Metallic thin film on sige si substrates 4

Chapter 4 Ni Growth on Si, Ge & Si0.8Ge0.2 Substrates at Room Temperature: Interfacial Reaction and Growth Mode 4.1 Introduction In this chapter, we will study the growth dynamics of Ni

Chapter 4 Ni Growth on Si, Ge & Si0.8Ge0.2 Substrates at Room Temperature: Interfacial Reaction and Growth Mode

4.1 Introduction

In this chapter, we will study the growth dynamics of Ni on Si(001), Ge(001)

Si0.8Ge0.2(001) surfaces using in-situ XPS and ex-situ AFM In particular, we will compare the growth mode, interfacial reaction and surface morphology of Ni thin films grown on hydrogen-terminated surfaces with those on clean surfaces

4.2 XPS and AFM results of Ni growth on Si substrates

4.2.1 Ni deposition on H-terminated and clean Si surfaces at RT

Figure 4.1(a) and (c) show the typical Ni 2p3/2 spectra versus Ni coverage on both hydrogen terminated and clean Si(001) surface at RT using Mg Kα X-ray source, respectively At 0.1% Ni deposition on H-Si(001) surfaces, the Ni 2p3/2 peak was observed at a binding energy (B.E.) of 854.0 ± 0.1eV (Fig 4.1(a)), which is about 1.2eV shift away from that of pure metallic Ni 2p3/2 peak (852.8 ± 0.1 eV) 5 This value

is very close to that of NiSi (854.0 ± 0.1eV) With increasing Ni coverage, the binding energy of Ni 2p3/2 shifted gradually towards the lower values Eventually, it reached and stayed at 852.8 ± 0.1 eV, a value as expected for metallic Ni thin film The B.E of

Si 2p from the substrate as shown in Fig 4.1(b) and 4.1(d) stayed consistently around

99.2 ± 0.1 eV throughout and did not show any change in shape or position Identical XPS spectra evolution can be also observed for Ni deposition on both clean and hydrogen terminated Si(111) surfaces (not shown here)

Fig 4.1 Ni 2p3/2 and Si 2p spectra obtained from Ni deposited on (a-b) hydrogen terminated and (c-d) clean Si(001) surface at RT with increasing Ni coverage In comparison, Ni 2p3/2 spectra from bulk Ni and Si 2p spectra from pure Si(001) substrate were included

At low metal coverages, a shift in BE of overlayers to higher values has previously been observed in the growth of Ir, Pd and Au on carbon180, Al and Cu on

x1.0 0.1%Ni

x102

x19 x6.3 x3.6 x2.2 x1.4 x1.1

x1.0 Pure Si

89%Ni 70%Ni

53%Ni 40%Ni

29%Ni 13%Ni 5%Ni

16% Ni 9% Ni Pure Si

27% Ni 35% Ni 45% Ni 54% Ni 61% Ni

x1.0

Ni 2p3/2

x200 0.1%Ni

x1.0 x1.9 x1.8 x1.7 x1.8 x2.1 x4.0

graphite as well as TiO2 surfaces181-183 In these systems, the shifts in BE has been attributed to the result of incomplete screening of the final electronic (hole) state This occurs if small clusters are formed on the surfaces (termed the cluster effect)184-185 In the present work, the shift of Ni 2p3/2 to a higher binding energy on both hydrogen-terminated Si(001) and Si(111) at low Ni coverage is unlikely to be attributed to cluster effects This is because the surface morphology after deposition of 10% or more Ni of H-Si(001) and H-Si(111) (Fig 4.2(a)&(b)) are relatively smooth and the RMS remained as ~2.1±0.5 Å throughout out the experiment Moreover, no clusters were observed to form on the Si surface

Fig 4.2 represent 1 µm × 1 µm AFM images after 10% Ni deposited on H-terminated (a) Si(001) and (b) Si(111) surfaces at RT, respectively

Experimentally, we observed that the B.E of Ni 2p3/2 (854.0 ± 0.1eV) even at

Ni coverage of 0.1% is very close to the value associated with bulk NiSi (854.0 ± 0.1 eV) Therefore, we attribute this to the formation of a NiSi phase and this occurs on both clean and hydrogen terminated Si(001) and Si(111) surfaces The formation of Ni-Si bond on clean Si surfaces is not unexpected since there are unsatisfied Si dangling bonds present at the surface186-187 However, it is more difficult to rationalise the observation of a Ni/H-Si reaction and formation of a Ni-Si bond on H-terminated

Si surfaces since all the dangling bonds are now passivated In this case, there are no available reactive Si dangling bonds to facilitate direct bond formation between Ni and surface Si atoms For this reaction to occur it would imply a reaction between Ni and the H-Si bond or with the Si back bonding at the H-Si-Si surface

The interstitial model188 proposed for growth on clean Si surfaces may explain how Ni silicide reaction can take place at RT on both clean and H-terminated surfaces Results from experiments by employing strain-sensitive X-ray diffraction59 and first principle calculations189-190 have suggested that at the initial deposition stage, Ni atoms can diffuse into Si crystal by either preferably occupying the interstitial sites of the Si host or diffusing to the off-center bridge site (B site) in second layer between the dimer rows189-190 The consequence of Ni occupying interstitial sites or B sites in Si is the weakening of the Si-Si covalent bond Electrons in the neighboring Si-Si covalent bonds will no longer remain in their localized states and will have to share between the interstitial and Si atoms, leading to substantial Si-Ni p-d hybridization191-192 This can result in the formation of a Ni-Si covalent bond192 at the growth front of Ni/Si or Ni/H-

Si surfaces Given the limited mobility of adatoms at room temperature, the NiSi phase formed at RT would likely be different from the stable crystalline orthorhombic MnP structure191, and it has been widely agreed in literature that thick Ni silicide layers grown at RT are amorphous46,48,50 We therefore described this phase as a “NiSi-like” film The observation of spontaneous silicide reaction also contradicts the assumption that there is a threshold thickness for silicide reaction to be feasible193 Our results showed that silicide reaction can occur spontaneously at RT and at a low Ni coverage

of ~0.1% Ni with respect to Si Therefore, the presence of hydrogen on Si surfaces does not suppress NiSi formation at RT as previously reported49,58-59,67,189-190

Nevertheless, NiSi2 is not observed on either clean or hydrogen-terminated Si surface, although it was reported to form when Ni is deposited on clean Si surfaces46,48,50

More interestingly, with increasing Ni coverage, the binding energy of Ni 2p3/2

progressively shifted from 854.0eV to 853.8eV, indicating a change from NiSi phase

to a Ni-rich silicide (Ni2Si)5 Eventually it reached and stay at the signature value of a metallic film of 852.8eV The silicide formation sequence at RT is illustrated as following:

Ni + Si
NiSi
Ni2Si
Ni

The implication is that this silicidation reaction only occurs at the initial Ni/Si or Si-Si interface and thereafter more metallic silicide phases were produced until a pure metallic Ni layer was grown above it It is known that hydrogen desorption from the Si surface is only significant above 300oC, and that the Si-H bond is particularly strong with a bond energy of 3.9 eV compared to the Si-Si bond energy of 3.2 eV157 Therefore, the silicide phase formed is most likely also H-terminated, which is consistent with Hirose’s observation that the hydrogen atoms terminating the original

Ni/H-Si surface are still present after the Ni deposition58 Unlike clean surfaces, further deposition of Ni/H-Si at RT will therefore occur above a H-terminated silicide layer instead Irrespectively, we therefore have growth of metallic nickel above initial silicide layers on both clean and H-terminated surfaces as Ni coverage increases The existence of the hydrogen atoms at Ni/H-NiSi interface may subsequently have an effect on the diffusion and nucleation behavior of following deposited Ni adatoms This in turn may lead to a different surface morphology compared to the clean surface, which will be discussed in the next section

It is interesting to note that during Ni deposition, the B.E of Si 2p as shown in Fig 4.1 (b) and (d) remains at 99.3 ± 0.1 eV regardless of Ni coverages This is similar

to the binding energy derived from analysis of bulk Ni2Si, NiSi and NiSi2 samples of 99.39 ± 0.12 eV191, and also similar to clean H-Si(001) substrates (99.2 ± 0.1 eV) The

Si 2p spectra from Ni silicides also do not display any noticeable energy shift during annealing194 For the Si signal detected at low Ni coverage range, the main contribution would come primarily from the Si substrate (Si 2p photoelectron escape depth ~30 Å); hence a shift in Si 2p binding energy due to formation of Ni-Si bonds may not be obvious However, as seen in Fig 4.1(b & d), this remains true even in the Ni coverage

of 89% We rationalize this observation as follows As shown by both semiempirical linear combination of atomic orbital (LCAO) extended Hückel scheme and linear muffin-tin orbital (LMTO) scheme, substantial Si-Ni p-d hybridization takes place for NiSi191 Therefore, Si forms covalent bonds with Ni even after forming Ni silicide192, which exercises little effect on Si’s electronic structure and hence the binding energy

of Si 2p However, the bonding effects on Ni are much more significant since there is a change in bond type from a metallic bonding geometry to a covalent bonding structure when metal silicides are formed A shift in Ni 2p3/2 BE to a higher value is therefore expected since there is a charge transfer from Ni to Si192

4.2.2 The growth mode of Ni on H-terminated Si surfaces

A standard method for the determination of growth mode using XPS is to measure the adsorbate and the substrate signals changes while increasing the exposure

of overlayer metal The S-t plots (photoelectron intensity signal versus evaporation time) should have a characteristic shape depending on the growth modes of adsorbate metal: layer by layer growth (FM growth), three-dimensional growth mode (Volmer-Weber growth, VW), the Stranski-Krastanov (SK) growth mode

In this experiment, Ni was evaporated from the e-beam evaporator onto the Si substrates while maintaining a flux of 20nA through out the experiments The growth was paused after every 10minutes for transferring the sample into XPS analysis chamber for scan without exposing the samples to air Mg Kα X-ray with Large Area (LA) lens mode was adopted to do the scans due to its relatively stable intensity over time Fig 4.3 shows the evolution of the atomic fraction of Ni (Ni 2p3/2) and Si (Si 2p)

as a function of Ni deposition time on both H-terminated Si(001) and Si(111) surface

at room temperature The plots obtained on the H-terminated silicon substrates appear

to be exponential-like, which implies that we may have layer by layer growth mode

10 20 30 40 50 60 70 80 90 100

model; (b) Another possible deviation from the ideal case could be attributed to a growth whereby the second layer starts to grow before the first layer is completed This mode of growth is termed pseudo-layer-by-layer or "simultaneous multilayer" growth mode, and has been similarly observed in the cases of Ni growth on clean Ge(111)72and on Ag(111) surfaces195 (c) Alternatively, we could also be observing Stranski-Krastanov growth mode instead

It is therefore not easy to distinguish the growth mode further without either resorting to some simple fitting or the use of other experimental techniques to probe the surface morphology evolution We will attempt to resolve this observation in two ways: (i) curve fitting the XPS data obtained through the used of known equation that describes pseudo-layer-by-layer growth39 and (ii) examining the surface morphology at various stages of growth using the AFM

For pseudo layer by layer growth mode, the decay of XPS signal from substrate (I ), i.e area under of Si 2psn peak, as a function of material thickness (d) is given by:

)exp(

0

s

a s

n

a

ndI

whereIa∞ is the signal intensity associated with a clean bulk-like film of the adsorbate,

λa is the adsorbate’ photoelectron IMFP in the adsorbate Equation 4.1 and 4.2 can also

be expressed as a function of time, where the product (n×da) can be written as a product of growth rate (G, in Å/min) and time (t, in minutes) The atomic fractions of

CSi and CNi can thus be expressed as a function of time For atomic fractions of Si, we have,

Ni

2p3 2

deposition) and INi∞ (clean Ni foil) are determined on the day of the experiments

Hence, the values of o

Si Si

G t100% exp

Ni Ni

tGt

GK

tGt

C

λλ

λ

exp1exp

1

exp1

%100

)

In order to determinate the nature of the growth mode, equations 4.4 and 4.5 were used to fit the experimental data as given in Fig 4.3 The electron inelastic mean free path (IMFP), λSi and λNi, is a constant for a given material and kinetic energy197

As for the choice of λSi and λNi, we will use the empirical equation previously developed by M.P Seah and W.A Dench197 Their equation relates the IMFP (λ, in units of nm) to the corresponding photoelectron kinetic energy (Ekin, in units of eV) and the type of material through which the photoelectrons traverses and is given below

Ni film is therefore 0.22nm by having atomic weight (A) of 58.69, bulk density of

8908 kg m-3, number of atoms 1, Avogadro’s number of 6.022×1023 The value as determined by equation 4.6 for λSi and λNi will thus be ~ 15Å and 9Å, respectively

Figure 4.4 shows the experimental data for Ni% and Si% evolution on Si(001) and H-Si(111) surfaces re-plotted together with the fits (dashed lines) given by equation 4.4 and 4.5 A reasonable fit through the experimental data points can clearly

H-be obtained for both the atomic fraction of Ni and Si at various deposition times A value of G~0.32 ± 0.01 Å/min is found for the Ni growth on H-Si(001) surface while for Ni growth on the H-Si(111) surface we obtained a value of ~ 0.42 ± 0.01 Å/min The implication of the above results is that Ni appears to follow a pseudo-layer-by-layer growth mode on these H-terminated Si surface

The resulting morphology of the film following this growth mode will likely be smooth Fig 4.5 and Fig 4.6 shows the morphological evolution of the surfaces before and after Ni deposition on hydrogen-terminated Si(001) and Si(111), respectively

(a) (b)

Fig 4.4 The dashed lines represent the simulated pseudo-layer-by-layer growth mode calculated according to equation 4.4 and 4.5 with λSi = 15 Å and λNi = 9 Å Diagrams (a) and (b) represent Ni and Si atomic fraction for Ni/H-Si(001) surface as a function

of Ni deposition time Diagrams (c) and (d) represent Ni and Si atomic fraction for Ni/H-Si(111) surface as a function of Ni deposition time

10 20 30 40 50 60 70 80 90 100

10 20 30 40 50 60 70 80 90 100

(a) (b) (c)

Fig 4.6 1µm × 1µm AFM images of (a) bare H-Si(111) and various Ni coverages equivalent to atomic fractions of (b) ~10%, (c) ~30%, (d) ~54%, (e) ~70% and (f)

~90% Ni on H-terminated Si(111) surface at room temperature

The surface morphology in both Fig 4.5 and Fig 4.6 appears to be decorated with continuous close-packed shallow 2-D domes across the whole Ni coverage range The domes typically have a size ranging from ~9±1 to 14±1 nm with a height ranging from ~1.5±0.2 to 2.4±0.2 Å Hence, the height to size aspect ratio is ~1:60, which suggests a significant wetting of the H-Si surfaces by Ni More importantly, no 3D islands or clusters are observed The surface appears to be smooth and flat with an average RMS roughness of ~ 2.1±0.2 Å and 1.5±0.2 Å remaining the same for all Ni coverages on H-Si(001) and H-Si(111), respectively The observation of smooth surface is consistence with the XPS modeling result that the growth of Ni on the H-terminated surface follows a pseudo-layer-by-layer growth mode

The growth rates on both H-Si(001) and H-Si(111) are very close (~ 0.27 ± 0.01 Å/min) when the same flux of Ni was used for deposition Therefore, the substrate orientation exercises little effect on the growth of Ni In order to probe the influence of incoming Ni flux on pseudo-layer-by-layer growth mode, we also vary the incoming

Ni flux for growth but only on H-Si(001) surfaces

In the following set of experiments, Ni was again evaporated from the e-beam evaporator onto H-Si(001) substrates but using three different incoming fluxes of 5nA, 10nA, and 20nA separately Figure 4.7 shows the evolution of the atomic fraction of

Ni (Ni 2p3/2) as a function of Ni-deposition time with three different fluxes on Si(001) surface at room temperature

H-Fig 4.7 Evolution of the atomic fraction of Ni during Ni evaporation on H-Si(001) surfaces with three fluxes of 5nA, 10nA & 20nA

Figure 4.8 shows the experimental data for Ni% and Si% evolution on Si(001) surfaces with three fluxes re-plotted together with the fits (dashed lines) given

H-by equation 4.4 and 4.5 Again reasonable fits through the experimental data points for all of the three fluxes can be clearly obtained for both the atomic fraction of Ni and Si

at various deposition times Values of G equal to ~0.09 ± 0.01 Å/min, ~0.27 ± 0.01

0 10 20 30 40 50 60

Å/min & ~0.40 ± 0.01 Å/min are found for the Ni flux of 5nA, 10nA, and 20nA, respectively The simulated results imply that Ni appears to follow a pseudo-layer-by-layer growth mode on H-Si(001) surface at all these fluxes Likewise, the resulting morphology of the film under three fluxes will likely be flat and smooth Consequently, the morphology of three surfaces at the end of growth were also examined by the AFM and shown in Fig 4.9

The surfaces in Fig 4.9 appear to be relatively flat and uniform for all these three samples using different fluxes without the presence of any 3D islands The surface appears to be smooth with close-packed 2-D domes The average RMS roughness was of only ~ 2.0±0.2 Å, which agreed with the XPS simulation result that the growth of Ni on the H-Si(001) surface follows a pseudo-layer-by-layer growth mode for the three difference fluxes Hence, varying the incoming Ni flux seems to have little influence on the growth mode of Ni on hydrogen-terminated Si surfaces

10 20 30 40 50 60 70 80 90

10 20 30 40 50 60 70 80 90

(a) (b)

(c) Fig 4.9 1µm × 1µm AFM images of (a) 54%Ni/H-Si(001), (b) 58%Ni/H-Si(001) & 64%Ni/H-Si(001) grown with Ni fluxes of 5nA, 10nA and 20nA, respectively

4.2.3 Comparison with Ni growth mode on clean Si(001) surfaces

In the previous experiments, hydrogen has consistently been used as a surfactant to terminate the reactive dangling bonds on Si surfaces in order to increase the diffusion length of the arriving Ni adatoms, and hence is greatly likely to influence the Ni growth mode It is therefore of interest to look into the growth mode of Ni on clean Si surface and compare it with that on hydrogen-terminated Si surfaces In this aspect, a clean Si(001) surface was prepared by heating the H-Si(001) to 500oC in-situ

to desorb the hydrogen After the Si has cooled down to RT, Ni was again evaporated

on it while maintaining a flux of 20nA through out the experiments Figure 4.10 shows the evolution of the atomic fraction of Ni (Ni 2p3/2) and Si (Si 2p) as a function of Ni-deposition time on both clean and H-terminated Si(001) surface at room temperature There is no significant difference in the evolution of Ni and Si atomic fractions

Fig 4.10 shows the evolution of the atomic fraction of Ni (square) and Si (circle) as a function of deposition time on clean (open symbols) and H-terminated (solid symbols) Si(001) surfaces at room temperature

It is tempting to conclude that the growth mode of Ni on clean Si(001) surface proceeds in the similar fashion as that observed previously for Ni on H-Si(001) surface, although curve fitting of the data for Ni/clean Si(001) yielded a G value of

~0.39 ± 0.01 Å/min, very close to G value of ~0.40 ± 0.01 Å/min for Ni/H-Si(001) However, when the surface of clean Si(001) and the surface at various stages of growth were examined by AFM, a marked difference in morphology was noticed Figure 4.11 shows the AFM scans taken from the surface of clean Si and those with Ni coverages equivalent to atomic fractions of 10%, 20%, 50% and 60% on clean Si(001) surfaces

0 10 20 30 40 50 60 70 80 90 100

Fig 4.11 1µm × 1µm AFM images (column I), 500nm × 500nm AFM images (column II) of surface morphology and line profiles (column III) of (a) clean Si(001) surface and compact Ni islands growth at Ni atomic fractions of (b) 10%, (c) 20%, (d) 50% and (e) 60%

0 0.2 0.4 0.6 0.8 1

X[nm]

While the bare hydrogen-terminated Si(001) in Fig 4.5(a) was flat and less, the surface of clean Si (001) in Fig 4.11(a) appeared to be rougher with the presence of small domes in typical size of ~10-20nm and height of ~0.3-0.5nm The 1µm×1µm AFM image of the bare clean Si(001) surface had a RMS of 2.5Å, compared to a value of 1.3Å previously observed on H-Si(001) surface

At 10% Ni coverage, (Fig 4.11(b)) the surface is decorated with Ni islands with size of ~10-20nm and the height between 0.6 to 1.0nm (the height measurement for all coverages is measured from the maximum to the height position marked by the island boundary) The islands thus have a height to size aspect ratio equivalent to 1:26, which is significantly smaller than the 1:60 observed when Ni was grown on hydrogen-terminated surface at the same coverage

Besides the Ni islands, two types of dark voids started to appear on the surface, type-A (marked with circle box) and type-B (marked with square box) The type-A voids were normally wider than 25nm and deeper than 2nm The width of type-B voids were generally ~10-20nm and their depths were ~0.7-1.1nm Both types of voids were attributed to areas not covered by the deposited Ni, namely incomplete growth of the top Ni layer However, since the depths of the type-A voids are higher than twice the height of the Ni domes while the type-B voids are close to or less than the height of the

Ni domes, we assign type-A and type-B voids to area not covered by Ni domes for more than two consecutive layers and less than one layer, respectively The incomplete coverage of Ni on the surface has led to an increase of RMS from 2.5Å for clean Si (001) to 5.3Å for 10%Ni/clean Si(001)

At a Ni coverage of 20%Ni (Fig 4.11(c)), line scan analysis again reveals that the islands were still ~10-20nm in size and the heights of these islands is typically between 0.5 to 1.0nm Thus there is no increase in height or size even though the

coverage has increased The further deposited Ni seemed to wet the previous layer, filling in both the type-A and type-B voids and then growing on top of it, which were evidenced by the reduced density of both types of voids and a decrease in RMS to 4.0Å

At a Ni% of 50% (Fig 4.11(d)), line scan analysis revealed that the surface seemed to have two layers of islands: the lower layer appeared to be around 0.5-0.7nm high and the top one was taller by another 0.5-0.7nm, adding to an overall height of

~1.0-1.4nm The top layer was stacking up above the layer formed previously Although there is no increase in the lateral size, with the growth of top islands and the formation of new type A and type-B voids in between, an increase in the RMS roughness to ~ 6.2Å was observed

By increasing the Ni coverage equivalent to 60% (Fig 4.11(e)), we again find that the surface was still decorated by the two-layer structure as seen previously at 50% (see Fig 4.11(d)) The density of islands having a height between 1.0-1.4nm has increased while the density of lower islands with height of ~0.5-0.7nm has significantly decreased However, no islands having a height between 1.5-2.1 nm are spotted The implication of these fixed size islands at various stages of deposition was that the growth mode of Ni on clean Si(001) is close to “layer by layer” growth; i.e the

2nd stack of islands starts to grow when the 1st stack of islands has almost completely packed the initial surface Similarly the 3rd stack of islands is seen when the 2nd layer is nearly covered by islands with dimension of ~10-20nm and height of 0.5-0.7nm In doing so, there is no significant 3D growth and film is also relatively flat, although it will continue to leave both types of voids on the previous layers and lead to a porous film We therefore describe the growth of Ni on clean Si(001) as compact-island growth, similar to that of Co on clean Si surfaces57

An aspect ratio of 1:26 would also imply some wetting of the clean Si surface

by Ni during deposition, although it is smaller than the 1:60 observed when Ni was grown on H-Si (001) These factors together with the compact growth of Ni islands make the distinction of growth mode by XPS between pseudo layer by layer and compact islands growth difficult Without further analysis of the surface by AFM, we would not have been able to realise there is a difference in the growth dynamics and resulting films’ quality observed between H-terminated and clean Si (001) surface

4.2.4 Role of surfactants on growth mode

On a Si substrate, the driving force for a Ni film to grow in a 2D layer by layer mode requires the sum of surface free energy of Ni and interfacial energy of Ni/Si to

be less than the surface free energy of Si substrate; i.e

where γNi/v is the surface energy of Ni-vacuum interface, γNi/Siis the interfacial energy between Ni and Si, and γSi/v is the surface energy of Si-vacuum interface This driving force may stem qualitatively from consideration of bond enthalpies It has been reported that Ni-Ni, Ni-Si and Si-Si bond enthalpies are 204 kJ mol-1, 318 ± 17 kJ mol-

1

and 310 kJ mol-1, respectively179.Based on the above values, Ni-Ni bond appears to

be significantly weaker than Ni-Si bond, which implies a preferred bond formation between Ni and Si in the substrate, rather than forming Ni-Ni bonds Consequently, the

Ni atoms landing on the surface will spontaneously react with Si For Ni atoms that land on existing Ni, some may have sufficient energy to diffuse to the edge of the incomplete 2D Ni layer and then react with the remaining exposed Si, while others

may simply adsorb and diffuse on top of the Ni layer In this way, pseudo 2D layer of

Ni will form on the surface Once the Si surface is fully covered by Ni, additional Ni atoms can only react with and grow on the underlying Ni layer Bond enthalpy considerations suggest the presence of a driving force for the 2D growth mode of Ni at room temperature The present XPS experiment results also suggest that there is spontaneous bond formation between Ni and Si prior to any thermal treatment on both H-terminated and clean Si surfaces Thus wetting of Ni on both clean and H-terminated surface is energetically favorable

The morphology obtained on H-Si (001) (Fig 4.5) and clean Si (001) (Fig 4.11) is rather different as shown by AFM examinations Smooth and continuous layer growth was seen on H-terminated surfaces while compact island growth was observed

on clean Si surfaces It should be noted that while the Ni layers grown on either clean

or on H-terminated surface oxidize when the samples are exposed to air, it will occur similarly on both types of surfaces with the same Ni coverage Furthermore, the surface morphology observed also does not change in size or form with increasing exposure time to air It is therefore unlikely that the oxidation process is the reason why compact islands are observed on the clean Si surfaces while a smooth morphology

is observed on the H-terminated surfaces The difference in morphology is attributed more to the different growth kinetics related to the presence of reactive Si dangling bonds on the surface

The difference in morphology observed when Ni grows on H-Si(001) (Fig 4.5) and clean Si(001) (Fig 4.11) can be rationalized as follows The passivation of Si dangling bond by hydrogen on the surface appears to promote diffusion of adatoms on the surface in such a way that arriving Ni atoms have sufficient energy to either diffuse along the substrate surface so as to join existing 2D islands laterally, or diffuse to the

edge of such islands and get incorporated there if they first land on top of these existing 2D islands In doing so, layer-by-layer growth mode is facilitated and the resulting surface of the film is thus smooth and flat As the formation of Ni-Si bonding occurs on both bare and H-passivated Si surfaces, the above results can only be reconciled if only H “floats” to the surface above the initial deposited Ni layer This has been supported by Higai’s first principle calculation that when Ni is deposited onto the Si dimer row, it can capture H from dimer Si and later returns H to Si before Ni moves to the most stable B site190 The next layer of Ni is then added to this H-NiSi-like layer at room temperature When this happen, continuously wetting is allowed and the morphology thus obtained will be flat and smooth

On a clean Si surface however, the Si dangling bonds can easily and immediately react with the deposited adatoms and therefore restrict the metal adatoms migration on the Si surfaces62-64,198 As a consequence Ni adatoms do not have sufficient energy to either diffuse along the substrate surface to join existing 2D islands laterally, or to diffuse to the edge of such islands and get incorporated there if they first land on top of these existing 2D islands Such a decrease in Ni diffusion length will lead to the formation of Ni domes at growth front of the Si surface, and constant presence of the dark voids/trenches since Ni adatoms are difficult to diffuse and fill in those voids due to a reduced mobility Thus, the surface obtained is rougher

4.2.5 Summary

It is evident from the above study that Ni reacts strongly with the Si substrates

to form a thin, amorphous NiSi-like layer at room temperature, which occurs on both clean and hydrogen terminated Si surfaces Unlike Al/H-Si(111) interface where

hydrogen suppressed the formation of interfacial AlSi which promoted island growth64, the presence of hydrogen on Si surfaces did not suppress the formation the Ni-silicide phase at room temperature With further Ni deposition, the interfacial layer becomes richer in Ni and eventually a metallic Ni film forms on top XPS curves appear to suggest that Ni grows via a pseudo-layer-by-layer mode on both clean and hydrogen terminated Si surfaces AFM images showed a smooth surface for Ni deposition on hydrogen terminated Si, while small close-packed island growth on the clean Si surface H-termination appears to play a beneficial role for a smooth surface morphology during Ni growth on Si surface This effect has kinetic rather than energetic origin and appears to be a consequence of an increased surface diffusion rate

of Ni adatoms The XPS curve method was unable to distinguish between very packed small island growth mode from pseudo layer-by-layer growth mode

closed-4.3 XPS and AFM results of Ni growth on Ge substrates

4.3.1 Ni deposition on H-terminated and clean Ge (001) surfaces at RT

Fig 4.12(a) shows the typical Ni 2p3/2 spectra evolution versus Ni coverage grown on hydrogen-terminated Ge(001) surface at RT with Mg Kα X-ray source (the spectra development when depositing Ni on clean Ge(001) was similar, and hence is not shown here) At 19%Ni deposition on H-Ge(001) surfaces, the Ni 2p3/2 peak was observed at a binding energy of 853.3 ± 0.1 eV, which is about 0.5 eV shift away from that of bulk metallic Ni 2p3/2 peak (852.8 ± 0.1 eV) With increasing Ni coverage, the binding energy of Ni 2p3/2 shifted gradually towards the value expected for bulk Ni

film The Ge 3d peaks from the substrate as shown in Fig 4.12(b) showed a shift of 0.1eV from 29.6 ± 0.1 eV to 29.5 ± 0.1 eV after 19% of Ni deposition With increasing

Ni coverage, the binding energy of Ge 3d shifted gradually towards 29.3 ± 0.1 eV, which is 0.3eV lower than the bulk value of Ge 3d in Ge substrates The shift in Ge 3d B.E is attributed to the band-bending effect, because the Ge substrate is n-type while the shift to lower values is gradual

Similar to the behaviour of Ni deposition on Si surfaces, the shift of Ni 2p3/2 to higher binding energy at low Ni coverage on both clean and hydrogen-terminated Ge(001) is also unlikely to be attributed to cluster effects This is because the surface morphology of Ni/H-Ge(001) surface (Fig 4.13) after 19% of Ni deposition was relatively uniform and smooth (RMS ~ 1.8Å) Moreover there was no formation of clusters on the Ge surface

85%Ni 75%Ni 65%Ni 56%Ni 44%Ni 30%Ni 19%Ni

Fig 4.13 represents 1 µm × 1µm AFM images of 19% Ni deposited on H-terminated Ge(001) RT

The shifts in BE of Ni 2p3/2 and Ge 3d suggest that there is a reaction to form

Ni germanide phases upon initial Ni deposition on H-Ge(001) surface at RT There is currently no information about RT reaction between Ni and H-Ge(001) & clean Ge(001) surface in the literature However, it has been pointed out in literature that Ni reacted immediately with the underlying clean Ge(111)72-73,199 and hydrogen-terminated Ge(111)74-75 upon deposition at RT to form an amorphous NiGe phase Thus, we attribute the shift of Ni 2p3/2 to the formation of a Ni germanide phase, most likely NiGe, which occurs on both clean and hydrogen terminated Ge(001) surfaces

To our best knowledge, there is lack of binding energies reference for Ni 2p3/2, Ge 2p3/2 and Ge 3d for NiGe phase in the literature Hence, we have to prepare NiGe phase in order to determine the characteristic B.E for Ni 2p3/2 in NiGe

NiGe is usually prepared by annealing a Ni thin film grown on Ge substrates to high temperature (>300oC) During annealing Ni2Ge was first formed within 150-

250oC9,200 Further increase of the temperature led to formation of Ni5Ge380 or

Ni3Ge2200 depending on the heating rate Eventually, NiGe as the last phase formed between 250 and 600oC9,13,76,80,199-200 Using this method of preparation, NiGe is

prepared by annealing 15%Ni grown on H-Ge(001) to 500oC Several similar experiments were repeated and a consistent value for Ni 2p3/2 of 853.5 ± 0.1eV is obtained Since the Ni 2p3/2 value of 853.3 ± 0.1eV at 19% is close to that in NiGe (853.5 ± 0.1 eV), we attribute the initial shift of Ni 2p3/2 away from metallic value to the formation of NiGe phase

The formation of Ni-Ge bond on clean Ge surfaces is not unexpected as unsatisfied Ge dangling bonds are present at the surface However, it is more difficult

to rationalise the observation of a Ni/Ge reaction and Ni-Ge bond formation on terminated Ge surfaces since all the dangling bonds are now passivated In this case, there are no available reactive Ge dangling bonds to facilitate direct bond formation between Ni and surface Ge atoms Similar to Ni behaviour on H-Si surfaces, it would imply a reaction between Ni and the H-Ge bond or with the Ge back bonding at the H-Ge-Ge surface in order for this reaction to occur

H-Once again, we will resort to the interstitial model188 as proposed for growth on clean Si surfaces in order to explain how Ni germanide reaction can take place at RT

on both clean and H-terminated Ge surfaces In the initial deposition stage, Ni atoms can incorporate into the interstitial sites of the Ge host lattice The consequence of Ni occupying interstitial sites in Ge is the weakening of the Ge-Ge covalent bond Electrons in the neighboring Ge-Ge covalent bonds will no longer remain in their localized states and will have to share between the interstitial and Ge atoms, leading to strong Ge p and Ni d hybridization201 This can result in the formation of a Ni-Ge covalent bond81 at the growth front of Ni/Ge or Ni/H-Ge surfaces Given the limited mobility of adatoms at room temperature, the structure thus formed would likely be different from the equilibrium stable crystalline MnP-type NiGe81 structure We therefore described this phase as a ‘‘NiGe-like’’ film As the germanide reaction can