1. Introduction

Femtosecond laser is a reliable tool for fabricating surface structures on metals, semiconductors and insulators. It draws people’s great interests because of the flexibility to create versatile surface structures especially in the micro/nanometer scale range [1–5]. These structures have shown huge potential in applications include enhancing the responsivity of detectors [6, 7], colorizing metals [8–13], tailoring tribological properties of the surfaces [14], and etc. Moreover, surface structure fabrication serves as a platform to study the laser-matter interaction because of the rich physics involved in the process [4, 15–19]. The most investigated structures are laser induced periodic surface structures (LIPSS), which has been an active research topic for over half a century since Birnbaum reported them in 1965 for the first time [2, 20]. The formation of low spatial frequency LIPSS (LSFL, also mentioned as ripples), which have a periodicity close to the laser wavelength and direction perpendicular to the laser polarization, can be well explained by the framework named Sipe-Drude model proposed by J. Sipe and J. Bonse [15, 21]. The non-uniform energy deposition caused by the interference between the incident laser and the scattered surface electromagnetic wave is considered the origin of LSFL formation. The excitation of surface plasma polaritons (SPPs) and transient change of the dielectric function of the material upon irradiation of femtosecond laser are also been included in this model [16, 22]. The micrometer size grooves, which are parallel to the polarization of the light, attract people’s attention recently [18, 23–28]. A result based on finite-difference time-domain (FDTD) method shows that not only the LSFL, but also the grooves and the high spatial frequency LIPSS (HSFL) can be understood in the framework of the electromagnetic approach [18, 27, 28]. Besides, a hydrodynamics-based scenario has also been suggested by G. D. Tsibidis to explain the mechanism of the surface structure formation [19, 25].

Based on these researches, the fluence, pulse numbers, polarization, wavelength of the laser, pulse sequences and the ambient environment have been used to control the morphology of the surface irradiated by femtosecond laser [1,2]. Previously we showed that substrate temperature could be used to tailor the properties of the surface structures and enhance the crystallinity of these structures [29, 30]. Other researches also showed that elevated substrate temperature could reduce the roughness in femtosecond laser processing and increase the drilling efficacy [31–34]. These results suggest that substrate temperature is an additional important parameter for studying the rich physics involved in the fabrication process and is a new way to control the surface morphology. Here we give the comparison of the surface morphology created by femtosecond laser and its evolution at a higher substrate temperature and at room temperature in detail. A combination of two-temperature model and FDTD method are used to discuss the temperature dependence of surface structures.

2. Experimental setup

The experiments were performed on an n-type (100) silicon wafer. Acetone and methanol were used to clean the wafer surface in an ultrasonic bath for 5 minutes prior to the laser irradiation. Then the wafer was rinsed with distilled water and blew dried with nitrogen. The temperature of the silicon wafer was monitored by a thermal imager (NEC H2640) and controlled by an electric heater up to 700 K. A linear polarized femtosecond Ti: sapphire laser operated at 10 Hz with a central wavelength of 800 nm was used to irradiate the silicon wafer at normal incidence in air. The femtosecond laser had pulse energy up to 3.6 mJ and pulse duration around 50 fs. The spot of the laser beam, analyzed by a CCD beam analyzer, had a Gaussian intensity distribution. It was focused with a 0.2 m focal length lens to a diameter of 196 μm (as measured at 1/e² of the maximum intensity) on the sample surface. The energy was fixed at 30 μJ in all the experiments presented here. The fluence in the center of the beam is 2 kJ/m². The pulse number was controlled by a mechanical shutter (Thorlabs, SH1). The irradiation at room temperature was carried out first before the one at 700 K. Then the silicon wafer was cooled down to room temperature in air. The morphology of the irradiated samples was analyzed by using a scanning electron microscope (JSM-7500F).

3. Results and Discussions

The electron micrographs of the surface structures on silicon irradiated with 4 femtosecond laser pulses at 300 K and 700 K are presented in Figs. 1(a)-1(b), respectively. Traditional LSFL, which are perpendicular to the laser polarization, are observed in both cases. The insets in Figs. 1(a) and 1(b) are the Fourier transformations of Figs. 1(a) and 1(b) respectively. Here we follow the notations used in [18] to describe the features present in the frequency domain. The black circles correspond to the position with a normalized wave vector equal to one. Two sickle like red spots, referred to as type-s features indicating the formation of LSFL, are observed in these two insets along the direction of the laser polarization. The position of these spots are very close to but outside the circle, implying that the periodicities are slightly smaller than the laser wavelength. The red elliptical spot in the center, named the type-g feature, represents the micrometer-size structures formation in Fig. 1(b). The longer axis of the type-g spot is perpendicular to the laser polarization, indicating that the micrometer-size structures are parallel to the laser polarization [28], which are the same as traditional micrometer-size grooves. The modulation of surface structures parallel to the laser polarization is also clearly observed in the space domain shown in Fig. 1(b). Figs. 1(c) and 1(d) show the LSFL in Fig. 1(a) and 1(b) at higher magnification, respectively. One major difference we observed is that more nanoparticles are present in Fig. 1(c), which are the surface structures formed at 300 K. The average size of the nanoparticles in Fig. 1(c) are much bigger than those in Fig. 1(d). Moreover, the LSFL ripples formed at 700 K [Fig. 1(d)] are regular while the width of the LSFL ripples formed at 300 K is not uniform. This result indicates that the surface roughness of the ripples is greatly reduced at higher substrate temperature. We also notice that at 300 K the entire structured area is covered by ripples [Fig. 1(a)] while the center of it is flat [Fig. 1(b)].

 

Fig. 1 Scanning electron micrographs of structures formed with 4 laser shots at: (a) and (c) 300 K; (b) and (d) 700 K. The insets in (a) and (b) are the Fourier transforms of Fig. 1(a) and 1(b) respectively. The black circles correspond to the radius with normalized wave vector equal to one. The arrow indicates the direction of the laser polarization.

Download Full Size | PDF

Figs. 2(a) and 2(d) shows the electron micrographs of the surface structures formed with 50 femtosecond laser shots irradiation at 300 K and 700 K, respectively. The periphery of the structured area formed at 300 K is covered by ripples, which are more uniform compared with the one obtained with 4 laser shots [Figs. 1(a) and 1(c)], while the center is covered by random structures [Fig. 2(b) and 2(c)]. The co-existence of ripples and grooves is present at the periphery of the structured area formed at 700 K shown in Fig. 2(e). The grooves cover almost the entire structured area with a periodicity in the range of 1.5 µm to 3 µm. Well organized nanometer size holes, with an average diameter around 500 nm and average distance around 770 nm are in the valleys of the grooves shown in Fig. 2(f).

 

Fig. 2 Scanning electron micrograph of structures formed with 50 laser shots: (a) Half of the crater and its (b) edge and (c) center formed at 300 K; (d) Half of the crater and its (e) edge and (f) center formed at 700 K. The arrow indicates the direction of the laser polarization.

Download Full Size | PDF

After irradiated with 500 laser shots, the morphology changes dramatically, as shown in Fig. 3. The structures formed at 300 K are consistent with the previous published results [2, 25, 35], shown in Figs. 3(a), 3(c) and 3(d). Micrometer-size grooves and cones are the major structures while ripples form in the periphery. In contrast, grooves parallel to the polarization of laser light still dominate the structures formed at 700 K as shown in Fig. 3(b). Interestingly, micrometer-size holes are present in the valleys of the grooves. The dimension and depth of these holes are much bigger than those observed in Fig. 2. We also notice that nanoparticles cover the inner walls of the micrometer size holes.

 

Fig. 3 Scanning electron micrograph of structures formed at different temperature and magnification with 500 laser shots with a viewing angle at 30°: (a), (c) and (d) 300 K; (b), (e) and (f) 700 K. The arrow indicates the direction of the laser polarization.

Download Full Size | PDF

The local fluence changes as a function of position in the structure areas shown in Figs. 1-3 because of the Gaussian distribution of the laser light adopted. The morphology chart versus fluence is shown in Fig. 4. The fluences for LSFL ripples and grooves generation decreases as the number of the pulses increasing for both temperatures, which indicates an incubation effect. The thresholds for LSFL ripples formation at room temperature and 700 K are almost identical after irradiation with 4 laser shots. It increases at elevated temperature when the number of the pulse is 50, consists with our previous result [29]. The outer part of the structured area is covered by nanoparticles with 500 laser shots irradiation at room temperature, which make it impossible to get the accurate lowest fluence for LSFL ripples formation. The fluence marked with red dotted circle showed in Fig. 4 is the local fluence of the outmost position where the LIPSS can be clearly observed. We notice that the fluence for LIPSS generation is 1.87 kJ/m², 1.4 kJ/m², 1.3 kJ/m² for 4, 50 and 500 pluses irradiation respectively at room temperature. It’s lower than the previously reported results (2 J/m², 100fs, and 200 pulses) [5]. Because of the nonlinear absorption, the carrier densities excited by these three fluences with a pulse duration 50 fs correspond to the one excited by a fluences at 2.5 kJ/m², 1.85 kJ/m², and 1.71 kJ/m² respectively when the pulse duration is 100 fs, calculated by the two temperature model proposed in this work. Those fluences are reasonably consistent with the previously published results.

 

Fig. 4 Morphology chart versus pulse number and fluence

Download Full Size | PDF

Upon irradiation of a femtosecond laser pulse, electrons in the valence band are excited to the conduction band. The electrons achieve a very high temperature and density because of the high intensity of ultrashort femtosecond laser pulse. This process is described by the two-temperature model proposed by G. D. Tsibidis [19]. The transient change of the dielectric function of silicon upon the excitation of a femtosecond laser pulse is calculated by combining it with the Drude model [15]. In this work, we follow the parameters used in [19, 23, 26]. Moreover, the optical property of silicon is also a function of temperature [36]. More specifically, the real and imaginary part of the complex refraction index are experimentally determined to be a linear and exponential function of temperature, respectively. A seventh-order polynomial fits is used to obtain the coefficients for the refraction index calculation at 700 K. In order to study the temperature dependence of surface structures especially the formation of grooves at higher substrate temperature, 1.5 kJ/m² was chosen as the excitation fluence. This fluence corresponds to the formation of surface structures shown in Fig. 2(e), in which ripples and grooves coexist. The dielectric functions are 2.88091 + 0.7015i and 0.39907 + 1.2353i at 300 and 700 K, respectively when the free-carrier collision frequency is 1.5 × 10⁻¹⁴ s⁻¹. A significant phonon–electron scattering occurs at high substrate temperature and leads to the increasing of the collision frequency [29, 37–39]. The dielectric function is 0.62874 + 2.36119i when the collision frequency is 3.3 × 10⁻¹⁴ s⁻¹.

The fluence used in the experiments presented here is above the melting threshold and below the ablation threshold in the structured area. The modulated absorption of the laser light, which will be discussed later in this work, leads to the non-uniform melting on the surface. Surface capillary waves with a periodicity equal to the period of the light absorption modulation are generated in the melted surface layer. The surface wave becomes permanent surface structures when the liquid silicon layer re-solidifies [5]. More energy is absorbed in a thinner layer at elevated temperature because of the increased energy absorption coefficient. At 700 K substrate temperature, the liquid silicon can be heated up to a higher temperature. Because of the Gaussian intensity distribution fact of the beam, the melting duration is extended to longer than the life time of the capillary waves in the center part of the light first at 700 K. The melted liquid silicon re-solidifies as a flat surface in this case as shown in Fig. 1(b). Meanwhile, nanoparticles melt easier at 700 K compared to those at room temperature, leaving a surface with decreased roughness, as observed in Fig. 1(d).

The modulated energy deposition is crucial for the origin of the surface structures. The η map calculated by the Sipe-Drude model is usually used to calculate the non-uniform energy deposition when the laser light interact with material’s rough surface. The symbol η is referred to the efficacy factor describing the efficacy of the inhomogeneous absorption at k(wave vector) with presented surface [15]. This analysis, however, forbids energy deposition with normalized wave vector smaller than one and does not support grooves formation. A recent published work shows that FDTD-η map is in good agreement to the η map theory and the possibility to have an energy deposition with a periodicity larger than the wavelength of the laser [18, 28]. This method is used in this work to investigate the inhomogeneous energy absorption of the laser light below a rough surface. The random roughness is introduced via the binary function proposed in [18]. The filling factor is set to 0.1. The dimensions of the Yee cells were set to 40nm, 40nm, and 20nm in X, Y and Z directions respectively. The lateral dimensions of the calculated area are set to 19 µm × 19 µm which contains 23 wavelengths of the femtosecond laser. The dimension in Z direction is set up such that the electric filed distribution below the surface up to 200 nm could be analyzed. We chose a linear polarized plane wave propagates along the Z direction as the light source. The optical properties of silicon is calculated by a combination of two-temperature model and Drude model mentioned above. To reduce the noise in the FDTD-η maps, an average of 10 FDTD-η maps calculated by using different binary functions was used.

The FDTD-η maps of silicon irradiated with a 1.5 kJ/m² femtosecond laser pulse at different substrate temperatures are shown in Fig. 5. We observe three different kinds of features, namely type-s, type-r and type-d features in the FDTD-η map on the surface of silicon at room temperature in Fig. 5(a). The type-s feature corresponds to the periodical energy absorption with a periodicity slightly smaller than the laser wavelength, which leads to the formation of LSFL ripples. The type-r feature is related to the surface structures with periodicity quite small compared to the laser wavelength. The type-d feature corresponds to the formation of surface structures with a periodicity around λ/Re(n) in space domain, in which λ is the wavelength of the laser and n is the complex refractive index of the material. The orientations of the structures related to the type-s and type-r features in space domain are perpendicular to the laser polarization while the structures induced by the type-d features are parallel to it. We should notice that the type-g feature is not observed here. The features in FDTD-η map change dramatically when the temperature of the substrate increases to 700 K. The type-d feature vanishes while the type-g feature, which corresponds to an energy absorption with a periodicity bigger than the wavelength leading to the formation of grooves, dominates. Meanwhile, the type-r feature becomes weaker and narrower. The type-s feature, however, disappears almost entirely, and this is not consistent with the experimental results where the ripples and grooves are coexist in this situation. The phonon–electron scattering is much stronger at 700 K compare to at 300 K, which increases the damping rate [29, 37, 38]. We show the FDTD-η map at the surface obtained with υ = 3.3 × 10⁻¹⁴ s⁻¹ in Fig. 5(e). The type-s feature occurs and merges with type-r feature while the type-g feature is still intense. We noticed that the type-g feature is not circular but elliptical. The major axis is perpendicular to the laser polarization, and it indicates the grooves related to type-g feature are parallel to it [28]. The intensity distributions on the silicon surface of the FDTD-η maps in Figs. 5(a) and 5(e) in the horizontal and the vertical direction are shown in Figs. 6(a) and 6(b). To reduce noise, five lines in the center of the FDTD-η maps are averaged together. The normalized intensity in the range where feature-g exist is very close to zero at 300 K while it is much higher at 700 K. Light absorption is greatly enhanced at 700 K and is consistent with the fact that the imaginary part of complex refractive index increases at 700 K. The increasing absorption leads to more energy deposited in to a thinner layer as mention above. We also show the FDTD-η maps 140 nm below the silicon surface in Figs. 5 (b), 5(d) and 5(f). The intensities decrease for all the three cases here because of the absorption in silicon. The type-r features shown in these three figures decrease the most and are consistent with the previously reported results [18, 24, 27, 28]. The intensity distributions of the FDTD-η maps in Figs. 5(b) and 5(f) in the horizontal and the vertical direction are also shown in Figs. 6(c) and 6(d). The type-d feature has higher intensity and the type-s feature is still present at 300 K. The most striking result is that only type-s and type-g features are present in Fig. 5(f), and Figs. 6(c) and 6(d) shows they have almost identical intensity. This simulation result well explains the coexistence of ripples and grooves formed in Fig. 1(b) and Fig. 2(e).

 

Fig. 5 FDTD-η maps of silicon irradiated with 1.5 kJ/m² femtosecond laser at different substrate temperatures and collision frequencies: 300 K, υ = 1.5 × 10⁻¹⁴ s⁻¹ (a) Z = 0, (b) Z = −140 nm; 700 K, υ = 1.5 × 10⁻¹⁴ s⁻¹ (c) Z = 0, (d) Z = −140 nm; 700 K, υ = 3.3 × 10⁻¹⁴ s⁻¹ (e) Z = 0, (f) Z = −140 nm. A linear scale was used ranging from blue (lowest value) to red (highest value) to present the normalized intensity in FDTD-η maps. The white dotted circles correspond to the position with a normalized wave vector equal to one. The arrow indicates the direction of the laser polarization.

Download Full Size | PDF

 

Fig. 6 The intensity distribution of the FDTD-η maps shown in Fig. 5: (a) horizontal (b) vertical direction at Z = 0; (c) horizontal (d) vertical direction at Z = −140 nm.

Download Full Size | PDF

People have proved the importance of the inter-pulse feedback mechanism in the formation of grooves previously [24, 27]. The energy absorption is affected by the surface structure created by the prior laser pulses. A previous report shows that the local light intensity is greatly enhanced in nano-holes [40]. The grooves and nanometer-size holes formed at 700 K with 100 laser shots are shown in Fig. 7(a). The distances between the holes are almost the same as in Fig. 2(d), in which a sample was irradiated by 50 laser shots. The average diameter increases slightly to 600 nm, and the surface is covered by more nanoparticles. To calculate the electric field distribution modulated by the surface structures, Fig. 7(a) is converted to a 3D structure based on the gray scale. The depth of the nano-holes is set to 250 nm. The electric field is normalized to the electric field we simulated with a smooth surface. The optical parameter is set to the same as Fig. 5(e). The normalized electric field on the bottom of the holes calculated by FDTD is shown in Fig. 7(b). It’s obviously that the local electric field is greatly enhanced in the nano-holes. The holes grow deeper and bigger with the enhanced local electric field when they are irradiated by the subsequent laser pulses. The adjacent nano-holes then merge together when the holes grow big enough and eventually form the structures shown in Fig. 3(b). Our proposed mechanism is further supported by the fact that the inner walls of the micrometer size holes shown in Fig. 3(f) are covered by nanoparticles, which we believe are the re-solidified liquid droplets produced by the locally enhanced energy absorption. To get a better understanding the formation mechanism especially the evolution of grooves, simulation based on the feedback will be investigated in the future.

 

Fig. 7 (a) Groove and nanometer size holes formed at 700 K with 100 laser shots; (b) The normalized electric field when the laser light irradiates the surface presented in Fig. 7(a). The arrow indicates the direction of the laser polarization.

Download Full Size | PDF

4. Conclusion

In summary, we compare the surface structures formed at different temperature irradiated with femtosecond laser pulses in detail. Traditional LSFL ripples, random structures, grooves and cones form successively as the number of laser pulse increases at a substrate temperature of 300 K. LSFL ripples with reduced surface roughness are generated at 700 K while modulations along the direction parallel to the polarization of the laser light are observed with 4 laser shots. The micrometer size grooves are the dominated surface structures after irradiated with 50, 100 and 500 laser shots at 700 K. The nanometer size holes are arranged along the valley of the grooves when the irradiated laser shots are 50 and 100. The holes expand and merge together to form micrometer size holes and valleys of grooves when the pulses number increases to 500. A combination of the two-temperature model, Drude model and FDTD method are used to calculate the FDTD-η maps at different temperatures. The results show that in the FDTD-η map at 700 K, the type-g feature, which corresponds to energy absorption modulation in the direction parallel to the laser polarization with a periodicity bigger than the wavelength, is the origin of the formation of micrometer size grooves observed in the experiments. Furthermore, the light absorption is enhanced at higher temperatures, which leads to the reduction of surface roughness. This work proves the possibility of using substrate temperature to tailor the properties of the surface structures induced by femtosecond laser irradiation. It can benefit both surface structures based applications and mechanism study of femtosecond laser-matter interaction.