Topological transition from deeply sub- to near-wavelength ripples during multi-shot mid-IR femtosecond laser exposure of a silicon surface

1. Introduction

Femtosecond (fs) laser ablation is the key enabling process in ultraprecise scribing [1], slicing of silicon and solid dielectrics [2,3], optical memory direct recording [4], drilling of microfluidic elements [5] and many other well-established or emerging technologies. In a low-fluence spallation regime, nanometer-deep (∼10 nm/shot) [6,7] fluence-independent flat craters can be produced, while in the succeeding high-fluence phase-explosion regime ablation rate per pulse monotonically increases versus fluence [6–9]. Under sub- or near-threshold conditions and multi-shot fs-laser exposures, fs-laser surface ablation proceeds inhomogeneously – either via low-fluence plasmon-polaritonic generation of deeply sub- or near-wavelength relief ripples (laser-induced periodical surface structures, LIPSS) [10–18], or via regular surface microstructuring (arrays of micro-spikes) as a result of hydrodynamic instabilities [10,19,20], sometimes accompanied by homogeneous ablation across the craters [20]. In both these cases of inhomogeneous fs-laser ablation, mass transfer mechanisms and rates per pulse are challenging and not experimentally studied yet, implying either non-ablative melt displacement from relief trenches toward its ridges [11], or ablative removal of melt from relief trenches [12], accompanied by redeposition of ablation products onto its ridges [10,15]. Moreover, both shallow spallation and deep phase-explosion removal mechanisms can be locally involved in ablation during LIPSS or micro-spike formation, resulting not only in shallow and deep regular surface reliefs, but also in corresponding different surface nanotopographies [12]. Despite some enlightening predictions [21], systematic quantitative studies of LIPSS reliefs, in particular – drastic topological relief transformations under prolonged fs-laser exposures – are still missing.

In this paper we report on spatially-resolved MIR fs-laser micro-patterning and ablation of silicon surfaces as a function of laser exposure, with the MIR wavelengths used to upscale LIPSS till micrometer dimensions and visualize their nanotopography by UV-laser confocal scanning profilometry. As a result, an exposure-dependent topological transition from shallow deeply-subwavelength ripples to deep near-wavelength ripples, oriented parallel and perpendicular to the laser polarization, respectively, the related strong enhancement in laser coupling to the relief and ablative LIPSS depth, as well as the underlying mass transfer mechanisms were experimentally characterized for the first time.

2. Experimental details

An amplified Ti:sapphire laser system Astrella (Coherent) provides the fundamental harmonic output at the wavelength of 800 nm (pulse duration τ ≈ 35 fs, maximum pulse energy ≤7 mJ, beam quality M² > 1.2, maximum repetition rate – 1 kHz), converted to MIR pulses at λ ≈ 2.6, 3.0, 3.5, 4.0 and 4.5 µm (τ ≈ 80 fs) using an optical parametric amplifier (Topas Prime, Light Conversion) [22]. In the two-step conversion scheme, first, signal and idler were generated, using a collinear OPA, with some residual 800-nm pump. In the second step, a non-collinear difference frequency generator (LightConversion Inc.) produced the mid-IR pulses tunable in the range of 2.6–5 µm, with the signal, idler, and MIR radiation spatially separated, leaving at different beam angles and no fundamental radiation impinging the sample surface.

The surface irradiation of a commercial weakly B-doped Si wafer (10¹⁴ cm⁻³) arranged on a motorized translation stage, by the MIR pulses with the peak fluence F₀ ≈ 1 J/cm² (about 5–6F_mod, for the nearly λ-independent single-shot modification (melting+ amorphization, cf.g., F_mod ≈ 0.27 J/cm² at 800 nm [23]) thresholds F_mod(λ ≈2.6–4.5 µm) ≈ 0.17-0.19 J/cm²) was performed at 1-Hz rate and normal incidence, using a parabolic silver mirror with a 50-mm focal length. The focal radius determined by the Liu method along with the F_mod values, was wavelength-dependent in the range of 13–23 `µm (see the supporting Supplement 1 for the details). The spatially-resolved topology of the craters versus the wavelength and exposure N_p was analyzed by 405-nm laser scanning microscopy (Lext OSL 5000, Olympus) (Figs. 1–5). The periodicity of the resulting LIPSS relief structures was measured by analyzing the 2D spectral intensity patterns of their Fast Fourier Transform (FFT) spectra.

Numerical calculations were performed to identify the electrical hot-spot locations within the actual relief profiles, taken from the profilometric measurements (Figs. 1–5), using finite-element method (FEM) and the dielectric function values of photo-excited silicon in the surface plasmon resonance and ablation states [24]. These states were distinguished for the reason that in the MIR-range wavelength-dependent surface plasmon resonance, favorable for ultrafine surface nanopatterning [24], is achieved on the photoexcited Si surface at electron-hole plasma (EHP) densities ρ ∼10²⁰−10²¹ cm⁻³ (see below), which are not supporting sufficient energy deposition on the surface, once persisting over the laser pulse. The surface reliefs were introduced into the computational cells with air ambient. The illumination of the structure was provided by a linearly-polarized plane wave. A perfectly matched layer (PML) and periodic boundary conditions with the self-consistent adaptive meshes were used along the axis of wave propagation and perpendicular axes respectively.

3. Experimental results and discussion

In our studies, at different MIR fs-laser wavelengths we observed the universal physical picture (Figs. 1–5), related to exposure-dependent onset, development and topological transition of surface ripples. They ranged from deeply-subwavelength ripples (wavelength-dependent periods Λ ≈ 1 µm ≤ (0.2–0.4)λ) with their stripes oriented parallel to the laser polarization (abnormal ripples) to near-wavelength ripples (wavelength-dependent periods Λ ≈ 2–3 µm ≤ (0.7–0.8)λ) with their stripes oriented perpendicular to the laser polarization (normal ripples) [21,24–25]. The abnormal ripple periods are well consistent with the previously observed ones (λ/5) for MIR-LIPSS in semiconductors [16], while the normal ripple periods are also consistent with the previous observations [10–18]. According to the classification predicted in [21,26–27], these ripples are surface patterns [28], rather than buried patterns [17,18,29] with either λ/2n (0.3–0.6 µm for the Si refractive index n ≈ 4 at λ ≈ 2.6–4.5 µm [30]), or λ/n (0.6–1.1 µm) [31]. Other alternative mechanisms – relief self-organization [32,33], (sub)microscale capillary [12,34] or oxidative feedback [35,36] effects may also take place.

 

Fig. 1. (a) Laser scans of surface topology induced by multiple fs-laser exposure at N_p = 5 for different wavelengths λ and F₀ ≈ 1 J/cm². The double-side arrow indicates the laser polarization. (b) Wavelength dependence of periods for well-developed near-wavelength (NWR) and deeply-subwavelength (DSWR) ripples.

Download Full Size | PPT Slide | PDF

 

Fig. 2. 2D-relief map of crater on Si surface ablated by 4 - µm laser pulses at Np = 2 and F₀ ≈ 1 J/cm², its 2D-FFT spectrum and cross-sectional profiles perpendicular and parallel to the laser polarization, respectively, with the zero-amplitude (black dashed) bottom line of the ablation craters. The circle with the dot and the bilateral arrows indicate the laser polarization state.

Download Full Size | PPT Slide | PDF

 

Fig. 3. Same as in Fig. 2, at Np = 3.

Download Full Size | PPT Slide | PDF

 

Fig. 4. Same as in Fig. 2, at Np = 5

Download Full Size | PPT Slide | PDF

 

Fig. 5. Same as in Fig. 2, at Np = 10.

Download Full Size | PPT Slide | PDF

Specifically, at N_p = 2–5 the only abnormal fine ripples appear and become more pronounced in the laser confocal 2D-maps of the silicon relief and its FFT-spectra on the bright irradiated spots, resulting in ≈ 10–20 nm/shot deepening of their trenches (Figs. 2–6). One can expect that at the normal laser incidence such deeply-subwavelength ripples will appear via interference of oppositely-directed surface plasmon-polaritons (SPPs), driven on the photo-excited Si (Si*) surface in air and providing the interferential surface standing wave with the twice smaller period [15,24,28,37–39]. Specifically, in the first extreme case such SPPs will possess the minimal possible wavelength under the surface plasmon resonance (SPR) condition (i.e., when the dielectric function value ℜe[(ɛ_Si*(4 µm)] ≈ −1 for the homogeneous surface EHP densities ρ ≈ 1 × 10²¹ cm⁻³), as predicted in [24] and demonstrated in [16,24]. At the circumstances, SPPs appear as quasi-static, almost non-propagating (dω/dk ≈ 0), high-wavenumber longitudinal surface plasmons (see, e.g., SPP dispersion curves in photoexcited Si in [15,24]), yielding in deeply-subwavelength ripples on Si surface at other wavelengths too [15]. Effectively, here we consider λ/2n*-dependence for the deeply-subwavelength ripple periods, where the effective refractive index of the photoexcited Si/air interface n*=[ɛ_Si*ɛ_air/(ɛ_Si*+ ɛ_air)]^1/2 [24]. The underlying MIR fs-laser photo-excitation processes in dielectrics were recently considered in [40].

 

Fig. 6. (a) Ablation depth Z_abl inside LIPSS at different wavelengths versus Np. (b) Ablation rate (dZ_abl/dN_p).

Download Full Size | PPT Slide | PDF

Comparing to visible-NIR (400–800 nm) fs-laser patterning of Si surfaces, in the MIR-range we demonstrate deeply-subwavelength ripples (similarly to the previous MIR [16,41] and NIR (1030 nm) [24,28] studies), while in the visible-NIR range the only near-wavelength ripples were reported [24,42]. The possible reason is the SPR, as a pre-requisite, is achieved on the photoexcited flat Si surface at MIR-wavelengths at much lower absolute values of EHP densities (still near-critical ones at the given wavelengths). However, then much higher EHP densities could be produced in the plasmon-mediated interferential patterns on the seeding surface relief, making the material very lossy at high EHP densities and resulting in lower positive or even negative optical feedback (see below for deeply-subwavelength and near-wavelength ripples versus N_p). Moreover, as described in [24], transient bandgap renormalization in Si at high EHP densities could strongly increase ℑm[ɛ_Si*] via approaching E_1,2 absorption bands and EHP ohmic heating, thus damping the SPR [24], with the effect becoming more pronounced at photon energies closer to E_1,2 bands (peaks at 3.4 and 4.3 eV [30], respectively). This circumstance could explain the known trend for the wavelength-scaled ripple period to be roughly constant for near-wavelength ripples [42], but to show a decreasing trend for deeply-subwavelength ripples [41]. Thus, the advantage of MIR-range fs-laser pulses in surface patterning of semiconductors by deeply-subwavelength ripples could be justified to explain the present evidences [16,41].

However, at slightly higher, intermediate exposures N_p ≥ 5 normal coarse (near-wavelength) ripples become also distinct on the Si relief and in its FFT spectra (Figs. 1–5). This topological transformation of the relief structure indicates that the local effective optical properties of the relief, which are dependent also on its height [43,44], changed to be beneficial (stronger positive feedback) for the orthogonal normal ripples, not abnormal ones (Figs. 2–5). This transition becomes more pronounced at N_p = 10, when the normal ripples predominate on the visibly dark irradiated spot (Fig. 2–5, 8). Simultaneously, the trench depths start to increase non-linearly versus N_p, approaching 40–60 nm/shot (Figs. 2–6). One can assume that owing to much higher electrical field intensity in the interference maxima, much higher supercritical EHP density ρ ≈ 5 × 10²¹ cm⁻³ [24] is locally induced later during the laser pulse and spread around, representing a MIR highly-reflective Si state (ℜe[(ɛ_Si*(4 µm)] << −1, the other extreme case) and providing sufficient volume energy density for local ablative formation of nano-deep trenches (the mechanism will be specified below). Meanwhile, within the experimental error bars there is no any wavelength dependence of the ablation rate per pulse in Fig. 6 and this is apparently consistent with the basic optics of Si in this MIR-range (normal dispersion far from absorption bands) [30].

Our supporting FEM simulations were run at the 4-µm MIR wavelength for the experimentally acquired reliefs (Figs. 2–5, direction across the normal ripples). They indicate that the quasi-static electrical field distributions exhibit no maxima (laser coupling) on the deeply sub-wavelength surface relief of MIR highly-reflective Si at the supercritical plasma density ρ ≈ 5 × 10²¹ cm⁻³, when ɛ_Si* (4 µm) ≈ −731 + i68 [24] (Fig. 7, right column, top).

 

Fig. 7. Calculated electrical field distribution for λ = 4 µm on the deeply subwavelength abnormal ripples and near-wavelength normal ripples, using the dielectric function of silicon taken at the SPR condition with ɛ_Si* (4 µm) ≈ −1 + i0.2 and in the ablation regime with ɛ_Si* (4 µm) ≈ −731 + i68. The dotted circles and the bilateral arrows indicate the incident electrical field polarization.

Download Full Size | PPT Slide | PDF

The situation improves for higher exposures N_p ≥ 5, when considerable laser coupling to trenches occurs on the normal ripples, supporting their development via the moderate positive feedback (Fig. 2–5, 7). In contrast, for the near-resonance conditions on the Si surface at ɛ_Si*(4 µm) ≈ −1 + i0.2 considerable laser coupling to ripple trenches proceeds at lower exposures (Fig. 7, left column, top), while at higher exposures Np ≥ 5 the laser coupling becomes much stronger, but at the ridges, providing the strong negative feedback for the sub-wavelength abnormal ripples. To bring these simulation results in correspondence with our experimental observations, one should assume that at lower exposures the near-resonance conditions on the Si surface (ℜe[(ɛ_Si*(4 µm)] ≈ −1) predominate during the fine ripple development (Fig. 1), while for N_p ≥ 5 the dielectric function of the MIR highly-reflective Si state with supercritical EHP density ρ ≈ 5 × 10²¹ cm⁻³ (ℜe[(ɛ_Si*(4 µm)] << −1) manage the coarse ripple development.

Such weak and strong coupling of MIR fs-laser to the different surface relief topographies – shallow abnormal and deep normal ripples – could be illustrated both experimentally and theoretically (Fig. 8). First, the white-light optical microscopy images of the irradiated spots appear bright at lower exposures, becoming darker at higher ones (Fig. 8(a)) due to lower specular reflection and potentially stronger absorbance/scattering. This is consistent with the experimentally measured cumulative drop of the specular reflectance of the 550-and 800-nm fs-laser probe pulses right after the 4-µm pump pulse (Fig. 8(b)). Similarly, our FEM simulations also indicated the exposure-dependent significant monotonic reduction of far-field MIR reflectance on the experimentally acquired reliefs (Fig. 8(b)), if the laser exposure starts at the near-resonance conditions (ℜe[(ɛ_Si*(4 µm)] ≈ −1), but not for initial MIR highly-reflective Si state with supercritical plasma density ρ ≈ 5 × 10²¹ cm⁻³ (ℜe[(ɛ_Si*(4 µm)] << −1). Hence, despite the ultrafast strong changes of EHP density during the MIR fs-laser pump pulses, it is the initial near-resonance condition on the Si surface ℜe[(ɛ_Si*(4 µm)] ≈ −1 at ρ ≈ 1 × 10²¹ cm⁻³ that earlier during the laser pulse supports the initial relief development versus N_p via generation of the shallow deeply sub-wavelength abnormal ripples, latter versus N_p accompanied by the much deeper near-wavelength normal ripples, fed via their enhancing the MIR laser coupling to the relief.

 

Fig. 8. (a): Optical micro-graphs of Si versus N_p with the double-side arrows indicating the laser polarization. (b): Experimental 550-nm and 800-nm probe reflectance at 4-µm pump pulse end and 4-µm probe reflectance calculated for the SPR and ablation regime and the corresponding relief cross-sections in Figs. 2–5.

Download Full Size | PPT Slide | PDF

To the end, we will discuss the exposure-dependent LIPSS formation mechanisms in their final, mass transfer stage in the interplay with the simultaneous homogeneous ablation of Si surface in the micro-crater form (Figs. 2–5). The cross-sectional laser-scanning profilometry of the fine ripples (low exposures N_p < 5) indicates (Figs. 2–5) that these structures emerge within the crater (N_p = 2) near its edges and develop to the center via non-ablative (probably, capillary [11]) mass transfer, as shown by the symmetric oscillations of the relief regarding their corresponding crater bottom line (N_p = 3, less distinctly – for N_p = 5). These fine ripples are reproduced from one laser shot to another shot despite the overall surface receding via spallation by 10–20 nm/shot (Fig. 6). This apparently occurs as a result of strong and rather deep imprinting of their underlying surface-plasmon interference pattern on the crater bottom during each incident MIR-fs pump pulse, which excites EHP in Si in the maxima to the ablation level ρ ≈ 5 × 10²¹ cm⁻³ [24] via the intermediate homogeneous SPR value ρ ≈ 1 × 10²¹ cm⁻³ during the laser pulses.

Similarly, cross-sectional profilometry of the normal ripples demonstrates that at low exposures these structures emerge within the spallation craters with the quite symmetrical ripple relief regarding the crater bottom line and rather homogeneous positive feedback shown by their even amplitudes, i.e., in non-ablative manner (hydrodynamic melt displacement toward ripple ridges [11]) despite the flat and shallow homogeneous spallative ablation. However, for higher exposures (N_p ≥ 5) the crater bottom shifts down much – four- or five-fold – faster (Figs. 2–5, 6), with ripples occupying the entire crater [45], but developing very inhomogeneously and without unique bottom line. Apparently, the achieved stronger laser coupling within the ripples drives now phase-explosion ablation with its much higher ablation rates per pulse (>> 10 nm/shot) [5–8]. Meanwhile, the coarse ripples are reproduced again from the shot to shot, apparently, mostly due to in-crater spatial interference of low-wavenumber, well-propagating surface plasmon-polaritons, running from the residual ripples at the crater edges, rather than due to rather weak imprinting of the underlying interference pattern on the crater bottom during each MIR fs-laser pump pulse.

4. Conclusion

In conclusion, during multi-shot exposure of silicon surfaces in air by mid-IR (2.5–5 microns) femtosecond laser pulses a distinct topological transition between two different periodical relief patterns was revealed on a (sub)micron scale. The first pattern, shallow bipolar parallel deeply-subwavelength ripples, appearing at lower surface exposures and fluences, was for the first time firmly identified as non-ablative melt displacement feature, induced apparently by interfering deeply-subwavelength surface plasmons. The second pattern, near-wavelength perpendicular ripples with ultra-deep trenches, replacing the first pattern at higher fluences during prolonged exposure, was related to a drastic increase in optical feedback-induced laser coupling to the developing relief and in phase-explosion ablation rate per pulse.