**2. Experimental Method**

A 2024-T3 aluminum alloy with thickness of 3 mm was used. The chemical composition of this alloy is shown in Table 1. The original alloy had a 0.2% proof stress of 334 MPa, tensile strength of 464 MPa, and elongation of 21.8%.

**Table 1.** Chemical composition of 2024-T3 aluminum alloy used in this study (mass%).


A single-mode fiber laser (IPG Photonics, YLS-2000-SM, Japan, wavelength: 1070 nm, CW) was used for full-penetration bead-on-plate welding of the aluminum alloy, as shown in Figure 1a. The fiber diameter was 14 μm and we used a laser power of 2.0 kW. The laser was focused on the alloy surface with a spot size of 54 μm. Ar was used for shielding gas with a flow rate of 30 L/min. A welding speed of 2.5 m/min was used. The top and bottom surfaces of the laser-welded specimens were observed using an optical digital microscopy (Hirox, KH-7700, Japan). The cross-section of the weld bead was observed using optical microscopy (Olympus, SZX7, Tokyo, Japan).

**Figure 1.** Schematic illustrations of the experimental procedures. (**a**) Full-penetration bead-on-plate laser welding and preparation of fatigue test specimens cut from the laser-welded plate. (**b**) Dry laser peening (DryLP) process using femtosecond laser pulses. (**c**) Geometry of the fatigue test specimens with the weld bead located in the center. DryLP was performed using an x–y automatic stage which sequentially moved the specimen in a serpentine pattern, as indicated by the red arrows.

Then, the laser-welded specimens were subjected to DryLP in air. The peening was performed 15 months after welding to allow the completion of natural aging. As shown in Figure 1b,c, femtosecond laser pulses with a wavelength of 800 nm, pulse duration of 130 fs, and pulse energy of 0.6 mJ (Spectra-Physics, Spitfire, Japan) were focused using a plano-convex lens with focal length of 70 mm onto the specimen. The laser pulses were overlapped, with a coverage of 692%, which was shown to be the most effective condition for DryLP of 2024-T3 aluminum alloy [9]. A detailed description of the DryLP process was provided in the previous study.

For the preparation of specimens for hardness tests, the weld reinforcement was removed and electropolished in 20% sulfuric acid-methanol electrolyte for 30 s to remove the work-strained layer before DryLP treatment. The hardness of the top surface was measured using a Vickers hardness tester (Mitsutoyo, HM-221, Kawasaki, Japan) with a load of 1.96 N and loading time of 15 s.

For the preparation of specimens for residual stress measurement, DryLP treatment was conducted on as-welded specimens without removing the weld reinforcement. Depth profiling of the residual stress which was normal to the weld bead in the specimens was conducted nondestructively using the BL22XU beamline at SPring-8 [32], using the strain scanning method [33] with monochromatic X-rays with a photon energy of 30.013 keV, as shown in Figure 2. A CdTe detector was used for the measurements. The residual stress σ was estimated using σ = *E*(*d* − *d*0)/*d*0, where *E* is the Young's modulus of 61.7 GPa, *d* is the d-spacing of the (311) plane of aluminum in the welded or DryLPed specimens, and *d*0 is the d-spacing of the (311) plane in the BM of 0.12196 nm. The d-spacing of the (311) plane parallel to the weld bead in the gauge volume was measured, as shown in Figure 2a,b. The widths of both the incident and receiving slits were 0.2 mm. For the surface measurements, the heights of these slits were 50 μm. For depth profiling, the slit heights, which determine the depth resolution, were 10 μm from the surface to a depth of 40 μm, and the slit heights were 30 μm deeper than a depth of 40 μm. The d-spacings of the (311) plane of the WM, below the weld toe, and in the HAZ were measured, as shown in Figure 2c.

**Figure 2.** Schematic illustrations of (**a**) overview, (**b**) top, and (**c**) side views of residual stress measurements using the strain scanning method with synchrotron X-rays.

Four kinds of specimens for fatigue testing were prepared: (i) As-welded specimen; (ii) reinforcement-removed welded specimen; (iii) DryLPed welded specimen; and (iv) DryLPed reinforcement-removed welded specimen. The stress concentration influenced the fatigue properties of the as-welded specimen due to both reinforcements and undercuts. Hence, to investigate the stress concentration only influenced by the undercuts, the reinforcements were removed. The reinforcements were removed using diamond pastes with a particle size of 1 μm. These specimens were cut from the laser-welded specimen, as shown in Figure 1a. DryLP was conducted on both surfaces of the

laser-welded specimen, as shown in Figure 1b. Plane bending fatigue tests (PBF-30, Tokyo Koki, Tokyo, Japan) were conducted at a cyclic speed of 1400 cycles/min with a constant strain amplitude and a stress ratio of *R* = −1 in air at room temperature based on Little's method [34]. The stress ratio of *R* = −1 was selected to indicate the effectiveness of the DryLP more clearly because both surfaces were treated. The fracture surfaces were observed using optical microscopy (Olympus, SZX7, Tokyo, Japan) and scanning electron microscopy (SEM; Hitachi, S-3000H, Tokyo, Japan).

The microstructures were observed to estimate dislocation densities in the specimens using a transmission electron microscopy (TEM; JEOL JEM-2010, Tokyo, Japan) with an acceleration voltage of 200 kV. For TEM observations, a small piece of the cross-section was thinned using a 30-keV-focused Ga-ion beam (Hitachi, FB-2000A, Tokyo, Japan).
