Femtosecond laser-induced dewetting of sub-10-nm nanostructures on silicon in ambient air

Laser nanoscale manufacturing technology has emerged as a highly precise method for the fabrication and processing of semiconductor devices.^1–6 The precision of laser manufacturing is closely related to the size of the laser focal spot.^7–10 Therefore, traditional laser manufacturing methods face significant challenges as semiconductor device sizes continue to decrease. To enhance laser manufacturing resolution, it is often necessary to minimize the size of the laser focal spot.^11,12 According to the Rayleigh formula D = 1.22λ/NA (where λ is the laser wavelength and NA is the numerical aperture of the objective lens used), reducing the focal spot size can be achieved by decreasing the laser wavelength or increasing the numerical aperture of the focusing objective lens.^13,14 On one hand, increasing the numerical aperture of the lens often requires immersion in a high refractive index medium, limiting the application scenarios of high-precision laser manufacturing. On the other hand, using short-wavelength incident lasers, such as deep ultraviolet or extreme ultraviolet lasers widely applied in photolithography, can significantly improve laser manufacturing resolution.^15–17 However, the high cost of short-wavelength light sources restricts the application scenarios of laser manufacturing. Therefore, achieving nanoscale surface structure fabrication in the far field and in the air with traditional light sources remains a critical challenge in current high-precision laser manufacturing.

Ultrafast laser processing, based on nonlinear absorption effects, offers a new approach with advantages such as high precision and a wide range of applicable materials for manufacturing beyond the diffraction limit of nanostructures.^18–20 In 2011, Joglekar utilized the nonlinear absorption effect of materials to achieve the ∼30 nm structures on glass surfaces.²¹ In 2016, by combining the nonlinear absorption effect of femtosecond lasers with stimulated emission loss, Gan achieved approximately 9 nm feature structures in specific photoresists.²² In 2019, by exploiting the far-field induced near-field breakdown characteristics of femtosecond lasers, ∼18 nm feature size structures on silicon substrates were realized.²³ In 2020, Lin utilized the orthogonal polarization incubation effect of dual-beam lasers to produce structures with a minimum size of ∼12 nm on Si surfaces.²⁴ These techniques have significantly promoted the development of femtosecond laser technology in the extreme micro–nano-manufacturing field, providing new insights for the application of femtosecond lasers in the semiconductor industry. However, these methods require high polarization characteristics of the processing beam, and the polarization properties of the beam have a significant impact on the processing results. It is challenging to apply these methods in applications where the polarization characteristics of the laser beam cannot be controlled in real-time, limiting the application scenarios of femtosecond laser extreme micro–nano-manufacturing.

In this study, we propose a femtosecond laser far-field induced dewetting method based on microsphere lenses in ambient air, achieving structures with a minimum half-width of ∼9 nm. Compared with a microsphere, after the tightly focused beam passes through the microsphere, the spot diameter is reduced by twice without changing the location of the focused spot, which results in a smaller, thinner molten pool. Theoretical analysis and numerical simulation of temperature fields during laser processing are conducted based on a two-temperature model, and the simulation results are consistent with experimental results. By alternating the interval of two consecutive pulses, it gives rise to three distinct nanostructures. The proposed laser-induced dewetting extreme manufacturing method provides a new approach and means for far-field ultra-high-resolution processing of silicon surfaces in the semiconductor field.

Figure 1(a) illustrates the experimental setup. Notably, in contrast to most microsphere-assisted lithography studies that place the microsphere in the near-field to directly process photon nano-jets,²⁵ in this study, a high-refractive-index barium titanate microsphere with a diameter of 20 μm was positioned ∼30 μm above the sample. Higher refractive indices bring photon nano-jets generated by the microsphere closer to its center. A photon nano-jet for focused light with an aperture angle of 30° was generated inside the microsphere. Figure 1(d) shows the optical field distribution after considering the interaction between the microsphere and tightly focused light, calculated using commercial finite-difference time-domain (FDTD) software. Photon nano-jets are located inside the microsphere, ∼4 μm from its lower end. The focal point of the focused light obstructed by the microsphere barely moves vertically compared to that without the microsphere. Figure 1(e) compares the intensity distribution of the direct incident light with the focused light coupled by the microsphere, which has a smaller waist size and a longer focal depth. Nanostructures can be directly formed on the material surface with ∼0.75 nJ, which is 15% less than the ablation threshold energy [Fig. 1(b)]. The width of the two nanowires was 20 nm, with a spacing of 197 nm. The specific formation mechanism of the nanostructures is discussed in the following section. Figure 1(c) shows the experimental system. A sequence of femtosecond pulses with a pulse width of 280 fs and wavelength of 515 nm, filtered using a spatial filter, provided a clean processing light source. The objective lens used had a numerical aperture of 0.6, which converted the incident light into focused light with an aperture angle of 30° and a working distance of 11 mm. The theoretical focal spot size is 1.04 µm, according to D = 1.22λ/NA. The microsphere was fixed on an atomic force microscope cantilever. Both the microspheres and the sample were controlled using piezoelectric ceramic positioners (PI, P-562.3CD, and P-621.2CL) to provide nanometer-level displacement (Fig. S1). The working distance between the combined microsphere and sample was ∼30 µm. Therefore, the working distance of the system when combined with the microsphere was considered to be ∼30 µm.

After irradiating the material with a single femtosecond pulse, the acoustoelectric coupling process ended at ∼4 ps, and the electronic and lattice temperatures reached 2350 °C [Fig. 3(c)]. Energy loss due to thermal radiation from the base was not considered throughout the calculation process. Figures 3(a) and 3(b) show the steady-state temperature field distribution and morphology of the molten area. The molten pool has a depth-to-diameter ratio of ∼1:3, with the center temperature higher than the edge, resulting in a surface tension gradient σ in the outward direction due to ∂s/∂T < 0 of the liquid metal.³² To minimize the surface free energy, the surface tension of molten silicon pulls the surface toward the center. Figure 3(d) shows the results of the molten pool melting after solidification, which was formed by the joint action of the frictional force and gravity between the molten silicon liquid and solid silicon during the process of surface tension and molten silicon liquid retraction. The diameter of the molten pool was ∼100 nm, and the molten material shrank inward, forming a nanogroove at the solid–liquid interface with a width as low as 9 nm. The highest point exceeded the substrate surface by 1 nm. The hydrodynamics simulation is shown in Fig. S2. The element distribution of the molten pool was characterized using EDX, with Si accounting for 99.9% of the total chemical composition, indicating that the nanostructures induced by the low-energy single-pulse femtosecond laser were physical property changes [Fig. 3(e)].

The distance between adjacent laser pulses plays a crucial role in the formation of nanostructures. By controlling this distance, three different nanostructures were obtained, as shown in Figs. 4(a)–4(c). The platform velocity was kept constant at 10 μm/s, while the laser repetition frequency was set to 500, 200, and 100 Hz, resulting in pulse distances of 20, 50, and 100 nm, respectively. The 3D and 2D AFM scanning results of the nanostructures are shown in Figs. 4(d)–4(f) and 4(g)–4(i), respectively. When the distance between pulse points was set at 100 nm, the nanorings formed by a single pulse were adjacent to the previous ones, and their formation was centered on the midpoint of the pulse. The resulting nanoring had a radius of 50 nm. The full-width half-maximum of the resulting nanowire was only 9 nm, and the distance between adjacent nanowires was ∼100 nm. During the transition from the solid to the molten state, non-equilibrium forces act on the material owing to the constant displacement speed of the platform. This causes the molten liquid material to pile up toward the other end, resulting in an eccentric cone structure. The eccentric direction is opposite to the direction of the platform movement. When the distance between pulses exceeded 100 nm, the radius of the molten pool was smaller than the distance between the two adjacent pulses, and continuous independent nanoring structures appeared, similar to Fig. 2(d). As the pulse distance decreased, the distance between the centers of the melt induced by a single pulse gradually decreased, and the overlap area increased. The destruction threshold of the material irradiated by a single-pulse femtosecond laser decreased. Therefore, when the distance between pulses was 50 nm, the distance between the two nanowires significantly increased to 200 nm. Meanwhile, owing to the increase in the volume of the molten pool, the linewidth promoted by dewetting increased to 18 nm. The eccentric cone structure evolved into a curved moon-shaped ripple structure, with adjacent ripples spaced 50 nm apart, which is the distance between the pulses. When the distance between the pulses was further reduced, the result of conventional laser direct writing appeared. The spot size is 490 nm, according to the simulation. At this point, the overlap between adjacent pulses exceeded 95%, and the material between the two nanowires was completely removed, resulting in an ablated conventional channel with a width of 242 nm. Owing to the nonlinear absorption effect of the material, ablation did not affect the surface morphology of the surrounding substrate. Figure 5 shows the effect of pulse intervals of 20–100 nm on the width of the laser-induced nanostructures. When the pulse width was above 100 nm, the width of the nanowires hardly changed because the nanowire structure formed individual nanodots. The decrease in pulse interval leads to an increase in the overlap area, directly leading to an increase in the laser damage area. Consequently, the width of the nanowires and the distance between them increased as the material was completely removed. The polarization direction is perpendicular to the scanning direction. However, the maintained structure preparation regardless of the laser polarization direction can be ascribed to the circular shape instead of an ellipse,²³ induced by a single pulse. As such, the processing direction can affect the orientation of the half-moon ring in the middle of the structure (Fig. S3).

Finally, we explored the potential of femtosecond laser-induced far-field dewetting for nanoscale direct writing. Figure 6(a) shows a curved structure with a full-width half-maximum of less than 20 nm. The dark areas visible in the AFM image are attributed to fluctuations in the laser energy, which either eroded or melted more of the material. Figures 6(c) and 6(e) demonstrate that femtosecond laser-induced far-field dewetting can maintain consistent uniformity over a large area, producing feature structures smaller than 20 nm. Moreover, the material’s nonlinear absorption effect results in clear boundaries for the nanostructures, reducing the thermal influence area and eliminating material accumulation near the processed area. By appropriately controlling the laser energy and scanning speed, thermal deformation can be greatly suppressed, leading to smooth processing quality. This is demonstrated in Fig. 4(e), which clearly shows the effect of the appropriate laser energy and scanning speed on thermal deformation.

We utilized femtosecond laser-induced far-field dewetting to prepare structures with a minimum feature size of 9 nm on the substrate surface, providing a novel approach for sub-10-nm processing with femtosecond laser-induced far-field dewetting and achieving arbitrary structure patterning with a half-width of 18 nm. The precise control of energy for femtosecond laser-induced far-field dewetting is highly demanding, and energy fluctuations can significantly affect the morphology of the nanostructure. By comparing the energy distribution of the direct femtosecond laser incidence, we believe that reducing the diameter of the focused light spot by other approaches can manufacture similar structures. Benefiting from an increased working distance, the application of femtosecond lasers to induce material dewetting demonstrates potential in the intravital modification of glass, particularly in endeavors such as the fabrication for optical waveguides with reduced diameters.³³ Furthermore, the approach has potential application scenarios of affordable, ultra-high-precision, high-speed laser direct writing in semiconductors.

See the supplementary material for a detailed description of the model of hydrodynamics, the control of the piezoelectric ceramic positioner, and the effect of processing direction on the structure.

The authors wish to acknowledge the funding provided by the National Key R&D Program of China (Project No. 2022YFB4700100), the National Natural Science Foundation of China (Grant Nos. 61925307, 61927805, and 61973298), and the CAS Interdisciplinary Innovation Team (No. JCTD-2019-09). The Innovation Promotion Research Association of the Chinese Academy of Sciences (No. 2022199). The Applied Basic Research Program of Liaoning Province, China (No. 2023JH2/101600037).

The authors have no conflicts to disclose.

Hao Luo: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Writing – original draft (lead). Xiaoduo Wang: Conceptualization (equal); Formal analysis (equal); Methodology (lead); Writing – original draft (equal); Writing – review & editing (lead). Yangdong Wen: Conceptualization (equal); Data curation (equal); Validation (equal). Ye Qiu: Writing – original draft (equal). Lianqing Liu: Conceptualization (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). Haibo Yu: Conceptualization (lead); Supervision (lead); Writing – review & editing (lead).

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