Experimentally, these oscillations are clearly visible in Fig. 4c that represents a (x, z) SNOM scan of the line at y = const. indicated by letters A and A′ in Fig. 4b. No oscillations are visible if a bare glass substrate is scanned, which confirms that the forward-scattered near field that produces them must be attributed to the Cu-NPs. From Fig. 4c, it can also be noted that the light intensity is relatively uniform along the x-axis, with a relatively lower intensity for the corresponding largest particles. This is a consequence of the fact that even in this sample, in which the Cu-NPs are relatively sparse, the inter-particle distances are not significantly larger than a wavelength and, therefore, the Cu-NPs cannot be considered to be isolated. Fig. 4d shows the integrated intensity of forward-scattered light along the A–A′ cross-section. Maxima corresponding to constructive interference and minima corresponding to destructive interference are clearly visible. It can be seen that for this specific sample, the first maximum of forward-scattered light intensity was observed at Δz = 1200 nm above the glass surface, while the absolute maximum occurs at about Δz = 4400 nm, after which the intensity of scattered light drops to very low values as can be expected from the poor green light transmittance of Cu-NPs in the far field.13
Using this very same procedure, we performed 3D-SNOM scans on random arrays of Cu-NPs of different diameters, obtained from a set of samples deposited at increasing Ar pressures from 9 to 23 Pa (see the ESI† for detailed characterization). In Fig. 5 we show the variation of the first maximum of forward-scattered light intensity as a function of the average particle diameter and we observe that the light scattered from Cu-NPs with a larger diameter is more closely coupled to the surface of the sample than in smaller Cu-NPs. For instance, Cu-NPs with 95 ± 10 nm diameter shows scattered wave peaks at Δz1-max = 220 ± 50 nm, compared to Δz1-max = 1400 ± 100 nm for the particles with 58 ± 10 nm diameter. This can be attributed to the dependence on the particle size of the cross-sections for near-field optical absorption and scattering of light.5–7 The scattering cross-section is proportional to the square of the nanoparticle volume34 and, subsequently, for sufficiently large nanoparticles it can be considered to be much greater than the cross-section for optical absorption. Therefore, for Cu-NPs with a diameter greater than 80 nm, multipolar effects due to Mie scattering manifest themselves as strong coupling of the scattered waves in the vicinity of the NP surface.
For any particle diameters, the values of Δz1-max measured by us are too large to use the radiation that is forward scattered by Cu-NPs in the near-field for harvesting light in organic thin film solar cells if the active layer is immediately overlaid in contact with the nanoparticle array. Typically, organic solar devices are no thicker than 200 nm to minimize recombination of charges due to the high resistivity in the active layers.35,36 Decreasing the device thickness to a few hundred nanometers or less is also required to overcome recombination due to the short diffusion length of excitons generated within the active layer of the cell. As such, light absorption is generally limited by the thickness of the active layer. In order to use Cu-NPs for increasing the organic solar cell performance, it is essential to position the active layer at a distance Δzi-max at which the maximum intensity of light is forward-scattered, for efficient concentration of light into the device. These considerations mean that it would be beneficial to employ dielectric materials for which the desired scattering effects are still present but for which surface plasmon resonances do not occur.
We explored the prospects of this concept by developing a more structured plasmonic architecture in which a SiO2 thin film, sputtered in the same chamber used to prepare our Cu-NPs, functions as a spacer (see the ESI† for the detailed sputtering operation of the glass target). The AFM images, x–y SNOM images and 3D-SNOM images obtained with 100 nm diameter Cu-NPs with a 200 nm SiO2 spacer and without a spacer are reported in Fig. 6a–c and 6d–f, respectively. On the one hand, from the AFM image in panel (a), it can be observed that the surface of the sample is entirely flat after sputtering the SiO2 spacer and the Cu-NPs are entirely embedded in SiO2; on the other hand, the x–y transmission SNOM image in panel (b) clearly reveals the presence of Cu-NPs below the sample surface in this case. Conversely, in the absence of the spacer, Cu-NPs are visible at the very same locations in the AFM (panel d) and the x–y transmission SNOM image (panel d) in Fig. 6.
More interesting results arise from the comparison of the 3D-SNOM images with and without a spacer, which are presented in panels (c) and (f), respectively. In the presence of the spacer, we observe a broad region of constructive interference of forward-scattered light, from zero to approximately 550 nm in the vicinity of the SiO2 surface. This broad enhancement region may be assigned to coupling of plane waves normally incident to the SiO2 surface with laterally propagating waves parallel to the SiO2 surface, which results in the introduction of a lateral wave vector component.37 Superposition of the lateral and normal components of the wave vector may also lead to the increased intensity of the scattered light in addition to constructive interference in the normally incident wave. These results indicate that this architecture is extremely promising for building a near-field plasmon-enhanced solar cell on top of its surface. Conversely, in the absence of the spacer, a destructive interference peak is located at approximately 160 ± 30 nm from the sample surface, which results from considerable near-field depletion in the region at which the active layer of a solar cell would be placed in this case. Fig. 7a shows the intensity profile of light scattered in the near-field along the z-axis, extracted from the 3D-SNOM images in Fig. 6c and f. It can be observed that, in addition to a broader constructive interference peak, the structure with the spacer produces a significantly higher intensity of forward-scattered light.
Bilayer metallic contacts of calcium (20 nm) and aluminum (100 nm) were fabricated in a thermal evaporator that was also installed in the same chamber housing the sputtering system, which resulted in four complete solar cells assembled on top of each of our four plasmonic architectures. The schematic of these devices is shown in Fig. 8a and b, respectively.
Device testing was also carried out in the glove box and the current–voltage (I–V) curves produced are shown in Fig. 8c and d, respectively. It can be observed that, although the dark characteristics of the two devices are basically the same, a significantly larger photocurrent could be extracted from the devices built on the architecture with the spacer, which produced an average short-circuit current Isc = 1.44 mA, more than twice the value (Isc = 0.67 mA) obtained in the absence of the spacer. Conversely, the open circuit voltage (Voc ≈ 0.56 ± 0.02 V) and fill factor (52 ± 2%) of the two types of devices did not show any noticeable differences. These results are a strong indication of the fact that the resulting increase in AM 1.5 photoconversion efficiency (η = 3.08% on average in the presence of the spacer vs. η = 1.38% in the absence of the spacer) is a consequence of the enhancement of light conveyed within the active layer of the device due to near-field radiation forward-scattered by Cu-NPs. In order to further substantiate these findings an identical control cell was also fabricated without any underlying array of Cu-NPs, which resulted in an efficiency (η = 2.45%) that was intermediate between the two cases, further corroborating the usefulness of our 3D-SNOM method for effectively identifying specific conditions for plasmonic solar cell enhancement.
By introducing a novel 3D scanning near-field optical microscope technique, we obtained the various positions of constructive interference peaks of plasmonic waves arising from multipolar contributions to the scattering cross-section by Cu-NP ensembles. The best coupling of the scattered light to the sample surface was achieved by incorporating a 200 nm SiO2 spacer between the Cu-NPs and the sample surface. With this configuration, the scattered light waves extended up to 500 nm from the Cu-NP/SiO2 surface. The observed improvement in the scattering enhancement from Cu-NP/SiO2 surfaces is ideal for the improved performance in plasmonic devices such as thin film solar cells that are normally no more than 200 nm in thickness.
In order to verify the effect of oxidation on thermally treated Cu-NPs, we performed energy dispersive X-ray spectroscopy (EDX) analysis of the samples deposited under the same conditions but annealed in a different environment using a LEO (Zeiss) 1540 field emission scanning electron microscope (SEM). This sample was sputtered on the Si substrate for 3 minutes at 75 W RF power, 0 V dc bias voltage and 9 Pa Ar pressure and divided into two parts that were annealed separately at 300 °C for 1 hour. The glove-box annealed sample was taken out immediately before being measured with a dwell time no more than 10 minutes before being admitted to the SEM/EDX chamber. The UV-VIS transmittance of the NPs was measured at normal incidence in the range of wavelengths between 350 nm and 800 nm using a Varian DMS80 spectrophotometer. The transmittance data were obtained at a step scan rate of 10 nm per second.
Near field optical study and morphological mapping on the nanoscale were performed using a Witec Alpha 300S atomic force microscope. The Witec Alfa 300S, an aperture-type, AFM-integrated SNOM instrument can be used to record SNOM images and, simultaneously, AFM topographic images of a sample. The system can record SNOM images in ‘transmission mode’ or in ‘collection mode’ that differ in the way the sample is illuminated and the scattered light is collected. In the collection mode operation described schematically in Fig. 1a, light from a 532 nm green laser operated up to 50 mW (Excelsior, Spectra Physics Inc., serial no. 10398) is directed, by a system of optical fibers, into an inverted microscope, and the optical response of the sample is collected by the nanohole (d ≈ 80 nm) located at the end of an AFM hollow tip, mounted below a high-resolution confocal microscope. The collected light is subsequently launched via the confocal microscope into an optical fiber that is connected to a photomultiplier tube (U-68000, Hamamatsu) operating in photon-counting mode. The sample is positioned on a 100 × 100 μm piezo-scanner that has a maximum excursion of 10 μm in the z-direction. In addition, the mechanical arm on which the confocal microscope is mounted can also be moved in the z-direction for optimizing the focal plane at the level of the AFM tip. If the hollow tip is sufficiently close to the sample surface, only the near-field optical response from the sample surface will be collected. However, if the hollow tip is lifted up at a controlled distance z from a nanoparticle that is located at the sample surface, all of the other normal modes representing the propagating component of the scattered waves can be detected.
In our SNOM collection mode operation, the sample surface was scanned along the (x, y) plane at z = 0, or in the (x, z) direction at y = constant, in order to obtain relevant information about the amount of light scattered by the particle at a nonzero distance from its surface, as demonstrated in Fig. 1b. With this three-dimensional optical imaging procedure, specifically designed for this study, the collection mode SNOM analysis was carried out first to obtain the (x, y) nano-optical and topographic (contact-mode AFM) images of Cu-NPs, which were recorded simultaneously. We then chose a line at y = const. from the (x, y) images and changed the distance z from the tip to the sample surface. This can be done reproducibly because the piezo-scanner stage on which the sample is mounted allows us to control the sample–tip distance with ±1 nm reproducibility.
The preparation of two identical organic solar cells with and without a 200 nm SiO2 spacer on top of Cu-NPs was carried out following the same procedure that was reported previously.38 Both the fabrication and device testing were performed in the glove-box, loaded with high-purity nitrogen, with oxygen and moisture levels less than 5 ppm. A 100 nm indium-tin oxide (ITO) thin film was spun directly on top of the two surfaces, with and without the spacer, from a colloidal suspension of ITO in isopropanol. The samples were then fast-annealed at 400 °C for 5 min. An anhydrous Poly(3,4-Ethylene Di-Oxy-Thiophene):Poly-Styrene Sulfonate thin film (PEDOT:PSS, Aldrich cat. no. 483095) in isopropanol was spun at 3000 rpm on the top of ITO to form a 30 nm hole transport layer, and then baked on a hot plate at 140 °C for 30 min. The solar cell active layers were assembled by spinning on the top of PEDOT:PSS a mixture (15:15 mg) of regioregular poly(3-hexylthiophene-2,5-diyl) (P3HT) and [6,6]-phenyl-C61-butyric acid methyl ester (PCBM) (Aldrich, cat no. 698989 and 684430, respectively) in chlorobenzene, followed by thermal annealing at 120 °C for 15 min.
The solar cells were then transferred into the deposition chamber for thermal evaporation of Ca/Al bilayer backing electrodes. The metallic contacts (20 nm Ca thickness and 100 nm Al thickness, measured using a Sycom STM-2 thickness monitor) were evaporated on each solar cell using a patterned shadow mask,38 resulting in a device area of 0.245 cm2 each. The solar cell I–V characteristics were measured directly in the glove box using a Newport 9600 AM 1.5 solar simulator at 1 sun. The solar simulator was calibrated using a Sciencetech SC-LT standard cell with certification accredited by the National Institute of Standards and Technology (ISO-17025).
Footnote |
† Electronic supplementary information (ESI) available: Detailed description of the deposition system and a complete characterization of the set of samples deposited at different chamber pressures, including AFM images, optical transmission spectra, collection-mode SNOM images and 3D-SNOM profiles of forward-scattered light. See DOI: 10.1039/c4nr05094k |
This journal is © The Royal Society of Chemistry 2015 |