3.1. Dynamics of the oscillating keyhole
In contrast to the steady-state linear scan strategy, the oscillating laser strategy is a dynamic process which operates at continuously and periodically varying thermal conditions. Each of the four key positions on an oscillation unit, as shown in Fig. 1, represents a different thermal condition and, thus, a different melting response of the material. To date the dynamics of oscillating keyholes are difficult to capture while most of the available understanding in the field derives from metallographic characterization or top-down observation 25,27,34. Here, DXR offers a unique capability for visualizing the spatiotemporal evolution of an oscillating keyhole.
Figure 2a shows the net result of the linear and circular oscillation at 400 mm/s linear scan speed and 0.12 mm circular oscillation amplitude. The oscillation units were slightly distorted by the linear scan vector to form an oval shape. As the scan speed increases, both the remelting and the number of oscillation units within a constant path should decrease. Instead of a constant speed, the transient speed of the laser oscillation can be described by a continuous and periodic wave function as shown in Fig. 2b. Given that the oscillating frequency is constant, the amplitude and the mean of the transient speed function correlates positively with the oscillation unit diameter and the linear scan speed, respectively.
Figure 3a shows a series of exemplar radiographs for process parameters of 400 mm/s linear scan speed and 0.12 mm circular oscillation amplitude. Figure 3c shows a comparative fluctuating keyhole during processing without circular oscillation resulting in the formation of a keyhole pore. Given the same laser power and speed, it is rather easy to distinguish the differing responses to laser oscillation from those to the conventional strategy. As evident in the depth profiles in Fig. 3e, with no oscillation, the keyhole depth exhibited a smaller short-range variation while the oscillating keyhole resulted in up to 125 µm difference between the deepest and the shallowest conditions within the same oscillation unit. Owing to the higher transient laser speed, the average keyhole depth in the oscillation experiment (99 ± 39 µm) is significantly shallower than its straight counterpart (267 ± 28 µm). The depth profiles of the simulated keyholes show reasonable agreement with the experimental data, especially in the absence of oscillation where the average keyhole depth is 284 ± 7 µm. In the experiment, a few instances of keyhole pore formation, which resulted in the outliers in the depth profile, contributed to the larger depth variation. No keyhole pore formed in the simulation; however, some protrusions were found within the keyhole as shown in Fig. 3d. They are often considered as the precursors for the formation of keyhole pores 35.
In an oscillating melt pool, the local pre-heat conditions at the point of laser-matter interaction vary continuously throughout the oscillation unit as highlighted in Fig. 1. The keyhole fluctuation can be visualized in Fig. 3a where the sequence started at position A (3.28 ms) then passed through positions B, C, and D at 3.32 ms, 3.36 ms, and 3.40 ms. At position C, the laser traverses through the hotter underlying material which may still be molten from the previous pass. Conversely, the underlying material at position A is cooler because of being further away from the active melt pool. The transient laser speeds are identical at the two positions where the scan vector of the oscillation unit is normal to the linear scan direction. As shown in Fig. 3e, the depth evolution curve corresponding to the case in Fig. 3a suggests that the keyhole penetrated > 100 µm deep at position C (3.36 ms) thanks to the higher pre-heat. The simulation captured the keyhole fluctuation showing an average keyhole depth of 146 ± 22 µm. Better agreement occurred at position C (3.36 ms) where the keyhole was at its deepest; but at position A (3.28 ms), the simulated keyhole is deeper by around 50 µm. Since the experiments were performed on thin plates, we speculate that the edge effect might cause this discrepancy by limiting the fluid flow and thermal diffusion along the X-ray direction.
At positions B and D, the melt pool depths are similar although the difference in transient speeds is maximum as the oscillation vectors are opposite to each other and also parallel to the linear scan vector. This implies that pre-heat condition affects the keyhole depth more than the transient speed. However, the keyhole shape was different at these two positions which was confirmed in both the experiment and the simulation. A protrusion formed on the opposite side walls of the keyhole due to the opposite oscillation vectors (see 3.89 ms vs. 3.93 ms in Fig. 3b).
3.2. Chevron pattern formation
Many studies have reported the observations of the chevron pattern on solidified melt tracks. The pattern contains information about the cooling length and the geometry of the melt pool which are dictated in turn by the laser processing parameters 36. In a linear scan, the appearance of the chevron pattern is often dense, with a lack of periodicity as shown in Fig. 4d; however, with a oscillating laser, as shown in Fig. 4b-c, the chevron units showed long-range patterns which were also observed in the DXR experiments and confirmed by the simulations. This link is further evidenced by the fact that the number of the chevron units matches the number of the oscillation cycles. The two events were synchronized by the surface swelling caused by the keyhole movement which was observed experimentally, Fig. 3a. Figure 4a shows a complete sequence of the formation of a surface swelling wave highlighted in orange. The sequence started at 7.56 ms when was right before the laser reached position C of the oscillation unit. The hotter liquid gained backward momentum here as the laser had traveled only from position B to position C. At 7.58 ms, the surface swelling wave quickly built up since the hotter fluid met the cooler recoil fluid from the tail wall of the melt pool induced by the previous oscillation. Meanwhile, a valley (the depletion of the molten metal as a manifestation of the forward moving recoil fluid) formed near the tail of the melt pool and caused local cool down which accelerated solidification. Figure 4b&e quantifies this phenomenon by tracking the distances between the reference line and the melt pool features. As the peak of the swelling wave traveled backward, the tail of the melt pool moved further away from the reference line suggesting the formation of a chevron unit. As shown in Fig. 4e, the formation of the new chevron unit was completed at 7.64 ms and for the second half of the oscillation unit, i.e., from position A to position C, the solidification front only advanced a much shorter distance.
One of the potential benefits of an oscillating laser is to control the shape of the melt pool. Comparing Fig. 4b to d, the oscillating beam reduced the length-to-width ratio of the melt pool at the same laser power and speed. Moreover, the unique chevron pattern also implies that the laser oscillation could introduce periodicity into the solidification process. This unique property of the laser oscillation may have potential applications in controlling microstructural/surface texture and enabling different scan strategies.
3.3. Keyhole depth analysis
The welding experiments surveyed three laser powers 300 W, 400 W, and 500 W. At each power level, we tested 5 circular oscillation diameters (0 mm, 0.09 mm, 0.12 mm, 0.18 mm, and 0.24 mm) and 6 linear scan speeds (0.2 m/s, 0.4 m/s, 0.6 m/s, 0.8 m/s, 1 m/s, and 1.2 m/s) resulting in total of 90 parameter sets as shown in Table S1. Figure 5 summarizes the average keyhole depth and the standard deviation at each parameter set which were extracted from the DXR videos. The keyhole dimensions follow the same trend across all three laser powers. The keyhole depth decreased as the circular oscillation diameter or laser scan speed increased. This is understood by knowing that the keyhole depth is mainly controlled by the energy density. Faster laser scan speed leads to lower energy density, thus, smaller keyhole depth. By contrast, the effects from the oscillation could be more subtle. Increasing the oscillation diameter distributes the same amount of laser energy over a wider area, which in turn reduces the laser energy density received by the top sample surface and results in a shallower keyhole. The oscillation strategy achieves similar impacts as beam defocusing or annular beam shaping but in a dynamic manner and without the need for complex diffractive optical elements. The standard deviations of the keyhole depths shown in Fig. 5d-f are indicators of keyhole stability. When excluding the linear scan cases, an increased oscillation diameter generally decreased the fluctuation of the keyholes. Note that the longer-range periodic fluctuation, primarily due to the different depths between position A and position C, is the major contributor to the larger deviations. Interestingly, as the oscillation diameter continues to increase, we start to observe that the standard deviations drop below the nominal levels of the corresponding linear scans, i.e., the stochastic fluctuations. That said, laser oscillation has the potential to compete with the conventional linear strategy to further reduce keyhole instability in an appropriate parameter space.
To elaborate further, Fig. 6 shows two groups of DXR series which were collected at 400 W laser power and 0.2 m/s laser scan speed. The top series captures the evolution of the keyhole shape for linear scan and the bottom one shows the keyhole evolution for circular oscillation with an amplitude of 0.12 mm. As shown in Fig. 5e, at 400 W and 0.2 m/s, a reduction in the standard deviation of keyhole depth was achieved by adding the laser oscillation. The top DXR series highlights the more severe keyhole fluctuation seen in the linear scan compared with their oscillation counterparts. The oscillation series sampled keyholes at many positions with respect to a single oscillation unit; yet, the fluctuation of the keyhole depths appears to be less even when the longer-range fluctuation was present. Additionally, the linear scan resulted in many keyhole pores which interact dynamically with the keyhole and contribute to the depth deviation. We believe that this reduction in the formation of keyhole pores might be facilitated by the reduction in the keyhole instability introduced by laser oscillation.
Five processing conditions were modeled in our simulation for which the parameters are shown in Table 1. These include two linear scan cases and three oscillating cases. As a quick validation of the simulations, Fig. 7 compares the maximum keyhole depths from the experiments with those from the simulations. There is good agreement between the simulations and the experiments based on the datapoints being close to the diagonal reference line. This highlights the ability of the model to capture the complex physics in an oscillating melt pool and suggests that more in-depth studies of the laser oscillation induced fluid flow will be feasible in the future.
3.4. Porosity reduction
The DXR videos also provide valuable insights on porosity formation as the two examples in Fig. 8 show. The main finding is that laser oscillation can mitigate or potentially eliminate the formation of keyhole porosity. Figure 8a shows the post-solidification fusion zones of four different welds processed by the same laser power and speed, but different oscillation settings. The linear scan in Fig. 8a resulted in a large cluster of keyhole pores. As the oscillation diameter increased, even the smallest amplitude (R = 0.09 mm) eliminated all the pores. One possible reason is that laser oscillation simply decreased the laser fluence impinging on the sample top surface and thus contributed to the occurrence of a less violent melt pool during solidification, which in turn reduced the formation of keyhole porosity. The linear scan case has the highest energy density (12.5 J/mm2) among the four. Laser oscillation reduced the energy density down to 8 J/mm2 at 0.09 mm amplitude, 7.14 J/mm2 at 0.12 mm amplitude, and 5.88 J/mm2 at 0.18 mm amplitude. Note that the energy density was derived from \(\frac{P}{vd}\) where \(P\) is the laser power, \(v\) is the laser speed, and \(d\) is the melt track width, i.e., laser spot size plus oscillation diameter.
To further investigate, one DXR result for the linear scan case at 400 W and 0.4 m/s was chosen as the baseline to compare with two other oscillating cases of the higher energy densities shown in Fig. 8b. Porosity only occurred in the no-oscillating case despite the higher energy densities used in the other two oscillating cases. That said, the laser energy density is only one of many critical factors that control the formation of keyhole porosity. Here, the results emphasize the importance of keyhole/melt pool shape. We speculate that the formation of keyhole porosity is influenced by 1) keyhole shape, as the deeper and narrower geometries tend to incur keyhole collapse and pore separation; 2) keyhole depth, which dictates the initial position and the mean escape path length of the keyhole pores; 3) melt pool length, which affects the average escape time. In terms of mitigating keyhole porosity, the oscillating keyhole is more stable than its linear counterpart because it has a more equiaxed morphology and resides in a larger molten volume. Additionally, the formation of keyhole porosity depends on several other factors, such as fluid flow and thermal distribution, which determine the force equilibrium and the mobility of a pore. Involving laser oscillation in the welding inevitably changes all the factors mentioned above.