1. Introduction

Cutting and drilling of materials, especially metals, is well established in industry. Because laser processing is typically slower and more costly than conventional processes, it is typically used in high value applications where precision is required. These applications are important drivers for laser source development, motivating improvements in cost, throughput, and efficiency of short pulse lasers. The desire to improve the throughput and efficiency of laser drilling and cutting also drives the exploration of new approaches which optimize the laser material interactions involved in the ablation process itself. Throughput at a given power is related to the removal per laser pulse, and efficiency is related to the energy required to remove a given amount of material. Throughout this text, we will use Δ_p to designate the removal per pulse. We will use two quantities that reflect laser processing efficiency, the specific energy of ablation and the ablation efficiency. The specific energy of ablation, E_A, is the energy required to remove a given volume of material and has units of J/cm³; it is related to the laser energy required to perform a given processing task. It can also be easily compared to relevant thermodynamic quantities including the enthalpy of vaporization; for instance, the goal for a very “efficient” process could have the specific energy of ablation lower than the enthalpy of vaporization. We define the ablation efficiency, η, as the reciprocal of E_A. An efficient ablation process will naturally have a higher value of η. We will present η in units of μm³/μJ for easy comparison to recent laser processing work in literature. Some recent examples to improve removal per pulse and the ablation efficiency in drilling include use of double pulses [1–3], pulse bursts [4–7], and very high repetition rate low energy pulses [8].

When a laser pulse interacts with a metal surface, it vaporizes a thin (10 to 100 nm) layer of material leaving behind a melt layer. The energetic cost of removing material as a vapor is high, since much of the energy of the laser pulse goes into heating the vapor and into the enthalpy of vaporization which is in general high. For short pulses (picosecond to femtosecond), the remaining melt layer is also typically very shallow; it can be expelled by explosive boiling, although especially for higher laser fluences or when using shorter wavelengths [9], it can also be expelled by cavitation. The hot melt has low tensile strength and even weak tensile stress produced on the tail of laser-induced shock ejects (cavitates) the layer of melt [9,10]. Material removal by explosive boiling requires heating to the critical point temperature (near 10,000K) taking up to 10⁶ Joules to remove a cubic centimeter of material (Ε_Α = 10⁶ J/cm³), and is thus, an inefficient removal mechanism [11]. As we show in a recent study of picosecond ablation in metals and semiconductors [9], removal by cavitation of the melt can be more efficient because most of the material is ejected as a liquid droplets (E_A in the mid 10⁵ J/cm³ range), but ultimately, much energy is still wasted to excessive heating.

For longer pulses (nanosecond and longer) in which the laser intensity is much less than for the shorter pulses, vaporization and explosive boiling can be reduced and more energy goes to creating a deeper melt layer. The most efficient processes avoid vaporization, setting up conditions to remove material as a lower temperature melt. Such processes could have a specific energy of ablation lower than the enthalpy of vaporization. However, the melt is held tightly to the surrounding material by surface tension, and in deep channels created by high aspect ratio drilling, the melt can be very difficult to remove. This limitation can be overcome in some industrial processes through use of high-pressure gases, however, the effectiveness of this approach is reduced in high aspect ratio channels, for very fast drilling or cutting, for thick materials, or in cases where some stand-off is required such that the laser is not proximate to the surface. Instead of using a gas-assist, researchers have investigated use of a second laser pulse to expel the melt, e.g [1,2]. A key finding of our paper is the identification of novel conditions for optimal melt removal.

The initial motivation for double pulse drilling was to enhance melt ejection, but mechanisms for efficiency enhancement were not clear and only minimal pulse parameter optimization was discussed. In Lehane et al., for example, a dual-head Q-switched Nd:YAG laser was used to produce two pulses with adjustable timing: a 22.5 J ms-long laser pulse used to create a deep melt in stainless steel (SS) with little to no vaporization, and a second, 2.5 J 100 μs-long laser pulse timed to interact with the melt surface after a maximal melt depth was reached. The authors argued that the second pulse was intense enough to vaporize the melt surface, expelling the molten steel via the resulting recoil pressure at the liquid surface. The dynamics of melt ejection were not explored. They found that this double pulse combination could increase removal by up to a factor of about ten; based on the data reported, we estimate a reduction of the specific energy of ablation, E_A, to 1.5x10⁵ J/cm³ for a 550 μm diameter, 1 mm deep channel in SS corresponding to an ablation efficiency, η = 6.7 μm³/μJ.

Following this work, a number of researchers explored double-pulses using smaller lasers more appropriate for precision processing. For example, Forsman et al., investigated the removal process in SS and Aluminum (Al) using double pulses from the same 532 nm 3 ns laser (pulse separations up to 150 ns), also finding improvements to removal per pulse [2]. Based on the data reported, we estimate E_A = 7.5x10⁵ J/cm³ (η = 1.3 μm³/μJ) for SS, and E_A = 2.4x10⁵ J/cm³ (η = 4.2 μm³/μJ) for Al for 900 μm deep, 40 μm diameter channels using 2.4 mJ pulses (~100 J/cm²). While it is possible that some of the improvement reported could have been due to melt ejection, the authors ascribe the increased efficiency to the interaction of the second pulse with the ejecta plume from the first pulse. Wang et al., found that double pulses from the same 20 ns, 1064 nm laser also improved removal in SS compared to a single pulse of the same fluence [3]. Like Lehane, they ascribed the improvement to melt ejection by the second pulse. For this work, we estimate a reduction of removal energy to Ε_Α ~2x10⁵ J/cm³ (η ~5 μm³/μJ) for 1 mm deep, 50 μm diameter channels using 2 mJ pulses (~100 J/cm²).

Very recently, pulse bursts and the use of very high repetition rates in (≥ 1 MHz) have been employed to increase the processing rate and efficiency of short pulse (ps) lasers. Hu et al., found that a 5-pulse burst of 10 ps pulses separated by 20 ns increased the removal rate of shallow pits in copper (we estimate E_A ~4x10⁵ J/cm³ with η ~2.5 μm³/μJ); they propose a mechanism which like melt ejection relies on residual heated material from a previous pulse but also involves enhanced hot electron diffusion rather than explicit melt ejection [4]. Recent pulse burst studies from [6] give similar results with η = 4.9 μm³/μJ in Si and 2.3 μm³/μJ for Cu, while burst mode studies from [7] report η = 2.5 μm³/μJ in Cu, all using 1 μm light. Kerse et al. have explored laser ablation using ultrafast bursts (GHz rate) of 800 fs pulses separated by less than a nanosecond to achieve efficient ablation with very low residual heating in the workpiece (“ablation cooled” removal) [5]. Based on their results, we estimate their best values of specific energy of ablation at, E_A = 10⁵ and 2x10⁵ J/cm³ (η = 10 μm³/μJ and 5 μm³/μJ) respectively for small (25 μm diameter deep) single burst ablation pits in Cu and Si. Like the work of Hu et al., this burst mode approach involves heat accumulation between bursts, but ultimately, removal occurs at high temperatures near the critical point. Finger et al. explicitly invoke residual heat and melt ejection to explain increased drilling efficiency in making ~80 μm deep pits in SS at MHz rep-rates (10 ps 1064 nm pulses) [8]. Here, E_A = 3x10⁴ J/cm³ (33 μm³/μJ) is quite good, however, it can’t be directly compared to the values for the much deeper channels described above since removal rates drop dramatically with channel depths and increased aspect ratios. The same is true for the removal energies quoted above for the small, shallow pits in Kerse. (Most approaches using very high repetition rates or ultra-fast pulse bursts operate with very small pulse energies, small spot sizes, and correspondingly small ablation pits; it is difficult to create deep, high aspect ratio channels under these conditions).

Here, we present a multi-laser (ML), double pulse approach which maximizes removal rate and removal efficiency for deep (up to multi-mm) hole drilling in metals – SS and Al – and in carbon fiber composites. Like [1], we employ a long heating pulse to create the melt, but instead of using a Q-switched laser, we employ a compact, high wall-plug efficiency gated CW fiber laser. More importantly, instead of using a nanosecond pulse (or longer) for the second pulse as in the previous double-pulse studies, we use a short pulse (20 ps) laser to generate high-impulse melt ejection. In particular, we chose to use UV pulses (355 nm = 3ω, frequency converted from 1064 nm = 1ω) for this work because we have found in previous studies [9] that in the picosecond regime, UV (355 nm) pulses couple better to the melt, providing greater hydrodynamic coupling (impulse) and concomitantly greater removal (more than 3 times greater) compared to IR (1064 nm) pulses for SS, Al and Si [9]; peak ablation efficiencies (minimum E_A) was also achieved in [9] for 355 nm light and for laser fluences above 10 J/cm², conditions which maximize hydrodynamic impulse and melt cavitation for a single 20 picosecond laser pulse. The power and pulse energy delivered by the UV picosecond laser is more than an order of magnitude lower than our CW fiber laser. We present a comprehensive study of the removal process using high-impulse ML ablation including a broad parametric investigation of ablation rate (fluence, power, material, spot size, and sample thickness). We compare the ML removal rate using a picosecond laser (high impulse) to that using a nanosecond laser (lower impulse), and together with the high-speed video images and multi-physics hydrodynamic simulations of the melt ejection process, provide insight into the removal mechanisms. X-ray computed-tomography (CT) images of through-holes were obtained to determine the shape, quality, and removal volume of the holes. We find that efficient removal by the short pulse laser is due to the high-impulse excitation of both melt cavitation and surface wave-induced liquid jet ejection. Removal rates and efficiency for this multi-laser approach are more than an order of magnitude greater than for either the pulsed laser or CW fiber laser, either operated in gated or true CW mode. Our approach exceeds the efficiency of those described above, with η as high as 100 μm³/μJ for SS and 20 μm³/μJ for Al for approximately 1 mm deep 10:1 aspect ratio channels. As expected, the average specific energy of ablation needed to create these channels, 10⁴ J/cm³ for SS and 5x10⁴ J/cm³ for Al, is lower than the enthaply of vaporization adjusted for reflection. Similar improvements are also found for high-impulse melt ejection in carbon fiber composites.

2. Experimental methods and materials

2.1. Multi-laser experimental layout and hole-drilling experiments

The multi-laser drilling system consists of a pulsed laser and CW laser as shown in the schematic in Fig. 1. In this study, we have used either a picosecond (ps) or nanosecond (ns) pulsed YAG laser, and a gated CW fiber laser (IPG photonics). Both of the pulsed YAG lasers operate at 355 nm (3ω). The ~20 ps laser (Ekspla) has a variable repetition rate up to 50 Hz and a pulse energy up to 14 mJ. The repetition rate of the 7 ns laser (Bigsky) is up to 10 Hz and the pulse energy is up to 70 mJ. The 500W CW laser (IPG Photonics) operates at 1064 nm (1ω) and is synchronized to the pulsed laser. We vary the gate width from 25 μs to 400 μs in this study effectively using it as a very long pulse laser. This setup allows the generation of a pulse train consisting of a long heating CW pulse followed by a short ejection pulse timed at the end of the gated CW pulse as shown in Fig. 2. For hole-drilling experiments, these two lasers are focused with separate lenses and then combined colinearly using a dichroic mirror onto the input surface of a sample. We determine the 1/e² diameter of the laser spot by measuring the single shot laser damage spot as a function of the pulse energy at the sample since the area of the damage scales linearly with the natural log of the total pulse energy [12]. The pulse energy and spot size of both lasers can be independently varied in our study. Two photodiodes are placed on the exit side of the sample; they begin recording at the start of laser drilling experiments and detect laser light from either of the two lasers marking the onset of sample drill-through. Since the sample thicknesses are often varied in our study, we define an average removal rate per pulse Δ_p (μm/pulse) as the sample thickness divided by the number of pulses required to drill through the sample.

 

Fig. 1 Experimental layout for multi-laser drilling. The 3ω short pulse laser (with either 20 ps or 7 ns pulses) is combined with a gated CW laser at 1ω and focused onto the input surface of a sample. Drill through is determined by detecting the onset of laser light at either wavelength using photodiodes on the exit side of the sample. A 810 nm laser is set up to provide illumination for fast video shadowgraph of the ejected material. Transmitted light through drilled channel can monitored using a CCD camera.

Download Full Size | PDF

 

Fig. 2 Multi-laser double pulse format. A long CW heating pulse at 1ω is followed by a short picosecond or nanosecond ejection pulse at 3ω.

Download Full Size | PDF

2.2. Samples

Commercially available 316L stainless steel (SS), 6061 aluminum (Al), and 110 copper (Cu) sheets of varying thicknesses (up to ~3 mm) were purchased (McMaster-Carr) and cut to size for the experiment. We also tested laser drilling of 3-mm thick engineered twill prepreg carbon fiber composite sheets. These samples are as received in terms of surface finish.

2.3. X-ray computed-tomography images of through holes

The samples were scanned using a commercial micro Computed Tomography (CT) system, Zeiss Xradia 510 Versa. This system has an x-ray tube with a tungsten target anode, which operates between 30 and 160 kV with maximum power between 2 and 10 W depending on the voltage. The spot size varies with the amount of power selected ranging from 2 to 4μm FWHM. The CT geometry is cone beam with a maximum angle of 34°. The CT setup for the Aluminum alloy sample was 80 kV, 7W and LE#6 filter while the stainless steel was scanned using 160 kV, 10 W and a HE# 6 filter. The voxel size was approximately 3 μm using the 4X lens and minimal geometric magnification on a 2.5 x 2.5k CCD with a physical pixel size of 13.5 μm. CT images were used to determine the general shape of channels drilled and to provide a quantitative estimate for the volume of material removed under various conditions.

2.4. High-speed imaging of ejecta from ablation and hole drilling

The dynamics of the ejection of ablated material during laser hole drilling was captured using high-speed video imaging to help determine material removal mechanisms. A diode laser (Cavilux) was synchronized with the ablation laser to provide the illumination from the side of the sample perpendicular to the ablation laser beam for shadowgraph imaging. The high-speed video camera (Shimadzu) was set up on the opposite side of the sample to image the ejected material using a 10x objective lens (Mitutoyo). The camera operates at a rate up to 10 M fps records and up to 256 frames can be recorded allowing us to capture motion of the ejecta from a given single laser shot during the drilling process in our study. Ablation videos from both SS and Al samples were recorded at varying laser conditions.

2.5. Multi-physics hydrodynamic simulation

In order to investigate the relevant physical processes that govern multi-pulse laser ablation, laser-material interaction simulations were performed using LLNL’s multi-physics hydrocodes HYDRA [13] and ALE3D [14]. HYDRA is an arbitrary Lagrangian-Eulerian (ALE) finite element code that couples a ray-tracing laser-plasma interaction (LPI) model with hydrodynamics and thermal diffusion solvers. Nanometer-scale spatially resolved HYDRA simulations were used to investigate laser energy deposition and laser-driven shock loading of the melt over the timescale of nanoseconds. Of particular interest was whether pulses of picosecond and nanosecond width could drive melt ejection through cavitation, as observed in [9]. The HYDRA simulations also provided the impulsive drive conditions for three-dimensional, microsecond-scale ALE3D simulations of the ensuing melt hydrodynamics, which were intended to investigate material ejection through cavitation, splashing, and out-jetting. ALE3D is an ALE finite element code with coupled hydrodynamics and thermal diffusion solvers that provides dynamic interface tracking for modeling temperature-dependent, non-uniform surface tension effects. The objective of the modeling effort was to provide physical insight into the mechanisms of material removal during picosecond and nanosecond pulsed laser irradiation of a melt pool, and to study how those mechanisms evolve during channel drilling.

2.6. Channel light propagation

Light propagation through drilled channels was investigated by imaging the transmitted beam using a CCD camera. The camera was set up to image the laser beam at the exit plane of the sample in order to determine the local fluence. Laser beam transmitted through a drilled channel was imaged and compared with the reference beam without a sample to assess laser energy loss to the channel walls for different sample thicknesses and channel aspect ratio.

3. Experimental – parametric studies

3.1. Parametric studies in thin (250 μm thick) stainless steel (SS) and aluminum (Al)

Figure 3 shows the average thickness removed per laser pulse in drilling through 250 μm thick plates of SS steel (Fig. 3(a)) and Al (Fig. 3(b)) as a function of short pulse laser fluence using the multi-laser approach (laser conditions shown in the Figs.). The average removal rate Δ_p was calculated as the sample thickness divided by the pulse numbers required to drill through. The width of the gated CW pulses is 100 μs. For the short pulse laser alone (no CW laser power), the average removal rate Δ_p grows rapidly for fluences between 5 and 10 J/cm² reaching ~0.2 μm/pulse for SS and ~3 μm/pulse for Al at 30 J/cm².

 

Fig. 3 Average removal per pulse Δ_p for 250 μm thick (a) SS and (b) Al plates using the multi-laser approach. In both cases, the drill-through rate for a given pulsed laser fluence increase dramatically as the gated-CW laser exceeds the power levels required to form a melt in each material (determined by SEM studies). Temperature estimates based on an analytic model are shown for some power levels.

Download Full Size | PDF

The power of the CW laser needed to produce melt depends on the sample material. We took SEM images of the CW only laser irradiated sites at varying laser power. The surface of the sample starts to change and show signs of material flow and re-solidify once the CW power is high enough to create melt shown in Fig. 4. Based on the morphology, we estimate that the thresholds for melt pool formation for 100 μs CW pulses are ~50W for SS and ~300W for Al. At lower power levels, no evidence of re-solidification is seen. At higher levels – 100W and 470W respectively – clear evaporation pits are seen.

 

Fig. 4 SEM images of the onset of melt formation. (a) SS with 100 μs 50 W CW irradiation and (b) Al with 100 μs 300 W CW irradiation.

Download Full Size | PDF

Simple estimates of the threshold power (F_TH) for melting can be obtained assuming temperature independent thermophysical parameters and a one-dimensional model. The absorbed laser energy AF_th is spent to heat up to the melt temperature T_m the slab with thermal diffusion length √Dτ [15,16].

The threshold estimates for melt pool formation based on the morphology for 100 μs CW pulses are not very different from the prediction of our crude model. At lower power levels, no evidence of re-solidification is seen. At higher levels – 100W and 470W respectively – clear evaporation pits are seen. The temperature in the metal as a function of depth z within the simple one-dimensional model (see details in [15], pp. 47) can be estimated as,

A great enhancement in rate is seen when the two lasers work together as seen in Fig. 3. When the CW laser power is above the melt threshold, the average removal rate Δ_p for multi-laser drilling improves significantly for both materials. We observe a greater than 10x enhancement for drilling SS and a 5x enhancement for Al using the two lasers together over the average pulsed laser removal rate Δ_p alone for these 250 μm thick samples. The enhancement due to this multi-laser approach is seen most strongly up to a 20 J/cm² pulsed laser fluence. The removal rate enhancement begins to saturate beyond 20 J/cm².

3.2. Sample thickness dependence on hole drilling

3.2.1. Hole drilling with a 160 μm diameter spot

Figure 5 shows the average removal rate for the multi-laser approach using a 160 μm 1/e² diameter 20 picosecond laser spot combined with a 100 μs gated CW laser compared to the removal rate for the gated CW laser alone and the 20 ps short pulse laser alone. The 1/e² diameter of the CW laser spot is ~320 μm. For all sample thicknesses less than 2 mm, the removal rate using the multi-laser approach is much greater than the removal using either laser alone. For thicknesses less than 1 mm, removal rates of the multi-laser approach are more than 10x higher. We find similar improvements for Al samples.

 

Fig. 5 Average removal rate Δp for the multi-laser high-impulse format as a function of thickness for SS. Here, the diameter of the picosecond laser spot (1/e²) is 160 μm and the diameter of the CW laser spot is 320 μm.

Download Full Size | PDF

3.2.2. Spot size dependence and the role of light propagation through the channel

The average removal rates Δ_p of all three modes (pulsed laser alone, CW laser alone, and multi-laser mode) begin to diminish as a function of thickness as shown in Fig. 5; channel depth impacts the multi-laser approach the most with the improvement in removal rate diminishing rapidly beyond 1 mm, disappearing for thicknesses greater than 2 mm. This could be due to at least two reasons: either ejecta produced by the multi-laser approach is easy trapped in the drilling channel, or the light from one or both laser sources is absorbed by the channel walls [16]–problems which become worse for higher aspect ratio, deeper channels. Since enhancements in the multi-laser approach depend strongly on the strength of both sources as shown in Fig. 3, it is reasonable to expect that absorption in the channel walls will, as seen in Fig. 5, affect the multi-laser approach more strongly than it does either source alone. To test this hypothesis, we imaged the fluence of the pulsed laser light at the exit-plane of the channel as a function of sample thickness in Al as seen in Fig. 6 and found that the relative fluence of the laser light at the exit surface of the sample decreases strongly with increasing sample thicknesses, dropping significantly for samples 1 mm or thicker, consistent with the observed reduction in the average removal rate Δ_p shown in Fig. 5. Absorption in the channel walls will affect both the gated CW melt laser and the short pulse impulse laser. A reduction of CW laser intensity with channel depth would decrease the melt pool depth and reduce the removal per pulse as argued in section 3.3, which would reduce removal for deeper channels. A reduction of the pulsed laser intensity with channel depth would reduce the impulse to drive out the melt.

 

Fig. 6 Cross-sectional line out of the short pulse only laser fluence measured through laser drilled channels at the exit plane of the samples (hole center ~125 μm) as a function of sample thicknesses/channel length. A Gaussian function is used to fit the beam profiles.

Download Full Size | PDF

One solution to reduced absorption in the channel walls is to increase the diameter of the channel, thus reducing the aspect ratio of the hole. Since the 160 μm diameter pulsed laser spot is smaller than the CW laser spot, it is likely that the pulsed laser sets the diameter of the hole. To test this, we increased the size of the pulsed laser spot from 160 μm to 210 μm and confirmed that the hole dimeter increases from approximately 180 μm to 220 μm shown in the images of Al samples with laser drilled channels in Fig. 7. The size of the channel diameter also increases on the exit side. Figure 8 shows the average removal rate Δ_p as a function of sample thickness for channels drilled using 160 μm, 210 μm, and 315 μm diameter (1/e²) pulsed laser spots for Al and SS. The removal rates for large and small spot sizes are very similar for the thinner samples which are not affected as strongly by absorption in the channel walls, but removal rates improve dramatically for the deeper channels using the larger 210 μm and 315 μm diameter pulsed laser spot. Hence, we expect that further modest increases in the channel diameter (requiring increased pulsed laser energy) will result in enhanced removal for samples thicker than 3 mm. It is also possible that increasing the CW laser spot size or power could mitigate reduction of the melt pool depth for deeper channels and increase the removal rate.

 

Fig. 7 Microscope images of the input and exit surfaces of channels produced with ML approach in Al. The ps-laser fluence is kept at 30 J/cm² while beam spot size varies from 160 μm to 210 μm. CW beam conditions are the same for both pulsed beam sizes – 320 μm spot size, 450 W power and 100 μs gate width.

Download Full Size | PDF

 

Fig. 8 Improvements in the removal rate per pulse Δ_p using the larger pulsed laser spot (315 μm and 210 μm) compared to a 160 μm pulsed laser spot (a) Aluminum, (b) stainless steel. The 355 nm short pulse laser pulse duration was 20 ps. The CW laser power was 450 W, the CW pulse was 100 μs long, and the spot size was fixed at 320 μm.

Download Full Size | PDF

This multi-laser drilling approach was also applied to carbon fiber composite material which has been known to be very different to cut using a laser. Using the short pulse laser at 30 J/cm² combined with the 450W CW laser beam with a 100 μs gate width, channels are produced in carbon fiber composite plates 3 mm thick. The removal rates improved significantly compared with short pulse only and CW only ablation, and are comparable to that for Al and SS.

3.3. Removal as a function of gated CW pulse length

It is clear that an optimal CW laser gate length (t_G) exists for the long melt-pulse; if pulse duration is too short, the melt is shallow, and the high impulse of ps-pulse is not used efficiently. If t_G is too long, the melt depth is deeper than the ps-pulse is able to eject. Changes in t_G are expected to both affect the peak melt temperature and the melt depth through (see relationships in Section 3.1). Figure 9 shows the ML average removal rate Δ_p for 760 μm SS and Al plates as a function of t_G combined with a short pulse laser at a fluence of 30 J/cm² and compares them to gated CW ablation using the same CW laser parameters.

 

Fig. 9 Multi-laser average removal rate per pulse Δ_p for 760 μm SS and Al plates as a function of CW laser gated pulse length using a pulsed laser fluence of 30 J/cm². Dashed fit lines are shown for proportionality to the gated pulse length, t_G for evaporative CW-only removal, and to the sqrt(t_G) for the ML removal process.

Download Full Size | PDF

It is natural that the CW laser-only average removal increases with gated pulse duration (t_G) since the removal per pulse (R_p) in this case is increases with the surface temperature (T_S (t_G)) dependent evaporation rate, R_evap(T_S (t_G)), and the duration of the pulse: R_p ∝ t_G *R_evap(T_S (t_G)). In fact, for t_G > 100 μs for Al and t_G > 50 μs for SS, R_p is proportional to t_G shown in Fig. 8 as dashed fit lines, which indicates that over these time ranges, the surface temperature, T_S ~T_S(t_G), is approximately constant.

3.3.1. Melt depth plays an important role in ML enhancement

The removal enhancement for the ML approach is not proportional to t_G, since ML removal is not driven by evaporation but by the ML short-pulse-driven enhancement. While the peak surface temperature is insensitive to t_G under these conditions, the melt depth should still increase with sqrt(t_G) in correspondence with the thermal diffusion depth (e.g., see Eq. (3)). The removal rate for Al and SS is correlated to the melt depth, also increasing by a factor roughly proportional to sqrt(t_G) (see Fig. 9 dashed fit lines): ML removal rates increase by a factor of about 2 (1.8 for Al and 2.0 for SS) going from t_G = 25 μs to 100 μs, and by another factor of two (1.7 for Al and 2.7 for SS) going from 100 μs to 400 μs. This correlation suggests that melt depth plays an important role in the ML enhancement. While it is possible that ML approach here removes the entire melt layer, it is also possible that the removal tracks melt depth. It is clear that the amount of ML removal cannot exceed the melt depth. The entire melt depth is clearly not always removed by the pulsed-laser when the ML removal rate is a function of pulsed-laser fluence for a fixed melt depth (fixed t_G and CW laser power) as for the thin samples of Fig. 3, Finally, note that melt depth is not constant as laser drilling proceeds deeper in a sample for a given set of laser parameters. As we have seen, as laser-drilled channels get deeper and the aspect ratio increases, fluence and irradiance decrease; this results in a shallower melt depth which can contribute to the reduced average removal rate Δ_p and drilling efficiency for thicker samples as seen in Fig. 8.

4. Multi-laser channel shape and quality

X-ray computed tomography was used to inspect the laser drilled channels. Cross-sectional views of typical channels made in a 2-mm thick Al plate produced using multi-laser approach (Fig. 10(a)), short pulse only (Fig. 10(b)), and gated CW only (Fig. 10(c)) are shown below. The ML drilled channel walls appear smooth with little collateral damage visible. The channel produced with short pulse laser only is also good quality in general, although the drill-through rate was roughly ten times slower than in Fig. 10(a). The rate using the gated CW alone was much slower than the other two. A 3D rendering of the channel in Fig. 10(a) using the ML approach is shown in Fig. 10(d) showing the smooth, slowly tapering walls. The tapering is likely a by-product of wall absorption as described below. Channel quality in the SS samples is similar to that shown here for Al. We use the full 3D-rendered CT scans of holes later to accurately compute the volume of material removed under various conditions.

 

Fig. 10 Computed tomography of channels drilled in 2 mm thick Al: cross-sectional views using (a) multi-laser; (b) ps laser alone; (c) gated CW laser alone (note, the vertical and horizontal axis scales are different as shown in the Figs). The depth to width aspect ratio of the laser-drilled hole in 10a is greater than 10:1. (d) shows a 3D rendered image corresponding to the hole drilled using the multi-laser approach in (a) (vertical and horizontal axis scale are different as shown in the Figs). The pulsed laser spot was 160 μm 1/e², and the CW laser spot was 320 μm. Note, the channel drill-through in (b) took about ten times as long to complete compared to the channel drill-through using the ML approach in (a); we did not complete the drill-through using the gated CW laser alone as the rate was much slower than either of the other two approaches.

Download Full Size | PDF

There are two non-ideal aspects associated with the region around the rim of the hole on the input side when using the ML approach as described here: there is some minor build-up of ejecta around the rim shown in Fig. 10(d) and Fig. 7, and there is a discoloration surrounding that shown in Fig. 7. It will be shown later that the small ejecta mound occurs on the first shot; we believe this can be greatly suppressed or eliminated by using the pulsed laser alone for the first several shots. The discoloration is due to very shallow ablation by the low fluence wings of the pulsed laser and to surface heating (oxidation) from the lower intensity wings of the gated CW laser beam. Both can be reduced by improvements to the beam quality. The later can be reduced by using shorted CW laser pulses – e.g., by reducing the gate from 100 μs to 25 μs as was done in Fig. 9.

5. Energy requirements and efficiency

Laser parameters used in drilling studies published in literature can vary significantly making it difficult to compare drilling rate. One meaningful metric is to calculate the total energy needed to produce a channel through a sample of a given thickness – the energy to penetrate. Another is to compute the laser energy per unit volume removed, which as we defined earlier, is the specific energy of ablation, E_A. In both cases, comparison should be made for the same thickness of sample as the depth of the channel and hole aspect ratio dramatically impact removal rate due to wall absorption. For the same reasons, comparisons should ideally be made using the same laser spot size (channel width) as well.

5.1. Pulsed laser excitation as an assist to pulsed CW laser ablation

Using the first metric, we compare the total CW energy to drill through samples of a given thickness for the CW laser alone and for the ML approach. Here, we view the ML approach as a pulsed laser assist to CW laser drilling. Note, the pulsed laser energy for these cases is much less than the CW laser energy, so this is tantamount to a total energy metric. The results are shown in Fig. 11(a). For both Al and SS, the pulsed laser dramatically reduces the CW laser energy (and total energy) required to drill through samples up to 2 mm thick (SS) and beyond 3 mm thick (Al). Figure 11(a) also includes data for a “true” CW laser drill through for 1.5 mm thick samples – the CW laser is run in continuous mode with no gating but with the same spot size as using for the other approaches. In both SS and Al, lower bounds on the required energy are shown; the experiment was stopped before a complete channel formed because the samples heated up significantly during the drilling test and started to warp. Using a pulsed laser to assist CW laser drilling greatly increases throughput and reduces total energy.

 

Fig. 11 Total energy required to create channels as a function of thickness for the multi-laser approach compared to that for gated CW laser alone, true CW laser (see text), and the 20 ps 355 nm laser alone: (a) CW laser drill through energy (pulsed laser as an assist to the CW laser), also includes lower bound for energy required to drill through using an un-gated “true” CW laser exposure; (b) Pulsed laser drill-through energy (gated CW laser as an assist to the pulsed laser).

Download Full Size | PDF

Higher process efficiency greatly increases the quality of the ML drilling compare with CW processing. More deposited energy in this situation rsults in increasing heat affected zone and the thermally induced cracks around the hole [18].

5.2. CW laser heating as an assist to pulsed laser ablation

Figure 11(b) shows the total drill-through energy from the point of view of using the CW laser as an assist to the pulsed laser. The CW laser reduces the required pulsed laser energy more than an order of magnitude for all cases. This improvement is significant from the system point of view. Pulsed laser with similar power to a CW laser are both much more expensive and much larger. Hence, using the CW laser for a ML assist, can dramatically reduce system cost and size.

5.3. ps-ML leads to unprecedented removal efficiencies

The second metric, the specific energy of ablation, E_A, is the best way to compare the efficiency of different approaches. The CT images of the channels allow us to calculate the total volume of material removed by treating the channel as a truncated cone. In Fig. 12(a), we compute this metric by dividing the total laser energy required to drill through 760 μm thick SS and Al samples as a function of the CW laser gated pulse length. Here, E_A is compared for the ML approach to gated CW laser drilling alone, and in all cases, the ML approach is about 100x more efficient. In Fig. 12(a) we also compare E_A for the ML approach to the enthalpy of vaporization per unit volume, the energy required to take each material from liquid to vapor. For this comparison, we have corrected the enthalpy of vaporization by the reflectivity, R, of these metals for our CW laser conditions (1 μm wavelength), since almost all of the energy used for removal is CW laser energy; we have used R = 0.5 for SS and R = 0.9 for Al. Ablative approaches which remove material in the vapor state will require at least this much energy. Note that for both materials and all values of t_G, E_A is significantly smaller than the reflectivity corrected enthalpy of vaporization, confirming that the ML approach removes material in the liquid state (note that E_A for the gated CW laser alone cases is much higher than the enthalpy of vaporization as expected). In Fig. 12(b), we plot the same data as removal efficiency in μm³/μJ, the reciprocal of E_A. To the best of our knowledge, the high-impulse ML approach demonstrated here leads to better removal efficiencies than the best approaches in the literature, both double pulse approaches, for channels through Al (see Forsman et al. [2],) with η = 4.2 μm³/μJ and SS (see Lehane et al [1],) with η = 6.7 μm³/μJ through samples of similar thickness (1 mm and 900 μm respectively).

 

Fig. 12 The average specific energy of ablation (a) and corresponding removal efficiency (b) for multi-laser (ML) laser drilling through 760 μm thick plates of Al and SS as a function of gated CW pulse duration, t_G along with comparisons to gated CW laser drilling alone. Figure 12(a) shows that ML specific energy of ablation is lower than the vaporization enthalpy corrected for reflection losses indicating that most of the ML removal occurs as liquid droplets, not vapor.

Download Full Size | PDF

The reduction of E_A and increase of efficiency as the gated pulse length, t_G, is reduced can be understood as follows: the total laser energy used for material removal is proportional to t_G, but the removal rate per pulse Δ_p is roughly proportional to sqrt(t_G) (see Section 3.3 and Fig. 9), so energy per removed volume must go as sqrt(t_G), decreasing as t_G is reduced. This is confirmed in Fig. 12 as the CW laser pulse gate time is reduced from 400 μs to 25 μs, the ML E_A drops by a factor of 4 for Al and 2.2 for SS. Ultimately, that wasted energy goes into undesirable heating of the sample with an increased probability of collateral damage as undesirable heat diffusion occurs both deeper below the melt pool and laterally around it as the gate pulse length is increased. We expect that a more optimal ML protocol would use a higher power CW laser and a shorted gated pulse length. The optimal ML strategy for a given problem requires understanding the physics of material remove in this regime.

6. Removal mechanisms

In our double pulse experiments, the CW laser creates a melt pool on the surface of the substrate, and the short pulse 355 nm laser provides a pressure impulse. The high irradiance of the short pulse laser rapidly heats the melt and produces a shock wave which propagates into the liquid and a resulting recoil pressure on the surface. A release wave follows the shock wave providing a source of tension in the melt. Under certain conditions, the shock can also reflect from the melt-solid interface at the bottom of the melt pool producing a counter-propagating wave which can also result in tension, either at the melt-solid interface or in reflection from the free surface. The melt is unstable to tension; removal by cavitation can follow under these conditions as the liquid under tension nucleates vapor voids which separate and remove a thin layer of liquid from the surface of the sample (see [9,10,19]). The recoil pressure created by the ps-laser pulse can also excite surface waves (surface tension waves in the melt) in the melt [18]. Laser-induced surface waves can eject material in two ways: corona ejection and liquid jet ejection. As the surface wave propagates from the center of the melt out to the perimeter, it can overshoot, shedding liquid droplets in a circular corona. As the surface tension pulls the wave back towards the center, the liquid can gather up forming a liquid jet which necks-off from the bulk of the liquid and leaves the sample. We believe we see evidence of both ejection mechanisms in the high-speed video of ML ablation (Section 6.1). Examples of these removal mechanisms are also seen in hydrodynamic simulations in Section 6.3).

6.1. The nature of ejecta

High-speed video is used to capture the shadowgraph of ejected material during multi-laser drilling. There are two types of ejecta usually observed on the first shot. Material can be seen ejected from the sample surface as a sheet tumbling in air as shown in top time sequence in Fig. 13. We believe this is the result of the cavitation of a melt layer. We also see examples of surface wave excitation in the middle panel of Fig. 13 which forms a crown at the perimeter of the laser ablation crater and then breaks into droplets which are ejected from the sample later in time. This type of corona ejection is most often seen on one of the first shots in drilling a channel. Material ejection from subsequent shots after a partial channel has been formed appears as droplets from the center of the hole as shown in the bottom panel of Fig. 13. Here, we see a canonical example of surface-wave induce liquid-jet ejection.

 

Fig. 13 Frames from high-speed video images of ejecta during laser drilling channel formation showing both ejecta types: melt ejection by cavitation and melt ejection through the formation of a liquid jet via surface wave excitation. Panels 1 (cavitation) and 2 (surface-wave corona droplet ejection) are from the first laser shot, and panel 3 (surface-wave liquid-jet ejection) is from the second laser shot.

Download Full Size | PDF

Fast video images of later laser shots usually show much less visible ejecta coming out of the partially drilled channel. Beyond shot number 5, noticeable ejecta is rarely seen in the video, although the CT images of the channels clearly indicate the depth of channels are continuing to increase, and in fact, they increase in a continuous way – the removal from shot-to-shot is not episodic but continuous. The volume of removed material is shown in Fig. 14(b) estimated from both the CT-rendering of partially drilled holes and estimated from high speed video images. Both estimation methods are qualitatively similar for the first 5 shots, but beyond that, there is no noticeable removal from the video images. The spatial resolution of the video is ~3 μm. It is possible that during later shots, material ejected through either cavitation or surface wave motion is broken into sub-micron sized droplets before exiting the channel making it difficult to detect on the camera.

 

Fig. 14 (a) Average removal depth as a function of laser shot number measured from cross-sectional CT images of ML laser drilled channels in SS (inset) using 20 ps, 30 J/cm² laser pulses combined with 450 W CW laser with a 100 μs gate width. (b) Volume of removed material of a single laser shot as a function of shot number estimated from the CT images and fast video frames.

Download Full Size | PDF

6.2. Direct comparison of removal with high-impulse (ps pulses) and low-impulse (ns pulses) laser drivers

To help understand the nature of the ejection process, we compared the average removal rate Δ_p as a function of the length of the short melt-excitation pulse – in particular, we compared melt ejection using our 20 ps laser pulse with removal using a 7 ns laser pulse. The short-pulse wavelength (355 nm) and fluence (30 J/cm²) were kept the same, as was the CW laser spot size, t_G, and power. The ps-laser spot and the ns-laser spot were also matched at 315 μm and 300 μm respectively. Thus, the only difference between the ps-ML and the ns-ML was the pulse length of the short pulse laser – 20 ps versus 7 ns (see Table 1). Figure 15 compares the average ML removal rate per pulse Δ_p using the 7 ns pulsed laser to the ML enhancement using the 20 ps laser as described above along with the CW only and ps-laser only removal rates for SS samples up to 1500 μm thick. The effect of the 7 ns laser excitation is seen for the thinnest sample (250 μm), but rapidly disappears for samples 500 μm thick and above for which the ns-ML removal converges to the CW laser only case.

Table 1. Comparison of ps-ML and ns-ML showing short-pulse (SP) laser parameters and average removal rate in SS along with the relative estimated ablation pressure and total impulse. The CW laser parameters were the same for both methods (1064 nm, 320 μm spot, t_G = 100 μs). The much higher removal rates for ps-ML drilling correlate to its much higher ablation pressure (impulse rate).

View Table

 

Fig. 15 Average removal per pulse Δ_p for the multi-laser approach as a function of material thickness comparing excitation by a 355 nm picosecond laser (20 ps) and a nanosecond laser (7 ns). Average removal rates for the CW laser alone, the ps-laser alone, and the ns-laser alone are shown for comparison.

Download Full Size | PDF

The main factor which separates ps-laser excitation and ns-laser excitation of the melt is the peak ablation pressure (impulse rate). A simple model, well-tested over a broad range of laser parameters, has been found to reproduce instantaneous ablation pressure, P_a, imparted to the melt, and the total impulse density, J_tot = P_a*t_p, resulting from the application of that pressure over the duration of the laser pulse [20]. According to the Phipps model, P_a ~I_p^0.7 t_p^-0.15, where I_p is the laser intensity; I_p ~Φ_p / t_p, where Φ_p is the pulsed laser fluence. Using these relationships, P_a ~Φ_p^0.7 t_p^-0.85. In the comparison in Fig. 15, Φ_p for both the ns-laser and ps-laser excitations was set to 30 J/cm² so P_a(20 ps) should be about 145 times (7 ns / 20 ps)^0.85 larger than P_a(7 ns). The peak pressure is more than 100x greater for the ps-laser pulse, and as we will see in Section 6.3, the tension generated to initiate cavitation is much higher. The total impulse density is approximately, J_tot ~Φ_p^0.7 t_p^-0.85 t_p = Φ_p^0.7 t_p^0.15, so that J_tot(20 ps) is only about 2.4x larger than J_tot(7 ns). The results of this analysis summarized in Table 1. This observation helps to clarify the dominant removal mechanism.

Cavitation requires tension, and the magnitude of tension produced is strongly related to the peak shock pressure related to the ablation pressure (P_a), so it is not surprising that the hydrodynamic simulations in Section 6.3 show cavitation only occurs for the 20 ps-pulse. The differences seen in the experimental data of Fig. 15 could be explained by the loss of cavitation for longer pulse durations. Because large-scale fluid motion associated with surface waves occurs on time-scales much longer (microseconds) than either of the pulse lengths here (Section 6.3), surface waves respond to the total impulse delivered (J_tot), not to the ablation pressure. Therefore, we would expect that removal rates by surface wave ejection would be similar for both pulse lengths, which contradicts the experimental evidence in Fig. 15. Based on the arguments here, we conclude that cavitation is more likely to be the dominant removal mechanism, particularly for thicker samples after the several shots where for the ns-pulse excitation, the ML enhancement disappears. This is also consistent with the observation that channel and exit hole diameters scale with the short-pulse laser spot-size as shown in Fig. 7. Cavitation is a local removal process; a liquid sheet separates from the bulk of the liquid only where tension is applied, and that happens only where the peak pressure/impulse is high (the laser spot where the pulsed laser energy is absorbed). Surface wave ejection can be non-local; fluid motion occurs throughout the liquid, not only where the total laser impulse is delivered, so that the resulting removal area can extend beyond where the pulsed laser energy is absorbed as the liquid redistributes and re-solidifies following the pulse.

While cavitation is likely the dominant mechanism, especially when the removal comes from deeper in the channel, we have clear evidence from the high-speed video that surface wave ejection also occurs. This removal mechanism is only obvious for the first few shots before the channel becomes deep. The fact that we no longer see obvious material removal after about the fifth shot as shown in Fig. 14 is consistent with removal by cavitation. As the thickness of the removed liquid layer drops, it becomes unstable and easily breaks up into small droplets, invisible in the high-speed video. Ultimately, the enhanced ablation and laser-drilling described here is due to high-impulse, multi-laser melt ejection.

6.3. Multi-physics hydrodynamic simulations of short pulse-melt pool interaction

In order to provide qualitative insight into laser energy coupling during the pulse, and the response of the melt pool to laser-driven shock and release wave propagation, 3D HYDRA models, similar to those presented in [7], were developed to simulate short pulse-melt pool interaction over the duration of nanoseconds. The simulations considered 330 nm, 30 J/cm² pulses with widths of 20 ps and 7 ns (FWHM). The corresponding spot sizes for the pulses were consistent with the experiments. For simplicity, an initialized 10 μm deep near surface Al melt with a uniform temperature of 1,500 K was used for the study. It is noted that this melt depth is consistent with the melt penetration predicted by heat transfer calculations for the 100 μs, 450 W preheating CW pulse.

Figure 16 shows the instantaneous free surface velocity as a function of time during the pulse, considering both dynamic motion and ablative recession. The more than 20x decrease in free surface velocity for the nanosecond pulse is directly related to the reduction in peak irradiance and more pronounced shielding from the ablative plume. The corresponding pressure distributions showing shock and release wave propagation are provided in Fig. 17. Note that the peak simulated shock pressure for the 20 ps pulse is more than an order of magnitude larger than the peak shock pressure for the 7 ns pulse, in qualitative agreement with the estimates made above. For the 20 ps pulse, the release (rarefaction) wave that follows the initial laser-driven shock puts the melt into tension and triggers cavitation near the melt-solid interface, as indicated by the low density expanding vapor cavity shown in Fig. 18. Cavitation during release is not observed for the 7 ns pulse, which can be attributed to the substantially weaker pressure drive and to destructive wave interaction during the pulse that lowers the imposed tensile strain within the melt. The HYDRA simulations, therefore, show melt cavitation, leading to deep removal of the intialized melt pool, for the 20 ps pulse but not for the 7 ns pulse, which is consistent with the interpretation in Section 6.2 that only the 20 ps pulse provides sufficient laser energy coupling to drive multi-micron scale cavitation.

 

Fig. 16 Instantaneous free surface velocity, considering both dynamic motion and ablative recession, for an Al target with an initial 100 μm dia. x 10 μm deep near surface melt based on HYDRA simulation: (a) 355 nm, 20 ps (FWHM), 30 J/cm² pulse, (b) 355 nm, 7 ns (FWHM), 30 J/cm² pulse.

Download Full Size | PDF

 

Fig. 17 Pressure distribution as a function of time for an Al target with an initial 100 μm dia. x 10 μm deep near surface melt based on HYDRA simulation: (a) 355 nm, 20 ps (FWHM), 30 J/cm² pulse, (b) 355 nm, 7 ns (FWHM), 30 J/cm² pulse. Initial surface position located at 0 μm depth. Plots show laser-driven shock and release wave propagation in the target, as well as ablative plume expansion.

Download Full Size | PDF

 

Fig. 18 Relative density (ρ_ref/ρ) distribution as a function of time for an Al target with an initial 10 μm deep near surface melt irradiated with a 355 nm, 20 ps (FWHM), 30 J/cm² pulse based on HYDRA simulation. Relaxation following the 20 ps pulse leads to cavitation near the melt-solid interface, as noted by the low density expanding vapor cavity. Cavitation was not observed for the 7 ns pulse.

Download Full Size | PDF

The surface velocity histories from the HYDRA simulations were then used as input for microsecond scale simulations of the ensuing melt hydrodynamics developed in ALE3D. The ALE3D models considered Al and SS specimens measuring 600 μm x 600 μm x 500 μm thick and incorporated a 600 μm thick air region to allow for the visualization of material motion and ejection out of the drilled channel. A non-reflecting outflow condition was used along the upper boundary of the air region to accommodate material ejection out of the simulation space (i.e. to prevent artificial reflection at the model boundary). The base of the specimen also utilized a non-reflecting boundary to study response in the absence of a strong backside reflection. For simplicity, the melt pool was idealized with an elliptic geometry and a uniform material temperature (1500 K for Al and 2500 K for SS).

The impulsive drive of the ablative recoil pressure was modeled by prescribing a surface velocity history at the melt-air interface based on the HYDRA simulations. The simulations utilized an experimentally anchored multi-phase equation of state model, as well as nonlinear temperature-dependent models for surface tension and cooling (evaporative and radiative heat loss), and incorporated viscous and gravitational effects. Both near and sub-surface conditions for pulse-melt pool interaction were investigated, where the latter was approximated by positioning the melt 250 μm within an idealized cylindrical channel.

Figure 19 shows a range of melt ejection processes for selected melt pool geometries considering a 210 μm dia., 355 nm, 20 ps, 30 J/cm² pulse, which are consistent with the experimentally observed melt ejection described in Section 6.1 and discussed throughout this section. The simulations are intended to provide qualitative insight into the mechanisms for material ejection and the relative influence of geometric effects. For Al, the 100 μs, 450 W CW pre-heating pulse creates a relatively shallow melt pool (~100 μm dia. x 10 μm deep). When this shallow melt pool is irradiated with the picosecond pulse, the release wave induces sufficient tension to cavitate a relatively thin sheet of liquid, as illustrated in Fig. 19(a) for a near surface melt and in Fig. 19(d) for a melt recessed within a 250 μm deep cylindrical channel. This cavitation process is consistent with the HYDRA simulation and the experimental observation shown in Fig. 13, panel 1. For the deeper melt pools generated in SS (~300-400 μm dia. x 100-200 μm deep), the response to the picosecond pulse is dominated by corona splash and the ensuing breakup into secondary droplets as seen in Fig. 19(b), which is similar to the experimental observation depicted in Fig. 13, panel 2. Asymmetric offset of the pulse from the centerline of the melt pool can also drive lateral sloshing in Fig. 19(c). Finally, for the notional case of a deeper Al melt pool (~200 μm) recessed within a drilled channel, the lateral confinement of the channel leads to canopy collapse of the splash corona seen in Fig. 19(d) and the subsequent outjetting of material (Fig. 19(f)), which is in qualitative agreement with the images in Fig. 13, panel 3. Thin sheet cavitation was also observed in this scenario. Splash ejection processes involving large scale motion of the melt were also seen in melt ejection simulations driven by the 7 ns pulse, but cavitation was not observed. This supports the contention above that the large removal enhancement for the picosecond drive pulse compared to the nanosecond pulse shown in Fig. 15 is due to cavitation, which is likely the dominate removal mechanism in deeper channel drilling.

 

Fig. 19 ALE3D short pulse-melt pool interaction simulations depicting a range of melt ejection mechanisms for a 210 μm dia., 355 nm, 20 ps, 30 J/cm² pulse (plots show relative density ρ_ref/ρ distribution); note ρ_ref is the reference solid-state density, so that ρ_ref/ρ = 1 indicates a solid or solid-density liquid region, and ρ_ref/ρ > 1 indicates a vapor or liquid-droplet rich vapor region: (a) cavitation of a 100 μm dia. x 10 μm deep, near surface Al melt; (b) corona splash breakup of a 300 μm dia. x 100 μm deep, near surface SS melt; (c) corona splash breakup and sloshing of a 400 μm dia. x 200 μm deep, near surface SS melt with an asymmetric pulse offset of 100 μm; (d) cavitation of a 100 μm dia. x 10 μm deep Al melt within a 250 μm deep channel; (e-f) canopy collapse, outjetting, and cavitation of a 400 μm dia. x 200 μm deep Al melt within a 250 μm deep channel.

Download Full Size | PDF

7. Discussion

Laser material processing for drilling and cutting has become more popular with the continuous advancement in laser technology. It is non-contact and offers improved precision over conventional machining techniques. However, for laser machining to be truly practical, one needs to overcome its relatively low efficiency which directly impacts cost and system complexity. Traditional laser ablation relies on the removal of a thin surface layer through explosive boiling when short pulse lasers are used and evaporation for longer pulses and CW lasers. These physical processes take place in a shallow layer which limits the material removal rate. They are also inefficient uses of laser energy. Heating a layer of material to the critical temperature (6,000 – 8,000K) invariably requires heating the surface well beyond that. Removal by evaporation requires expending the very significant latent heat of vaporization.

It is a paradigm change to employ a CW laser to create a deep melt layer followed by a short pulse, high-impulse laser to eject the melt. We have used a gated CW laser in our study to produce enough melt while minimizing unnecessary collateral heating. The high efficiency, lower cost CW laser is used for the more energy intensive phase – creation of the melt. The UV picosecond laser is ideal for providing the high-impulse needed for effective melt ejection [7]. Only moderate pulse energy and power are required from this more complex laser. The multi-laser approach presented in our paper takes advantage of both types of lasers, and as a result, improves material removal rate and efficiency significant. Another benefit of this technique is that it is less sensitive to material types.

Real life applications often put constraints on processing speed, power consumption, and instrument size, portability, cost and ease of use. Increased material removal rate and efficiency leads to faster processing speed and lower total power consumption. High power CW fiber lasers are compact, durable turn-key systems which are readily available. When taking into consideration pulse energy and repetition rate, system footprint, and cost, commercial picosecond lasers available today are better suited for high-impulse melt ejection during material processing compared to femtosecond lasers. The efficiency of this approach is also attractive for portable applications.

The ML approach to laser drilling presented here can be generalized. For example, while we did not explicitly test this here, we expect that a picosecond laser operating in the green (2ω) or near IR (1ω) could also be used for effective ML melt ejection for some materials, though they may be less efficient and will likely required higher pulsed laser fluence to match the material removal rates obtained with our UV picosecond laser. The approach could also benefit from a higher power CW laser. Our study shows a higher material removal efficiency using CW with shorter gate width to produce the melt layer. This may be further improved with the use of a higher power CW laser which will allow us to reduce the gate width even further and still generate the melt layer. This ML for efficient, high removal rate laser drilling approach for high aspect ratio channels can be also be applied to non-metals with the proper choice of wavelength for the melt laser. Here, we have used a commonly available CW fiber laser operating at 1064 nm to melt Al and SS. For metals which don’t readily absorb near IR light, a different wavelength for melt generation can be chosen. For instance, Cu is highly reflective at 1064 nm but absorbs readily at 500 nm. A gated high power CW laser (or quasi-CW laser) operating in the green could be used as the melt laser for Cu; a UV picosecond laser could still be used for melt ejection (cavitation). Optically transparent materials such as glasses can be melted by a CO₂ laser operating further in the IR at a wavelength near 10.6 μm and have been recently used in conjunction with picosecond lasers to process glasses [21]. A gated CO₂ laser could be used for melt generation. The electronic structure of liquid glasses well above their melt temperature becomes quasi-metallic and readily absorbs UV or even near IR light [22,23], hence combining a CO₂ laser with a UV picosecond laser could be an effective ML drilling approach for glasses.

8. Summary

We have demonstrated a multi-laser approach for high efficiency drilling and cutting of metals that combines a picosecond UV laser with a gated CW near-IR laser. This approach offers significant improvement over existing single laser approaches (CW or pulsed) for metal drilling. To our knowledge, the removal efficiencies described here exceed the best laser removal efficiencies reported by other means. Our approach can produce high aspect channels in a range of materials including metal, semiconductor, and carbon fiber composite. High-speed ejecta imaging, CT-scanning, and hydrodynamic simulations were performed to help identify the important physical processes critical for the high efficiency drilling including melt formation and ejection through cavitation and surface wave excitation. In particular, we find that these high-impulse multi-laser enhancements are likely due to both laser-induced surface wave instabilities and cavitation of the melt for shallow holes and melt cavitation for deeper channels.