RLG Buffer Composition Profile

The thickness of each layer was confirmed through TEM observation and analysis. The overall thickness of each buffer, incorporating the Ge underlayer and the SiGe cap, was obtained in the (000) diffraction mode and defined from the Ge underlayer / Si(001) substrate interface to the surface of the sample. For this investigation the graded region is defined within the boundaries of the misfit dislocation network (Figure 4.6). This will give rise to a small experimental uncertainty, which could be quantified by performing a SIMS measurement on each sample, however, limited resources make this impractical for the current investigation.

A list of RLG samples and their graded layer thickness are given in Table 4.1 along with SiGe cap compositions and corresponding grading rates. Each sample composition was measured by HR-XRD RSMs, and an example of the dislocation network formed is shown in Figure 4.6. The grading rate (GRRLG

thickness layer

n compositio GR_RLG = ( −1)

) for each sample is

calculated by . An effective grading thickness of 10 nm

was assigned to the sample with no graded region (sample 3695) as this the calculated diffusion length for a drop of 1% Ge composition in the Ge underlayer at a growth temperature of 850°C (see Mehrer [117]).

Wafer Number

Reverse Grading Thickness (±2 %) (nm)

Top Layer SiGe Composition (±0.005) (-) Grading Rate (±1 %) (%µm-1 4054 ) 4594 0.785 4.69 4053 1885 0.786 11.4 4046 931 0.785 23.1 4050 736 0.787 29.0 4051 347 0.787 61.3 4128 193 0.759 124 4129 136 0.766 172 4052 117 0.797 174 4130 71 0.768 324 3695 10 (effective) 0.793 2069

Table 4.1: The RLG samples under investigation with the graded region thickness, top layer composition and corresponding grading rate. The grading layer thickness was measured from

TEM photos. Compositions were calculated from XRD RSMs.

Figure 4.6: An example of an RLG structure sample 4050. Lines shown are only guides for the eye.

High energy SIMS using a caesium (Cs+) ion beam with an energy of 14.5 keV was

performed on samples 4050, 4046 and 4053 with RLG thicknesses (tRLG

Figure 4.7

) of 1885 nm, 931 nm and 736 nm respectively. , Figure 4.8 and Figure 4.9 show TEM micrographs of these samples with their corresponding SIMS profiles. The SIMS profile shown in Figure 4.8 was measured more rapidly than the others

and shows random fluctuations of composition which is attributed to the measurement, rather than the growth technique. The linearity of the graded regions when analysed through SIMS gives coefficient of determination (r2) values of ~0.995

indicating highly linear grading profiles. A composition drop within the Ge underlayer (effect number 2) is seen in all three samples and is observed to increase with graded layer thickness.

Figure 4.7: A SIMS profile over a TEM micrograph to show a profile comparison of sample 4050 with a linear graded region of thickness 736nm with a grading rate of 29.0%µm-1.

Figure 4.8: Composition profile of sample 4046 with a linear graded region of thickness 931nm with a grading rate of 23.1%µm-1.

Figure 4.9: Composition profile of sample 4053 with a linear graded region of thickness 1885nm with a grading rate of 11.4%µm-1.

Wafer Number (Figure) Reverse Grading Thickness (±2 %) (nm) Ge Composition Drop (±0.5) (%) Distance of Composition Drop (±2 %) (nm) 4050 (Figure 4.7) 736 0.9 279 4046 (Figure 4.8) 931 1.7 329 4053 (Figure 4.9) 1885 1.0 424

Table 4.2: The measured Ge composition drop and distance over which the drop acts within the Ge underlayer labelled as composition drop 2 in Figure 4.7, Figure 4.8 and Figure 4.9, as

Table 4.2 shows the measured composition drop and the distances over which it occurs within the Ge underlayer for samples shown in Figure 4.7, Figure 4.8 and Figure 4.9. Within samples 4050 and 4046 the drop is linear. These drops in composition are speculated to be due to Si diffusion from the RLG region into the pure Ge underlayer [118].

Figure 4.9 shows a slightly different diffusion profile for sample 4053 where the drop is then followed by a rise in composition. This indicates that Si first diffuses into the Ge underlayer then a segregation process occurs. These diffusion and segregation effects are thought to be due to the high growth temperature of 850°C which is close to the melting point of pure Ge (See Table 2.1). When the reverse graded region is thicker the grading rates are lower, this means the surface of the buffer is kept at a higher Ge composition for a longer time during growth of the graded region. It is speculated that this is the main cause of enhanced diffusion, surface melting and segregation in the Ge underlayer.

The maximum observed composition drop due to diffusion or segregation of 1.7 % is equivalent to a misfit ε = -0.0008 when compared to the Ge underlayer. If the critical thickness for a constant composition layer with an equivalent misfit is found on Figure 2.14 the thickness required for glide of dislocations is 80nm and the thickness required for nucleation of dislocations is 32 µm. This shows that the composition drop into the Ge underlayer will induce dislocation glide and dislocations are

In all three samples for which composition profiles are presented two CVD chamber etches were performed within the growth. This leads to an inevitable change of the growth kinetics due to the control loop fluctuation (see Section 2.2.4). One of these etches was performed in-between the deposition of the Ge underlayer and the RLG layer and the other was performed in-between the RLG layer and SiGe cap. As the RLG/SiGe cap interface is grown at a higher temperature than the Ge underlayer a thicker chamber deposition occurs which will cause a larger reduction in growth temperature. When the deposition is removed, growth rate of the Si within the structure will increase as the SiH2Cl2

Figure 4.7

precursor is in a temperature limited growth mode. This results in a sudden decrease in Ge composition, which is rapidly quelled when the growth chamber is again coated with a few monolayers of deposition, resulting in an overall abrupt dip in composition. This is shown in the composition profiles of the uncalibrated structures (effect number 1 in and Figure 4.9) as dips of maximum 2 % over 110 nm. The associated misfit induces dislocation glide, but not nucleation of dislocations (Figure 2.14). The calibrated sample 4046 (Figure 4.8) was designed to have no mismatch at any interface. As the fluctuations in the profile shown are attributed to SIMS artifacts generated during the rapid measurement, it can be assumed that no composition drops occur during deposition.

The SIMS profiles in Figure 4.7, Figure 4.8 and Figure 4.9 were performed with a caesium (Cs+

4.2.3 ) ion beam with a high energy of 14.5 keV, which allowed rapid lower vertical resolution measurements. With this SIMS technique, only the thicker RLG buffers could be analysed. Due to mass-flow considerations (Section ) the thinner RLG buffers required profiling to ensure surface segregation was not a major effect and that a graded region was actually achieved. An oxygen (O+) ion beam of a

lower energy (500 eV) was used to profile these buffers, the result of which is shown in Figure 4.10.

Figure 4.10: A comparison of all SIMS profiles for the RLG buffer samples. The depth shown is from the SiGe cap / RLG interface. The effect of diffusion can be seen to increase as buffer

thickness is increased. Low energy SIMS measured by Dr. R. Morris. High energy SIMS measured by Dr. A. Simons.

Figure 4.10 shows the Ge profile taken from the SiGe cap / RLG interface into the Ge underlayer. Lower energy SIMS indicates the high linearity of the profiles and the extent of the diffusion at the RLG / Ge underlayer interface. The composition drop due to diffusion within the thinner graded layers is less than that of the thicker layers and is again found to be insignificant in terms of strain, only to induce dislocation glide. The anomalous drop in composition within sample 4129 (Figure 4.10) is speculated to be an effect of surface segregation where the high rate of increase of SiH2Cl2 precursor flow does not allow Si adatom incorporation.