2.1.1.

Step 1: Photoresist and substrates

The low line edge roughness of the metal frame after lift-off is a prerequisite for achieving a consistent gap width along hundreds of a micrometer length. When the sidewalls of the photoresist are undercut shape, lift-off is easier without remaining residues and prevents the resist sidewalls from being coated, which eventually make the line edges of the metallic patterns clearer. Therefore, photoresist of the right type and height should be chosen for the very specific gap shape one desires. Types of photoresist determine the metal profile after lift-off processes, and there is an optimal photoresist height for the chosen thickness of the metal. For this reason, we used commercial photoresists, AZ5214E and AZ5206E, an image reversal type. In general, an image reversal type ends up in an inverted trapezoidal shape much in the same fashion as a negative type, which shows straight line metal edges with graded sidewall profiles.

After selecting the photoresist type, we define photolithography conditions for patterned metal films with desired edge angles. The conditions of the photoresist height of AZ5214E and AZ5206E were targeted to 1.2 and $0.6\mu \mathrm{m}$, respectively. The recipes of both image reversal type photoresists are revealed in this section depending on silicon, sapphire, quartz, and gold film. When each condition of this process is optimized, the deposited metal cliffs can have a clearer edge with a specific sidewall angle.

For target metal thickness of 300 to 400 nm, AZ5214E photoresist is well suited by following a rule of thumb in photolithography; the resist thickness should be at least three times greater than the metallic film thickness. The photoresist patterns are made on bare substrates (sapphire, quartz, and silicon) and on metallic films. The sample is cleaned with acetone, isopropanol, and deionized water and spin-coated at 4000 rpm for 60 s, followed by soft-baking at 90°C for 90 s. For sapphire or quartz substrates, Hexamethyldisilazane is spin-coated at 4000 rpm for 60 s before spin-coating the photoresist in order to promote adhesion between the photoresist and the substrate. We used 365-nm i-line mercury arc lamp in a mask aligner (Karl Süss MJB-3, Süss MicroTec) for mask exposure, where the sample is illuminated through Cr mask with an exposure dose of $140{\mathrm{mJ}/\mathrm{cm}}^2$. Then, the sample was postbaked at 115°C for 85 s and exposed to UV lamp again without the mask for with exposure dose of $300{\mathrm{mJ}/\mathrm{cm}}^2$. Finally, we developed the sample in MIF (Metal Ion Free) 300 developer (AZ Electronic Materials) for 30 s. The development time depends on pattern size: $10\mu \mathrm{m}(l_x)\times 10\mu \mathrm{m}(l_y)$ takes 90 s, $50\mu \mathrm{m}\times 50\mu \mathrm{m}$ takes 30 s, and $3\mathrm{mm}\times 5\mathrm{mm}$ takes 20 s. AZ5214E is used for methods 2, 3, and 5 (see Table 1). In the case of method 5, although the maximum thickness of the first metallic film is lower than the case of methods 1 and 4, a 185-nm-thick sacrificial layer is added on top of the metallic film, which is suitable to use AZ5214E.

Table 1

Fabrication methods.

Meanwhile, AZ5206E is suitable for smaller metal thickness of 150 to 200 nm used for methods 1 and 4. The recipe is similar but the reversal bake time, and the first and the second exposure dose are different. The UV exposure dose with a Cr mask is $120{\mathrm{mJ}/\mathrm{cm}}^2$, the reversal bake time is 120 s at 115°C, and the second UV exposure dose without the mask is $800{\mathrm{mJ}/\mathrm{cm}}^2$; other conditions are the same as used in photoresist AZ5214E. The samples are developed in MIF 300 for 3 s and then gently soaked in deionized (DI) water without shaking or sonication. This is repeated once more until the patterns appear and it is distinguishable by looking through the optical microscope. If the sample is developed for 6 to 7 s without this interruption, then the photoresist might vanish before the patterns emerge.

During the second UV exposure without a mask, the entire resist film is exposed to make the resist areas—up to now, the unexposed areas—developable. The exposure dose should be high enough to fully transmit through the bottom of the resist transferring a positive type to image reversal type. For this reason, the second exposure dose is longer for the thick resist case. Moreover, the thickness of the resist film also affects the develop time. A thick resist film (5214E) is developed longer than the thinner one, AZ5206E. The second exposure dose is $300{\mathrm{mJ}/\mathrm{cm}}^2$ for the AZ5214E case while the case for AZ5206E is $800{\mathrm{mJ}/\mathrm{cm}}^2$ because the develop time is long enough to erase the unwanted photoresist with a reversal bake time of 85 s. However, the thinner photoresist develop requires only 7 s in MIF 300. The thinner the resist thickness is, the longer the baking time is, to maintain the cross-sectional sidewall. Therefore, increasing the reversal bake time from 85 to 120 s and the second UV expose dose from 300 to $800{\mathrm{mJ}/\mathrm{cm}}^2$ are necessary for a short-time development.

2.4.1.

Sample characterization

The nanogap fabrication techniques using the chemical etching process is already well-known by the previous studies, and the uniformity of the gap in a millimeter-scale square area has been confirmed by scanning electron microscope (SEM) images and transmitted optical microscope images at low magnification.^12,13 To confirm the uniformity of the nanogap sample after the tape-peeling process, we observed a sample with field emission scanning electron microscopy at low magnification in Figs. 3(a) and 3(b). The in-plane dimension of $10\text{-}\times 10\text{-}\mu \mathrm{m}$-squared loop alumina-line array is used for the SEM images to quantitatively confirm the uniformity around the nanogap after tape peeling. The cross-sectional images in Figs. 3(c) and 3(d) show that the nanogap is formed with the thickness of 230 nm and the width of 5 nm with an average sidewall angle as 76 deg. The yield of the sample after tape peeling is as high as the chemical etching process, as long as the sidewall angle of the patterned metallic film is higher than 70 deg. These images show that the nanogap structures are well covered in same structural dimensions over a millimeter-scale square area.

Fig. 3

(a) A top-view SEM image of nanogap sample after tape peeling at low magnification. (b) An enlarged image of (a). The inset shows the enlarged version of (b), showing a few $10\text{-}\times 10\text{-}\mu \mathrm{m}$-squared loop of alumina line. (c) A tilted SEM image of a left side of the square loop. The Pt deposition layer is used for protecting the surface damage while etching the groove by focused ion beam. (d) A cross-sectional image of an enlarged image of (c), focused on the nanogap.

In addition, the topographic images of a sample after tape peeling and a sample after chemical etching are observed by atomic force microscopy (AFM) in Figs. 4(a) and 4(b). In both cases, the first metallic frame thickness is 15 nm higher than the thickness of the secondary metallic film to increase the yield of the nanogap structure. The two samples can be distinguished by different patterns, for instance, $10\text{-}\mu \mathrm{m}\times 10\text{-}\mu \mathrm{m}$-squared patterns for the tape-peeling case and $10\text{-}\mu \mathrm{m}\times 40\text{-}\mu \mathrm{m}$-patterns for the chemical etching case. The line profiles of Figs. 4(c) and 4(d) are obtained from the red-colored area in the corresponding topographic images of Figs. 4(a) and 4(b), respectively. The tape-peeled sample shows some defects the size of 8 to 15 nm near the nanogap line as in Fig. 4(c). The tilted view of the SEM image in Fig. 3(c) ensures that the defects occurred when splitting the connections of the secondary metallic layer at the gap line, separating it into the parts that are on the patterned frame and the portions that are plugged into the holes of the patterned frame. However, after etching, a part of the reacted material remaining as particles is observed at some of the gap lines, the contours of the squares [Fig. 4(b)]. The remaining can be removed by cleaning the sample again because the secondary metallic film contains an adhesion layer [see Fig. 9(b)]. The average surface roughness of the peeled surface is found to be 0.625 nm after chemical etching and 1.089 nm after tape peeling. These results show that the surface uniformity of the nanogap structure made by chemical etching is superior on the sample made by tape peeling, but both exfoliation processes are reliable because of the high yield.

Fig. 4

AFM topographic images of nanogap sample (a) after tape peeling and (b) after chemical etching. The line profiles (c) after tape peeling obtained from the red-colored area of (a) and (d) after chemical etching obtained from the red-colored area of (b).

3.1.1.

Method 1: Tilted MIM nanogaps using AZ5206E

The cross-sectional schematics of the method 1 procedure are shown in Fig. 5(a). The fabricating conditions and parameters applied at each step are listed in Tables 1 and 2, respectively. Specifically, Cr 3 nm and Au 150 nm are deposited with an e-gun evaporator (Vacuum Tech., Korea) for the first metal frame. After the lift-off, alumina layer of 5 nm is coated over the metal frame with ALD. The condition of using ALD is mentioned in Sec. 2.2. Third, the second metallic films of Au 150 nm without adhesion layer are deposited by e-gun evaporator again. Finally, an adhesive tape is used for the exfoliation process. The average film thickness of the first and second metallic films and the gap thickness of both metallic films are all 150 nm, showing a flat surface of the MIM nanogaps.

Fig. 5

(a) Cross-sectional schematics of the procedure of method 1, (b) a top view of SEM images of nanogap sample with large areas, and (c) an enlarged image of (b), focused on an alumina line of the rectangular-shaped loop in between metallic films. (d) A cross-sectional SEM image of (c).

Top views of the SEM images are shown in Figs. 5(b) and 5(c). The dark-colored rectangular-shaped arrays (metallic film) are spaced in between the bright gray-colored surface (metallic film covered with alumina) over an area of $200\mu \mathrm{m}\times 150\mu \mathrm{m}$. Nanogaps that are in between the different colored sections in Fig. 5(c) can be seen by a cross-sectional SEM image [Fig. 5(d)]. The gap thickness is 150 nm, the same as the metallic film thickness. The sidewall angle of the first metallic film is 70 deg, and the schematics in Fig. 5(a) show that the reason is due to the photolithography in step 1, and that the tape-peeling process in step 4 does not matter.

3.1.2.

Method 2: An asymmetric V-groove structure using AZ5214E

In this method, the first metallic frame formation are done by photolithography with photoresist AZ5214E with Cr 3 nm and Au 320 nm as in Fig. 6(a) (see Table 2). After lift-off and an alumina layer coating with ALD, a secondary metallic film is deposited at 320 nm [Fig. 6(b)] and 220 nm [Fig. 6(c)] at different samples. An adhesive tape is used for the exfoliation process. An asymmetric V-groove shape appears due to the different height of the first and second metallic films and the direction of the peel depending on the edge shape of the first metallic frame.

Fig. 6

(a) Cross-sectional schematics of fabricating the asymmetric V-groove type nanogap structure followed in method 2. (b) Cross-sectional SEM images of (i) before and (ii) after peeling off the unnecessary metal films that are atop of first metallic pattern. (c) Cross-sectional SEM images of (i) the left side and (ii) the right side of a single rectangular-ring shape after peeling off the unnecessary metallic films from right-to-left direction.

As shown in Fig. 6(b), the formation of an asymmetric V-groove gap shape is observed by the cross-sectional SEM images, which are taken before and after applying an adhesion tape. The first metallic frame forms a pointed tail at the bottom edges due to an undercut-shaped photoresist while depositing the Au film [see Fig. 6(b)(ii) cross-sectional view from the lift-off result]. On top of the point tail of the first Au metallic film, a cross-link composed of metallic grains is formed, unlike in smoother secondary metallic film away from the gap region [Fig. 6(b)(i)]. These cross-linked metallic grains and the unnecessary films that are on top of the first metallic frame are removed by an adhesion tape, forming a V-groove shape of the gap [Fig. 6(b)(ii)].

When the difference of the first metallic frame and the secondary film thickness is increased to 100 nm, a tilted MIM structure and an asymmetric V-groove both appear depending on the direction of peel during the exfoliation process. The V-groove shows the gap thickness of about 35 nm, and the sidewall angle at the point tail is 60 deg while the sidewall angle of the secondary metallic film is 80 deg. As shown in Fig. 6(c)(i), the nanogap with the gap thickness of 320 nm is the same as the thickness of the first metallic frame. The cross-sectional view of SEM images that are from (i) the left and (ii) the right side of the rectangular $3\mathrm{mm}\times 5\mathrm{mm}$ is shown in Fig. 6(c) after peeling the adhesive tape in the direction of right-to-left. A cross-link part from both SEM images in Fig. 6(c), the secondary metallic regions on top of the point tail at the first metallic film bottom edges, shows two weak-bonding points: the top [see Fig. 6(c)(i)] and bottom edge of the first metallic film [see Fig. 6(c)(ii)]. These weak-bonding points correlate with the direction of peeling and the cross-sectional slope and size of the point tail apart from the first metallic film. For instance, the adhesion tape-peeling direction comes toward the sidewall of the first metallic frame in Fig. 6(c)(i) while the peeling direction is opposite in Fig. 6(c)(ii). Moreover, the slope of the point tail is steeper than in Fig. 6(c)(ii), and the size of the point tail is larger.

3.1.3.

Method 3: An asymmetric Y-groove structure using ion beam etching

In method 3, an ion beam etching is added to make the first metallic frame. The argon (Ar) ion beam trims the edge of the metallic frame, including the point tail, and ends up with a thickness of 320 nm. After the ALD process, the secondary metallic film is deposited 100 nm lower than the thickness of the first metallic film. The exfoliation process is carried out by applying an adhesive tape as before. Figure 7(a) describes the fabrication procedures in a cross-sectional view.

Fig. 7

(a) Cross-sectional schematics to show the procedure of nanometer gap fabrication, including ion beam etching process in method 3. (b) A schematic of a typical design of the sample mount and an ion beam source inside the Ar ion miller chamber; Ar ion beam size covers 6 of 4-in. wafers. The sample mount is connected to the axis of rotation, and it can tilt the mount angle. (c) Tilted-view and (d) cross-sectional SEM images of (i) before and (ii) after trimming the edges of the first metallic film using ion beam etching process. (e) Top-view and (f) cross-sectional SEM images that are taken after the exfoliation process followed in (i) method 2 and (ii) method 3.

The etching process is carried out by an Ar ion beam miller (Hitachi Tech, Ltd., Japan). The accelerating voltage is fixed at 300 V, and the Ar gas flow rate is set at 20 sccm under the pressure of 5 micron Torr. The sample is attached to the sample wafer of the mount and the Ar beam size covers 6 of 4-in. wafers as shown in Fig. 7(b). The sample mount is rotated in a clockwise direction and the target angle can be tuned by tilting the central axis of the mount. When the tilting angle of the mount axis is set at 15 deg for 3 min and additionally set at 75 deg for 4 min, the etch depth of the Au film is about 37 nm (see Table 3). This etching process is additionally repeated three more times to remove the point-tail part of the first metallic film. The two different tilting angles are set to avoid the shadow effect and an unexpected deposition from etched metallic clusters.

Table 3

Ion beam etching rate.

The tilted and cross-sectional SEM images of (i) before and (ii) after ion beam etching are shown in Figs. 7(c) and 7(d), respectively. The point-tail part of the first metallic film is removed after etching as in Figs. 7(c). In Fig. 7(d), 250 nm of the first metallic film decreased to 100 nm, and the point-tail part is clearly removed showing the sidewall angle as 70 deg with rounder edges than in methods 1 and 2. Furthermore, the oft-existing undesired remaining photoresist layers are removed by ion beam etching as can be seen by the comparison of the top and cross-sectional view of SEM images in Figs. 7(e) and 7(f). By removing the point-tail part of the first metallic film and the undesirable remaining photoresists, a nanogap can be seen above the substrate in between the Au metallic films in the cross-sectional view of the SEM image of Fig. 7(f)(ii). The gap thickness of this MIM structure, 150 nm, is much larger than the V-groove case of Fig. 7(f)(i), where the nanogaps exist only at the bottom near the substrate.

3.1.4.

Method 4: MIM nanogaps using a photoresist mask

The photolithography and ion beam etching are carried out on a metallic film to fabricate the first metallic frame using the photoresist AZ5206E as a sacrificial layer [Fig. 8(a)]. Etching conditions are followed as in method 3. After stripping the photoresist by acetone, ALD process, secondary film deposition and exfoliation process by tape are carried out as in method 1, 2, and 3. Unlike methods 1, 2, and 3, the thicknesses of the first and the secondary metallic films are both 120 nm with the sidewall angle of 73 deg.

Fig. 8

(a) Cross-sectional schematics to show the procedure of method 4; AZ5206E photoresist is a shadow mask of an Ar ion beam etching. (b) Cross-sectional SEM images of (i) before and (ii) after milling Au layer through photoresist shadow mask. The Ar ion beam etches 100-nm-thick Au layer, while it etches AZ5206E photoresist about 60 nm. (c) A cross-sectional and (d) a top-view SEM image after the exfoliation process in method 4.

Fig. 9

(a) Cross-sectional schematics of the procedure of method 5, chemical etching instead of adhesive tapes in the exfoliation process. A patterned Cr–Cu–Cr layer is a shadow mask while using ion beam miller. (b) Top view of SEM images; a dark gray shows the remaining Cr layer on top of the first metallic frame (left); an enlarged part of the left SEM image is shown after an extra wet-etching process and ultrasonic agitation (right). (c) A cross-sectional SEM image of the sample part of (b). (d) A normalized transmitted amplitude and the FE of the sample.

Photolithography and lift-off processes (methods 1, 2, and 3) have thickness limitations because of the relation between the photoresist and an appropriate deposition thickness (see Sec. 2.1.1). For instance, AZ5206E (or AZ5214E) is used for 150 to 200 nm (or 300 to 400 nm) of metallic thickness due to a good rule of thumb.

However, the thickness limitation can be solved using the photoresist as a shadow mask, also as a sacrificial mask, for patterning the first metallic frame. As shown in Fig. 8(b), the cross-sectional view of SEM images of (i) before and (ii) after ion beam etching a metallic film through a photoresist mask shows the possibility of fabricating a thin metallic frame. The Ar ion beam etches the photoresist about 60 nm while etching 100-nm-thick Au film. We suspect that the Pt deposition, a protection layer while slicing the sample for a cross-sectional view, presses down the photoresist resulting in a trapezoidal shape.

Furthermore, the thickness of an edge part of the photoresist after ion beam etching is decreased by an ion beam etching and shows the maximum thickness of the metallic film with photoresist AZ5206E around 100 nm [Fig. 8(b)(ii)]. A shadow mask AZ5206E that has a smaller undercut-depth than AZ5214E is suitable for avoiding the undesired secondary deposition on the shadow region of the photoresist.

A sub-10 nm of alumina layer is sandwiched between 120-nm Au films with the sidewall angle of 73 deg as shown in a cross-sectional and a top view of the SEM image in Figs. 8(c) and 8(d), respectively. As the top edge of the first metallic frame is not rounded as in method 3, it is possible to deposit the secondary Au film with the same height as the first metallic film as in method 1 without making cross-links between the unnecessary Au films that are on top of the first metallic frame and the secondary Au film, which is plugged in the rectangular holes of the first metallic frame.

Unlike the previous methods mentioned above, the thickness of a nanogap MIM film is a thinner version of method 1. A thin and flat nanogap metallic structure has shown advantages over the recent studies in that it can enhance the electric field stronger than a thick and/or a tapered-edge type of nanogap structure.²⁹

3.1.5.

Method 5: An asymmetric Y-groove structure using metallic shadow mask

In method 5, an ion beam etching process is performed through a metallic shadow mask, and the sacrificial layer is exfoliated by a chemical etchant [Fig. 9(a)]. A 100-nm-thick metallic film with Cr 3 nm of adhesion layer is prepared. Cr 5 nm–Cu 160 nm–Cr 20 nm metallic film is patterned by photolithography and lift-off. This metallic film acts as a shadow mask for an Ar ion beam to pattern the first metallic frame. After the ALD process, the secondary 100-nm-thick metallic film with an adhesion layer Cr 3 nm is deposited. Finally, the exfoliation process is performed by a chemical etchant for removing the shadow mask, which is also a sacrificial layer.

The explanation of the first and the second layer of the shadow mask and the chemical etching process is given in Sec. 2.4. The top layer of the shadow mask, 20-nm Cr layer, is coated to prevent the edge of the mask film being rounded from ion beam etching. The Cr dry etching rate is $1\mathrm{nm}/\min$, which is much slower than Au ($9\mathrm{nm}/\min$) and Cu ($5\mathrm{nm}/\min$), so the Cr layer of the top of the shadow mask can endure while milling a thin Au film. This work is to form the sidewall angle of the first metallic frame as 80 deg [See Fig. 9(b)]. The undesired particles after wet etching are washed out by sonication in acetone.

A top view of a $50\text{-}\mu \mathrm{m}\times 50\text{-}\mu \mathrm{m}$ sized square-ring array with 100-nm-thick MIM structure of the SEM image is shown in Fig. 9(b). The dark gray-colored region shows the remaining Cr layer on top of the first metallic frame that can be distinguished from the bright gray-colored square regions (secondary-deposited Au films). After an extra wet-etching and cleaning process, the enlarged SEM image of a part of the side of the ring array shows that a dark line (alumina, nanogap) passes through the bright gray (Au film) surface. This explains that the Cr layer on top of the first frame is removed.

In Fig. 9(c), a cross-sectional view of an SEM image shows the nanogap MIM structure that has minimum 5-nm gap width with the gap thickness of 70 nm. This structure has an adhesion layer, a dark gray line sandwiched between the bright gray-colored area (the substrate) and the white-colored area (the Au films). Underneath the right side of the Au film, the gray line is thicker than the left side. This results from the two layers: an alumina layer and a Cr layer. The alumina layer is coated over the first metallic frame, including the substrate, and subsequently, the Cr layer, the part of the secondary film deposition process, is deposited.

To confirm that the sample is an optically active nanogap, the FE spectra of the sample are shown in Fig. 9(d), by estimating from normalized THz transmitted spectra. The transmitted spectrum from the sample is normalized by the spectrum of the bare substrate. Using the Kirchhoff integral formalism,⁴³ the enhanced field in near-field, FE, can be calculated as dividing the effective gap coverage ratio. Since the incident THz beam is polarized, the sides that are perpendicular to the incident beam strongly enhance the field inside the nanogap. Therefore, the effective gap coverage ratio can be defined as $A_{\text{gap}}/A_{\text{metal}}$, where $A_{\text{gap}}$ indicates the area of the parallel sides of the square-ring arrays that are perpendicular to the incident beam and $A_{\text{metal}}$ implies the area except the ${\mathrm{A}}_{\text{gap}}$. As a result, up to 2700-fold FE at around 0.4-THz resonant frequency is obtained. The experimental data made from other methods are reported in Refs. 29 and 45.

The samples made by the etching method can be washed again unlike those made by an adhesive tape, because the secondary metallic film is deposited after the adhesion layer. Moreover, the exposed alumina layer on top of the first frame does not exist in this method due to the sacrificial layer. This results in a top view of nanogap ring arrays in metallic films nonexisting alumina layers and a cross-sectional view of metal/alumina/metal structure, which is the candidate for sensing biomolecules and SERS.