Another factor that contributes to the lack of consensus between bc8/r8 annealing studies is the similarity of hd-Si, Si-XIII, and nc dc-Si as revealed by the key techniques often used for phase identification, namely Raman microspectrocopy and SADP. The Raman peaks associ- ated with the different exotic phases and thed-spacings that form the characteristic reflections in SADP (among other diffraction-based measuring techniques) are given in detail in Chap- ter 3. However, the relevant sections are reproduced here in Table 6.1 and Table 6.2 for convenience.
As Raman shift is sensitive to crystal size [208], the Raman peak from nanocrystalline (nc) dc-Si will be shifted relative to bulk dc-Si [207]. In particular, the TO peak of nc dc-Si may be down-shifted from 520 cm−1 such that there is considerable overlap with the main hd-Si peak at 514 cm−1, potentially obscuring the presence of hd-Si. The magnitude of this shift is dependent on the crystallite size. Similarly, the primary Si-XIII peak at 497 cm−1 will overlap with the hd-Si peak at 496 cm−1. The deconvolution of these overlapping peaks is further complicated by the presence of a-Si which has a broad Raman peak centred at∼480 cm−1.
SADP is commonly used to support Raman results in providing phase identification. However, apart from a [102] reflection, all hd-Sid-spacings observed in SADP can be attributed to other 1as long as the transformations to dc-Si from the crystalline phases occur at different temperatures to the
Raman Shift (cm−1) Phase Phonon Mode Reference
301.9 dc 2TA [173]
520.3 dc TO [173]
510 - 520 dc (nc) TO [207]
480 a-Si Broad Peak [45]
496 hd TO [174] 514 hd A1g [174] 200 Si-XIII [44] 330 Si-XIII [44] 475 Si-XIII [44] 497 Si-XIII [44]
Table 6.1: The Raman peaks associated with the dc, hd, a-Si, and Si-XIII phases of Si. The primary peak of each phase is highlighted in bold text. A nanocrystalline material may result in a change in Raman shift. A relevant shift in the dc-Si TO peak is also included.
dc-Si bc8-Si hd-Si
hkl d-spacing (˚A) hkl d-spacing (˚A) hkl d-spacing (˚A)
111 3.125 110 4.695 100 3.29 220 1.920 200 3.318 002 3.14 311 1.637 211 2.709 101 2.91 220 2.346 102 2.27 321 1.774 110 1.90 400 1.659 103 1.77 200 1.65 112 1.63
Table 6.2: The d-spacing for dc-Si, bc8-Si, and hd-Si. The unit cell parameters used for the calculation are taken from Ref. [184] (dc-Si), Ref. [32] (bc8-Si), and Ref. [142] (hd-Si). The difference in significant figures for the values presented for hd-Si are due to the difference in significant figures in the reported unit cell parameters.
phases that may be present alongside hd-Si. That is, if no reflections at 2.27 ˚A are observed2, it is possible to incorrectly classify a hd-Si sample as a mixture of dc-Si and bc8-Si instead. In particular, the [100], [002], and [101] reflections from hd-Si can easily be mistaken for the [111] reflections of dc-Si (especially if each individual reflection spot in the SADP is quite large). It is also possible to miss Si-XIII in samples thinned for TEM. For example, it was reported that the presence of Si-XIII was dependent on sample thickness, which suggests Si-XIII may not be stable in thinned samples [44]. Therefore, the absence of Si-XIII reflections in SADP may not necessarily mean the absence of Si-XIII in the pre-thinned sample.
The results presented in this chapter aim to establish a method for unambiguous identification of hd-Si. After this, measurements from indents annealed using furnace annealing, incremental annealing, and laser annealing are presented to form a more complete picture of the transfor- mation pathways of the bc8/r8 mixed structure. This includes measurements from previously published works that are re-analysed using this new method for identification. Such cases have been indicated in the text. In particular, the stable temperature range of hd-Si, the transformations to and from Si-XIII, and the differences between laser annealing and thermal annealing are explored.
6.2
Experimental Details
Indents were made using an Ultra-Micro-Indentation System (UMIS) 2000 using a ∼40 µm diameter spherical tip. Indentation to a maximum load of 700 mN with a load rate of ∼5 mN/s (50 increments) and an unload rate of∼1.3 mN/s (200 increments) were undertaken. It should be noted that the UMIS load/unload rate is controlled by the number of incremental steps taken to load/unload rather than an exact load rate. These conditions were chosen as they are known to produce circular regions of the bc8/r8 structure∼10µm wide and∼400 nm deep [39, 41, 93, 112]. The size of these indents is comparable to the “large” indents presented in the work by Ruffellet al. [41].
Furnace annealing was performed at temperatures between 50 - 750◦C for 2 hours in a N2
atmosphere unless otherwise stated. A different sample was used for each temperature. Stage annealing was performed in 10◦C increments, from 30 - 240◦C, in a N2 atmosphere. A single
sample was used for the entire temperature range. The sample was brought up to temperature and held for 200 s. After each step, the sample was allowed to cool back to RT before Raman spectra were taken to prevent temperature related shifts in the spectra. Laser annealing was performed using a continuous wave (CW) 532 nm laser operating at 400 kW/cm2 with a spot size of∼1 µm in diameter. To ensure a relatively homogeneous transformed region, the laser
2
was raster scanned across the surface at a rate of 1 µm/s. All samples were imaged using XTEM and SADP after annealing.
The peak positions measured using Raman microspectroscopy were compared with those calcu- lated using theab initioframework of the density functional theory (DFT). These calculations were provided by Prof. Andres Mujica from Universidad de La Laguna as a part of a joint project.