Scalable Fabrication of Metallic Nanogaps at the Sub-10 nm Level

Abstract

Metallic nanogaps with metal-metal separations of less than 10 nm have many applications in nanoscale photonics and electronics. However, their fabrication remains a considerable challenge, especially for applications that require patterning of nanoscale features over macroscopic length-scales. Here, some of the most promising techniques for nanogap fabrication are evaluated, covering established technologies such as photolithography, electron-beam lithography (EBL), and focused ion beam (FIB) milling, plus a number of newer methods that use novel electrochemical and mechanical means to effect the patterning. The physical principles behind each method are reviewed and their strengths and limitations for nanogap patterning in terms of resolution, fidelity, speed, ease of implementation, versatility, and scalability to large substrate sizes are discussed.

Keywords: nanoelectronics; nanofabrication; nanogap; nanophotonics; plasmonics.

Figure 1

a) Schematic of shadow‐mask photolithography, in which i) a substrate is coated with a target metal and a photoresist, ii) the resist is selectively exposed with UV light via a photomask to modify the solubility of the resist, and iii,iv) the entire stack is immersed in a developer to remove the soluble parts of the resist. In the case of an (initially insoluble) positive resist, the exposed parts are rendered soluble (iii), while in the case of an (initially soluble) negative resist, the exposed parts are rendered insoluble. The exposed metal may be subsequently removed by chemical etching prior to removal of the resist. b) The smallest spot size D to which a point source may be focused by a mirror or lens is ∼λ/(2nsinθ), where λ is the wavelength of the illuminating light, n is the index of refraction of the surrounding medium, and θ is the half‐angle of the light cone entering the optic. c) Schematic of an Extreme‐UV projection lithography (EUV‐PL) system showing a laser‐produced plasma source that generates light of wavelength 13.5 nm, a collector mirror that focuses the light, illuminator optics that direct the light onto a reflective mask, and projection optics that redirect the light reflected by the mask to a wafer. The entire system operates in vacuum.^{[ 36 ]} Inset of (c) shows 7‐nm transistors with a 30 nm pitch fabricated by EUV projection lithography. c) Main image: Reproduced with permission.^{[ 36 ]} Copyright 2010, Springer Nature. Inset image is courtesy of IBM.

Figure 2

a) Schematic of extreme UV interference lithography (EUV‐IL), in which coherent light from an EUV light‐source strikes two closely spaced gratings, resulting in interference fringes where the diffracted beams overlap. The periodicity of the first‐ and second‐order fringe patterns are equal to one‐half and one‐quarter the grating periodicity, respectively.^{[ 38 ]} b) SEM images of line arrays with pitches of i) 16 nm, ii) 14 nm, and 12 nm, obtained by EUV‐IL using respective grating pitches of 64, 56, and 48 nm, and second‐order overlap.^{[ 39 ]} c) Schematic showing glancing‐angle metal deposition onto an EUV‐IL patterned line‐array of HSQ resist. d) Schematic showing appearance of line‐array after glancing‐angle deposition. e) SEM image showing nanogap line‐array obtained by glancing‐angle metal deposition, using an EUV‐IL patterned line‐array with an initial pitch of 250 nm. After metal deposition, the gap‐size has been reduced to approximately 12 nm.^{[ 40 ]} a) Reproduced with permission.^{[ 38 ]} Copyright 2011, IOP Publishing Ltd. b) Reproduced with permission.^{[ 39 ]} Copyright 2016, Proceedings of SPIE. e) Reproduced with permission.^{[ 40 ]} Copyright 2011, American Institute of Physics.

Figure 3

a) Schematic of Holographic Optical Element (HOE) used for three‐beam interference lithography,^{[ 47 ]} comprising a 6‐cm × 6‐cm quartz slide with three etched phase‐gratings (X, Y, and Z) arranged at 120° to one another. b) Schematic showing experimental set‐up for three‐beam interference lithography. The HOE is illuminated with an expanded 266‐nm laser beam, and an SU8‐coated substrate is placed a distance of 3.95 cm from the HOE where the three first‐order diffraction beams overlap, generating an interference pattern with hexagonal symmetry. c) Schematic of the three‐beam interference pattern, in which zones A, B, C, and D correspond to regions of high, medium, low, and very low intensity according to the degree of constructive or destructive overlap between the beams. d) Typical scanning electron micrograph showing the SU8 pattern due to a short exposure time, in which cross‐linking occurs only in the highest intensity zones A and B, generating isolated, cylindrical rods of SU8. e) Typical scanning electron micrograph showing the SU8 pattern due to a longer exposure time, in which additional cross‐linking occurs in zone C, generating bridges between the cylindrical rods; weakly exposed SU8 in zone D is removed during developing, creating holes on each side of the bridges. f,g) Scanning electron micrographs showing a typical shadow‐mask template obtained by three‐beam interference lithography and the corresponding gold nanoarray. Adapted with permission.^{[ 47 ]} Copyright 2020, Royal Society of Chemistry.

Figure 4

a) Schematic showing experimental set‐up for four‐beam interference lithography, in which three of the beams are oriented at 120° to one another (when projected onto the sample plane), while the fourth beam is oriented at 180° to one of the other three beams.^{[ 48 ]} b–d) Simulated interference patterns showing the effects of introducing a phase‐delay of 0°, 75°, or 90° into one of the four laser beams. The regions inside the green contours correspond to areas where the intensity is too low to induce cross‐linking of SU8, and hence holes are expected to form when the resist is developed. For zero phase‐delay, a hexagonal array of equally sized singlet holes is expected; while, for a 90° phase‐delay, a hexagonal array of equally sized doublet holes is expected. e–g) Simulated interference patterns showing—for a fixed phase‐delay of 90° in one of the laser beams—an increase in the size of the low intensity regions as the beam intensity is increased from low (e) to high (g). h–j) Scanning electron micrographs showing—for a fixed phase‐delay of 90° in one of the beams—the formation of doublet holes of increasing size and decreasing separation as the beam intensity is increased from low (h) to high (j). At the highest intensity an average hole separation of ≈20 nm is obtained. The SEM images are in broad agreement with the simulated results shown in (e–g). k,l) Low magnification scanning electron micrographs showing a typical shadow‐mask template (k) obtained by four‐beam interference lithography with a phase‐delay of 90° in one of the beams, and the corresponding gold nanoarray (g). Adapted with permission.^{[ 48 ]} Copyright 2011, American Chemical Society.

Figure 5

a) Illustration of nanogap fabrication using two closely spaced visible‐light laser beams and a thermally‐activated negative photo‐resist.^{[ 45 ]} The activation area due to each laser beam is smaller than the ≈200‐nm spot‐size since only resist that is located near to the center of the spot is heated sufficiently to undergo chemical conversion. By scanning the laser beams in a line and then immersing the sample in a developer to remove the unexposed parts of the resist, two parallel ridges of insoluble resist are generated. The width of the ridges—and hence the separation between them—is controlled by changing the laser intensity, with higher intensities yielding wider ridges with narrower separations. b) AFM image showing reduction in gap‐width in Ti/SiO₂ bilayer from 200 nm (far left) to 18 nm (far right) as the laser intensity is increased from 40 to 80 mW. c) Top‐view optical images of a metallic nanogap electrode array, obtained using the thermally activated negative photo‐resist. d) Micrograph and SEM images of one nanogap within the array in (c). Reproduced with permission.^{[ 45 ]} Copyright 2020, American Chemical Society.

Figure 6

a) Schematic of EBL using a positive resist, in which i) the selected parts of the resist are rendered soluble by exposure to the scanning e‐beam, ii) the exposed resist is removed using a developer, iii) the exposed parts of the target layer are etched away, and the resist is removed, leaving a patterned layer of the target material. b) Low‐magnification SEM image showing gold nano‐disc arrays with ≈5 nm gaps, obtained using EBL and the negative photoresist HSQ.^{[ 60 ]} c,d) Two and three‐way electrodes separated by <2 nm fabricated by electron beam milling, that is, by using a high beam‐current electron beam to directly ablate atoms from c) a platinum target and d) a silver target.^{[ 67} , ^{68 ]} b) Reproduced with permission.^{[ 60 ]} Copyright 2011, American Chemical Society. c) Reproduced with permission. Copyright 2007, American Chemical Society. d) Reproduced with permission.^{[ 68 ]} Copyright 2007, American Chemical Society.

Figure 7

a) Schematic of FIB milling technique, where a nanogap is formed by using a scanning ion beam to directly sputter atoms from the target material. b) SEM image of a gold dimer antenna obtained by Ga FIB milling. The two gold islands at the center are separated by a ≈12 nm gap.^{[ 73 ]} c,d) Top‐ and tilted‐view SEM images of bowtie shaped air‐gaps in gold with ≈4‐nm minimum separation, obtained by Ga FIB milling.^{[ 72 ]} e) Example of a bow‐tie shaped gold dimer with a gap‐width of ≈6 nm obtained using a combination of Ga and He ion FIB milling for coarse‐ and fine‐resolution patterning, respectively.^{[ 75 ]} b) Reproduced with permission.^{[ 73 ]} Copyright 2013, Nature Publishing Group. c,d) Reproduced with permission.^{[ 72 ]} Copyright 2015, American Chemical Society. e) Reproduced with permission.^{[ 75 ]} Copyright 2014, American Chemical Society.

Figure 8

a) Schematic of electromigration‐based break formation, in which a current is passed through a narrow strip of metal that has been lithographically patterned with a central notch. The current induces necking and eventual splitting of the notch due to migration of metal atoms away from the notch region. b) AFM image of gold electrodes obtained by electromigration with a gap width of 1–2 nm. c) Schematic of mechanically controlled breaking, in which a narrow strip of metal with a lithographically patterned central notch is bent on a flexible substrate until the metal splits at the pinch‐point of the notch. By further bending or relaxing the substrate, the width of the gap can be dynamically adjusted. d) Gold source (S) and drain (D) electrodes fabricated by MCB method, with a side‐electrode (G) for gating the current between the other two electrodes. e) Schematic of strain‐induced cracking, in which a rigid substrate is successively coated with a sacrificial layer such as amorphous silicon, followed by a metallized, brittle target material. The target layer is patterned into an array of etched bridges with notches at their mid‐points, and the bridges are then released from the substrate by undercut chemical etching of the sacrificial layer. Removal of the sacrificial layer allows the released target material to contract, and in so doing it splits at the pinch‐point of the notch where the stress is highest. f) SEM image of a sub‐10 nm silver nanogap electrode obtained by stress‐induced cracking. b) Reproduced with permission.^{[ 86 ]} Copyright 2002, Nature Publishing Group. d) Reproduced with permission.^{[ 96 ]} Copyright 2013, f) reproduced with permission.^{[ 106 ]} Copyright 2018, The Royal Society of Chemistry.

Figure 9

a) Schematic of atomic layer lithography (ALL) process, in which i) a first metal M1 is deposited on a substrate, ii) a conformal layer of Al₂O₃ is deposited over the metal and the substrate using atomic layer deposition, iii) a second metal M2 that adheres weakly to Al₂O₃ is deposited on top of the coated substrate, iv) a rigid adhesive material is applied to the upper surface of the stack, making contact with the (uppermost) parts of M2 that lie on top of M1, v) the adhesive is peeled away from the stack, taking with it the unwanted parts of M2, and vi) the oxide layer is removed by reactive ion etching, leaving M1 and M2 side‐by‐side on the substrate with a separation equal to the width of the Al₂O₃ layer. b) Top‐view SEM image of a 5‐nm‐wide annular gap in a 200‐nm‐thick silver film obtained by ALL. c) Photograph showing a wafer‐scale array of silver nanogap features in a Si wafer. Adhesive tape has been peeled away from the right side of the Si wafer (see inset). d) Optical micrograph of a sample containing approximately 150 000 silver nanogap rectangles (gap size = 5 nm, total ring length = 0.7 mm). b–d) reproduced with permission.^{[ 116 ]} Copyright 2013, Nature Publishing Group.

Figure 10

a,b) Procedure for generating smooth metal nanogap arrays by atomic layer lithography. The standard ALL method is followed up to and including the deposition of M2. Then, by choosing an adhesive that conformally coats M2 and binds strongly to it, the complete metal nanogap array is removed from the substrate in the process of peeling away the adhesive, leaving behind only those parts of the Al₂O₃ that are directly bound to the substrate. Inverting the adhesive yields a smooth upward‐facing metallic nanogap array, in which the heights of M1 and M2 differ only by the few nanometer thickness of the Al₂O_3. c) SEM image of line‐arrays obtained by the modified ALL procedure, with spacings of approximately 5 nm between the Al lines (dark) and the gold lines (light). d,e) Alternative procedure for generating smooth metal nanogap arrays by atomic layer lithography, in which the standard ALL method is followed up to and including the deposition of M2 (d). By subjecting M2 to glancing‐angle ion polishing until the upper surface of M1 is exposed, an ultrasmooth surface is obtained in which the upper surfaces of M1 and M2 are level (e). f) SEM image of resulting nanoring arrays using cylindrically patterned gold for M1 and uniformly deposited gold for M2. The width of the nanorings is approximately 10 nm. a–c) Reproduced with permission.^{[ 21 ]} Copyright 2015, American Chemical Society. d–f) Reproduced with permission.^{[ 120 ]} Copyright 2016, American Chemical Society.

Figure 11

a–f) Schematic showing key processing steps in conventional and self‐peeling adhesion lithography. The conventional procedure comprises the following steps: first, metal M1 is deposited on a substrate and patterned as appropriate (a); second, M1 is selectively coated with a metallophilic SAM (b); third, metal M2 is deposited uniformly over M1 and the exposed substrate (c); fourth, an adhesive film is applied to the surface of M2 (d); fifth, the film is peeled away from the substrate, selectively removing M2 from those regions located directly above the SAM (e‐i); finally, the SAM is removed by UV/ozone or O₂‐plasma treatment, leaving M1 and M2 sitting in a complementary arrangement side‐by‐side on the substrate (f), separated in the limiting case by the length of the SAM. The self‐peeling procedure follows the conventional method up to step (d), except the peeling layer comprises a polymer with a high coefficient of thermal expansion α and a high Young's modulus Y, spin‐coated onto M2 from a heated solution. As the polymer film cools, tension builds inside the film until it is sufficient to induce spontaneous peeling of the polymer from the coated substrate, taking with it those parts of M2 that are located directly above the SAM (e‐ii). The SAM is removed as before by UV/ozone or O₂‐plasma treatment. SEM images of Au‐Al nanogaps obtained with g) matched and h) unmatched metal heights, using a single layer of ODT as the SAM, 50‐nm gold for M1 and 50‐nm Al or 30‐nm Al for M2. i) High resolution SEM image of a sub‐3 nm Al/Au nanogap. j) SEM image showing an example of large‐area patterning using adhesion lithography. The separation between the light regions (Au) and dark regions (Al) is approximately 5 nm. k) Photograph showing formation of Al/Au nanogap arrays by self‐peeling adhesion lithography. The photograph was taken midway through the peeling step, with the peeling layer partially detached from the metal‐coated substrate. (a‐f,k) Adapted under the terms of the Creative Commons CC‐BY license.^{[ 130 ]} Copyright 2019, The Authors. Published by Wiley‐VCH. g‐j) Adapted under the terms of the Creative Commons CC‐BY license.^{[ 131 ]} Copyright 2021, The Authors. Published by Wiley‐VCH.

Figure 12

Fabrication of massively parallel nanoring arrays using a combination of NSL and size‐tunable a‐lith.^{[ 131 ]} a–d) Schematic of the fabrication procedure, in which: first, a monolayer of close‐packed polystyrene nanospheres is deposited on a substrate (a); second, the nanospheres are “shrunk” by oxygen plasma treatment, leaving voids between them (b); third, metal M1 = Au is deposited on the substrate through the nanosphere template and the template is removed, leaving a hexagonal array of nanoholes in the gold film (c); and, fourth, the holes are “filled” with a second metal (M2 = Au) using size‐tuneable adhesion lithography, resulting in a hexagonal array of ring‐shaped nanogaps (d). e) 20‐µm × 40‐µm SEM image of an Au–Au nanoring array, obtained using a molecular ruler of length N = 1. f–h) High magnification SEM images of Au‐Au nanoring arrays, obtained using molecular rulers of length N = 1 (f), N = 2 (g), and N = 5 (h). Each array has a pitch of ≈500 nm and a ring‐diameter of ≈380 nm, defined by the nanosphere diameters before and after etching. Reproduced under the terms of the Creative Commons CC‐BY license. Copyright 2021, The Authors. Published by Wiley‐VCH.

Figure 13

a) Schematic of SPLprocess,^{[ 141 ]} in which i) a layer of HSQ negative e‐beam resist is deposited on a substrate, ii) a circular pattern is defined in the resist by EBL, iii) a thin gold layer is evaporated onto the resist‐coated substrate, iv) an adhesive polymer is drop‐cast onto the substrate and cured under UV irradiation, v) the adhesive is peeled from the substrate, leaving gold trapped inside the ring of resist, and vi) the resist is removed by a developer, leaving a disk of gold. b) SEM images of plasmonic gold structures obtained by using EBL to define “figures of eight” (i.e., pairs of touching gold rings) in the resist at step (a‐ii). c) SEM images of plasmonic gold structures obtained by using EBL to define clusters of seven touching rings in the resist at step (a‐ii). The gaps between the gold discs are approximately 15 nm. Reproduced with permission.^{[ 141 ]} Copyright 2016, American Chemical Society.

Figure 14

a) Schematic showing SPL by FIB milling,^{[ 142 ]} in which i,ii) a narrow ring‐shaped trench is milled directly into a metal film, iii) a layer of adhesive is applied to the metal film, and iv) the adhesive is peeled from the metal film, taking with it all metal outside of the trench. b,c) SEM images of gold dimers formed by using helium FIB‐milling to define “figures of eight” (i.e., pairs of touching gold rings) in a gold film at step (i). The images show the situation b) before and c) after peeling. The separation of the discs is approximately 15 nm. Reproduced with permission.^{[ 142 ]} Copyright 2016, American Chemical Society.

Figure 15

a) Schematic of thermal nanoimprint lithography (NIL) process, in which i,ii) a rigid mold is pressed into a resist to create a thickness contrast, iii) the mold is mechanically separated from the thermally‐cured resist, and iv) the compressed resist is anisotropically etched away, leaving resist only in the locations where imprinting did not take place. b) Hole array imprinted in poly(methyl methacrylate) (PMMA) with ≈10 nm diameter holes.^{[ 151 ]} c) SEM image of Si mold used to generate the pattern shown in (b), comprising a regular 2D‐array of columns of approximate width 10 nm. b,c) Reproduced with permission.^{[ 151 ]} Copyright 1997, American Institute of Physics.

Figure 16

a) Schematic of photocurable nanoimprint lithography, in which i) a transparent silica mold is pressed into a bilayer of poly(methyl methacrylate) (PMMA, transfer polymer) and a negative resist, ii) the resist is exposed to UV light, causing it to harden, iii) the mold is removed, iv) the resist is subjected to reactive ion etching until the PMMA layer is exposed, v) the exposed PMMA is etched away by treatment with an oxygen plasma, and vi) the patterned bilayer of PMMA and resist is used as a shadow mask for metal deposition. The bilayer may alternatively be used as an etch mask. b) SEM images of silicon oxide molds used for photocurable NIL. c) SEM images of contact shadow masks, obtained using the molds in (b). d) SEM images of gold nanogaps, obtained using the contact masks in (c). b–d) Reproduced with permission.^{[ 150 ]} Copyright 2004, American Institute of Physics.