CHAPTER 7 

FEMTOSECOND EXCIMER LASER MICROMACHINING OF AL₂O₃-TiC CERAMICS. INFLUENCE OF THE LASER PULSE DURATION

The micromachining of Al₂O₃-TiC ceramics using femtosecond duration laser pulses (l=248 nm) is studied. The results are compared with those obtained when processing is carried out using nanosecond laser pulses at the same radiation wavelength, in order to investigate the influence of the laser pulse duration on the ablation process.

7.1 INTRODUCTION

Progress in the field of laser materials processing relies on the development of innovative laser systems. Recently a new class of ultrashort pulse duration lasers has appeared and is being used to investigate laser radiation-material interaction phenomena and their applications in materials processing [1]. Two laser systems of interest in this context are the Ti:sapphire (l=735-1053 nm) and excimer-dye (l=220-300 nm, l=380-760 nm) laser configurations. With the advent of chirped pulse amplification [2] and pulse compression techniques [3], these lasers offer pulse durations in the femtosecond range and pulse energies up to 125 and 15 mJ for Ti:sapphire and excimer-dye laser configurations, respectively. The ultra-short pulses opened new perspectives for the shaping of difficult to machine materials with nanosecond duration pulses and to improve the accuracy of laser machined parts [4]. Materials that are difficult or impossible to machine with nanosecond duration pulses include materials for which the photon energy of excimer lasers is not high enough for direct (single-photon) bandgap excitation or for efficient defect generation [5], and metals, which melt extensively, leading to poor finish and low accuracy [6].

For pulses with duration t longer than a few tens of picoseconds, the generally accepted picture of bulk damage to defect-free dielectrics involves heating of conduction-band electrons by the incident radiation and transfer of this energy to the lattice [7]. The electrons and lattice reach thermal equilibrium within the first tens of picoseconds of the laser pulse, and damage typically occurs when the deposited heat is sufficient to melt, boil or fracture the dielectric material. The controlling rate of damage is that of thermal conduction through the lattice, and since the heat penetration depth , this model predicts a t^1/2 dependence of the threshold damage fluence [8, 9]. A deviation from this “long-pulse” scaling of the breakdown threshold was first reported by Soileau et al. [10] for pulses with duration in the range of 4 to 10 ps. Du et al. [11] studied the breakdown of fused silica induced by 780 nm laser pulses with duration in the range of 150 fs to 7 ns, and determined that the fluence threshold for breakdown varies with the square root of pulse duration t only for pulses longer than 10 ps. Several studies on laser-induced damage thresholds of dielectric materials for pulse duration ranging from 140 fs to 1 ns [12, 13], confirmed a change in the damage mechanism and morphology for pulses shorter than 10-20 ps, and a deviation from the long-pulse t^1/2 law (Figure 7.1.1). This deviation has been attributed to a transition from the long pulse duration, thermally dominated regime, to a regime dominated by multiphoton and collisional ionisation for ultra-short pulses [12]. The ultra-short pulse duration allows to reach extremely high intensities (>10¹⁴ W/cm²), leading to significant field-induced multiphoton ionisation [14, 15]. Once a high free electron density is produced by multiphoton ionisation, the material no longer has the properties of a dielectric, and it will absorb the laser energy via inverse Bremsstrahlung heating, similar to a metal [13]. For these ultra-short and intense pulses, free electrons gain energy from the laser field much faster than it is transferred from the electrons to the lattice. The energy of electrons increases until they have sufficient energy to collisionally ionise neighbouring atoms thereby producing more free electrons [16]. According to this mechanism the damage threshold occurs when a sufficiently high electron density is reached [12, 13]: a lower limit electron density occurs when the energy density of conduction electrons equals the binding energy of the lattice, while a higher limit is the critical electron density at which the plasma becomes reflective, which depends on the radiation wavelength [16]. Calculations indicate that the theoretical damage threshold is only logarithmic dependent on the choice of the electron density value, as defined above [12]. The actual damage occurs after the pulse has passed, when the dense plasma expands away from the surface [17]. The deviation from the t^1/2 dependence of the threshold damage fluence described above for dielectrics, has also been observed for metals [13, 18].

Ablation of oxide ceramics by subpicosecond pulse duration lasers has been studied with respect to various applications, including micromachining and deposition of thin films [19]. The enhancement of non-linear optical effects along with reduced thermal effects and high precision, allows the manufacturing of 3D microstructures in transparent materials such as sapphire, quartz, and CaF₂ [5, 20]. Processing of transparent materials using a Ti:sapphire laser and pulse duration of 4.5 ps and 200 fs showed that machining with the shorter 200 fs laser pulses greatly improves surface finish [20]. Ilhemann et al. [21] compared femtosecond and nanosecond pulse duration excimer laser ablation of oxide ceramics, using pulses with a duration of 500 fs and 30 ns at l=248 nm. For alumina and magnesia the results were similar, with femtosecond pulses showing improved coupling of radiative energy to the material due to enhanced non-linear absorption effects. A lower ablation threshold and increased ablation rate were observed for 500 fs laser pulses as compared to 30 ns pulses. Zirconia was shown to have similar ablation thresholds for both the 248 nm pulse duration pulses. This was attributed to zirconia’s low bandgap, which decreases the importance of non-linear absorption processes. In all cases, femtosecond pulses showed a larger ablation rate at identical laser fluence. The benefits of 350 fs duration pulses over 1.4 ns pulses was demonstrated for hole drilling of tooth enamel [13]. In the case of 1.4 ns pulses, a fluence of 30 J/cm² was required to drill the holes, and was followed by cracking and significant thermal damage. In contrast, the 350 fs pulses required only 3 J/cm², and material removal did not involve any thermal damage. Temperature measurements showed that in the case of the 1.4 ns pulses, the bulk temperature of 1 mm slices of tooth increased over 40°C, whereas for the 350 fs pulses, the temperature rise was only 2°C.

Similarly to that observed when laser treatment is carried out with nanosecond pulses, the growth of surface structures has also been observed using pulse lengths in the femtosecond range. Rubahn et al. [22] investigated the ablation of crystalline muscovite mica, and observed that processing with 500 fs laser pulses at l=248 nm results in cone formation. Nanosecond laser pulses lead to different morphologies, depending on the radiation wavelength. For 248 nm radiation, extensive melting results in extremely irregular surfaces, but cones do not develop. In the case of 193 nm radiation, cone-like structures develop after irradiation with consecutive pulses for fluences just above threshold (~1 J/cm²), but do not form at higher fluences (~10 J/cm²). No explanation for the morphology dependence on the pulse length was provided. Various micron-sized surface structures have been observed on silicon after irradiation with ion beam or pulsed laser radiation. For example, mounds and columns are formed when silicon is sputtered by energetic Ar ions [23]. In chapters 4 and 5, several reports on structure formation in silicon with nanosecond duration laser pulses are described [24-28]. Formation of arrays of conical spikes when silicon surfaces are irradiated with high fluence femtosecond laser pulses in SF₆ or Cl₂ was reported by Her et al. [29, 30]. The authors suggest that the asymmetric shape of the base of spikes results from the laser polarisation, leading to a growth mechanism dependent on the local energy density due to differential absorption of s- and p-polarised light. However, blunt spikes develop in vacuum and sharp spikes in halogen-containing gases [29], indicating that laser-induced chemical etching is also involved in the removal of material. The authors observed that the height of the spikes decreases with increasing pulse duration or decreasing laser fluence. Since the cross sections of laser-induced etching and ablation are strongly dependent on the laser fluence and pulse duration [21, 31], the authors considered that both processes could explain the observed variation.

7.2 EXPERIMENTAL

The ablation experiments were carried out at the Laser Laboratorium Göttingen, Germany, and performed with a femtosecond pulse duration KrF laser system (l=248 nm) developed in-house. The pulse duration of this system is t=500 fs and the pulse energy is about 10 mJ. The laser system consists of a hybrid excimer-dye laser, capable of generating subpicosecond pulses at most of the excimer laser wavelengths [32]. The experimental arrangement is shown in Figure 7.2.1. An excimer laser is used as a pump laser for pumping a special subpicosecond dye-laser amplifier arrangement, and as an amplifier for the amplification of the frequency doubled output pulses of the dye laser set-up. The oscillator channel of an excimer laser delivers 15 ns pump pulses at 308 nm. The pump energy is distributed among the various dye cells as indicated in Figure 7.2.1. The pump beam coupled out by the first two quartz plates is used for pumping a dye laser setup. This makes use of two cascade dye lasers, two amplifier stages, and a gated saturable absorber (GSA) in between the amplifiers, resulting in an output pulse duration of ~8 ps at 365 nm. The purpose of the dye lasers is to sharpen the leading edge, while the GSA plus the saturated amplifiers cut the trailing part of the pulses. The first dye laser yields a pulse at 340 nm with ~100 ps risetime. The quenching mirror at the end of the first dye laser has no effect on the final pulse duration, but cuts the unwanted several ns long trailing component of the pulses already at the first stage and avoids eventual optical damage of the following stages by excessive pulse energies. The second dye laser has a ~10 ps risetime, and the 1 nm broad emission is centred at 365 nm. The gated saturable absorber consists of a solution of BBOT and a tilted cavity formed by two identical dichroic mirrors having approximately 10% reflectivity at 365 nm and 100 % reflectivity at 430 nm, acting at the same time as dye cell windows. The combined pulse forming effect of the oscillators, saturated amplifiers and the GSA results in an output pulse duration of ~8 ps at 365 nm, when 1 mm cavity length, with a solution of 1´10^-3 BBOT in ethanol is applied in the GSA. These pulses have an energy of typically ~4 mJ and are then used for pumping a distributed feedback dye laser master (DFDL) oscillator. The DFDL is a tuneable, achromatic arrangement, utilising a transmission grating, a microscope objective and a special dye cell. In the DFDL the interference fringes, which are necessary for DFDL operation, are created by imaging a coarse transmission grating onto the inner surface of the dye cell by the use of a microscope objective (Figure 7.2.1). The pump beam is first expanded by a cylindrical telescope to permit a pencil-like illumination of the grating and the active medium. Only the two diffracted first orders are used for the formation of the interference pattern, while the zeroth order is blocked by a stop. By proper choice of the distance between the grating and the objective, the wavelength of the DFDL is set to twice the KrF wavelength. The output pulse is then amplified in a two stage amplifier of standard design, reaching about ~20 mJ after the second amplifier. The output pulse is then frequency doubledin a 0.2 mm thick, 52° cut BBO crystal, with losses leading to a slightly divergent ultraviolet pulse with 2-3 mJ. This pulse is then amplified in a first pass through the second channel of the excimer laser filled with a KrF mixture. After the first pass the beam is magnified by a beam expanding telescope and spatially filtered by a pinhole. Then, the pulse is sent through the amplifier in a second pass, increasing the energy up to 10 mJ, with no more than 15% amplified spontaneous emission (ASE) background (nanosecond component). The cross section of the output beam is ~11´25 mm².

The experiments were performed on Al₂O₃-34 wt. % TiC ceramics. For comparison purposes, a few experiments were performed on targets of titanium carbide and alumina, containing 99.5 wt % of TiC and 96 wt. % of Al₂O₃, respectively. The laser beam was shaped by inserting in its path a square aperture, which was projected on the sample surface. A lens of 100 mm focal length in a 45:1 demagnification set-up was used to achieve a beam spot size at the sample surface of about 130´130 mm². For higher laser fluences, a 60 mm focal length lens was used, leading to a spot size of 80´80 mm². Machining of periodic line structures was performed using a Schwarzschild type reflective objective. This objective guarantees no front distortion for the femtosecond pulses and a high UV transmission. The objective has a nominal demagnification of 18´ and a numerical aperture of 0.3, leading to a resolution of »413 nm for 248 nm radiation. A transmission grating with 55 lines/mm was used as a mask. In order to increase the line density by a factor of 2, the zero order of the diffraction pattern was blocked. With this arrangement, a pattern of lines spaced 520 nm was created. The beam dimensions were reduced using a telescope in order to increase the laser fluence. The experiments were performed in vacuum (£10^-1 mbar) to prevent air breakdown at the focal point. The pulse energy was measured by pyroelectric detectors, and the fluence adjusted using a variable dielectric attenuator.

By using mode locking very short pulses can be achieved [Lasers and optical engineering,Das]. Active mode locking is achieved by forcing the longitudinal modes to maintain a fixed-phase relationship using a shutter inside the cavity (acousto-optic or electro-optic modulation). Mode locking can be obtained without any shutters by using a saturable absorber inside the cavity. This form of mode locking is called passive mode locking. Saturable absorbers, mostly dyes in solvents, absorb the light passing through them for low power, but pass light through with no attenuation for high power. Thus, a saturable absorber will act as a shutter. Since, in practice, the laser medium itself acts as a saturable absorber, for many cases the laser can mode lock by itself.

- Output duration after the DFDL

- Nanosecond trail percentage

FOR RESULTS, DISCUSSION, CONCLUSIONS AND BIBLIOGRAPHY CHECK THE THESIS MAIN PAGE