GeSn and GeSi are promising channel materials for gate-all-around (GAA) devices to achieve the high ION thanks to the high mobility [1, 2]. The undoped channels are used to reduce the impurity scattering and to further enhance the mobility. The GAA structure with high number of stacked channels can further enhance the ION for a fixed footprint. The dopant diffusion and segregation play an important role on GAA device structure design. In this work, the dopant diffusion and segregation effects of B and P in highly stacked Ge0.9Sn0.1 and Ge0.95Si0.05 epilayers are investigated, respectively.The Ge:B/Ge0.9Sn0.1/Ge:B epilayers were grown on the Ge-buffered SOI substrate at 320°C and 100 torr by chemical vapor deposition (CVD) using SnCl4, Ge2H6, and B2H6 precursors. The 160nm undoped Ge buffer was grown with an 800℃ annealing to confine the misfit dislocations at Ge/Si interface. For the 8 stacked Ge0.9Sn0.1 epilayers, 36 nm Ge0.9Sn0.1 layers sandwiched by 24 nm Ge:B layers were grown. Transmission electron microscopy (TEM) image of Ge:B/Ge0.9Sn0.1/Ge:B epilayers is shown in Fig. 1(a). The Ge:P/Ge0.95Si0.05/Ge:P epilayers were grown on the Ge-buffered SOI substrate at 350°C and 100 torr by SiH4, GeH4, and PH3 precursors using similar Ge buffer layers. For the 8 stacked Ge0.95Si0.05 epilayers, 24 nm Ge0.95Si0.05 layers sandwiched by 25 nm Ge:P layers were grown. TEM image of Ge:P/Ge0.95Si0.05/Ge:P epilayers is shown in Fig. 1(b). Note that both Ge0.9Sn0.1 and Ge0.95Si0.05 channel layers are unintentionally doped. In the 8 stacked Ge0.9Sn0.1 epilayers, the [B] in the Ge:B layers is as high as ~2.0 ×1021 cm−3 (Fig. 2(a)). In the 8 stacked Ge0.95Si0.05 epilayers, the [P] in the Ge:P layers is as high as ~2.2 × 1020 cm−3 , and the minimum [P] in the Ge0.95Si0.05 channel layers is from ~4.2 ×1017 cm−3 to ~1.5 ×1018 cm−3 (Fig. 2(b)).The B and P diffusion phenomena in the GeSn and GeSi layers are investigated, respectively. The deep undoped GeSi layers have higher [P] due to higher thermal budget (Fig. 3). Note that the P atoms diffuse from the Ge:P layers into GeSi channels during the epitaxial growth. The similar [B] ~3x1016cm-3 in the 2nd to 8th GeSn layers is due to the B detection limit of SIMS analysis (Fig. 3). Note that the decay length (nm/decade) is defined in Fig. 4 and Fig. 5. The decay length of [B] on the top side (4.3 nm/decade) is larger than that on the bottom side (2.9 nm/decade) due to SIMS knock-on effect [3] (Fig. 4). On the other hand, the P segregation [4] increases the decay length on the bottom side of GeSi layers (Fig. 5). Both the decay length of [B] and [P] increase with the increasing depth (Fig. 6) due to the higher thermal budget of the deeper layers in the CVD chamber. In addition, the difference of [B] decay length between top and bottom sides increases with the increasing depth of epilayers (Fig. 6). For the Ge:P/GeSi/Ge:P epilayers, the difference between top and bottom [P] decay length (less than 0.9 nm/decade) is smaller than that of [B] decay length in Ge:B/GeSn/Ge:B epilayers (larger than 1.4 nm/decade) due to P segregation, and the difference decreases for deep layers.AcknowledgmentThis work is supported by the Ministry of Science and Technology, Taiwan (MOST 110-2622-8-002-014-, 110-2218-E-002-030-, and 110-2218-E-002-042-MBK) and the Ministry of Education, Taiwan (NTU-CC-110L892601). The support by Taiwan Semiconductor Research Institute (TSRI), Hsinchu, Taiwan, is also highly appreciated.
Read full abstract
Jul 13, 2022
Shujun Zhang 
Open Access