**1. Introduction**

Silicon carbide (SiC) is regarded as one of the most important wide-band gap semiconductors due to its excellent physical, electronic, and optical performances, i.e., a high melting temperature, a high strength, a high thermal conductivity, a large breakdown voltage, and a high electron mobility [1,2]. Much effort has been made to develop the potential applications of SiC devices, such as Schottky barrier diodes in next-generation, large-scale integrated circuits. Although SiC has more than 200 polytypes, the hexagonal 4H- and 6H-SiC are particularly promising due to their advanced physical properties.

To reduce the high cost of SiC wafers and improve SiC-devices, SiC-on-insulator (SiCOI) structures have been proposed because of their excellent performance, such as the low-power dissipation to save energy and the high radiation resistance to use in space [3]. Similar to many other semiconductors, SiCOI structures can be fabricated by "smart-cut" technology, which was first reported by Bruel [4] in 1995, to achieve silicon layer transfer for the fabrication of silicon-on-insulator (SOI) materials. The "Smart-Cut" technology contains three main processes, initially hydrogen or helium ion implantation with a fluence of the order of 10<sup>16</sup> to 10<sup>17</sup> cm<sup>−</sup><sup>2</sup> at room temperature, then wafer bonding to another rigid substrate (handling wafer) before thermal annealing, and finally fracture to achieve thin layer transfer at elevated temperatures [5–7]. The initiation and propagation of micro-cracks in H-implanted SiC play a critical role in exfoliation of the wafer surface. The formation of micro-cracks depends on the growth of platelets. These platelets are composed of vacancy-hydrogen compounds. The formation of vacancy-hydrogen compounds is due to the interaction between implantation-induced vacancies and implanted hydrogen. Therefore, it is critical to investigate the formation and growth of platelets in SiC implanted with H ions under di fferent experimental conditions, such as the implantation fluence, temperature, and annealing treatment. It is well known that the growth of micro-cracks inside the SiC wafer can induce surface blisters when the SiC wafer is not bonded to a substrate, and the same activation energy between blister formation and layer splitting is argued by Tong et al. [8]; therefore, it is a convenient way to evaluate the smart-cut threshold condition via observation of surface blisters and exfoliation. Our recent study involved 6H-SiC implanted by 134 keV H2 + at room temperature [9]. The maximum exfoliation e fficiency was achieved for the sample implanted with a fluence of 1.5 × 10<sup>16</sup> H2 +/cm<sup>2</sup> followed by 1100 ◦C annealing for 15 min. A further increase in implantation fluence was found to retard the exfoliation e fficiency due to the negative e ffects of implantation-induced lattice damage on the growth of vacancy-hydrogen clusters, consistent with the report of Gregory et al. [10] that the threshold fluence for exfoliation in H-implanted 4H-SiC decreases with increasing implantation temperature (room temperature to 600 ◦C). Up to now, most of the published reports aimed at H-implanted SiC were concerned with implantation at a low temperature and then annealing at a high temperature [11–17]. The exfoliation e ffect in H-implanted SiC without the annealing treatment was, to our knowledge, not investigated. Many open questions are concerned with the nature of the H implantation-induced defects and their influence on micro-crack growth. In this paper, we studied the exfoliation e fficiency of 6H-SiC implanted at 450 ◦C and subsequently annealed at 1100 ◦C for 15 min, compared with 6H-SiC implanted at 900 ◦C without annealing.

#### **2. Experimental Process**

For the experiments to study the exfoliation e fficiency of SiC as a function of implantation temperature, bulk SiC samples, 6H polytype <sup>&</sup>lt;0001>Si orientation, purchased as research grade material from HF-Kejing Company, Heifei, China, were implanted with 194 keV H2 + to a fluence of 5 × 10<sup>16</sup> H2 +/cm<sup>2</sup> at 450 and 900 ◦C. Hydrogen implantation experiments were performed on a 320 kV high-voltage platform equipped with ECR (Electron Cyclotron Resonance) ion sources in the Institute of Modern Physics, Chinese Academy of Sciences (CAS). The beam was rastered using an electrostatic scanner with fixed frequencies of 993 and 990 Hz in horizontal and vertical directions, respectively, to provide uniform ion fluence across the sample. The ion fluence was in-situ measured using a Faraday cup assembly in front of the sample. The beam flux was kept at 2.3 × 10<sup>13</sup> ions/cm<sup>2</sup> s. The implantation temperature was measured by a thermocouple, and the deviation of the implantation temperature was less than 1 ◦C. The wafers were tilted 7–8◦ from the direction of normal incidence during the implantation. According to the Stopping and Range of Ions in Matter (SRIM-2013) [18], the expected H peak concentration was approximately 11 at.% at 576 nm below the sample surface, as shown in Figure 1. To observe exfoliation on the surface of H-implanted 6H-SiC at 450 ◦C, thermal annealing at 1100 ◦C for 15 min in air atmosphere was performed.

**Figure 1.** Depth distributions of the displacements per atom (dpa) and the projected range of 194 keV H2<sup>+</sup>-implanted 6H-SiC to a fluence of 5 × 10<sup>16</sup> H2+/cm<sup>2</sup> simulated using the SRIM-2013 code (density of 3.21 g/cm<sup>3</sup> and displacement energies of C = 20 eV and Si = 35 eV).

Lattice damage before and after annealing was investigated by Raman spectroscopy and transmission electron microscopy (TEM) using a Tecnai G20 operated at 200 kV. Confocal Raman spectra were recorded at room temperature in a *z*(*xx*)*z* backscattering geometry using an HR-800 spectrometer from France. The 532 nm line of an argon ion laser was focused on a 1 × 1 μm<sup>2</sup> spot and collected through a 50× objective lens. A 100 μm confocal pinhole diameter was used, and 600 lines/mm grating were performed. The acquisition time for each spectrum was 30 s for one accumulation. The spectra were measured ranging from 150 to 1800 cm<sup>−</sup>1. A double tilt goniometer stage was used, in order to tilt the TEM sample to satisfy different diffraction vectors. The lattice defects were detected by weak-beam dark-field (WBDF) with (*g*, 3*g*), *g* = 0002 and *g* = 2110 near *z* = 0110, where *g* is the diffraction vector and *z* is the zone axis. To study the depth distribution of implantation-induced defects, cross-sectional samples were prepared. The fabrication process of the cross-sectional transmission electron microscopy (XTEM) samples was described as follows. Initially, XTEM samples were prepared by mechanical thinning up to approximately 30 μm in thickness, followed by ion milling with Ar ions in two steps. In the first step, the ion milling energy was 5 kV with a glancing angle of ±5◦ until optically controlled perforation occurred in the middle of the XTEM sample. In the second step, ion milling energy decreased to 2 kV with a glancing angle of ± 3◦ for 1 h to minimize radiation damage induced by the Ar ions [14,15]. The surface morphology was measured by scanning white-light interferometry (SWLI), and surface roughness after exfoliation was measured by atomic force microscopy (AFM).
