1. Introduction

The needs for high-speed material processing with higher precision and lower mechanical and thermal damage have increased in the fields of mass production for modern device [1–4]. While traditional mechanical processes like diamond sawing, drilling, or scribing have come to their limitations in the course of size reduction with less damages, laser based material processing has been presented to offer an alternative method to tailor microelectronic and micromechanical components. However, even nanosecond laser is not free from any thermal as well as mechanical damage to the material. Micromachining based on ultrafast lasers can be used to process biological samples, including live cell and thin films as well as bulk materials for many fields including micro-optics, electronics, and even biology [5–11].

Ultrafast laser techniques have been known to exhibit potential applications in high precision processing because of their minimizing heat affected zone (HAZ), as compared to the nanosecond laser technologies [12–16]. Other processing technology based on the high energy particle such as an electron or beam ion, which needs a special environment like vacuum, shows a limitation in its application [17]. So, ultrashort pulse laser techniques have been investigated widely to develop the counterpart techniques. However, the ultrafast laser techniques have a very weak point in terms of the processing speed, which is much slower than that of the traditional mechanical method currently adapted in the device production. It is urgent to overcome those issues for real industrial applications of ultrafast laser processing. Meanwhile, adaptive optics, which is generally used in nanosecond laser processing, may be applied to overcome the limitation of low processing speed [18]. However, it is not convenient to directly apply adaptive optics to ultrafast laser processing due to the deterioration of the performance caused by a possible change in the ultrafast laser beam characteristics.

In this report, we present a novel optical processing technology to improve the processing performance of femtosecond micromachining. The precise coupling of ns laser pulse into an ultrafast laser pulse both in the spatial and temporal domain, in short, hybridization processing, results in remarkably higher processing speeds compared to the case where each laser is used in processing by itself. Furthermore, we have observed better performance in material processing with minimizing the surface roughness, which is usually observed in ultrafast laser processing [19,20], as well as mechanical stress caused by the illumination of nanosecond laser pulses with high fluence on a substrate [21,22].

2. Experiments

Figure 1 shows a schematic diagram of the experimental setup. The femtosecond amplifier system consists of a mode-locked Ti:Sapphire oscillator and a regenerative amplifier, which delivers 150 fs laser pulses at 810 nm with a repetition rate of 1 kHz [23]. The nanosecond laser has a pulse width of 250 ns with a repetition rate of 1 kHz at 532 nm. A delay generator was used to vary the time delay between the nanosecond laser and femtosecond laser pulse.

By using the stabilization system consisting of the polarization beam splitter and the half wave plate, we achieved better long-term stability in nanosecond laser output. A variable neutral density filter was used to attenuate the femtosecond laser power. A p-doped Si (100) wafer with a thickness of 650 µm was mounted on the motorized translation stage that was used to translate the sample with a constant speed of 6 mm/sec. The distance between laser spot is 6 µm/pulse. Both the femtosecond and nanosecond laser were focused on the wafer with a bi-convex lens with a focal length of 150 mm, and the beam diameters at the focal plane were 40 µm and 80 µm for the femtosecond laser and nanosecond laser, respectively.

The ablation depth, as well as surface morphology, was characterized by atomic force microscopy (AFM, XE-120, PSIA, Korea) and scanning electron microscopy (SEM W-800, Hitachi, Japan)

 

Fig. 1. Schematic diagram of the hybridization laser processing setup (DM: dichroic mirror, BS: beam splitter, PD: photo diode, PBS: polarization beam splitter, HWP: half wave plate, ND: neutral density filter)

Download Full Size | PDF

3. Results and discussions

Figure 2 exhibits the images of AFM and SEM of silicon surface processed at three different time delays (-300, 0, and 300 ns) in the hybridization configuration in addition to ns- and fs-laser pulses only. The time delay is defined in terms of the time difference between the femtosecond laser and the nanosecond laser. At the negative delay, the nanosecond laser follows the femtosecond laser, and the nanosecond laser arrives before the femtosecond laser at the positive time delay. The fluences of the nanosecond laser and femtosecond laser pulses are kept at 1.0 J/cm² and 0.13 J/cm², respectively. The fluence of both lasers are close to the damage threshold of the silicon substrate as shown in Fig. 2(a) and (b) [24,25]. First, we have processed the silicon wafer with the nanosecond laser or femtosecond laser by itself (we have denoted this as ‘ns only’ or ‘fs only’, respectively) to compare the performance of the hybridization processing to that of normal microprocessing. It should be noted that the apparent features of a micro crack can be observed in the case of the ns laser pulse only exposure. The micro-crack extends longer than about 100 µm from the center of laser spot. Optical microscopic images on the processed area also clearly show the appearance of the micro crack. Meanwhile, in case of the fs laser only, the laser illumination results in a prominent surface roughness with neither actual ablation of the silicon substrate nor the micro crack.

 

Fig. 2. AFM and SEM images of the processed Si wafer surfaces for (a) the nanosecond laser only and (b) the femtosecond laser only, and hybridization processing with a time delay of (c) -300, (d) 0, (e) +300 ns.

Download Full Size | PDF

In hybridization processing, the ablation of the Si wafer was investigated with varying time delay from -450 ns to 350 ns. For the time delay of -300 ns, as in the ns only and fs only cases, the ablation depth is only about 0.2 µm. It is interesting to note, however, that both the surface roughening and micro-crack formation is substantially lowered even if the two laser pulses are completely isolated from each other in the temporal domain (Fig. 2(c)). When one sets the time delay between the two different laser pulses to be zero while keeping all the other exposure conditions including laser fluence constant, the valley-like morphology appears in both the AFM and SEM images with a remarkable enhancement of the ablation depth to about 3.0 µm (Fig. 2(d)). An additional interesting observation is the substantial reduction of the surface roughness and micro-crack. With a further increase of the time delay to 300 ns, the ablation depth slightly decreased to about 1.8 µm with a little, but observable, micro-crack on the processed surface of the silicon as shown in Fig. 2(e). It should be noted that the width of the processed area is almost constant, even though the depth of processing is strongly dependent on the time delay.

In order to have more quantitative information on the observed large enhancement of the ablation depth in hybridization processing, we have measured the cross-sectional profile of the processed regime by AFM and show the results in Fig. 3(a) at the various time delays. The pulse trails observed in a digital oscilloscope are also shown in Fig. 3(b). When the delay time is set to zero, we observed the largest ablation depth. To estimate the process efficiency quantitatively, we have measured the ablation area from the cross-sectional depth profile and denote the results as a function of delay time in Fig. 4. The ablation efficiency at a delay time of zero is more than 12 times that of the femtosecond laser only. At the negative delay, the dependence of the ablation efficiency on the time delay between the fs- and ns laser pulses is very close to the temporal profile of the ns laser. At the positive time delay, however, we observed an apparent delayed decaying component with the time to decrease to 50 % of the peak of several hundreds of nanosecond in the ablation efficiency curve.

 

Fig. 3. (a) AFM cross sectional depth profile of the processed surface of a silicon wafer for various time delays. (b) Pulse trail of the laser pulses used in the hybridization process of a silicon wafer.

Download Full Size | PDF

No apparent enhancement in the ablation efficiency at the negative time delay reveals that the precedent fs-laser illumination on the silicon substrate does not affect on the ablation by the following ns-laser pulse. This observation could be rationalized in terms of the different ablation mechanisms of thermal and nonthermal processes for ns laser and ultrafast laser pulses, respectively. Furthermore, the ablation threshold of 0.8 J/cm² for 10 ns laser pulses is about four times higher than that for 100 fs laser pulses. [24,25] Therefore, the precedent fs laser pulse energy close to the ablation threshold scarcely alters the local temperature of the materials, and then dose not result in any prominent changes in the overall processing efficiency.

 

Fig. 4. Cross-sectional ablation area of a silicon wafer as a function of delay time. The temporal feature of nanosecond laser pulse (dotted line) is shown for comparison.

Download Full Size | PDF

To explain the observed rather large nonlinear enhancement of the ablation efficiency at the positive time delay, it is valuable to consider the general picture on the mechanism related to the interaction between the laser energy and materials. When the laser energy is locally deposited into the materials, the generation and subsequent acceleration of free carriers, including electrons and holes follows. A prominent temperature rise in the focal volume occurs during thermalization of the laser energy through collisions with heavy particles, like atoms and molecules, and non-radiative recombination processes of the carriers with a time constant of a few picoseconds to tens of picoseconds, depending on the materials.

At a given nanosecond laser fluence close to the ablation threshold, the nanosecond laser pulse transiently changes the material states of either its local temperature or carrier densities. Meanwhile, it is somewhat insufficient to explain the dependence of the ablation efficiency as a function of the time delay in terms of free carrier densities through direct generation by the nanosecond laser itself. This supposition was supported by the observation of the presence of the delayed decaying components of ablation efficiency as shown in Fig. 4. As mentioned before, the relaxation time constant of the free carriers is much shorter than the nanosecond laser pulse width. The ultrafast relaxations of the free carrier result in hot phonons, which eventually relax without any prominent change the in material surfaces if the fluence is less than the ablation threshold. At high temperatures before the onset of bulk melting, incomplete surface melting has been characterized by the formation of a liquid-like layer on the surface of the semiconductor [26,27]. This behavior of the Si(100) surface at higher temperatures has been investigated using various techniques, including high resolution electron energy loss spectroscopy, ultraviolet photoemission spectroscopy, and X-ray absorption studies. These works suggested that the incomplete surface melting dynamics play a significant role in the semiconductor to metal transition below the bulk melting temperature of 1685 K. Accompanied by the increase in surface metallicity, the density of states at the Fermi level increased with increase in temperature. If successive fs-laser pulses illuminate the area, however, the effects of the precedent ns-laser on the material states remarkably affect the nature of the ablation process produced by the latter laser pulse, as well as the ablated surface. The increase in temperature of the materials should play a significant role in the enhancement of the ablation efficiency under the configuration of ns- and fs-laser hybridization. The previous theoretical investigation reported that the ablation threshold fluence strongly depends on the substrate temperature and is significantly lower for irradiation with 15-ps pulses than for 150-ps pulses [28]. In fact, the statically heated silicon substrate at 900 K exhibits an apparent reduction in its ablation threshold with a little increase in the ablation efficiency [29].

The temporal coupling of the femtosecond laser and nanosecond laser induces the local temperature change on the sample during processing, which produces a remarkable decrease in the ablation threshold fluence and increases the processing speed accompanied of the surface metallization of Si(100). Increasing of the nanosecond laser energy causes a change in the temperature of the target material. The energy of the nanosecond laser is controlled so as not to cause the irreversible change with the nanosecond laser fluence by itself. At the same time, the femtosecond laser pulse, which is synchronized in the time domain, irradiates the same space and irreversibly ablates a large amount of material with a lower fluence. Such a progress in femtosecond laser micro-processing makes it possible to maximize the processing speed and reduce the processing threshold energy, which decreases the various high order nonlinear effects which are confronted when we focus high-power femtosecond laser pulses on the target under atmospheric conditions [30].

4. Conclusions

We have demonstrated a novel hybridization method to overcome the inherent lower processing speed of ultrafast laser micromachining. While keeping the good characteristics of the femtosecond laser micromachining, such as a reduction of the thermal and mechanical damage area, we can improve the processing speed by transiently changing the nature of the target substrates before the femtosecond laser shot. By synchronizing the nanosecond laser with the femtosecond laser in the temporal and spatial plane, it is possible to increase processing performance with a small amount of femtosecond laser fluence.