Metallic thin film on sige si substrates 3

The peak intensity/area of Ge 2p3/2 and Ge 3d remained constant from room temperature RT to 500oC, but decreased significantly when the annealing temperature was above 500oC.. Apart from

Chapter 3 Thermal Stability of Si0.8Ge0.2 Virtual Substrates

3.1 Introduction

In this chapter, the thermal stability of Si0.8Ge0.2 virtual substrate was studied using XPS, AFM and TEM Two temperature regimes have been identified; in region I (T<500oC), Ge segregation was observed while in region II (T≥500oC), preferential desorption of Ge occurred with formation of pits/holes on the surface A model was proposed to describe this observed behaviour in terms of Ge diffusion-desorption kinetics occurring at the surface region of the Si0.8Ge0.2 virtual substrate

3.2 In-situ composition change of Si0.8Ge0.2 substrates as a function of annealing temperature and time

A hydrogen terminated Si0.8Ge0.2 VS (prepared as described in Chapter 2) was annealed inside the XPS chamber from room temperature to 650oC At each targeted temperature, Ge 2p3/2, Ge 3d and Si 2p spectra from the sample were scanned Quantitative calculation of the Ge and Si’s atomic ratio (at%) were also tabulated based on the spectra for Ge 2p3/2 and Si 2p as well as Ge 3d and Si 2p At each targeted annealing temperature, the Ge% and Si% compositions were also monitored until there were no further changes For annealing temperature above 500oC, this process usually took ~2 hours After which, the substrate temperature was then raised to the next targeted value and the process repeated until 650oC Figures 3.1(a)-(c) shows the typical spectra of Ge 2p3/2, Ge 3d and Si 2p after annealing the H-Si0.8Ge0.2 substrate for 2 hours to various temperatures

Fig 3.1 XPS spectra of (a) Ge 2p3/2, (b) Ge 3d and (c) Si 2p after annealing

H-Si0.8Ge0.2 for 2 hours at various temperatures as indicated

It can be seen from Fig 3.1 that the binding energy (B.E.) and full width at half maximum (FWHM) of Ge 2p3/2, Ge 3d and Si 2p remained nearly the same through out these temperatures However, significant difference is noticed when we compare the peak intensity/area change of Ge and Si The peak intensity/area of Ge 2p3/2 and Ge 3d remained constant from room temperature (RT) to 500oC, but decreased significantly when the annealing temperature was above 500oC In contrast, there was

no discernable change in the peak intensity/area of Si 2p from RT to 650oC Apart from the variation of peak intensity/area with temperature since above 500oC, the behavior of Ge and Si spectra as a function of annealing time at each targeted temperature is also worth noting One such example is shown in Fig 3.2 for data collected at 530oC It can be seen from Fig.3.2 (a) and 3.2 (b) that while both the binding energy and FWHM of Ge 2p3/2 and Ge 3d remained unchanged, the peak intensity decreased as annealing time increases The fall in peak intensity occurred

only initially as it appeared to stabilize as annealing time increased Again, no obvious

change in binding energy, FWHM or peak intensity was noticed for Si 2p spectra

Fig 3.2 The spectra of (a) Ge 2p3/2, (b) Ge 3d and (c) Si 2p of H-Si0.8Ge0.2 as a

function of annealing time at 530oC

The decrease in peak intensity/area of Ge 2p3/2 and Ge 3d spectra with

temperature and time would indicate a reduction in the Ge concentration In order to

investigate this further, the relative Ge and Si concentration determined during each

annealing step is plotted as a function for both temperature and time as shown in Fig

3.3 As described previously, the sample was annealed at each temperature until there

was no further change in compositions before increasing the annealing temperature

further Therefore, data points obtained from the same temperature were grouped

together in a given column as shown in figure

Two distinctive trends are clearly observed and they can be divided into two

different temperature regions Region-I corresponds to the temperature range from

50oC to 450oC (i.e Fig 3.3(a) & (c)) while region-II corresponds to the temperature range from 500oC to 650oC (i.e Fig 3.3(b) & (d))

Fig 3.3 shows the Si and Ge atomic ratio (%) as a function of progressive annealing temperatures and time Plot (a) and (b) is based on analysis of Ge 2p3/2 and Si 2p spectra while (c) and (d) is based on analysis of the Ge 3d and Si 2p spectra respectively The data points grouped within the same column represent data obtained from the same temperature

In region-I (i.e Fig 3.3(a)), a noticeable increased in the Ge% (Ge 2p3/2: Si 2p) from a value of 22% to 26% was observed when the annealing temperature increases from 50oC to 450oC However, in the same temperature regime (i.e Fig 3.3(c)), the

5 10 15 20 25 70 75 80 85 90 95 100

0 5 10 15 20 25 70 75 80 85 90 95 100

Temperature (oC)

650 o C

630oC

610oC

590 o C

570 o C

550oC

530 o C

5 10 15 20 25 70 75 80 85 90 95 100

0 5 10 15 20 25 70 75 80 85 90 95 100

Ge% (Ge 3d : Si 2p) shows no discernable change and remained relatively constant at

a value of 23% throughout As Ge 2p3/2 photoelectrons have a lower kinetic value of 268.6eV compared to that of Ge3d photoelectrons (1458.6eV), Ge 2p3/2 photoelectrons have a shorter inelastic mean free path than that of Ge 3d photoelectrons, indicating that Ge 2p3/2 photoelectrons are more surface sensitive than Ge 3d photoelectrons The higher Ge% (26%) measured at 450oC compared to 22% at 50oC (based on Ge 2p3/2 :

Si 2p) and compared to 23% from Ge 3d : Si 2p suggests that the surface region has become more Ge rich as the annealing temperature increases This is not unexpected as preferential Ge segregation has been widely reported not only during growth of Si1-

xGex layer or Si/Si1-xGexsuperlattices but also for postgrowth annealing of strain Si

1-xGex layers160-170 This process is believed to be driven by Ge having a larger covalent radius and lower surface energy compared Si It therefore effectively reduces the strain energy by preferentially having it at surface rather than in the bulk160-164 Thus, as Ge surface segregates during annealing, an increase in Ge% near surface region occurs

In region-II, a noticeable decrease in the Ge% is observed in Fig 3.3(b) (or Fig 3.3(d)) as the annealing temperature is increased from 450oC to 500oC and beyond At

650oC, the steady state Ge composition as shown in Fig 3.3(b) (Fig 3.3 (d)) for example had fallen to ~2% (4%) compared to the starting value of 22% (23%) Two other interesting features are also worth noting At a given annealing temperature, the Ge:Si ratio initial decreases stabilized with time after ~ 2hours at each temperature This result can be related to the observation that at temperature above 500oC, the Ge peak intensity decreased initially and stabilized later while Si peak intensity did not change (Fig 3.2) This observation can be seen more clearly in Fig 3.4 whereby another set of experiment was performed at the same condition except this sample was kept for longer time at 650oC and 700oC

Fig 3.4 Composition change of H-Si0.8Ge0.2 (100) with annealing time at 650oC and

700oC

Contrary to formation of a Ge-rich layer in region-I (T<500oC), the experimental data suggests a depletion of Ge at Si0.8Ge0.2 surface in region-II (T≥500oC) through formation of either a silicon rich or Ge poor layer Such layers can be formed in three ways Firstly, Ge present on the surface may diffuse into the bulk of Si0.8Ge0.2, which leaves a Si-rich layer on the top surface and therefore Ge% decreases Secondly, annealing at high temperatures is known to cause Ge surface segregation160-170, which has been similarly observed in region-I in our work Ge surface segregation can lead to the formation of Ge-rich 3D islands on the surface, which was recently observed171-174but on strain Si1-xGex layers with the average island diameter at the bottom about 20-50nm172 If such Ge islands grow beyond the core level photoelectron attenuation length (3λ), Ge below the depth of 3λ effectively does not contribute to Ge XPS signal and results in a decrease in the Ge signals when compared to an initially flat surface This process also exposes more Si within the analysis area, which in turn magnifies the effect of a decrease Ge composition The last possibility is that Ge may desorb from surface, which leads to a direct loss of Ge from the surface region The chance for the

680 700 720 740 760 780 800 820 840 860 880 900 920 0

1 2 3 4 5 6 7 8 9 90 91 92 93 94 95 96 97 98 99

first possibility to materialize is low because Ge always tends to surface segregate in

Si1-xGex system160-170 Nevertheless, to probe the first possibility, depth distribution of

Ge in Si0.8Ge0.2 samples after annealing was investigated by sputtering and the results are presented in the next section

3.3 In-situ XPS Depth profiling of Si0.8Ge0.2 substrates before and after annealing

XPS depth profiling was performed on a Si0.8Ge0.2 sample after annealing it for 2hrs at 700oC The resulted plot is also compared to a sample not subjected to thermal treatment The sputtering time can be related to the depth from the surface (i.e., multiplying the sputtering rate) The longer the sputtering time, the deeper is the distance from surface A gradual increase of Ge concentration from surface to the bulk was clearly seen for the annealed Si0.8Ge0.2 while the concentration for the Si0.8Ge0.2

without annealing stayed relatively the same throughout There is no build-up of Ge in the bulk of annealed Si0.8Ge0.2, and the Ge concentration in the bulk was measured to

be nearly the same as that of the Si0.8Ge0.2 without annealing Therefore, the results indicate a net loss of the Ge from the material and the depletion of Ge at the surface is not attributed to the Ge inward diffusion from the surface region to the bulk of the substrate

0 50 100 150 200 250 300 350 400 450 500 0

5 10 15 20 75 80 85 90 95

100

Si% before annealing Ge% before annealing Si% after annealing Ge% after annealing

The implication of a deficiency of Ge on the surface is that there may be Ge surface segregation and formation of 3-D Ge-rich islands on the surface Alternatively,

it could also attribute to Ge desorption, which will cause a loss of materials from the surface and thus produce depression/holes on the surface Since both of them modify the surface but in different ways (islands versus holes), we would be able to identify the actual mechanism by examining the surface morphology of the Si0.8Ge0.2 after annealing through AFM

3.4 Ex-situ surface morphology of Si0.8Ge0.2 as a function of annealing temperature

In order to probe the surface morphology carefully, a series of H-Si0.8Ge0.2 (001) samples were prepared in a similar condition after annealing them for 3 hours at

200oC, 300oC, 450oC, 500oC, 620oC, 640oC, 660oC & 700oC, respectively To compare the surface morphology, one H-Si(001) and one H-Ge(001) samples were annealed up

to 640oC and 500oC, respectively (Because severe desorption from Ge substrate

happens when temperature is higher than 500oC, we did not anneal Ge substrate to above 500oC) AFM images of the surface morphology were immediately captured after the samples were cooled down and taken out of UHV chamber

For samples annealed in region-I (T<500oC), it can be seen from the 10µm×10µm AFM images that the surface morphology throughout annealing (<500oC) was similar to the hydrogen-terminated substrate at RT (Fig 3.6 (a)) The RMS roughness remained as low as ~2.5nm based on 10µm × 10µm AFM images throughout (Fig.3.8(a)) There was clear absence of the reported 3D islands with width

of 20-50nm172 However, when the surfaces were zoomed into the areas marked by the black square boxes in column (i), the surface changed from featureless to close-pack dome-structure based on the 500nm × 500nm high resolution AFM images in column (ii) Roughness analysis again revealed that RMS remained stable at 0.2nm below

400oC but increased to 0.4nm at 450oC (Fig.3.8(b)) Such an increase in RMS at microscopic scale is attributed to thermal roughening, which is more severe at higher temperatures Cross-section analysis of the domes from 100oC to 450oC revealed that the domes’ width and height remained as 11.0nm and 0.2nm throughout, respectively (Fig.3.9) Hence, there was no change in domes’ size and height during annealing in region-I Besides the similar morphology in region-I, the Ge : Si ratios also did not change significantly during region-I (Fig 3.3 (a) & (c))

In region-II (T>500oC, Fig 3.7), the surface morphology surprisingly did not show the formation of Ge-rich 3D islands with width of 20-50nm as anticipated Therefore, the decrease in Ge% in region-II as shown by XPS (Fig 3.3 (b) & (d)) was not due to Ge-rich 3D island formation on the surface, which will leave Ge desorption the only mechanism to account for the decrease of Ge in region-II

Fig 3.6 (i) 10µm×10µm and (ii) 500nm×500nm AFM images of the (a) as-received

Si0.8Ge0.2, and the Si0.8Ge0.2 after annealing at temperatures of (b) 100oC, (c) 200oC, (d)

300oC, (e) 400oC & (f) 450oC, respectively The profiles of the lines indicated in column (ii) were shown in column (iii)

Fig 3.7 (i) 10µm × 10µm and (ii) 500nm × 500nm AFM images of the Si0.8Ge0.2 after annealing 3hours at temperatures of (a) 500oC, (b) 620oC, (c) 640oC, (d) 660oC & (e)

700oC, respectively Images in column ii were zoomed from the area marked by square boxes in column (i) The profiles of the lines indicated in column (ii) were shown in column (iii)

ii) i)

Instead of island formation, visible square-like and rectangular-likes holes were clearly observed on the surface under 10µm × 10µm AFM images when annealed above 620oC Formation of holes implies a loss of materials from the substrates Since XPS sputtering results indicate a loss of Ge from the substrate in region-II, we attribute the formation of both square and rectangular holes to the loss to Ge, which can only occur through desorption We have thereafter defined the square and rectangular holes

as type-A and type-B holes, respectively The holes formation has degraded the surface morphology, leading to an increase in RMS from 4.2nm at 620oC to 9.4nm at 700oC based on 10µm × 10µm AFM images (Fig 3.8(a)) Furthermore, the dome size and height significantly increased compared in region-I, implying a more server thermal roughening results Within the same region-II, dome size still continued to increase from 11.1nm at 500oC to 21.7nm at 700oC, while the dome height remained roughly the same at 0.8nm (Fig 3.9) The observation suggests that the domes formed at lower temperature tend to agglomerate together at high temperature By doing so horizontal size is increased but vertical dimension is retained the same

Temperature (oC)

(a) (b) Fig 3.9 Evolution of (a) dome size and (b) dome height as a function of annealing temperature based on 500nm × 500nm AFM images

As for comparison, the surface of H-Si(001) and H-Ge(001) remained smooth and hole-free after annealing up to 640oC and 500oC, respectively (Fig 3.10(a) & (c)), while the RMS of both surfaces were less than 1nm However, severe desorption happened on Ge substrate when temperature was raised further, which led to a decrease in sample size and a complete disappearance of sample at high temperatures (>500oC) Hence, it is reasonable to relate the holes formation in Fig 3.7(b
e) to the

Ge component in Si0.8Ge0.2 substrate rather than the Si component

Fig 3.10 10µm × 10µm AFM images of (a) H-Si(001) and (c) H-Ge(001) after annealing to 640oC and 500oC, respectively Their high resolution 500nm × 500nm AFM images are shown in (b) and (d), respectively

In order to quantify the size of both the type A and type B holes, we have again resorted to cross-section line profile analysis of the holes A typical example of line-profiles across four holes labeled as i), ii), iii) and iv) at 620oC, 640oC, 660oC and

700oC in Fig 3.7(b)
3.7(e) is shown in Fig 3.11 It can be seen that all of their section display a “V” shape along one pair of holes’ edges Their cross-section along the other pair of edge also exhibited the “V” shape (not shown here), and hence the holes have the 3D structure of inverted pyramids Beside the shape, we can also quantify the holes’ width, depth and slope angles from the line-profiles, and the slope facet can be calculated thereafter

Fig 3.11 Line profiles of holes at (i) 620oC, (ii) 640oC, (iii) 660oC and iv) 700oC as indicated in Fig 3.7(b), (c), (d) and (e), respectively The widths for holes of i, ii, iii and iv are 176nm, 202nm, 223nm, 531nm, respectively The depths for holes of i, ii, iii, and iv are 44nm, 52nm, 101nm, 138nm, respectively

0 50 100 150 200 250 300 0

10 20 30 40 50

20 40 60 80 100 120 140 160

X[nm]

After measuring 153, 185, 138 and 103 holes of both type A and type B from four 10µm × 10µm AFM images for 620oC, 640oC, 660oC and 700oC, respectively, the stacked histogram of holes’ width, depth, and facet are showed in Fig 3.12 It can be seen that the holes generally became bigger and deeper at higher temperatures, while the slopes’ facet remained constant about {11n} (n: 3-5) Since the annealing time was same for all these samples, bigger size of the holes indicates a greater loss of material from substrates annealed at higher temperatures

It was evident that type-A (square shape) and type-B (rectangular shape) pits were the predominantly features of the surface morphology at 620oC, 640oC, 660oC and 700oC from column (i) of Fig 3.7(b) to 3.7(e) These macroscopic pits were at least 0.030µm2 in size and ~ 0.012µm in depth within the AFM limits Besides these two types of holes, by scanning areas of the surface (marked by square boxes) which appeared flat in 10µm × 10µm AFM images, occurrence of the third distinctive holes were also observed from 620oC to 700oC They were randomly distributed and were only visible when the scan size was reduced to 500nm × 500nm (see column (ii) of Fig 3.7(b) to 3.7(e)) or lower These holes are labeled as type-C and exist concurrently with both type-A and type-B pits Usually they were much smaller compared to either type-A or type-B pits, and were irregular in shape, although occasionally they exhibited a hint of having a square-like outline It should be noted that none of these three types of holes/pits were observed on samples at RT or annealed to temperature below 450oC (Fig 3.6)

(a) (b)

(c) Fig 3.12 Stacked histogram of the (a) width, (b) depth and (c) n value for facet of {11n} for type-A and type-B holes appeared at 620oC, 640oC, 660oC and 700oC, respectively

0.05 0.10 0.15 0.20

0.10 0.20 0.30 0.40

0.05 0.10 0.15 0.20

0.05 0.10 0.15 0.20 0.25 0.30