The development of nanofabrication technology forms the cornerstone of modern nanoscience; achieving nanoscale manufacturing with finer patterns is thus essential for broad applications such as nano-electronics, metrologies, and molecular biology. The most commercialized technology, photolithography, provides the platform for large-area and high-throughput production of electronic and photonic devices. Owing to the optical diffraction limitation, shrinking the wavelength of the light source has been continuously investigated to achieve higher resolution. The deep/extreme ultraviolet lithography technique and soft X-ray [1,2] have been significantly expanded over the decades to further push the nanofabrication limit. However, the high cost and complex process bound their application in large-scale production, which can hardly be popularized in the nanofabrication of small and customized prototype devices. The maskless alternative schemes, such as electron beam lithography [3] and focused ion beam [4], are widely available in nanoscale fundamental research, but the similar challenge to balance the accuracy and the cost still holds. Direct laser writing [5] is a low-cost lithography technique with high processing speed, while the resolution is still restricted by the diffraction limit. Therefore, a high-precision nanofabrication technology with low cost and high flexibility is still highly desired.

Surface plasmons are recently employed for investigating plasmonic lithography (PL) in the near field [6], which provides a promising solution to reduce the feature sizes with conventional ultraviolet light sources. Utilizing the localized surface plasmon (LSP) induced by either space light or SPPs, PL breaks the intrinsic limit of optical diffraction and confines the light at deep-subwavelength scale [7–10]. Nanoscale patterning with half-pitch resolutions less than 50 nm has been confirmed using a 365 nm wavelength light source [11,12], demonstrating the potential of PL as a viable and low-cost alternative lithography technology with high resolution. The LSP nanostructure is an essential element in PL to generate sub-diffraction-limited light spots with elaborately designed nano-apertures or nanoparticles [13,14]. For instance, the PL resolution strongly depends on the ridge gap of the bowtie aperture [15,16]. The light spot is commonly equal to or slightly larger than the minimum size of the LSP nanostructure. To achieve a smaller spot, the key size of the LSP nanostructure has been continuously reduced to approach the limiting fabrication resolution of focused ion beam milling (<20nm), which faces the problem of time consumption and low qualification rate. On the other hand, though the conventional photoresist material is applicable to PL, it has been reported that utilizing the heat from LSP as well as the thermal type resist can further compress the feature size down to half of the light spot [17]. However, a bulky and expensive picosecond pulsed laser is needed to suppress thermal diffusion and limit the exposed feature. Therefore, a scheme to realize smaller feature sizes with an easy-fabrication nanostructure and inexpensive continuous-wave laser should be explored, which would greatly reduce the tool cost.

To overcome the above obstacles, herein we establish and experimentally demonstrate a thermal probe nanolithography system based on an infrared surface plasmon (ISP). With four exquisitely designed annular slits on an atomic force microscope (AFM) tip, the radially polarized illumination is efficiently converted to ISP, which is then guided and focused at the tip apex to excite an ultra-strong LSP spot. The lower surface plasmon loss at the infrared band leads to higher near-field enhancement compared with the case of ultraviolet light illumination. Besides, a longer wavelength ensures the relatively large key size of the nano-annuli, which reduces the fabrication difficulty. The deep-subwavelength light field squeezing of infrared LSP and optical-thermal effect further leads to the increase of tip temperature above the writing threshold of the thermal resist. We emphasize that the annular slits greatly improve the local thermal concentration even under the continuous-wave laser illumination, which not only allows for heating the tip with reduced laser power, but also minimizes the heating zone without affecting the neighboring region. With these advantages, subwavelength features with a linewidth down to 13 nm are achieved using a 1064 nm continuous-wave semiconductor laser source of only several tens of milliwatts radiation power. In comparison with other PL technologies, our scheme may further lower the cost and technical threshold of nanofabrication.

Figure 1 depicts the schematic diagram of the proposed ISP-assisted thermal probe nanolithography. A commercially available AFM tip (Nanosensor, uniqprobe) is used as the base of our thermal probe. The cantilever made of a quartz-like material allows the infrared laser to propagate inside the probe with little absorption. A 50 nm thick Au film is coated on the tip side, with the same material and approximate thickness as that of reflective coating on the backside. Such design is beneficial for relieving the bending of the cantilever caused by the nonuniform thermal expansion under laser irradiation [18]. The incident light (red beam) impinges the probe from the cantilever side where a 10 μm aperture is designed as the incident window. The AFM feedback laser (purple beam) is reflected by the remaining Au film into the quadrant photodiodes detector without affecting the AFM measurement functions. The lithography laser transmits through the cantilever and arrives at the four concentric annular slits etched on the Au film. To prevent the direct transmission of the lithography laser, the slit width should be subwavelength level and is set as 200 nm. Due to the rotational symmetry of annuli, the radially polarized beam is employed to effectively excite SPPs as indicated by the inset in Fig. 1. The coupled SPPs are then directed downwards to ultimately boost the LSP with an ultra-strong spot at the tip apex. The dense energy accumulation further heats the tip, which can be utilized for the thermal probe nanolithography as the probe scans above the thermal resist in the near-field condition. Note that the locations of each annular slit are deliberately tailored to satisfy the phase-matching condition so that the SPP excited by each slit interferes constructively. The radii of the four annuli are 1.4 μm, 2.6 μm, 3.3 μm, and 3.8 μm. Increasing the number of annuli is expected to couple more light into SPPs and may further enhance the LSP intensity. However a larger annulus may exceed outside the edge of the cantilever, due to the taper sides [Fig. 3(b)] and thus only four annuli are employed here.

To visualize the SPPs’ excitation and their conversion into LSP, we first perform electromagnetic simulations using the commercial software COMSOL Multiphysics. Two coated plasmonic probes with and without annular slits working at the excitation wavelength of 1064 nm are compared. A refractive index of 1.45 is assumed for the quartz-like probe, while the refractive index for Au film is introduced from Ref. [19]. Under the illumination of the radially polarized beam from the top, the electric field intensity distributions are plotted in the insets of Fig. 2(a). As expected, in both cases electromagnetic energy is concentrated at the tip apex, inducing local field enhancement. But differently, the unpatterned conical probe simply serves as a plasmonic waveguide with a gradually decreasing cross section [20]. It should be pointed out that the space light cannot reach the tip apex, while only the surface plasmon with a reduced wavelength during propagation can be guided without experiencing mode cut-off and finally compressed down to the nanometer scale. The smooth surface of metallic film results in low surface plasmon conversion efficiency. Most incident optical energy is reflected and the focused LSP intensity is relatively low. In contrast, the annular slits can be regarded as a grating coupler to effectively raise the surface plasmon conversion efficiency. The phase difference between ISP excited by two arbitrary annular slits equals 2nπ, which satisfies the condition of Fabry-Perot resonance. It is hard to derive the optimal annular radii solely through the analytical method due to the irregular geometry of the probe. We construct the geometrical model of the probe based on the scanning electron microscopy (SEM) images and implement the parameter sweep to sequentially decide the optimized radii. Besides, the ISP mainly propagates along the outside surface of the metallic film and the optical energy is then transferred to the outside of the waveguide. As the tip radius decreases, the wave number of ISP gradually increases, leading to a decreasing group velocity and finally achieving the LSP at the tip apex [21]. The corporate action of raised conversion efficiency, constructive interference, energy transfer, and decreased group velocity generates the super-focused spot with significantly enhanced intensity. The field intensity distribution at the plane 10 nm distance away from the tip is shown in Fig. 2(a). The central focal spot of the patterned probe has a peak intensity 35 times that of the unpatterned probe with a 62 nm FWHM spot size.

Besides the slit design, the near-field enhancement and spot size also strongly depend on the radius of the tip apex (see details in Appendix A). A conservative radius of 40 nm is used in our simulation. In fact, a much sharper tip with a radius less than 5 nm has been experimentally obtained [9], which would further compress the focusing spot and increase the local field at the tip apex. We also notice that side illumination has been utilized to realize plasmonic nanofocusing, in which the linearly polarized laser impinges the tip from the lateral side with polarization parallel to the tip axis [22,23]. However, the focusing efficiency is much lower than the top incidence method we adopt here. For the coated plasmonic probe without slits, the peak light intensity of top incidence is 36 times that of the side incidence under the same irradiation power (see details in Appendix B).

The inherent loss caused by the metal absorption further heats the probe. We note that the super-focused mode not only applies to the light field, but also effectively improves the local thermal concentration by minimizing the heating zone. Figure 2(b) shows the simulated temperature profiles under heating from the optical field for the above two kinds of plasmonic probes. Here a continuous-wave laser is employed as the light source and a threshold temperature of 350°C is assumed, which corresponds to the decomposition temperature of the thermally sensitive polyphthalaldehyde (PPA) [24]. In the simulation only the heat transfer due to conduction is considered since the contributions of free convection and radiation can be negligible in our investigated temperature region below 400 K [25]. With the annular slits, the hot spot is more focused with a heating feature size of 34 nm, about two-thirds that of the probe without slits, while the required laser power is greatly reduced from 190 mW to merely 20 mW. Since the dissipation loss of a surface plasmon is the only heat source, which is proportional to the local light intensity, the heating region of the patterned plasmonic probe is mainly restricted at the tip apex, while the probe without annular slits is more uniformly heated, as indicated by the insets in Fig. 2(b). Therefore, by combining SPP focusing, LSP conversion, and nanoscale thermal management, the super-focused mode efficiently squeezes light into the deep sub-wavelength scale and is expected to achieve nanolithography with tens of nanometers resolution using a 1064 nm continuous-wave laser source.

To verify the lithography performance, next we establish a thermal probe nanolithography system by equipping a modified AFM with the fabricated plasmonic probe. As indicated in Fig. 3(a), a diode-pumped solid-state laser serves as the lithography source. A linear polarizer is first applied to ensure the horizontal polarization, and then the emergent light is converted to a radially polarized beam with a radial polarization converter (see Fig. 7 in Appendix C for the polarization detection). After multiple reflections, the laser is guided into a parallel direction with respect to the tip axis and finally focused on the tip cantilever by an objective lens (5×, NA=0.14). To realize the optical alignment between the incident laser and tip, the mirror inside the AFM and the objective lens are mounted onto a fine-tuning XY stage, which allows for precise control of the illuminating position (see Fig. 8 in Appendix C for the experiment setup). Figure 3(b) shows the SEM images of a fabricated plasmonic probe with four concentric annular slit structures. To obtain this sample, the 5 nm Cr film for adhesion and the 50 nm Au film are deposited on the tip side via DC magnetron sputtering. Then the optimal annular silts are milled by a Zeiss triple-beam focused-ion-beam (FIB, Zeiss Orion NanoFab, Jena, Germany) microscope using a 23 pA current with a 30 kV Ga ion beam [Fig. 3(c)]. On the backside, an aperture is milled to allow the transmission of an incident laser [Fig. 3(d)].

With the nanolithography system and plasmonic probe at hand, we then perform the nanolithography experiment. First, the PPA solution with a mass concentration of 2.0% is spin-coated on a silicon wafer at 3000 r/min. The substrate is then baked on a heating plate at 110°C for 2 min to completely remove the resist solvent and obtain a 30 nm thick polymer film. Next, we scan the resist surface using the plasmonic probe at a speed of 1 μm/s with the lithography laser on to write the predesigned patterns. Finally, the laser beam is turned off and the AFM measurements are done in the same region to observe the surface topography. Soft-tapping mode is chosen to avoid mechanical damage to the tip. A closed-loop configuration of the scanner stage is enabled to increase the scan accuracy of lateral positions. When heated above the threshold temperature, the polymer is decomposed into volatile monomer and evaporated at a microsecond time scale. Therefore, the lithography patterns can be measured in situ without the development process. Figure 4(a) shows the image of the lithography line with a 13 nm FWHM at a laser power of 50 mW. This is the threshold power and no surface modification is observed at lower laser irradiation. We note that a sharper tip with radius of 10 nm can realize a much smaller lithography linewidth than the simulation prediction of 34 nm in Fig. 2(b) (see Table 1 in Appendix A). Moreover, the difference of threshold power between simulation and experiment can be attributed to three parts. First, it is hard to evaluate the heat required for the decomposition and evaporation of the resist in simulation. Second, the nanostructure imperfection of the fabricated probe and the scattering losses related to the metal surface roughness reduce the focused light intensity at the tip apex, and the optical components cause the decay of laser power. Third, the distance between the tip and the substrate surface that varies between zero and tens of nanometers in tapping mode may also affect the heating of the resist.

Above the threshold, the feature size of the writing pattern can be controlled by regulating the laser power. Under the laser irradiation of 75 mW, the linewidth and depth of the pattern are increased to 29 nm and 3 nm, respectively, as shown in Fig. 4(b). Besides, this nanolithography system shows high flexibility of arbitrary 2D pattern processing by designing the scan trajectory. The logo “BJAST CTIC” is written on the resist as shown in Fig. 4(c), demonstrating the advantage of maskless lithography. We note that to minimize the heat-affected zone to the nanoscale, the ultrashort pulsed lasers have been commonly used as the light sources for thermal direct-write lithography [17,23]. In this method, very high transient light intensities can be obtained and the average power is kept to a relatively low level. Therefore, thermal diffusion into a neighboring region is effectively prevented, while in our strategy, the continuous-wave laser is demonstrated to enable the nanolithography with a feature size down to 13 nm. Such superiority can be ascribed to the design of the super-focused mode, which ensures good thermal confinement and reduces the heated features. Based on the meticulously designed ISP thermal probe, a smaller pattern size may be further expected by applying the pulsed lasers as the light source (see Appendix D for the improved feature size using a pulsed laser).

In summary, we design and fabricate a plasmonic probe with four concentric annular slits working at 1064 nm wavelength. This approach is universal and applicable to a broad wavelength range in the near-infrared spectrum (see Appendix E). The long wavelength of infrared light leads to a 200 nm width of the slit nanostructures, which greatly reduces the fabrication difficulty. The constructive interference of infrared surface plasmons excited by the slits forms the super-focused mode for both the light and thermal fields. With such key components, a thermal direct-writing nanolithography system is established and we demonstrate how a continuous-wave laser can serve as the lithography laser source. Benefiting from the nanoscale thermal confinement at the tip apex, lines and 2D patterns written on the thermal resist are recorded with feature sizes down to 13 nm. The presented results proved that our strategy promises a low-cost, high-resolution, and maskless lithography system.

Besides the photothermal method, another thermal lithography technology based on a scanning probe is well-established by heating the probe with a hot resistor [26,27]. The hot current flows through the cantilever legs, resulting in joule heating in the base region of the tip. A nanoscale topographical pattern is generated using a specially designed three-terminal probe integrated with two separated resistive heaters and a capacitive platform [28]. However, such heatable probe requires complicated fabrication processes, which are usually designed for a specific device or fabricated in house [29]. Moreover, the thermoelectrical scans are performed through the resistive read sensor to measure the surface topography. The different mechanism makes it incompatible with the common AFM that relies on the quadrant photodiodes detector feedback system, and a major revamp on the device is thus unavoidable, while our strategy bridges the gap between a nanolithography system and an AFM by readily introducing a top incident laser, which reduces entry barriers and may facilitate the widespread adoption nanolithography technology. Such cost-effective PL technique is expected to find potential applications in fundamental research and nano-devices prototyping.