复旦大学：《纳米线材料和功能器件》课程教学资料_纳米线的组装_A nanoscale combing technique for the large-scale assembly of highly aligned nanowires

nature LETTERS nanotechnology PUBLISHED ONLINE: 21 APRIL 2013 I DOL: 10.1038/NNANO 2013.55 a nanoscale combing technique for the large-scale assembly of highly aligned nanowires Jun yao hao yan and charles m. lieber.2* The controlled assembly of nanowires is a key challenge in the applied force and speed. During contact and translation, the pro development of a range of bottom-up devices 2. Recent truding nanowire ends first stick to the surface. then 如mh8o9:rm图h4 pulled over the the anchoring a misalignment angle of +5 (refs 5, 12-14)and allowed force, is maximized due to the weak interaction with the various multiwire devices to be fabricated, 10-13, 9. However combing surface hese methods still create a significant number of crossing We first demonstrate the essential feature of the nanocombing defects, which restricts the development of device arrays and technique by transferring silicon nanowires. we used a functiona circuits based on single nanowires/nanotubes. Here, we show lized SiO, surface and a resist-layer surface as the anchoring and hat a nanocombing assembly technique, in which nanowires combing surfaces, respectively. The target substrate was patterned are anchored to defined areas of a surface and then drawn with an organic resist layer(Microposit $1805) where the exposed out over chemically distinct regions of the surface, can yield SiO2 regions, which were modified with tetramethylammonium arrays with greater than 98.5% of the nanowires aligned to ions during pattern development, served to improve the anchoring within +1 of the combing direction. The arrays have a crossing of silicon nanowires to the exposed SiO, regions during transfer(see defect density of 0.04 nanowires per um and efficient end Methods). The combing process was then carried out by sliding th registration at the anchoring /combing interface. With this nanowire growth substrate over the target substrate at a constant ve technique, arrays of single-nanowire devices are tiled over ocity and pressure, with high-viscosity hydrocarbon liquid added chips and shown to have reproducible electronic properties. between the two surfaces Scanning electron microscopy (SEM) images of the silicon nano nanowires to within +1 and to assemble suspended and array of anchoring regions(Fig. 1b-d)re a substrate with an deterministic assembly, to align ultralong (millimetre-scale) wire arrays obtained from nanocombing onto eral important crossed nanowire arrays. points. First, for constant 15-]m-length anchoring regions the Nanowires can be assembled in solution using electric or mag- nanowires show excellent alignment in the combing regions over netic fields 4.7, fluid flows., the Langmuir-Blodgett technique-10 lengths varying from 15 to 80 um and a total single array length and bubble- blown films, but creating high-density individually of 500 um. Qualitatively, the SEM images show extremely straight connected nanowire arrays efficient and scalable manner and parallel nanowires regardless of position within the array remains challenging. Mechanical printing 6 overcomes a number shown in Fig. 1b. Second, this high degree of nanowire alignment these problems and has been used to make multi-nanowire and was observed over macroscopic dimensions in the nanocombing multi-nanotube devices, including circuitry for artificial skin20 and process; that is, nanocombed nanowires in 40 arrays over a integrated circuits on plastic substrates. However, the alignment 3 mm x 11 mm substrate chip(Supplementary Fig. S1) ratios obtained with this approach(90-95% at a +5 misalign- exhibit alignment similar to that in Fig. 1b-d. Third, comparison ment angle. 3)can still lead to crossing defects that limit the appli- of the nanowires in the anchoring and combing regions cation of printing for single-nanowire devices (Supplementary Fig S2) clearly shows the lower alignment, higher were due to the fact that forces used to anchor and align nanowires ing region. All three observations are consistent witf. in the anchor- our model for exist in the same substrate(printing) regions; that is, the anchor- nanocombing and previous studies 23. Fourth, the upper portion of ing force necessary for the attachment of nanowires to the device Fig. 1b, where the length of the combing region(dark) is 80 ur substrate can oppose the aligning force and result in less effective (versus 15 um for the anchoring region), shows nanowires ar orientational control. Hence, improvements in the alignment of only present starting from and closer to the anchoring region( direc nanowires during assembly might be achieved by separating the tion of combing is upward in the image), and are largely absent from anchoring location of nanowires from the aligning region, an the remainder of the combing region. These observations highlight approach termed nanocombing in analogy to combing/ straighten- the stronger interaction of the anchoring region acting to trap nano- ing of hair anchored to the scalp or the molecular combing reported wires versus the weak interaction between the nanowires and th previously for DNA2. combing surface. There are several essential features of the nanocombing process Quantitative analysis of alignment determined from (Fig. la). First, separate anchoring regions with strong nanowire more than 1, 600 nanowires from arrays over the substrate chip interactions and aligning regions with weaker nanowire interactions (Fig. le)showed that -98.6% of the combed nanowires are aligned terned on the target substrate. Second, the to within +1 of the combing direction, representing better nanowire growth substrate is brought into contact with the target than tenfold improvement in alignment compared with previous substrate and translat he desired direction with a defined reports5 12-4.8(see Methods). Moreover, the high degree of alignme Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02138, USA, School of Engineering and Applied Science, Harvard University, Cambridge, Massachusetts 02138, USA. *e- mail: cml@cmliris harvard ed NatureNanotEchnOlogYIVol8iMay2013Iwww.nature.com/naturenanotechnology o 2013 Macmillan Publishers Limited. All rights reserved

A nanoscale combing technique for the large-scale assembly of highly aligned nanowires Jun Yao1 , Hao Yan1 and Charles M. Lieber1,2* The controlled assembly of nanowires is a key challenge in the development of a range of bottom-up devices1,2. Recent advances2–19 in the post-growth assembly of nanowires and carbon nanotubes have led to alignment ratios of 80–95% for a misalignment angle of +58 (refs 5,12–14) and allowed various multiwire devices to be fabricated6,10–13,19. However, these methods still create a significant number of crossing defects, which restricts the development of device arrays and circuits based on single nanowires/nanotubes. Here, we show that a nanocombing assembly technique, in which nanowires are anchored to defined areas of a surface and then drawn out over chemically distinct regions of the surface, can yield arrays with greater than 98.5% of the nanowires aligned to within +18 of the combing direction. The arrays have a crossing defect density of ∼0.04 nanowires per mm and efficient end registration at the anchoring/combing interface. With this technique, arrays of single-nanowire devices are tiled over chips and shown to have reproducible electronic properties. We also show that nanocombing can be used for laterally deterministic assembly, to align ultralong (millimetre-scale) nanowires to within +18 and to assemble suspended and crossed nanowire arrays. Nanowires can be assembled in solution using electric or mag￾netic fields3,4,7, fluid flows5,6, the Langmuir–Blodgett technique8–10 and bubble-blown films14, but creating high-density individually connected nanowire arrays in an efficient and scalable manner remains challenging. Mechanical printing16 overcomes a number of these problems and has been used to make multi-nanowire and multi-nanotube devices, including circuitry for artificial skin20 and integrated circuits on plastic substrates21. However, the alignment ratios obtained with this approach (90–95% at a +58 misalign￾ment angle12,13) can still lead to crossing defects that limit the appli￾cation of printing for single-nanowire devices. We hypothesized that the alignment limitations in previous work were due to the fact that forces used to anchor and align nanowires coexist in the same substrate (printing) regions; that is, the anchor￾ing force necessary for the attachment of nanowires to the device substrate can oppose the aligning force and result in less effective orientational control. Hence, improvements in the alignment of nanowires during assembly might be achieved by separating the anchoring location of nanowires from the aligning region, an approach termed nanocombing in analogy to combing/straighten￾ing of hair anchored to the scalp or the molecular combing reported previously for DNA22. There are several essential features of the nanocombing process (Fig. 1a). First, separate anchoring regions with strong nanowire interactions and aligning regions with weaker nanowire interactions are lithographically patterned on the target substrate. Second, the nanowire growth substrate is brought into contact with the target substrate and translated in the desired direction with a defined applied force and speed. During contact and translation, the pro￾truding nanowire ends first stick to the anchoring surface, then the remaining length is pulled over the combing surface, where the aligning force, which is diametrically opposed to the anchoring force, is maximized due to the weak interaction with the combing surface. We first demonstrate the essential feature of the nanocombing technique by transferring silicon nanowires23. We used a functiona￾lized SiO2 surface and a resist-layer surface as the anchoring and combing surfaces, respectively. The target substrate was patterned with an organic resist layer (Microposit S1805) where the exposed SiO2 regions, which were modified with tetramethylammonium ions during pattern development, served to improve the anchoring of silicon nanowires to the exposed SiO2 regions during transfer (see Methods). The combing process was then carried out by sliding the nanowire growth substrate over the target substrate at a constant vel￾ocity and pressure, with high-viscosity hydrocarbon liquid added between the two surfaces. Scanning electron microscopy (SEM) images of the silicon nano￾wire arrays obtained from nanocombing onto a substrate with an array of anchoring regions (Fig. 1b–d) reveal several important points. First, for constant 15-mm-length anchoring regions the nanowires show excellent alignment in the combing regions over lengths varying from 15 to 80 mm and a total single array length of 500 mm. Qualitatively, the SEM images show extremely straight and parallel nanowires regardless of position within the array shown in Fig. 1b. Second, this high degree of nanowire alignment was observed over macroscopic dimensions in the nanocombing process; that is, nanocombed nanowires in 40 arrays over a 3 mm × 11 mm area of a substrate chip (Supplementary Fig. S1) exhibit alignment similar to that in Fig. 1b–d. Third, comparison of the nanowires in the anchoring and combing regions (Supplementary Fig. S2) clearly shows the lower alignment, higher crossing defects and slightly higher nanowire density in the anchor￾ing region. All three observations are consistent with our model for nanocombing and previous studies12,13. Fourth, the upper portion of Fig. 1b, where the length of the combing region (dark) is 80 mm (versus 15 mm for the anchoring region), shows nanowires are only present starting from and closer to the anchoring region (direc￾tion of combing is upward in the image), and are largely absent from the remainder of the combing region. These observations highlight the stronger interaction of the anchoring region acting to trap nano￾wires versus the weak interaction between the nanowires and the combing surface. Quantitative analysis of alignment determined from analysis of more than 1,600 nanowires from arrays over the substrate chip (Fig. 1e) showed that 98.6% of the combed nanowires are aligned to within +18 of the combing direction, representing better than tenfold improvement in alignment compared with previous reports5,12–14,18 (see Methods). Moreover, the high degree of alignment 1 Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02138, USA, 2 School of Engineering and Applied Science, Harvard University, Cambridge, Massachusetts 02138, USA. *e-mail: cml@cmliris.harvard.edu LETTERS PUBLISHED ONLINE: 21 APRIL 2013 | DOI: 10.1038/NNANO.2013.55 NATURE NANOTECHNOLOGY | VOL 8 | MAY 2013 | www.nature.com/naturenanotechnology 329 © 2013 Macmillan Publishers Limited. All rights reserved.

ETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO201355 Transfer substrate Misalignment angle Figure 1 I Schematics and demonstration of nanocombing a, Schematics of the nanocombing process. The blue arrow indicates the travelling direction of the growth substrate with respect to the target substrate, which yields a combing/aligning force that is parallel and opposite to the anchoring force. The dashed window at the right bottom shows a side view of the nanocombing process. b-d, SEM images of silicon nanowires on the combing (resist)surface at different magnifications. The thickness of the resist(S1805) layer was -70 nm. Scale bars: 50 pm(b), 10 um(c), 2 um(d). e, Angle distribution of the rrays combed over a 3 mm x 11 mm chip, where the logarithm of nanowire(nw) number is plotte results in a low crossing defect density, -0.04 nanowire per um(see A summary of nanowire density as a function of length of anchoring Methods), which is also tenfold better than values achieved at region(Fig. 2a) shows that the density increases and then saturates similar nanowire densities by contact printing. Importantly and at-1.5 nanowires per um as the anchoring length increases from 2 as will be discussed further in the following, the highly aligned to 15 um; no further density increases were observed as the anchor nd low crossing defect density allows for the fabrication of high ing length was increased to 30 um. Interestingly, the average nano ields of reproducible individual nanowire devices. wire length in the combing region(Supplementary Fig. $3)is or We also explored several directions to determine key factors weakly dependent on anchoring length. Our analysis shows that, controlling nanowire density, as well as factors important to the for an average as length of 30 um, the average combing nanocombing process. We first investigated how the length of the length is x7 um( Supplementary Fig. S3). Considering the -50% anchoring region affects nanowire density in the combing region. reduction in nanowire length resulting from breakage by friction laTureNanotEchNologYIVol8IMay2013Iwww.nature.comnaturenanotechnoloy o 2013 Macmillan Publishers Limited. All rights reserved

results in a low crossing defect density, 0.04 nanowire per mm (see Methods), which is also tenfold better than values achieved at similar nanowire densities by contact printing18. Importantly and as will be discussed further in the following, the highly aligned and low crossing defect density allows for the fabrication of high yields of reproducible individual nanowire devices. We also explored several directions to determine key factors controlling nanowire density, as well as factors important to the nanocombing process. We first investigated how the length of the anchoring region affects nanowire density in the combing region. A summary of nanowire density as a function of length of anchoring region (Fig. 2a) shows that the density increases and then saturates at 1.5 nanowires per mm as the anchoring length increases from 2 to 15 mm; no further density increases were observed as the anchor￾ing length was increased to 30 mm. Interestingly, the average nano￾wire length in the combing region (Supplementary Fig. S3) is only weakly dependent on anchoring length. Our analysis shows that, for an average as-grown length of 30 mm, the average combing length is 7 mm (Supplementary Fig. S3). Considering the 50% reduction in nanowire length resulting from breakage by friction Anchoring region Combing region Transfer substrate + Growth substrate (flip over) a b d e c Growth substrate Transfer substrate 0 5 10 15 20 1 10 100 1,000 NW number Misalignment angle (°) Figure 1 | Schematics and demonstration of nanocombing. a, Schematics of the nanocombing process. The blue arrow indicates the travelling direction of the growth substrate with respect to the target substrate, which yields a combing/aligning force that is parallel and opposite to the anchoring force. The dashed window at the right bottom shows a side view of the nanocombing process. b–d, SEM images of silicon nanowires on the combing (resist) surface at different magnifications. The thickness of the resist (S1805) layer was 70 nm. Scale bars: 50 mm (b), 10 mm (c), 2 mm (d). e, Angle distribution of the combed nanowires obtained from analysis of nanowire arrays combed over a 3 mm × 11 mm chip, where the logarithm of nanowire (NW) number is plotted with respect to misalignment angle (see Methods for details). LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2013.55 330 NATURE NANOTECHNOLOGY | VOL 8 | MAY 2013 | www.nature.com/naturenanotechnology © 2013 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOl: 10.1038/NNANO201355 LETTERS window for nanocombing. At higher contact pressures, L>4m) c10 N cm, we observe a decrease in density, which 's that simply forcing the two surfaces together cannot yield an increased number of nanowires per unit area. Additional studies are needed to define the origin of the decreased density at higher pressures( for example, enhanced fragmentation of nano- wires on the growth substrate and or anchoring region), but the density-independent range of pressures is indicative that surface interactions within the anchoring region are dominant. Third, we have carried out several experiments to better illumi nate factors important to the nanocombing process. when the pat terned SiO2 surface is developed for a minimum amount of time (8 s versus 50 s)and then more extensively rinsed with deionized water(30 s versus 10 s)to minimize the effect of tetramethyl- ammonium hydroxide in the developer, we find a much lower nanowire density(0.3 nanowires per um), using the optimum anchoring length and overall resist thickness discussed previously Anchoring length (um) Furthermore, when the developed SiO, surface of the anchoring regions is rinsed with dilute potassium KOH solution(50 s, 1.5 wt% KOH), we observe a further increase in the combin density to almost 2 nanowires per um(Fig. 2b, star). Contact angle measurements(Supplementary Fig. S4)further show that higher densities can be associated with the more hydrophilic anchoring region, and also confirm that the combing regions are hydrophobic. Moreover, measurements carried out using a mineral oil/octane mixture as well as isopropyl alcohol as the transfer liquid( Supplementary Fig. S5)suggest that the liquid polarity, but not viscosity, is most important in enhancing the of nanowires in na studies suggest that further es in density may be achieved through variations in the anchoring-surface chemistry 4. Finally, experiments carried out with nanocombing rate varied by factor of ten( Supplementary Fig S6)show that the density and alignment ratio are approximately independent of the combing peed, thus suggesting that following attachment in the anchor region the necessary/required alignment force is lower than the minimum shear value accessed (consistent with a weak nano- Wire-Co combing (resist)surface with respect to different anchoring lengths The highly aligned nanowires produced over large areas by nano The statistics are based on nanowires with lengths >4 um. Data are from combing open up the potential to fabricate hierarchical arrays of the sample shown in Fig. 1b-d. b, Nanowire density with respect to different single-nanowire devices. Compared with previously reported assem resist/combing layer thicknesses. The anchoring surfaces were modified bly methods,8-1, which generally lack end-to-end registrationfor using the same conditions in all cases. The star indicates the combed individual nanowires, nanocombing ensures that the aligned nano- nanowire density by anchoring-surface modification using 10 s development wires in an array start from the anchoring/corr \ o can be and 50 s KOH treatment. Inset: nanowire density with respect to different patterned by lithography. Hence, registration of one combing pressures. The nanowire density comparisons in b are all based readily controlled in arrays patterned on a substrate, as shown in on the same anchoring length of 30 um. Fig 3a, b With the'start' of nanowires in the arrays defined by nano- combing, it is then straightforward to define a uniform 'end edge by reported previously, this yields an estimated ratio of - 1: I between lithography-based trimming to yield well-aligned and unifor anchoring length and combing length. Furthermore, we note that length nanowire arrays over the substrate surface(Fig. 3c, the apparent density decrease for anchoring lengths <10 um Supplementary Fig. S7). Obtaining uniform-length nanowires in Fig 2a)shows the critical role of the anchoring region in trapping every array using this approach places constraints on this length, nanowires during the nanocombing process. We also find that much because there are variations in the nanowire length from the anchor longer nanowires yield a corresponding increase in aligned, nano- ing/combing interface(Fig. 1),as already discussed. Despite thi combed length(discussed in the following limitation, we can readily obtain well-aligned and uniform nanowire Second, investigations of the nanowire density versus resist thick- arrays with lengths of 2-5 um, and the average combing length can less(Fig. 2b)show that the combing density is approximately con- be increased by increasing the length of nanowires on the growth tant up to a value of -70 nm, and decreases with increasing resist substrate. To estimate the device density for applications thickness above this value. This result is consistent with the nano-( Supplementary Information, Nanowire density analysis), an alter combing model (Fig. la), as an increased separation of the growth nating pattern of 10 um anchoring and combing lengths with 4 um urface reduces the efficiency of nanowire contact with and trapping resultant trimmed nanowire length yields 20% nanowire surface by the anchoring surface region. Below this value, the combing coverage and a device density of 7 x 10 cm This could density is largely independent of resist thickness. Moreover, increased to 27% nanowire surface coverage and a device density studies of nanowire density as a function of contact pressure of 1 x 10 cm- if a less conservative combing length of 5 ur (Fig. 2b, inset)show that the density is largely independent of were used. In either case, the device density values are 10-100x pressure between 2 and 6N cm which provides a relatively higher than that achieved previously 5. 7. NatureNanotEchnOlogYIVol8iMay2013Iwww.nature.com/naturenanotechnology o 2013 Macmillan Publishers Limited. All rights reserved

reported previously13, this yields an estimated ratio of 1:1 between anchoring length and combing length. Furthermore, we note that the apparent density decrease for anchoring lengths ≤10 mm (Fig. 2a) shows the critical role of the anchoring region in trapping nanowires during the nanocombing process. We also find that much longer nanowires yield a corresponding increase in aligned, nano￾combed length (discussed in the following). Second, investigations of the nanowire density versus resist thick￾ness (Fig. 2b) show that the combing density is approximately con￾stant up to a value of 70 nm, and decreases with increasing resist thickness above this value. This result is consistent with the nano￾combing model (Fig. 1a), as an increased separation of the growth surface reduces the efficiency of nanowire contact with and trapping by the anchoring surface region. Below this value, the combing density is largely independent of resist thickness. Moreover, studies of nanowire density as a function of contact pressure (Fig. 2b, inset) show that the density is largely independent of pressure between 2 and 6 N cm22 , which provides a relatively broad window for nanocombing. At higher contact pressures, above 10 N cm22 , we observe a decrease in density, which shows that simply forcing the two surfaces together cannot yield an increased number of nanowires per unit area. Additional studies are needed to define the origin of the decreased density at higher pressures (for example, enhanced fragmentation of nano￾wires on the growth substrate and/or anchoring region), but the density-independent range of pressures is indicative that surface interactions within the anchoring region are dominant. Third, we have carried out several experiments to better illumi￾nate factors important to the nanocombing process. When the pat￾terned SiO2 surface is developed for a minimum amount of time (8 s versus 50 s) and then more extensively rinsed with deionized water (30 s versus 10 s) to minimize the effect of tetramethyl￾ammonium hydroxide in the developer, we find a much lower nanowire density (0.3 nanowires per mm), using the optimum anchoring length and overall resist thickness discussed previously. Furthermore, when the developed SiO2 surface of the anchoring regions is rinsed with dilute potassium KOH solution (50 s, 1.5 wt% KOH), we observe a further increase in the combing density to almost 2 nanowires per mm (Fig. 2b, star). Contact angle measurements (Supplementary Fig. S4) further show that higher densities can be associated with the more hydrophilic anchoring region, and also confirm that the combing regions are hydrophobic. Moreover, measurements carried out using a mineral oil/octane mixture as well as isopropyl alcohol as the transfer liquid (Supplementary Fig. S5) suggest that the liquid polarity, but not viscosity, is most important in enhancing the density and alignment of nanowires in nanocombing. These studies suggest that further increases in density may be achieved through variations in the anchoring-surface chemistry24. Finally, experiments carried out with nanocombing rate varied by a factor of ten (Supplementary Fig. S6) show that the density and alignment ratio are approximately independent of the combing speed, thus suggesting that following attachment in the anchoring region the necessary/required alignment force is lower than the minimum shear value accessed (consistent with a weak nano￾wire–combing surface interaction). The highly aligned nanowires produced over large areas by nano￾combing open up the potential to fabricate hierarchical arrays of single-nanowire devices. Compared with previously reported assem￾bly methods5,6,8–14, which generally lack end-to-end registration9 for individual nanowires, nanocombing ensures that the aligned nano￾wires in an array start from the anchoring/combing interface patterned by lithography. Hence, registration of one end can be readily controlled in arrays patterned on a substrate, as shown in Fig. 3a,b. With the ‘start’ of nanowires in the arrays defined by nano￾combing, it is then straightforward to define a uniform ‘end’ edge by lithography-based trimming to yield well-aligned and uniform￾length nanowire arrays over the substrate surface (Fig. 3c, Supplementary Fig. S7). Obtaining uniform-length nanowires in every array using this approach places constraints on this length, because there are variations in the nanowire length from the anchor￾ing/combing interface (Fig. 1), as already discussed. Despite this limitation, we can readily obtain well-aligned and uniform nanowire arrays with lengths of 2–5 mm, and the average combing length can be increased by increasing the length of nanowires on the growth substrate. To estimate the device density for applications (Supplementary Information, ‘Nanowire density analysis’), an alter￾nating pattern of 10 mm anchoring and combing lengths with 4 mm resultant trimmed nanowire length yields 20% nanowire surface coverage and a device density of 7 × 106 cm22 . This could be increased to 27% nanowire surface coverage and a device density of 1 × 107 cm22 if a less conservative combing length of 5 mm were used. In either case, the device density values are 10–100× higher than that achieved previously15,17. 0 5 10 15 20 25 30 0.0 0.5 1.0 1.5 2.0 2.5 a b Density (μm−1) Anchoring length (μm) (L > 4 μm) 0.0 0.5 1.0 1.5 2.0 2.5 0 4 8 12 16 0.0 0.5 1.0 1.5 2.0 Density (μm−1) Resist thickness (nm) 20 50 100 200 500 Density (μm−1) Pressure (N cm−2) Figure 2 | Nanowire density control. a, Silicon nanowire density on the combing (resist) surface with respect to different anchoring lengths. The statistics are based on nanowires with lengths .4 mm. Data are from the sample shown in Fig. 1b–d. b, Nanowire density with respect to different resist/combing layer thicknesses. The anchoring surfaces were modified using the same conditions in all cases. The star indicates the combed nanowire density by anchoring-surface modification using 10 s development and 50 s KOH treatment. Inset: nanowire density with respect to different combing pressures. The nanowire density comparisons in b are all based on the same anchoring length of 30 mm. NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2013.55 LETTERS NATURE NANOTECHNOLOGY | VOL 8 | MAY 2013 | www.nature.com/naturenanotechnology 331 © 2013 Macmillan Publishers Limited. All rights reserved.

ETTERS NATURE NANOTECHNOLOGY DOI ANO2013.55 d 10-12⊥a=-100mV 圆■ igure 3 I Nanowire device arrays. a, Dark-field image of silicon nanowire arrays. The anchoring windows were defined by photolithography, and the resist layer(S1805) thickness was -70 nm Scale bar, 100 um. b, SEM image of one of the combed nanowire arrays on the resist layer Scale bar, 2 um. c, Dark-field image of trimmed nanowire arrays(resist layer removed). Scale bar, 40 um. d, Optical image of nanowire device arrays connecting to electrode arrays. Scale bar, 200 um. e, Representative SEM image of one of the device arrays. Scale bar, 2 um. f, ss-Vs characteristics (Vas=0.1 V) from 20 top-gated Ge/Si nanowire devices assembled by nanocombing. The channel length of the devices is 3. um, with AlO3(7 nm) serving as the dielectric layer for the top gate(Cr/Au= 5/50 nm). The electrical characterizations were performed in an ambient environment. As an example of the fabrication of device arrays, patterned, devices assembled by direct contact printing(1.2+ trimmed nanowire blocks were used to fabricate arrays of single- Moreover, the scaled on current of 0.40 mA um nanowire field-effect transistors (FETs). Photolithography was channel length of 1 um)is similar to that reported for singl used to define a periodic array of electrodes around each block of high-performance Ge/Si nanowire devices without consideration aligned nanowires(Fig. 3d). The individual nanowires were then of arrays. We note that the variation of on current(Fig. 3f)is onnected using electron-beam lithography. SEM images(Fig. 3e) similar to that seen in isolated nanowire devices, and probably demonstrate the high yield of individual nanowire devices for results from variations in contact resistance and individual nanowire each block, which is readily achieved due to the excellent nanowire properties2. Taken together, these results indicate that the nano- alignment and end-to-end registration. In other words, the trim- combing process preserves the critical electrical transport properties ming step, which is registered to the anchoring/ combing interface, of nanowire devices makes the fabrication nearly deterministic with respect to the We have also explored the potential of extending nanocombing y-coordinate(where y is defined as the nanowire axis). The typical to laterally deterministic assembly(and an overall nearly determi- yields of single-nanowire FETs achieved by this approach are nistic fabrication process)by defining the x-coordinates of nand 90%, and exceed the -60% achieved in our recent work using wires during nanocombing. Conceptually, laterally deterministic direct printing". We also applied nanocombing to assemble assembly can be enabled by controlling the widths of the anchoring e/Si core-shell nanowire arrays so as to directly compare the windows( Fig. 4a, left)such that narrow, periodic anchoring reproducibility of FET properties with previous work. Current windows produce a periodic or other predetermined pitch nanowire versus gate-voltage data from 20 representative devices(Fig. 3f) array. A representative SEM image of the silicon nanowire array pre how that the individual nanowire FETs have reproducible behav- duced by this method( Fig 4a, right) shows that a majority of the iour. Of particular importance to the development of integrated anchoring windows yield single, highly aligned nanocombed circuits is the threshold voltage, which has an average +1 s d (stan- silicon nanowires. Significantly, individual nanowire devices can dard deviation) of 1.7+0.2 V, where the variation in threshold is then be fabricated without registration to the nanowires, because substantially better than that reported recently for Ge/Si nanowire the pattern of contact electrodes simply matches the original AtuReNanotEchnOlogYIVol8iMay2013www.nature.com/naturenanotechnology o 2013 Macmillan Publishers Limited. All rights reserved

As an example of the fabrication of device arrays, patterned, trimmed nanowire blocks were used to fabricate arrays of single￾nanowire field-effect transistors (FETs). Photolithography was used to define a periodic array of electrodes around each block of aligned nanowires (Fig. 3d). The individual nanowires were then connected using electron-beam lithography. SEM images (Fig. 3e) demonstrate the high yield of individual nanowire devices for each block, which is readily achieved due to the excellent nanowire alignment and end-to-end registration. In other words, the trim￾ming step, which is registered to the anchoring/combing interface, makes the fabrication nearly deterministic with respect to the y-coordinate (where y is defined as the nanowire axis). The typical yields of single-nanowire FETs achieved by this approach are 90%, and exceed the 60% achieved in our recent work using direct printing25. We also applied nanocombing to assemble Ge/Si core–shell nanowire arrays so as to directly compare the reproducibility of FET properties with previous work. Current versus gate-voltage data from 20 representative devices (Fig. 3f ) show that the individual nanowire FETs have reproducible behav￾iour. Of particular importance to the development of integrated circuits is the threshold voltage, which has an average+1 s.d. (stan￾dard deviation) of 1.7+0.2 V, where the variation in threshold is substantially better than that reported recently for Ge/Si nanowire devices assembled by direct contact printing25 (1.2+0.7 V). Moreover, the scaled ON current of 0.40 mA mm21 (scaled channel length of 1 mm) is similar to that reported for single high-performance Ge/Si nanowire devices26 without consideration of arrays. We note that the variation of ON current (Fig. 3f ) is similar to that seen in isolated nanowire devices, and probably results from variations in contact resistance and individual nanowire properties23. Taken together, these results indicate that the nano￾combing process preserves the critical electrical transport properties of nanowire devices. We have also explored the potential of extending nanocombing to laterally deterministic assembly (and an overall nearly determi￾nistic fabrication process) by defining the x-coordinates of nano￾wires during nanocombing. Conceptually, laterally deterministic assembly can be enabled by controlling the widths of the anchoring windows (Fig. 4a, left) such that narrow, periodic anchoring windows produce a periodic or other predetermined pitch nanowire array. A representative SEM image of the silicon nanowire array pro￾duced by this method (Fig. 4a, right) shows that a majority of the anchoring windows yield single, highly aligned nanocombed silicon nanowires. Significantly, individual nanowire devices can then be fabricated without registration to the nanowires, because the pattern of contact electrodes simply matches the original −6 −4 −2 0 2 4 6 10−12 10−10 10−8 10−6 Ids (A) Vg (V) Vds = −100 mV a d e f b c Figure 3 | Nanowire device arrays. a, Dark-field image of silicon nanowire arrays. The anchoring windows were defined by photolithography, and the resist layer (S1805) thickness was 70 nm. Scale bar, 100 mm. b, SEM image of one of the combed nanowire arrays on the resist layer. Scale bar, 2 mm. c, Dark-field image of trimmed nanowire arrays (resist layer removed). Scale bar, 40 mm. d, Optical image of nanowire device arrays connecting to electrode arrays. Scale bar, 200mm. e, Representative SEM image of one of the device arrays. Scale bar, 2 mm. f, I ds2Vg characteristics (Vds ¼ 0.1 V) from 20 top-gated Ge/Si nanowire devices assembled by nanocombing. The channel length of the devices is 3.8mm, with Al2O3 (7 nm) serving as the dielectric layer for the top gate (Cr/Au ¼ 5/50 nm). The electrical characterizations were performed in an ambient environment. LETTERS NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2013.55 332 NATURE NANOTECHNOLOGY | VOL 8 | MAY 2013 | www.nature.com/naturenanotechnology © 2013 Macmillan Publishers Limited. All rights reserved.

NATURE NANOTECHNOLOGY DOl: 10.1038/NNANO201355 LETTERS Anchoring window 04(mm) First combing and cleaning Second combing igure 4 I Nanocombing applications a, Left: schematic of periodic silicon nanowire array by nanocombing. Right: SEM image of the resulting periodic anowire array on the resist(PMMA) surface. Each anchoring window has a 300 nm x 15 um(W x L) Sio, surface defined by electron-beam lithography, with the PMMa layer(thickness of 50 nm) serving as the combing surface. The Sio, surface was functionalized in MF- 319(50 s) and cleaned (20 s in deionized water) before combing Scale bar, 2 um. b, SEM image of a periodic silicon nanowire device array made from the combing method shown in a. c, Dark-field image of ultralong silicon nanowires on the combing surface (70 nm $1805). d, a millimetre- long combed silicon nanowire on the resist surface e, SEM image of a suspended silicon nanowire array. The resist layer has been removed. Scale bar, 1 um. f, Left: schematics of the two consecutive combing steps used to define a crossed nanowire array. The first layer of combed nanowires, produced in the standard manner, is treated as a substrate and processed in the standard manner for nanocombing the second crossed array of nanowires. Note that during the second perpendicular combing process, thicker resist layer (80 nm) was used, and the window was cleaned for 10 s in buffered oxide etch(BOe, 1:7, Transene)and modified with developer (MF-CD-26)before nanocombing. Right: SEM image of a silicon nanowire crossbar array. The resist layer has been removed. Scale bar, 1 um. The first combing was horizontal (from left to right)and the second combing was vertical (from bottom to top NatureNanotEchnOlogYIVol8iMay2013Iwww.nature.com/naturenanotechnology o 2013 Macmillan Publishers Limited. All rights reserved

0 0.2 0.4 (mm) 0.0 0.5 1 (mm) Anchoring window a b c d f e First combing and cleaning Second combing Figure 4 | Nanocombing applications. a, Left: schematic of periodic silicon nanowire array by nanocombing. Right: SEM image of the resulting periodic nanowire array on the resist (PMMA) surface. Each anchoring window has a 300 nm × 15mm (W × L) SiO2 surface defined by electron-beam lithography, with the PMMA layer (thickness of 50 nm) serving as the combing surface. The SiO2 surface was functionalized in MF-319 (50 s) and cleaned (20 s in deionized water) before combing. Scale bar, 2 mm. b, SEM image of a periodic silicon nanowire device array made from the combing method shown in a. c, Dark-field image of ultralong silicon nanowires on the combing surface (70 nm S1805). d, A millimetre-long combed silicon nanowire on the resist surface. e, SEM image of a suspended silicon nanowire array. The resist layer has been removed. Scale bar, 1 mm. f, Left: schematics of the two consecutive combing steps used to define a crossed nanowire array. The first layer of combed nanowires, produced in the standard manner, is treated as a substrate and processed in the standard manner for nanocombing the second crossed array of nanowires. Note that during the second perpendicular combing process, a thicker resist layer (80 nm) was used, and the window was cleaned for 10 s in buffered oxide etch (BOE, 1:7, Transene) and modified with developer (MF-CD-26) before nanocombing. Right: SEM image of a silicon nanowire crossbar array. The resist layer has been removed. Scale bar, 1 mm. The first combing was horizontal (from left to right) and the second combing was vertical (from bottom to top). NATURE NANOTECHNOLOGY DOI: 10.1038/NNANO.2013.55 LETTERS NATURE NANOTECHNOLOGY | VOL 8 | MAY 2013 | www.nature.com/naturenanotechnology 333 © 2013 Macmillan Publishers Limited. All rights reserved.