Semiconductor Nanowires: A Platform for Exploring Limits and Concepts for Nano-Enabled Solar Cells

: Over the past decade extensive studies of single semiconductor nanowire and nanowire array photovoltaic devices have explored the potential of these materials as platforms for a new generation of efficient and cost-effective solar cells. This feature review discusses strategies for implementation of semiconductor nanowires in solar energy applications, including advances in complex nanowire synthesis and characterization, fundamental insights from characterization of devices, utilization and control of the unique optical properties of nanowires, and new strategies for assembly and scaling of nanowires into diverse arrays that serve as a new paradigm for advanced solar cells.


Table of Contents Entry:
Advanced synthetic control over the electronic and optical properties of semiconductor nanowires enables testing new paradigms for advanced solar cells.

Broader Context Box:
The solar power received by the earth dwarfs global power demands by several orders-ofmagnitude. Photovoltaics convert light to electrical energy and have the potential to partially replace current energy technologies that rely on carbon-based fuels. At present, however, a lack of infrastructure and the high costs of photovoltaics prevent this. New ideas and materials are being explored to develop next-generation solar cells that could operate more efficiently and cheaply. Nanowires have emerged as one promising platform to explore such new concepts.
Their small dimensions allow for efficient charge separation and light absorption properties unique as compared to bulk materials. Furthermore, the synthesis and fabrication of nanowire devices differs significantly from traditional wafer-based technologies, thus presenting new opportunities such as use of less abundant materials or cheaper substrates. Here, we discuss the benefits and remaining challenges of nanowires for PV and review progress towards understanding and optimizing the electrical and optical performance of nanowire devices. We focus on single nanowire studies that can define limits for what is achievable when multiple nanowires are assembled. Challenges and initial progress towards scaling are presented, and, for the first time, we articulate unique capabilities of solar cells derived from multiple, distinct NWs.

Introduction
The energy received by the earth from the sun in one hour could provide 80% of the energy that was consumed globally in 2008 at an average rate of ~17 TW. 1 Photovoltaic (PV) solar cells are the most common form of light to electrical power converter. For over two decades, single-and poly-crystalline silicon solar cells have dominated the world market for PVs with recent average annual growth in capacity exceeding 70%. 2 Nevertheless, installed PV capacity levels (40 GW in 2010) 2 represent only ~0.2% of total energy consumption, and remain low largely due to a PV efficiency to cost metric that is uncompetitive 3 with conventional power generation strategies such as fossil fuel or nuclear. Although there are lower cost technologies based on thin film solar cells such as CuIn x Ga 1-x Se 2 , CdTe, and amorphous silicon, 4-7 these solar cells sacrifice efficiency and have not reached grid parity. 3 Fundamentally, the high cost of solar power 3 is driven by expensive feedstock (e.g. high purity Si), scarcity of feedstock (e.g. rare semiconductors), and expensive material processing/purification (e.g. high temperature or high vacuum syntheses), while low efficiencies result from incomplete light absorption, carrier recombination, and thermalization losses. 8,9 Seeking to break this impasse, research in nextgeneration solar cells aims to achieve greater than 20% solar to electric power conversion efficiencies at a specific cost of less than $0.50 per peak Watt. 3 While conventional technologies have steadily improved and decreased cost, there are ample opportunities to explore new materials and concepts that could lead toward high-efficiency and low-cost solar cells.

Nanowires as a PV Platform
The field of nanoscience has provided scientists and engineers with a diverse set of materials including nanoparticles, [10][11][12][13] one-dimensional (1D) [14][15][16][17][18] and 2D 19,20 carbon materials, and metallic and semiconductor 1D nanowires [21][22][23][24][25] (NWs). Over the past decade, several motivations have fueled intensive research in NWs and how these materials could increase efficiency and reduce cost in solar cells. One proposed motivation is that bottom-up syntheses could yield new device architectures difficult or impossible to realize by conventional planar fabrication strategies. For example, it has been shown that larger lattice mismatches can be accommodated during heteroepitaxial growth of nanoscale semiconductor heterostructures. 26,27,55 Second, it is possible to prepare NWs with lower thermal budgets and with less complex fabrication steps than typically required for top-down manufacture of conventional solar cells. [28][29][30] Third, the use of flexible and/or transparent substrates could reduce cost and expand the ubiquity of solar cells in applications ranging from building-integrated to fabric-integrated PV. Fourth, the unique characteristics of light interaction with subwavelength NWs and periodic NW arrays can yield absorption comparable to conventional semiconductor films of much greater material volume. 31 Finally, because of the inherently short charge separation distances in NWs with coaxial p-n junctions, these structures could be used to explore new material combinations or to realize third-generation devices exploiting hot carriers, carrier multiplication, or intermediate bands, concepts which aim to overcome the Shockley-Queisser limit. 8,9,32 Despite substantial progress in developing NW PVs, a number of questions remain regarding the benefits and limitations of using NWs for higher efficiency and/or more costeffective solar cells but also as compact power sources that could be integrated seamlessly with nanoelectronics. This feature review discusses strategies for implementing NWs in solar energy applications, advances in NW synthesis, insights from detailed device studies, design and control of the optical properties of NWs, and new strategies for assembly and scaling of NWs into arrays which can form the basis of solar cells.
One strategy for NW solar cells relies on synthesis of NW building blocks and the subsequent assembly of these into complex multi-component devices ( Figure 1). In this paradigm, sophisticated NW building blocks are synthetically encoded with distinct electrical and optical properties and then tested on the level of single devices. Results from single NW device characterization can subsequently provide feedback for the design of new functional NWs. Finally, NWs can be assembled to form highly differentiated structures that have properties distinct from monolithic or top-down processed semiconductor architectures.
To illustrate the logic of this paradigm, we propose a unique NW-enabled solar cell: the assembled NW 'multi-junction' solar cell. In an example of our concept ( Figure 1C), 3 distinct NWs are synthesized to absorb efficiently blue, green, and red light. By stacking and interconnecting these horizontally-oriented NWs in the appropriate vertical order, one can realize a multi-junction solar cell with efficient absorption of solar spectrum, while simultaneously eliminating many of the significant challenges and costs associated with traditional multijunction thin-film solar cells, such as epitaxial growth or wafer-bonding of III/V materials. 33,55 This bottom-up approach 41,42 has been used previously to realize numerous nanoscale functional devices, [34][35][36][37][38][39][40] and also has advantages 43 for the development of next-generation solar cells: 1) synthesis can encode functionality in the form of chemical (e.g. doping) and structural (e.g. interfaces, grain boundaries, morphologies) compositions which can be tuned from the micro-, nano-, to atomic length-scales; and 2) assembly can combine diverse NW materials and structures on various substrates (e.g. glass, foil, plastic) to build a complex multi-component system with properties distinct from the individual building blocks.
Simultaneously, numerous groups have investigated an alternative strategy based on growth of large-area NW or microwire arrays wherein the wires are oriented with their long axis perpendicular to the substrate and where the device configuration is fixed during subsequent device fabrication. 44 Such an approach can deliver several benefits including: 1) device fabrication which is potentially scalable to large-area devices/modules; and 2) enhanced absorption due to long optical path-lengths and photonic effects which can persist in periodic vertically-oriented NW arrays. 45,46 As an example, research groups investigating arrays of Si microwires have demonstrated light scattering/trapping effects from periodic and random arrangement of high-refractive index semiconductor microwire arrays. 45 As another example, a ZnO NW photoanode was utilized to enhance the current density in a dye-sensitized solar cell due to increased loading of Ru-dye onto ZnO NW surfaces and improved charge percolation. 47 There are also challenges associated with this vertical NW array approach, including formation of good electrical contacts to both sides of the PV junction and uncertain cost competitiveness relative to traditional technologies stemming primarily from reliance on preparation of NW arrays from crystalline wafers. Detailed discussion of vertically-oriented arrays is beyond the scope of this review but has been recently reviewed. 31,93

Nanowire Synthesis, Characterization and Control
Nanowire PV devices can incorporate p-type/intrinsic/n-type (p-i-n) dopant modulation for carrier separation in 2 distinct motifs: (1) axial geometry with junction encoded along the growth direction of the NW, and (2) radial geometry with junction encoded between the core and coaxial shells. Schematics of a single p-i-n axial and radial nanowire diode are shown in Figure   2, where pink, yellow, and blue regions denote p-type, intrinsic, and n-type diode segments, respectively. In these structures, electron-hole pairs are generated upon absorption of photons whose energies are equal to or greater than the band-gap of silicon (E g = 1.12 eV for singlecrystal silicon). Carrier generation and separation are most efficient within the depletion region due to the built-in field established across the p-i-n junction. [34][35][36]48 Photogenerated holes (electrons) drift in the presence of the electric field through the p-type (n-type) regions and are ultimately collected as photocurrent by ohmic metal contacts. In the axial configuration, the ptype and n-type regions can be made arbitrarily short since their main purpose is to provide contact to the junction encoded within the nanowire. One attractive consequence of this is that the active device area can be kept small so as to enhance integration with other nanodevices or nanoelectronic circuits. In the radial configuration, the p-i-n interface extends along the length of the nanowire allowing carrier separation to take place in the radial versus the longer axial direction. In this structure, carrier collection distances are smaller or comparable to the minority carrier diffusion length 4,46 meaning that photogenerated carriers can reach the p-i-n junction with high efficiency without substantial bulk recombination, even in poor-quality or low-mobility materials.

Synthesis of Nanowires
Semiconductor NWs are typically synthesized using the now well-established metal nanocluster-catalyzed vapour-liquid-solid (VLS) growth process. 21,22,41,49 In this process ( Figure   3A), a vapour-phase semiconductor source dissolves in the liquid metal/semiconductor droplet maintained at a temperature near or above the metal/semiconductor eutectic point. Continued supply of the gaseous source supersaturates the droplet and the chemical potential between the gas-liquid and liquid-solid interfaces acts as a driving force for nucleation of the solid semiconductor. During growth, the liquid-solid interface is the site for continued incorporation of crystalline solid material into the elongating NW.
Axial NWs are obtained through VLS growth as illustrated in Figure 3C and Figure 3E.
Growth of NW axial heterostructures entails use of various reactant sources (e.g. dopant or other semiconductor precursors) and it is therefore important to select a nanocluster catalyst that can maintain a stable eutectic under compatible conditions. Radial or core/shell heterostructures are obtained by shifting growth conditions from VLS to a vapour-solid growth regime where dissociation and deposition of reactants is promoted on the NW surface ( Figure 3B) resulting in concentric shells. 27 Radial NWs encode function in the form of coaxial shells ( Figure 3D, F) and afford several advantages as compared to VLS-grown axial NWs, including: (1) orthogonalization of carrier separation and collection into radial and axial components, (2) atomically sharp interfaces are accessible due to the layer-by-layer vapour-phase growth mechanism, and (3) greater freedom to synthesize new structures because of the broad range of parameters for shell growth, which will be discussed below.
A diversity of solution-phase and gas-phase techniques exists for the synthesis of semiconductor NWs. Chemical vapor deposition (CVD) has been broadly adopted for growth of group IV, III/V, and II/VI semiconductor NWs and, in a few instances, molecular beam epitaxy (MBE) from gas 50 and solid-state precurors has also been utilized. Metal-organic chemical vapour deposition (MOCVD) 51 has been extensively applied for growth of compound semiconductor NWs. [54][55][56] Silicon core/multi-shell NWs with high-quality electronic interfaces were recently synthesized using a low-pressure CVD method tailored to elicit a polymorphic crystal structure with very low defect densities. 34 A scanning electron microscopy (SEM) image of such a Si NW ( Figure 3G) reveals smooth facet surfaces distinct from the disordered and unfaceted nanocrystalline morphologies previously reported. 36 In this work, the high-quality NW material led to the first demonstration of Si NW solar cells with large V OC 's of 0.5 V. In addition, low-temperature, solution-phase doping of CdS NWs with Cu ( Figure 3H) has been used to form active devices with good electrical properties, including V OC 's of = 0.6 V and efficiencies of 5.4%. 52 In addition, the Samuelson group has pioneered the use of molecular organic vapour phase epitaxy (MOVPE) to grow III/V NWs. In a recent work, the Samuelson group demonstrated the ability to selectively incorporate polytypic and twin-plane superlattices in InAs NWs ( Figure 3I) and showed how these crystallographic manipulations can be tuned with welldefined synthetic parameters including temperature and diameter. 53 Together, these reports underscore the range of synthetic systems available for growth of semiconductor NWs.

Nanowires: Diversity in Material, Structure, and Morphology
Synthetic advances in the growth of group IV, III/V, and II/VI semiconductor NWs have correspondingly enabled substantial progress in transistor, laser, and solar-cell devices. For example, Ge/Si core/shell nanowires were first prepared by nanocluster-catalyzed VLS-growth of crystalline Ge cores followed by low-temperature deposition of an a-Si shell that was subsequently annealed. 27 The resulting high-resolution transmission electron microscopy (TEM) images and energy dispersive x-ray spectroscopy (EDS) maps of Si and Ge ( Figure 4A) confirmed that the nanowire heterostructures possessed abrupt and crystalline interfaces.
MOCVD synthesis of defect-free InGaN/GaN NWs with 26 multiple quantum well (MQW) structures 54 ( Figure 4B, C) further demonstrated the potential for implementing complex structures at the nanoscale using materials whose large lattice mismatches could not be easily accommodated in thin films. 57 Finally, it has been demonstrated recently that axial NWs can be induced to kink at specific points through transient modulation of growth pressure and that multiple kinks can be encoded through iterative application of such modulations. 58 In addition to Si and Ge kinked NWs, group II/VI multiply kinked CdS NWs were synthesized ( Figure 4D), thus demonstrating the generality of this approach.
In addition to the many semiconductor materials and structural types accessible via controlled synthesis, unique morphologies are possible through tailored synthesis of Si NWs. It is possible to synthesize NWs with hexagonal, octagonal, and rectangular cross-sections ( Figure   5A) through modification of the aforementioned core/shell synthesis (e.g. shell growth temperature, silane partial pressure, etc.). Specifically, bright-field TEM images of Si core/multishell NWs sectioned ca. perpendicular to their axes reveal a hexagonal cross-section 34 with welldefined surface facets ( Figure 5B). By growing the final, n-type doped, shell of this NW at  Figure 6B). The phosphorous profile exhibits two peaks coincident with the left and right sides of the outer n-shell which have peak widths, 20-30 nm, in good agreement with the expected n-shell thickness of 25 nm based on measured growth rates. 34 Such measurements verify that well-defined, nanoscale dopant variations can be encoded in core/shell NWs and, more generally, demonstrate the control possible with bottom-up synthesis.

Device Electrical Transport and Performance
Fundamental studies of NWs for next-generation solar cells involves two important steps: (1) synthesis of NW components with precisely defined structural and chemical properties, and (2) elucidation of PV properties at the single NW level, which permits determination of the optoelectronic properties, intrinsic limits, and potential benefits of NWs for solar cells.
For device fabrication, NWs must be transferred from their growth substrates to a host substrate that will support the NW in its final device configuration. Core-shell Si NWs are typically transferred to Si 3 N 4 substrates using dry transfer 34 or contact printing techniques. 60,61 Subsequently, a portion of a single NW is protected with a lithographically defined etch mask and then immersed in etchant so that the exposed portion can be removed to the p-type core. 34 Once metallic contacts are selectively defined to n-and p-type portions of the core/shell device, the single NW device is complete ( Figure 7A). Axial devices 35 follow a very similar fabrication process, but without the need for etch masks. NWs containing other semiconductors may require different substrates, etch masks, etchants, etching processes, or metal contacts from those developed successfully for Si NWs. [34][35][36] Investigation of the I-V transport characteristics ( Figure 7B) of single core/multi-shell NW devices 34 and core/nanocrystalline-shell NW devices 36 revealed several important trends.
First, the higher V OC value, 0.48 V, was obtained from the optimized core/multi-shell geometry and represents a substantial, two-fold improvement over core/nanocrystalline-shell 36 NWs, and is the highest V OC reported to-date for a single Si NW device. Significantly, one can attain V OC 's > 0.47 V from devices with diameters as small as 200 nm, demonstrating that much larger micronscale diameters 62 are not necessary to achieve a good V OC . Second, the core/multi-shell device exhibits a larger fill factor (FF) of 73%, which is substantially higher than the 55% FF reported previously. 36 Third, while the core/nanocrystalline-shell geometry attains the higher J SC of 23.8 mA/cm 2 , core/multi-shell NW devices exhibit J SC values, 8-10 mA/cm 2 , substantially larger than expected for an equivalent thickness of bulk Si. 63 The physical origin of this striking observation will be discussed in Section 6.
Recently, finite element method (FEM) simulations 96 showed that both axial and radial NW PVs are capable of achieving high V OC values of ~0.7 V. For axial NWs, simulations show that surface recombination near the exposed junction is primarily responsible for limiting V OC below 0.7 V. For radial NWs, bulk minority carrier lifetimes < 10 ns are primarily responsible for limiting V OC . Experimental and calculated results clarify the fundamental device physics limits that must be surmounted to achieve high V OC NW PVs.
To assess the influence of shell architecture on electronic properties of core/multi-shell NWs, we examined the junction leakage current for four distinct types of structures. A plot of V OC versus the logarithm of the ratio of short-circuit current (I SC ) to dark saturation current (I 0 ) for sixteen total devices ( Figure 7C) is linear, as expected, with a monotonic increase in V OC with decreasing I 0 . 63 Representing a 1000-fold improvement over previous nanocrystalline-shell NW devices, 36 the p/in device with V OC = 0.48 V has a dark saturation current I 0 = 1.1 fA and dark saturation current density J 0 = 0.34 fA/µm 2 . This dramatic reduction in leakage current directly correlates with the two-fold improvement in V OC . Furthermore, the steady reduction in I 0 (increase of V OC ) from the p/n to p/pn to p/in device underscores how electrical performance is a sensitive function of synthetically encoded junction properties. Notably, the temperature coefficient of V OC extracted from I/V data acquired down to 60 K is -1.9 mV/K and close to the value dV OC /dT = -1.7 mV/K calculated for bulk single-crystalline Si solar cells, 64 which attests to the good transport characteristics of core/multi-shell NW devices. Interestingly, data obtained on core/multi-shell NW devices with lengths spanning over 1 order of magnitude ( Figure 7D) exhibit linear scaling of I SC and nearly constant J SC . This result confirms that NW device performance is preserved over a large range of NW lengths and encourages use of these building blocks in future large-area arrays which could be fabricated solely by standard photolithography.
Together, these data demonstrate that precise synthetic control of NW structures is a powerful determinant of improvements in V OC , FF, and I 0 .
Several groups have proposed that metal nanocluster catalysts, especially gold, used in NW growth could enhance recombination and adversely affect V OC in NW devices. 65 To further investigate this possibility, NWs were synthesized using Al catalyst 65 which, unlike Au, does not act as a mid-band gap trap state in Si. 66 Nevertheless, the best illuminated I-V curve for the core/multi-shell diode geometry ( Figure 8A, grey curve) yields a V OC , 0.23 V, ~2 times smaller than the value for analogous Au-catalyzed devices. Furthermore, even devices where Au was removed by wet-chemical etching of as-grown NW cores (Figure 8, red curve) yielded a V OC , 0.44 V, similar to the best results obtained on core/multi-shell devices prepared without removal of the catalyst. Together, these results suggest that proposed Au impurities do not significantly lower V OC and again affirm that the overall quality of the core/multi-shell structure, as dictated by synthesis, is the most important driver of good electrical performance.
Several reports have also focused on the role of NW surface states in recombination of charge carriers. Using near-field scanning optical microscopy (NSOM) to locally excite charge carriers in a reverse-biased axial NW device, researchers found that a 10 nm a-Si shell can serve as effective passivation with a 100-fold reduction in surface recombination. 67 62 Together, these results argue that with proper passivation it is possible to improve the electrical performance of NW devices.

Light Absorption in Nanowires
In addition to good electrical performance, a solar cell must absorb significant light in order to achieve high efficiency. Conventional semiconductor devices, such as bulk crystalline or thin-film solar cells, achieve greater light absorption by use of thicker absorbing layers 68 and/or anti-reflection techniques (e.g. surface texturing [69][70][71] or a quarter-wave dielectric coating 72,73 ).
However, increased material usage and additional fabrication steps typically increase cost. 74 Absorption of light by NWs is significantly distinct from bulk material absorption. Several groups have shown that subwavelength size, high-refractive index NWs can absorb a significant amount of light using a fraction of the volume needed with bulk semiconductor material. 34,59,75 To develop more efficient NW PV devices, significant effort has focused on how the absorption properties of NWs depend on material, size, and cross-sectional morphology. 34,59,[75][76][77] Synthetic control over these parameters may allow the design of NW devices with absorption characteristics distinct from traditional crystalline wafer or thin-film structures. 34,59 In addition to light absorption at the single NW level, horizontally 75,78 -or vertically-oriented 45,79 NW arrays with appropriately chosen pitch size have been shown to exhibit broadband anti-reflection as well as enhanced light-scattering. The following sections discuss the origin of distinct light absorption, and strategies to tune light absorption in single NWs and their assembled structures.
We emphasize that synthesis affords a high degree of control to tune and optimize light absorption in these materials for PV.

Introduction and Basic Theory
Semiconductor NWs are subwavelength optical cavities that can support resonant modes.
To exploit the optical properties of NWs it is necessary to quantify their optical resonances through calculation 34 Experiments to verify and quantify simulated optical properties of NWs have been carried out by measuring photocurrent spectra of single NW PV devices 34,59,62 or photodiodes [75][76][77]85 . For example, external quantum efficiency (EQE) values of ~0.15 have been reported for micro-wire devices based on Al-Si Schottky junctions. 85 Relative EQE values have been also reported for Si 75,77 and Ge 76 nanowire devices acting as photodetectors. Significantly, measurement of the absolute EQE from Si NW PV devices has been reported in several instances. 34,59 Shown in Figure 9C is a representative result of the absolute EQE versus wavelength acquired for a p-i-n core/multi-shell NW device and a simulated spectrum of the same structure. Both the experimental and simulated spectra are highly structured, with good agreement in the peak wavelengths and amplitudes, where the only adjustable parameter in the FDTD simulation was the size of the NW. By comparing the experimental and simulated spectra, peaks in the experimental spectrum are assigned to specific resonant absorption modes ( Figure   9B). Finally, the EQE spectrum approaches and even exceeds unity for wavelengths in the range of 400-500 nm which is a manifestation of the optical antenna effect. 34,62,83 This effect is purely classical and caused by the sub-wavelength diameter of the NW. To calculate EQE from photocurrent or from simulated absorption data, the projected area of the NW (normal to substrate plane) was used. In the end, EQE values greater than unity can be observed because the absorption cross-section of a single NW can exceed its physical cross-section for some photon energies. Figure 9D presents the normalized absorption spectra for Ge NW photodetectors with diameters ranging from 10 to 110 nm. 92 Taken together, these findings demonstrate that semiconductor NWs can sustain distinct wavelength-dependent absorption characteristics across a large range of the solar spectrum.

Size and Morphology Effects
One attractive feature of subwavelength size cavities is that fine changes to their structure can lead to dramatic alteration of their optical properties. [86][87][88] Recent studies have revealed how morphological changes (i.e. size and cross-sectional morphology) in NWs influence their absorption characteristics. 34,59,[75][76][77] In general, the ability to significantly tailor absorption in NWs as compared to conventional planar structures could drive development of efficient photovoltaic devices as well as other photonic applications.
The absorption behavior of p-i-n NWs with hexagonal cross-sections was investigated as a function of diameter. EQE spectra for the core-shell Si devices with diameters of 170, 280, and 380 nm ( Figure 10A, black straight line) and simulated spectra from the same structures ( Figure   10A, red dashed line) highlight several features. First, the number of measured peaks increases with increasing NW diameter. Second, identical absorption modes shift to longer wavelengths with increasing NW size. For instance, the peaks at 445, 620, and 795 nm for the small, intermediate and large sized devices, respectively, correspond to the same Fabry-Perot type modes. For PV purposes, enhanced current density (J SC ) can be achieved by tuning the absorption peaks so that they are better matched in wavelength to the maximum irradiance of the solar spectrum. 34,59 The J SC of a NW is calculated by integrating the product of its EQE spectrum with the solar spectrum. A plot of the total photocurrent per unit area ( Figure 10B, dashed red) and per unit volume ( Figure 10B, dashed black) as a function of NW diameter is obtained from simulation. From this data it is evident that J SC increases gradually with increasing diameter except for a local maximum at a diameter of 140 nm. 75 The J SC obtained from simulation agrees well with the J SC obtained from the photocurrent measured on a single NW device ( Figure 10B, red points). The overall increase in J SC for larger NWs results from an increased number of absorption peaks due to additional resonant modes emerging at longer wavelengths ( Figure 10A).
Smaller NW cavities support larger optical antenna effects, and this effect accounts for the local maximum in J SC ( Figure 10B, dashed black). In addition, the photocurrent per unit volume increases steadily as the size of a NW becomes smaller and, in particular, this value increases dramatically for devices with diameters less than 200 nm ( Figure 10B, dashed red). The noticeable increase in the total photocurrent per unit volume is due to an increasingly larger ratio of absorption cross-section to physical cross-section. These results illustrate that NW structures can efficiently localize light in nanoscale volumes. Indeed, it is possible that optical concentration effects within NW photonic cavities might lead to higher injection levels and concomitant increases in photovoltage, provided that undesirable levels of Auger recombination can be avoided. One can thus envision further exploiting this feature to realize photovoltaic devices that save cost through reduced use of absorber material.
In addition to size, the morphology-dependent absorption properties of NW devices have been investigated. It has been recently shown that NWs with rectangular cross-sections can be synthesized by growing the last n-type doped shell of a p/in Si NW at higher temperatures (cf. Section 4). 89 Polarization-resolved EQE spectra for devices fabricated from such rectangular NWs reveal several new features. In the TE spectrum of the 260 nm diameter NW ( Figure 11A, black straight), a peak centered at 570 nm shows a nearly-unity EQE amplitude, whereas bulk Si would require ~3.3 µm of material to produce the same EQE value at this wavelength.
Furthermore, by comparing the EQE spectrum of a rectangular NW to the spectrum of a hexagonal NW with an equivalent size, the distinct absorption properties of the rectangular NW become apparent. While the peaks centered at 565 nm (TE) and 680 nm (TM) are observed at similar wavelengths in both hexagonal and rectangular NWs ( Figure 11B, *), the peaks in the rectangular NW have much larger amplitudes. To understand the origin of these higher amplitude peaks their accompanying absorption mode profiles were investigated. Unlike a normal peak in the rectangular NW ( Figure 11A, profile 1), the higher amplitude peak exhibits a highly symmetric mode profile ( Figure 11A, profile 2). Since the latter mode matches with the cross-sectional morphology of the rectangular NW it experiences lower optical loss. 90 In conclusion, design of cross-sectional morphology can provide a feasible method to enhance absorption efficiency at specific wavelengths.
Recently, the morphology-dependent absorption properties of NWs with various crosssections (i.e. square, circular, hexagonal, and triangular) were explored by performing full-field electromagnetic simulations, as shown in Figure 11C-E. 75 In this simulation, amorphous Si NWs were considered. The results uncovered several trends that depend on the cross-section of a NW, including a wavelength shift of peaks between 650 -800 nm and a variation of J SC . However, this simulation does not reproduce the pronounced peaks at longer wavelengths for a NW with rectangular cross-section as discussed in the preceding paragraph. The discrepancy between these two reported results lies in the absorber material: crystalline Si is inherently less absorptive than amorphous Si. For more absorptive materials, the short attenuation depth of photons reduces optical feedback within the cavity, leading to a weak dependence on morphological differences.
In this case, differences in absorption are due to differences in cross-sectional area of the simulated structures rather than their morphology. In general, designs for efficient NW photovoltaics could include a combination of highly absorptive and highly tunable NW optical components, where the latter could be achieved with synthetic control of morphology.
From these studies, it is clear that NWs absorb light differently from thick semiconductor films that are traditionally used for photovoltaics. The NWs subwavelength size and high-refractive index allow for these structures to absorb light from outside their projected area and sustain optical resonant modes with distinct spatial mode profiles. Taken together, light absorption in NWs is determined by the following: (i) absorption coefficient of the NW material, (ii) spatial profile of NW cavity modes, (iii) optical antenna effect, and (iv) wavelength matching between NW cavity modes and irradiance of the solar spectrum. We see considerable promise in further tuning absorption in NWs through synthetic manipulations directed at controlling not only the cross-sectional but also the internal morphology of NWs.

Assembly and Scaling of Nanowires for Photovoltaic Applications
Single NW photovoltaics are an ideal platform for elucidating fundamental factors dictating performance 34 and can be used to power nanoelectronic elements. 36 Nevertheless, one must consider the potential for assembly and scaling of these functional elements into arrays that will be required for general purpose solar cells. As outlined at the beginning of this review, we have proposed a 'multi-junction' solar cell consisting of stacked NWs (Figure 1), each of which has been synthetically optimized to absorb a significant fraction of the solar spectrum.
Significantly, recent experimental results have begun to address the prospects for integration and assembly of large numbers of NWs into efficient photovoltaic arrays. 60,61,78 As an initial step, core/multi-shell NW elements were integrated in parallel and it was shown that this multi-NW device ( Figure 12A) retains good 'cell' characteristics (e.g., V OC and FF are nearly constant) with increasing number of NW elements. Indeed, the FF for a device consisting of 8 NWs, 69.3%, is similar to the best value attained on single NW elements, 72.0%. Central to the potential for scaling, an increase in I SC from 191 to 1723 pA was observed as the number of NWs in a device was increased from 1 to 8, respectively ( Figure 12A). Critically, this increase in absolute I SC was accompanied by preservation of J SC to within 3% of that for the best single NW device. Furthermore, recent FDTD simulation results across a range of NW diameters show that the current density of a single-layer array of these close-packed NW building blocks is within 15% of that of a single NW ( Figure 12B). Interestingly, this correspondence holds even for NW arrays with up to 30% void space, 34 meaning that efficient NW solar cells could be tolerant of incomplete filling fraction. Moreover, a recent work simulated the absorption properties of a-Si NW arrays 78 and, in particular, determined that absorption (J SC ) can be enhanced by appropriate choice of pitch size ( Figure 12C). Together, these studies show that there is considerable promise in exploring light coupling and propagation within NW arrays. One significant advantage of an assembled NW solar cell is that NW size, morphology, and material composition can be tuned in addition to and independent of the array geometry/spacing, thus offering several strategies for control of light absorption in these systems. Development of larger arrays should be able to exploit reported advances in assembly of dense parallel nanowires on the micrometer to several inch scale, where the nanowires have also exhibited good electronic properties. 60,61 For purposes of enhancing current density and thereby efficiency, we propose a new paradigm which involves layering and interconnecting optimized core/multi-shell NW components. Demonstrating the potential of this new concept, recent work has shown how a well-aligned vertical stack of 2 p/pin core-shell NWs can be assembled on a quartz substrate and connected electrically in parallel ( Figure 12D, inset). Transport measurements for such a device yield a J SC of 14.0 mA/cm 2 , a 1.4x increase compared to the best single NW device.
Significantly, the experimental EQE spectrum for this double NW device ( Figure 12D) shows that peaks coincide in wavelength with those for a single NW device and that EQE amplitudes for the former are increased by a factor of 1.0 -2.0 across the spectrum. Notably, FDTD simulations reproduce the EQE enhancement for the double NW stack configuration and predict a comparable 41% increase in J SC . Finally, additional simulations have shown that a layered stack of five NWs can yield J SC values of 24 and >40 mA/cm 2 without and with a BSR, respectively ( Figure 12E). These values coupled with the high V OC values reported for the single core/multi-shell NW devices (Section 5), which have been shown to be preserved upon parallel integration ( Figure 12A), could yield power-conversion efficiencies >15%.

Conclusions and Prospects for Future Research
Future studies must not only explore synthesis of new nanoscale materials 34,59,94 but also develop robust methods for assembly of large area arrays for photovoltaic applications.
Challenges and opportunities exist for the development of assembled NW photovoltaic arrays.
For example, while existing transfer techniques 60,61,95 can pattern NWs over large-areas and at high density, more desirable methods should provide control over array pitch regardless of NW size or morphology. Furthermore, to reduce packing defects during assembly of NW arrays, methods should maximize end-to-end registry of NWs and reject small NW fragments leading to voids. In addition, NW design 96 and synthesis efforts should minimize NW heterogeneity with regards to size and doping, because variations in these parameters can limit the power conversion efficiency of the system 97 . Finally, although the total systems cost of an assembled NW photovoltaic is difficult to quantify rigorously, synthetic advances including nanowire growth by aerotaxy 98 , use of cheaper metal catalysts and lower temperature processes 52 will likely be a primary driver of cost reductions. Ultimately, the assembly and large-scale integration of diverse NW building blocks could become a viable strategy for development of efficient and costeffective solar cells.