Multiscale Hierarchical Micro/Nanostructures Created by Femtosecond Laser Ablation in Liquids for Polarization-Dependent Broadband Antireflection

In this work, we present the possibility of producing multiscale hierarchical micro/nanostructures by the femtosecond laser ablation of transition metals (i.e., Ta and W) in water and investigate their polarization-dependent reflectance. The hierarchical micro/nanostructures are composed of microscale-grooved, mountain-like and pit-rich structures decorated with hybrid laser-induced periodic surface structures (LIPSSs). The hybrid LIPSSs consist of low/high and ultrahigh spatial frequency LIPSSs (LSFLs/HSFLs and UHSFLs). LSFLs/HSFLs of 400–600 nm in a period are typically oriented perpendicular to the direction of the laser polarization, while UHSFLs (widths: 10–20 nm and periods: 30–50 nm) are oriented perpendicular to the curvatures of LSFLs/HSFLs. On the microstructures with height gradients, the orientations of LSFLs/HSFLs are misaligned by 18°. On the ablated W metasurface, two kinds of UHSFLs are observed. UHSFLs become parallel nanowires in the deep troughs of LSFLs/HSFLs but result in being very chaotic in shallow LSFLs, turning into polygonal nanonetworks. In contrast, chaotic USFLs are not found on the ablated Ta metasurfaces. With the help of Fourier transform infrared spectroscopy, it is found that microgrooves show an obvious polarization-dependent reflectance at wavelengths of 15 and 17.5 μm associated with the direction of the groove, and the integration of microstructures with LSFs/HSFLs/UHSFLs is thus beneficial for enhancing the light absorbance and light trapping in the near-to-mid-infrared (NIR-MIR) range.


Introduction
Femtosecond laser ablation (fs-LA) is a versatile technique that enables the production of a large variety of surface structures [1][2][3][4][5], and the structures' diversity can be further enriched in combination with other techniques [6,7]. Laser-induced periodic surface structures (LIPSSs) [1,[8][9][10][11][12][13] are the most typical nanoscale structures that are uniquely achievable by fs-LA, whose periods can be manipulated from tens of nm to hundreds of nm by changing the laser properties [14], processing parameters [8] and ablation environments [15]. Generally, LIPSSs are categorized into low and high spatial frequency LIPSSs (LSFL/HSFL) according to the ratio of LIPSS periods (Λ) to the fs laser wavelength (λ) [8]. The periods of LSFLs and HSFLs are defined as ranging from about the wavelength to half of the laser wavelength (λ/2 ≤ Λ LSFL ≤ λ) and less than half of the wavelength (Λ HSFL < λ/2), respectively [8,16]. Recently, our group has shown the necessity to define sub-100 nm [17][18][19][20] periods as a new category Figure 1a displays the typical microstructures obtained by fs-LA of Ta in water. The microstructures are not uniform, consisting of microholes, curved troughs and mountain-like bump structures. The formation of curved troughs is attributed to the refraction and reflection by the bubbles generated during fs-LAL [17]. The diameter of microholes ranges from 2 to 5 µm. Enlarged images (Figure 1b,c) indicate that LIPSS nanostructures with periods of 400-600 nm are located on all microstructures. Although LIPSSs observed on the microstructures are categorized into the mixture of LSFLs/HSFLs (according to the definition of LSFL (λ/2 ≤ Λ LSFL ≤ λ) and HSFL (Λ HSFL < λ/2), together with the laser wavelength of 1045 nm used for fs-LAL), the formation mechanism should be the same due to the small difference in the period. The orientations of LSFLs/HSFLs are perpendicular to the direction of light polarization. Bifurcations of LSFLs can be obviously seen on the microstructures (Figure 1c). The depths of bifurcated LSFLs/HSFLs are much shallower than other parts (Figure 1d). A large amount of UHSFLs of 30-50 nm per period are located in the troughs and joint regions of LSFLs/HSFLs. The periods of UHSFLs are much narrower than those prepared on Ti by fs-LA in air [27][28][29]. In a relatively deeper trough of LSFLs/HSFLs, UHSFLs are characterized by extra stretched nanowires from their middle parts (pointed out by green arrows in Figure 1e), so that it is difficult to differentiate them from main UHSFLs (Figure 1e). In contrast, UHSFLs at the shallow region are much easier to identify, and some UHSFLs are decorated with nanodots (pointed out by green arrows in Figure 1f). Left-bent, right-bent and zigzag UHSFLs (pointed out by pink arrows) are all observed (Figure 1f), indicating that an extra turbulent factor is triggered during the formation of UHSFLs. Figure 1g demonstrates the possibility of producing microgrooves decorated with both tilted and horizontal LSFLs/HSFLs (magnified images are also shown in Figure 1h,i). The orientation of LSFLs/HSFLs pointed out by a green arrow is perpendicular to the direction of light polarization (Figure 1h), while the tilted LSFLs/HSFLs deviate by 18 • from the microgroove direction, as expected (Figure 1i), resembling the herringbone structures on copper surfaces obtained by s-polarized fs laser ablation at a large incident angle [45]. Schwarz et al. reported that an inclined fs laser can significantly alter the orientations of LSFLs on fused silica [46]. Zheng et al. also proposed that the inclined angle at the boundary of a Gaussian beam can lead to the formation of slantwise-oriented LIPSSs [47]. Hence, the formation of herringbone structures shown in Figure 1i is deemed to be caused by an inclined fs laser ablation arising from bubble reflection/refraction [17]. A higher magnification of LSFLs/HSFLs structures allows for the observation of the states of UHSFLs on the side walls of LSFLs/HSFLs, as shown in Figure 1j-l. The orientations of UHSFLs depend on the LSFLs/HSFLs where they are located. At the tip of LSFLs/HSFLs where LSFLs/HSFLs change their curvature, UHSFLs can be even aligned from left to right (indicated by a green arrow in Figure 1k). On the ridge of LSFLs/HSFLs, UHSFLs are always aligned perpendicular to the ridge of LSFLs (Figure 1l) along the ridge's inclination. This finding is in accordance with our previous report, which showed that USHFLs are not parallel to the direction of light polarization but strongly related to the LSFLs/HSFLs curvatures [18]. The simultaneous formation of LSFLs/HSFLs may originate from the excitation and interference of counter-propagating surface plasmons with extreme wavenumbers during fs-LAL [48], while the formation of UHSFLs is attributed to Marangoni bursting during the formation of LSFLs/HSFLs molten layers [18].  Figure 2g). In the yellow region of Figure 2g, UHSFLs with periods of 30-50 nm are parallel to each other, and their orientations are not parallel to the direction of light polarization but perpendicular to the curvatures of the underlying LSFLs/HSFLs (Figure 2i). In the regions where LSFLs/HSFLs have both deep troughs and inclined ridges (Figure 2j), UHSFLs are ordered in different ways. UHSFLs are highly ordered in the deep region of LSFLs/HSFLs, while they become very disordered on the outermost inclined ridges (indicated by white arrows in Figure 2k,l). This means that some factors that can arouse the disturbance of UHSFLs are triggered during the solidification of UHSFLs but that such turbulent factors have a limited influence on the deep troughs of LSFLs/HSFLs. The turbulent effect generated during fs-LAL of W is much stronger than that generated during fs-LAL of Ta. Similarly, Zhao et al. observed the formation of chaotic nanospikes (diameter 10-100 nm, length up to 250 nm) between LSFLs after the single-pulse fs laser irradiation of W in air [49]. Although the mechanism is still unclear and deserves further investigations, we speculate the following: W is a material with a  Figure 2g). In the yellow region of Figure 2g, UHSFLs with periods of 30-50 nm are parallel to each other, and their orientations are not parallel to the direction of light polarization but perpendicular to the curvatures of the underlying LSFLs/HSFLs (Figure 2i). In the regions where LSFLs/HSFLs have both deep troughs and inclined ridges (Figure 2j), UHSFLs are ordered in different ways. UHSFLs are highly ordered in the deep region of LSFLs/HSFLs, while they become very disordered on the outermost inclined ridges (indicated by white arrows in Figure 2k,l). This means that some factors that can arouse the disturbance of UHSFLs are triggered during the solidification of UHSFLs but that such turbulent factors have a limited influence on the deep troughs of LSFLs/HSFLs. The turbulent effect generated during fs-LAL of W is much stronger than that generated during fs-LAL of Ta. Similarly, Zhao et al. observed the formation of chaotic nanospikes (diameter 10-100 nm, length up to 250 nm) between LSFLs after the single-pulse fs laser irradiation of W in air [49]. Although the mechanism is still unclear and deserves further investigations, Nanomaterials 2020, 10, 1573 5 of 15 we speculate the following: W is a material with a high imaginary part of permittivity (ε": 32.58 at 1060 nm [50]) that enables a very high absorbance of infrared light. Considering that an infrared pulse laser (wavelength: 1045 nm) was used for fs-LAL, the as-prepared surface structures effectively absorbed the beam energy, which caused an increase in the liquid temperature near the structures and may also have aroused additional magnetic and electrical effects near the surface structures [50]. In consequence, the orientations of UHSFLs were greatly disturbed during their solidification.
Nanomaterials 2020, 10, x FOR PEER REVIEW 5 of 14 high imaginary part of permittivity (ε'': 32.58 at 1060 nm [50]) that enables a very high absorbance of infrared light. Considering that an infrared pulse laser (wavelength: 1045 nm) was used for fs-LAL, the as-prepared surface structures effectively absorbed the beam energy, which caused an increase in the liquid temperature near the structures and may also have aroused additional magnetic and electrical effects near the surface structures [50]. In consequence, the orientations of UHSFLs were greatly disturbed during their solidification. To compare the difference between our results and previous reports, LIPSSs obtained on Ta and W by fs-LA in both air and liquids are summarized in Table 1. In our previous report, we showed that, under the same ablation condition, the periods of LIPSSs (including both LSFLs/HSFLs and UHSFLs) obtained by fs-LAL in acetone [18] were almost the same as those produced by fs-LA in water shown in this work. Studies on Ta-LIPSS are very limited. Barmina et al. obtained LSFL/UHSL by fs laser ablation in water, but the SEM images of structures shown in their work were captured from an inclined angle, making it very hard to identify the periods of LSFLs [51]. Jorge-Mora et al. produced LSFLs on Ta surfaces by fs-LA in air with LIPSS periods which were slightly smaller than the laser wavelength (λ = 1030 nm vs. Λ = 780 nm) [52] but much larger than the periods (370-600 nm) of Ta-LIPSSs obtained in liquids (this work and ref. [18]). This means that liquid is beneficial for achieving narrower periods of LIPSSs, a trend that is also confirmed by the summary of the periods of Si-LIPSSs obtained in different environments [15]. Recently, Kudryashov et al. proposed that the factors of the squared optical refraction index (n) of environments can be used to estimate the environment-dependent periods (Λ) of LIPSSs following the trend of Λ ∝ 1/n 2 , which indicated that liquids with a higher refraction index than air normally produced smaller LIPSS periods [48]. The refractive indices of air, water, ethanol and acetone are 1, To compare the difference between our results and previous reports, LIPSSs obtained on Ta and W by fs-LA in both air and liquids are summarized in Table 1. In our previous report, we showed that, under the same ablation condition, the periods of LIPSSs (including both LSFLs/HSFLs and UHSFLs) obtained by fs-LAL in acetone [18] were almost the same as those produced by fs-LA in water shown in this work. Studies on Ta-LIPSS are very limited. Barmina et al. obtained LSFL/UHSL by fs laser ablation in water, but the SEM images of structures shown in their work were captured from an inclined angle, making it very hard to identify the periods of LSFLs [51]. Jorge-Mora et al. produced LSFLs on Ta surfaces by fs-LA in air with LIPSS periods which were slightly smaller than the laser wavelength (λ = 1030 nm vs. Λ = 780 nm) [52] but much larger than the periods (370-600 nm) of Ta-LIPSSs obtained in liquids (this work and ref. [18]). This means that liquid is beneficial for achieving narrower periods of LIPSSs, a trend that is also confirmed by the summary of the periods of Si-LIPSSs obtained in different environments [15]. Recently, Kudryashov et al. proposed that the factors of the squared optical refraction index (n) of environments can be used to estimate the environment-dependent periods (Λ) of LIPSSs following the trend of Λ ∝ 1/n 2 , which indicated that liquids with a higher refraction index than air normally produced smaller LIPSS periods [48]. The refractive indices of air, water, ethanol and acetone are 1, 1.33, 1.36 and 1.36, respectively, which would give ratios of LIPSS periods in air/water/ethanol/acetone (Λ air /Λ water /Λ ethanol /Λ acetone ) of 1/1.33 2 /1.36 2 /1.36 2 . According to the fact that 780 nm is the period of Ta-LIPSS obtained by fs-LA in air at λ = 1030 nm, the periods of Ta-LIPSSs obtained in water and ethanol/acetone at the wavelength of 1030 nm should be 441 and 421 nm, respectively. This calculation fits well with the period range of 400-600 nm experimentally obtained by fs-LA in water and acetone at a close wavelength of 1045 nm, as presented in this work and our previous work [18]. Hence, it can be concluded that the theoretical trend of Λ ∝ 1/n 2 proposed by Kudryashov et al. [48] is at least applicable to some cases. Compared to Ta-LIPSSs, W-LIPSSs have been studied much more. Roughly, the LIPSS periods obtained by fs-LAL in liquids such as ethanol and water are smaller than those obtained in air (Table 1). However, varying laser parameters such as pulse duration, wavelength, repetition rates and laser fluences can significantly change the periods of W-LIPSSs in an air environment. Some of W-LIPSSs' periods are within the period range of W-LIPSSs obtained in liquids. The most noteworthy involves using an fs laser with an ultrahigh repetition rate of up to 80 MHz: the periods of W-LIPSSs are as small as 150-185 nm [53], equal to half of the W-HSFL periods obtained in liquids [18,[54][55][56]. Similarly, LAL of Si in oil at a repetition rate of 90 MHz allows for the formation of abnormal LIPSSs with sub-100 nm periods [21]. The reason for this phenomenon is unknown, and it deserves a much deeper investigation. However, it is not within the scope of the present work.  (Figure 3a,e). The maximal depths of the grooves are~30 µm, as indicated by the cross-sectional profiles shown in Figure 3b,f. The heights of the microstructures fluctuate heavily, indicating the randomness of the microstructures. In contrast, a crater-rich region is observed on the W surface (Figure 3c). The depths of the microcraters are very shallow, less than 5 µm. Some inhomogeneous grooves with a maximal depth of 23.8 µm are located near the crater-rich region (Figure 3c). The observation of another ablated region, shown in Figure 3g, indicates that the ablation depth on the W surface can still reach 30 µm (the highest region indicated by red color in Figure 3g corresponds to the unablated region). On the W surface, fs-LAL also induced the formation of different groove widths (Figure 3g,h). Nanomaterials 2020, 10, x FOR PEER REVIEW 7 of 14 of the grooves are ~30 μm, as indicated by the cross-sectional profiles shown in Figure 3b,f. The heights of the microstructures fluctuate heavily, indicating the randomness of the microstructures. In contrast, a crater-rich region is observed on the W surface (Figure 3c). The depths of the microcraters are very shallow, less than 5 μm. Some inhomogeneous grooves with a maximal depth of 23.8 μm are located near the crater-rich region (Figure 3c). The observation of another ablated region, shown in Figure 3g, indicates that the ablation depth on the W surface can still reach 30 μm (the highest region indicated by red color in Figure 3g corresponds to the unablated region). On the W surface, fs-LAL also induced the formation of different groove widths (Figure 3g,h).  Figure 4 displays the EDS spectra of structures from the orange rectangles in each inset SEM image obtained by fs-LAL of Ta and W in water. The atomic percentages of metal, carbon and oxygen were tested for two different regions of the hierarchical micro/nanostructures of both Ta and W samples, as shown in Table 2, while also being tested for unablated metal substrates. The atomic percentages of Ta, C and O are 48.79%, 26.34% and 24.87% for the first region of the Ta-structure (Figure 4a, inset image) and 46.40%, 25.39% and 28.22% for the second region (Figure 4b),  Figure 4 displays the EDS spectra of structures from the orange rectangles in each inset SEM image obtained by fs-LAL of Ta and W in water. The atomic percentages of metal, carbon and oxygen were tested for two different regions of the hierarchical micro/nanostructures of both Ta and W samples, as shown in Table 2, while also being tested for unablated metal substrates. The atomic percentages of Ta, C and O are 48.79%, 26.34% and 24.87% for the first region of the Ta-structure (Figure 4a, inset image) and 46.40%, 25.39% and 28.22% for the second region (Figure 4b), respectively; and the   Despite a large difference in carbon percentages between the unablated Ta and W targets, the carbon percentages are almost the same for different structures on all ablated Ta and W samples. However, the oxidation rates are distinctly different for the Ta and W hierarchical structures and are speculated to be associated with different thermal properties of Ta and W, as listed in Table 3. Since it is well known that the formation of UHSFLs is related to the material melting phenomenon during fs-LA [18,31], the melting periods of the ablated materials must strongly depend on the thermal properties of the ablated materials. Both W and Ta have almost the same specific heat of 0.13 and 0.14 J/g K [65]. However, other thermal properties are very different. The thermal diffusivity of ~70 mm 2 /s for W [66] is much higher than that of 24.2 mm 2 /s for Ta [67] at room temperature, and both decrease with an increasing temperature [66]. The thermal conductivities of  Despite a large difference in carbon percentages between the unablated Ta and W targets, the carbon percentages are almost the same for different structures on all ablated Ta and W samples. However, the oxidation rates are distinctly different for the Ta and W hierarchical structures and are speculated to be associated with different thermal properties of Ta and W, as listed in Table 3. Since it is well known that the formation of UHSFLs is related to the material melting phenomenon during fs-LA [18,31], the melting periods of the ablated materials must strongly depend on the thermal properties of the ablated materials. Both W and Ta have almost the same specific heat of 0.13 and 0.14 J/g K [65]. However, other thermal properties are very different. The thermal diffusivity of~70 mm 2 /s for W [66] is much higher than that of 24.2 mm 2 /s for Ta [67] at room temperature, and both decrease with an increasing temperature [66]. The thermal conductivities of W and Ta are thus calculated to be 170 and 57 W/m K, respectively. As can be seen, both the thermal diffusivity and thermal conductivity of W are much higher than those of Ta, indicating that W should tend to cool more rapidly than Ta after laser heating. Additionally, the melting temperature of 3422 • C for W is relatively higher than that of 3017 • C for Ta [65,68]. For Ta, the smaller thermal conductivity and thermal diffusivity with a lower melting temperature must endow a relatively longer melting period than W so as to enable the interaction with surrounding oxygen species during fs-LAL. The states of laser pulses such as repetition rates, pulse energy and incident angle may all be modulated by random bubble reflection/refraction [17], so that the thermal states in different positions change a lot. Meanwhile, the shorter periods of the thermal effect of W during fs-LAL may be attributed to a great discrepancy between the oxidation rates at different regions (Figure 4c,d). Nevertheless, despite different oxidation rates in different positions, when compared with the oxygen percentages of the unablated W samples, fs-LAL in water is indeed capable of enhancing the oxidation of the ablated surfaces.

Polarization-Dependent Reflectance
The optical properties of the structures strongly depend on the morphology of the structured surfaces. Anisotropic structures such as parallel grooves [69] and L-shape structures [70] generally possess polarization-dependent reflectance. In this regard, two different regions of the multiscale hierarchical micro/nanostructures of the ablated Ta and W samples were measured for the polarization-dependent reflectance study. Figure 5a,b and Figure 5c,d show the reflective spectra in two different regions on the ablated Ta and W targets in the wavelength range of 1.25-25 µm, respectively. The red curves were measured with the probe light whose polarization direction was perpendicular to the laser scanning direction for fs-LA and was defined as Y-reflectance (as indicated in Figure 5b), while the blue curves were parallel to the horizontal direction (X-reflectance).
Nanomaterials 2020, 10, x FOR PEER REVIEW 9 of 14 W and Ta are thus calculated to be 170 and 57 W/m K, respectively. As can be seen, both the thermal diffusivity and thermal conductivity of W are much higher than those of Ta, indicating that W should tend to cool more rapidly than Ta after laser heating. Additionally, the melting temperature of 3422 °C for W is relatively higher than that of 3017 °C for Ta [65,68]. For Ta, the smaller thermal conductivity and thermal diffusivity with a lower melting temperature must endow a relatively longer melting period than W so as to enable the interaction with surrounding oxygen species during fs-LAL. The states of laser pulses such as repetition rates, pulse energy and incident angle may all be modulated by random bubble reflection/refraction [17], so that the thermal states in different positions change a lot. Meanwhile, the shorter periods of the thermal effect of W during fs-LAL may be attributed to a great discrepancy between the oxidation rates at different regions (Figure 4c,d). Nevertheless, despite different oxidation rates in different positions, when compared with the oxygen percentages of the unablated W samples, fs-LAL in water is indeed capable of enhancing the oxidation of the ablated surfaces.

Polarization-Dependent Reflectance
The optical properties of the structures strongly depend on the morphology of the structured surfaces. Anisotropic structures such as parallel grooves [69] and L-shape structures [70] generally possess polarization-dependent reflectance. In this regard, two different regions of the multiscale hierarchical micro/nanostructures of the ablated Ta and W samples were measured for the polarization-dependent reflectance study. Figure 5a,b and Figure 5c,d show the reflective spectra in two different regions on the ablated Ta and W targets in the wavelength range of 1.25-25 μm, respectively. The red curves were measured with the probe light whose polarization direction was perpendicular to the laser scanning direction for fs-LA and was defined as Y-reflectance (as indicated in Figure 5b), while the blue curves were parallel to the horizontal direction (X-reflectance).  Compared to the reflectance, reported elsewhere, of flat hierarchical LSFL/UHSFLs structures obtained by fs-LA in acetone [18], the reflectances in both X-and Y-directions decrease by 60~70% for both the Ta and W samples. This indicates that the endowment of LSFL/UHSFLs hybrid structures on microstructures can significantly enhance the light absorbance and light trapping in the NIR-MIR range, especially in the MIR range. Each X-and Y-reflectance of the Ta measure at two different positions in the wavelength range of 1.25-20 µm shows almost the same characteristics, although they have different reflectance intensities, indicating that the as-prepared Ta structures are similar but with different structure heights (Figure 5a,b). Many small X-and Y-reflection peaks appear in the range of 15-20 µm for Ta, which are ascribed to the existence of LSFLs/HSFLs/UHSFLs [18]. The spectra fluctuations in the wavelength of 20-25 µm of both two Ta samples may be due to inhomogeneous microspikes, as shown in Figure 4a,b. Figure 5c,d displays variations of the polarization-dependent reflectance on the relatively flat hierarchical structure and the grooved hierarchical micro/nanostructure of W in the wavelength range of 1.25-25 µm, respectively. The reflectance of both samples is almost the same in the range of 1.25-5 µm (Figure 5c,d), with the exception of some small peaks appearing on the grooved sample. The small peaks at~3.33 µm (as pointed out by a green arrow in Figure 5d) are attributed to the CH stretching from the alkyl groups [71,72]. Such CH stretching mode peaks do not appear on the relatively flat structure (Figure 5c), indicating the structure-dependent surface chemistry induced by fs-LAL. As the wavelength increases from 5 to 20 µm, the X-and Y-reflectances of the relatively flat hierarchical structures look almost same, and both gradually decrease. Broad peaks appear at 14 µm in the X-reflectance and at 18 µm in the Y-reflectance, but the maximal value of each of these two peaks is only slightly higher than the reflectance of the other direction, which is ascribed to the flatness of the micro/nanostructures. Unlike this relatively flat structure, due to the existence of aligned grooves (Figure 5d), the reflectance of the grooved hierarchical micro/nanostructure shows basin-like curves, with new peaks appearing at 15 µm for the X-reflectance and 17.5 µm for the Y-reflectance. Due to the structure difference in the two samples, the polarization-dependent reflectance in the wavelength range of 17.5-25 µm is also different (Figure 5c,d). According to the theory (λ/4 = period of trenches) that the reflectance minima appear at λ = 20 µm in FTIR spectra (Figure 5c,d) [12,13], the trench of the W structures is estimated to be 5 µm in width. From these results, it can be concluded that anisotropic hierarchical micro/nanostructures prepared by fs-LAL can be used to investigate structure-dependent reflectance, a topic that, to date, has been seldom studied in relation to laser-structured surfaces. From the reflectance of the two regions of both the Ta and W structures, it can be concluded that fs-LAL generates more homogenous structures on Ta than on W, which in turn can help explain the huge difference in the region-dependent W structural morphologies shown in Figure 3c,g and oxidation rates shown in Figure 4c,d.
The EDS analysis (Figure 4) indicates that the ablated surfaces may not be fully oxidized. Moreover, the formation of LIPSSs, including all LSFLs/HSFLs/UHSFLs, is confirmed as resulting from the melting behavior of metal materials [31]. The hierarchical structures should be a mixture of both metal and metal oxides. Since metal oxide materials, including both tantalum oxide and tungsten oxide, are normally insulators [73,74], one should consider composing the as-prepared hierarchical structures out of a lot of metal-insulator-metal (MIM) structures [75], which are good NIR-MIR absorbers due to the fact that their electronic and magnetic resonances occur synchronously at a certain wavelength. These MIM absorbers could contribute to reflectance variations at different wavelengths for different structures on both Ta and W surfaces.

Conclusions
This work presented the feasibility of developing a hierarchical micro/nanostructure by fs-LAL of Ta and W in water. Due to random light reflection and refraction induced by bubbles, random microstructures, including microgrooves, microcraters and mountain-like microstructures, were formed. All microstructures were fully decorated with hybrid LSFL/HSFL/UHSFLs nanostructures. The periods of LSFLs/HSFLs were in the range of 500-800 nm, and UHSFLs had periods of 30-50 nm. As compared to UHSFL structures obtained on Ta, which had a very limited structure disorder, those prepared on the W surface were very chaotic, especially in the shallow troughs of LSFLs/HSFLs and their outermost ridges, which may be associated with the difference in thermal properties of Ta and W, including the melting temperature, thermal diffusivity and thermal conductivity. Using two different micro/nanostructures on the Ta and W surface as representatives, it was proven that the introduction of a microgroove rendered the polarization-dependent reflectance at wavelengths of 15 and 17.5 µm along the X-and Y-directions, respectively. Ta and W multiscale micro/nanostructures presented in this work possessed lower polarization-dependent reflectances than those of LSFLs/UHSFLs hybrid nanostructures prepared by fs-LAL in acetone [18], which one may ascribe to the presence of microstructures. Integrating hybrid LSFLs/HSFLs/UHSFLs with microstructures into multiscale hierarchical micro/nanostructures can significantly enhance the high absorbance and light trapping in the NIR-MIR range, especially in the MIR range. This work may inspire the preparation of polarization-dependent reflective metasurfaces by fs-LAL, along with studies on the corresponding structure-dependent reflectance.