Three-Dimensional Electrochemical Axial Lithography on Si Micro- and Nanowire Arrays.

A templated electrochemical technique for patterning macroscopic arrays of single-crystalline Si micro- and nanowires with feature dimensions down to 5 nm is reported. This technique, termed three-dimensional electrochemical axial lithography (3DEAL), allows the design and parallel fabrication of hybrid silicon nanowire arrays decorated with complex metal nano-ring architectures in a flexible and modular approach. While conventional templated approaches are based on the direct replication of a template, our method can be used to perform high-resolution lithography on pre-existing nanostructures. This is made possible by the synthesis of a porous template with tunable dimensions that guides the deposition of well-defined metallic shells around the Si wires. The synthesis of a variety of ring architectures composed of different metals (Au, Ag, Fe, and Ni) with controlled sequence, height, and position along the wire is demonstrated for both straight and kinked wires. We observe a strong enhancement of the Raman signal for arrays of Si nanowires decorated with multiple gold rings due to the plasmonic hot spots created in these tailored architectures. The uniformity of the fabrication method is evidenced by a homogeneous increase in the Raman signal throughout the macroscopic sample. This demonstrates the reliability of the method for engineering plasmonic fields in three dimensions within Si wire arrays.

Silicon micro- and nanowires have outstanding tunable optoelectronic properties.[1−3] Their one-dimensional geometry is particularly attractive because it decouples minority charge-carrier collection from the light-absorption process[2] while providing a high surface area for catalyst loading.[4,5] Because Si wire arrays can be easily doped and are compatible with microelectronic fabrication,[6,7] they have been important candidates for sensors,[8] photoelectrodes,[1,4,5] batteries,[9] photocatalytic systems,[10−12] solar cells,[13] nanoscale electronics,[6,7] and nanoelectrodes for neurological studies.[14] For most of these applications, the Si wires need to be properly interfaced with metal nanostructures that can act either as catalysts,[4,5,10,11] electrical contacts,[6,7,14] or plasmonic materials.[12,15−17] Unfortunately, synthesizing hybrid single crystalline Si (c-Si) nanowire architectures with defined metal nanostructures is challenging and can only be achieved by performing three-dimensional lithography within the wire arrays, which is not possible with current techniques.

To date, multiphoton microfabrication techniques have been the preferred method for performing lithography in three dimensions.[18] However, they are usually of low throughput, are not able to produce features smaller than 10 nm, and can only be used to modify polymeric substrates. Various techniques have been used to create complete shells around Si wires, such as chemical vapor deposition,[19] electrodeposition,[20,21] organometallic precursor pyrolysis,[16] or wet chemical synthesis.[22] Alternative approaches, based on Rayleigh instability[23] and selective etching of Si,[24,25] have provided elegant ways to control Si wire morphology. However, these methods are either strictly limited to pure Si architectures[24,25] or have constraints in terms of shell composition, dimension, and location (>400 nm range)[23] that restrict their potential use. Until now, electron-beam lithography has been the only method to pattern metals on Si nanowires.[6] It is very slow and expensive and cannot be used for patterning three-dimensional nanoscale systems. The ability to locate well-defined metal nanostructures at specific locations within Si micro- and nanowire arrays could accelerate fundamental studies on hybrid nanostructures and add advanced properties to existing nanowire architectures. For example, recent studies have shown that the position of metal catalysts within nanostructured TiO2 and Si photoelectrodes can strongly influence photocatalytic activity.[5,26] Nanometer control over catalyst location within Si nanowire photoelectrodes could aid in the understanding and optimizing of catalyst loading to improve photoelectrode performance. Additionally, because metallic nanostructures can greatly enhance the incident electric field (E-field) under plasmonic excitation,[27,28] their integration within Si wire arrays could be used to precisely define plasmonic hot spots in three dimensions within the array.[15−17] Such enhanced E-fields are of fundamental and technological significance because they can be used to modify important processes, such as light emission and absorption, Raman scattering via the surface-enhanced Raman spectroscopy (SERS) effect, and photochemical reactions via hot electron generation and photothermal effects.[15−17,29−37] Until now, coaxial lithography (COAL) has been the most versatile technique used to engineer core–shell nanowires.[33] COAL is built upon the seminal works of Martin, Penner, and Moskovits, who developed electrochemical deposition within porous membranes to synthesize a variety of metal and semiconductor nanowires.[38−40] Based on the sequential electrodeposition and selective etching of materials with different chemical and mechanical stabilities within porous anodic aluminum oxide membranes, COAL is a powerful method for generating metal shells around nanowires. However, the process is limited to low-quality semiconductor materials that need to be grown from solution, while the straight morphology of the wire and its size and periodicity is fixed and inherited from the rigid alumina template. Additionally, COAL cannot be used to modify pre-existing nanostructures.

Herein, we present a benchtop method, termed three-dimensional electrochemical axial lithography (3DEAL), to pattern vertically aligned crystalline silicon micro- and nanowire arrays (diameters ranging from 160 nm to 1 μm; Figure ) with tailored metal architectures. We show the flexibility of the method to coat straight as well as kinked wire arrays with multisegmented metal shells, yielding Si core–metal shell (Si@metal) wires with positive and negative features down to 40 and 5 nm, respectively. The thickness of the metal shells can be controlled in the 30–150 nm range. The shell can be composed of a variety of metals (Au, Ag, Ni, and Fe), while the shell length and position are controlled electrochemically. Using sacrificial shells combined with selective etching, well-defined metal rings around the nanowires can be fabricated. The homogeneity of the structures produced was verified by scanning electron microscopy (SEM), high-angle annular dark field (HAADF) scanning transmission electron microscopy (STEM), and confocal Raman microscopy. We further show that Si nanowire arrays patterned with gold rings show significant enhancement of the Raman signal due to the enhanced electric field around the metal rings.

Figure 1

Three-dimensional electrochemical axial lithography within Si micro- and nanowire arrays. (a) Scheme showing the successive synthetic steps. (1) A loosely packed hexagonal array of polystyrene spheres (in blue) yields a gold nanohole film array (in yellow) on top of the Si wafer (in gray) after colloidal lithography. (2) Si nanowires (in gray) are synthesized via metal assisted chemical etching using the gold nanohole film array as an etching mask. (3, 4) Si nanowire arrays (in gray) coated with a SiO2 shell (in blue) are embedded within a polymer film (orange). (5) The dissolution of the SiO2 shell generates annular pores. (6) The pores guide the electrodeposition of multisegmented shells around the Si nanowires starting from the gold film at the bottom (here, nickel is shown in black and gold in yellow). (7) Dissolution of the polymeric membrane and (8) selective etching of the sacrificial shell (here, nickel) leads to a well-defined metal shell (here, a gold ring). (b–d) Secondary electron SEM images of the various Si-metal microwires synthesized. Scale bars: 1 μm. (b) Si wires with a conformal Ni–Au shell and (c) after selective etching showing well-defined gold rings. (d) A low-magnification image showing an array patterned with a gold ring with similar dimensions but located at a different position along the wires.

Square-centimeter-scale Si wire arrays with well-defined diameters and spacing are prepared using a combination of colloidal lithography[41,42] and metal-assisted chemical etching (MACE),[43−45] as is schematically depicted in Figure a. Colloidal lithography is a low-cost technique that combines the large-scale self-assembly of polymeric spheres on planar surfaces (Figures S1 and S2) with a physical evaporation technique, which can be used to generate large-scale gold nanohole arrays.[45,46] Such gold nanohole arrays can be used to etch silicon via MACE. During the etching process, the metal film sinks into the Si, forming Si wires. This combined method allows for the preparation of Si wire arrays with tunable wire diameter, length, and array periodicity (pitch).[43] After the MACE process, the gold nanohole film (used as a mask during chemical etching) is located at the bottom of the wires.

The 3DEAL starts with the synthesis of a porous membrane with tunable dimensions around the Si wires (Figure a). The membrane synthesis requires three steps: the coating of the Si wires with a conformal SiO2 shell of tunable thickness grown via a sol–gel process (Figures S3 and S4); deposition of a conformal polycarbonate film to embed the Si@SiO2 wires (Figure S5); and selective etching of the SiO2 shell. This leads to the formation of a continuous membrane composed of open annular pores around each Si wire. At this stage of the synthesis, the nanostructured gold film lays flat at the bottom of the nanowires and can be externally connected to a potentiostat and used as a working electrode. This allows for electrodepositing multisegmented shells around the silicon wires, similar to the growth of metal shells around polymeric cores during COAL.[33] Prior to the electrodeposition, a thin protective SiO2 shell (∼3–4 nm thick) is grown around the embedded wire arrays to avoid parasitic growth of metal particles on the silicon wires. EDX analysis shows that the passivating layer present at the Si–metal interface can be removed with HF after 3DEAL (Figure S6). This result suggests that the electrical contact between the metal and the Si wire should be of acceptable quality. The shell segment thickness (i.e., outer diameter) is controlled by the SiO2 shell thickness, which can be adjusted in increments of ca. 10 nm (Figures S3 and S7). The shell height depends on the number of coulombs used during electrodeposition, which is set by the deposition time under potentiostatic conditions.[32,33,47] After dissolving of the polycarbonate, selective etching of the sacrificial shell using appropriate etchant solutions that leave the target material shell intact (Table S2) yields metal rings with well-defined dimensions and positions along the silicon wires (Figure ).

We demonstrate the synthesis of a variety of metal rings (Au, Ag, Ni, and Fe) around c-Si micro- and nanowire arrays (Figures 1, 2, and S8–S15) with controlled feature size, spacing, and position along the wires (listed in Table S1). There is no limit to the number and height of rings that can be prepared along the wires. For example, we synthesized Au ring heptamers (i.e., 7 rings) separated by 20 nm gaps around Si nanowires (230 nm diameter, Figure a). Multimetallic hybrid structures can also be synthesized, such as the Si@Ni–Fe nanowires shown in Figure f, where the Ni ring is located at the bottom of the wire and the Fe ring is at the top. The range of materials compatible with 3DEAL could be extended to metal oxides (TiO2, ZnO, and MnO) and conjugated polymers [polypyrrole, poly(3-hexylthiophene), polythiophene, and poly(3,4-ethylenedioxythiophene] that are chemically stable and can be electrochemically deposited.[30,32,33,48,49] Less-stable materials could potentially be integrated by using a polyaniline sacrificial shell that dissolves readily in acetone.[33] Additionally, the metal base layer can be selectively etched: For example, we prepared Si@Ni wire arrays without a metal base layer by selectively etching the gold film in a KI/I2 aqueous solution (Figure S16). Because MACE can be performed with a variety of metals,[44] it should be possible to find an appropriate combination of base layer metal and selective etchant for a specific ring target material. Unlike coaxial lithography that is limited to the modification of straight wires,[33] our method can be used to pattern kinked nanowires (Figure b), which can be produced via a modified version of MACE.[50] The gap length between the rings can be reliably controlled down to ca. 20 nm over large areas (Figure a), and the smallest negative feature achieved to date is a 5 nm gap synthesized between two gold rings (Figure S9); however, at these small dimensions, a significant fraction (i.e., > 50%) of the dimers were bridged, presumably due to local electrode inhomogeneities. The gap length and ring height dispersity is in the ∼15% range, which is typical for such electrochemically controlled processes.[30,31,33,47] Atomic force microscopy (AFM) measurements show that the Au shell surface is quite smooth (Figure S17) and has a root-mean-square roughness of 0.6 nm. Such a low roughness is most likely inherited from the smooth surface of the templating sacrificial SiO2 shell (Figures S3 and S4), which is properly transferred into the polycarbonate membrane after etching. This suggests that the Si@metal structures will be great candidates for plasmonic applications that require smooth metal films to obtain high-quality surface plasmon resonances.[51]

Figure 2

Materials and morphology library of Si@metal nanowire arrays (diameters shown ranging from 160 to 310 nm with a 590 nominal pitch). All scale bars: 250 nm. (a) Au ring heptamer (gap length: 22 ± 3 nm). (b) Au ring dimer on kinked nanowires. (c) Ni ring located at the bottom. (d) Fe ring located at the top. (e) Ag nano-ring dimers. (f) Fe ring at the top and an Ni ring at the bottom. (a–f) From left to right: structure scheme, STEM elemental map (white, Si; yellow, Au; blue, Ag; red, Fe; and green, Ni), HAADF STEM image, and cross-section SEM images showing (a) backscattered electron signal and (b–f) secondary electron signal.

We demonstrate the ability of the technique to generate optically active nanostructures within crystalline two-dimensional silicon nanowire arrays by patterning arrays with single gold rings and gold ring dimers and characterize them via confocal Raman microscopy (Figure ). The fast Fourier transform of an SEM image acquired at intermediate magnification shows that the Si@Au arrays prepared are crystalline and hexagonally packed (Figure S15). A total of four samples were prepared using the same Si nanowire array batch (nominal array pitch of 590 nm, nanowire diameter of ∼195 nm, and height of ∼3 μm): native Si nanowire arrays as produced after MACE with a gold film at the bottom of the wires, Si nanowire arrays prepared by etching the gold film in a KI/I2 solution, Si nanowire arrays patterned with single Au rings (Figure S14), and Si nanowire arrays patterned with gold ring dimers (Figures 3b and S15). The rings had an average height of ∼63 nm and thickness of ∼47 nm. The gap length between the two rings for the dimer was ∼30 nm. The photograph shown in Figure c demonstrates that the Au/Ni multisegmented shell deposition used to prepare the Au ring dimer sample is homogeneous over macroscopic dimensions (∼38 mm2). 1,4-Benzenedithiol (BDT) was used as the SERS target molecule due to its high affinity for gold surfaces and its well-defined Raman peaks.[31,36] Raman measurements were performed by focusing a 785 nm laser (2.5 μm diameter, 1 mW power, 3 × 20 s exposures) on top of the Si@Au arrays, expected to irradiate an average of 16 Si@Au wires. Noble-metal nanostructures can dramatically increase the incident and scattered E-field in the visible near-infrared range.[27,28] This can lead to large enhancement of the Raman signal, which approximately scales as |E|4, where E is the electric field at the metal surface.[34] Si nanowire arrays with and without a nanostructured gold film showed no measurable Raman signal, apart from the peaks between 900 and 1000 cm–1 that are due to the second-order Raman scattering of Si (Figure e).[52] The samples containing Au rings, however, showed a large Raman signal, which was highest for the Au ring dimers (Figure e). The homogeneity of the Au ring dimer sample was investigated by performing a Raman map over an area of ∼7.5 × 103 μm2 (3 μm step size, 900 spectra) by extracting the Raman signal of the 1563 cm–1 peak (C=C stretching vibration of the benzene ring structure).[53] We found an average signal of 177 ± 23 counts per second, corresponding to a relative standard deviation of 13% (Figures 3f and S18). Considering the complexity of the hybrid structures and the fact that SERS is highly sensitive to minute structural changes, the signal showcases remarkable uniformity. Full three-dimensional electromagnetic simulations, using the finite difference time domain (FDTD) method, suggest that the maximum near-field intensity enhancement at the 785 nm laser excitation wavelength is generated at the Si wire–Au ring interface and is around ∼780 and ∼930 for the single Au ring and the Au ring dimer, respectively (Figure ). For simplicity, only one Si@Au nanowire was modeled (see Figure S19 for the full maps). Assuming that most of the Raman signal at a 785 nm excitation wavelength originates from the BDT molecules that are adsorbed in the hot spot regions, located near the Si–Au interface (Figure a), we would expect the Au ring dimer sample to provide a Raman signal approximately 1.6 times larger than the signal measured at the single Au ring sample. Experimentally, we measured an increase in the Raman signal of ∼1.8 between the two samples, which matches the simulations. Our simulations also suggest that the Au rings concentrate the light within specific parts of the Si nanowires (Figure S19): the E-field intensity in the Si located above the rings (top 1.5 μm) is increased by a factor of ∼2 for a single Au ring and ∼3 for a Au ring dimer compared to a pristine Si nanowire, while it is nearly suppressed below the rings (bottom 1.5 μm of the Si wire). These results demonstrate the reliability and potential of the method to engineer plasmonic fields and tune light absorption in three dimensions within Si wire arrays over large areas.

Figure 3

Surface-enhanced Raman scattering on Si@Au nanowire arrays. (a) Schematic showing the geometrical parameters that can be adjusted. (b–d) Arrays patterned with Au ring dimers (30 nm gap). (b) Left: secondary electron STEM image. Center and right: HAADF STEM images. (c) Photograph of the sample after the electrodeposition of the sacrificial Ni and target Au shells, highlighting the large area covered with the hybrid nanowire array. (d) Low-magnification cross-sectional SEM images showing ∼100% yield of Si@Au ring dimers. (e) Typical unprocessed Raman spectra after the BDT functionalization of (from bottom to top) Si nanowires (black curve), native Si nanowires with a gold film at the bottom (magenta curve), a single-ring array (blue curve), and a 30 nm gap ring dimer array (red curve). The spectra are offset for clarity. (f) Smoothed two-dimensional Raman map at 1563 cm–1 (10× objective, NA of 0.25, and step size of 3 μm, corresponding to 900 spectra over a 87 μm × 87 μm area), showing that the average signal is homogeneous with a mean value of 177 ± 23 counts per second.

Figure 4

FDTD simulations. (a) E-field intensity enhancement maps at the laser excitation wavelength (i.e., 785 nm) for a single Au ring (left) and a Au ring dimer (30 nm gap, right) around a Si nanowire in air. The ring region only is shown for clarity. The ring dimensions are 70 nm height, 190 nm inner diameter, and 45 nm thickness. Both single rings and ring dimers are located at the middle of the Si wire (i.e., at 1.5 μm from the bottom of the wire). The bottom end of the top ring correspond to the position y = 0. The plane wave source propagates along the y axis (from top to bottom), and the electric field is polarized along the x axis. The logarithmic color scale of field intensity enhancement is the same for both maps. (b) E-field intensity enhancement line scan at the surface of the nanowire (1 nm distance from the surface) in the ring region, schematically depicted by the black arrows. Left: single ring (blue curve). Right: ring dimer (red curve), linear scale. The line scans are offset for clarity.

To conclude, we demonstrate the synthesis of complex hybrid c-Si@metal micro- and nanowire architectures. The method is based on sacrificial SiO2 coating and the formation of a porous membrane with tunable dimensions, which serves as a template for the electrochemical deposition of metal shells around the Si nanowires. Defined ring architectures can be produced by selective etching of the sacrificial metal shells. This is a shift in paradigm for templated syntheses that have, until now, relied on the direct replication of a template structure.[54] 3DEAL can be used to pattern well-defined pre-existing three-dimensional nanostructures with metal features as small as 5 nm. The technique is compatible with unconventional morphologies such as kinked nanowires and could be easily generalized to a variety of Si nanostructures with, for example, different cross-section morphologies[45] or conical shapes.[55] The remarkably small standard deviation obtained from Raman measurements underlines the high uniformity of the hybrid structures and demonstrates the reliability of the technique to control and shape electromagnetic near-fields in three dimensions. Our simulation show that this could be used to spatially control light absorption. The current experimental design requires a metal film at the bottom of the wire array, which is conveniently present after MACE. For this reason, our approach should be compatible with other important semiconductors that can be nanostructured via MACE such as GaAs or InP.[56,57] Because it is possible to selectively deposit a continuous and conductive metal film at the bottom of “bare” Si microwire arrays (i.e., without a metal base layer in the first place),[58] we believe that 3DEAL will be also compatible with other types of Si wire arrays made, for example, via vapor–liquid–solid growth[19] or photolithography and dry-etching processes.[21] Ultimately, our work is a proof-of-concept of a universal benchtop approach for the patterning of three-dimensional nanostructured architectures that have a continuously open pore system.