Could this revolutionize the semiconductor industry?

Over the past few decades, silicon nanowires (SiNWs) have been widely studied in nanoelectronics, optoelectronics, sensing/detection, biotechnology, and energy systems. It is worth noting that as the physical size of electronic devices becomes smaller, the impact of quantum confinement effects on electronic properties becomes more and more obvious. Theoretical and experimental studies have confirmed that when the diameter of SiNW approaches the carrier de Broglie wavelength (1nm for electrons), the indirect band gap of silicon can be adjusted to the direct band gap. The band gap of these small nanowires can be increased by several eV from the bulk value (Eg = 1.12 eV), placing their photoluminescence in the visible range.

Current SiNW synthesis methods use catalyst-nanoparticle-assisted vapor-liquid-solid (VLS) growth method or wet chemical etching process, and their diameter is limited by the size of the catalyst nanoparticles. As a result, typical SiNWs formed are quite large in diameter (10-100nm), where the impact of one-dimensional confinement is partial at best. Although valuable success has been achieved in the preparation of further reduced diameter SiNWs by using VLS growth of small catalyst nanoparticles, oxide sheaths of etch-shell SiNWs, or supercritical solution phase growth of Au nanoclusters, most of these synthetic methods Catalyst nanoparticles are still used, requiring complex purification processes to remove them, damaging and doping the SiNWs in the process. In addition, the produced SiNWs are not neatly arranged and have low growth density. This may be one of the key reasons why one-dimensional crystalline silicon at sub-5nm quantum confinement dimensions is experimentally poorly understood.

Here, we report a chemical vapor etching (CVE) process that can form high-density and vertically aligned ultranarrow silicon nanowires (SiNWs). Sub-5 nanocrystalline SiNWs (aspect ratio greater than 10,000) up to tens of microns long can be formed directly in Si wafers without any catalyst nanoparticles. The formation mechanism is different from catalyst-assisted growth and etching methods. The SiNWs produced by the CVE process reported in this article have a lattice reduction rate of up to 20% exceeding that of conventional silicon, and have excellent oxidation stability compared with conventional silicon. We also report ultranarrow silicon nanowires with rich quantum confinement properties under strong one-dimensional confinement.

The scanning electron microscopy (SEM) images in Figure 1a–c demonstrate striking examples of ultra-high density and vertically aligned SiNWs, where the resulting arrays closely resemble a forest of high-density and vertically aligned single-walled carbon nanotubes. Surface-oriented SiNW forests were fabricated directly from Si wafers, vertically etched on a 9 μm height plane using SiCl gas for 1 h of high-temperature CVE in a highly controlled Ar and H environment ( Fig. 1d and Supplementary Fig. 1 ).

Close inspection at the bottom of the SiNW array (Fig. 1e, f) shows that very narrow nanowires are densely packed with high uniformity. After etching for 2 hours, the nanowires can be further extended to a height of 37 μm (Figure 1g). These SiNWs can be dispersed in solvents such as ethanol, and isolated nanowires can be obtained by simple sonication (Supplementary Figure 2). Figure 1h shows the process of oxidation-induced CVE synthesis of ultranarrow SiNWs. Based on equilibrium concentration calculations, the main etchants responsible for etching Si at 1400 K at normal pressure are SiCl and HCl (Supplementary Note 1 and Supplementary Tables 1, 2). HCl vapor etchant is first produced by the decomposition of SiCl 4 , with H 2 participating at around 1400 K , as well as other gas compounds (SiCl 3 , SiCl 2 and SiHCl 3 ).

The initial silicon substrate has a native oxide layer of 2-2.5 nm thickness (Supplementary Figure 3a, d). When the temperature increased to 1150°C in Ar atmosphere, we observed thermal desorption of the native oxide layer into a thickness of 1–2 nm (Supplementary Figure 3b, e). When SiCl 4 is introduced into the CVE reactor at a temperature of 1150°C, the oxide layer is further unevenly etched by HCl into a highly porous and high-density clustered oxide structure, which acts as a mask and allows a controlled amount of etchant to diffuse into Si (Supplementary Fig. 3c, f and Supplementary Table 3). From a representative set of cross-sectional TEM images, we extracted the diameter distribution (violin plot) of pores (average 2.55 nm) and oxide clusters (average 1.92 nm) (Supplementary Figure 3g).

Figure 1: Morphology of vertically aligned high-density SiNWs formed on (100) silicon substrate.

a Tilt, b low, and c high magnification plan views of vertically aligned SiNWs. d SEM image of vertically aligned SiNWs on (100) Si wafer after 1 h of etching. e High magnification SEM image of the bottom interface between SiNWs and the etched Si substrate. f Magnified image of the dashed rectangular area in (e). g Tilt view of vertically aligned SiNWs after 2 h of etching. h Schematic diagram of the oxidation-induced etching process for fabricating ultra-narrow Si nanowires.

The silicon substrate is then etched into nanowire structures by two main anisotropic etchants, SiCl 4 and HCl vapor, using porous silicon oxide as a mask, while producing by-products SiCl 2 , SiHCl 3 and H 2 . The continuation of the CVE process described above results in the extension of well-aligned nanowires. We speculate that as the oxidant concentration (>5.5ppm) increases, the oxidizing gas will oxidize and passivate the SiNWs surface, promoting the survival of the nanowires from the etchant vapor. The emergence of high-density nanowires will reduce the etchant concentration penetrating the nanowire array, easing the lateral etching of the nanowires and forming an extended nanowire array (Supplementary Figure 4a). However, when the oxidizing gas concentration exceeds the threshold, the entire Si surface will be oxidized, forming a continuous dense oxide film that is resistant to etching (Supplementary Figure 4b). Furthermore, we found that at very low O2 / H2O concentrations (<5ppm), major etching of Si occurred without nanowire formation (Supplementary Fig. 4c).

Direct evidence of the crystal structure of vertically aligned SiNWs with diameters less than 5 nm and lengths up to several microns was obtained using high-resolution transmission electron microscopy (HRTEM). The HRTEM image of the SiNW beam in Figure 2a shows the clear lattice fringes of the SiNW, reflecting the highly crystalline nature of the nanowire surface without obvious oxidation (Supplementary Note 2 and Supplementary Figure 5). The structural characteristics of the nanowires have been further investigated using selected area electron diffraction (SAED) mode. Diffraction rings of randomly dispersed SiNWs can be used to characterize the structure, with each radius corresponding exactly to the inter-plane distance dhkl. Three diffraction rings observed at 2.50 Å, 1.53 Å and 1.31 Å show the presence of {111}, {220} and {311} planes of the diamond cubic lattice (space group Fd3m) 35. The calculated lattice parameter is 4.33 Å, which is 79.7% of bulk Si (5.43 Å). The large anisotropic compression of the diamond cubic Si lattice under normal pressure and room temperature conditions is surprising.

By analyzing a large number of representative HRTEM images (Fig. 2b), we extracted the diameter distribution (violin plot) of nanowires in the range of 2-5 nm (average 3.44 nm), with a narrow diameter size distribution (relative standard deviation ~20.7 %). The diameter of most nanowires is equivalent to or smaller than the exciton Bohr radius (~5nm), which is the size range over which quantum confinement effects of tunable electronic and optical properties can be achieved. To confirm this special lattice reduction, we performed x-ray diffraction (XRD) analysis on SiNW samples. The XRD spectrum of SiNWs arranged vertically on the (100) Si substrate shows that the 2θ of the {111} plane moves from 28.45° (bulk Si, JCPDS card number 65-1060) to 32.96°, and the 2θ of the (220) plane moves from 47.31° moves to 54.99° (Fig. 2c). The strong (400) peak at 69.12° comes from the (100) silicon wafer substrate. The corresponding SiNW lattice constant is approximately 4.70 Å, again confirming the lattice reduction (13.4%).

Figure 2: Analysis of the vertically aligned high-density SiNW crystal structure

a HRTEM image and SAED pattern of polycrystalline SiNW. bViolin plot showing diameter distribution determined from HRTEM. X-ray diffraction pattern of vertically aligned SiNWs with C on (100) Si substrate. The strong (400) peak comes from the (100) silicon wafer substrate. dThe relationship between lattice reduction rate and SiNW diameter. e HRTEM and corresponding FFT images of a single silicon nanowire. f Schematic diagram of the crystal orientation of vertically aligned (100) SiNWs on a (100) Si substrate as observed with a relatively rough edge direction <110>.

Past research has shown that the etch rate of silicon wafers is very sensitive to local strain. Specifically, tensile strain results in significantly faster etching rates. Therefore, we hypothesize that continued formation of nanowires after initial etching requires an exposed surface with tensile strain, stabilized by lattice contraction of the bulk lattice. The large lattice shrinkage we observe may arise from the increased demand on the tensile strained surface as the nanowire size decreases. For low-dimensional crystals, surface stresses and associated surface reconstructions (such as dimerization in silicon) lead to residual strains that become significant as dimensions approach a few nanometers. In extreme cases, they can lead to nonlinear elastic deformation, superelasticity and even lattice phase transitions. The size dependence of lattice deformation is a characteristic of these effects, and to see whether the shrinkage we observed is related to the nanowire size, we analyzed several HRTEM images of SiNWs along their length. It can be seen from Figure 2d that the lattice reduction rate (%) increases with the decrease of the local diameter D of the nanowire, and the increase is approximately linear, that is,

The net reduction is between 13-20%. Note that at a fixed diameter, there are considerable fluctuations in the measured lattice shrinkage, which may be attributed to the wobbles in the morphology of the randomly positioned nanowires. In addition, the etchant actively attacks and removes surface Si atoms, which results in slightly uneven diameters of the nanowires and also increases the complexity of the evaluation. The size dependence shows that the surface effect is the origin of the lattice shrinkage stabilization, and the stress-induced etching path is appropriately used to form Si surface nanowire clusters and atomically thin Si subvalent oxide sheaths. When the nanowire diameter approaches several nanometers, the sheath is Zoom in (Supplementary Figure 5). A combination of XRD characterization, TEM measurements of lattice stripes along the axial and radial directions, and the size dependence of shrinkage indicate that the nanowire units maintain a cubic state as they shrink. Therefore, the nanowire is compressed uniformly along the axial and radial (and therefore also azimuthal) directions, with the resulting strain scaling by Poisson's ratio. We attribute this hydrostatic stress state to the etching reaction chemistry, most likely stabilized by the oxide layer formed after the surface reaction, and entrust a more detailed stability analysis of the compressed nanowires to future studies.

For further probing, HRTEM images along the [110] band axis of the crystalline SiNW and perpendicular to the long axis of the nanowire revealed fringes from the {111} plane. HRTEM of multiple individual nanowires revealed vertically aligned SiNWs with corresponding interplanar spacing d111 of approximately 0.26 nm (Fig. 2e and Supplementary Fig. 6). The corresponding fast Fourier transform (FFT) image shows that the nanowires were formed in the [100] direction, the same direction as the original silicon wafer. This etching technique has little dependence on the doping characteristics of the silicon substrate, i.e., SiNWs can be obtained from n-type, p-type and highly doped p-type Si wafers (Supplementary Figure 7). As shown in Figure 2e, atomic-level roughness on the SiNW surface can usually be observed. The rough morphology of SiNW sidewalls may lead to scattering of electrons and/or phonons. Anisotropic etching of silicon surfaces depends on the removal rate of silicon atoms related to bond strength theory. Specifically, the {100} surface has only two bonds connected to the substrate, while other surfaces, such as the {111} surface, have three bonds. Since the surface Si atoms are bonded to Cl−, the weakening effect on the Si back bond on the {100} surface is more obvious. In addition, the surface energies of Si{100}, {110} and {111} are 1.99, 1.41 and 1.15 J cm ^{− 2} respectively . The Si{100} plane with higher surface energy and surface bond density (1.36×1015cm ^{− 2} ) provides more reaction sites, resulting in faster Si atom removal speed and anisotropic <100> etching (Fig. 2f ).

Structural fingerprinting of ultra-narrow silicon nanowires was performed using Raman spectroscopy at an excitation wavelength of 532 nm. All measurements were performed at low laser power and room temperature to eliminate thermal effects (Supplementary Figures 8, 9 and Supplementary Tables 4, 5). Figure 3a compares the Raman spectra of bulk Si and SiNWs characterized under the same excitation energy but different exposure times. Due to the scattering of first-order optical phonons, a representative Raman peak appears at 520cm−1 in bulk crystal Si. However, in our SiNWs, the corresponding Raman peak is red-shifted by nearly 15 cm ⁻¹ compared with bulk Si . Its line width becomes wider (FWHM is 12.4cm ⁻¹ ), and the line shape becomes asymmetric; transverse acoustic phonon mode (2TA, from 302 to 290cm ⁻¹ ) and transverse optical phonon mode (2TO, Second-order spectral redshift from 969 to 933cm ⁻¹ ). The wavenumber decrease of this Raman peak is mainly due to the phonon confinement effect, mainly the diameter reduction. When the crystal size shrinks to the nanometer scale, the momentum conservation rules are relaxed, and phonon scattering is not limited to the center of the Brillouin zone, and phonon dispersion near the center is also considered. The smaller the crystal size, the greater the wave number of the Raman peak shifts downward, the greater the asymmetry of the Raman peak, and the greater the width of the Raman peak.

Figure 3: Optical properties and band gaps of vertically aligned ultranarrow SiNWs.

a Raman characterization results show an obvious red shift of the SiNW peak band. b Photoluminescence spectrum of ultra-narrow dispersion in ethanol with excitation energy of 5.17 eV. c Photographs of SiNWs prepared on silicon and SiNWs dispersed in ethanol under UV light. d Combined UPS and IPES spectra showing the quasiparticle band gap of SiNWs with conduction/valence band bias. e UV-visible absorption spectrum of SiNW. Tauc diagram represents the direct band gap transition of SiNW. fSiNW energy level diagrams derived from UPS, IPES and PL. Eg,t, Eg, opt and Eb are the transmission band gap, optical band gap and binding energy respectively.

When the diameter of SiNW approaches the carrier de Broglie wavelength, the band gap of SiNW is renormalized due to the quantum confinement effect. We also expect that subcritical diameter SiNWs have direct band gaps that increase as the nanowire diameter decreases, regardless of surface endpoints. As the diameter decreases, the band gap of the nanowire gradually broadens and deviates from the band gap of the bulk silicon. As shown in Figure 3b, using an excitation wavelength of 240nm (5.17eV) at room temperature, the ultranarrow nanowire exhibits a strong photoluminescence (PL) peak center at 3.50eV and a weak shoulder at 3.8eV. When the indirect band gap of bulk Si is 1.12 eV, the PL peak undergoes an obvious blue shift, which proves the band gap renormalization of SiNW. From Figure 3c, it can be determined that the blue light emission amount of SiNWs prepared on Si substrate and SiNWs dispersed in ethanol under ultraviolet light (4.88 eV) corresponds to the emission energy of PL spectrum.

The density of states (DOS) of ultranarrow SiNW and the energy positions of its valence band maximum (VBM) and conduction band minimum (CBM) were studied using ultraviolet light emission spectroscopy (UPS) and inverse light emission spectroscopy (IPES). In order to determine the DOS characteristics of ultra-narrow SiNWs and the characteristics of their surface oxidation, native SiO2 was used as a reference. In both cases, obvious O 2p valence characteristics were observed around 7-9ev and -3~-4ev for UPS and IPES respectively in Figure 3d. At the combination energy of VBM of 4.35eV and CBM of 0.40eV, their energy level initiation centered on the Fermi level was found, indicating that the quasi-particle band gap of ultra-narrow SiNW is 4.75eV. The combination of UPS and IPES spectra again confirms the band gap renormalization as it directly indicates the DOS of SiNWs.

Furthermore, we were unable to observe clear features between VBM and CBM onsets, which may be due to low or no gap state density due to defects or impurities within the experimental error range. However, further research is needed to fully understand defects and impurities in SiNWs. To determine the optical bandgap energy, Tacu plots were obtained from the UV-visible absorption spectra (Fig. 3e, Supplementary Note 3). By extrapolating the linear region to the abscissa of photon energy, a direct optical band gap of 4.16 eV can be extracted. Furthermore, in Figure 3f, by comparing the optical band gap of SiNW obtained from the absorption spectrum with the quasi-particle band gap (Eb=Eg,t−Eg,opt), a large exciton binding energy of 0.59 eV is estimated, which is the quantum limit One of the fingerprints of the effect. Surprisingly, it is about 100 times higher than bulk Si (0.0055 eV). We believe that these greatly increased quasiparticle band gaps of 4.75eV and exciton binding energies of 0.59eV may be due to a combination of factors, including efficient quantum confinement dimensions, extraordinary lattice reduction (the energy gap caused by shortening of the distance between atoms) increase) and the dielectric shielding effect from the Si/SiOx core-shell structure of SiNW.

To investigate the stability of SiNWs in air, SiNWs were exposed to ambient air (room temperature 22 °C, relative humidity 40-50%) for up to two months without any surface modification/termination, and HRTEM images of SiNWs were recorded accordingly (Figure 4a). The freshly prepared SiNWs had clear lattice fringes without obvious amorphous oxide shell. After immersion in air for 7 days, 30 days, and 60 days, the same SiNW surface was slowly oxidized, and the estimated oxide thickness (and corresponding silicon core diameter measurement) was 5.0Å (3.75nm), 12.8Å (3.03nm), and 14.9Å (2.84nm), respectively. As shown in Figure 4b, the surface oxidation rate of SiNWs decreased with time, which may be due to the self-limiting oxidation effect caused by the Si/SiOx interface. It has been reported that when bulk Si is cleaved and exposed to air (25 °C, relative humidity 30-50%), the surface of Si is immediately oxidized, forming 11-13Å thick silicon oxide within 24h. Furthermore, hydrogen-passivated Si oxidizes by up to 7.6 Å after 24 h in air and by up to 11 Å within two weeks in air. During the first 7 days in ambient air, our ultra-narrow SiNWs have an average oxidation growth rate of about 0.7 Å/day, which is at least 58% and 25-43% lower than the oxidation rates of bulk Si and hydrogen-passivated Si, respectively.

Figure 4: Oxidation stability of vertically aligned ultra-narrow SiNWs.

a HRTEM image showing the stability of ultranarrow silicon nanowires under room temperature air exposure. Note that the same SiNWs are observed in HRTEM for different durations. The dimensions of the silicon core were measured. b Oxide layer thickness as a function of ambient air exposure time.

In this study, we demonstrate high density, alignment, and catalyst-free synthesis of sub-5nm SiNWs by developing a vapor phase silicon etching process. SiNWs are oriented along the [100] direction and the lattice is reduced by 13-20%, which may improve the stability of the nanowires against corrosion and oxidation. These sub-5nm SiNWs with special lattice reduction exhibit significant phonon and electron confinement effects, which may facilitate potential applications in nanoelectronics and optoelectronics, such as transistors and biosensors. Furthermore, macroscopic films containing billions of such small SiNWs in a highly aligned manner could also become promising material systems for gas/chemical sensors, anodes for lithium-ion and lithium-sulfur batteries, and solar cells, where Strong quantum confinement and ultrahigh silicon surface area are both desirable.

By developing a chemical vapor silicon etching process using silicon tetrachloride (SiCl 4 ) gas in a highly controlled argon (Ar) and hydrogen (H2) environment at 1000-1150°C, silicon wafers with diameters smaller than Preparation of high-density and well-aligned SiNWs of 5 nm and lengths up to tens of microns. The silicon wafer is located in the center of the furnace. In a typical manufacturing process, the reaction chamber is pumped down to a base pressure of 10-2 Torr and backfilled with ultra-high purity Ar gas (500 sccm) until normal pressure. When the chamber temperature reaches 1000-1150°C, introduce SiCl4 vapor into the chamber by bubbling 20 sccm of 5-10% H2 balanced Ar gas through a glass bubbler . After etching, a flow of 500 sccm of ultra-high purity Ar gas was passed through the chamber to help cool the chamber.

The morphology of SiNWs was characterized using a thermal field emission scanning electron microscope (Supra 25 FE-SEM, Zeiss). For structural analysis, an aberration-corrected transmission electron microscope (TEM, FEI Titan Themis 300) was used to obtain direct crystalline silicon structures. The chemical composition of the nanowires was determined by an energy-dispersive X-ray spectrometer (EDS) attached to the TEM. SiNWs were dispersed in ethanol (Reagent Alcohol, anhydrous, ≤0.005% water, Sigma-Aldrich) and dropped onto an ultrathin carbon layer (3 nm) supported by Lacey carbon and copper grids (Ted Pella, inc. ). A scanning electron microscope (SEM)/focused ion beam (FIB) dual-beam system was used to prepare a transmission electron microscope (TEM) cross-section sample with Ga ion beam as the beam source. To protect the upper surface of the cross-sectioned sample, a 200 nm thick Pt layer was first deposited using an electron beam and then a 2 μm thick Pt layer was deposited using an ion beam. Cross-sections rough-cut and cleaned using a 30kV gallium ion beam.

Finally, a 5kV Ga ion beam was used for cross-section polishing. High-resolution TEM (HRTEM) images were taken at 300kv, and EDS mapping was performed in STEM mode. In order to study the oxidation resistance of nanowires in air, HRTEM images were recorded on the same nanowires after exposing the above grids to ambient air for different times from 1 week to 2 months. Use ImageJ to perform background noise reduction processing on HRTEM images. The XRD spectra of SiNWs on Si substrate were recorded using a high-resolution X-ray diffractometer (Rigaku SmartLab) under Cu Kα radiation (λ = 1.54 Å) using theta-2 θ scanning. X-ray photoelectron spectroscopy (XPS) analysis of the nanowires was performed using an XPS spectrometer (Thermo Fisher Scientific K-Alpha+). SiNWs were first dispersed in ethanol and then dropped onto a highly oriented pyrolytic graphite (HOPG) substrate. Optical measurements were performed using a Raman spectrometer (Jobin Yvon HR 800, Horiba) and photoluminescence (Hitachi F7000 fluorescence spectrophotometer).

For the Raman sample, a stainless steel razor blade was chosen as the substrate, which does not have any strong peaks near the typical Si peaks near 300, 520, and 960 cm−1. SiNWs were then collected from the etched silicon substrate surface, dispersed in ethanol using a weak ultrasonic process, and placed on the substrate for Raman measurements. In order to reduce the impact of laser Raman scattering on heating, the data were collected in the low laser flux P=1.87mW region, and the band shape in this region has nothing to do with P. For PL samples, SiNWs were dispersed in ethanol by ultrasound and excited at a wavelength of 240 nm. Luminescence detection was performed on SiNW dispersion in a UV quartz test tube under an ultraviolet light source (UV lamp, 254 nm) in a dark room. The silicon wafer was used as a reference, and SiNWs on the silicon substrate and SiNWs dispersed in ethanol solvent were used as samples. The UPS measurement uses a PHOIBOS 150 hemisphere (SPECS GmbH), in which a He I (hν=21.22eV) discharge lamp is used as the excitation source. IPES was performed in isochromatic mode using a low-energy electron gun with a BaO cathode and a bandpass filter (SrF2+NaCl) of 9.5 eV .

By measuring the Fermi edge of clean Au films, resolutions of 100meV for UPS and 750meV for IPES were determined. The spectral binding energies given are calibrated to the Fermi level. In order to avoid sample charging, SiNWs were prepared on HOPG substrates using an ethanol-dispersed spin coating method. The transmittance of the nanowires was measured using a JASCO V-770 UV-visible spectrophotometer. To prepare the sample, a thin SiNWs film was prepared on a quartz substrate using a drop-cast ethanol dispersion method, and then its signal was subtracted from the spectrum.