Ming
Ming
abc,
Fei
Gao
ag,
Jian-Huan
Wang
ace,
Jie-Yin
Zhang
ace,
Ting
Wang
acd,
Yuan
Yao
a,
Hao
Hu
cf and
Jian-Jun
Zhang
*abcd
aBeijing National Laboratory for Condensed Matter Physics and Institute of Physics, Chinese Academy of Sciences, Beijing 100190, China. E-mail: jjzhang@iphy.ac.cn
bCollege of Materials Science and Opto-electronic Technology, University of Chinese Academy of Sciences, Beijing 100049, China
cHefei National Laboratory, Hefei 230088, China
dSongshan Lake Materials Laboratory, Dongguan 523808, China
eBeijing Academy of Quantum Information Sciences, Beijing, 100193, China
fFrontier Institute of Science and Technology, Xi'an Jiaotong University, Xi'an 710054, China
gQilu Institute of Technology, Jinan 250200, China
First published on 21st March 2023
Ge/Si nanowires are predicted to be a promising platform for spin and even topological qubits. While for large-scale integration of these devices, nanowires with fully controlled positions and arrangements are a prerequisite. Here, we have reported ordered Ge hut wires by multilayer heteroepitaxy on patterned Si (001) substrates. Self-assembled GeSi hut wire arrays are orderly grown inside patterned trenches with post growth surface flatness. Such embedded GeSi wires induce tensile strain on the Si surface, which results in preferential nucleation of Ge nanostructures. Ordered Ge nano-dashes, disconnected wires and continuous wires are obtained correspondingly by tuning the growth conditions. These site-controlled Ge nanowires on a flattened surface lead to the ease of fabrication and large-scale integration of nanowire quantum devices.
The scalability of the NW-based quantum devices, however, is still a major challenge. It is a prerequisite to obtain NWs with controlled positions and geometries for NW-based device applications. Conventional vapor–liquid–solid (VLS) methods are able to form uniform NWs and NW-networks.19,20 However, these NWs typically grow out of the substrate plane and require to be transferred to a second substrate for device fabrication, which is an obstacle for scalability. The incorporation of a metal catalyst in VLS growth is another challenge for quantum devices, as it leads to decreased carrier mobility, parasitic channels and charge noises.21,22 The direct growth of in-plane Ge NWs and NW-networks has been demonstrated via selective area growth23 recently, but the crystalline imperfections such as dislocations and stacking faults still remain unsolved. By combining top-down fabrication and bottom-up self-assembly methods, site-controlled Ge hut wires (HWs) with perfect crystal quality have been previously demonstrated along the edges of pre-patterned trenches.16 But these trenches generally create certain challenges for subsequent device fabrication. Tremendous efforts have been made to grow wires inside the shallow trenches; however, only GeSi HWs with low Ge concentrations were obtained.24
Strain-driven formation of nanoscale islands on lattice-mismatched layers in hetero-epitaxy offers an effective way to obtain coherent islands25–29 and nanowires.30 Liu et al. found that27 the strain field induced by embedded strained islands in multilayer heteroepitaxial films leads to the formation of island columns with common size; their 2D model can be generally applied for the formation of multi-layer nanowires; however, due to the large lattice mismatch between Si and Ge, it is kinetically difficult to form long Ge wires.31,32 Here, we demonstrate uniform Ge HW arrays on a flattened surface by multi-layer growth of strained Ge(Si) layers separated with Si spacer layers on top of site-controlled GeSi HWs. We first fabricate shallow trenches to obtain GeSi HWs with fully controlled positions followed by a surface flattening process via Si spacer layer growth under optimized growth conditions. The embedded GeSi HWs generate medium tensile strain on the surface of the Si spacer layers which results in the vertical stacking growth of ordered (Ge)Si HWs. Furthermore, by tuning the growth conditions, other ordered nanostructures such as nano-islands and dumbbells can also be obtained.
Fig. 2 shows the growth process of Ge HWs. Fig. 2a is the AFM image showing the 1st HWs obtained after the deposition of a 6 nm thick Ge0.33Si0.67 alloy with subsequent 1 h in situ annealing. These 1st HWs are located inside the pre-patterned trenches. They are bound by {105} facets with approximately 9 nm in height and 2 μm in length. At both sides of the wires, there are still two shallow trenches existing with a depth of ∼5 nm, as observed from the AFM line-scan across the HWs (black line in Fig. 2f). The surface is flattened after the 21 nm thick Si spacer layer growth. The AFM image in Fig. 2b shows the 2nd HWs after 1.8 nm Ge0.41Si0.59 deposition with subsequent 1 h in situ annealing. They exhibit an identical wire length of 2 μm and a wire height of ∼8 nm. Shallow trenches appear at both sides of the 2nd HWs with a depth of ∼2 nm, as shown in the blue line of Fig. 2f. The appearance of these shallow trenches is attributed to the diffusion of Si from highly strained regions (edges of the HWs33) into the HWs, which is well known in the case of large Ge islands obtained at a high growth temperature.34,35 Outside these trenches, we observe satellite-islands with a height of ∼6 nm. These satellite-islands intend to evolve into satellite wires at a higher annealing temperature (not discussed in this work). The growth of the satellite islands can be attributed to the spread tensile strain over the surface of the Si space layer induced by the embedded HWs, where islands tend to nucleate next to the trench due to the attractive force from trench–island interactions.36
The 2nd spacer layer of 14 nm caps the 2nd HWs with an almost flattened surface. After subsequent deposition of the 6 Å Ge layer at 610 °C, the Ge HWs with a height of ∼7 nm are obtained, as shown in Fig. 2c. They have a sidewall angle of 11.3° (red line in Fig. 2f) corresponding to the {105} facets, as further confirmed by the HAADF-STEM image in Fig. 2d. The strain-induced trenches and satellite-islands are also observed at both sides of the Ge HWs. It needs to be mentioned that although a pure Ge layer is deposited, the Ge HWs actually consist of GeSi alloys due to the Si–Ge intermixing.15 The vertical alignment of the Ge(Si) HWs is characterized by the HAADF-STEM image in Fig. 2d. These HWs show a decreased lateral size with an elevated Ge content. We also see the truncated top of the 1st and 2nd layers of HWs, which are attributed to the Si–Ge intermixing during the spacer layer growth. Sharp interfaces between the HWs and the spacer layers are observed with no dislocations found in the cross-sectional area (Fig. 2d and e), indicating a high-quality crystal growth of Ge(Si) HWs. The morphological evolution of the three-layer growth is summarized with a line-scan as shown in Fig. 2f.
The overall site-controlled growth of the three-layer HWs is summarized as below. For the 1st layer HWs: the growth of HWs inside the shallow trenches with sidewall angles less than 10° (Fig. 2d and f) is driven by both the elastic relaxation and the exposed-surface minimization.37 For the upper layer HWs: the nucleation and growth of Ge(Si) nanostructures in the tensile strain regions induced by the embedded HWs is energetically favourable,26,27 in addition, the atomic diffusion of Ge contents over the Si spacer surface is also enhanced due to the tensile strain31,32 induced by the embedded HWs, which leads to the lateral growth of long Ge(Si) HWs.
Firstly, all parameters of the sample in Fig. 2c are kept constant except lowering the Ge growth temperature from 610 °C to 590 °C and 570 °C, respectively. The corresponding AFM images are shown in Fig. 3a and b. At a growth temperature of 590 °C, we observe disconnected HWs with a height of ∼6 nm and a length ranging from hundreds of nanometres to micrometres (Fig. 3a). They are all oriented along the 〈100〉 direction, identical to the pre-patterned trenches. At the growth temperature of 570 °C, ∼5 nm high nano-dashes with a length of tens or hundreds of nanometres are obtained. Furthermore, by performing 15 min in situ annealing at 570 °C after the Ge layer growth, these nano-dashes merge together into complete wires, as shown in Fig. 3d. They exhibit enlarged sizes with a height of ∼11 nm, which indicates a stronger Si–Ge intermixing and a lower Ge content. Dumbbells or matchsticks30 with islands positioned at the HW ends are also observed after the annealing. This is attributed to the induced tensile strain maxima on top of the HW endpoints by the embedded HWs, which is in line with the calculations in the following section.
Fig. 3 AFM images of the Ge nanostructure on the top of embedded HWs. (a) Disconnected wires and (b) nano-dashes are obtained under the same growth conditions as Fig. 2(c) except that the temperature was lowered to 590 °C and 570 °C, respectively. (c) Dot chains are obtained under the same growth conditions as in (a) except that the 2nd spacer thickness was increased to 21 nm. (d) The same growth parameters as (b) but follows a 15 min in situ annealing at 570 °C after the Ge layer growth. Scale bar: 1 μm. |
The influence of the spacer thickness on the formation of Ge nanostructures has also been studied. After the deposition of the Ge layer under identical conditions as indicated in Fig. 3a, we increased the 2nd spacer layer thickness to 21 nm. Ordered 1D chains of Ge islands are obtained, as shown in Fig. 3c. These Ge chains consist of ∼10 nm high hut islands and dome islands with a diameter of ∼110 nm and a height of ∼15 nm. By changing the growth temperature, chains of uniform dome islands can also be obtained. These results can be explained on the basis of the reduced tensile strain from the increased spacer thickness, which will also be discussed in details in the following section.
The surface strain field induced by both the 1st and 2nd layers of HWs on the 2nd spacer surface is shown in Fig. 4a. The 1st layer HWs have a Ge content of 0.33 with w1 = 90 nm (9 nm in height), l1 = 2000 nm and d1 = 35 nm. The 2nd layer HWs exhibit an increased Ge content of 0.41 with w2 = 80 nm (8 nm in height), l2 = 2000 nm and d2 = 14 nm. The HW dimensions are extracted from the sample in Fig. 2f. The embedded HWs generate tensile strain on the surface of the Si spacer layer right above the HWs, with existing compressive strain located surrounding the tensile strain area, as is further shown by the line-scan of the surface strain along the width and length directions passing through (0,0,0) orientations (red lines in Fig. 4c and d), respectively. It should be noted that tensile strain regions are energetically favorable for the nucleation and growth of Ge(Si) nano islands. Under certain growth conditions, these Ge(Si) nano islands can also merge into wires in intact form, as schematically illustrated in Fig. 4b. In addition, surface strain calculations are also performed at different Si spacer thicknesses. At an increased spacer thickness of d2 = 21 nm and d1 = 42 nm, the tensile strain drops dramatically and distributes more widely along the width direction (black lines in Fig. 4c and d). Ge on such a low tensile strained Si spacer layer has a large diffusion barrier, which leads to the formation of islands.
It is interesting to note that most of our results on HW formation (non-uniform distribution of the strain field and modification of the spacer thickness to the surface strain) are perfectly in line with previous works25–27,40–43 that focus on the growth of islands. Their model and analysis can be applied to multilayer nanowire formation universally. However, for Ge growth on a Si layer, the large lattice mismatch leads to a large diffusion barrier,31,32 making it difficult to form long Ge wires. In our approach, we first grow GeSi, which can easily form wires due to the medium lattice mismatch strain; the tensile strain in the Si spacer layer induced by the embedded GeSi wires favours the Ge diffusion and leads to long wire formation. The growth of the 1st and 2nd layers of GeSi HWs serves as somehow a “physical catalyst”, by lowering the kinetic barrier for the growth of the desired long Ge HWs.
Finally, it should be noted that the growth of self-organized strain induced HWs is a complex process that involves a number of conditions such as the growth temperature, the deposition rate, and the growth thickness. In our model, we focus on the growth regime where the chemical potential is mainly determined by the surface strain field. Once at a high growth temperature, the surface islands nucleate and grow continuously, leading to the formation of multifaceted domes with steeper facets. The surface energy44 and island–island repulsion36 would have a more pronounced influence on the nanostructure growth. A more advanced model (e.g., an extension of the approach proposed in ref. 44) would clearly be required for these growth regimes to complete the entire mechanism of lateral nanostructure growth by tackling subtle issues such as determining the critical spacer layer thickness of whether small islands transform into intact wires or domes. In addition, the Si–Ge intermixing45 during the growth would also need to be considered more carefully. However, some qualitative conclusion can be reached at the current stage. At a lower growth temperature (Fig. 3b), the diffusion length of surface atoms is reduced, therefore, an insufficient amount of Ge atoms is accumulated in the preferential growth regions, leading to the formation of only small islands or nano-dashes. Due to the island–island repulsion36 force and insufficient diffusion, these small islands or nano-dashes cannot merge into short wires but remain separated as island/dash arrays. But these closely spaced islands or dashes will merge together into disconnected wires (Fig. 3a) and eventually form intact wires only at an increasing growth temperature (Fig. 2c) or optimized in situ annealing (Fig. 3d).
This journal is © The Royal Society of Chemistry 2023 |