FIG 6 - uploaded by Thomas Kuech
Content may be subject to copyright.
Dislocation are observed in the XTEM micrographs obtained from the 1200 nm SiGe films grown on ͑ a ͒ bulk Si and ͑ b ͒ 330 nm SOI substrates. An arrow indicates the placement of the SiGe–Si interface. The stacking fault in ͑ b ͒ has accumulated at least six dislocations that have glided to it from the right-hand side.
Source publication
The strain relaxation behavior of Si0.82Ge0.18 films on silicon-on-insulator (SOI) substrates was investigated for films grown beyond the critical thickness and strain-relaxed during growth and metastable films, grown beyond the critical thickness, which relaxed during subsequent thermal annealing. The thickness of the top silicon layer of the SOI...
Contexts in source publication
Context 1
... 8. AFM micrographs obtained from the in situ relaxed SiGe layers which were 1200 nm thick and grown on substrates consisting of ͑ a ͒ bulk Si, ͑ b ͒ a 10 000 nm SOI substrate, ͑ c ͒ a 330 nm SOI substrate, and ͑ d ͒ a 40 nm SOI substrate. The long, even trenches apparent in Fig. 6 ͑ a ͒ are contrasted with the much shorter shallower trenches in ͑ c ͒ and ͑ d ͒ .
...
Context 2
... when the oxide serves to form the compliant medium within the compliant substrate structure. After annealing at 1050 °C, the diffraction peak from the Si–SOI layer diminished greatly in intensity and the peak shifted toward the SiGe diffraction peak. This peak shift is due to significant interdiffusion between the SiGe and Si– SOI layer that is apparent after annealing at 1050 °C. Nu- merical simulations of the interdiffusion process confirm that the interdiffusion would lead to a Ge compositional gradient across the Si–SOI layer from 18% Ge at the SiGe–Si–SOI interface to at least 1% Ge at the oxide interface after a 1050 °C anneal. The X-ray diffraction and AFM measurements have con- firmed the presence of dislocations that relax the SiGe film. XTEM was also carried out on samples annealed at 950 °C to determine the distribution of these dislocations within the material. Dislocations are easily observed in films on both bulk Si and SOI substrates as shown in Fig. 4. For films grown on the bulk Si substrate ͓ shown in Fig. 4 ͑ a ͔͒ dislocations loops appear at least 2 m into the Si substrate. Pairs of associated dislocations on parallel glide planes are observed in the substrate as previously described in Ref. 20. The substrate dislocations in the SOI substrate simply cross the 40 nm Si–SOI layer to the buried oxide. There is no dislocation- based contrast along the Si–SOI/SiO 2 interface in the TEM micrographs. This lack of contrast in the TEM micrographs has been directly observed at room temperature for dislocations at metal/amorphous material interfaces. 21–23 In these previous studies, the dislocation core was found to dissipate by core-spreading at the interface, reducing the elastic dis- tortion and energy of the dislocation core. This core- spreading is likely to be taking place at the Si/SiO 2 interface in this study. The amorphous layer is able to reorient, on the atomic scale, to absorb the displacement of the core leading to a final result that is very similar to the behavior of a dislocation reaching or interacting with a free surface. While a larger density of dislocations is observed extending into the Si–SOI layer than into the companion bulk Si substrate under similar growth conditions, no film threading dislocations were observed in the small areas of the Si– SOI layer probed by the XTEM, putting an upper limit on the threading dislocation density near 10 7 per cm 2 . An equivalent density of threading dislocations is observed in the film grown on the bulk Si substrate. In the previous series of samples, the 150 nm films grown on the bulk Si or SOI substrates remain fully strained following growth and cooling to room temperature. An increase in the film thickness to 340 nm however leads to the initiation of film relaxation on all the substrates during the growth process. The Pendellosung thickness fringes present in the RSMs of Figs. 2 ͑ a ͒ and 2 ͑ d ͒ diminish in intensity and the SiGe peak broadens as a result of the in situ strain relaxation. The misfit dislocation density is still too low, however, to cause a measurable shift in the diffraction peak position. Film growth to 765 nm and 1200 nm did result in observable amounts of strain relaxation as indicated by peak shifts in the respective RSMs. The triple crystal RSMs from these films, grown on a bulk Si substrate, 330 nm and 40 nm SOI substrates, are presented in Fig. 5. The degree of film relaxation, as indicated in the peak shift and peak width, is nearly identical for the 765 nm and 1200 nm thick films grown on all the various substrates. The only variation in the RSMs with a substrate is the shift in the position of the Si–SOI diffraction peak. The peak from the bulk Si substrate is also broadened upon the growth of the SiGe film, similar to the 330 nm Si–SOI layer, whereas the 40 nm Si–SOI diffraction peak is below the detection limits of the x-ray measurement system. It is notable from these RSMs that the peak position of the 330 nm Si–SOI peak is shifted to positive Bragg angles. Analysis of the 40 nm and 70 nm SOI samples, with the lower resolution but higher signal intensity detector slit, allows the shift of the diffraction peak from the thin Si–SOI layer to positive Bragg angles to also be observed. This peak shift was not detectable in the 10 000 nm SOI sample. A larger Si–SOI peak shift is observed for the 1200 nm film over the 765 nm film as expected from the greater strain relaxation in the 1200 nm film. This positive shift of Bragg angle indicates a reduction in the Si–SOI out-of-plane lattice constant and a resultant tensile strain. Table II contains a summary of the x-ray data for the in situ relaxed samples. The full width at half maximum is measured in the tilt direction of Fig. 5. TEM micrographs from the in situ relaxed samples confirm the presence of dislocations. Dislocations are also indicated by the crosshatch and the increased x-ray peak breadth observed for these samples. XTEM micrographs in Fig. 6 show a high density of misfit dislocations at the SiGe/Si interface. Film threading dislocations are observed in all the samples as well as a high density of dislocations that extend and terminate at the buried oxide. As with the annealed samples, the dislocations that reach the oxide do not show any contrast in the TEM micrograph indicating that the strain field associated with the core of the misfit dislocations at the Si/SiO 2 interface has been largely dissipated. Additionally, stacking faults are present in all the thick relaxed films grown on both bulk Si and SOI substrates. The presence of stacking faults has been previously reported when relaxing highly strained films, such as Ge on relaxed Si Ge , 24 and in a graded SiGe buffer layer grown with a grading rate above Ͼ 27% per m. 25 In the XTEM micrograph of Fig. 6 ͑ b ͒ , a stacking fault with at least six associated dislocations is observed. The density of dislocations indicates that the stacking fault strongly interacts with the threading dislocations. The PVTEM micrographs of the 1200 nm films re- laxed on bulk Si, 300 nm SOI, and 40 nm SOI, are shown in Fig. 7. These specimens were prepared by thinning from the substrate side so that only the SiGe film is imaged by TEM. All three samples have stacking faults and threading dislocations present. The blocking of dislocation motion by stacking faults, prevalent on all the substrates, increases the TDD which ranges from 10 cm on bulk Si, to 5 10 cm on 330 nm SOI, and 2 ϫ 10 7 cm Ϫ 2 on 40 nm SOI. Many of the dislocations in the PVTEM micrographs lie on common glide planes. This arrangement is the product of a dislocation multiplication process during film relaxation, which has been previously observed in SiGe films. 20,26 The buried oxide of the SOI substrate does not appear to interfere with these multiplication processes. The AFM micrographs of all the in situ relaxed films exhibit the dislocation-derived crosshatch, as shown in Fig. 8. These films have a similar rms roughness ranging from 4.9 to 6.4 nm. Morphological differences are observed, however, on a finer scale. All of the surface trenches run the full width of the scan range for the SiGe films grown on the bulk Si substrates. This behavior is also present on the 10 000 nm SOI, which also has a high density of surface mounds. On the 330 nm and 40 nm Si–SOI substrates, only the deepest trenches run the length of the 10 m scans. Many shallow trenches run only a short distance. This is contrasted with the crosshatch structure of the films that relaxed during the ex situ annealing shown in Fig. 3. All of these ex situ annealed films exhibited very long uniform trenches, which were only present on the in situ relaxed bulk Si and 10 000 nm SOI samples. Similar in situ observations were made using low- energy electron microscopy, as described in Ref. 27. The dependence of the SiGe film relaxation on both the temperature and film thickness for Si and SOI substrates, whether induced by high-temperature annealing or during growth, is essentially identical. In all cases, a large density of misfit dislocations at the Si/SiGe interface are inferred from the increased x-ray diffraction peak width, the surface crosshatch, and, more directly, in the TEM micrographs. These results cannot be attributed to a compliant substrate mechanism, which would relax the film without the introduction of dislocations in the SiGe film. In addition, the amount of strain transfer to the Si–SOI layer, required by the compliant substrate mechanism, was not observed in the x-ray diffraction. Compliant substrate theory predicts a tensile strain in the Si–SOI layer. This tensile strain is only observed in this study after extensive dislocation-based film relaxation. In both the ex situ and in situ samples, an appreciable dislocation density is also apparent from the diffraction peak width. Compliant substrate strain sharing between the SiGe film and the Si–SOI layer was not present in any of the growth or annealing experiments. Film relaxation was only observed due to dislocation nucleation and propagation. Dislocations nucleate and propagate across the sample relaxing the film, with segments extending into the substrate. The dislocations in either the bulk Si or Si–SOI layers are the product of the strong dislocation interactions, which push dislocations into the underlying layers. These dislocation interactions have been observed and modeled for a variety of cases including dislocation multiplication processes during perpendicular and parallel crossings of dislocations. 20 The strain field interaction between dislocations provides the driving force to move the dislocations into the Si–SOI layer and has a far greater influence on dislocation motion than the dislocation image force attributed to the buried oxide. 15 On the SOI substrates, the dislocations penetrating into the Si–SOI layer can reach the buried amorphous Si–SOI/oxide interface. The substrate ...
Context 3
... distribution of these dislocations within the material. Dislocations are easily observed in films on both bulk Si and SOI substrates as shown in Fig. 4. For films grown on the bulk Si substrate ͓ shown in Fig. 4 ͑ a ͔͒ dislocations loops appear at least 2 m into the Si substrate. Pairs of associated dislocations on parallel glide planes are observed in the substrate as previously described in Ref. 20. The substrate dislocations in the SOI substrate simply cross the 40 nm Si–SOI layer to the buried oxide. There is no dislocation- based contrast along the Si–SOI/SiO 2 interface in the TEM micrographs. This lack of contrast in the TEM micrographs has been directly observed at room temperature for dislocations at metal/amorphous material interfaces. 21–23 In these previous studies, the dislocation core was found to dissipate by core-spreading at the interface, reducing the elastic dis- tortion and energy of the dislocation core. This core- spreading is likely to be taking place at the Si/SiO 2 interface in this study. The amorphous layer is able to reorient, on the atomic scale, to absorb the displacement of the core leading to a final result that is very similar to the behavior of a dislocation reaching or interacting with a free surface. While a larger density of dislocations is observed extending into the Si–SOI layer than into the companion bulk Si substrate under similar growth conditions, no film threading dislocations were observed in the small areas of the Si– SOI layer probed by the XTEM, putting an upper limit on the threading dislocation density near 10 7 per cm 2 . An equivalent density of threading dislocations is observed in the film grown on the bulk Si substrate. In the previous series of samples, the 150 nm films grown on the bulk Si or SOI substrates remain fully strained following growth and cooling to room temperature. An increase in the film thickness to 340 nm however leads to the initiation of film relaxation on all the substrates during the growth process. The Pendellosung thickness fringes present in the RSMs of Figs. 2 ͑ a ͒ and 2 ͑ d ͒ diminish in intensity and the SiGe peak broadens as a result of the in situ strain relaxation. The misfit dislocation density is still too low, however, to cause a measurable shift in the diffraction peak position. Film growth to 765 nm and 1200 nm did result in observable amounts of strain relaxation as indicated by peak shifts in the respective RSMs. The triple crystal RSMs from these films, grown on a bulk Si substrate, 330 nm and 40 nm SOI substrates, are presented in Fig. 5. The degree of film relaxation, as indicated in the peak shift and peak width, is nearly identical for the 765 nm and 1200 nm thick films grown on all the various substrates. The only variation in the RSMs with a substrate is the shift in the position of the Si–SOI diffraction peak. The peak from the bulk Si substrate is also broadened upon the growth of the SiGe film, similar to the 330 nm Si–SOI layer, whereas the 40 nm Si–SOI diffraction peak is below the detection limits of the x-ray measurement system. It is notable from these RSMs that the peak position of the 330 nm Si–SOI peak is shifted to positive Bragg angles. Analysis of the 40 nm and 70 nm SOI samples, with the lower resolution but higher signal intensity detector slit, allows the shift of the diffraction peak from the thin Si–SOI layer to positive Bragg angles to also be observed. This peak shift was not detectable in the 10 000 nm SOI sample. A larger Si–SOI peak shift is observed for the 1200 nm film over the 765 nm film as expected from the greater strain relaxation in the 1200 nm film. This positive shift of Bragg angle indicates a reduction in the Si–SOI out-of-plane lattice constant and a resultant tensile strain. Table II contains a summary of the x-ray data for the in situ relaxed samples. The full width at half maximum is measured in the tilt direction of Fig. 5. TEM micrographs from the in situ relaxed samples confirm the presence of dislocations. Dislocations are also indicated by the crosshatch and the increased x-ray peak breadth observed for these samples. XTEM micrographs in Fig. 6 show a high density of misfit dislocations at the SiGe/Si interface. Film threading dislocations are observed in all the samples as well as a high density of dislocations that extend and terminate at the buried oxide. As with the annealed samples, the dislocations that reach the oxide do not show any contrast in the TEM micrograph indicating that the strain field associated with the core of the misfit dislocations at the Si/SiO 2 interface has been largely dissipated. Additionally, stacking faults are present in all the thick relaxed films grown on both bulk Si and SOI substrates. The presence of stacking faults has been previously reported when relaxing highly strained films, such as Ge on relaxed Si Ge , 24 and in a graded SiGe buffer layer grown with a grading rate above Ͼ 27% per m. 25 In the XTEM micrograph of Fig. 6 ͑ b ͒ , a stacking fault with at least six associated dislocations is observed. The density of dislocations indicates that the stacking fault strongly interacts with the threading dislocations. The PVTEM micrographs of the 1200 nm films re- laxed on bulk Si, 300 nm SOI, and 40 nm SOI, are shown in Fig. 7. These specimens were prepared by thinning from the substrate side so that only the SiGe film is imaged by TEM. All three samples have stacking faults and threading dislocations present. The blocking of dislocation motion by stacking faults, prevalent on all the substrates, increases the TDD which ranges from 10 cm on bulk Si, to 5 10 cm on 330 nm SOI, and 2 ϫ 10 7 cm Ϫ 2 on 40 nm SOI. Many of the dislocations in the PVTEM micrographs lie on common glide planes. This arrangement is the product of a dislocation multiplication process during film relaxation, which has been previously observed in SiGe films. 20,26 The buried oxide of the SOI substrate does not appear to interfere with these multiplication processes. The AFM micrographs of all the in situ relaxed films exhibit the dislocation-derived crosshatch, as shown in Fig. 8. These films have a similar rms roughness ranging from 4.9 to 6.4 nm. Morphological differences are observed, however, on a finer scale. All of the surface trenches run the full width of the scan range for the SiGe films grown on the bulk Si substrates. This behavior is also present on the 10 000 nm SOI, which also has a high density of surface mounds. On the 330 nm and 40 nm Si–SOI substrates, only the deepest trenches run the length of the 10 m scans. Many shallow trenches run only a short distance. This is contrasted with the crosshatch structure of the films that relaxed during the ex situ annealing shown in Fig. 3. All of these ex situ annealed films exhibited very long uniform trenches, which were only present on the in situ relaxed bulk Si and 10 000 nm SOI samples. Similar in situ observations were made using low- energy electron microscopy, as described in Ref. 27. The dependence of the SiGe film relaxation on both the temperature and film thickness for Si and SOI substrates, whether induced by high-temperature annealing or during growth, is essentially identical. In all cases, a large density of misfit dislocations at the Si/SiGe interface are inferred from the increased x-ray diffraction peak width, the surface crosshatch, and, more directly, in the TEM micrographs. These results cannot be attributed to a compliant substrate mechanism, which would relax the film without the introduction of dislocations in the SiGe film. In addition, the amount of strain transfer to the Si–SOI layer, required by the compliant substrate mechanism, was not observed in the x-ray diffraction. Compliant substrate theory predicts a tensile strain in the Si–SOI layer. This tensile strain is only observed in this study after extensive dislocation-based film relaxation. In both the ex situ and in situ samples, an appreciable dislocation density is also apparent from the diffraction peak width. Compliant substrate strain sharing between the SiGe film and the Si–SOI layer was not present in any of the growth or annealing experiments. Film relaxation was only observed due to dislocation nucleation and propagation. Dislocations nucleate and propagate across the sample relaxing the film, with segments extending into the substrate. The dislocations in either the bulk Si or Si–SOI layers are the product of the strong dislocation interactions, which push dislocations into the underlying layers. These dislocation interactions have been observed and modeled for a variety of cases including dislocation multiplication processes during perpendicular and parallel crossings of dislocations. 20 The strain field interaction between dislocations provides the driving force to move the dislocations into the Si–SOI layer and has a far greater influence on dislocation motion than the dislocation image force attributed to the buried oxide. 15 On the SOI substrates, the dislocations penetrating into the Si–SOI layer can reach the buried amorphous Si–SOI/oxide interface. The substrate dislocations, pictured in the XTEM of the 40 nm Si–SOI layer in Fig. 4 ͑ b ͒ , all directly cross the layer and terminate at the oxide. In the 330 nm Si–SOI layer in Fig. 6 ͑ b ͒ , the majority of the dislocations continue to reach the buried oxide layer. Throughout these series of samples, no dislocation-based contrast was observed at the Si/SiO 2 interface in the TEM micrographs. Therefore, the growth temperature of 630 °C is sufficient for substantial dislocation core-spreading to occur during the film growth at this interface This change in the dislocation structure occurs with no observed changes in the onset of film strain relaxation or the relaxation rate. This result suggests that the rate limiting steps governing film relaxation are largely taking place in the film and not in the substrate or at the film–substrate ...
Context 4
... of dislocations that extend and terminate at the buried oxide. As with the annealed samples, the dislocations that reach the oxide do not show any contrast in the TEM micrograph indicating that the strain field associated with the core of the misfit dislocations at the Si/SiO 2 interface has been largely dissipated. Additionally, stacking faults are present in all the thick relaxed films grown on both bulk Si and SOI substrates. The presence of stacking faults has been previously reported when relaxing highly strained films, such as Ge on relaxed Si Ge , 24 and in a graded SiGe buffer layer grown with a grading rate above Ͼ 27% per m. 25 In the XTEM micrograph of Fig. 6 ͑ b ͒ , a stacking fault with at least six associated dislocations is observed. The density of dislocations indicates that the stacking fault strongly interacts with the threading dislocations. The PVTEM micrographs of the 1200 nm films re- laxed on bulk Si, 300 nm SOI, and 40 nm SOI, are shown in Fig. 7. These specimens were prepared by thinning from the substrate side so that only the SiGe film is imaged by TEM. All three samples have stacking faults and threading dislocations present. The blocking of dislocation motion by stacking faults, prevalent on all the substrates, increases the TDD which ranges from 10 cm on bulk Si, to 5 10 cm on 330 nm SOI, and 2 ϫ 10 7 cm Ϫ 2 on 40 nm SOI. Many of the dislocations in the PVTEM micrographs lie on common glide planes. This arrangement is the product of a dislocation multiplication process during film relaxation, which has been previously observed in SiGe films. 20,26 The buried oxide of the SOI substrate does not appear to interfere with these multiplication processes. The AFM micrographs of all the in situ relaxed films exhibit the dislocation-derived crosshatch, as shown in Fig. 8. These films have a similar rms roughness ranging from 4.9 to 6.4 nm. Morphological differences are observed, however, on a finer scale. All of the surface trenches run the full width of the scan range for the SiGe films grown on the bulk Si substrates. This behavior is also present on the 10 000 nm SOI, which also has a high density of surface mounds. On the 330 nm and 40 nm Si–SOI substrates, only the deepest trenches run the length of the 10 m scans. Many shallow trenches run only a short distance. This is contrasted with the crosshatch structure of the films that relaxed during the ex situ annealing shown in Fig. 3. All of these ex situ annealed films exhibited very long uniform trenches, which were only present on the in situ relaxed bulk Si and 10 000 nm SOI samples. Similar in situ observations were made using low- energy electron microscopy, as described in Ref. 27. The dependence of the SiGe film relaxation on both the temperature and film thickness for Si and SOI substrates, whether induced by high-temperature annealing or during growth, is essentially identical. In all cases, a large density of misfit dislocations at the Si/SiGe interface are inferred from the increased x-ray diffraction peak width, the surface crosshatch, and, more directly, in the TEM micrographs. These results cannot be attributed to a compliant substrate mechanism, which would relax the film without the introduction of dislocations in the SiGe film. In addition, the amount of strain transfer to the Si–SOI layer, required by the compliant substrate mechanism, was not observed in the x-ray diffraction. Compliant substrate theory predicts a tensile strain in the Si–SOI layer. This tensile strain is only observed in this study after extensive dislocation-based film relaxation. In both the ex situ and in situ samples, an appreciable dislocation density is also apparent from the diffraction peak width. Compliant substrate strain sharing between the SiGe film and the Si–SOI layer was not present in any of the growth or annealing experiments. Film relaxation was only observed due to dislocation nucleation and propagation. Dislocations nucleate and propagate across the sample relaxing the film, with segments extending into the substrate. The dislocations in either the bulk Si or Si–SOI layers are the product of the strong dislocation interactions, which push dislocations into the underlying layers. These dislocation interactions have been observed and modeled for a variety of cases including dislocation multiplication processes during perpendicular and parallel crossings of dislocations. 20 The strain field interaction between dislocations provides the driving force to move the dislocations into the Si–SOI layer and has a far greater influence on dislocation motion than the dislocation image force attributed to the buried oxide. 15 On the SOI substrates, the dislocations penetrating into the Si–SOI layer can reach the buried amorphous Si–SOI/oxide interface. The substrate dislocations, pictured in the XTEM of the 40 nm Si–SOI layer in Fig. 4 ͑ b ͒ , all directly cross the layer and terminate at the oxide. In the 330 nm Si–SOI layer in Fig. 6 ͑ b ͒ , the majority of the dislocations continue to reach the buried oxide layer. Throughout these series of samples, no dislocation-based contrast was observed at the Si/SiO 2 interface in the TEM micrographs. Therefore, the growth temperature of 630 °C is sufficient for substantial dislocation core-spreading to occur during the film growth at this interface This change in the dislocation structure occurs with no observed changes in the onset of film strain relaxation or the relaxation rate. This result suggests that the rate limiting steps governing film relaxation are largely taking place in the film and not in the substrate or at the film–substrate interface. Dislocation core-spreading has been observed, in situ , at interfaces between metals and amorphous layers. 28 –31 The lack of an extended crystal structure in the amorphous layer enables atomic motion, transforming the large and very local displacement of the dislocation core into smaller displacements spread along the interface through plastic deformation of the amorphous layer. 32,33 This accommodation of the dislocation by the oxide reduces the strain, and ultimately removes the strain associated with the dislocation core and the accompanying dislocation line tension. This reduction in the dislocation line tension drives dislocation motion through the Si–SOI layer. The amount of misfit dislocation length transferred from the SiGe/Si interface to the Si/SiO 2 interface can be experimentally determined from the change in strain of the layers. Each dislocation changes the lattice constant by an amount equal to the pure edge component of the Burger’s vector in the plane of the interface. The 60° dislocations found in the SiGe system have a Burger’s vector of the type a /2 ͓ 011 ͔ , and the relevant edge component is a / & ͓ 110 ͔ ϭ 0.193 nm. The average spacing of these dislocations can be determined from the change in in-plane strain of the layer: ...
Similar publications
Рассматривается проблема проектирования электроприводов с асинхронными двигателями, обеспечивающими оптимальные пусковые и энергетические характеристики в динамических режимах работы электропривода.Проведен анализ исследований и публикаций, где начато решение этой проблемы. В настоящее время в качестве электрооборудования для привода технологически...
Purpose
Past research has found meaningful differences between job and organisational engagement. However, research to date has mainly focused on job engagement, whereas the construct of organisational engagement has been largely neglected. The purpose of this study was to investigate the antecedents of organisational engagement and its influence o...
Citations
... SOI is one of the most studied compliant substrates that is readily available commercially. However, the buried oxide layer (BOX) restrains the flexibility of top silicon layer (Fig. 1a) and may introduce extra thermal stress during cooling process [21][22][23]. Besides, according to the critical thickness condition theory, if the substrate thickness is less than its critical thickness of relaxation, the substrate can allow the growth of infinitely thick epilayers without misfit dislocations [24][25][26]. To be an efficient compliant substrate, the top silicon layer should be thin enough that majority of the strain is confined inside itself and also less bonded with the supporting layer. ...
The material quality of III-nitrides is severely limited by the lack of cost-effective substrates with suitable lattice and thermal expansion coefficients. A suspended ultrathin silicon membrane substrate ([Formula: see text]16 nm), fabricated by an easy process on SOI substrates, is thus designed for nitride epitaxial growth, which can effectively release the strain in the epi-layers, and has demonstrated large-area (Al)GaN growth with a smooth surface and greatly reduced defect density. This research provides a promising CMOS-compatible method for growing cost-effectively high-quality III-nitrides that can be used for the development of high-performance devices.
... Characteristics of the GaN growth on SOI substrates have been reported by many research groups which further demonstrate the compliant feature of SOI substrate [9,10]. Yet, the potential of SOI substrate as a compliant substrate remains unrealized: (1) the top silicon layer is restrained by the buried oxide layer which intercepts the concept of complaint [11]; (2) the buried oxide layer can introduce extra thermal stress during cooling process. Besides, it has been proposed that the reduction of the substrate thickness can increase the critical thickness of the epitaxial layer, but the approach to fabricate a suspended or fully compliant silicon layer under 30nm is difficult to fabricate and is rarely studied. ...
A 30nm-thick suspended-ultrathin silicon membrane based on SOI (silicon-on-insulator) substrate for GAN epitaxial growth is fabricated and analysed. We demonstrate that thermal stress of GaN in the effective area on this novel substrate can be reduced 44% more than that in normal SOI substrate, which can reduce the dislocation and crack density in GaN. The absence of BOX (buried oxide) layer are confirmed as main factors to result in reduction of thermal stress in GaN layer. This work provides a promising approach to obtain high-quality GaN by utilizing mechanical structure.
... It is found in Fig. 2(b) that the 40-nmthick SiGe sample has higher compressive strain (e cmax ¼ 1.8%) than the 60-nm-thick one (e cmax ¼ 1.45%) in the Ge fraction of 100%. In general, strain relaxation of GOI films during Ge condensation zcan be initiated through crystal defect generation 15,16 and/or slippage between GOI and buried oxides, 27 caused by excessive strain energy. Thus, this SiGe thickness dependence of strain relaxation can be explained by the difference in the total strain energy in films. ...
... 22 This cooling time dependence can be associated with the difference in the thermal expansion coefficient between Ge and SiO 2 . 15,16,27 A rapid change in additional strain related to the difference in the thermal expansion coefficient under quick cooling might trigger dislocation generation and/ or slippage of GOI films. Further studies to identify the mechanism of strain relaxation would be needed to comprehensively understand the present SiGe thickness and the cooling time dependencies of strain relaxation during Ge condensation. ...
We study the impact of the SiGe thickness in starting substrates composed of Si/Si 0.25 Ge 0.75 /SOI(100) structures for the Ge condensation process on the resulting Ge-on-insulator (GOI) film properties. We evaluate the physical properties of the GOI films using AFM and Raman spectroscopy. It is found that 10-nm-thick GOI films with higher compressive strain (ϵ c = 1.75%) and more uniform spatial strain distribution are obtained for 40 nm-thick-Si 0.75 Ge 0.25 through a Ge condensation process with slow cooling than 60 nm-thick-SiGe. This suppression of strain relaxation is due to the lower total strain energy by the thinner SiGe layer. By using this GOI substrate, 10-nm-thick GOI p-channel metal-oxide-semiconductor field effect transistors (pMOSFETs) are demonstrated with the high performance of μ h = 467 cm ² V ⁻¹ s ⁻¹ and I on /I off > 7.2 × 10 ⁵ . The effective hole mobility of the 10 nm-thick GOI pMOSFET increases significantly with reducing measurement temperature from 298 K to 100 K, indicating the high contribution of phonon scattering to the mobility.
... SOI has been criticized since the Si layer is not free-standing. Rehder et al. reported that the SiO 2 layer underneath the Si which is also called the BOX intercepts the compliance of the Si layer [16]. According to their calculation, compared to a freestanding Si layer, SOI reduces the critical thickness achievable for the overlying layer such as GaAs. ...
Compliant substrates are one of the many promising lattice engineering approaches for hetero-epitaxy. There are many approaches to realize such compliant substrates. Silicon-on-insulator (SOI) wafer is a good option as it is readily available from industry. Previous studies on SOI using silicon thicker than 10 nm on insulator reported limited or no compliance. However, the effect of ultrathin silicon thickness (<10 nm) on the compliance of SOI wafer has not previously been studied. In this work, we investigate the quality of GaAs hetero-epitaxial films on SOI as a function of ultra-thin Si layer thickness. Results show that the quality of GaAs films deposited by Metal-Organic Chemical Vapour Deposition improves with decreasing Si thickness. When Si thickness is <4 nm, the resultant GaAs film is single-crystalline. When Si thickness is >4 nm, the GaAs film is multi-crystalline. These films have been characterized by X-ray diffraction measurement, atomic force microscopy and transmission electron microscopy. These findings show the potential for SOI wafers as compliant substrates at ultrathin Si thicknesses.
... The deposited dislocations activate STZs in amorphous layers, and further deformation increases the density of STZs, resulting in shear banding in amorphous layers. Since the metal-amorphous interfaces are less ordered, the incoming lattice dislocations are smeared along the interface, reducing the local strain/stress concentration [82,[90][91][92][93][94][95][96]. The maximum strength that metal-amorphous nanolaminates can reach is mainly determined by the yield strength of the single amorphous phase [97][98][99][100], although the nanolaminate composites exhibit plasticity. ...
The mechanical behavior of nanolaminates is dominated by interfaces that act as sources, barriers, and preferred sites for storage and dynamic recovery of glide dislocations. In this article, the deformation mechanisms of a variety of metal-based nanolaminates are reviewed with emphasis on unusual mechanical properties such as ultra-high flow strength without loss of plastic deformability.
... 11−14 The second problem comes from the Si/SiO 2 interface which is not fully slippery at temperatures of growth 15 and started to glide only at very high temperatures in conjunction with the formation of misfit dislocations. 16 Moreover, several studies focus on the formation and compositional evolution of SiGe layers epitaxially deposited on SOI during thermal oxidation. It is commonly reported that at a high oxidation temperature SiGe interdiffusion is dominant and allows the formation of SiGe layer with smooth but gradually decreasing Ge concentration profile. ...
In this work, we compare the morphological and structural features of SiGe membranes fabricated by three different processes: direct deposition of Si0.5Ge0.5 on Si(001) nominal substrate, direct deposition of Si0.5Ge0.5 on silicon on insulator, and deposition of SiGe with low Ge concentration on silicon on insulator followed by Ge enrichment by condensation. We show that the formation of fully strained Ge rich layers free of defects with flat surface is only possible by the two-steps epitaxy/condensation process. We demonstrate that the condensation based process enables the total inhibition of the morphological instability, together with the hindering of dislocations for critical thickness much larger than the ones commonly obtained by direct deposition. These behaviors could be explained by the injection of self-interstitials in the Ge rich layers during condensation. Such remarkable properties could be generalized to many other systems using similar condensation process.
... The only barrier that then prevents dislocation formation is kinetic: the energy required to move the atoms along the slip plane. The hypothesis of Kästner and Gösele [45] was later adopted by Rehder et al [46]. ...
... α is a factor, with value from b/(t Si + t SiGe ) to 1, that represents the energy reduction caused by core spreading [50]. For example, α is b/(t Si + t SiGe ) if we assume that core spreading totally eliminates the line energy, as previously assumed [45,46]. On the other hand, α is unity if no core spreading is present and SiO 2 is as stiff as the Si and SiGe layers (hence the maximum possible dislocation line energy). ...
... must be applied as well to the initially unstrained Si layer ( figure 7(b)). The more SiGe relaxes by this configuration, the more strain is injected into the Si template, and if the line energy is zero, the equilibrium strain state of the SiGe/Si structure is similar to that of free-standing strain-sharing nanomembranes [46]. The strain energy in the Si layer added to the system is Strain energy in Si template layer E Si strain = ...
Extremely thin single-crystal sheets have unique mechanical properties, which may influence the generation and behaviour of extended defects during heteroepitaxial growth. Using low-energy electron microscopy (LEEM) we investigate the earliest stages of inelastic strain relaxation in SiGe grown heteroepitaxially on Si-on-insulator (SOI) at a sensitivity not possible with other methods, employing a structure that forces dislocations, if they can form at all, to reside at the Si/oxide interface of SOI(0 0 1). LEEM confirms a lower dislocation line energy at the Si/amorphous oxide (SiO2) interface than at the crystalline SiGe/Si interface. The line energy is, however, nonzero, in contrast with earlier assumptions. The lower line energy makes the thermodynamic critical thickness for growth on SOI(0 0 1) lower than on bulk Si(0 0 1) for otherwise identical growth conditions. Nevertheless we can grow heteroepitaxial SiGe films on SOI(0 0 1) that are much thicker than even the thermodynamic critical thickness for growth on bulk Si(0 0 1), suggesting high kinetic barriers for dislocation formation or motion.
... 1 The full exploitation of a ultrathin SOI requires the control of its structural properties. In particular, the possibility of manipulating the deformation state, with the introduction of an in-plane strain , can increase the carrier mobility in the top silicon layer. 2 Moreover, SOI is one of the principal candidates for the realization of compliant substrates which could open novel possibilities for the integration of several lattice mismatched semiconductor heterostructures on Si. [3][4][5] The transfer of strain from an epilayer towards its substrate can follow two mechanisms. The first one simply relies on the elastic sharing of misfit strain between the epilayer and the substrate. ...
... An ultrathin SOI used as a substrate for the deposition of thicker SiGe epilayers is a particularly interesting case because the reduction of the silicon thickness allows to approach a regime where the strain induced by the two mechanisms is predicted to be comparable. 4 Within this framework, we investigated the relaxation mechanism of SiGe epilayers deposited on ultrathin-SOI substrates evaluating the possibility of using them as sacrificial stressors to plastically deform the underlying SOI layer, or to use the thin SOI layer as a compliant substrate to grow relaxed SiGe epilayers. ...
... Hence, in the case of a low dislocation density I 1 acts as a misfit dislocation getterer: While for the entire SiGe/Si bilayer the SiO 2 interface is far from being a free surface, each individual dislocation does feel the presence of SiO 2 as a soft interface and it is attracted to it. For thicker SOI layers, Rehder et al. 4 found that SiO 2 could be actually modeled as a completely free surface. Their model of a completely free surface overpredicts the residual strain that we measured in the SOI layers, and we attribute this discrepancy with the analytical model to the fact that considering the interface I 1 as a free surface overestimates the misfit dislocation getterer effect. ...
We studied the plastic deformation of an ultrathin silicon-on-insulator with epitaxial Si1−xGex by transmission electron microscopy, Raman spectroscopy, and finite-element method. We analyzed a top Si layer of 10 nm (testing also a 2 nm layer) with epitaxial Si0.64Ge0.36 stressors of 50 and 100 nm. SiGe plastically deforms the top Si layer, and this strain remains even when Si1−xGex is removed. For low dislocation densities, dislocations are gettered close to the Si/SiO2 interface, while the SiGe/Si interface is coherent. Beyond a threshold dislocation density, interactions between dislocations force additional dislocations to position at the Si1−xGex/Si interface.
In this paper we propose a strategy to achieve monolithic integration of III-Vs on Si for photonic integration through a simple process. By mimicking the SiO2/Si/SiO2 waveguide necessary to couple light from the gain medium on its top, we adopt a ~2 μm thick silicon dioxide mask for epitaxial lateral overgrowth (ELOG) of InP on Si. The ELOG InP layer as wells as the subsequently grown quantum wells (~1. 55 μm) have been analyzed by photoluminescence and transmission electron microscopy and found to have high optical quality and very good interface. The studies are strategically important for a monolithic platform that holds great potential in addressing the future need to have an integrated platform consisting of both III-Vs and Si on same chip.
