A distal end portion for pressing an object to be joined of one of a resonator and a reception jig is constituted as a protrusion having an outer face not including an angular sharp edge, such as a hog-backed shape or a semispherical shape, and during joining, while metal foils stacked are being laterally thinly stretched with lateral vibration from the resonator and atoms of the metal foils are being moved laterally without the protrusions forcibly holding down the metal foils, a metal amount corresponding to a volume stretched is smoothly discharged along a curved face without breaking the metal foils, and even when the stacked metal foils are used as the object to be joined, breaking of the metal foils is not generated during joining, so that a step of protecting the metal foils with a protection member is unnecessary.