Computational Investigation of the Adsorption and
Reactions of SiH
x
(x 5 0–4) on TiO
2
Anatase (101) and
Rutile (110) Surfaces
Wen-Fei Huang,
[a]Hsin-Tsung Chen,*
[b]and M. C. Lin*
[a]The adsorption and reactions of the SiHx(x ¼ 0–4) on Titanium dioxide (TiO2) anatase (101) and rutile (110) surfaces have been studied by using periodic density functional theory in conjunction with the projected augmented wave approach. It is found that SiHx(x¼ 0–4) can form the monodentate, bidentate, or tridentate adsorbates, depending on the value of x. H coadsorption is found to reduce the stability of SiHxadsorption. Hydrogen migration on the TiO2 surfaces is also discussed for elucidation of the SiHx decomposition mechanism. Comparing adsorption energies,
energy barriers, and potential energy profiles on the two TiO2 surfaces, the SiHxdecomposition can occur more readily on the rutile (110) surface than on the anatase (101) surface. The results may be used for kinetic simulation of Si thin-film deposition and quantum dot preparation on titania by chemical vapor deposition (CVD), plasma enhanced CVD, or catalytically enhanced CVD. VC 2013 Wiley Periodicals, Inc.
DOI: 10.1002/qua.24388
Introduction
Titanium dioxide (TiO2) has become more attractive recently with numerous theoretical and experimental studies because of its promising applications in photocatalysts and photoelec-trochemical processes.[1,2] The photophysics of TiO2 sensitized by dyes,[3–5] polymers,[6,7] and semiconductors[8–10] have been widely investigated with an aim towards solar energy conver-sion by the photovoltaic effect. The dye-sensitized solar cells (DSSC) were invented by O’Regan and Gr€atzel (1991)[11] and the conversion efficiency of the DSSC have reached over 10%.[3–5,12,13] Organic dye molecules were attached onto the porous TiO2 structures as adsorbate to provide an alternative photovoltaic device for electricity generation.[11,14] However, organic dyes are relatively expensive and less stable with shorter lifetimes than inorganic materials. Accordingly, many researchers[8–10] have attempted to use semiconductor quan-tum dots (QDs, such as InN, InP, InAs, PbS, and CdS) instead of dyes for solar cell applications. Wang and Lin[15]have success-fully deposited the InN films on TiO2 nanoparticle films by low-pressure organometallic chemical vapor deposition and found that the InN/TiO2films showed a broadband adsorption in the UV/visible range, covering 390–800 nm.
To improve the heterogeneous interface between a QD and TiO2and achieve higher efficiencies, there have been attempts to search for a variety of linkers (e.g., H3BO3,[16,17] BClx,[18] H2S,[19] H2O2,[20] and HNO3,[21]). An alternative material, the well-known silicon has been studied extensively for the solar cell applications.[22–31] Previously, the growth of Si on TiO2 ru-tile (110) surface has been reported by Abad et al.[32]In addi-tion, Lin et al.[33] have used SiH4 as a precursor with Ar as a carrier gas to grow silicon QDs over a TiO2 nanoparticle film by CVD. The Si QDs were protected by SiCx or SiNx thin films formed by CVD or plasma enhanced CVD (PECVD). The Si QDs distributed layer can generate electron-hole pairs to enhance
the photon–electron transfer efficiency. The underlying physi-cochemical processes are very complex and it is hard to obtain the information on the detailed reaction mechanism under very limited experimental conditions. Accordingly, to improve our understanding of the complex Si-QDs growth process, we use the density functional theory (DFT) calculation to elucidate the reaction mechanism involved in the formation of the first layer Si film over the TiO2nanoparticle film.
In this article, we study the mechanisms for the decomposi-tions of SiHxon the two TiO2surfaces, anatase (101) and rutile (110), in relation to the potential growth of Si thin films and QDs, which may also be used as linkers for other QDs to attach on TiO2. First, we study the adsorption energies and geometries of various SiHxspecies with and without hydrogen coadsorption, useing DFT. The potential energy profiles of the reactions are calculated by nudged elastic band (NEB) method to be discussed below. The results of this study may help understand the SiHx adsorption and decomposition mecha-nisms on the TiO2surface for fabrication of solar cells by CVD, PECVD, or catalytically enhanced CVD (Cat-CVD).
Computational Methods and Models
All calculations were performed by the spin-polarized DFT with the projected augmented wave method[34]as implemented in the Vienna ab initio simulation package.[35,36] The ionic cores were described by the generalized gradient approximation [a] W.-F. Huang, M. C. Lin
Department of Applied Chemistry and Center of Interdisciplinary Molecular Science, National Chiao Tung University, Hsinchu 300, Taiwan
[b] H.-T. Chen
Department of Chemistry and Center for Nanotechnology, Chung Yuan Christian University, Chungli 32023, Taiwan
E-mail: htchen@cycu.edu.tw or chemmcl@emory.edu
with the PW91[37] formulation which has been shown to work well for gas-surface reactions.[16,21,38,39] The electronic orbitals were represented by the plane-wave expansion including all the plane waves with their kinetic energies smaller than the chosen cutoff energy, hk2/2m \ Ecut, which ensures the con-vergence. The calculations were carried out with a 500 eV cut-off energy using the (4 4 4) and (3 3 1) Monkhorst– Pack k-points for the bulk and (101) surface of anatase, whereas the (2 4 2) and (4 3 1) Monkhorst–Pack k-points with the same cutoff energy were chosen for bulk and (110) surface of rutile, respectively. The NEB method[40] was applied to locate the transition states (TS). Two to eight images were used for each calculated TS. All the transition structures were verified by the frequency calculations. The adsorption energies were calculated by the following equa-tion:
Eads¼ ðEtotal Esurf EgasÞ
where Etotal, Esurf, and Egas are the electronic energies of the adsorbed species on the surface, a clean TiO2 surface, and a gas-phase molecule, respectively. In addition, we consider the Gibbs free energy, DG, of each gaseous species and their adsorption on the surfaces. The change of Gibbs free energy with finite temperature and pressure considered of the reac-tion can be calculated as follows:
DGadsðT; PÞ ¼ Eads Esurf Egasþ Fvib;adsðTÞ ½DHgð0K ! T; P0Þ TSgðT; P0Þ þ kTlnðp=P0Þ where the Gibbs free energies of solid surface have relatively small variation with respect to temperature and pressure and are approximated by the DFT calculated total energy.
FvibðTÞ ¼ X n i¼1 FvibðT; v iÞ ¼ Xn i¼1 hvi 2 þ kT lnð1 e bhviÞ
where b ¼ 1/kT and vi is the vibration frequency. The total vibrational correction can be expressed as
Fvib;adsðTÞ ¼ Fvib adsðTÞ F vib g ðTÞ where Fvib adsðTÞ, F vib g ðTÞ, and F
vib;adsðTÞ are vibrational Gibbs free energies of the adsorbed molecule, the free gas molecule, and their change arising from the adsorbate-surface interaction, respectively.
The TiO2 nanoparticle film is a polycrystalline material with different phases dependent on annealing temperature,[41–43] particle size,[44–46] and shape.[47] The rutile (110) and anatase (101) surfaces, which have the lowest surface energies with similar characteristics, may coexist in a nanoparticle film.[15,48] Theoretically, the (101)[49,50] and (110)[49] are the most stable surfaces of the anatase and rutile, respectively. Accordingly, both surfaces are considered in this calculation. The surface super cells are modeled as periodically repeated slabs each consisting of 24 units with three and four Ti layers for anatase
and rutile, respectively, separated by a vacuum space greater than 13 A˚ , which guaranties no interactions between the slabs, for both surfaces as shown in Supporting Information Figure S1. The lower layers of each slab are fixed to preserve the cal-culated bulk parameters, whereas the remaining layers are fully relaxed to simulate the surface behavior during the calculations.
As shown in Supporting Information Figure S1, four different adsorption sites on both rutile (110) and anatase (101) surfaces are labeled. They are five-fold coordinated titanium, six-fold coordinated titanium, two-fold bridging oxygen, and three-fold coordinated oxygen, corresponding to Ti5c, Ti6c, O2c, and O3c, respectively. The coordinating unsaturated O2c and Ti5c sites are more active than the coordinating saturated O3c and Ti5c sites.[51,52] Gas-phase molecules are simulated in a cubic box with 20 A˚ on a side, which is large enough to ignore interac-tions between each periodic gas molecules.
To shorten the nomenclature, HxSi–y(a) denotes SiHx adsorption on the y site with the Si atom. For y, O represents an O2csite and Ti represents a Ti5csite. If three O adsorption sites coexist, the third O is an O3csite. For example, HSi–3O(a) represents SiH adsorption on the O2c, O2c, and O3csites with the Si atom. Furthermore, HxSi–y, H(a) denotes that HxSi–y(a) coadsorbs with one hydrogen on an adjacent O2c site, and HxSi–y…nH(a) represents that SiHxcoadsorbs with n H atoms on the O2csite at a long distance.
Results and Discussion
To verify the reliability of the computational results, we first compared the calculated bulk lattice constants with experi-mental values. The predicted lattice constants are a¼ 3.824 A˚ and c ¼ 9.678 A˚ for anatase, which are in good agreement with the experimental values[53,54] of a ¼ 3.872–3.875 A˚ and c¼ 9.502–9.514 A˚ , whereas those for rutile are a ¼ 4.593 A˚ and c¼ 2.958 A˚ which are in reasonable agreement with the experimental values[53] of a ¼ 4.587–4.593 A˚ and c ¼ 2.954– 2.959 A˚ . The predicted fractional coordinates (0.208 and 0.304 for the anatase and rutile bulk, respectively) also agree well with the experimental values (0.208[55]and 0.305[56]). The geo-metries of gas-phase molecules, SiHx(g) (x ¼ 1–4), are also examined, and the results are summarized in Table S1 of the Supporting Information. Our results are in very good agree-ment with the experiagree-ment data.
Adsorption and reaction of SiHx(x 5 0–4) on the TiO2 anatase (101) surface
Adsorption of SiHxon the anatase (101) Surface. For each SiHx adsorption on the TiO2anatase (101) surface, there are several possible configurations. SiHxcan adsorb at the O2cor Ti5csite to form a monodentate adsorbate, a bidentate adsorbate at the (O2c)2 or (O2c)(Ti5c) site, or a tridentate adsorbate at the (O2c)2(O3c) site, depending on the value of x. All the adsorption energies are listed in Table 1, and the geometries are shown in Figures 1, 4, and S2 of the Supporting Information. Figure 1 presents the geometries for the most stable configurations.
The other geometries involved in the potential energy profiles (discussed in the next section) are shown in Figure 4, and the remaining ones are shown in Supporting Information Figure S2. The adsorption free energies of most stable SiH4fragments on the TiO2rutile (110) and TiO2anatase (101) Surfaces at the temperature of 298, 400, 600, and 1000 K, respectively, are cal-culated and listed in Table 2.
As shown in Figure 1, SiH4 molecule can interact with an O2c atom of the anatase surface by physisorption, using its H atoms. The lengths of H…O2care 2.778, 2.811, and 3.534 A˚ . Its adsorption energy is calculated to be 1.2 kcal/mol; see Table 1. To carry out mechanistic studies, we examined the adsorption sites and energies for SiH3, SiH2, SiH, and Si as well. The most
stable configuration for SiH3 adsorption is located at an O2c site with an adsorption energy of 57.6 kcal/mol. The bond length of SiAO2c is 1.670 A˚ . On the contrary, the SiH3 only weakly bonds with a Ti5csite with a small adsorption energy of 2.2 kcal/mol. For the SiH2 adsorption, the most stable con-figuration is located at the O2c, O2cbridging site (H2Si–2O(a)). The H2Si–2O(a) has an adsorption energy of 90.6 kcal/mol and the Si atom is bound to two O anions of the surface with SiAO distances of 1.673 A˚. In addition, the formation of the SiAO bonds produces much relaxation of the surface O anion bonded to the silicon atom, resulting in the protrusion by approximately 0.42 A˚ from the surface. The monodentate adsorption at the O2c site (H2Si–O(a)) and bidentate Table 1. Adsorption energies (kcal/mol) of SiH4fragments on the TiO2rutile (110) and TiO2anatase (101) surfaces with and without a coadsorbing H
as depicted in the specified figures.
Adsorbate Adsorption site
On anatase surface On rutile surface
With HAO2c(a)[a] With HAO2c(a)[a]
Figure Eads Figure Eads Figure Eads Figure Eads
SiH4 AO2c 1a 1.2 5a 2.6
SiH3 AO2c 1b 57.6 4b 53.4 5b 76.1 8b 68.4
ATi5c S2a 2.2 S2b 17.7 S7a 11.8 S7b 20.7
SiH2 AO2c 4e 38.6 4d 36.7 8e 50.5 8d 40.7 AO2c,ATi5c S2c 50.6 S2d 50.3 AO2c,AO2c 1c 90.6 S2e 85.7 5c 103.2 S7c 79.1 SiH AO2c S2f 52.0 S2g 51.8 S7d 69.5 S7e 64.3 AO2c,ATi5c S2h 52.1 S2i 45.6 AO2c,AO2c 4i 81.2 4h 78.2 8k 87.8 8j 75.4 AO2c,AO2c,AO3c 1d,4n 108.6, 98.5 S2j, S2k 96.7, 91.3 5d 125.8 S7f 109.8 Si AO2c S2l 61.0 S2m 56.2 S7g 79.0 S7h 65.4 AO2c,ATi5c S2n 66.0 S2o 61.9 AO2c,AO2c S2p 90.4 S2q 84.4 S7i 110.4 S7j 96.2 AO2c,AO2c,AO3c 1e, 4q 100.5, 91.3 4l, 4p 93.9, 83.8 5e 113.4 8p 100.8
[a] With hydrogen coadsorption on an adjacent O2csite.
Figure 1. The optimized geometries of adsorbed SiH4and its fragments, SiH3, SiH2, SiH, and Si, on the TiO2anatase (101) surface. [Color figure can be
adsorption at the O2c, Ti5c sites (H2Si–OT(a)) are less stable (see Table 1). For SiH and Si adsorption, five different configu-rations on different sites of the anatase surface are found. Dif-ferent from SiH2, the SiH and Si preferentially adsorb on two O2c and one O3c anions of the surface, forming HSi–3O(a)_1 (Fig. 1d), HSi–3O(a)_2 (Fig. 4n) and Si–3O(a)_1 (Fig. 1e), Si– 3O(a)_2 (Fig. 4q), respectively. As shown in Table 1, the HSi– 3O(a)_1 and Si–3O(a)_1 are more stable than HSi–3O(a)_2 and Si–3O(a)_2. The adsorption energy of HSi–3O(a)_1 is 108.6 kcal/mol and the SiAO2cand SiAO3cdistances are 1.633 and 1.711 A˚ , respectively. While the corresponding values for Si–3O(a)_1 are 100.5 kcal/mol and 1.748 and 1.884 A˚ , respectively.
In addition, we also consider the SiHx and H coadsorption. Their optimized structures are depicted in Figure 4 and Figure S2 of the Supporting Information. As summarized in Table 1, the adsorption energy with hydrogen attaching to a neighbor-ing oxygen site decreases in the range of 0.2–11.9 kcal/mol. These results indicate that the hydrogen coadsorption may destabilize the SiHxadsorption on the TiO2anatase (101) sur-face, unlike the enhancement effects of H-coadsorption on the binding of OH,[57,58] HS,[19]NO3,[21]and HOBO2[16]species with titania.
Reaction of SiHx on the anatase (101) Surface. Hydrogen
Migration and H2 Desorption. The decomposition reaction of SiH4 can be schematically expressed by the following mechanism:
H4SiðaÞ ! H3SiðaÞ þ H O2cðaÞ ! H2SiðaÞ þ 2H O2cðaÞ
! :::::: ! SiðaÞ þ 4H O2cðaÞ (1)
The dissociated H atoms can adsorb on adjacent O2csites to potentially block active sites and thus stop the subsequent decomposition reactions. Accordingly, the migration of H atoms has been studied first. Here, we only consider the hydrogen adsorption on the O2c site, H–O2c(a) because the adsorption energy of H on the O2csite (56.1 kcal/mol) is more stable than that on the Ti5csite (0.7 kcal/mol).
It is readily understandable that the distribution of hydrogen adsorption sites will affect the transition state energy barriers strongly. To solve this problem, we consider one of the SiAH bonds of SiHx firstly decomposes to a nearest O2c site, and then migrates to an adjacent O2csite. The hydrogen atom can effectively ‘‘jump’’ on adjacent O2c sites away from SiHx for it to undergo further decomposition. The scheme is shown in Supporting Information Figure S3. As depicted in Figure 2, one hydrogen adsorbs on the #1O2c site can migrate along two directions, one along \ 111 > to the #2O2c site and another one along \ 010 > to the #4O2c site. The hydrogen can migrate to the#2O2csite by overcoming TS2a with 13.1 kcal/ mol. The migration to the#4O2csite is achieved by a stepwise reaction. The hydrogen moves to the#3O2csite with an endo-thermicity of 15.4 kcal/mol via TS2b (26.7 kcal/mol), and then continues moving to the #4O2c site with an exothermicity of 15.4 via TS2c (11.7 kcal/mol). In conclusion, the hydrogen migration is easier along the \ 111 > than the \ 010 > direction.
The other possible reactions are also considered; two hydro-gen atoms adsorbed at two neighboring O2csites may recom-bine to give one H2 gas molecule, which is another possible reaction to vacate adjacent O2c sites. As shown in Figure 2b, the energy barrier (TS2d) for forming H2(g) is 43.2 kcal/mol, Table 2. Adsorption free energies (kcal/mol) of the most stable SiH4fragments on the TiO2rutile (110) and TiO2anatase (101) surfaces with and
without a coadsorbing H.
Adsorbate Temperature (K)
On anatase surface On rutile surface
With HAO2c(a)[a] With HAO2c(a)[a]
Figure DGads Figure DGads Figure DGads Figure DGads
SiH4 298 1a 7.8 – – 5a 6.4 – – 400 9.7 – 8.3 – 600 14.1 – 12.7 – 1000 22.2 – 20.8 – SiH3 298 1b 45.4 4b 41.2 5b 63.9 8b 56.2 400 42.5 38.3 61.0 53.3 600 35.7 31.5 54.2 46.5 1000 22.6 18.4 41.1 33.4 SiH2 298 1c 76.0 S2e 71.1 5c 88.6 S7c 64.5 400 72.8 67.9 85.4 61.3 600 65.5 60.6 78.1 54.0 1000 51.4 46.5 64.0 39.9 SiH 298 1d 95.6 S2j 83.7 5d 112.8 S7f 96.8 400 92.3 80.4 109.5 93.5 600 85.2 73.3 102.4 86.4 1000 71.4 59.5 88.6 72.6 Si 298 1e 92.6 4l 86.0 5e 105.5 8p 92.9 400 90.1 83.5 103.0 90.4 600 85.0 78.4 97.9 85.3 1000 75.0 68.4 87.9 75.3
which is much higher than the energy barrier (TS2a, 13.1 kcal/ mol) of the lowest hydrogen migration path. Hence, hydrogen atom migration is considered as the main process for vacating two adjacent O2c sites by the SiHx adsorbate. Other possible reaction paths for forming H2(g), for example, HAO2c,HAO2c(a) –> H2–O2c(a) –> H2(g) þ TiO2(s), are also considered; however, all the energy barriers involved are higher than TS2d of 43.2 kcal/mol.
SiHx Decomposition. Figure 3 shows the schematic potential energy profiles for the multipath on the TiO2 anatase (101) surface, and the related decomposition energy barriers and reactions energies are listed in Table 3. As seen in Figure 3a, the reaction starts from the SiH4 adsorption at an O2c site, denoted by SiH4:O2c (Fig. 1a), with an exothermicity of 1.2 kcal/mol. The SiH4decomposition can be realized by breaking one of the SiAH bonds with an energy barrier (TS3a) of 22.1 kcal/mol. The dissociating H atom primarily attaches to an ad-jacent bridging O2c anion forming an OAH bond of 0.978 A˚ and co-adsorbs with SiH3, producing H3Si–O,H(a) (Fig. 4b) with an exothermicity of 15.4 kcal/mol. The coadsorbing hydrogen in H3Si–O,H(a) can jump along the nearest O2csite to a long distance, producing H3Si–O…H(a), with an exother-micity of 4.2 kcal/mol.
As shown in Table 4, the energy barriers of hydrogen migra-tion on the anatase (101) surface with different coexisting
adsorbates differ only by 0.4–0.5 kcal/mol. Therefore, we can consider all the energy barriers between HxSi–O,H(a) and HxSi–O…H(a) to be approximately 13.1 kcal/mol. The dissocia-tion energy for SiH4(g) –> SiH3(g) þ H(g) is 90.4 kcal/mol,[59] which is much larger than the energy barrier at TS3a, 20.9 kcal/mol, for SiH4decomposition on the anatase (101) surface as one would expect.
As shown in Figure 3b, the second decomposition process takes place by overcoming a 44.1 kcal/mol reaction barrier at TS3b, producing hydrogen coadsorbed H2Si–O,H(a) with an endothermicity of 39.6 kcal/mol. The Si…H distance is 3.653 A˚ in H2Si–O,H(a) (Fig. 4d), in which the coadsorbates, SiH2 and H, interact weakly, hence the energy difference between H2Si– O,H(a) and H2Si–O…H(a) is only 1.9 kcal/mol. The SiH2radical can adsorb on the surface with monodentate and bidentate configurations; therefore, three possible reaction paths are also considered as follows:
H2SiOðaÞ ! HSi O:::HðaÞ ! SiO:::2HðaÞ (2) ! H2SiOTðaÞ ! HSiOT:::HðaÞ ! SiOT:::2HðaÞ (3) ! H2Si2OðaÞ ! HSi3O:::HðaÞ ! Si3O:::2HðaÞ (4) The related potential energy profiles are shown in Figure 3 and Supporting Information Figure S4. The path (4), named multipath, is the most likely reaction pathway due to the most Figure 2. Schematic potential energy profiles for a) the H migration and b) the H2(g) formation on the TiO2anatase (101) surface. [Color figure can be
stable intermediates and the smaller energy barriers involved. The paths (2) and (3), named O2c-path and O2c,Ti5c-path, respectively, are less important; they are shown in Supporting Information Figure S4. In the following, we only discuss the most likely reaction pathway (4).
As illustrated in Figure 3c, the SiH2 bound at the two O2c sites, H2Si–2O(a) (Fig. 1c), can be formed from H2Si–O(a) by overcoming an energy barrier (TS3c) of 9.6 kcal/mol with an
Figure 3. Schematic potential energy profiles for the multipath on the TiO2anatase (101) surface. The detail geometries are shown in Figure 4, and the
re-ferring adsorption energies are shown in Table 1. (For brevity, in Figs. 3b, 3c, and 3d, one, two, and three H(a) co-adsorptions are, respectively, omitted although they are included in the energy calculations. Hence, the reactant H3Si–O(a) represents H3Si–O…H(a) and H2Si–O(a) denotes H2Si–O…2H(a)).
[Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
Table 3. The decomposition energy barriers and the reaction energies for SiHx(a)fi SiHx-1...H(a) (unit: kcal/mol).
x¼ 4 x¼ 3 x¼ 2 x¼ 1
On anatase Energy barrier 20.9 44.1 26.5 33.4
Reaction energy 20.8 37.7 23.1 4.1
On rutile Energy barrier 28.8 38.6 22.9 28.7
Figure 4. The optimized geometries along the multipath on the TiO2anatase (101) surface. [Color figure can be viewed in the online issue, which is
exothermicity of 52.0 kcal/mol. Comparing with the formation of another intermediate, H2Si–OT(a) (Supporting Information Fig. S2c), TS3c is 2.6 kcal/mol higher than TSs3 (Supporting Information Fig. S4) because the O2c…O2c distance (3.823 A˚ ) is larger than the O2c…Ti5c distance (1.839 A˚ ). However, H2Si–2O(a) is 40 kcal/mol more stable than H2Si–OT(a), and the SiAO2c bond length (1.673 A˚ ) in H2Si–2O(a) is shorter than the SiAO2c and SiATi5c bond lengths (1.685 and 2.522 A˚ ) in H2Si–OT(a). Overcoming a 42.9 kcal/mol energy barrier at TS3d, one of the HASi bonds of H2Si–2O(a) can dissociate to the adjacent O2c site to form the coadsorption HSi– 2O,H(a) (Fig. 4h) with an endothermicity of 31.9 kcal/mol. The HSi–2O(a) adsorbate (Fig. 4i) can be formed with 3.0 kcal/mol exothermicity from the HSi–2O,H(a) by H migration. The tridentate adsorbate, HSi–3O(a)_1 is formed from the bidentate adsorbate, HSi–2O(a), by passing over TS3e with an exothermicity of 27.4 kcal/mol, as shown in Figure 3d. The SiAO3c bond (1.711 A˚ ) is 0.078 A˚ longer than SiAO2c bond (1.633 A˚ ) in HSi–3O(a)_1 due to the coordination with the saturated O3c site. The H-migration from HSi–3O(a)_1 pro-ducing Si–3O,H(a)_1 (Fig. 4l) with an endothermicity of 29.9 kcal/mol via TS3f has a high barrier of 60.8 kcal/mol. The coadsorbed hydrogen can move from the Si–3O,H(a)_1 con-figuration to produce the single Si adsorption, Si–3O(a)_1
(Fig. 1e), with an exothermicity of 6.6 kcal/mol. By overcom-ing TS3g with a 15.2 kcal/mol barrier, the bidentate adsorb-ate, HSi–2O(a), also can bond with another O3csite to form HSi–3O(a)_2 (Fig. 4n) with an exothermicity of 17.3 kcal/mol. The tridentate adsorbate, HSi–3O(a)_2, is 10.1 kcal/mol less stable than HSi–3O(a)_1. Different from HSi–3O(a)_1, the dehydrogenation of HSi–3O(a)_2 has a very high barrier of 78.4 kcal/mol because the distance of H to the closest O2c site in HSi–3O(a)_2 (5.175 A˚ ) is too long for hydrogen disso-ciation. However, the HSi–3O(a)_2 can isomerize to the HSi– 3O(a)_1 by passing through an energy barrier of only 17.4 kcal/mol (TS3h).
The mechanisms for the dehydrogenation process, HxSi– O,H(a) –> H2(g)þ Hx-1Si–O(a) for x¼ 1–3, are also considered for SiHx decomposition. The energy barriers for x ¼ 1–3 are calculated to be 45.3, 33.9, and 52.8 kcal/mol, respectively, which are much higher than the energy barrier of hydrogen migration (TS2a, 13.1 kcal/mol). The potential energy profiles for these three reaction paths are shown in Figure S6 of the Supporting Information. Obviously, the reaction paths of HxSi– O,H(a) –> H2(g)þ Hx-1Si–O(a) are less likely to take place than those of HxSi–O,H(a) –> HxSi–O…H(a) –> Hx-1Si–O,H…H(a) –> Hx-1Si-O…2H(a), due to their higher energy barriers.
Adsorption and reaction of SiHx(x 5 0–4) on the TiO2rutile (110) surface
Adsorption of SiHxon the rutile (110) Surface. Figure 5 shows the geometries for the most stable configurations on the rutile (110) surface. The other geometries involving in the potential energy profiles (discussed in the next section) are shown in Figure 8, and the remaining ones are shown in Supporting
In-formation Figure S7. Similar to the results on the anatase (101) surface, the SiH4 mole-cule physisorbs on the rutile (110) surface with an adsorption energy of 2.6 kcal/mol, and the most stable configura-tions for each SiHx are H3Si–O(a), H2Si–2O(a), HSi–3O(a), and Si–3O(a), with their correspond-ing adsorption ener-gies: 76.1, 103.2, 125.8, and 113.4 kcal/mol.
Comparing with the
results on the anatase (101) surface, the SiHx adsorbates on the ru-tile surface are bound
more strongly with
higher energies by 6.6– 20.0 kcal/mol. Notably, the HxSi–OT(a) structures Table 4. Energy barriers (kcal/mol) of H migration affected by adjacent
adsorbates.
None SiH3 SiH2 SiH
On anatase 13.1 12.7 13.6 12.7
On rutile 31.1 28.3 28.3 27.6
Figure 5. The optimized geometries of adsorbed SiH4and its fragments, SiH3, SiH2, SiH, and Si, on the TiO2rutile (110)
cannot be formed on the rutile (110) surface because of the dif-ferent topography.
The SiHx and H coadsorption is also considered here, and the optimized structures are depicted in Figure 8 and Figure S7 of the Supporting Information. Similar to the results on the anatase (101) surface, the H coadsorption destabilizes the SiHx adsorption. The decreases in adsorption energies by H coad-sorption are in the range of 5.2–24.1 kcal/mol, which are more significant than those on the anatase (101) surface.
Reaction of SiHx on the rutile (110) Surface. Hydrogen
Migration. Similar to the results on the anatase (101) surface, the hydrogen adsorption at the O2c site (66.3 kcal/mol) is more stable than that at the Ti5csite (1.6 kcal/mol). Hence, we also consider hydrogen migration on the rutile (110) surface to understand the SiHxdecomposition on the surface. As shown in Figure 6, one hydrogen at#5O2csite can migrate along two directions on the rutile (110) surface, one along \001 > to the #6O2c site and another one along \ 110 > to the
#7 O2c site. To realize the migration along the \ 111 > direction, we double the size of the surface model in the \ 111 > direction. As shown in Figure 6, the hydrogen migrates to the#6O2csite with a 31.1 kcal/mol barrier at TS6a. However, the long dis-tance (6.586 A˚ ) between#5O2cand
#7
O2cincreases the difficult for migration along the \ 111 > direction. Hence, the only
possible direction for hydrogen migration is along the \ 001 > direction. Comparing to the anatase (101) surface, hydrogen migration occurs less easily on the rutile (110) sur-face due to the lower energy barrier for the former process (13.1 vs. 31.1 kcal/mol). Hence, the SiHx (x ¼ 1–4) stepwise decomposition may be easier on the anatase (101) surface due to the vacating of the nearest O2csite by hydrogen migration.
Similar to the anatase case, the reactions of H2(g) formation cannot compete with the hydrogen migration reactions. As shown in Figure 6b, the energy barrier (TS6b) for H2(g) pro-duction is predicted to be 64.9 kcal/mol, which is much higher than the energy barrier (TS6a, 31.1 kcal/mol) of the lowest hydrogen migration path.
SiHxDecomposition. Similar to the reactions on the TiO2anatase (101) surface, the reactions of SiHxon the TiO2rutile (110) sur-face can be divided into two paths: O2c-path and multipath. However, the O2c, Ti5c-path does not exist on the rutile (110) surface since the HxSi–OT(a) configuration does not exist. The multipath shown in Figure 7 is the main path in the rutile case, and the structures involved are shown in Figure 8. The related decomposition energy barriers and reaction energies are listed in Table 3. The O2c-path is the less important reac-tion path, as shown in Figure S8 of the Supporting Information.
Similar to the reactions on the anatase surface, the first step of the reactions on the TiO2 rutile (110) surface is the SiH4 physisorbed on the O2csite, named SiH4:O2c(Fig. 5a), with an exothermicity of 2.6 kcal/mol as alluded to above. By overcom-ing an energy barrier of 31.4 kcal/mol from SiH4:O2cvia TS7a, H3Si–O,H(a) (see Fig. 8b structure) can be formed exothermi-cally by 32.7 kcal/mol. The migration of the H atom from the O2csite in H3Si–O,H(a) producing H3Si–O…H(a) is exothermic by 7.7 kcal/mol. As shown in Figure 7b, the H2Si–O,H(a) (see Fig. 8d structure) can be formed endothermically with 37.3 kcal/mol by breaking an HASi bond in H3Si–O(a) via TS7b (38.6 kcal/mol). The H2Si–O…H(a) can be formed by hydro-genation migration in H2Si–O,H(a) with an exothermicity of 9.9 kcal/mol. The HASiAH line of the bonds is along \ 110 > direction in H2Si–O,H(a). The SiH2molecule rotates in concur-rent with the hydrogen migration.
The bidentate adsorbate, H2Si–2O(a), can be formed from the monodentate adsorbate, H2Si–O(a), by overcoming a very small bending barrier of 0.9 kcal/mol at TS7c with a large exo-thermicity of 52.7 kcal/mol attributable to the formation of a new bond. Different from the case on the anatase (101) sur-face, the formation of HSi–3O(a) on the rutile (110) surface is a stepwise path. The migration of one of the H atoms in H2Si– 2O(a) may be accomplished by forming the intermediate M8a (see Fig. 8h structure) first with a large endothermicity of 46.8 kcal/mol via TS7d by overcoming a very high barrier of 73.3 kcal/mol. Comparing with the energy for hydrogen dissocia-tion from H2Si–2O(a) producing HSi–2O(a)þ H(g) (95.0 kcal/ mol), the transition state TS7d is 21.7 kcal/mol lower in energy as one would expect. Overcoming TS7e with a barrier of 28.8 kcal/mol, the H on the O2csite in M8a can migrate to the clos-est O2csite forming HSi–2O,H(a) (see Fig. 8j structure), with an Figure 6. Schematic potential energy profiles for a) the H migration and
b) the H2(g) formation on the TiO2rutile (110) surface. [Color figure can be
Figure 7. Schematic potential energy profiles for the multipath on the TiO2rutile (110) surface. The detail geometries are shown in Figure 8, and the
refer-ring adsorption energies are shown in Table 1. (For brevity, in Figs. 7b, 7c, and 7d, one, two, and three H(a) co-adsorptions are, respectively, omitted although they are included in the energy calculations. Hence, the reactant H3Si–O(a) represents H3Si–O…H(a) and H2Si–O(a) denotes H2Si–O…2H(a) here.) [Color
exothermicity of 5.7 kcal/mol. The migration of the detached
H atom from the Si atom in HSi–2O,H(a) producing
HSi–2O…H(a) (see Fig. 8k structure) is exothermic by 12.4 kcal/mol.
Similar to the case on the anatase (101) surface, the HSi– 2O(a) can adsorb on the surface with a tridentate configura-tion forming the HSi–3O(a) (see Fig. 4d structure) with an exo-thermicity of 38.0 kcal/mol via TS7f having a small bending Figure 8. The optimized geometries along the multipath on the TiO2rutile (110) surface. [Color figure can be viewed in the online issue, which is available
barrier of 0.2 kcal/mol. The O2cATi6c bond of HSi–3O(a) is almost broken with a long distance of 2.848 A˚ . Similar to the case of H2Si–2O(a), hydrogen migration from HSi–3O(a) can occur by a stepwise pathway. First, the hydrogen migrates to the O2csite, which binds with the SiH molecule, forming M8b (see Fig. 8n structure) with an endothermicity of 29.5 kcal/mol via TS7g by overcoming a large barrier of 66.7 kcal/mol. For HSi–3O(a), the direct hydrogen dissociation into the gas-phase occurs with an endothermicity of 79.6 kcal/mol, which is higher than TS7g. Overcoming an energy barrier of 36.5 kcal/ mol via TS7h, the hydrogen of M8b can migrate to the closest O2c site with an exothermicity of 2.9 kcal/mol, forming Si– 3O,H(a) (see Fig. 8p structure). The hydrogen migration pro-cess from Si–3O,H(a) produces the tridentate adsorbate, Si– 3O…H(a) (see Fig. 5e structure), with an exothermicity of 13.3 kcal/mol.
Similar to the anatase case, the mechanisms for the dehy-drogenation reactions, HxSi–O,H(a) –> H2(g)þ Hx-1Si–O(a) for x ¼ 1–3, are also considered. The energy barriers for x ¼ 1–3 are 48.6, 14.3, and 51.4 kcal/mol, respectively; see Figure S10 of the Supporting Information. Similar to the anatase case, the reactions for H2(g) formation cannot compete with the hydro-gen migration reactions due to their higher energy barriers.
Comparing with the results on the anatase (101) surface, the energy barriers of the H2Si–O(a) –> H2Si–2O(a) and HSi– 2O(a) –> HSi–3O(a) processes on the rutile surface are higher by 11 kcal/mol on average. Conversely, the intermediates and products are more stable on the rutile surface than those on the anatase surface with much larger exothermicities of 42 kcal/mol on average (see Figs. 3 and 7).
The reaction energies for SiHxdissociation reactions in the gas phase are known as follows:[59]
SiH4ðgÞ ! SiH3ðgÞ þ HðgÞ DE¼ 90:4 kcal=mol SiH3ðgÞ ! SiH2ðgÞ þ HðgÞ DE¼ 68:1 kcal=mol SiH2ðgÞ ! SiHðgÞ þ HðgÞ DE¼ 73:6 kcal=mol SiHðgÞ ! SiðgÞ þ HðgÞ DE¼ 72:8 kcal=mol
These energies are significantly higher than the barriers for their dissociation reactions on the two titania surfaces accord-ing to the potential energy profiles depicted in Figures 3 and 7, reflecting the catalytic effects of the TiO2 surfaces which may help the growth of Si thin films and QDs upon annealing.
Conclusions
The mechanisms for SiHx decomposition on the two TiO2 surfaces, anatase (101) and rutile (110), have been investigated by periodic DFT calculations. Our results show that SiH4 physi-sorbs weakly on the two surfaces, whereas the SiH3 preferen-tially adsorbs at an O2c site with monodentate configuration. SiH2binds strongly at two O2csites with a bidentate configu-ration and SiH and Si adsorb more favorably at the (O2c)2(O3c) sites with tridentate configurations. Their adsorption energies are predicted to be 1.2, 57.6, 90.6, 108.6, and 100.5 kcal/mol on the anatase (101) surface and 2.6, 76.1, 103.2, 125.8, and
113.4 kcal/mol on the rutile (110) surface, respectively. Hydro-gen coadsorption with these species was found to reduce the stability of SiHx adsorption, contrary to the known electron withdrawing groups such as OH and SH, among others. Hydro-gen migration on the TiO2 surfaces has also been studied in conjunction with SiHx decomposition. We find that hydrogen migration is easier on the anatase (101) surface. In addition, the potential energy surfaces for the reactions of SiHxon the two TiO2 surfaces have been constructed. The stable inter-mediates on the two surfaces are bidentate adsorbates, H2Si– 2O(a), with binding energies >90 kcal/mol and tridentate adsorbates, HSi–3O(a) and Si–3O(a), with binding energies >100 kcal/mol. Comparing the adsorption energies on the two TiO2surfaces, the SiHxspecies bind more strongly on the rutile (110) surface by as much as 42 kcal/mol on average, whereas the barriers for their decomposition reactions are higher than those on the anatase surface by 11 kcal/mol on average. The results of this study may be used, for example, for kinetic sim-ulations of Si thin-film deposition and QD growth upon annealing on titania by CVD, PECVD, or Cat-CVD.
Acknowledgments
The authors are grateful to Taiwan’s National Center for High-per-formance Computing for the CPU’s facility and to Taiwan National Science Council under Grant Numbers NSC 101-2113-M-033-009-MY3 and NSC 101-2632-M-033-001-MY2 for the research support. MCL also wants to acknowledge Taiwan Semiconductor Manufac-turing Co. for the TSMC Distinguished Professorship and Taiwan National Science Council for the Distinguished Visiting Professorship at the Center for Interdisciplinary Molecular Science, National Chiao Tung University, Hsinchu, Taiwan.
Keywords: titanium dioxide
gas-surface reactionsSiHx decompositiondensity functional theoryHow to cite this article: W.-F. Huang, H.-T. Chen, M. C. Lin, Int. J. Quantum Chem. 2013, 113, 1696–1708. DOI: 10.1002/qua.24388 Additional Supporting Information may be found in the online version of this article.
[1] K. I. Hadjiivanov, D. G. Klissurski, Chem. Soc. Rev. 1996, 25, 61. [2] A. L. Linsebigler, G. Lu, J. T. Yates, Chem. Rev. 1995, 94, 735. [3] M. Gratzel, Nature 2001, 414, 338.
[4] M. K. Nazeeruddin, A. Kay, I. Rodicio, R. Humphry-Baker, E. Muller, P. Liska, N. Vlachopoulos, M. Gratzel, J. Am .Chem. Soc. 1993, 115, 6382. [5] B. O’Regan, M. Gratzel, Nature 1991, 353, 737.
[6] C. L. Huisman, A. Goossens, J. Schoonman, Chem. Mater. 2003, 15, 4617.
[7] D. Li, C. Gu, C. Guo, C. Hu, Chem. Phys. Lett. 2004, 385, 55.
[8] P. Yu, K. Zhu, A. G. Norman, S. Ferrere, A. J. Frank, A. J. Nozik, J. Phys. Chem. B 2006, 110, 25451.
[9] J. L. Blackburn, D. C. Celmarten, A. J. Nozik, J. Phys. Chem. B 2003, 107, 14154.
[10] R. Vogel, P. Hoyer, H. Weller, J. Phys. Chem. 1994, 98, 3183. [11] B. O’Regan, M. Gr€atzel, Nature 1991, 353, 737.
[12] A. Yella, H. -W. Lee, H. N. Tsao, C. Yi, A. K. Chandiran, M. K. Nazeerud-din, E. W. -G. Diau, C. -Y. Yeh, S. M. ZakeerudNazeerud-din, M. Gr€atzel, Science 2011, 334, 629.
[13] C. Chen, M. Wang, J. Li, N. Pootrakulchote, L. Alibabaei, C. Ngoc-le, J. Decoppet, J. Tsai, C. Gr€atzel, C. Wu, S. Zakeeruddin, M. Gr€atzel, ACS Nano 2009, 3, 3103.
[14] W. R. Duncan, O. V. Prezhdo, Annu. Rev. Phys. Chem. 2007, 58, 143. [15] J. -H. Wang, M. C. Lin, ChemPhysChem. 2004, 5, 1615.
[16] P. Raghunath, M. C. Lin, J. Phys. Chem. C 2008, 112, 8276. [17] P. Raghunath, M. C. Lin, J. Phys. Chem. C 2009, 113, 3751.
[18] J. -G. Chang, J. -H. Wang, M. C. Lin, J. Phys. Chem. A 2007, 111, 6746. [19] W. -F. Huang, H. -T. Chen, M. C. Lin, J. Phys. Chem. C 2009, 113, 20411. [20] H. Wen-Fei, P. Raghunath, M. C. Lin, J. Comput. Chem. 2011, 32, 1065. [21] C. -Y. Chang, H. -T. Chen, M. C. Lin, J. Phys. Chem. C 2009, 113, 6140. [22] A. Shah, P. Torres, R. Tscharner, N. Wyrsch, H. Keppner, Science 1999,
285, 692.
[23] R. B. Bergmann, Appl. Phys. A 1999, 69, 187.
[24] L. Stalmans, J. Poortmans, H. Bender, M. Caymax, K. Said, E. Vazsonyi, J. Nijs, R. Mertens, Prog. Photovolt. Res. Appl. 1998, 6, 233.
[25] B. Hamilton, Semicond. Sci. Technol. 1995, 10, 1187. [26] R. L. Smith, S. D. Collins, J. Appl. Phys. 1992, 71, R1.
[27] A. G. Cullis, L. T. Canham, P. D. J. Calcott, J. Appl. Phys. 1997, 82, 909. [28] P. R. McCurdy, L. J. Sturgess, S. Kohli, E. R. Fisher, Appl. Surf. Sci. 2004,
233, 69.
[29] V. G. Erkov, S. F. Devyatova, E. L. Molodstova, T. V. Malsteva, U. A. Yanovskii, Appl. Surf. Sci. 2000, 166, 51.
[30] P. G. Karlesson, J. H. Richter, M. P. Andersson, J. Blomquist, H. Sieg-bahn, P. Uvdal, A. Sandell, Surf. Sci. 2005, 580, 207.
[31] S. Takabayashi, R. Nakamura, Y. Nakato, J. Photochem. Photobiol. A 2004, 166, 107.
[32] J. Abad, C. Rogero, J. Mendez, M. F. Lopez, J. A. Martı´n-Gago, E. Roman, Appl. Surf. Sci. 2004, 234, 497.
[33] M. C. Lin, T. -N. Yang, S. -M. Lan, T. -Y. Wei, J. -P. Chiu, W. Y. Ma, Method of fabricating silicon quantum dots-sensitized layer of silicon quantum dots-sensitized solar cells, Issued/Publication Date: 2007/09/ 01; Patent Number: 200733406; IPC: H01L-031/042(200601), Taiwan. [34] P. Blochl, Phys. Rev. B 1994, 17, 953.
[35] G. Kresse, J. Furthmuller, J. Comput. Mater. Sci. 1996, 6, 15. [36] G. Kresse, J. Furthmuller, Phys. Rev. B 1996, 54, 11169. [37] J. P. Perdew, Y. Wang, Phys. Rev. B 1992, 45, 13244.
[38] J. -H. Wang, M. C. Lin, J. Phys. Chem. B 2006, 110, 2263.
[39] J. -H. Wang, M. C. Lin, Y. -C. Sun, J. Phys. Chem. B 2005, 109, 5133. [40] G. Mills, H. Jonsson, G. K. Schenter, Surf. Sci. 1995, 324, 305. [41] K. D. Kim, S. H. Kim, H. T. Kim, Colloids Surf. 2005, 254, 99. [42] G. L. Li, G. H. Wang, Nanostruct. Mater. 1999, 11, 663.
[43] Y. Hwu, Y. D. Yao, N. F. Cheng, C. Y. Tung, H. M. Lin, Nanostruct. Mater. 1997, 9, 355.
[44] H. Zhang, J. F. Banfield, J. Mater. Chem. 1998, 8, 2073. [45] H. Zhang, J. F. Banfield, J. Phys. Chem. B 2000, 104, 3481.
[46] M. R. Ranade, A. Navrotsky, H. Z. Zhang, J. F. Banfield, S. H. Elder, A. Zaban, P. H. Borse, S. K. Kulkarni, G. S. Doran, H. J. Whitfield, Proc. Natl. Acad. Sci. 2002, 99, 6476.
[47] A. N. Enyashin, G. Seifert, Phys. Status Solidi B 2005, 242, 1361. [48] W. Li, C. Ni, H. Lin, C. P. Huang, S. I. Shah, J. Appl. Phys. 2004, 96, 6663. [49] F. Labat, P. Baranek, C. Adamo, J. Chem. Theory Comput. 2008, 4, 341. [50] M. Lazzeri, A. Vittadini, A. Selloni, Phys. Rev. B 2001, 63, 155409. [51] A. Vittadini, A. Selloni, F. P. Rotzinger, M. Gratzel, Phys. Rev. Lett. 1998,
81, 2954.
[52] M. Ramamoorthy, D. Vanderbilt, R. D. King-Smith, Phys. Rev. B 1994, 49, 16721.
[53] J. K. Burdett, T. Hughbanks, G. J. Miller, J. W. Richardson, Jr., J. V. Smith, J. Am. Chem. Soc. 1987, 109, 3639.
[54] C. J. Howard, T. M. Sabine, F. Dickson, Acta Cryst. 1991, 47, 462. [55] M. Horn, C. F. Schwerdtfeger, E. P. Meagher, Z. Kristallogr. 1972, 136,
273.
[56] H. Perron, J. Vandenborre, C. Domain, R. Drot, J. Roques, E. Simoni, J. -J. Ehrhardt, H. Catalette, Surf. Sci. 2005, 72, 518.
[57] J. Zhao, B. Li, K. D. Jordan, J. Yang, H. Petek, Phys. Rev. B 2006, 73, 195309.
[58] K. Onda, B. Li, J. Zhao, K. D. Jordan, J. Yang, H. Petek, Science 2005, 308, 1154.
[59] NIST Computational Chemistry Comparison and Benchmark Database, NIST Standard Reference Database Number 101, Release 15b, August 2011, Editor: Russell D. Johnson III, Available at: http://cccbdb.nist.gov/
Received: 14 October 2012 Revised: 11 December 2012 Accepted: 17 December 2012 Published online on 8 January 2013
