Copper-based intersectional nanofabrication of optical nanoantennas for volatile organic compound sensing

Volatile organic compounds (VOCs) are chemicals that can vaporize easily at atmospheric pressure and room temperature. The presence of VOCs is associated with many industrial and human activities. Therefore, VOCs can act as indicators of the occurrence of such activities, such as benzene and some alcohol, that can be used to monitor human health and the natural environment as well.^1,2 The determination of VOCs’ concentrations is thus of great importance in various fields. However, most VOCs are typically non-reactive gases making them hard to detect.³ Conventionally, mass spectrometers in combination with liquid chromatography or gas chromatography are used for VOCs sensing with a wide detection range and low detection limit, but the instruments are expensive and bulky, making them unsuitable for on-site detection.^4,5 A variety of VOC sensors, including electrical and optical ones, have been proposed to be able to trace amounts of VOCs. Although electrical VOC sensors have developed rapidly, they generally require high operating temperatures making them unsuitable for the demands of ultracompact sensors used in the future Internet of Things (IoT). By contrast, optical VOC sensors can work at room temperature with a small footprint. In particular, sensors with plasmonic nanoantennas have shown great potential for ultrasensitive detection of VOCs thanks to the remarkable field enhancement brought by those nanoantennas.^6–9 

Metallic nanoantennas can support localized surface plasmon resonances (LSPRs) that can concentrate incoming light into nanoscale volumes, termed “hot spots,” where significantly enhanced near-field intensity occurs.¹⁰ The resonance wavelengths of these plasmonic nanoantennas are sensitive to the nanoantenna shape, size, and surroundings providing various feasible methods to tune the resonances for different applications.^11–15 In particular, plasmonic sensors have attracted a lot of attention for their considerably enhanced sensing performance. When target molecules are immobilized onto the nanoantenna surface, especially those situated inside the “hot spots,” the plasmon resonances shift to longer wavelengths due to the increase in the environmental refractive index. Usually, a linear relationship can be established between the wavelength shift and the number (or concentration) of the molecules, which represents the sensing principle for most plasmonic sensors that essentially detect the changes in the effective refractive index of the surroundings of the nanoantennas.^16–18 

From the viewpoint of practical applications, it is highly desirable to be able to fabricate these plasmonic sensors in a cost-effective fashion.¹⁹ Conventional lithography tools, such as e-beam lithography (EBL) and focused ion beam (FIB), which are both expensive and time-consuming, cannot meet this challenge. Colloidal lithography (CL)^20,21 is known as a high-throughput nanofabrication technique that can produce isolated nanoantennas (nanotriangles,^22–24 nanodisks,²⁵ and nanocrescents²⁶), ordered arrays,²⁷ and even sophisticated nanostructures.^28,29 Recently, we have demonstrated a new extension to conventional CL, colloidal lithography assisted intersectional nanopatterning (CLAIN), that can fabricate disk-shaped metallic and all-dielectric nanoantennas without using dry etching, eliminating any potential environmental hazards of the etching gases and the residues.³⁰ We utilized intersectional metal deposition and the poor adhesion between gold and oxide to create disk-shaped nanoantennas with tapes.³¹ While CLAIN removes expensive dry etching, the use of gold as the sacrificial layer still represents a considerable cost.

In this work, we demonstrate an upgraded version of CLAIN that uses copper (Cu) as the sacrificial material. The use of Cu significantly reduces the overall cost, as the price of Cu is ∼8000 times less than that of Au. Optical nanoantennas can be readily fabricated with different materials and on different substrates. An additional and convenient advantage of using Cu is that simple thermal heating can effectively shrink the Cu nanoholes providing a facile means to tailor the sizes of the nanodisks. We then design plasmonic VOC sensors by combining the fabricated Au nanodisks with zeolitic imidazolate framework-8 (ZIF-8) materials and demonstrate highly-sensitive detection of ethanol vapor. We believe that the proposed new method enriches the nanofabrication technique CLAIN and opens avenues for the cost-effective mass-production of plasmonic VOC sensors.

Figure 1 illustrates the process of the Cu-based CLAIN method, which involves only colloidal lithography and metal deposition. First, polystyrene (PS) nanospheres are spin-coated onto an ITO substrate forming a non-close-packed sub-monolayer. Next, a Cu film is deposited onto the sample via electron-beam evaporation. Subsequently, the Cu-coated nanospheres are gently scraped away using a piece of polydimethylsiloxane (PDMS) slab immersed in isopropyl alcohol, leaving nanoholes in the Cu thin film.³² Then, the second deposition (Al) is carried out, forming a discontinuous metal layer over the top surface as well as onto the substrate inside the nanoholes. Finally, the sample is immersed in a Cu etching solution to selectively dissolve the Cu layer and simultaneously remove the Al layer leaving Al nanodisks on the substrate.

The material in the second deposition decides the composition of the nanodisks, as illustrated in Fig. 1 with Al nanodisks. For other common plasmonic materials such as Ag and Au, an adhesion layer of chromium (Cr) or titanium (Ti) should be included in the second deposition. It is clear that Cu-based CLAIN shares the same advantage as Au-based CLAIN in that the whole nanofabrication process does not involve dry etching, which is usually required to fabricate disk-shaped nanoantennas with conventional CL.²⁸ Moreover, Cu-based CLAIN shows the same high-throughput fabrication capability but at a considerably reduced cost compared with Au-based CLAIN. It is noted that chemical wet etching is used in Cu-based CLAIN, as we notice that scotch tapes are not as efficient to remove the Cu film as they are to remove the Au film, which can be attributed to the much higher free energy of oxide formation associated with gold.^31,33 Therefore, chemical etching, which is more aggressive than physical peeling, is used. The detailed procedure for the fabrication can be found in the section “Methods.” As we show below, our proposed nanofabrication method shows high versatility in fabricating nanodisks of different sizes, morphologies, and composition materials.

For Cu-based CLAIN, selective etching of Cu with regard to other materials is key to achieving different metallic nanoantennas. By choosing the right etching solutions, different nanodisks can be fabricated with the diameters determined by the PS nanospheres. Figure 2(a) shows the white light dark-field (DF) scattering spectra of Al nanodisks with different diameters (i.e., 150, 200, and 250 nm). A red-shift of the plasmon resonance is observed as the diameter of the Al nanodisks increases from 150 to 200 nm, and the resonance then shifts to shorter wavelengths as the nanodisk diameter becomes bigger. Finite-difference-time-domain (FDTD) simulations confirm such spectral evolution, as shown in Fig. 2(b). The major plasmon peak first shows a red-shift until the disk diameter reaches 250 nm, after which the peak slowly shifts to the shorter wavelength side.³⁴ Additional simulations (Fig. S1 in the supplementary material) corroborate further blue-shifts of the plasmon resonances as the nanodisks become even larger.

Figure 2(d) shows the DF scattering spectra of the fabricated Au nanodisks. A well-defined single plasmon peak is observed for the Au nanodisk with a diameter of 150 nm, while a broader peak with multiple shoulders appears when the disk diameter increases to 250 nm. As the nanodisk diameter increases, high-order modes can be easily excited in the shorter wavelength range, while the plasmonic dipole modes shift to longer wavelengths.³⁵ Therefore, the observed spectra are a combination of the plasmonic dipole modes and high-order modes. The dark field images in Figs. 2(c) and 2(f) show that a large number of Al and Au nanodisks can be obtained by this method, and the density of the nanodisks can be readily controlled by the colloidal concentration and the spin-coating speed. The AFM images on the right side show the good morphology of individual nanodisks. It is noted that periodic nanodisk arrays can also be fabricated with this method. To that end, non-close-packed PS nanosphere arrays are required as the initial templates, which can be obtained by simple O₂ plasma etching or mechanical expansion of close-packed nanosphere arrays.³⁶ 

We also fabricated nanodisks with different diameters on generic SiO₂ substrates, which are widely used as substrates for various plasmonic sensors. Figures 3(a) and 3(b) show Au and Al nanodisks, respectively, obtained with 200 nm PS spheres on glass slides. It is clear that a large number of well-defined metal disks can also be fabricated on glass substrates. Both the Au nanodisks and the Al nanodisks show distinctive plasmon peaks at 763 and 512 nm, respectively, as shown in Fig. 3(c), which shows good agreement with the numerical simulations shown in Fig. 3(d). The plasmon resonances of the Al and Au nanodisks span the visible wavelength range providing good tunability for different plasmonic sensing applications. With its excellent stability and biocompatibility in various environments, gold is still the most widely used plasmonic material, especially for plasmonic sensors. We thus used Au nanoantennas in the following experiments.³⁷ 

As in Au-based CLAIN, the diameters of the metallic nanodisks can be readily controlled by using different PS nanospheres, which are demonstrated in Figs. 2 and 3. However, the diameters of these commercially available PS spheres often come with discrete values separated by at least 50 nm. Although customer-made spheres can be obtained, the tuning range is still limited, and the corresponding cost is significantly higher. Here, we show another practical advantage of Cu-based CLAIN that it is capable of tuning the diameters of the final nanodisks that are made from PS spheres with the same diameter. It is noted that Cu can be easily transformed in the air into Cu₂O or CuO via simple thermal heating. The addition of oxygen increases the volume of Cu films.^38,39 Therefore, we can heat the Cu nanoholes on a laboratory heating plate to deliberately oxidize the copper leading to an effective reduction of the diameters of the nanoholes. These downsized nanoholes can serve as new deposition masks for nanodisk antennas with diameters much smaller than the original PS spheres, as shown in Fig. 4.

The final Cu nanohole diameter and hence the Au nanodisk diameter can be easily tailored by controlling the heating time. Figure 4(a) shows the changing geometries of the Cu nanoholes with thermal heating. As heating time increases, more oxide grows, and the nanoholes become smaller. After 30 s, the film surface becomes rough and granular. Eventually, the nanohole can be completely filled with oxide. Figure 4(b) shows SEM images of the corresponding Au nanodisks fabricated from the reduced Cu nanoholes, together with the corresponding dark-field scattering images of individual nanodisks. Each individual nanodisk in the SEM micrograph corresponds to the DF scattering image on the right with subtle color changes.^40,41 The diameter decreases from top to bottom as the heating time increases. Figure 4(c) shows the corresponding scattering spectra of the nanodisks. The spectral evolution is distinctly presented as the nanodisk diameter gradually increases from bottom to top. For nanodisks with diameters smaller than 178 nm, a clear single peak is observed, which corresponds to the dipole resonance mode of the nanodisk. As expected, the resonance shows a consistent red-shift with increasing disk diameters. As the disk diameter becomes much larger, the spectra in the visible range become more complex, and multiple peaks appear, which can be attributed to the excitation of high-order modes in the nanodisks. The numerical simulation results, shown in Fig. 4(d), show good agreement with the experimental results. In our experiments, the heating temperature is set at 250 °C, which is above the temperature of 150 °C required to form Cu₂O.^39,42 Heating at other temperatures is also possible to control the nanohole diameters. However, it is noted that the thermal oxidation of Cu is actually a complicated process, as the oxidation involves the formation of two oxides at different heating temperatures.⁴³ By choosing a relatively higher temperature, we can simply vary the heating time to control the sizes of the Au nanodisks providing a facile means to tune the plasmon resonances of the disk nanoantennas. Compared with Au-based CLAIN, Cu-based CLAIN not only significantly reduces the overall cost of the method but also provides an additional degree of freedom to tune the nanodisk diameters and hence their optical properties.

Finally, we applied the fabricated Au nanodisks for plasmonic VOC sensing. Au nanodisks (D = 200 nm) were first prepared on ITO substrates and then immersed in a growth solution for the in situ synthesis of ZIF-8 nanoparticles. The coverage and amount of ZIF-8 materials can be controlled by the number of growth cycles.^44,45 Figure 5(b) shows the AFM and SEM images of samples after two cycles (left) and four cycles (right) of growth. Au nanodisks are still perceivable, with scattered ZIF-8 observed between the nanodisks after two cycles. By contrast, the Au nanodisks are barely visible after four growth cycles of ZIF-8. Figure 5(c) shows the x-ray diffraction (XRD) spectrum of the synthesized ZIF-8, and a good agreement with the database is achieved, indicating the good quality of the ZIF-8 materials. X-ray photoelectron (XPS) spectroscopy was also carried out on the samples confirming the in situ growth of ZIF-8 onto the Au nanodisks (Figs. 5 and S2 in the supplementary material).

To evaluate the sensor’s performance, we chose ethanol vapor as the target, as it is one of the most common VOCs in the environment. Over-inhalation of ethanol vapor could lead to neurological disease.⁴⁶ The samples were placed inside a closed chamber with glass windows on the top and bottom of the chamber. With a microsyringe, various volumes of ethanol (1, 2, 3, 4, or 5 µl) were injected into the chamber, and the ethanol was then allowed to diffuse for 10 min. We recorded the transmission spectra of the samples under different ethanol concentrations at room temperature. With no ethanol, the plasmon resonance appears at 996.1 nm, which is much larger than that of bare Au nanodisks due to the deposition of ZIF-8 (four cycles). When ethanol is adsorbed into the porous ZIF-8, the effective refractive index of the surroundings of the samples increases, leading to a red-shift of the resonance, as shown in Fig. 6(a). Figure 6(b) shows the values of the wavelength shift under different ethanol concentrations. A linear relationship is observed as expected, and a sensitivity of 3.1 pm/ppm is achieved. For the sample with two cycles of ZIF-8, a smaller sensitivity of 0.5 pm/ppm is obtained. As shown in Fig. 5(b), two cycles of ZIF-8 do not give a complete coverage of ZIF-8 over Au nanodisks. As the ethanol-collecting layer, four cycles of ZIF-8 thus provide more sites for ethanol adsorption and bring more ethanol molecules close to the hot spots of the Au nanodisks leading to a larger sensitivity.

Figure 6(c) shows the role of the ZIF-8 layer thickness in the sensing performance. An ethanol concentration of 3943.8 ppm is used for the evaluation. Due to the low coverage of ZIF-8, two cycles of ZIF-8 show a slightly larger response than the bare Au nanodisks. As the ZIF-8 increases to four growth cycles, the wavelength shift becomes six times larger than that of two ZIF-8 cycles suggesting good transportation of ethanol molecules to the vicinity of the Au nanodisks. Interestingly, as the ZIF-8 becomes thicker (six growth cycles), the wavelength shift decreases, which can be attributed to the possibility that the ethanol molecules are likely accumulated on the top part of the film away from the enhanced near-field of the plasmon modes.^47,48 Figure 6(d) shows the selectivity of the sensors to ethanol vapor at 3943.8 ppm. The sample with four cycles of ZIF-8 is exposed to three different kinds of alcohol (ethanol, methanol, and isopropanol) and shows a higher response to ethanol than the other two types. The selectivity of the sensor comes from the physical pore size of ZIF-8. Smaller molecules are easier to penetrate into the ZIF-8 adsorption layer. It is noted that the sensors still show a certain degree of nonspecific detection of other VOCs. It is possible to enhance selectivity by combining different metal-organic framework (MOF) materials in a single sensor design.

In summary, we have demonstrated highly-sensitive plasmonic VOC sensors with an upgraded version of the CLAIN nanofabrication technique that uses Cu as the sacrificial material. Cu-based CLAIN inherits all the advantages of Au-based CLAIN, and it significantly reduces the overall cost of the method, making it even more cost-effective. Furthermore, the use of Cu actually allows great tunability of the sizes of the plasmonic nanodisks by straightforward thermal heating, providing an additional and important advantage of Cu-based CLAIN for practical applications. Highly sensitive plasmonic VOC sensors are demonstrated by combining the fabricated Au nanodisks with porous ZIF-8 materials. It is anticipated that Cu-CLAIN can be used to fabricate various optical nanoantennas over large areas. The combination with other functional materials makes Cu-CLAIN a feasible and attractive technique to fabricate highly sensitive plasmonic sensors.

Colloidal solutions of PS spheres were purchased from Beijing Zhongkeleiming Technology Co., Ltd. (diameters: 150 and 250 nm) and Tianjian BaseLine, Inc. (diameters: 200 and 300 nm). They were diluted in deionized water to 100th–300th of the original concentration and then sonicated for 3 min. Then, they were spin-coated (1500 rpm for 60 s) on clean glass and ITO substrates. Afterward, copper films with a thickness of 75 or 100 nm were deposited onto the substrates by an electron beam evaporator (Explorer 14, Denton). After scraping off the PS spheres from the sample with a PDMS sheet, we performed a second metal deposition on the samples. In particular, for the preparation of gold nanodisks, 2 nm Cr was deposited as an adhesion layer before the deposition of Au film. Finally, the Cu film was etched off with a mixture of NH₄OH and H₂O₂ (volume ratio 1:1), and the metal nanodisks remained on the substrate.

After repeating the above steps with 300 nm diameter PS spheres, we heated the copper hole film samples at a constant temperature of 250 °C using a laboratory heating plate. After that, we performed 2 nm Cr and 30 nm Au film deposition and etched off the Cu and Cu₂O mixed films. We varied the heating time to obtain different nanodisks with reduced sizes.

We grew the MOF material ZIF-8 in situ on the prepared densely distributed gold nanodisks in a methanol solution containing zinc nitrate hexahydrate (12.5 × 10⁻³ m) and 2-methylimidazole (25 × 10⁻³ m). Then samples were placed in a sealed gas chamber and injected with different volumes of ethanol. The gas chamber was placed under a fiber-coupled spectrometer (Ocean Optics) to measure the transmission spectra of the samples before and after the ethanol evaporation.

The morphology of the prepared samples was analyzed by atomic force microscopy (Dimension FastScan, Bruker) and scanning electron microscopy (Thermo Scientific FEI, Czech Republic) operating at 5 kV. White light dark-field scattering spectra of individual nanodisks were obtained using a transmission dark-field optical microscope (Eclipse Ti-2, Nikon) integrated with a spectrometer (Kymera 328i-A, Andor) and a charge-coupled device camera. We used a white light source (a halogen lamp, 12 V, 100 W) to illuminate the sample through a dry dark-field objective (NA: 0.8–0.95). The scattered signal was collected by a 60× objective (NA = 0.70) and passed through a slit in front of the spectrometer. X-ray photoelectron spectroscopy was recorded using Axis Ultra DLD (Britain, Kratos) with a monochromated Al Kα source (1486 eV) for excitation. X-ray diffraction patterns were recorded on an X-ray diffractometer (Rigaku Ultima IV) at a rate of 5°/min using Cu Kα radiation (1.5418 Å).

A commercial software package (FDTD Solutions, Lumerical) is used to simulate the scattering spectra of the nanodisks. The simulated nanostructures are excited by a Total-Field Scattered-Field (TFSF) source with the wavevector along the z axis and the polarization along the x axis. The dielectric functions of the metals are extracted from the work of Palik.⁴⁹ The thickness of the ITO layer is set to 150 nm.

See the supplementary material for information about the scattering spectra of Al nanodisks with larger diameters and additional XPS spectra of the sample.

The authors acknowledge the support from the National Natural Science Foundation of China (Grant Nos. 61975067, 62005100, and 12004140), the Department of Science and Technology of Guangdong Province (Grant Nos. 2019A1515110246, 2020A151501905, and 2021A1515012352), the Guangdong Provincial Innovation and Entrepreneurship Project (Grant No. 2016ZT06D081), and Special Funds for Chinese Central Universities (Grant No. 12819011) that supported this work.

The authors have no conflicts to disclose.

Meng Li: Investigation (lead); Methodology (lead); Software (lead); Writing – original draft (supporting). Yan Huang: Investigation (supporting); Methodology (supporting). Lipeng Sun: Investigation (supporting). Zhaoqiang Zheng: Investigation (supporting). Churong Ma: Investigation (equal). Xiangping Li: Supervision (supporting). Bai-Ou Guan: Supervision (equal). Kai Chen: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Funding acquisition (lead); Project administration (lead); Supervision (lead); Writing – original draft (lead).

The data that support the findings of this study are available from the corresponding author upon reasonable request.