a The energy band structure and b partial density of states PDOS of undoped TiO 2 anatase crystal from our calculations with respect to the I4 1 /amd unit cell.. They are quite different
Trang 1In fluences of metallic doping on anatase crystalline titanium dioxide:
sensitized solar cell (DSSC)
Computational Materials Science Laboratory, Faculty of Physics, Hanoi University of Science, 334 Nguyen Trai, Thanh Xuan, Hanoi, Viet Nam
h i g h l i g h t s
Ca, Al and W dopants strongly distort the lattice and narrowed the band gap
Nb negatively shifts while the others positive shift the conduction band bottom
Nb and W dopants reduce Ti4þto Ti3þwithout forming oxygen vacancy
Be, Mg, Ca, Zn and Al dopants induce oxygen vacancy without Ti3þ
Nb and W inhibit the surface defects while the others do the reversed manner
a r t i c l e i n f o
Article history:
Received 24 April 2013
Received in revised form
4 September 2013
Accepted 21 December 2013
Keywords:
Ab initio calculations
Band structure
Electronic structure
Defects
a b s t r a c t
In this work, we examined the influences of metallic X dopants (X ¼ Be, Mg, Ca, Zn, Al, W and Nb) on the electronic structure of anatase TiO2in the framework of density functional theory (DFT) The dopant-induced electronic structure modifications are believed to directly change the photovoltaic (PV) be-haviors of the X-doped TiO2based DSSCs The dopants are shown to either directly inhibit the intrinsic
Ti3þand oxygen vacancy surface defects of TiO2or enhance these defects depending on their valence states These dopant-induced defect modifications, in turn, strongly affect the PV behaviors of the DSSCs The combined effect of electronic structure and surface-defect modifications determined the photo-electric efficiency of the device
Ó 2013 Elsevier B.V All rights reserved
1 Introduction
Titanium dioxides, TiO2, have been a productive plot for
appli-cation researches due to their stability, high refractive index, strong
UV light absorbing capability and photo-activities Recently,
numerous works have been devoted to TiO2 films with anatase
nanocrystalline structures as working electrodes in dye-sensitized
solar cells (DSSCs) Although many metallic oxides have been
tested for DSSC electrode[1e5], efforts on improving DSSCs’
per-formance were focused on TiO2anatase which leads to the highest
efficiency w7.4%[6] However, this is still lower than that of
con-ventional silicon solar cells Improving DSSC efficiency is still a
major ongoingfield of research
Doping the TiO2electrode with metallic elements is a useful way
to improve the photovoltaic (PV) performance of DSSCs The
Nb-dopant was shown to raise up the short-circuit current JSCof the Nb-doped TiO2based DSSCs owing to the dopant-induced up-shift
of the Fermi level and the lowering of the conduction band (CB) bottom as well as the formation of the intra-band states[7,8] The Fermi level up-shift and the formation of the intra-band states in-crease the electron diffusion coefficient while the lowering of the
CB bottom improves the electron injection possibility The best cell
efficiency, h ¼ 8.0%, was obtained for 2.5% Nb doping (for the undoped case,h¼ 6.8%) The similar effects were observed for the Ta-doped case in which a PV efficiencyh¼ 8.18% was achieved owing to both the enhancement of electron transport driving force and the doping-induced increase of the electron concentration[9]
It was reported that Al and W dopants modified the electrical surface state of the Al and W doped TiO2anatase nanoparticles, which significantly changes the powder aggregation, charge-transfer kinetics, dye absorption characteristics and indirectly af-fects the quality of DSSCs [10] The Al-induced reduction of the surface Ti3þ defect concentration enhanced the dye absorption
* Corresponding author Tel./fax: þ84 4 3 5583980.
E-mail address: trangnguyenphys@gmail.com (T.T Nguyen).
Contents lists available atScienceDirect Materials Chemistry and Physics
j o u r n a l h o m e p a g e : w w w e l s e v i e r c o m / l o c a t e / m a t c h e m p h y s
0254-0584/$ e see front matter Ó 2013 Elsevier B.V All rights reserved.
http://dx.doi.org/10.1016/j.matchemphys.2013.12.025
Materials Chemistry and Physics xxx (2014) 1e8
Trang 2X ¼ Be, Mg, Ca, Zn, W, Al and Nb in the framework of density
functional theory (DFT) Our results suggest that the PV behaviors
of metallic doped-TiO2electrodes in previous studies resulted from
the combined effect of electronic structure and surface-defect
modifications induced by dopants We also distinguished the
sur-face defect Ti3þformed with oxygen vacancy (Ti3þ/OV) from that
one without oxygen vacancy The differences between them led to
the opposite changes in the VOCof the DSSC when the TiO2
elec-trode was doped with W and Nb at low concentration (2%) despite
of the similarity in valence
2 Calculation details
The equations for ground states of investigated systems were
built up and solved in the DFT framework (these equations are
colloquially called KohneSham equations) [12] with the Dmol3
package[13] DFT methods deal with the most complicated
po-tential term, the electroneelectron interaction by splitting it into
two separated parts, i.e the classical Coulomb potential and the
exchange-correlation potential In our calculations, the former was
evaluated by solving the Poisson equations for charge density in a
completely numerical approach [13] The later was formulated
within the local density approximation (LDA) which is based on the
well-known exchange-correlation energy of the uniform electron
gas by J.P Perdew and Y Wang and parametrized by Ceperley and
Alder as a functional of electron densityePWC functional[14] All
electrons of the examined systems, i.e core, semi-core and valence
electrons, were treated equally in building up electroneelectron
interaction potential which is so-called all electron potential (AE)
In order to solve KohneSham equations, a self-consistent field (SCF)
procedure is provided by the Dmol3package because of the mutual
dependence between electron density and electroneelectron
interaction potential[13] The initial wave function was produced
by linear combination atomic orbital (LCAO) method with the
nu-merical atomic-like basis set named DNP[13] Within this basis set,
each occupied atomic orbital is represented by one numerical
atomic-like wave function A second set of wave-functions is added
for the valence orbitals The polarization d-functions are also
added The orbital cut off was 5.2 A and the energy convergence
criterion was 106eV/atom
Examined systems were the bulk and the surface of 6.25%
X-doped anatase TiO2compounds with X¼ Be, Mg, Ca, Zn, Al, W and
Nb In order to simulate these systems, we started from the unit cell
of anatase TiO2 crystal which belongs to I41/amd space group
(Fig 1a) It should be noted that in anatase crystalline form, the
titanium atoms are put within the octahedral ligandfield of its six
neighboring oxygen atoms We labeled the corner oxygen atoms of
the octahedron O(1) and apical oxygen atoms O(2) A super-cell of
the size 2 2 1 I41/amd unit cells with space group P1 was built to
3.1 Undoped TiO2anatase Electronic structure of un-doped anatase TiO2has been inten-sively studied in the density functional theory (DFT) framework
[16e20] A wide range of the DFT methods spanning from tradi-tional one-electron local density approximations (local density approximatione LDA and generalized gradient approximation e GGA)[16,17,19,20], hybrid functional (B3LYP)[18] to many-body corrected DFT method GW[19]has been applied for TiO2anatase using various types of basis sets including Gaussian type orbital (GTO, 6-31G) [16,18], augmented plane wave (APW)[17], plane wave (PW) with pseudopotential (PP)[19,20]and all-electron po-tential (AE)[16e18] The most common problem of conventional DFT methods (LDA and GGA) is the band-gap underestimation The band-gap values of TiO2anatase extracted from LDA and GGA cal-culations are about 2 eV[16,17,19,20] Our calculation indicates a band-gap of 2.04 eV, which well agrees with the previous results The exact exchange functional correction of the hybrid functional B3LYP method and the many-body correction of the GW method, otherwise,w15% overestimated the band-gap value (Egcalw3.7 eV while Egexpw3.2 eV)[18,19] Another band gap problem for TiO2is the nature of the band gap, i.e whether it is direct or indirect and where it occurs in the k-space Although all calculations agreed that the global minimum of the conduction band (GMCB) is at G point
k ¼ [0 0 0], the global maximum of the valence band (GMVB) po-sitions from different methods are various, i.e.Gk ¼ [000],M k ¼ [½
½ 0] or near X pointk ¼ [0 0.44 0] point Our result shows the GMVB near the M pointk ¼ 5/6[½ ½ 0] so that the band gap is indirect (seeFig 2a) Unfortunately, we could notfind any experi-mental evidence to clarify this confusion On the other hand, the position of GMVB is sensitive to bond-lengths and lattice parame-ters The divergence of the calculation results may be due to the disagreement in optimizing the lattice structure
Fig 2presents the band structure, the density of states (DOS) diagram and some orbitals at theGpoint located near the top of VB and the bottom of CB The k-path was chosen with respect to the I41/amd space group of which the reciprocal space (k-space) con-tains the following high symmetry points: G pointk ¼ [0 0 0], X pointk ¼ [0 ½ 0], Z point k ¼ [0 ½ 0], M point k ¼ [½ ½ 0], R point
k ¼ [0 ½ ½] and A point k ¼ [½ ½ ½] The energy off-set is at the Fermi level denoted by the horizontal dash line inFig 2a and b The band structure from our calculation is in good agreement with that one produced by R Asahi et al on the base of their LDA/AE/LAPW calculation [17] We can see that an oxygen 2p-like-band was observed at about 16.3 eV below the Fermi level within a narrow energy rangew 1.9 eV The valence band (VB) spans in an energy range ofw 5 eV which agrees quite well with the X-ray photo-emission spectroscopy measurement (XPS)[21] The CB and VB are
Trang 3formed by the overlap between O 2p, Ti 3d and Ti 4s orbitals.
Because of the strong overlap between O 2p and Ti 4s orbitals, the O
2peTi 4s bonding states are located at the bottom of the VB and the
anti-bonding counterpart should be at the higher energy region of
CB which falls out of the our calculated energy region The small contribution of Ti 4s orbital to the Ti 4seO 2p bonding states ex-hibits the ionicity of the electron Ti 4seO 2p transfer The bondinge anti-bonding interaction between Ti 3d and O 2p orbitals seems to
Fig 1 Unit cells used in our calculations (a) The I4 1 /amd unit cell of anatase crystalline titanium dioxide; (b) The P1 unit cell of 6.25%-X-doped TiO 2 , X 0.0625 Ti 0.9375 O 2 , where X ¼ Be,
Al, Nb, Mg, Zn and Ca; (c) The P1 vacuum-slab supercell to model the (101) surface.
Fig 2 (a) The energy band structure and (b) partial density of states (PDOS) of undoped TiO 2 anatase crystal from our calculations with respect to the I4 1 /amd unit cell The high symmetry points in the k-space corresponding to this unit cell areGpoint k ¼ [0 0 0], X point k ¼ [0 ½ 0], Z point k ¼ [0 ½ 0], M point k ¼ [½ ½ 0], R point k ¼ [0 ½ ½] and A point
k ¼ [½ ½ ½] The horizontal dash line denotes the Fermi level which is also the energy off-set (b) Some orbitals around HOMO and LUMO atGpoint k ¼ (0 0 0) including HOMO, LUMO, LUMO þ 3 and LUMO þ 5.
T.T Nguyen et al / Materials Chemistry and Physics xxx (2014) 1e8 3
Trang 4The octahedron bond-lengths at impurity site, i.e XeO(1,2) bonds,
increase with respect to the increase of the impurity ionic radius
The bond-lengths of surrounding TieO6 octahedrons slightly
change, except for the cases of Ca, W and Al doping in which the
bond-lengths suddenly increase regardless of the impurity ionic
radius Therefore, the lattice constants alter similarly to the change
in ionic radius of impurity in the cases of Be, Mg, Zn and Nb doping
but strongly increase in the cases of Ca, W and Al doping In order to
get a deep insight into the anomalous changes of TieO
bond-lengths induced by Ca, W and Al dopants, electronic structure
as-pects should be involved as discussed below
Figs 4and5show the band structures and the partial DOS of the
un-doped and doped TiO2 anatase compounds with a k-path
through high symmetry points of P1 space group, i.e G pointk ¼ [0
0 0], F pointk ¼ [0 ½ 0], Q point k ¼ [0 ½ ½] and Z point k ¼ [0 0 ½]
It is noticed that the higher part of CB was not calculated with the
P1 unit cell used in case of doped compounds The energy off-set
was put at the Fermi level There are several important points
inferred from these energy band structures:
- All of our dopants produce no separated impurity bands
occurring in the VB-CB band gap They are quite different from
transition metal dopants which introduces 3d impurity states
into the VB-CB band gap of anatase crystalline TiO2[22]
How-ever, it is worth-while to note that there are some
impurity-induced bands occurring in the other band gaps in the case of
Ca, Al and W doping They are three pure Ca 3p bands located at
0.68 eV below the O 2s band and six impurity-surrounding
oxygen 2s bands located 0.27 eV above the O 2s band of the
host lattice in the case of Ca doping; one Al 3seO 2p bonding
band situated 0.35 eV below the VB in the case of Al doping and
five W 5deO 2p bonding bands locates at 0.49 eV below the VB
in the case of W doping We suggest that these additional bands put a strong Coulomb repulsion on VB which enhances Coulomb repulsion of octahedral ligandfield on Ti Consequently, the Tie
O bond-lengths are anomalously increased in the cases of Ca, Al and W doping In particular, the appearance of the 3p Ca bands right below the O 2s band of the host lattice also puts the strong Coulomb repulsion on the 2s states of oxygen atoms sur-rounding impurity sites so that they are up-shifted, resulting in
6 impurity-induced O 2s bands above the main O 2s band This interaction is also observed in CaO, where the O 2s semi-core band is slightly shifted up (w1 eV) in comparison with other alkaline earth oxide (MgO, SrO and BaO) due to the Ca 3peO 2s Coulomb interaction[23]
- The contribution of alkaline earth (AE) metallic impurities (X¼ Be, Mg, Ca) to the VB is so small that the nature of XeO bonds can be considered to be strong ionic The AE oxides are well-known as typical ionic crystals with wide band gaps ranging from 5 to 10 eV (10 eV for BeO, 7.8 eV for MgO and 6.9 eV for CaO)[24,25] Their VBs are primarily composed of O 2p states and their CBs are s states of AE metal Therefore, the wide band gaps are corresponding to the large energy difference between
O 2p and AE metallic s states We suggest that the absence of AE compositions in the VB and the strong ionic nature of XeO bonds are due to the large energy difference between O 2p and
Table 1
A summarization of lattice parameters and electronic structure information of undoped and doped anatase TiO 2
Lattice parameters
Octahedral bond-lengths
Ionic radius of impurity ( A) 0.605 (Ti4þ) 0.45 (Be2þ) 0.72 (Mg2þ) 1.00 (Ca2þ) 0.74 (Zn2þ) 0.62 (W5þ) 0.535 (Al3þ) 0.64 (Nb5þ) Electronic structure information
Fig 3 The variation of the lattice constants a, b and c and TieO(1), TieO(2), XeO(1) and XeO(2) bond-lengths upon the ionic radius of the atom X at the doped site in the P1 unit cell shown in Fig 1 b with X ¼ Be, Al, Ti, W, Nb, Mg, Zn and Ca (ascending order
in ionic radius).
Trang 5AE metallic s states The contributions of the AE s states will
appear in the higher part of the CB (w 4 eV above GMVB or even
higher, primarily depending on the O 2pemetallic s gap) which
is not included in our calculations The same situation is
observed in Al doping case even though Al2O3were shown to be
less ionic than AE oxides with a large band gap of 8.8 eV[26]
- The contributions of impurities to CB and VB become more
significant in the case of Zn, W and Nb doping Those
contri-butions primarily come from 3d states of Zn, 5d states of W and
4d states of Nb It should be noted that the Zn 3d states contribute only in the VB while the W 5d and Nb 4d contribu-tions appear both in VB and CB This difference originates from the fact that the W 5d and Nb 4d electrons take part in crystal binding as valence electrons but Zn 3d electrons do not (semi-core electrons) Therefore, the contributions of W 5d and Nb 4d states to the VB and the CB correspond to the formation of the bonding and anti-bonding states caused by the overlap between the metallic d orbitals and the O 2p orbitals Meanwhile, it was shown that the VB-CB energy gaps of zinc oxides result from the bondingeantibonding interaction between O 2p and Zn 4s states[27] This interaction energetically pushes down the O 2p dominated bonding states and pushes up the Zn 4s dominated antibonding states The lowered O 2p states energetically reach
to the Zn 3d states so that they can hybrid with the Zn 3d states Consequently, Zn 3d-like band is broadened and joins the VB Although the p-d hybridization was shown to be weakened when the ligandfield transforms from tetrahedron to octahe-dron (corresponding to the crystal structure transition of ZnO from Zinc-blende or wurtzit to rock-salt)[28], we still imply a
p-d hybrip-dization for our Zn-p-doping case in which Zn impurity atoms are put inside octahedral ligandfield of TiO2anatase host lattice
- As expected from the oxidization states of impurity elements, the dopants Be (II), Mg (II), Ca (II), Zn (II) and Al (III) shift the Fermi level downwards to overlap the top of the VB which is mainly composed of oxygen 2p states while the dopants W (V) and Nb (V) shift the Fermi level upwards to overlap the bottom
of the CB By this way, holes corresponding to acceptors are formed and located on the oxygen 2p like orbitals in the former
Fig 4 The band structures of undoped and 6.25%-X-doped anatase TiO 2 with X ¼ Be, Al, W, Nb, Mg, Zn and Ca The energy off-set was put at Fermi level which was denoted by the red dash lines (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig 5 The DOS of dopant within the host lattice of anatase TiO 2 : (a) Be-DOS, (b)
Al-DOS, (c) W-Al-DOS, (d) Nb-Al-DOS, (e) Mg-Al-DOS, (f) Zn-DOS and (g) Ca-DOS The energy
off-set was put at Fermi level which was denoted by the vertical dash lines.
T.T Nguyen et al / Materials Chemistry and Physics xxx (2014) 1e8 5
Trang 6so the VB is shrunken Meanwhile, the d-p anti-bonding states,
which is the main composition of the CB, is down shifted, hence,
CBeVB gap is narrower This is the case of Ca, W and Al doped
compounds Besides, the large expansion of TiO6octahedron in
Ca, W and Al doped TiO2, corresponding to the extension of Tie
Ti distance, reduces the 3dxye 3dxyoverlap, hence reduces the
3dxye 3dxyhopping elements between Ti sites Accordingly, the
non-bonding bands Ti 3dxysituated at the bottom of the CB
become less dispersive L Thulin and J Guerra proposed to apply
a strain to modify the band gap and carrier effective mass of
anatase TiO2, which agrees with our observations of the
bond-lengtheband gapeband dispersion relation [19] Other
impu-rities (Be, Mg, Zn and Nb) induce a slight increase of the band
gap (<5%), a reductions of 3dxyband dispersion (<7%) except
expansion (<12%) In particular, the dopant-induced band gap
observed from the optical absorption measurements, i.e the
band gap increase w5% at 3% doping concentration [8] Our
calculations reasonably agree with the experimental result
which shows a band gap enlargement of 4% for 6.25% Nb-doped
compound Since these impurities do not noticeably affect the
TieO bond-lengths, it is believe that neither ligand field nor Tie
O bondingeantibonding interaction modifications play any
significant role in these effects We propose here various reasons
for the changes of the VB width For Zn, the p-d hybridization
mentioned above expands the VB to the negative energy For the
others, it seems that the strong Be 2se O 2p e Ti 3d and Nb 4d e
O 2pe Ti 3d overlaps add bonding states to the bottom of the
VB, which extend the VB of Be and Nb doped material The triple
bondingeantibonding interaction in the case of Mg doping was
weaker so that the VB expansion is quite small (w0.02 eV)
3.3 The dopant-induced electronic structure modification effects on
the PV behavior of DSSC
In order to understand the effect of doping on energy
trans-formation efficiency of the DSSC, it is necessary to remind that the
power conversion efficiency h of a PV cell is estimated by the
formula:
h ¼ FFJSCVOC
where FF is thefill factor, JSCis the short circuit current, VOCis the
open circuit voltage of the solar cell and Pinputis the power of the
incident light For the DSSC, the JSCstrongly depends on the
elec-tron diffusion coefficient and the injection possibility of
photo-excited electrons from the highest occupied molecular orbital
band contraction caused by Be (0.05 eV), Mg (0.07 eV), Ca (0.18 eV),
Zn (0.06 eV), Al (0.22 eV) and W (0.25 eV) doping raises the CB bottom Thus, the electron photo-injection and photocurrent are improved by Nb dopant but suppressed by Be, Mg, Ca, Zn, Al and W dopants Especially, Ca, Al and W dopants definitely inhibit the injection possibility due to the strong contraction of 3dxyband The practical situations are much more complicated Since the electron transport oxide layers are usually nanocrystalline, the roles of electron recombination at the nanoparticle/electrolyte and nanoparticle/dye interface and surface defects become extremely important The surface Ti3þ defects are believed to mediate the electron recombination at the interfaces, hence, form the dark current IDC which significantly reduces the VOC of the nano-crystalline electrode DSSCs The surface-induced band tail states within the band gap also play the primary role in the electron transport via multi-trapped processes (the band tail states are so-called transport levels)[29] Otherwise, the dopant itself induces
Ti3þdefects without the accompanying oxygen vacancies or neu-tralizes the inherent Ti3þdefects within the anatase nanocrystal-line TiO2 Moreover, the dopant not only directly modifies the electronic structure of the material but also indirectly varies the electronic structure by neutralizing or inducing surface Ti3þ de-fects, inhibiting or enhancing the formation of oxygen vacancy OV The combined effect of electronic structure and surface-defect modification causes the observed modifications in PV activity of doped electrodes In the following part, we will discuss the in-fluences of the metallic dopants on the PV behaviors via defect modifications They are considered as indirect influences
3.4 Mechanism of defect formatione the influences of metallic dopants
Oxidization state of Ti in TiO2 is Ti4þ.The Ti3þ defect can be formed by the Ti4þreduction via two typical groups of processes, i.e (i) the reductions without oxygen vacancy including the photo-induced reduction and dopant-photo-induced reduction, and (ii) the ox-ygen vacancy accompanied reductions The processes of thefirst group produce Ti3þdefects while the second ones produce Ti3þ/OV
defects
In the photo-induced reduction process, one appropriate light quantum hnof the incident irradiation to TiO2surface is absorbed to excite one electron e-at the top of the VB (composed of O 2p like states) to the bottom of the CB (composed of non-bonding Ti 3dxy
states) This excitation corresponds to the light induced charge transfer from O2to Ti4þ The photo-excited electron tends to be trapped there, reducing the surface Ti4þto defect Ti3þand leaving a hole hþon O site It is well-known that TiO2is strongly photoactive under ultraviolet (UV) irradiation in order to overcome the band gapw3.2 eV, creating Ti3þdefects which are important reactive
Trang 7agents for photo-catalytic processes [30] Therefore, the
dopant-induced changes in band gap noticeably affects the
photo-induced Ti4þeTi3þ reduction The band gap enlargement shifts
the photoactive region of the material from the UV range toward
the higher energy region while the band gap contraction shifts
toward the lower energy region Otherwise, the solar spectrum
strongly declines in the region from the UV lights to the higher
energy As a result, Ca, Al and W dopants will significantly enhance
the photo-induced Ti4þeTi3 þreduction under solar light but Mg, Zn
and Nb will suppress it Concerning the dopant-induce reduction,
the W and Nb induced partialfilled Ti 3dxyband at the bottom of
the CB corresponds to the Ti4þeTi3 þreduction.
In order to examine the influences of metallic dopants in the
surface Ti3þ/OVdefects, we calculated the formation energy ETi3 þ =O V
of the Ti3þ/OVdefect on the (101) anatase TiO2surface using the
slab model inFig 1c
ETi3þ =O V ¼ EsurfþTi3þ =O Vþ EO Esurf
Where Esurfis the total energy of the slab, EOis the energy of free
oxygen atom and EsurfþTi3þ =O Vis the total energy of the slab with an
oxygen vacancy in the surface The formation energies are
sum-marized inTable 2and schematically compared inFig 6 The W and
Nb dopants seem to inhibit the formation of surface Ti3þ/OVdefects
due to the increase of formation energy while the others exposes
reversed effects
3.5 Combined effect of electronic structure and surface-defect
modification in practical doped TiO2electrode
Firstly, we make a comparison between ourfirst principle
ex-aminations on the PV behaviors of the metallic-doped electrodes
with available experimental observations for Al, W and Nb doping
case On the base of X-ray photoemission spectroscopy (XPS), T
Nikolay et al reported negative shifts of the Fermi level and a
reduction of the free electron concentration (corresponding to Ti3þ
concentration) for 0.5% and 1.5% Nb-doped TiO2 electrodes and
positive shifts of the Fermi level for higher concentrations 2.5% and
3% with an increase of Ti3þconcentration[8] The negative shifts of
Fermi level at low concentrations are unexpected as shown by our
calculations It should be noted that the creation of Ti3þ/Ovand Ti3þ
defects positively shifts the Fermi level Our results suggest three
processes of Ti3þ/Ovand Ti3þ defect modifications in Nb doped
anatase nanocrystalline TiO2, which are:
(i) The inhibition of the formation of surface Ti3þ/OVdefects by
Nb dopants negatively shifts the Fermi level
(ii) The Nb-induced band gap enlargement suppresses the
photo-induced surface Ti4þeTi3 þdefects, and hence,
nega-tively shifts the Fermi level
(iii) The Nb dopant itself reduces Ti4þto Ti3þ, positively shifting
the Fermi level
At low dopant concentrations, the combination of processes (i)
and (ii) dominates the process (iii), negatively shifting the Fermi
level and the reducing of Ti3þconcentration At high doping
con-centrations, when the effects of the intrinsic Ti3þand Ti3þ/OVdefect
inhibition is saturated, the process (iii) dominates, raising the Fermi
level up and increasing the Ti3þconcentration Concerning the VOC, the measured value was changed in the discordance with the Fermi level shifts In order to interpret these contradictory observations, the dark current IDCshould be involved At low concentrations 0.5 and 1.5%, the lowered surface Ti3þconcentration reduces IDC, hence, increases the VOCwhile the negative-shift of EFtends to reduce the
VOC With these dopant concentrations, the domination of dark current effect produces the enhanced VOC The opposite effect oc-curs for higher concentrations The experimental observations agree quite well with our calculation that the lowering of the CB bottom improves the electron injection and increases the JSCat all studied Nb concentrations
Despite of the similarity in valence between W and Nb (V), ac-cording to our calculation, W dopant strongly narrows the band gap
of TiO2 Hence, it should enhance the photo-induced Ti4þeTi3 þ
reduction on the surface of the nanoparticle, i.e the process (ii) of
W case is in the opposite trend to that one of Nb case Consequently, the process (i) is dominated by the combination of processes (ii) and (iii) so that the Ti3þconcentration should be increased in the case of W doping even at low concentration This is in good agreement with K H Ho et al observation based on XPS mea-surement for 1.6% W-doped anatase TiO2[10] The VOCis decreased but the ISCis increased by W dopant at concentrations from 0.1 to 2.0% because of the strong domination of the defect effect with enhanced transport level density[10,11]
In the case of Al doping, the processes (i) and (iii) occur in the opposite trend with the Nb and W cases and the processes (ii) is in the same trend with W doping case Consequently, the process (iii)
is dominated by the combination of the two others, increasing the
Ti3þ concentration, positively shifting the Fermi level This is in contrast with the XPS observation in Ref [10] which showed a decrease of Ti3þconcentration in 3.3% at Al-doped TiO2synthesized via hydrolysis but agrees with that one on samples prepared via a single-step direct combination of vaporized Ti, Al and O2using a
6 kW thermal plasma system [31] In the later experiment, the observation of the increased Ti3þ/OVconcentration and the nar-rowed optical band gap due to the substitution of Al3þfor Ti4þ matched quite well with our calculation results In order to clarify the disagreement between the two XPS measurements, it should be noted that there are some possible positions for dopant atom when doped into TiO2 nanoparticle, i.e interstitial site of TiO2, near-surface substitutional site and deep-in-bulk substitutional site
On the base of lattice deformation and photocatalytic activity
Table 2
Summarization of formation energies of oxygen vacancy on the (101) anatase TiO 2
surface.
ETi3þ =O V (eV) 6.55 0.69 0.51 0.43 0.79 3.64 6.99 6.80
Fig 6 Formation energies of Ti3þ/O V surface defect on the (101) anatase TiO 2 surface with various X atoms at the doped site in the P1 vacuum-slab supercell shown in Fig 1 c (X ¼ Be, Al, Ti, W, Nb, Mg, Zn and Ca) The horizontal dot line denotes the formation energies of the Ti3þ/O V surface defect on the undoped surface (X ¼ Ti).
T.T Nguyen et al / Materials Chemistry and Physics xxx (2014) 1e8 7
Trang 8reducing of defect concentration or by the up shift of CB bottom as
shown by our results
4 Conclusions
In conclusion, the effects of X dopants (X¼ Be, Mg, Ca, Zn, Al, W
and Nb) on the electronic structure and surface defects of anatase
TiO2were systematically examined The impurity bands occurring
within the below-VB energy gaps in case of Ca, Al and W strongly
extend the TieO bond, narrow the band gap and Ti 3dxy band
width The donor and acceptor states locate on oxygen (acceptor
states) and titanium (donor states) sites of the host anatase TiO2
lattice, indicating that the quinquevalence dopants Nb and W
directly reduce Ti4þ to Ti3þ without forming OV and the
lower-valence dopants Be, Mg, Ca, Zn and Al induce OVwithout Ti3þ
The calculated formation energies of oxygen vacancy in the doped
and un-doped (101) surfaces expose that the W and Nb dopants
inhibit the formation of oxygen vacancy while the others do the
reversed manner
The PV behaviors of the metallic doped-TiO2based DSSCs are
determined by the combined effect of the dopant-induced
elec-tronic structure and defect modification According to this, the
previous experimental observations on the influences of Nb, W and
Al dopants on DSSC were well-explained The divergence between
experimental observations and our calculation results in the case of
Al dopant suggests an additional factor to determine the PV ef
fi-ciency of the doped TiO2electrode It is the position of the dopant
ion within the host lattice which is possibly the interstitial,
[7] P.S Archana, R Jose, M.M Yusoff, S Ramakrishna, Appl Phys Lett 98 (2011)
152106 [8] T Nikolay, L Larina, O Shevaleevskiy, B.T Ahn, Energy Environ Sci 4 (2011)
1480 [9] J Liu, H Yang, W Tan, X Zhou, Y Lin, Electrochim Acta 56 (2010) 396 [10] K.H Ko, Y.C Lee, Y.J Jung, J Colloid Interface Sci 283 (2005) 482 [11] X Zhang, S.T Wang, Z.S Wang, Appl Phys Lett 99 (2011) 113503 [12] W Kohn, L.J Sham, Phys Rev A 140 (1965) 1133
[13] B Delley, J Chem Phys 92 (1990) 508 [14] J.P Perdew, Y Wang, Phys Rev B 45 (1992) 13244 [15] O Ermer, Struct Bonding 27 (1976) 161
[16] S.D Mo, W.Y Ching, Phys Rev B 51 (1995) 13023 [17] R Asahi, Y Taga, W Mannstadt, A.J Freeman, Phys Rev B 61 (2000) 7459 [18] M Calatayud, P Mori-Sánchez, A Beltrán, A Martín Pendás, E Francisco,
J Andrés, J.M Recio, Phys Rev B 64 (2001) 184113 [19] L Thulin, J Guerra, Phys Rev B 77 (2008) 195112 [20] W.J Yin, S Chen, J.H Yang, X.G Gong, Y Yan, S.H Wei, Appl Phys Lett 96 (2010) 221901
[21] R Sanjinés, H Tang, H Berger, F Gozzo, G Margaritondo, F Lévy, J Appl Phys.
75 (1994) 2945 [22] T Umebayashi, T Yamaki, H Itoh, K Asai, J Phys Chem Solids 63 (1909)
2002 [23] B Baumeier, P Krüger, J Pollmann, Phys Rev B 76 (2007) 205404 [24] A.S Rao, R.T Keamey, Phys Status Solidi B 95 (1979) 243 [25] K.J Chang, S Froyen, M Cohen, J Phys C Solid State Phys 16 (1983) 3475 [26] J Koo, S Kim, S Jeon, H Jeon, Y Kim, Y Won, J Korean Phys Soc 48 (2006) 131
[27] P.D.C King, T.D Veal, A Schleife, J Zúñiga-Pérez, B Martel, P.H Jefferson,
F Fuchs, V Muñoz-Sanjosé, F Bechstedt, C.F McConville1, Phys Rev B 79 (2009) 205205
[28] J.E Jaffe, R Pandey, A.B Kunz, Phys Rev B 43 (1991) [29] D Monroe, Phys Rev Lett 54 (1985) 146
[30] A Si risuk, E Klansorn, P Praserthdam, Catal Commun 9 (2008) 1810 [31] C.-Y Tsai, T.-H Kuo, H.-C His, Int J Photoenergy (2012), http://dx.doi.org/ 10.1155/2012/874509
[32] Y Li, S Feng, F Jiang, G Lu, S Li, J Serb Chem Soc 72 (4) (2007) 393