DSpace at VNU: A computational study on the adsorption configurations and reactions of SiHx(x=1-4) on clean and H-covered Si(100) surfaces

Lin Please cite this article as: Thong N-M Le, P.Raghunath, L.K.Huynh, M.C.Lin, A Computational Study on the Adsorption Configurations and Reactions ofSiHxx = 1-4 on Clean and H-covered

Accepted Manuscript

Title: A Computational Study on the Adsorption

Configurations and Reactions of SiHx(x = 1-4) on Clean and

H-covered Si(100) Surfaces

Author: Thong N-M Le P Raghunath L.K Huynh M.C Lin

Please cite this article as: Thong N-M Le, P.Raghunath, L.K.Huynh, M.C.Lin,

A Computational Study on the Adsorption Configurations and Reactions ofSiHx(x = 1-4) on Clean and H-covered Si(100) Surfaces, Applied Surface Sciencehttp://dx.doi.org/10.1016/j.apsusc.2016.06.099

This is a PDF file of an unedited manuscript that has been accepted for publication

As a service to our customers we are providing this early version of the manuscript.The manuscript will undergo copyediting, typesetting, and review of the resulting proofbefore it is published in its final form Please note that during the production processerrors may be discovered which could affect the content, and all legal disclaimers thatapply to the journal pertain

A Computational Study on the Adsorption Configurations and Reactions of

Thong N-M Le1, P Raghunath2, L K Huynh3* and M C Lin2*

1

Molecular Science and Nano-Materials Laboratory, Institute for Computational Science and

Technology, Quang Trung Software Park, Dist 12, Ho Chi Minh City, Vietnam

Graphical Abstract

The barriers for hydrogen abstractions from the adsorbed speices are negligible comparing to the barriers for the decompositions

Abstract

Possible adsorption configurations of H and SiHx (x=1-4) on clean and H-covered Si(100) surfaces are determined by using spin-polarized DFT calculations The results show that, on the clean surface, the gas-phase hydrogen atom and SiH3 radicals effectively adsorb on the top sites, while SiH and SiH2 prefer the bridge sites of the first layer Another possibility for SiH is to reside on the hollow sites with a triple-bond configuration For a partially H-coverd Si(100) surface, the mechanism is similar but with higher adsorption energies in most cases This suggests that the surface species become more stable in the presence of surface hydrogens The minimum energy paths for the adsorption/migration and reactions of H/SiHx species on the surfaces are explored using the climbing image-nudged elastic band method The competitive surface processes for Si thin-film formation from SiHx precursors are also predicted The study reveals that the migration of hydrogen adatom is unimportant with respect to leaving open surface sites because of its high barriers (˃ 29.0 kcal/mol) Alternatively, the abstraction of hydrogen adatoms by H/SiHx radicals is more favorable Moreover, the removal of hydrogen atoms from adsorbed SiHx, an essential step for forming Si layers, is dominated by abstraction rather than the decomposition processes

Keywords: silane precursor, silicon surfaces, Si(100), hydrogen abstractions, plasma enhanced

Introduction

Silicon has played an important role in the fabrication of integrated circuits [1] The advances made from traditional silicon wafers to silicon thin films make it feasible for a wider range of applications related to future microelectronic [2] and photovoltaic [3] devices Plasma-enhanced chemical vapor deposition (PECVD) [4] is a commonly used technique to grow amorphous [5-8] or crystalline [2, 8-10] silicon thin films from silane precursors It has been reported that this technique significantly increased the growth rate of silicon against thermal depositions (CVD) [6] under identical conditions, although these processes were controlled by similar surface kinetics [9] In plasma-enhanced depositions, the reactive radicals are amplified by collisions of energetic electrons with the precursor molecule such as SiH4 Co-existing in the gas phase, besides SiH4 (silane), SiH2 (silylene) and silylidyne (SiH), hydrogen (H) and SiH3 (silyl) are known as the most abundant radicals in this regime [5, 7] Fast growth rates are continuously achieved by the adsorption and reactions of the generated radicals with the silicon substrate surface

The deposition techniques have evolved in practical applications with large-scale production, but the key process of silicon thin-film growth has not been fully understood at the molecular level, although the reactions of SiHx on silicon surfaces have been extensively studied both experimentally [8, 10-12] and computationally [13-21] Based on experimental studies, Gate and co-workers [11, 12] proposed the growth mechanism of silicon which became the preliminary source for a large number of later works In these studies, a steady-state surface kinetic model was also developed to estimate the silicon growth rate The model provided a rather good description on both mechanism and kinetics for CVD processes Gate’s mechanism not only contributed to the understanding of CVD but also PECVD because of their similarities in surface kinetic occurrences, although PECVD is more complex than CVD in the aspect of the plasma chemistry producing reactive species for surface reactions to occur However, the drawback of this model was to assume the overall growth rate to be equivalent to the dissociative adsorption of SiH4[12] Srinivasan [10] showed that hydrogen abstractions and hydrogen etching via Eley-Rideal (ER) [22]

C In another experiment, Srinivasan [8] found that there was a transition from amorphous silicon to crystalline silicon by hydrogen abstractions at temperatures lower than 250 oC Obviously, these experiments concentrated on abstraction processes for activating adsorption sites on the surface, which was absolutely necessary in the plasma-enhanced deposition

On the other hand, computational studies can critically cover various approaches from classical to ab

initio calculations conducted on both cluster and slab models These calculations, moreover, can extend

over a wide range of surface processes occurring throughout realistic deposition conditions involving abstractions of surface hydrogens, adsorptions of gas-phase radicals, diffusions of surface radicals, decompositions of adsorbed species, and abstractions of hydrogens from adsorbed radical species

Ramalingam et al [13] showed that there was no energetic difference for hydrogen abstraction by SiH3from the crystalline or amorphous Si(100)-(2x1) surface However, the mobility of the radicals was higher for the amorphous than the crystalline surface In another work, Ramalingam and co-workers [14] mentioned that the “valley-filling mechanism” accounts for the surface roughness, i.e., the mobile precursors such as SiH3 diffuse and react with dangling bonds on the valleys Cereda [15] estimated the probability of 60% for the barrierless abstraction of surface hydrogens by silyl radicals via the ER mechanism Bakos and coworkers [16] proposed three pathways for H abstraction by SiH3, i.e., ER, Langmuir-Hinshelwood (LH) [22] and precursor-mediated (PM) [22], where ER was barrierless, while the barriers for LH and PM models were 17.5-18.0 and 9.0 kcal/mol, respectively The calculation results from Kang and Musgrave [17] showed that the barrier for surface hydrogen abstractions was less than 1.0 kcal/mol As can be seen from the above summary, SiH3 radicals can diffuse on the surface and then abstract the surface hydrogens via Eley-Rideal mechanism without a significant barrier When the dangling bonds are available, further steps are the adsorptions and reactions of radicals on the surface The dissociative adsorption of silane was also calculated by Kang and his coworker [17] with a favorable barrier of 7.4 kcal/mol This barrier was also found to be 12-14 kcal/mol by Brown and coworkers [18, 19] Smardon [20] confirmed that silanes adsorbed dissociatively on Si(100)-(2x2); the fragments could

be either on the same dimer or on adjacent dimers with the energy difference of 4.2 kcal/mol Kang and

his coworker [17] determined that the barriers for hydrogen removals from adsorbed SiH3, along with transforming to SiH2 bridging structure, were 5.7 and 32.9 kcal/mol with and without H(g), respectively The full decomposition of SiH3 radicals to Si and H adatoms was also mentioned in Ceriotti’s calculations [21], which are in good agreement with the experimental data from Gate and coworkers [11] Overall, the deposition process extensively incorporates competitive surface processes including diffusions, abstractions and decompositions These surface processes should be taken into consideration for constructing a full thin-film growth mechanism It has been known that such a growth mechanism remains unclear Moreover, the kinetics for most of these processes is not fully characterized experimentally or computationally

The purpose of the study is to investigate the mechanisms of the adsorption and decomposition mechanisms of SiHx(x=1-4) radicals on the Si(100) surface leading to the thin-film growth using the density functional theory (DFT) The scope of this study includes: (1) hydrogen migrations on the clean surface, (2) hydrogen abstractions from the surface by either hydrogen atoms or silyl radicals, (3) decompositions of surface silicon-hydride species, and (4) hydrogen abstractions from the surface silicon-hydride species These surface reactions involve both ER and LH mechanisms that contribute to the thin-film growth by producing of surface dangling bonds and dehydrogenation of adsorbed species The microscopic understanding on the growth mechanism is essential to establish the kinetics as well as to develop a realistic simulation model which may help predict the evolution of each surface species under practical conditions When validated, the mechanism may provide a greater opportunity to effectively optimize the production of silicon thin films in industrial scales

Computational methods

All calculations have been carried out using the Vienna Ab initio Simulation Package (VASP) [23-26] based on periodic density functional theory (DFT) The frozen ionic cores are described by the projector augmented wave (PAW) method [27], and the Kohn-Sham valence states are expanded in the plane wave

basis sets up to 380 eV The exchange-correlation energy is described by the generalized gradient approximation with the Perdew-Burke-Ernzerhof (PBE) functional [28-30]

The p(2×2) cell of the Si(100) surface was modeled as periodically repeated slabs with six atomic layers and four Si atoms on each layer For the surface calculations, slabs were separated by a vacuum spacing greater than 17 Å in the direction perpendicular to the surface, which guarantees no interaction between the slabs The three top layers were allowed to relax for all geometry optimizations, while the three bottom layers were fixed at bulk positions with the experimental bulk lattice constant (5.43 Å) [31] The surface Brillouin zone was sampled with the Monkhorst-Pack scheme [32] using a (6×6×1) k-point mesh converging with respect to the electronic energy The ionic relaxation was stopped until the forces

on all free atoms were less than 0.02 eV/Å The Gaussian smearing method [26, 33] with a smearing parameter of 0.01 eV was applied The total energy of all gas-phase species were calculated in a box with dimensions of 15Å on each side, large enough to ensure negligible interactions between neighboring cells Spin-polarized calculations were performed to account for the magnetic properties of SiHx radicals It should be mentioned that we did not include zero-point energy corrections (ZPE) to the total energies presenting hereafter Climbing image-nudged elastic band (CI-NEB) method [34] was employed to search the minimum energy paths (MEP) and/or locate the transition state structures (TS) Vibrational frequencies of adsorbates, transition states and other complexes were also calculated Each TS was confirmed to have only one imaginary (negative) vibrational mode

Results and discussion

1 Adsorption configurations

a) Gas-phase H, H2 and SiHx Species

The spin-polarized calculations were selectively applied to each type of atoms and molecules to obtain minimum energies corresponding to the stable structures Figure 1 shows the optimized structures of gas-phase H2, SiHx species along with their optimum bond lengths and bond angles These calculated geometries are in good agreement with previous experimental observations and electronic structure calculations as compared in Table 1

H2(g) SiH(g) SiH2(g) SiH3(g) SiH4(g)

Figure 1 Optimized geometries of H2/SiHx (x=1,4) gas-phase The bond lengths (Å) and bond angles (o) are also included on each structure

Table 1 Comparing between bond distances, bond angles and vibrational frequencies with the

experimental and computational results for gas-phase H/SiHx species

No Species Bond distance (Å) Bond angle (

2263, 956, 956, 799)

2297, 2292, 991,

991, 955, 955, 955)

[a] Huber and Herzberg [35], [b] Irikura [36], [c] Gubte and Prasad [37], [d] Jacox [38], [e] Yamada and Hirota [39], [f] Boyd [40], [g] deduced by its symmetry, [h] Shimanouchi [41] The bond distances and bond angles involves H-H, Si-H separation and H-Si-H angle, respectively The numbers in parentheses

are from Raghunath et al [42] predicted by electronic structure calculations with the Gaussian program

b) Reconstructed Si(100) surfaces

Si(100) surfaces are widely used as substrates for microelectronic and optoelectronic device

fabrications [17, 19] In this periodic DFT calculations, the clean Si(100) surfaces either reconstruct to

p(2x1) or p(2x2) symmetries of the buckled dimers The alternating-buckled p(2x2) dimer has a lower energy of 2.0 kcal/mol than the buckled p(2x1) dimer These calculations are in good agreement with the results from Roberts and Needs [43] Figure 2 shows optimized configurations for different Si(100) surfaces, i.e., ideal clean (a), p(2x1)-2dimer reconstructed (b), p(2x2)-2dimer reconstructed (c) and

p(2x1)-2dimer hydrogen-covered (d) The dimer bond length for p(2x2) symmetry is found to be 2.37 Å, while the bond lengths between the up-site and down-site with the second layer are 2.39 and 2.34 Å, respectively (see Figure 2, (c)) These findings are close to the values of 2.27, 2.35 and 2.32 Å for the dimer bond length, up-site and down-site bonding to the second layer, respectively, using all-electron

method employed by Tang et al [44] Additionally, these results are also in good agreement with the experimental data from Tromp et al [45], i.e., 2.36, 2.40 and 2.34 Å for the dimer bond length, up-site

and down-site bonding to the second layer, respectively The small energy difference between the structure (2x1) in (b) and (2x2) in (c) indicates that these geometries co-exist locally on the reconstructed Si(100) surface In this study, we only focus on the adsorptions and reactions of SiHx on the p(2x2) unit cell and H-covered p(2x1), correspondingly

(a) Ideal clean Si(100)

Etot = -114.6 (eV)

(b) Clean Si(100)-2x1

Etot = -117.4 (eV)

(c) Clean Si(100)-2x2

Etot = -117.5 (eV) (d) H-covered Si(100)-2x1

Figure 2 The side view of the optimized geometries of different Si(100) surfaces: (a) ideal clean Si(100),

(b) reconstructed p(2x1), (c) reconstructed p(2x2) and (d) hydrogen-covered The bond lengths are given

in Å The structures (c) and (d) will be used for the current study The grey and white spheres are Si and

H atoms, respectively The total energies for the three clean Si(100) surfaces are also presented above The energy difference between the clean Si(100)-2x1 and Si(100)-2x2 structure is 0.1 eV, or about 2 kcal/mol

c) Hydrogen adsorption

Figure 3 shows two possible configurations for the adsorption of hydrogen on the surface sites, i.e.,

h-si-t1 and h-si-t2 for top-1 (up-Si atom) and top-2 (down-Si atom), respectively The bond length of the

adsorbed hydrogen and Si surface is typically around 1.5 Å It is shown that, on the clean surface, hydrogen atoms can strongly adsorb on both sites of a buckled dimer with comparable adsorption energies, 71.3 vs 69.8 kcal/mol for top-1 and top-2, respectively (cf Table 2) Because of the H Si

attraction, the surface Si atom moves up to higher positions out of the surface plane Consequently, the geometry of full hydrogen-covered surface, configuration (d) on Figure 2, becomes highly symmetrical comparing to the clean surface, configuration (c) onFigure 2

sih-si-b-intra2 sih-hsi-b-intra2 sih-si-b-inter1 sih-hsi-b-inter1

sih-si-b-inter2 sih-hsi-b-inter2 sih-si-hol sih-hsi-hol

sih 2 -si-b-intra2 sih 2 -hsi-b-intra2 sih 2 -si-b-inter1 sih 2 -hsi-b-inter1

sih 2 -si-b-inter2 sih 2 -si-b-inter2

SiH 3

sih 3 -si-t1 sih 3 -si-t2 sih 3 -hsi-t1,2

SiH 4

sih 4 -si-t1 sih 4 -si-t2

Figure 3 Adssorption configurations of H/SiHx on clean and H-covered Si(100) surface The notation

sih-si-b-intra1 indicates SiH (sih) adsorbs on clean Si surface (si) at bridge site (b) involving intra-dimer,

with configuration 1 (intra1), while the notation sih-hsi-b-inter1 represents SiH (sih) adsorbs on hydrogen-covered Si surface (hsi) at bridge site (b) involving inter-dimer, with configuration 1 (inter1)

The grey and white spheres are Si and H atoms, respectively

d) SiH adsorption

On the clean surface, SiH radicals adsorb favorably on to bridge sites of either the same dimer dimer) or two adjacent dimers (inter-dimer) of the first layer The most stable structure is found to be the inter-dimer bridging with the adsorption energy of 87.1 kcal/mol, see the geometry in Figure 3, sih-si-b-

(intra-inter1 and the energy in Table2, SiH, (a) (intra-inter1 There is another inter-dimer bridging with 73.7 kcal/mol,

i.e., sih-si-b-inter2, which is geometrically different from sih-si-b-inter1 with respect to the hydrogen position The longer bond length with the second layer of sih-si-b-inter1 against sih-si-b-inter2 (3.09 vs

2.55 Å, respectively) makes it more stable than its counterpart In the presence of adsorbed hydrogens, the

adsorption energies for sih-hsi-b-inter1 and sih-hsi-b-inter2 are 92.0 and 80.6 kcal/mol, recpectively

The higher adsorption energies in the presence of hydrogens suggested that the hydrogenation regime is energetically more favorable than the adsorption on the clean surface But the covered hydrogens can block the dangling bonds, resulting in a slower kinetics for adsorption processes Intra-dimer bridging,

sih-si-b-intra1 in Figure 3, is the second stabilizing structure with the adsorption energy of 79.7 kcal/mol

on the clean surface In this case, the dimer bond length stretches from 2.37 Å to 2.62 Å in the parallel direction to the dimer row, leaving an open place for SiH bridging The partially hydrogen-covered surface regime has a small effect on SiH intra-dimer adsorption because the energy difference is about 1.0 kcal/mol, i.e., 80.7 vs 79.7 kcal/mol for with and without hydrogen coverage, respectively When SiH

the adsorption energy is 76.0 kcal/mol In the hydrogen-covered regime, the adsorption energy of the

hollow-site, sih-hsi-hol, is 76.7 kcal/mol, which is a small difference from the adsorption on the clean

surface It is also shown that the adsorption of SiH on top sites is less stable than any other adsorption

sites On a dimer, SiH radicals only adsorb on to the up-site, sih-si-t1 in Figure 3, with the adsorption energy of 54.4 kcal/mol They directly transform into the inter-dimer bridging of sih-si-b-inter1 or sih-si-

b-inter2 geometries whenever their initial positions are in the down sites The adsorption behaviors of

SiH on a buckled dimer can be interpreted by taking into account the electronic properties of the dimer components It is widely known that the buckled-up, buckled-down dimer components perform as a nucleophilic and electrophilic site [20], respectively, while SiH is a reactive radical attributing to the richly unpaired electrons As a result, the SiH radicals weakly adsorb on the buckled-up dimer component, but strongly attach to the other site Moreover, the SiH radicals on the down sites can bridge

to the inter dimer components to further minimize the adsorption energies as shown on Figure 3,

sih-si-b-inter1, sih-si-b-inter2

In the hydrogen-covered regime with only one open site left for SiH adsorption, it forms an unstable

complex for the up-site adsorption; while the down-site, sih-hsi-t2 in Figure 3, has the adsorption energy

of 56.4 kcal/mol There is a possibility for SiH bridging between the first and second layer in the hydrogenation regime The Si-Si bond length, in this case, stretches from 2.34 to 2.50 Å for the

adsorption to occur with the adsorption energy of 66.4 kcal/mol, corresponding to the geometry

sih-hsi-b-intra2 in Figure 3

e) SiH2 adsorption

On the clean surface, SiH2 favorably captures the two dangling bonds either on the same dimer dimer) or the adjacent dimers (inter-dimer) of the first layer In the former case, in response to the interaction with SiH2, the original dimer bond length stretches out from 2.37 to 2.48 Å opposite to the

(intra-dimerization direction, giving sih 2 -si-b-intra1 in Figure 3 This relaxation is corresponding to the

adsorption energy of 73.7 kcal/mol, which is close to the results from Chen et al [46] with 77.0 kcal/mol

There are two ways for the inter-dimer bridging with respect to the two different sites of the adjacent dimers The first inter-dimer configuration is bonding between the up-site of one dimer with the down-site

of the adjacent dimer, i.e., bonding between the nearest neighbors, giving sih 2 -si-b-inter1 The other

inter-dimer configuration should be bonding between two up-site atoms of the adjacent inter-dimers, sih 2

-si-b-inter2 The former case is more stable than the other with the adsorption energy of 74.4, which is in good

agreement with the result of Chen and co-workers [46], i.e., 74.5 kcal/mol Similar to the adsorption of SiH on the top sites of the clean surface, SiH2 only exists on the up-site, not the down-site of the dimer The adsorption of SiH2 on the top sites is less stable than any other active sites Accordingly, they reconstruct to more stable adsorption configurations before breaking their hydrogen bonds On the hydrogen-covered surface, comparing to the clean surface, the energies are higher for the adsorptions on the top (50.7 vs 47.3 kcal/mol) and inter1 (81.6 vs 74.4 kcal/mol), but lower for the adsorption on intra-dimer (60.9 vs 73.7 kcal/mol) and inter2 (46.6 vs 54.2 kcal/mol) This trend is in agreement with the

conclusions from Lim et al [47], who reported that the inter-dimer configurations are more favorable in

the presence of adsorbed hydrogens It is also possible for SiH2 bridging between the first and second

layer in the hydrogenation regime, sih 2 -hsi-b-intra2 in Figure 3, corresponding to the adsorption energy

of 50.5 kcal/mol, Table 2, SiH2, intra2

f) SiH3 adsorption

The SiH3 radical is the most abundant species under PECVD conditions Therefore, the adsorption of SiH3 is significant for surface film growths On the clean surface, SiH3 radicals adsorb on both sites of the buckled dimers with the energy difference only 1.0 kcal/mol, 58.3 for top-1 and 57.3 kcal/mol for top-2,

cf Table 2 These results are comparable with the calculated values from Chen et al [46], 55.8 vs 57.0

kcal/mol for top-1 and top-2 configurations, respectively, using 6-dimer slab model The bond lengths between the Si radical and Si surface are the same for both cases, i.e., 2.35 Å, cf Figure 3, sih3 -si-t1 for

top-1 and sih 3 -si-t2 for top-2 On the hydrogen-covered surface, there are almost the same structures, cf

Figure 3, sih 3 -hsi-t1,2 with the adsorption energies of 61.9 and 62.0 kcal/mol for top-1 and top-2,

g) SiH4 adsorption

The SiH4 molecules are weakly bound to the clean Si(100) surface with similar adsorption energies of

2.6 and 2.4 kcal/mol on top-1 and top-2 adsorption sites (cf Figure 3, sih 4 -si-t1 for top-1 and sih 4 -si-t2 for

top-2), respectively These structures are mainly different in the relative position of the SiH4 molecules along the dimer direction For top-1, the SiH4 molecule is right above the dimer bond, while for top-2, the SiH4 molecule locates in the hollow site, between the two separate dimer rows The bond lengths between the Si surface and H in the SiH4 are also similar, i.e., 1.90 and 1.97 Å for top-1 and top-2, respectively These stable configurations will be considered for studying various decomposition pathways of SiH4 on the surface