Three-dimensional nanopillar-array photovoltaics on low-cost and substrates

Three-dimensional nanopillar-array photovoltaics on low-cost and substrates

(Parte 1 de 3)

Three-dimensional nanopillar-array photovoltaics on low-cost and flexible substrates

Solar energy represents one of the most abundant and yet least harvested sources of renewable energy. In recent years, tremendous progress has been made in developing photovoltaics that can be potentially mass deployed1–3. Of particular interest to cost-effective solar cells is to use novel devicestructuresandmaterialsprocessingforenablingacceptable efficiencies4–6. In this regard, here, we report the direct growth of highly regular, single-crystalline nanopillar arrays of optically active semiconductors on aluminium substrates that are then configured as solar-cell modules. As an example, we demonstrate a photovoltaic structure that incorporates three-dimensional, single-crystalline n-CdS nanopillars, embedded in polycrystalline thin films of p-CdTe, to enable high absorption of light and efficient collection of the carriers. Through experiments and modelling, we demonstrate the potency of this approach for enabling highly versatile solar modules on both rigid and flexible substrates with enhanced carrier collection efficiency arising from the geometric configuration of the nanopillars.

The ability to deposit single-crystalline semiconductors on support substrates is of profound interest for high-performance solar-cell applications7. The most common approach involves epitaxial growth of thin films by using single-crystalline substrates as the template8,9. In this approach, the grown material could be either transferred to another substrate by a lift-off or printing process3,10, or remain on the original substrate for fabrication of the solar modules. This epitaxial growth process, although highly useful for efficient photovoltaics, may not be applicable for cost-effective solar modules, especially when compound semiconductors are used. Recently, semiconductor nanowires grown by the vapour– liquid–solid (VLS) process have been shown to be a highly promising material system for photovoltaic devices4–6,1–14. Owing to their single-crystalline nature, they have the potency for highperformance solar modules. Although nanowires can be grown non-epitaxially on amorphous substrates, their random orientation on the growth substrates could limit the explored device structures. Here, we demonstrate the template-assisted, VLS growth of highly ordered, single-crystalline nanopillars on aluminium substrates as a highly versatile approach for fabricating novel solar-cell modules. This proposed approach could simplify the fabrication process of photovoltaicsbasedoncrystallinecompoundsemiconductorswhile enabling the exploration of new device structures.

To explore the potency of our proposed strategy, we synthesized highly ordered, single-crystalline nanopillars of n-CdS directly on

1Department of Electrical Engineering and Computer Sciences, University of California at Berkeley, Berkeley, California 94720, USA, 2Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA, 3Berkeley Sensor and Actuator Center, University of California at Berkeley, Berkeley, California 94720, USA. *

CdTe CdS


CdTe Cu/Au


CdS nanopillar ac b Electron Holen-CdS p-CdTe

Figure 1|CdS/CdTe SNOP cells. a, Energy band diagram of a CdTe/CdS photovoltaic. b, Cross-sectional schematic diagram of a SNOP cell, illustrating the enhanced carrier collection efficiency. c, SNOP-cell fabrication process flow.

aluminium substrate and embedded them in a thin film of p-CdTe as the optical absorption material (Fig. 1). Conventional thin-film photovoltaics rely on the optical generation and separation of electron–hole pairs (EHPs) with an internal electric field, as shown in Fig. 1a. Among different factors, the absorption efficiency of the material and the minority carrier lifetime often determine the energy conversion efficiency15. In this regard, simulation studies have previously shown the advantages of three-dimensional (3D) cell structures, such as those using coaxially doped vertical nanopillar arrays, in improving the photocarrier separation and collection by orthogonalizing the direction of light absorption and EHPs separation (Fig. 1b)16. This type of structure is particularly advantageous when the thickness of the device is comparable to the optical absorption depth and the bulk minority carrier lifetimes are relatively short. Under such circumstances, the optical




[312] Zone axis – ab c

2 µm 500 nm

Figure 2 | SNOP cell at different stages of fabrication. a,b, SEM images of an as-made AAM with perfectly ordered pores (a) and a CdS nanopillar array after partial etching of the AAM (b). c, Transmission electron micrograph of the interface between a single-crystalline CdS nanopillar and a polycrystalline CdTe thin film. Inset: The corresponding diffraction pattern for which the periodically symmetric spots and multi-rings can be found. The symmetric spots are originated from the single-crystalline CdS nanopillar and the multi-rings are originated from the polycrystalline CdTe thin film.

generation of carriers is significant in the entire device thickness and the 3D structure facilitates the efficient EHPs separation and collection. In addition, 3D structures have been shown to enhance the optical absorption efficiency of the material13,17,18. Specifically, photoelectrochemical studies of Cd(Se, Te) nanopillar arrays have shown that the nanopillar-array photoelectrodes exhibit enhanced collection of low-energy photons absorbed far below the surface, as comparedwithplanarphotoelectrodes17.Theseresultsdemonstrate the potential advantage of non-planar cell structures, especially for material systems where the bulk recombination rate of carriers is larger than the surface recombination rate. However, so far the conversion efficiency of the fabricated photovoltaics based on coaxial nanopillar arrays, grown by VLS, have been far from the simulation limits16, with the highest reported efficiency of ∼0.5% (ref. 1) arising from un-optimized nanopillar dimensions, poor nanopillar density and alignment, and/or low pn junction interface quality12,13, although single-nanowire devices have demonstrated better efficiencies5,14. Furthermore, controlled and cost-effective process schemes for the fabrication of large-scale solar modules that use highly dense and ordered arrays of single-crystalline nanopillar arrays have not been demonstrated. Here, some of the challenges summarized above are addressed through novel device structure engineering and fabrication process development.

The fabrication process of our proposed 3D solar nanopillar

(SNOP) cell uses highly periodic anodic alumina membranes (AAMs) as the template for the direct synthesis of single-crystalline nanostructures. This approach has been widely used for fabrication of dense arrays of metallic, semiconductor and organic 1D nanostructures, owing to the ease of membrane fabrication and nanostructure geometric control19–2. Highly regular AAMs with a thickness of ∼2µm and a pore diameter of ∼200nm were first formed on aluminium foil substrates (Fig. 1c) by using previously reportedprocesses(seeSupplementaryFig.S1)23,24.Figure 2ashows a scanning electron microscopy (SEM) image of an AAM with long-range and near-perfect ordering after the anodization. A barrier-thinning process was applied to branch out the pore channels and reduce the alumina barrier layer thickness at the bottom of the pores to a few nanometres19,21. A ∼300-nm-thick Au layer was then electrochemically deposited at the bottom of the pore channels with an alternating current method (see the Methods section). The AAM with the electrodeposited Au catalytic layer was then placed in a thermal furnace to carry out the synthesis of the CdS nanopillar array by the VLS process (see the Methods section). To form the 3D nanopillar structures, the AAM was partially and controllably etched in 1N NaOH at room temperature. Notably, this etch solution is highly selective and does not chemically react with the CdS nanopillars. Figure 2b shows a 3D nanopillar array with exposed depth, H ∼ 500nm. The exposed depth was varied by tuning the etching time (see Supplementary Fig. S2) to enable a systematic study of the effect of the geometric configuration on the conversion efficiency. A p-type CdTe thin film with ∼1µm thickness (see Supplementary Fig. S3) was then deposited by chemical vapour deposition (see the Methods section) to serve as the photoabsorption layer owing to its near-ideal bandgap (Eg =1.5eV) for solar energy absorption15. Finally, the top electrical contact was fabricated by thethermalevaporationofCu/Au(1nm/13nm),whichenableslow barriercontactstothep-CdTelayerowingtothehighworkfunction of Au. It is worth noting that although the deposited Cu/Au bilayer was thin, the optical transmission spectrum (see Supplementary Fig. S4) shows that it has only ∼50% transparency, which results in a major cell performance loss because light is shone from the top during the measurements. Further top-contact optimization is required in the future, for instance, by exploring transparent conductive oxide contacts. The back electrical contact to the n-type CdS nanopillars was simply the aluminium support substrate, which greatly reduces the complexity of the fabrication. The entire device was then bonded from the top to a transparent glass support substrate with epoxy to encapsulate the structures.

One of the primary merits of our fabrication strategy is the ability to produce high-density, single-crystalline nanopillar arrays on an amorphous substrate with fine geometric control, without relying on epitaxial growth from single-crystalline substrates. The single-crystalline nature of the grown CdS nanopillars is confirmed by transmission electron microscopy analysis with a near 1:1 stoichiometric composition observed by energy-dispersive X-ray spectroscopy (see Supplementary Fig. S5). Notably, abrupt atomic interfaces with the polycrystalline CdTe layer are observed (Fig. 2c). In addition, 3D nanopillar or nanowire arrays, similar to the ones used in this work, have been demonstrated in the past to exhibit unique optical absorption properties13,18. Similarly, we have observed reduced reflectivity from CdS nanopillar arrays especially when the inter-pillar distance is small (see Supplementary Fig. S6). This observation suggests that 3D nanopillar-based cell modules can potentially improve the light absorption while enhancing the carrier collection.

An optical image of a fully fabricated SNOP cell is shown in

Fig. 3a with an active surface area of 5×8mm. The performance was characterized by using a solar simulator (LS1000, Solar Light) without a heat sink. Figure 3b demonstrates the I–V characteristics of a typical cell under different illumination intensities, P, ranging from 17 to 100mWcm−2(AM1.5G). Specifically, an efficiency (η) of ∼6% is obtained with an open circuit voltage Voc ∼0.62V, short circuit current density Jsc ∼ 21mAcm−2 and fill factor F ∼ 0.43 under AM1.5G illumination. The I–V curves cross over each other


J (mA cm

Voltage (V)

100 mW cm–2


5 mW cm–2

35 mW cm–2 24 mW cm–2

17 mW cm–2

J sc (mA cm

Fill factor

Voc (V)

Intensity (mW cm¬2) Efficiency (%)

Glass Al substrate

Cu wire

CdS/CdTe solar cell ac b d

Figure 3 | Performance characterization of a representative SNOP cell. a, An optical image of a fully fabricated SNOP cell bonded on a glass substrate.

b, I–V characteristics at different illumination intensities. c, The short-circuit current density, Jsc, shows a near-linear dependence on the illumination intensity, whereas the fill factor, F, slightly decreases with an increase of the intensity. d, The open-circuit voltage, Voc, slightly increases with the intensity and the solar energy conversion efficiency is nearly independent of the illumination intensity for P=17∼100mWcm−2.

above Voc, which can be attributed to the photoconductivity of CdS (ref. 25). The dependency of the performance characteristics on the illumination intensity is shown in Fig. 3c,d. As expected,

Jsc exhibits a near-linear dependency on the intensity because in this regime the photocurrent is proportional to the photon flux with a constant minority carrier lifetime. On the other hand, Voc increases only slightly from 0.5 to 0.62 V with a linear increase of Jsc, which we attribute to a slight thermal heating of the device (see Supplementary Fig. S7) because a cooling chuck was not used during the measurements26. As the efficiency of a solar cell is expressedasη=Voc×Jsc×FF/P andFF slightlydecreaseswithlight intensity, the extracted η ∼6% shows minimal dependence on the illumination intensity as shown in Fig. 3d. It should be noted that thismodestefficiencyisobtainedwithouttheuseofanantireflective surface coating or concentrators.

Althoughtheconversionefficiencyofourfirst-generationSNOP cells reported here is already higher than most of the previously reported photovoltaics based on nanostructured materials11–13, further improvements are needed to meet the high-performance application requirements. Notably, the reported efficiency is higher than that of the planar CdS/CdTe cell with comparable CdTe film thickness27, but lower than those with optimal CdTe film thicknesses. As confirmed by simulation (see Supplementary Figs S8 and S9), we speculate that the efficiency can be readily enhanced in the future through further device and materials optimization, for instance, by using top contacts with higher optical transparency and lower parasitic resistances. Specifically, our top contacts result in a ∼50% efficiency loss owing to their low transparencylevel(seeSupplementaryFig.S4),whichcanbereadily improved in the future.

To further examine the effect of the geometric configuration of the nanopillars on the overall conversion efficiency, devices with different embedded CdS nanopillar lengths, H, (controlled by the etching time of the AAM, Supplementary Fig. S2) were fabricated and carefully characterized while maintaining the same overall CdTe thickness. As evident from Fig. 4a, the conversion efficiency drastically and monotonically increases with H. Specifically, η = 0.4% is obtained for H = 0nm. In such a case, only the top surface of the CdS nanopillars is in contact with the CdTe film. As a result, only a small space charge region is obtained with low carrier collection efficiency. Most of the photogenerated carriers are lost by recombination in the CdTe film, especially through non-radiative recombination at the defect-rich grain boundaries. By increasing H, the space charge region area is effectively increased with much improved carrier collection efficiency. In particular, the device conversion efficiency is increased by more than one order of magnitudewhenH isincreasedfrom0to∼640nm.

To interpret the observed trend of the efficiency dependency on the geometric configuration, 2D theoretical simulations were carried out by using a Sentaurus simulator (Fig. 4b–d). The details of the simulation can be found in Supplementary Information. The simulated efficiency as a function of H, shown in Fig. 4b, is in qualitative agreement with the experimentally observed trend. Meanwhile, the recombination rate for H = 0 and 900nm is visualized and plotted in Fig. 4c and d, respectively. It is clearly evident that the space charge and carrier collection region is

650 NATURE MATERIALS | VOL 8 | AUGUST 2009 | © 2009 Macmillan Publishers Limited. All rights reserved.


Efficiency (%) Nanopillar embeded height (nm)

CdTe TF CdS nanopillar AAM µ m)

(Parte 1 de 3)