II. EXPERIMENT

The samples were (001) silicon films of thickness 280 nm, deposited on a $(10\bar 12)$ sapphire substrate of thickness 600 μm at 900 °C by chemical vapor deposition (CVD). The in-plane alignments are Si [100]||Al₂O₃$[\bar 1101]$ and Si [010]||Al₂O₃$[11\bar 20]$. The as-deposited SOS samples were subjected to a Si⁺ implantation at room temperature under two implanting energy levels, 135 and 200 KeV, and three dose densities, 10¹⁴, 10¹⁵, and 5 × 10¹⁵ ions/cm². To avoid any self-annealing during implantation, the temperature of the samples were kept at 25 °C by active cooling with liquid nitrogen. The implanted samples were then annealed in a furnace for 1 h at 600, 700, 800, 900, and 1000 °C, respectively, and cooled down inside the furnace for half an hour before the final cooling in air.

The residual stresses of the implanted and annealed samples were measured at room temperature by Raman Scattering with a Renishaw Invia unit equipped with backscattering configuration. A standard setup of 514 nm Argon ion laser and 1800 L/mm grating was utilized to detect the silicon band at around 520.5 cm⁻¹. The incident light was focused on the (001) surface of the silicon film using a 20x microscope objective. The scattered light was collected by the same objective. The spot size of the incident beam was about 2 μm and the penetration depth in Silicon was about 0.77 μm.^{Reference DeWolf33} Using this setup, a spectral resolution of 1.7 cm⁻¹ was obtained. The residual stress in the epitaxial silicon film was assumed to remain equi-biaxial^{Reference Liu, Zhang, Brawley, Atanackovic and Duvall16} and could be measured by the shift of the Raman band Δω (cm⁻¹). To find the peak position of the crystalline silicon (c-Si) band, the Raman spectra were fitted with Gaussians. The minima of the second derivative of the fitted function were then searched at around 520.5 cm⁻¹. In doing so, the in-plane normal stress σ could be uniquely related to the frequency shift Δω by σ = −249Δω (MPa).^{Reference Englert, Abstreiter and Pontcharra34} It has been shown that the residual stress calculated by this equation is consistent with the XRD stress measurement.^{Reference Liu, Zhang, Brawley, Atanackovic and Duvall16}

The microstructures of the re-growth SOS samples were studied using the transmission electron microscopy (TEM), using the cross-sectional specimens along silicon [110] prepared by a Nova 200 Nanolab (FEI, Hillsboro, OR) focused ion beam (FIB) system. In the FIB milling process, the silicon film was protected by Pt deposition on the surface. The TEM specimens were milled by Ga ion beam at 30 KV and 0.63 nA. The AutoTEM software was used to reduce the thickness to less than 100 nm for a better lattice view. The TEM specimen was tilted around the silicon <110> zone axis by a double tilt holder in a Philips CM-200 TEM (Philips, Hillsboro, OR) to obtain the bright-field images and the high resolution lattice images under a 200 KeV electron beam.

III. EXPERIMENTAL RESULTS

A. Effect of implantation

The ion implantation can modify microstructure of the as-grown film by introducing the vacancies and interstitial defects (Frenkel pairs). The accumulation of these defects results in the dislocation loops. When the deformation energy reaches a critical value of 12 eV/atom, the crystalline-to-amorphous phase transformation occurs.^{Reference Narayan35,Reference Gibbons36} The depth and thickness of the amorphous phase depend on the dose density and implantation energy. In the Raman spectra of the implanted specimens, as shown in Figs. 1(a) and 1(b), the occurrence of the broadened peak at a small wave number indicates the phase transformation from c-Si to amorphous silicon (a-Si). The results showed that under both the energy levels, the amorphous fraction increases with increasing the Si⁺ doses. Narayan and Holland^{Reference Narayan35} have shown that the amorphous phase nucleates at the region of the maximum number of collisions and then expands toward the surface and the deeper region with the increase of implanted ion doses. With the same dose density, a higher energy (e.g., 200 KeV) resulted in a larger amorphous fraction. This is because an implanting ion, which carried more energy can penetrate deeper, collide with more lattice atoms, and have a longer mean-free path than the lower energy condition. The nominal distributions of collision events were simulated by the software TRIM (Transport of Ions in Matter), as shown in Fig. 1(c) for energy levels of 135 and 200 KeV. It reveals that under the extreme conditions in the present study, i.e., 200 KeV and 5 × 10¹⁵ ions/cm², the ion implantation induced damage should have significantly changed the silicon/sapphire interface. Moreover, in both of the conditions, the least collisions occur in the layers close to the surface, where the crystalline structure should be preserved. This provides seeds for the film re-grown in annealing.^{Reference Lau, Matteson, Mayer, Revesz, Gyulai, Roth, Sigmon and Cass14}

FIG. 1.

Raman spectra on the as-implanted samples subjected to the energies of (a) 135 KeV and (b) 200 KeV; (c) TRIM simulation of collision events. The curves E14, E15, and 5E15, respectively stand for Si⁺ dose densities of 10¹⁴, 10¹⁵ and 5 × 10¹⁵ ions/cm².

The variations of residual stresses in the specimens implanted with different energies and dose densities are summarized in Fig. 2. Clearly, the implantation brought about a significant reduction of residual stresses in the silicon film. The increase of both the ion density and implantation energy resulted in more stress relief. This is because the heavier doping conditions give rise to more vacancies, dislocations and amorphous transformations, so as to alleviate more elastic strain (compressive) in the lattice. However, it is surprising to note that the residual stress became tensile under the extreme condition—dose density of 5 × 10¹⁵ ions/cm² and implantation energy of 200 KeV. This is unexpected because the effect of amorphization should only alleviate the existing stress rather than induce a stress of opposite direction. This will be further discussed in Sec. IV.

FIG. 2.

Room temperature residual stresses measured on the as-implanted samples.

B. Effect of annealing

Annealing after implantation is to restore the crystalline structure through epitaxy re-growth. Fig. 3 shows the Raman spectrum of the annealed samples (implantation conditions 200 KeV and 5 × 10¹⁵). The annealing at 600 °C has smeared out most of the amorphous phase, leaving only a sharp peak at about 521 cm⁻¹. This indicates the domination of the c-Si after annealing. It is also noted that increasing the annealing temperatures brought about a better crystalline quality of the Si film, evidenced by the decrease of the FWHM (full width half maximum) of the c-Si Raman band as shown in Fig. 3(b). A possible explanation for such crystalline quality improvement is that the higher temperature has resulted in a higher mobility of dislocations, thereby brought about their higher rate of annihilation.^{Reference Aindow, Batstone, Pfeiffer, Phillips and Pond25}

FIG. 3.

(a) Raman spectrum of the implanted samples annealed at different temperatures, and (b) Variations of Raman FWHM with annealing temperatures. The samples were implanted with Si ⁺ at 200 KeV and 5 × 10¹⁵ ions/cm².

The apparent right shifts of Raman peaks with increased annealing temperatures, as demonstrated in Fig. 3(a), indicate the restore of compressive residual stresses in the silicon film. To better understand the effect, the variation of residual stresses versus annealing temperature is given in Fig. 4(a), where the continuous and dashed lines pertain to samples subjected to 200 and 135 KeV implantations, respectively. The result shows that a higher annealing temperature renders greater residual stress regeneration. For the samples implanted with smaller dose densities, the annealing at the deposition temperature (900 °C) can regenerate the residual stresses to the as-grown level. It is noted that the lighter the doping (i.e., smaller implantation energy or less dose density), the lower annealing temperature is needed to restore the residual stresses to the as-grown level.

FIG. 4.

Stress variations with annealing temperatures. The comparisons are to illustrate the effects of (a) implanting energies and (b) annealing method on residual stresses. Si⁺ dose densities are 10¹⁴, 10¹⁵ and 5 × 10¹⁵ ions/cm² for the curves E14, E15, 5E15, respectively.

The highest rate of residual stress regeneration was observed in the sample subjected to the heaviest implantation. As indicated by the solid black curve in Fig. 4(a), this rate is consistent with the variation rate of residual stresses obtained by taking into account only the mismatch of coefficients of thermal expansion (CTE).^{Reference Moridi, Ruan, Zhang, Liu and Haddad37} However, the whole stress-temperature curve shifts upwards with a rather constant offset of about 150 MPa. This indicates that heating to and annealing at different temperatures cannot remove the tensile stresses induced by implantation. If the tensile stress is due to the amorphization of the Si layer, it should be effectively removed after the annealing and crystal re-growth. This, however, did not take place. Therefore, the implantation-induced tensile stress must have its structural origin at the interface.

To further confirm the above findings, let us examine a 2-step annealing method,^{Reference Ohmura, Inoue and Yoshi27} with which an implanted sample was first annealed at 600 °C for an hour and then further annealed at temperatures ranging from 700 to 1000 °C. Figure 4(b) shows the stresses measured at room temperature after the 2-step annealing. For all the implanted samples studied, it was found that the variation of the stresses with the highest annealing temperature is consistent with that obtained from the single step annealing shown in Fig. 4(a). It is therefore reasonable to conclude that the final residual stresses are only dependent on the highest annealing temperature and are not affected by the 1st annealing at 600 °C.

IV. STRESS RELEASE MECHANISM

The above experimental results demonstrate that different implantation and annealing conditions indeed brought about different residual stresses. As such, the following questions need to be answered:

1. (i) Why do different ion densities and implantation energy levels render different residual stress relief?

2. (ii) Why do samples subjected to different implantation conditions exhibit dissimilar stress variations during their subsequent annealing?

3. (iii) What causes the tensile stress in the sample that was most heavily implanted?

It has been shown that with the increase of ion doses, the accumulation of vacancies and interstitial point-defects brings about dislocation loops and amorphous clusters, which results in more stress mitigation. With the same dose density, a higher energy further enhances the stress relief by two mechanisms, as illustrated in Fig. 5. Firstly, compared to the lower energy condition [Fig. 5(a)], the high-energy [Fig. 5(c)] implantation shifts the maximum number of the collisions toward the interface, leading to more amorphous phase in the vicinity of the interface. With a greater interface amorphization, the lattice mismatch stress could be more effectively alleviated. Secondly, as shown from the TRIM simulation of the ion ranges [Figs. 5(b) and 5(d)], silicon ions with a larger energy can penetrate into the substrate, resulting in substitutional point-defects in the sapphire substrate. These excessive ions eventually expand the sapphire lattice, and lead to the tension in the film. It can be expected that at a certain density of the ion dose, e.g., at about 4 × 10¹⁵ shown in Fig. 3, the two effects will bring about a full relaxation of the compressive residual stress and even the tensile stress in the film.

FIG. 5.

Schematics of implantation energy effects on the microstructures (a) and (c), and the ion resting ranges (b) and (d) simulated by TRIM software; the implantation energy was 135 KeV for (a) and (b) and 200 KeV for (c) and (d).

The effects of implantation of different energy levels were exhibited by the TEM images as shown in Fig. 6. For the low-energy implantation [Fig. 6(a)], owing to the limited penetration depth of the ions, the crystalline interface between film and substrate is still preserved, which is proven by the diffraction pattern [Fig. 6(c)] taken from the interface region “B” of Fig. 6(a). In Fig. 6(c), the overlapped diffraction patterns of the single crystalline silicon and sapphire are well discerned. Inside the silicon film, the high dose density of 5 × 10¹⁵ ions/cm⁻² has fully transformed c-Si to a-Si, as indicated by the diffraction pattern [Fig. 6(d)] taken from the region “A” of Fig. 6(a). In contrast, the higher-energy implantation [Fig. 6(b)] has completely transformed the interface to be amorphous and produced a defective zone in sapphire adjacent to the interface, whereas this defective zone was not observed under the low-energy condition.

FIG. 6.

TEM bright field images of as-implanted SOS subjected to the doping energies of (a) 135 KeV and (b) 200 KeV; (c) and (d) are respectively the diffraction patterns taken at the interface “B” and amorphous area “A” in (a). The ion dose density for both of the specimens was 5 × 10¹⁵ ion/cm². The samples were tilted to the sapphire <0112> zone axis.

Different implantation conditions give rise to different interface structures. For the lightly doped samples that preserve the as-grown interface structures, the primary effect of annealing was to recover the implantation defects (e.g., vacancies and dislocation loops). In this case, an annealing temperature as high as 900 °C was sufficient to smear out most of those defects, and restored the residual stress to the as-grown level. However, the heavy implantation could substantially modify the interface structure. As shown previously in Fig. 4(a), the high temperature annealing restored the crystalline structure and the subsequent cooling led to the thermal stress induced by the CTE mismatch. However, the constant offset of about 150 MPa from the reference curve seems to imply that the expansion in the substrate could not be removed by annealing.

The microstructural variations from the as-grown and re-grown SOS are illustrated in Fig. 7. Compared with the as-grown film [Fig. 7(a)], the TEM images of samples annealed at 600 °C [Fig. 7(c)] and 1000 °C [Fig. 7(e)] after the heaviest implantation (200 KeV, 5 × 10¹⁵ ions/cm²) exhibit the defective region in the sapphire substrate, which spreads about 100 nm beneath the interfaces. The depth of this defected zone is consistent with the ion resting range obtained from TRIM simulation [Fig. 5(d)], indicating the deep penetration of the high-energy Si⁺ ions into the substrate. The high temperature annealing does not recover this defected zone, henceforth, bringing about the variation of residual stresses in the above silicon layer. The Si⁺ ions deflect the sapphire lattice by introducing point-defects, edge dislocations, and most importantly, cause the local expansion in the sapphire lattice.

FIG. 7.

TEM bright field images of (a) as-grown (c) 600 °C and (e) 1000 °C annealed SOS; and the corresponding high resolution images for (b) as-grown, (d) 600 °C annealed and (f) 1000 °C annealed SOS to reveal the atomic structure of the silicon and sapphire interface; the samples were implanted under the energy of 200 KeV and ion density of 5 × 10¹⁵ ion/cm² prior to annealing.

Fig. 8 exhibits a pair of edge dislocations in the sapphire lattice, 20 nm beneath the interface. These dislocations are different from the periodical misfit dislocations shown in the as-grown SOS [Fig. 7(b)]. The latter was formed at the growth process to accommodate the large lattice mismatch through DME.^{Reference Narayan and Larson17,Reference Bayati, Molaei, Narayan, Narayan, Zhou and Pennycook18} They appear at the interface and are nucleated from the surface steps as shown in Fig. 7(b). In contrast, the implantation-induced dislocations in sapphire lattice were 20 nm away from the interface. These dislocations could be the evidence of the distortion of the sapphire lattice due to excessive Si⁺ ions since similar defects cannot be found in the as-grown specimen. This supports our deduction and explains why the residual stresses measured are smaller than the calculated thermal stresses.

FIG. 8.

A high resolution TEM image of the sapphire lattice at the depth of 10–20 nm beneath the interface. The sample was subjected to the 200 KeV and 5 × 10¹⁵ ions/cm² implantation and 1000 °C annealing.

Finally, our results showed that different annealing temperatures would lead to different interface structures too. Under a low annealing temperature (say 600 °C), the interface still carries the memory of the implanted structure and remains amorphous, as shown in Fig. 7(d). This amorphous region acts as a buffer layer to accommodate the lattice mismatch strain. A sufficiently high temperature of 1000 °C could restore well the crystalline structure of the interface, making it similar to that of the as-grown film [Fig. 7(b)]. The lattice and misfit dislocations could be clearly discerned under the high-resolution TEM [Fig. 7(f)].