3.1 The orthogonal experiment
The range analysis was used to determine the extent of the influence of each factor on the experimental results by calculating the average value and range of the experimental results for each level factor, as presented in Table 5. Additionally, to reduce experimental errors and validate the effectiveness of the experiments, we conducted a statistical analysis of variance of the experimental results, as presented in Table 6 [27, 28]. The analysis of range and variance range were used to analyze the influence of factors on the tensile strength of ultrasonic assisted laser welding joints. The purpose was to assess the primary and secondary relationships, rank them and select the optimal combination of welding parameters. The results of the orthogonal experiment are given in Table 4.
Table 4
The results of the orthogonal experiment
| | Laser power (W) | Welding speed (mm/s) | Ultrasonic power (W) | Tensile strength(MPa) |
1 | 1200 | 15 | 0 | 862.6 |
2 | 1200 | 20 | 500 | 940.6 |
3 | 1200 | 25 | 1000 | 966.3 |
4 | 1300 | 15 | 500 | 969.2 |
5 | 1300 | 20 | 1000 | 997.2 |
6 | 1300 | 25 | 0 | 875.1 |
7 | 1400 | 15 | 1000 | 956.8 |
8 | 1400 | 20 | 0 | 874.2 |
9 | 1400 | 25 | 500 | 913.6 |
The results of the tensile strength range analysis are presented in Table 5. The range of ultrasonic power was 102.8, which was 3.18 times of welding power and 5.41 times of welding speed, indicating that the order of influence on the tensile strength of the joint was C (ultrasonic power) > A (laser power) > B (welding speed).
Table 5
The results of the tensile strength range analysis
| Laser power (A) | Welding speed (B) | Ultrasonic power (C) |
K1 total | 2769.5 | 2788.6 | 2611.9 |
K2 total | 2841.5 | 2812 | 2823.4 |
K3 total | 2744.6 | 2755 | 2920.3 |
k1 average | 923.2 | 929.5 | 870.6 |
k2 average | 947.2 | 937.3 | 941.1 |
k3 average | 914.9 | 918.3 | 973.4 |
Range | 32.3 | 19 | 102.8 |
Sequence | C > A > B |
Optimized | A2 | B2 | C3 |
Optimized combination: A2B2C3 |
The results of tensile strength variance analysis were presented in Table 6. The P-value of factor C was 0.018, and the p-value was less than 0.05, indicating that the experimental results were statistically significant. Moreover, the p-value of C was smaller than the p-values of factors A and B. The results showed that ultrasonic power was a significant factor affecting laser welded joints, followed by laser power and welding speed. Therefore, the results of range analysis and variance analysis showed that the optimal combination of ultrasonic assisted welding parameters was laser power 1300 W, welding speed 20 mm/s, and ultrasonic power 1000 W.
Table 6
The results of the tensile strength variance analysis
Factor | SS | DOF | MS | F-value | P-value |
A | 1688.18 | 2 | 884.09 | 5.622 | 0.151 |
B | 547.28 | 2 | 273.64 | 1.823 | 0.354 |
C | 16581.38 | 2 | 8290.69 | 55.223 | 0.018 |
D(error) | 300.26 | 2 | 150.13 | | |
The above results showed that the ultrasonic power has the greatest influence on the tensile properties of Ti6Al4V and the optimal parameter combination was Sample 5. In order to further analyze the reasons for the influence of ultrasonic assistance on the tensile strength of Ti6Al4V welded joints, we set the corresponding control groups as Sample 10 and Sample 11 under the optimal welding parameters. The specific experimental parameters are shown in Table 7.
Table 7
The experimental parameters of control group
| Laser power (W) | Welding speed (mm/s) | Ultrasonic power (W) |
Sample 5 | 1300 | 20 | 1000 |
Sample 10 | 1300 | 20 | 500 |
Sample 11 | 1300 | 20 | 0 |
3.3 Porosity
Porosity is the most critical factor affecting the tensile strength of joints. There are two main types of pores in the welding process, including the process porosity and the metallurgical porosity [10, 11]. The formation of metallurgical pores is primarily attributed to the entrapment of hydrogen or other gases within the weld pool during the welding process, leading to their inability to escape in a timely manner. These pores are characterized by small size and distribution at various locations within the weld. The formation of process porosity is attributed to the unstable collapse of the laser keyhole, which exhibits a large and irregular shape, primarily concentrated at the base of the weld. To quantitatively analyze the impact of ultrasonic-assisted laser welding on porosity, the 3D-CT tomography was employed to observe the size and quantity of pores in the joints. As shown in Fig. 4a, within the CT-3D field of view, a large number of pores were distributed throughout the weld when without ultrasonic power supplied. Among them, most of them were small pores with a diameter of 100 to 200µm, showing blue and green, and a few were large pores (more than 500 µm) with a diameter of 600 to 900 µm, showing orange. The statistical analysis results of pores were shown in Fig. 4d, e and f. With the increase of ultrasonic power, both the number of pores and the porosity gradually decreased. The formula for porosity calculation was as follows:
$$\:{\text{P}}_{\text{r}}=\frac{\sum\:\text{V}\text{r}}{\text{V}}\times\:100\%\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:(1)$$
In the laser welded sample without ultrasonic, the number of pores and porosity were 277 and 3.06%, respectively. Moreover, the number of macropores was 6. When the ultrasonic power was increased to 500W, the number of pores and porosity were reduced to 101 and 0.71%, respectively, and the number of macropores was reduced to 1. When the ultrasonic power was further increased to 1000W, the number of pores and porosity were reduced to 10 and 0.08%, respectively, and no macropores were present in the weld joint. The porosity of the joint was reduced by 97.4% compared with the weld without ultrasonic application.
The reduction of porosity was mainly attributed to the cavitation effect of ultrasonic. When the ultrasonic was applied to the molten pool, steady-state cavitation occurred under specific conditions. As the steady-state effect is generated, the interface between the melt was broken, thereby producing unstable cavitation bubbles. A surge was generated around the cavitation bubble during rupture, which accelerated the flow of the surrounding melt, drove the motion of bubbles in the melt, and promoted collision and merging of the bubbles. The merging of bubbles thus effectively increased the volume of bubbles in the molten pool. The increase in the volume of the bubble would increase its rising speed in the molten pool. The cavitation threshold induced by the cavitation effect could be expressed as follows [30]:
$$\:{\:P}_{T}={P}_{0}-{P}_{v}+\frac{2}{3\sqrt{3}}{\left[\frac{{\left(\frac{2\sigma\:}{{R}_{0}}\right)}^{3}}{{P}_{0}-{P}_{v}+\frac{2\sigma\:}{{R}_{0}}}\right]}^{\frac{1}{2}}\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\left(2\right)$$
where P0 was 1 atm, which was equal to 0.1013 MPa [31]; Pv stood for the saturation vapor pressure, which was equal to 0 MPa[31]; σ was the surface tension coefficient, which was equal to 1.4 N/m [32]; R0 was the radius of the cavitation nucleus, which was equal to 10− 5 m [31]; The calculated result of formula (2) was 1.93×105 MPa; When the intensity of the ultrasound exceeded a threshold, a cavitation effect occurred inside the molten pool. The intensity of the ultrasonic PK in the molten pool could be expressed as follows [32]:
$$\:{P}_{K}=\:{\left[\frac{2P{C}_{l}{\rho\:}_{m}}{S}\right]}^{\frac{1}{2}}\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\left(3\right)\:\:$$
where P was the output power of the ultrasonic (0 W, 500 W and 1000 W); ρm was the density of melt, which was equal to 4208 kg/m3 [31]; Cl was the sound velocity of ultrasonic in titanium alloy melt, which was equal to 4407 m/s [32]; S was the area of horn end, which was equal to 7.854×10− 5 m2; In this experiment, when the ultrasonic output power was P1 = 500 W and P2 = 1000 W, the corresponding homologous values were 1.537×107 MPa and 2.250×107 MPa, respectively. They were both above the threshold value PT=1.93×105 MPa.
Therefore, significant cavitation effects were observed in the molten pool under the ultrasonic power of 500 W and 1000 W. Whether these bubbles can transform into pore defects depends on their ability to rise out of the molten pool before the keyhole closes. According to the Stokes' formula[21], the bubble escape velocity Ve could be expressed as:
$$\:{V}_{e}=\frac{2\left({\rho\:}_{m}-{\rho\:}_{g}\right)g{r}^{2}}{9{\eta\:}_{m}}\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\left(4\right)\:$$
where g was the acceleration of gravity, which was equal to 9.8 N/kg; r was the pore radius; ηm was the liquid metal viscosity, which was equal to 0.005 gcm− 1s− 1 [32]; ρg was the bubble density, which was equal to 1.784 kg/m3 [33]. The solidification rate was assumed to be equivalent to a welding speed of 20 mm/s or higher [34]. The pore critical diameter calculated by Eq. (4) was 206.8 µm. The results indicate that when the diameter of bubbles within the molten pool exceeds 206.8 µm, they are able to escape before the keyhole closes, whereas pores with diameters below this critical threshold are unable to exit the molten pool and consequently lead to the formation of pores defect. Additionally, as shown in Fig. 4f, this conclusion was further supported by the observation that all remaining pores were smaller than 200 µm at the ultrasonic power of 1000 W.
3.4 Microstructure
3.4.1 Microstructure in the central of FZ
Figure 5 shows the microstructure of FZ region under different ultrasonic powers. It could be seen that the crisply martensite full the interior of the prior-β grain boundary. When cooling from the melting state to the solidification state, the liquid molten pool will be transformed into coarse prior-β grains. The prior-β grain boundaries of grains were shown in Fig. 5a.
According to the continuous cooling transition curve (as shown in Fig. 6), the high temperature β phase will transform into a structure related to its cooling rate. The cooling rate in the FZ region exceeds 410°C/s, resulting in the transformation of the high-temperature β phase into fine needle α' martensite [29, 35]. With the cooling of the FZ, the primary acicular α1' martensitic phase grew at a 45° direction from the grain boundary in the prior-β grain and ceased its growth until reaching the other grain boundary. Then, as the molten pool continued to cool, a secondary precipitated α′ phase (α2') was formed in a direction perpendicular to the α1' martensite phase. The cross-growth of primary α1' martensite and secondary α2' martensite resulted in the formation of basket structures within the prior-β grains in the FZ region [36]. Figure 5b and c illustrated that, with the increase of ultrasonic power, the length of martensite α1' was significantly reduced and the content of secondary precipitated phase α2' was also increased, resulting in a more compact mesh basket structure.
3.4.2 Grain size
Grain size is one of the important factors affecting the tensile strength of joints. Due to the limitation of sampling range, the prior-β grain morphology was difficult to be observed completely,
Therefore, in order to ensure the rigor of scientific research, in this section, we expanded the sampling range (region 2 in Fig. 2b) and performed the parent grain crystallographic reconstruction (PGCR) for prior-β grains [37]. The Aztec-crystal software was employed to accurately assess the influence of ultrasonic treatment on the grain size of α' martensite and prior-β. As shown in Fig. 7a, e, i, it could be seen that there were large numbers of α' martensite with different orientations within the prior-β grains. However, the presence of a large number of α' martensite made it difficult to observe the grain boundary distribution and grain size of prior-β. Figure 7c, g and k show the reconstructed grain appearance, and no α’ martensite is observed at the reconstructed prior-β grain boundary. At the ultrasonic power of 0 W, it was observed that a large number of prior-β grains exhibited columnar crystal appearance. With the increase of the ultrasonic power, the number of prior-β columnar crystals was significantly decreased, while the number of small sized equiaxed crystals was increased. The statistical results of α' martensite are shown in Fig. 7b, f, and j. With the increase of the ultrasonic power, the average grain size of α' martensite was reduced by 10.4 µm to 10.0 µm and 9.7 µm respectively. Additionally, the proportion of small-sized grains (α' martensite size less than 9 µm) increased from 36.9–37.3% and 38.5% respectively. The degree of grain refinement for α' martensite reached 6.7%. The statistical results of prior-β shown in Fig. 7d, h, and l demonstrated that, with the increase of the ultrasonic power, the average grain size of prior-β was reduced by 83.6 µm to 73.6 µm and 71 µm respectively. Additionally, the proportion of small-sized grains (prior-β size less than 90 µm) increased from 62.54–67.34% and 70.14% respectively. The degree of grain refinement for prior-β reached 15.1%.
Ultrasonic vibration refines the grain size of prior-β through mechanical and cavitation effects. On the one hand, the application of ultrasonic to the molten pool increased the stirring frequency, which broke the epitaxial growth trend of prior-β columnar crystals. As a result, these coarse prior-β columnar crystals would subsequently be fragmented into numerous fine new dendrites. And these delicate dendrites offered a multitude of nucleation sites for the development of equiaxed crystals, thereby promoting grain refinement [19]. On the other hand, the ultrasonic cavitation effect generated a significant quantity of cavitation bubbles within the molten pool, the cavitation bubbles would expand, plug, and burst rapidly in the molten pool. Upon the collapse and closure of these bubbles, there would be instantaneous reduction in pressure surrounding the bubbles, leading to a rise in the melting point within the molten pool. It could be explained by the Clausius-Clapeyron equation [38]:
$$\:\frac{\varDelta\:T}{\varDelta\:P}={T}_{m}\frac{{V}_{1}-{V}_{2}\:}{\varDelta\:H}\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:(5)$$
where ΔT was the temperature change, Tm was the solidification temperature under atmospheric pressure, ΔP was the pressure change, ΔH was the latent heat of crystallization, V1 was the liquid volume, and V2 was the solid volume. The increase of melting point was equivalent to increasing the degree of undercooling in crystallization process, thereby increasing the nucleation rate of molten pool. The generation of a large number of new nuclei provides conditions for the subsequent growth of equiaxed crystals.
The reduction of α' martensite could be attributed to the following two aspects. On the one hand, the α' martensite nucleated and grew within the prior-β grain, thus the maximum length of the α' martensite was constrained by the prior-β grain boundary. The refined prior-β grain significantly reduced the overall size of the primary α1' martensite. On the other hand, the cavitation effect of the ultrasonic vibration increased the cooling of the molten pool, which implied that the phase transformation (solidification) started at a temperature lower than the equilibrium phase transition temperature (994℃). The enhancement of undercooling facilitated the growth of the α1' and α2' martensite phase within the prior-β grain [39]. Compared with α1' martensite, α2' martensite exhibited a smaller phase size, thus the growth of large amounts of α2' martensite perpendicular to α1' martensite reduced the overall size of the α' martensite, which contributed to the formation of a compact basket structure. Hall-Patch provided the relationship between weld grain size and strength [40]:
$$\:{{\sigma\:}}_{1}={{\sigma\:}}_{0}-\:\text{k}{d}^{-1/2}\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:(6)$$
where \(\:{{\sigma\:}}_{1}\:\)was the strength of the weld, \(\:{{\sigma\:}}_{0}\) was the material constant, k was the slope of Hall-Patch, and d represented the average grain size of the weld. Therefore, the joint strength of Ti6Al4V increased with the refinement of the average grain size of prior-β and α' martensite.
3.4.3 Grain boundary characteristics and dislocation density
The misorientation angle of grain boundaries reflects the atoms arrangement law, which greatly affects the microstructure characteristics and performances of materials [18]. Based on the orientation angle between adjacent grain boundaries, grain boundaries can be categorized into two types: high-angle grain boundaries (> 15°, HAGBs) and low-angle grain boundaries (2–15°, LAGBs). Figure 8a, b, and c show the grain boundaries distribution in the FZ region under different ultrasonic powers (region 1 in Fig. 2b), where the blue lines are LAGBs and the red lines are HAGBs. The results indicated that HAGBs were distributed across almost the entire region, with a limited LAGBs presented in the interior of α' martensite. Figure 8d, e and f illustrate the statistics charts of misorientation angle distributions. The results indicated that the grain boundary orientation exhibited multi-peak distribution characteristic under different ultrasonic powers, and mainly distributed around 2°, 10°, 60° and 90°. Furthermore, the majority of HAGBs were distributed around 60°. As the ultrasonic power increased, the proportion of LAGBs decreased from 10.8–8.95% and 7.03%, respectively, while the proportion of HAGBs increased from 89.2–91.05% and 92.97%, respectively. This indicated that ultrasonic assistance helped to promote the transition from LAGBs to HAGBs.
The phenomenon was attributed to the enhanced undercooling of the molten pool by ultrasonic cavitation effect. The decrease of the supercooling degree in the molten pool promoted the nucleation and growth of α1' and α2' martensite. A large number of α2' martensite and α1' martensite growing perpendicular to each other will form more HAGBs [41]. Compared with LAGBs, the HAGBs can impede the dislocation movement more effectively [42], and when the cracks expand to the HAGBs, it usually needs to consume a larger amount of energy, and thus plays a role in preventing crack expansion. Therefore, the sample of 1000 W had a stronger ability to retard brittle cracks, which significantly improved the joint strength.
During laser welding of Ti-6Al-4V alloys, the uneven of heat input inevitably results in an inhomogeneous distribution of residual stresses within the material, and this difference in distribution leads to inhomogeneity in grain deformation. As a result, many geometrically necessary dislocations (GNDs) are generated in the weld to maintain the coherence of grain deformation [43]. The distribution of GNDs in the weld can be reflected by the kernel average misorientation (KAM). Figure 9a, b and c show the KAM diagrams of 0 W, 500 W, 1000 W, where different colors reflect the distribution of GNDs within the material. The results indicated that at the ultrasonic power of 0W, the KAM diagram mainly showed green color with relatively high dislocation density. As shown in Fig. 9b and c, with the increase of ultrasonic power, the KAM map gradually transitioned from the original green to blue, and the average KAM value also decreased from 0.44° to 0.36° and 0.31° respectively. This indicated that the GNDs density of the weld was gradually reduced and the deformation of the weld grains were also reduced.
Ultrasonic assisted could reduce the dislocation density in the weld and improved the deformation resistance of the joint, which could be attributed to the following two aspects. From a microscopic point of view, residual stress caused the material lattice to be in a high-energy unstable state (dislocation state). When the ultrasonic wave was applied to the molten pool, the vibration effect it generated effectively increased the vibration frequency and amplitude of lattice atoms. This allowed the previously distorted atoms to return to their normal low-energy stable state after obtaining sufficient energy, thereby opening up dislocations and relaxing residual stresses [44]. From a macro point of view, the inhomogeneous of grain deformation is mainly affected by the uneven heat input. The stirring effect of ultrasonic waves in the molten pool promoted a more uniform temperature distribution, thus reducing the temperature gradient in the weld. Therefore, the residual stress was significantly reduced and the reduction in dislocation density significantly enhanced the resistance to grain deformation.
3.4.4 Texture strength
During the welding process, the uneven heat input leads to a significant temperature gradient inside the weld, which leads to the formation of coarse prior-β columnar grains in the weld pool. These prior-β columnar grains tend to have a specific grain orientation, which makes the joint strength of Ti6Al4V alloy significantly anisotropic in different loading directions[45]. Figure 10a, c and e show the polar figures of the α phase under different ultrasonic powers. It could be seen that the maximum texture strength of α phase all appeared on the crystal plane of {0001}, and with the increase of ultrasonic power, the maximum texture strength of α phase decreased from 9.48 to 6.71 and 6.34 respectively. The polar diagrams of reconstructed β phase are shown in Fig. 10b, d, f. It could be seen that the maximum texture strength of β phase under different ultrasonic power all appeared on the {100} crystal plane, and with the increase of ultrasonic power, the maximum texture strength of β phase decreased from 8.79 to 8.11 and 6.91, respectively. Previous studies have shown that the precipitation of α phase mainly distributed inside the prior-β grain boundary and followed the Burgers orientation relationship with the parent β phase, which makes the texture strength of the α phase consistent with that of the β phase [46]. The reduction of the texture strength of α and β phases by ultrasonic was mainly attributed to the refinement of prior-β grain. The stirring effect of ultrasonic broke the growth trend of prior-β columnar grains and promoted the formation of equiaxed grains, which reduced the anisotropy of the crystals. Therefore, the texture strength of α and β phases were both significantly reduced.
3.5 Mechanical properties
Figure 11 shows the tensile test results of the control sample and the base metal. The tensile strength of the base material was 1023.7 MPa and the tensile strength of the sample without ultrasonic was 920.1 MPa. When the ultrasonic power was gradually increased to 500 W and 1000 W, the tensile strength of the joint increased to 960.3 MPa and 997.2 MPa, respectively, representing a rise of 4.37% and 8.38%.
The joints under different ultrasonic power all failed in the FZ area. Figure 12a, d and g depict the macroscopic tensile fracture of the joint under different ultrasonic powers. It was evident that significant large-size pore defects were present on the fracture surface of the sample without ultrasonic treatment, but not on the fracture surface of the 500 W and 1000 W samples. Pore defects significantly deteriorate the mechanical properties of joints, as they lead to stress concentration during tensile loading, thereby accelerating the fracture of the weldment. Figure 12b and c depict the fracture surface (magnified in 500 times and 1000 times) under the ultrasonic power of 0 W. It was evident that a limited number of shallow dimples were distributed around the cleavage step, and the distinct characteristics of river patterns emerged on the cleavage step surface, indicating a mixed ductile and brittle fracture mode for the sample without ultrasonic treatment. Figure 12e and f show that, under the ultrasonic power of 500 W, the number of dimples distributed around the cleavage step increased, and the size of dimples also decreased significantly. As shown in Fig. 12h and i, under the ultrasonic power of 1000 W, a large number of dimples were observed in the fracture, and the number of dimples was larger and the size was also smaller, which was a typical ductile fracture feature. The large number of aggregated dimples significantly enhanced the strength of the joint.