1. Introduction

Since the first observation of laser-induced periodic surface structures (LIPSS) on various semiconductor surfaces by Birnbaum [1], laser-induced or laser direct writing of nanostructures on the surface and inside the volume of various materials has been intensively studied over the past two decades [2–20]. Most observations were limited to both low-spatial-frequency LIPSS (LSFL) and high-spatial-frequency LIPSS (HSFL) or nanogratings. For strongly absorbing materials such as metals or semiconductors, LSFL are primarily observed with a period close to or slightly smaller than the writing laser wavelength and an orientation of the planar structures perpendicular to the laser polarization [2–6,20]. However, HSFL are also observed but less common [2,3,21,22,25], especially in the case where the induced periodic structures are oriented in parallel to the laser polarization [3,25]. Under some circumstances such as irradiating the targets in different ambient conditions or with a confinement layer, periodic arrays of columns, nanoholes and other type of structures were formed [18,19]. In the case of water confinement or using high repetition rate laser, ultra-fine HSFL was induced with its orientation (groove direction) perpendicular to the laser polarization [23,24]. The formation mechanism of the above-mentioned periodic surface structures on semiconductors was mainly interpreted with the Sipe-Drude model or the laser plasma polariton interference model [4,10,11,26]. In this paper, we report new types of laser-induced ultra-fine nanostructures on the surface of silicon. To the best of our knowledge, there have been no reports on the generation of ultra-fine features, such as droplet array, grid array, and HSFL with its orientation either parallel or perpendicular to the laser polarization, by femtosecond laser direct writing at the silicon-air interface at relatively low repetition rate. These various nanostructures were found to co-exist, and sharp transitions from one to another were also observed along the scan track. Possible physical mechanisms including Coulomb explosion, interference and local field arrangement are proposed to interpret these structures. For the sake of clarity, we here define a few generic terms which will be used in the discussion section:

2. Experiments and discussion

The experiment was carried out with a 1 kHz, 45 femtosecond (fs), 800 nm Ti-Sapphire laser and a 1 kHz, 200 fs, 1550 nm ultrafast OPA. For the 800 nm experiment, an EO infinity-corrected 10X objective is used for focusing the beam whereas a Mitutoyo NIR 50X is used for the 1550 nm experiment. The focal spot diameters, D, are estimated to be 6 μm and 2.5 μm, respectively. The sample, Si(111), was mounted on a 3D translation stage with its front surface located at the focus of the laser beam. Pulse energy (E_p) was varied from 34 nJ (0.24 J/cm²) to 46 nJ (0.33 J/cm²) for 800 nm, and from 5 nJ (0.2 J/cm²) to 25 nJ (1.0 J/cm²) for 1550 nm, whereas pulse-to-pulse spacing, d, which is defined as a ratio of scan speed to laser repetition rate, was varied from 10 nm to 150 nm for both cases. The corresponding effective pulse number, N_eff, which can be simply calculated as D/d, was indicated in the figures. All laser irradiations were performed with the motion stage running at a constant speed unless otherwise stated, and the experiment was performed with only one parameter varied at a time and at a fixed interval. Following the irradiations, the sample was examined with a scanned electron microscope SEM (FEI Quanta 3D FEG).

In the 800 nm case, Type I structure was produced with E_p in the range between 38 nJ (0.27 J/cm²) and 46 nJ (0.33 J/cm²). As shown in Fig. 1(a) where the structure was inscribed at 44 nJ, the corresponding laser fluence is 0.31 J/cm². The period is slightly increased from 540 nm to 600 nm with the pulse-to-pulse spacing increased from 10 nm to 150 nm. Such dependence of Λ on d is similar to the one previously reported on the surface of fused silica [15], although a slight difference in the structure was present for small d which corresponds to large N_eff. The cause of this discrepancy will be explained later. In addition, in the transition from stationary to moving case which was carried out with pulse fluence fixed at 0.27 J/cm², the symmetrically distributed nanogrooves and the decrease in the number of nanogrooves when the structure is written with E ⊥ S (Fig. 1(b)), and the different extents of elongation of nanogrooves in terms of the overlap of different portion of the laser beam profile in the case of E ║ S (Fig. 1(c)), all these coincide with the fundamental criteria of the incubation-based nanoplasmonic model [15] and thus with which the type I formed on the silicon can be well interpreted. The physics behind this model is the local field rearrangment in accordance with the positive feedback from laser-induced defects. Local field enhancement and suppression are taking place around the ionization zones in the direction perpendicular and parallel to the laser polarization, respectively. Periodic distributed local intensity maxima are thus evolved in a certain sequence according to the feedback and thus result in the formation of nanogrooves. At small d where relatively more energy is deposited, thermal melting and resolidification take place in the case of silicon while severe ablation occurs and causes damage to both nanogrooves and nanoplanes in the case of fused silica [15]. This discrepancy is due to different thermal properties of these two materials. When the melt covers part of the groove on silicon, local field rearrangement [17] takes place and thus deform the leading side lobe, which leads to the formation of irregular patterns. A sharp transition due to the fluctuation of pulse fluence was observed at 0.27 J/cm² as shown in Fig. 1(d) where the nanostructure changes suddenly from the type I to the mixing of grid array and the type II structure. A detailed description for the formation of these nanostructures will be provided below.

 

Fig. 1 (a) Type I structure induced by 800 nm laser for E_p = 44 nJ (0.31 J/cm²) on the surface of silicon. (b) Transition from stationary to moving for E ⊥ S. (c) Transition from stationary to moving for E ║ S. (d) Transition from the Type I structure to mixed grid array and the type II structure. The scale bar on the bottom left is for (a), the one on the bottom right is for (b) to (d). N_eff is the effective pulse number. (b–d) E_p = 38 nJ (0.27 J/cm²).

Download Full Size | PDF

With the decrease of the pulse energy to 38 nJ or less, we observed a number of ultra-fine periodic nanostructures for different d. As shown in Fig. 2, melting droplets seem to be the main feature of the nanostructures which were produced at 38 nJ (0.27 J/cm²) with E ║ S. Surprisingly, the droplets are distributed in a very regular form as illustrated by the curved (Fig. 2(a)) and straight dashed lines(Fig. 2(c)). What is even peculiar is that the droplets are not only aligned periodically and parallel to the electric field, but also seemed to be driven by the laser field as evidenced by the tracks left behind. In addition, at the onset of the scan track, the melting droplet arrays are co-existing with other types of nanostructures such as the type I inside the rectangle in Fig. 2(b) and the type II inside the ellipses in Figs. 2(b) and 2(d). The periods of both the droplet arrays and the type II are about 200±25 nm. Note that the period of the type I structure inside the rectangle is about 600 nm, it is quite obvious that the type I is just about to develop as evidenced by the shot-to-shot evolution pattern (22 shots) shown in the inset. In the other case where the nanostructures were created at 36 nJ (0.25 J/cm²) or 38 nJ (0.27 J/cm²) with E ⊥ S, the topography, as shown in Fig. 3, is different from the previous case but bears some similarities with it. The melting droplet arrays are also regularly distributed but mainly along the edges (Figs. 3(a) and 3(d)) and with a band of randomly distributed droplets present at the center of the scan track for d = 10 nm only. Also, in the transition cases from stationary to moving, as shown in Figs. 3(d) and 3(e), the droplets are distributed in a ring structure, which could be interpreted as results of Coulomb explosion [29]. In addition, the droplets were confined to the surface due to the surface tension. As compared to the previous case (i.e. E ║ S), the orientation of the melting droplets is changed by 90° in agreement with the change of laser polarization, which may suggest that the ultrafast melting [30–33] is taking place within the pulse duration in this case. Moreover, ripples that consist of traces of melt are clearly observed in the tilted view (Fig. 3(b)) and its orientation is perpendicular to the laser polarization. The blow-up for the ripples and melting droplets is shown in Fig. 3(c). The spacing between the melt traces is about 210 nm and the droplets are about a few tens of nanometers in size. With a slight change in either pulse energy or pulse-to-pulse spacing, the nanostructures could be switched from droplet arrays (Figs. 3(a), 3(c) and 3(d)) to grid arrays (Fig. 3(f)) and to a type II structure (Fig. 3(g)) as well. Moreover, a wave-type structure was observed on the edges of the scan tracks within which either the grid array or the type II, or both are present. The pitch of this wave-type structure is the same as the enclosed nanostructures (Figs. 3(f) and 3(g).)

 

Fig. 2 Melting droplet array formation by 800 nm laser with E_p = 38 nJ (0.27 J/cm²) for different overlap of pulse number. The curved and straight dashed lines in (a) and (c) are given as guides to the eye. They indicate the regular distribution of the droplets. The nanostructures inside the ellipses in (b) and (d) are the type II structure, but the one in the rectangle in (b) is a type I structure at its initial stage. The inset shows the shot-to-shot evolution pattern of the type I structure at 22 shots. (c) and (d) are the tilted views at 52°, the real width of the nanostructures in vertical direction should be divided by a factor of cos(52°). The scale bar is for all the images except the inset.

Download Full Size | PDF

 

Fig. 3 (a–e) Formation of melting droplet array, (f) grid array and (g) type II structure. (b) is the tilted view of (a). (c) is the blow-up of (b). The scale bar is for (a), and (d) to (g).

Download Full Size | PDF

We tried to repeat this experiment and found that all the experimental results presented above can be replicated. However, it is difficult to observe all these different types of nanostructures in one experiment unless all the writing parameters are changed in very small steps. This is simply because these nanostructures are very sensitive to the total amount of deposited energy which is a function of the pulse fluence and pulse-to-pulse spacing. In one of our repeated experiments in which E is parallel to S, apart from the type II structure (Fig. 4(b)), we observed another ultra-fine HSFL which we call type III (Fig. 4(d)). The pitch of type III is the same as the type II, but its orientation is perpendicular to the laser polarization. Depending on the total deposited energy, the type III was found to coexist with either the droplet array or the grid array as shown in Figs. 4(a) and 4(c), respectively.

 

Fig. 4 Formation of the type III structure by 800 nm laser. Mixing of the droplets and the structure of type III (a) and type II (b). (c) Mixing of the grid array and the type III structure. (d) The type III structure. The scale bar is for all the images. Note that the slightly lower pulse fluence than the previous case is very likely due to the uncertainty of the alignment.

Download Full Size | PDF

In the experiment with the 1550 nm laser, the type I and type II structures were produced (type III was not observed) with E_p set in the range between 5 nJ (0.2 J/cm²) and 25 nJ (1.0 J/cm²) and d set in the range between 10 nm and 150 nm, respectively. As shown in Fig. 5, the pitch of the type I is about 1 μm (Figs. 5(a) and 5(b)), whereas the pitch of type II is about 320 to 340 nm (Figs. 5(c) and 5(d)). Both scale linearly with the laser wavelength respectively in comparison to those formed by the 800 nm laser. One notes that the type I is not as regular as the one produced by the 800 nm laser, but the type II is much more significant and regular. The melting droplets which are distributed in line with the nanogrooves were observed in the type II structure and only for d = 10 nm, and its size is much bigger compared to those formed by the 800 nm laser, as shown in Fig. 5(e). A sharp transition from the type II to the type I was also observed along the scan track, as shown in Fig. 5(f). One notes that the wave-type structure is not present in this case, which suggests that a pool of melt is not induced as evidenced by the unmodified regions between the nanogrooves (Figs. 4(c) and 4(d)). This is probably because the interaction with the 1550 nm laser relies on a two-photon process where the energy deposition is well confined compared to the single photon process taking place with the 800 nm laser.

 

Fig. 5 type I and type II structures formed by 1550 nm laser. The scale bar on the bottom left is for (a), (c) and (e), and the one on the bottom right is for (b), (d) and (f).

Download Full Size | PDF

As described above, a few ultra-fine nanostructures were observed and they are very sensitive to the irradiant fluence. So far there is no model that can satisfactorily interpret all of these phenomena observed in our experiments where many different physical processes such as Coulomb explosion, ultrafast melting, interference, material ablation, and thermal melting coexist. These processes contribute differently under different writing parameters and thus lead to the formation of various nanostructures. Therefore, it is reasonable to address the formation mechanisms empirically, i.e. in terms of each of the formed nanostructures.

We first attribute the formation of the melting droplet array and the grid array to the consequence of wave interference. This idea is inspired by the topography of the grid array shown in Fig. 3(f) in which the grid pattern resembles an interference pattern produced by waves from two sources. When a femtosecond laser pulse is incident upon the target surface, a pool of melted silicon is immediately produced via ultrafast melting accompanied with Coulomb explosion. This situation is similar to the effect of a thrown stone hitting the water. Some of the melt is displaced from the center as a result of the Coulomb explosion and propagates outwards but is mostly confined to the surface due to the surface tension as evidenced by the splashing of the droplets in Figs. 3(a–e). In the transition from stationary to moving as shown in Figs. 3(d) and 3(e), the droplets are distributed in a ring structure and the dashed lines in Fig. 3(e) clearly show that the droplets are originating from roughly the center of the ring which corresponds to the peak of the focused femtosecond laser beam. Electric force from the laser field also acts on the droplets as evidenced by the change of the droplet elongation direction following the laser polarization as shown in Fig. 2(c) and Fig. 3(d). This also suggests that the ultrafast melting takes place with the pulse duration. The wave nature is further confirmed by the multiple rings of droplets and the ripples, as shown in Fig. 2(a) and Fig. 3(a), respectively. The ripples are more obvious in the tilted view shown in Fig. 3(b). The spacing between the rings or ripples represents the wavelength of laser-excited electron plasma waves, which governs the period of the induced nanostructures. Again, the formation of a central band of randomly distributed droplets could also be easily understood by considering the interaction by analogy to a thrown stone hitting the water. In this scenario, the liquid (melted material) is pushed outwards (escaping radially in the plane) and downwards (normal to the plane) leaving a void at the center. The surrounding melted material moves quickly inward to fill the void. Some could be pushed upward above the surface into the air due to kinetic energy and then could settle down by deposition. A band of randomly distributed droplets can thus be formed(Fig. 3(a)).

The melt resolidifies before the arrival of the next pulse that hits on the surface at a distance of pulse-to-pulse spacing with respect to the previous pulse. Because of the memory effect or the surface topography induced by the previous pulses, the striking of the next pulses would cause more than one ionization spots on the surface. Therefore interference of the laser-excited electron plasma waves is generated on the melt. Depending on the amount of deposited energy, if the crests are lower than the “gentle” ablation threshold [27,28], the interaction zone would remain in the melted state within the pulse duration, no material is ejected. But the weak Coulomb explosion will drive the droplets outwards to the edges of the scan track. Therefore, less droplets are present in the middle except for those from redeposition. However, if the crests are slightly above the “gentle” ablation threshold, a very thin layer of material will be removed through the Coulomb explosion at the corresponding locations. The typical ablated spot size is about 150 nm in diameter (Fig. 3(f)). In the mean time, a certain amount of deposited energy is removed along with the material ejection. Therefore, there is not enough energy left for further melting and less material is pushed outwards. As a consequence, the droplets are less significant or almost disappear at the edges as shown in Figs. 3(e–g), only a melted profile (wave-type structure) is produced.

The formation mechanism for the type II structure is slightly more complex. Apart from the interference and the Coulomb explosion, the action of the electric force of the laser pulse on the charged particles at the crests of the interfering waves, which strongly pulls the charged material in the laser polarization direction during the ultrafast melting stage, changes the local field distribution and thus the interference pattern. This is evidenced by the evolution from the grid array to the type II structure shown in Fig. 3(f). This evolution is realized through the memory effect so that the crests line up along the laser polarization direction forming a nanogroove as evidenced by the joining of ablation spots shown in Fig. 1(d) and Fig. 3(f). As a consequence of that, periodic nanogrooves are formed with its groove direction parallel to the laser polarization.

As the deposited energy is slightly increased, the nanostructure could switch from the type II to the type III as shown in Figs. 4(b) and 4(a) (where the deposited energy is increased by decreasing the pulse-to-pulse spacing). We argue that in this moment the local field effect starts to set in and take control of the structure orientation via the local field enhancement around the direction perpendicular to the electric field [15]. This is evidenced by the link up of the grids in the direction perpendicular to the electric field shown in Fig. 4(c). However, the local field effect is not strong enough to play a role in adjusting the grating pitch, therefore, the period of the nanostructure is still governed by the interference.

As the deposited energy is further increased by increasing the pulse energy to above 38 nJ, the type I structure is created with its pitch three times greater than that of the type II or type III. The reason for this is that, at this moment, the local field effect is completely taking over the control of the interaction process. Now it not only governs the orientation of the nanogrooves, but also the period through the evolution of the side lobes [15]. Because the interaction is highly localized, the propagation of laser-excited electron plasma waves becomes impossible and thus no interference of the electron plasma waves could take place. This is why the mixing of the droplet array and the type I structure only exists in the transition from stationary to moving where the droplets are created before the formation of the type I structure (Figs. 1(b) and 1(c)). The transition from one nanostructure to another (e.g. from type I to type II as shown in Fig. 1(d) or vice versa, as shown in Fig. 5(f)) is dependent on the variation of the deposited energy along the scan track, and this transition could invert if the variation of the deposited energy is reversed. The sudden change in structure is simply a threshold effect. A systematic overview of these nanostructurings in terms of respective parameters and underlying physical mechanisms is summarized in Table 1.

Table 1. Overview of nanostructuring with 800 nm laser in terms of respective parameters and underlying physical mechanisms (d refers to pulse-to-pulse spacing, Λ is period).

View Table

3. Conclusion

We have produced novel ultra-fine nanostructures with pulse energy around the threshold for silicon ablation. With the deposited energy slightly increased, the nanostructures could change from the melting droplet array to the grid array, to the type II and type III structures and finally to the type I structure. The proposed corresponding mechanisms, interference with the contribution from Coulomb explosion for the formation of the droplet array and the grid array, interference and action of electric force for the formation of the type II structure, interference plus the local field effect for the formation of the type III structure, and the local field effect for the type I structure, satisfactorily explain most observations. Our experimental results highlight that the laser matter interaction on the surface of silicon can be more diversified than what has been reported on the surface of silica glass [15]. Many physical processes are involved in the interaction process and respectively play different roles depending on the writing parameters. The need for a systematic experimental study including theoretical modeling is highly demanded.