When a 100 nm thick Si layer was transferred onto a bare Si wafer by the hydrogen-induced-layer-transfer process, a spongy damage layer with microvoids was formed on the transferred layer because of hydrogen blistering. The surface-to-volume ratio of the damage layer was greater than that of the layer where blistering did not occur. Therefore, the damage layer was selectively etched by an HF/H 2 O 2 mixture and completely removed ͑the etching rates of Si at 70°C for the bulk layer and damage layer were 0.45 and 13.8 nm/min, respectively͒. Consequently, a smooth, damage-free Si layer ͑root-mean-square roughness = 2.74 nm͒ was obtained.The hydrogen-induced-layer-transfer process, also known as the ion-cut process or smart cut process, is employed to fabricate silicon-on-insulator ͑SOI͒ wafers used in the manufacture of highend integrated circuits. 1 In this process, H + ions ͑dose: 3.5 ϫ 10 16 -1.0 ϫ 10 17 cm −2 2 ͒ are implanted in a Si wafer capped with a dielectric layer to obtain a layer of H + ions at a certain depth ͑hereafter referred to as "buried layer"͒ forming the desired layer for layer transfer. Subsequently, the H + -implanted wafer and a handle wafer are bonded together by an appropriate wafer bonding technique.Annealing the bonded pair at 500-600°C results in the formation of hydrogen microbubbles, and the aggregation of these microbubbles in the buried layer leads to crack propagation around the damage peak formed during ion implantation. 2 The desired layer is then transferred to the surface of the handle wafer. The surface of the as-split layer is rough because the buried layer is fractured by hydrogen blistering; the root-mean-square ͑rms͒ roughness of the as-split layer is approximately 8-11 nm 3 at an implant dose of 10 17 cm −2 of H + . Most of the defects generated during H + ion implantation are distributed on one side of the transferred layer, which results in the formation of a spongy damage layer on the transferred layer after the annealing process. The cross-sectional transmission electron microscopy ͑TEM͒ image in Fig. 1 shows that most of the lattice damage layer is filled with hydrogen ions, which is introduced by the implantation ͑H 2 + ion dose: 4 ϫ 10 16 cm −2 , 160 keV͒ on one side of the transferred layer. The as-implanted silicon wafer is covered by a 300 nm thick polycrystalline silicon layer and a 100 nm thick SiO 2 layer.The device performance of a fully depleted ͑FD͒ transistor is strongly dependent on the thickness of the SOI layer. In FD nanodevices, the thickness of the Si films should be controlled within an accuracy of a few nanometers for a good parameter reproducibility. For example, to suppress short-channel effects ͑SCEs͒ in 70 nm FD SOI transistors, it is necessary to reduce the thickness of the SOI layer to within 30 Ϯ 5% nm. 4 However, it is difficult to maintain a uniform thickness when the well-defined layer is thinned by chemical mechanical polishing ͑CMP͒.Because of the low mass of H + ions, it is usually necessary to thin the transferred layer by CMP for ac...