Metallic thin film on sige si substrates 6

Chapter 6 High Temperature Reaction between Ni and Si001, Ge001 and Si0.8Ge0.2001 Virtual Substrates 6.1 Introduction In Chapter 6, we will use XPS to monitor different Ni silicides, ge

Chapter 6 High Temperature Reaction between Ni and Si(001),

Ge(001) and Si0.8Ge0.2(001) Virtual Substrates

6.1 Introduction

In Chapter 6, we will use XPS to monitor different Ni silicides, germanides and germanosilicides formation when ultra-thin Ni films (~2-6Å) were deposited on hydrogen-terminated Si(001), Ge(001) and Si0.8Ge0.2(001) substrates and annealed in-situ to different temperatures in vacuum Atomic force microscope (AFM) will also be used to study the surface morphology evolution along annealing and to correlate the observation with XPS results In particular, we will investigate the effect of annealing temperature on the Ni intensity, compound phases and their stability

6.2 The effects of temperature when annealing Ni/H-terminated Si(001), Ge(001) and Si0.8Ge0.2(001) surfaces

In this set of experiments, ~ 15% Ni was deposited on a series of H-terminated Si(001), Ge(001) and Si0.8Ge0.2(001) substrates at room temperature, which is equivalent to ~2-6 Å of Ni layer according to the growth rate measurement derived from Chapter 4 Each sample was then annealed directly to the target temperatures ranging from 100oC to 620oC for 1 hour The Ni 2p3/2,Si 2p and Ge 3d spectra were collected as a function of annealing time And the typical spectra after one hour annealing at various temperatures are shown in Fig 6.1, 6.2 & 6.3 for Si(001), Ge(001) and Si0.8Ge0.2(001) substrates, respectively

It can be seen from Fig 6.1 that when 6Å Ni was deposited on H-Si(001) surface at RT, binding energy (B.E.) of Ni 2p3/2 was at 853.8±0.1 eV, which corresponds to Ni in Ni2Si environment5 This is understandable as Ni initially forms a NiSi-like layer at the interface and further growth of Ni to 6Å result in a formation of Ni-rich-silicide-like layer From 100oC onwards, annealing the samples resulted in a shift in the binding energy to a higher value At 200oC, Ni 2p3/2 displayed a signature value of 854.0±0.1 eV, which was attributed to NiSi phase5,209 Between 400oC and

620oC, Ni 2p3/2 peak stayed around 854.5±0.1 eV, a value corresponding to Ni in the bulk NiSi2 crystal structure The high temperature annealing facilitates Ni diffusion into Si lattice and the subsequent reaction with Si, therefore the silicide phase has gradually developed from a Ni-rich Ni2Si phase to a Si-rich NiSi2 phase as the annealing temperature increased

Fig 6.1 Normalized (a) Ni 2p3/2 and (b) Si 2p spectra collected at normal photoelectron emission during annealing at different temperatures after depositing ~6 Å Ni on the H-terminated Si(001) surface at RT

x1.00 x1.00

Hence, NiSi2 is a more stable phase at high temperature, which started to form

at 400oC This formation temperature is much lower than the reported value of 500o

C-700oC138,142,144-145 We attribute this to the use of ultra-thin Ni film in our work, which facilitates the faster inter-diffusion of Ni & Si to form NiSi2

For the case of 2Å of Ni grown on H-Ge(001), the change of Ni 2p3/2 and Ge 3d B.E after annealed from 100oC to 500oC is shown in Fig 6.2 The B.E of Ni 2p3/2

was at 853.4±0.1 eV at RT Between 100oC and 500oC, Ni 2p3/2 peak remained at 853.5±0.1 eV throughout annealing The close B.E value suggests that only one phase was observed in our experiment from RT to 500oC Because NiGe is also the final phase in Ni-Ge system during annealing and no NiGe2 was observed even after annealing to 700oC (either thin film or bulk)9-10,77,149, the single phase from RT to

500oC was attributed to NiGe, which is the stable phase at high temperatures

x1.0 x1.0

For annealing 5Å Ni grown on H-Si0.8Ge0.2(001) from 100oC to 620oC, the change of Ni 2p3/2, Si 2p and Ge 3d B.E is shown in Fig 6.3 The B.E of Ni 2p3/2 was

at 853.6±0.1 eV upon deposition at RT When annealing from 100oC to 300oC, the B.E was observed to shift progressively to 853.8±0.1 eV Between the temperature region of 300oC and 500oC, the B.E further shifted from 853.8±0.1 eV to 854.5±0.1

eV Therefore, there is a significant shift in B.E from 300oC to 500oC Similar to the case of annealing Ni/H-Si(001) (Fig 6.1) where the shift of Ni 2p3/2 B.E during annealing was attributed to different Ni silicides phases, the shift of Ni 2p3/2 B.E in the current case can be similarly ascribed to different Ni germanosilicide formation

(a) (b)

(c) Fig 6.3 Normalized (a) Ni 2p3/2,(b) Si 2p and (c) Ge 3d spectra collected at normal photoelectron emission during annealing at different temperatures after depositing ~5

Å Ni on the H-terminated Si0.8Ge0.2(001) surface at RT

According to the description in Section 1.2.1.5 in Chapter 1, the sequence for

Ni germanosilicide formation when annealing thick Ni layer (>10nm) grown on hydrogen-terminated Si1-xGex(001) surface is as follows14-22,152,219,220:

x1.0 x1.0

x2.63 x1.22 x1.01 x1.00 x1.07

x1.10 x1.24

y

z z y y C

x x C

x x C

GeSiGeSiNi

GeSiGeNiSiGe

NiSiGe

SiNi

1 2 1

1 1

650 1

500 350

~ 1

2 325

)(

)(

where y<x<z Unfortunately, there are no reported B.E values for each specific Ni germanosilicide phases Therefore, we are going to relate the detected Ni 2p3/2 B.E at different temperatures to different Ni germanosilicide phases according to their reported formation temperatures However, it should be noted that the formation temperatures of different Ni germanosilicide phases in our work can be much lower than the reported value It is because the ultra thin nickel film (~5Å) in our work can lower the phase transformation temperatures

Starting from the last formed Ni germanosilicide phase, NiSi1-yGey or Ni(Si

1-yGey)2 (y<0.2), this phase has the Ni 2p3/2 B.E value of 854.5±0.1eV in between 500oC and 620oC This B.E value is very close to that of NiSi2 (854.5±0.1), hence we determine the last formed phase to be Ni(Si1-yGey)2 The phase prior to the Ni(Si1-

yGey)2 formation isNiSi0.8Ge0.2, which has the Ni 2p3/2 value of 853.8±0.1 eV in between middle temperature range of 100oC and 300oC Before the formation of NiSi0.8Ge0.2,the Ni-rich Ni2(Si0.8Ge0.2) phase forms below 100oC, corresponding to a

Ni 2p3/2 B.E value of 853.6±0.1 eV

The Ni 2p3/2 B.E shift as a function of annealing temperature on Si(001), Ge(001) and Si0.8Ge0.2(001) substrates are summarized and displayed in Fig 6.4 It is difficult to tell whether Ni preferred to bond with Si or Ge right after deposition at RT However, after annealed at high temperature (≥500oC), it can be seen that the B.E of

Ni 2p3/2 in Ni(Si1-yGey)2 was very close to that of NiSi2 but away from that of NiGe The implication is that Ni prefers to bond with Si rather than Ge in the Ni(Si1-yGey)2

islands at high temperatures15,16,22 Through this study, it is demonstrated that we are able to identify different Ni silicide & germanosilicide phases formation using XPS even when the ultra-thin Ni layers (~5-6Å) were annealed on Si & Si0.8Ge0.2 surfaces

It should be noted that the full width at half maximum (FWHM) of Ni 2p3/2

decreased from ~1.7 eV to ~1.2 eV when temperature increased from 300oC to 620oC for the 5Å Ni grown on H-Si and H-Si0.8Ge0.2 substrates, while it decreased from ~1.1

eV to ~0.7 eV when temperature increased from 100oC to 500oC for the 2Å Ni grown

on H-Ge The decrease in the FWHM coupled with the change of the Ni 2p3/2 peak shape from an asymmetric (metallic-like phase) to a symmetric shape (silicide, germanide and germanosilicide phase), indicates an improvement in the silicide, germanide and germanosilicide crystallinity Also, it is worth noting that during deposition of ~2 Å of Ni, the binding energy of Si 2p and Ge 3d remained at 99.2±0.1 and 29.0±0.1 eV This is the same as the binding energy derived from analysis of bulk NiSi2 and NiGe samples, and also similar to clean Si and Ge substrates (99.2± 0.1 eV)2, within our experimental error The Si 2p and Ge 3d spectra did not display any noticeable energy shift during annealing Given that only 2-6 Å of Ni layers are deposited, the Ni2Si, NiGe and NiSi0.8Ge0.2 layers were thin enough such that the Si 2p and Ge 3d signal from the pure Si, Ge and Si0.8Ge0.2 substrates were also detected

(a)

(b) Fig 6.4 Shift of Ni 2p3/2 binding energy when annealing (a) 15%Ni/H-Si(001) & 15%Ni/H-Ge(001) as well as (b) 15%Ni/H-Si0.8Ge0.2(001) from RT to 620oC

It is interesting to note from Fig 6.3 that while the crystallinity improved, the

Ge intensity increased slightly from RT to 400oC, and thereafter decreased from 400oC

to 620oC As explained in Chapter 3, Ge segregated to the surface region from RT to

450oC, which caused an increase in Ge intensity When the temperature was further raised to above 500oC, Ge started to desorb from the surface, leading to a loss of Ge

0 50 100 150 200 250 300 350 400 450 500 550 600 650 853.3

NiGeNiGe

Ni/H-Ge(001)

NiSi2NiSi

from surface and a decrease in Ge intensity At each temperature, Ge intensity decreased first and stabilized with time eventually

Besides the change in Ge intensity, it can be seen from Fig 6.1, 6.2 and 6.3 that the intensity of the Ni peak decreased while the temperature increased The implication

is that there is either an overall decrease in the Ni concentration or an increase in the substrates’ concentration when temperature increases In order to evaluate the relative change of Ni concentration at different temperature across several samples, we normalize the Ni% during annealing to the initial Ni% upon deposition (~15%) for each sample, and plot the normalized Ni% as a function of annealing temperature and time across three substrates in Fig 6.5

For annealing 15%Ni/H-Si(001) in Fig 6.5(a), this ratio was fairly constant and did not change significantly with annealing time in the temperature range between RT and 200oC However, a significant decrease in ratio with time was observed when the sample was annealed to higher temperatures (≥300oC) For 15%Ni/H-Ge(001), the normalized Ni% started to drop with annealing time even from 100oC onwards (Fig 6.5(b)) When the 15%Ni/H-Si0.8Ge0.2 was annealed below 100oC, this ratio was fairly constant and did not change significantly with annealing time (Fig 6.5(c)) However, a significant decrease in ratio with time was observed when the sample was annealed to temperatures to 200oC and above

(a)

(b)

(c) Fig 6.5 Normalized Ni% ratio versus annealing time at various temperatures after

~15% Ni deposited on H-terminated (a) Si(001), (b) Ge(001) and (c) Si0.8Ge0.2

substrates

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 0.0

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

0.2 0.4 0.6 0.8 1.0

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

500400oC

o C

In all these three cases as described, two regimes can clearly be found as separated by the dash vertical line for each of these three substrates Regime I is characterized by an initial rapid decrease at the beginning of the annealing This decrease in value however reaches a steady state in about ~18mins, ~35mins and

~20mins for Si, Ge and Si0.8Ge0.2 substrates, respectively The decrease in value is also more significant at higher temperatures Regime II on the other hand is characterized

by a characteristic steady-state value which only decreased when the annealing temperature was further increased

The values in the steady state regimes are re-plotted as shown in Fig 6.6 For Ni/H-Si0.8Ge0.2 system, similarly the values of Ge and Si atomic ratio in the steady state regimes were also plotted for comparison Clearly the normalized Ni% decreased when temperature decreased, while normalized Ge% increased below 500oC due to segregation and then decreased above 500oC due to desorption

(a)

(b)

(c) Fig 6.6 Normalized Ni% atomic ratio in the steady state regime when annealing 15%

Ni grown on (a) H-Si(001), (b) H-Ge(001) and (c) H-Si0.8Ge0.2(001) substrates to different temperatures

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

0.2 0.4 0.6 0.8 1.0 1.2 1.4

The decreases in the steady value ratio on all three substrates were clearly more significant when the annealing temperature was higher In Chapter 4, we concluded that thin amorphous “NiSi-like”, “NiGe-like” and “NiSi0.8Ge0.2-like” layers are formed

at the initial growth front between Ni and Si, Ge and Si0.8Ge0.2 substrates on both clean and H-terminated Si, Ge and Si0.8Ge0.2 surfaces More importantly, the growth of Ni on

Si, Ge and Si0.8Ge0.2 substrates follows a pseudo-layer-by-layer growth mode The films that are formed at room temperature are therefore continuous and for simplicity

we assume it to be “NiSi-like”, “NiGe-like” and “NiSi0.8Ge0.2-like” films on Si, Ge and

Si0.8Ge0.2 substrates, respectively, which are drawn schematically in Fig 6.7

Fig 6.7 A schematic drawing of the “NiSi-like”, “NiGe-like” and “NiSi0.8Ge0.2-like” layers upon Ni thin films deposition on Si, Ge and Si0.8Ge0.2 substrates at RT, respectively

The decrease in the normalized Ni% ratio with time and temperature would thus imply an overall decrease in the detected nickel concentration or an increase in the detected Si, Ge and Si0.8Ge0.2 concentration within the XPS analysis depth of 3λ In order to explain the occurrence of regime-I and more significantly regime-II, we consider the following possible scenarios

Si, Ge or Si0.8Ge0.2-substrates

NiSi-like, NiGe-like, or NiSi0.8Ge0.2-like layers 3λ

(i) Ni desorption from the surface

When the temperature increases, it is possible that Ni may desorb from the surface Desorption is an activated process, hence more will be loss if heated to higher temperatures While this may explain why the initial decrease in Ni concentration is faster at higher temperature (regime-I), it cannot account for the steady state value at a given temperature after annealing more than 30 minutes (regime-II) In the event of an isothermal Ni desorption, the concentration of Ni should keep decreasing and approach zero as time increases It would not remain constant and only change when temperature is changed

(ii) Simple Ni inward diffusion

Ni is found to be the dominant diffusing species during the formation of Ni2Si, NiSi, NiSi2 phases135,139 and NiGe phase149 as well as NiSi0.8Ge0.2 phase16, although the out-diffusion of Ge in Ni/H-Si0.8Ge0.2 system dominates at high temperatures17 Upon annealing, Ni atoms will therefore diffuse from the initial Ni/NiSi, Ni/NiGe, Ni/

Si0.8Ge0.2 interfaces into the Si, Ge and Si0.8Ge0.2 substrates Diffusion processes are strongly temperature dependent and it is also an activated process Higher the temperature faster is the diffusion rate and hence more Ni is lost from the surface into the bulk While this can account for the occurrence of regime-I, the observation of a steady state value which does not change with time (i.e regime-II) at a given temperature is harder to explain Since only a fixed amount of Ni is deposited onto the

Si surface, Ni inward diffusion will only result in a continuous decrease in Ni surface concentration, which will eventually approach zero instead of staying constant at each temperature This is again contrary to experimental observations

(iii) Reaction of Ni and Si, Ge and Si0.8Ge0.2 to form corresponding silicides, germanide and germanosilicide

Simple Ni diffusion can not explain the observation of regime-II, and it is thus possible that there are other coexisting mechanisms We have not, for example, considered the reaction between Ni and the underlying substrates, which is known to strongly depend on temperature and several possible phases have been identified below for three substrates

For thick Ni layer (>1000 nm) grown on Si surface, the typical silicide phases formation sequence is as following5,135:

Through monitoring the B.E shift of Ni 2p3/2 during annealing (Fig 6.4(a)), it

is clear that different Ni silicides are formed at different stage of annealing in our work For example, after Ni2Si are formed at RT, NiSi was formed upon annealing at

200oC, while NiSi2 appeared between 400oC and 620oC Each phase is stable at each corresponding temperature range until the temperature was increased into another range As the silicide phases evolve from Ni-rich Ni2Si to a Si-rich NiSi2, the Ni concentration on the surface decreased accordingly but stabilized with time until there

is a change in temperature range With Ni simple diffusion to account for the occurrence of regime-I, this phase transition would be able to account for the stabilized Ni% in regime-II In Fig 6.8, we schematically summarize our initial speculation for the Ni/Si interface evolution when annealing at different temperatures

Fig 6.8 Schematic representation of the Ni/Si interface evolution during annealing (a) Initial Ni2Si-like layer at temperature between 25oC-100oC, (b) progressive diffusion

of Ni and Si to form the NiSi phases at temperature between 100oC and 300oC finally (c) NiSi2 formation at temperature between 400oC-600oC

It is also interesting to compare the structure of NiSi with that of NiSi2 phase NiSi has a orthorhombic MnP structure (Fig.6.9(a))2,3 The NiSi2 phase, which is formed at higher temperature, adopts a cubic CaF2 structure (Fig.6.9(b))2,3 It is

Si-Substrate 3λ

Ni

Si

NiSi thin films

increase temperature

increase temperature

Si-Substrate

(c)

composed of eight body-centered cubes adjacent to one another with all the corner sites occupied by silicon atoms but only half of the body-center positions are occupied

by Ni atoms, providing an effective vacancy concentration of 50%4 The inherent structure of NiSi2 has a more open structure compared to NiSi so that formation of this kind of the body-centered lattice provides Ni atoms with highly efficient diffusion channel Hence excess Ni on the surface can diffuse readily to react with Si through the formed NiSi2 phase structure As a result, the initial decrease in the Ni to Si ratio is more rapid at higher annealing temperature due to the formation of more NiSi2

Fig 6.9 (a) MnP-type NiSi structure; (b) CaF2-type NiSi2 structure

For Ni/H-Si0.8Ge0.2 system, while Ni simple inward-diffusion can account for the rapid decrease of Ni% in regime-I, the steady-state value in regime-II can also be attributed similarly to a reaction between Ni and Si0.8Ge0.2 substrate and the subsequent

Ni germanosilicide formation in Ni/H-Si0.8Ge0.2 system It is know that the reaction between Ni and Si0.8Ge0.2 strongly depends on temperature For thick Ni layer (>100 nm) grown on Si0.8Ge0.2 surface, the typical germanosilicide phases formation sequence at difference temperatures is as follows152,219,220:

z z y

y C

x x C

x x C

GeSiGeSiNiGe

NiSiGe

SiNi

Ni< →o − − o → − > →650o 1− 2+ 1−

1 500

350

~ 1

2 325

)(

)(

The stability of each phase is temperature dependent Ni2Si1-xGex phase is more Ni rich than NiSi1-xGex, which in turn is more Ni rich than Ni(Si1-yGey)2 Thus, it is possible that the different steady-state Ni% values at different temperatures are attributed to the presence/coexistence of different Ni germanosilicide phases The reaction of Ni/Si1-

xGex to form Ni2Si1-xGex, NiSi1-xGex or Ni(Si1-yGey)2 is an activated process and would require Ni atoms to diffuse through initial NiSi1-xGex layers to form more of these phases The decrease in Ni/Si1-xGex ratio at higher temperatures could therefore be due

to diffusion and phase transformation to a less Ni rich or more Si1-xGex rich germanosilicide phase

Through monitoring the B.E shift of Ni 2p3/2 during annealing (Fig 6.4(b)), it

is clear that different Ni germanosilicides are formed at different stage of annealing in our work For example, after Ni2Si0.8Ge0.2 was formed at RT upon deposition, NiSi0.8Ge0.2 was formed upon annealing between 100 and 300oC, while Ni(Si1-yGey)2

(y<0.2) appeared between 500oC and 620oC Each phase is stable at each corresponding temperature range until the temperature was increased into another range As the germanosilicide phases evolve from Ni-rich Ni2Si0.8Ge0.2 to a Si1-yGey-rich Ni(Si1-yGey)2, the Ni concentration on the surface decreased accordingly but stabilized with time until there is a change in temperature range With Ni simple diffusion to account for the occurrence of regime-I, this phase transition would be able

to account for the stabilized Ni% in regime-II

As for Ni/H-Ge system, while Ni simple inward-diffusion also accounts for the rapid decrease of Ni% in regime-I, the steady-state value in regime-II can be partly attributed to Ni germanide formation in Ni/H-Ge system It is because only NiGe phase exists through 100oC to 500oC Therefore, the decrease in steady-state value as

temperature increased can not explained by the phase transition between different Ni germanides The only plausible scenario pertaining to single NiGe phase but with a reduction in Ni% when annealing temperature increased is due to a change in surface morphology from 2D flat surface into 3D structure This is because more Ge substrate area would be exposed as a result of the surface NiGe layer is breaking apart In addition, when the thickness of the 3D structure is bigger than 3λ of Ni 2p3/2, the Ni atomsbelow 3λ deep will not be detected and hence cannot attribute to the overall Ni intensity These two factors together bring down the steady-state Ni% value in regime-

II when temperature increased

In order to verify this postulation for the behavior in regime-II when annealing 2Å Ni/H-Ge(001) samples, an examination of the surface morphology before and after annealing is clearly necessary This was similarly done after annealing 5-6Å Ni on H-Si(001) and H-Si0.8Ge0.2(001) substrates since phase transition from NiSi into NiSi2

and from NiSi0.8Ge0.2 into Ni(Si1-yGey)2 may also result in a change in surface morphology

Figure 6.10 shows the surface morphology of the film prepared at room temperature and after annealing to 620oC (it was 500oC for Ge surfaces) While the initial surfaces were flat and featureless, we found that the surface morphology after annealing was not The Si(001), Ge(001) and Si0.8Ge0.2(001) surfaces were covered by rectangular islands after annealing at high temperatures Thus, annealing the initial flat NiSi-like, NiGe-like and NiSi0.8Ge0.2-like films destroyed the smooth and continuous morphology while instead resulted in formation of 3D islands

(a) (b) (c)

Fig 6.10 (a) to (c) shows 5µm × 5µm AFM images of the surface at room temperature after 2-6Å Ni was deposited on H-terminated Si(001), Ge(001) and Si0.8Ge0.2(001) surfaces, respectively (d), (e) & (f) show 5µm × 5µm AFM images after annealing the

Ni films grown on Si, Ge and Si0.8Ge0.2 surfaces for 1 hour at 620oC, 500oC & 620oC, respectively

In Fig 6.11, we schematically present the surface morphology evolution when annealing 2-6Å Ni grown at RT on Si(001), Ge(001) and Si0.8Ge0.2(001) surface to

620oC, 500oC and 620oC, respectively The formation of these Ni nanostructures at high temperatures will be discussed in more detail in the next section

Fig 6.11 Schematic representation of the processes occurring when annealing ~2-6Å

Ni grown on (a) Si(001), (b) Ge(001) and (c) Si0.8Ge0.2(001) surfaces

6.3 Surface morphology of 15%Ni/H-terminated Si(001), Ge(001) and

Si0.8Ge0.2(001) surfaces after annealing

6.3.1 Surface morphology of 15%Ni/H-terminated Si(001) surfaces after annealing

Figure 6.12 shows the surface morphology after annealing the Si(001) samples to 300oC, 400oC, 500oC, and 620oC For annealing temperatures not exceeded 300oC, the surface morphology remained the same as that at room temperature, which is decorated with small close-packed round islands with an average diameter of 12.5nm and an average height of 0.2nm The RMS remained as low as 0.1nm Since XPS results suggested a formation of NiSi phase below 300oC, this smooth morphology should correspond to the NiSi layer

15%Ni/H-NiSi

Si(001) NiSi2

(a)

Si(001)

620oCRT

NiGe

Ge(001) NiGe

(b)

Ge(001)

500oCRT

Fig 6.12 Columns (a) 5µm × 5µm and (b) 1µm × 1µm AFM images after annealing

~15%Ni/H-terminated Si(001) surfaces to 300oC, 400oC, 500oC and 620oC, respectively The surface morphologies after RT deposition and annealing below

300oC are similar to that shown for surfaces annealed to 300oC Images in column (b) zoom into the areas marked by while square boxes in corresponding images in column (a) Column (c) shows the line profiles of the islands indicated in column (b)

0.5 1 1.5 2 2.5

1 1.5 2 2.5 3

X[nm]

IV [110]

[110]

0 2 4 6 8 10

X[nm]

III

However, after annealing to 400oC and beyond, rectangular elongated islands are now occupying the surfaces At 400oC, shallow and flat rectangular area started to appear on the surface with height less than 1nm When the temperature further increased to 500oC, the average width and thickness of the island were 120.8nm and 0.3nm, respectively Such small heights suggested a 2D structure formation at 400oC &

500oC When the temperature increased further up to 620oC, the islands became much taller and bigger The average width and thickness were 199.6nm and 24.0nm, respectively, and the facets of the islands were mainly {11n} with n being 3, 4 and 5 The edges of the islands were all along [110] and [110] directions throughout annealing, but the 2D islands have been transformed into 3D islands as temperature increased from 400oC to 620oC XPS results imply the existence of NiSi2 on the surface between 400oC and 620oC, which was formed through nucleation and is known

to cause agglomeration and degrade the morphology The AFM images indeed show that the continuous NiSi film was broken into rectangular 2D islands at 400oC, while the 2D islands continued to grow into 3D islands at 620oC Such a surface morphology evolution obtained from AFM matches well with the transformation from NiSi into NiSi2 when temperature increased from 200oC to 620oC as observed by XPS

The observed formation of NiSi2 islands will expose more Si atoms from the substrates within the XPS analysis depth compared to a flat Ni thin film In this case, apart from the decrease in Ni signal due to inward diffusion, the overall Si 2p signal would also increase These two effects will clearly decrease the ratio of Ni% and increase that of Si% Visually the density of islands decreased as temperature increases, which would lead to a significant decrease in the Ni% Although not reported here, we have also observed that the density of islands at a given temperature