ARTICLE pubs.acs.org/JPCC
Origin of the Enhanced Visible Photocatalytic Activity in (N, C)-Codoped ZnS Studied from Density Functional Theory Honggang Sun,† Xian Zhao,† Liang Zhang,§ and Weiliu Fan*,†,‡ †
State Key Laboratory of Crystal Materials, Shandong University, Jinan 250100, China Department of Chemistry and Chemical Engineering, Shandong University, Jinan, 250100, China § State Key Laboratory of Molecular Reaction Dynamics, Dalian Institute of Chemical Physics of CAS, Dalian, 116023, China ‡
bS Supporting Information ABSTRACT: Density functional theory calculations are used to investigate the origin of the experimentally observed changes in visible photoactivity of hexagonal wurtzite ZnS induced by (N, C) codoping. The accurate comparative analyses of geometric and electronic structures for the different doping models have been discussed. For the mono N- or C-doped ZnS systems, the substituted doping just induces a slight band gap narrowing (less than 0.2 eV), while the interstitial doping leads to more significant optical absorption red shift. However, the calculated energy indicates the formation of the interstitial doping is difficult with the lower impurity concentration. For the NþC-codoped systems, the electron transition from the impurity states in the gap to the conduction band will induce a large red shift, suggesting the visible-light absorption. However, the impurity states are partially occupied character, which may act as recombination center and reduce the photoinduced current density. The calculated results of 2NþC-codoped ZnS indicate the N-C-N trimer structure was formed in the ZnS lattice, and the trimer doping induces a larger red shift of photoelectron transition and passivates the partially occupied states by the charge compensation effect in acceptor-donor-acceptor pair. Our work provides a solid basis for the rationalization of experimentally observed red shift of optical absorption in wurtzite ZnS as a consequence of (N, C) codoping and shows that 2NþC codoping will be a promising way for improving the visible-light activity of semiconductor photocatalysts.
1. INTRODUCTION Environmental problems and energy crises provide the impetus for sustained fundamental and applied research in the area of environmental remediation and renewable resource producing. Semiconductor photocatalysis offers the potential for degradation of organic pollutants and shifting of water to hydrogen energy through its efficiency and potentially broad applicability.1-5 To obtain more efficient and advantageous photocatalysts, various new compounds and materials for photocatalysis have been synthesized in the past few decades. Sulfide semiconductors such as ZnS6-14 have been reported as important semiconductor photocatalysts to remove organic pollutants and a highly efficient photocatalyst for water splitting even without any metal particles as cocatalyst due to its excellent properties of the rapid generation of electron-hole pairs and the highly negative reduction potentials of excited electrons.13 Yet, the spectral response range of ZnS is limited only to wavelengths under ultraviolet irradiation owing to the wide band gap energy of 3.7 eV.12,15 It has been envisaged that ZnS doped with transition metals such as Ni- or Cu-doped ZnS photocatalysts show a high activity for H2 evolution under visible-light irradiation even without a noble metal cocatalyst.6,7 However, semiconductor photocatalysis of transition metal doping has some problems for r 2011 American Chemical Society
transition metals doping such as thermal instability and an increase of charge carrier recombination centers; thus the application is limited for efficient photocatalytic reactions. Doping of nonmetal anions such as N, C, and F in the semiconductors has been proven to be very promising for efficient photocatalytic oxidation of organic compounds under visible-light radiation. Yet, experimental studies show that, for this kind of monodoping, the photogenerated current is low because the partially occupied impurity bands can act as recombination centers and reduce the photogenerated current. A possible way to improve the photocatalytic performance of doped semiconductors is to exploit the cooperative effect of introducing different dopants into the matrix. Di Valentin et al. using density fuctional theory (DFT) calculations have inferred that (N, F) codoping could enhance the photocatalytic activity of TiO2 under visible-light irradiation,16 as a consequence of the charge compensation between the nitrogen and the fluorine impurities. Theoretical studies by Gai et al. indicate that (C, Mo) passivated codopants in TiO2 would induce passivated defect bands and an improved visible-light Received: October 27, 2010 Revised: December 25, 2010 Published: January 13, 2011 2218
dx.doi.org/10.1021/jp110263e | J. Phys. Chem. C 2011, 115, 2218–2227
The Journal of Physical Chemistry C
ARTICLE
Figure 1. The optimized structures of (a) pure, (b) Ns-doped, (c) Cs-doped, (d) Ni-doped, and (e) Ci-doped ZnS. The yellow, gray, blue, red spheres represent sulfur, zinc, nitrogen, and carbon, respectively.
activity for photoelectrochemical water splitting.17 Therefore, a codopants pair for semiconductor may be a potential mean to guarantee a successful efficient visible photocatalysts. However, there are still some difficulties for the codopants. The experiment by Du et al. have confirmed the calculations of Di Valentin et al., which synthesize a good photocatalytic activity under visible light for (N, F)-codoped TiO2.18 Yet, in fact, the main effect of fluorine doping has been proved on the surface fluorination by many other experiments,19-22 which influences the adsorption of the hydroxyl on the semiconductor’s surface. Although the experiment by Zhang et al. achieved a better photocurrent of the CþMo-codoped TiO2, the optical absorption edge is only 32 nm red shift (the vis-light response is still weak).23 Thus, the choice of the codopant pair should be carefully analyzed and evaluated to get a better effect for the photocatalysis. Recently, Muruganandham et al. achieved a favorable visiblelight-responsive ZnS photocatalyst using (N, C) codoping.11 Previously, the codoping of carbon and nitrogen was also used to improve the photocatalytic activity of TiO2 by some experiments.24,25 These results show (N, C) codoping may be good codopants for improving photocatalytic activity. However, similar to the above-mentioned issue, the N or C doping will induce some partially filled impurity states, which is disadvantageous to the photogenerated current density. Yamamoto has noted the substituted N impurity for ZnS may induce some acceptor level in the valence band.26 Theoretical calculations by Yang et al. indicate that some partially occupied states are formed when C doping substitutes the lattice S atom of ZnS.27 Therefore, the form of (N, C) codoping in ZnS need a careful analysis to avoid the partially filled states. In addition, though Muruganandham et al. gave some speculation based on their experimentally observation, the detailed mechanisms of the band gap narrowing in (N, C)-codoped ZnS are still unclear. To the best of our knowledge, there was no reported theoretical work focusing on the synergistic effects of (N, C) codoping in ZnS. It is well-known that the photocatalytic process is strongly associated with the electronic structures of the semiconductor. So, the electronic structure of a photocatalyst plays a crucial role in determining the photocatalytic activity. Therefore, it would be necessary to relate the quantum calculations to experimental results and further reveal the origin
Table 1. Formation Energy for N- and/or C-Doped ZnS Configurations species
Ef (eV)
Ns
2.29
Cs
10.68
Ni
6.63
Ci NsþCs
13.60 8.74
NsþCi
7.97
2NsþCi
4.51
2CsþNi
12.84
of the experimentally observed better visible-light photocatalytic activity in (N, C)-codoped ZnS. In the present paper, to obtain microscopic insight to the effect of (N, C) codoping on the visible photocatalytic activity of ZnS, we have fabricated appropriate N-doped, C-doped, and (N, C)codoped models and carried out a comparative analysis of geometric and electronic structures using DFT calculations. The results provide a possibility of acceptor-donor-acceptor codopants as a “good pair” application on implementing an efficient visible-light activity of ZnS by the synergistic effects of (N, C) codoping. It is expected that this knowledge would aid the further design and construction of new effective visible-light photocatalysts.
2. COMPUTER DETAILS AND STRUCTURE ASPECTS The calculations were performed using the CASTEP code28 based on first-principles DFT. The exchange and correlation interactions were modeled using the generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof (PBE) functional.29 According to the experiment by Muruganandham et al., the hexagonal wurtzite ZnS (space group P63mc) was used in the present work. Crystal structure and lattice parameters (a = 3.822 Å and c = 6.261 Å) taken from a previous work30 were used as input for the structural optimization. The Vanderbilt ultrasoft pseudopotential was used to deal with the core electrons and the valence atomic configurations are 3d104s2 for Zn, 3s23p4 for S, 2219
dx.doi.org/10.1021/jp110263e |J. Phys. Chem. C 2011, 115, 2218–2227
The Journal of Physical Chemistry C
ARTICLE
considered as NsþCs, NsþCi, and NiþCs, respectively. In the NsþCs-codoped structure, the two lattice S atoms were replaced by impurity atoms. In the NsþCi- and NiþCs-codoped structures, one impurity atom was set at tetrahedral or octahedral interstitial sites, and the other doped atom substituted the vertex S atoms of the tetrahedron and octahedron. For 2NsþCi- and 2CsþNi-codoped ZnS supercells, the interstitial atoms were introduced to the tetrahedral or octahedral interstitial sites and the two substituted atoms replaced the corner S atoms of the tetrahedron and octahedron. For all the doped structures, the geometrical optimization was implemented at Γ. In fact, the calculated results of geometrical optimization indicated the optimized configurations were similar for both tetrahedral and octahedral interstitial sites, and the interstitial doping of octahedral sites is energetically favorable. And a 2 2 2 k mesh was used for electronic properties calculations based on the corresponding optimized geometries. To compare the relative stability of the N and/or C doped systems, we calculated the formation energies of the doped ZnS supercells according to the following formulas Ef ¼ Etot ðdefectÞ - Etot ðpureÞ - nN μN - nC μC - nS μS Here Etot(defect) is the total energy of the supercell with doping and Etot(pure) is the total energy of the ideal supercell. nN and nC are the number of N and C doping in the lattice of ZnS. nS is the number of S atoms being removed from the supercell, which is one for mono substituted doping, two for double substituted doping, and zero for interstitial doping. μN, μC, and μS represent the chemical potential of the N, C, and S atoms. It should be mentioned that most of the photocatalysts are prepared through thermal decomposition or calcinations under open atmosphere. So, we assumed the chemical potential of N (μN) is equal to the energy of half a N2 molecule and the chemical potential of C (μC) and S (μS) are calculated according to CþO2fCO2 and SþO2fSO2. Under O-rich conditions Figure 2. The total electron density maps (left) and the electron difference density maps (right) for (a), (a0 ) pure, (b), (b0 ) Ns-doped, (c), (c0 ) Cs-doped, (d), (d0 ) Ni-doped, and (e), (e0 ) Ci-doped ZnS. The cross section is the (001) plane.
2s22p3 for N, and 2s22p2 for C. The planewave cut off was set to 310 eV. The k mesh of 7 7 4 for pure phase was used for geometry optimizations using the Monkhorst-Pack scheme.31 In the geometrical optimization, all forces on atoms were converged to less than 0.03 eV/Å; the maximum ionic displacement was within 0.001 Å. and the total stress tensor was reduced to the order of 0.05 GPa. The optimized lattice parameters were a = 3.862 Å and c = 6.330 Å, in good agreement with the experimental value. This result indicates that our computational approach is reasonable. To study the doped systems, we employed a 3 3 2 ZnS supercell (72 atoms) to construct doped models. Two ways were used to introduce nitrogen or carbon atoms into the lattices of ZnS. One was that a sulfur atom was replaced with an impurity atom labeled Ns or Cs, and another was that an impurity atom was set at an interstitial site noted as Ni or Ci. In the present work, both tetrahedral and octahedral interstitial sites in ZnS have been considered for the interstitial doping (see the figures in Supporting Information), while only the most stable configuration was discussed in detail. When the N/C concentration ratio is 1:1 in codoped ZnS (labeled as NþC), three configurations were
μC ¼ μCO2 - μO2
μS ¼ μSO2 - μO2
Hereinto, μO2 is fixed at the energy of an O2 molecule; μCO2 and μSO2 are equal to the energy of one CO2 and one SO2 molecule, respectively.
3. RESULTS AND DISCUSSIONS 3.1. Mono N- or C-Doped ZnS. In this section, we have investigated a variety of configurations of mono N or C located at substituted and interstitial position in the ZnS lattice. Our main aim is to study the changes of the local microstructures and the electronic structures due to the doping atom and further to discuss whether the changes can narrow the band gap to visiblelight region and improve the photocatalytic activity. A. Geometric Structures. The optimized geometric structures taken from the doped ZnS supercells are shown in Figure 1. For comparison, the structures of pure ZnS are also shown. For the pure ZnS, as shown in Figure 1a, both Zn and S atoms are in the tetrahedron coordination, and four S-Zn bond lengths are 2.366, 2.366, 2.366, and 2.372 Å. For the Ns-doped structure (Figure 1b), the optimized N-Zn bond lengths shrink to about 1.992, 1.992, 1.992, and 1.978 Å compared to the original S-Zn bonds, which should be contributed to the stronger Pauling electronegativity of N (3.04) than S (2.58). For the Cs-doped model in Figure 1c, although the Pauling electronegativity of C 2220
dx.doi.org/10.1021/jp110263e |J. Phys. Chem. C 2011, 115, 2218–2227
The Journal of Physical Chemistry C
ARTICLE
Table 2. Mulliken Charge Population on the N, C, S, and Zn Atoms for Pure Phase and Doped Systems charge (e) species
N
C
pure
bond length (Å) S
Zn
S-Zn
-0.49
0.49
2.366(3)
-0.48
0.50
N-Zn
N-S
C-Zn
C-N
C-S
2.372 Ns
-0.93
1.992(3) 1.978
-0.87
Cs
-0.49
0.50
2.052(3) 1.977
Ni
-0.87
Ci NsþCs
-0.94
-0.70 -0.63
0.11/-0.49
0.49
-0.01/-0.49 -0.48
0.50 0.51
1.623 1.663 1.939
1.985
1.994
2.011
2.006
2.014
2.012 NsþCi 2NsþCi
-0.56 -0.63,-0.66
-0.27 0.16
-0.49 -0.49
0.50 0.50
2.091
2.107
2.109
2.115
1.248
2.128
1.242
2.131 2.153
1.259
2.197 2CsþNi
-0.28
-0.41, -0.57
-0.49
0.51
2.076
1.243
2.098
1.301
2.115 2.131
(2.55) is smaller than S atom, the C-Zn bond lengths also shorten to 2.052, 2.052, 2.052, and 1.977 Å, which is mainly due to the much smaller atomic radius of C (0.77 Å) than S atom (1.02 Å). Confessedly, the atomic electronegativity (and size) strictly defined only for atoms participating in a covalent bond, and here we have ionic character for S-Zn, N-Zn, and C-Zn bonds, but we can find some useful information from a qualitative analysis. For the Ni- or Ci-doped ZnS models, the impurity atom is inclined to bond with adjacent lattice S atom after geometrical optimization. The N-S bond length for Ni-doped ZnS is 1.623 Å slightly shorter than the C-S bond length for Ci-doped ZnS (1.663 Å), which is mainly due to the stronger electronegativity of N than C. The calculated values for the formation energy of the doped ZnS were shown in Table 1. The results suggest that N or C prefers to occupy the S site rather than interstitial site, which is agreement with the X-ray photoelectron spectrometer (XPS) result.11 Therefore, we conclude that the interstitial doping may be only for higher dopants concentrations in the experiment. To understand the charge redistribution induced by N or C doping, we calculate the electron density and electron density difference for the two doped ZnS supercells as shown in Figure 2. The maps of pure phase are also plotted for comparison. As shown in parts b and c of Figure 2, when the N or C atom substitutes the lattice S atom, after charge redistribution, the interactions of N-Zn/C-Zn bonds become stronger than the original S-Zn bonds. The density difference maps (parts b0 and c0 of Figure 2) indicate that N or C atom withdraws charges from Zn atoms. Furthermore, Mulliken charge population analyses are calculated and shown in Table 2. Confessedly, the absolute magnitudes of the atomic charges yield by the population analysis have little physical meaning, but we can find some useful information from the relative values of Mulliken population. The charges on N and
C ions are -0.93 and -0.87 e capturing from the adjacent Zn cations, respectively, which are much more than that on the S ions (-0.49 e) of the pure phase. Thus, we contribute the stronger N-Zn/C-Zn bonds to the more charge transfer from adjacent Zn. For the Ni- and Ci-doped structures, the total electron density maps in parts d and e of Figure 2 show the N and C doping atoms bonded with adjacent S are major covalentlike bonding interactions by common electron clouds. Parts d0 and e0 of Figure 2 show there are some charges transferring from the lattice S to the bonding impurity atoms. The total charges on N-S and C-S are -0.76 and -0.71 e, respectively, indicating that NS- and CS- species are formed in the interstitial doped ZnS. B. Electronic Structures. To investigate the changes of the electronic structures of the doped ZnS and further study the effect of these changes on determining the photocatalytic activity, the band structures and the density of states (DOS) of pure, N-, and C-doped ZnS are calculated and shown in Figures 3 and 4. For the pure phase, Figure 3a gives a band gap of about 2.11 eV. Experimentally, the pure ZnS has a large band gap of about 3.7 eV.12,20 The band gap underestimation of DFT always exists due to the well-known limitation of predicting accurate conduction band properties.32 However, it is still a widely accepted method to discuss the valence band in the electronic structure calculations, and this gives reasonable explanation for the experimental results, because only the relative positions of the occupied states and empty states need be taken into account. As shown in the Figure 4a0 , the valence band maximum (VBM) of the pure phase is predominantly contributed to by S 3p character. The conduction band minimum (CBM) basically originates from the Zn 4s states with small S 3p states. That is to say the electronic 2221
dx.doi.org/10.1021/jp110263e |J. Phys. Chem. C 2011, 115, 2218–2227
The Journal of Physical Chemistry C
ARTICLE
Figure 3. The band structures of (a) pure, (b) Ns-doped, (c) Cs-doped, (d) Ni-doped, and (e) Ci-doped ZnS. The Fermi level is set at zero energy.
transition from S 3p states to Zn 4s states is responsible for the optical absorption onset of pure ZnS. For the Ns-doped structure, the band gap has a slight decrease to about 1.98 eV (shown in Figure 3b). Some levels pass through the Fermi level, indicating that N impurity acts as an acceptor doping in ZnS. The calculated DOS and project density of states (PDOS) in parts b and b0 of Figure 4 show that the N 2p states overlap and mix with the valence band. Some mixing states extend beyond the Fermi level and thus induce the band gap narrowing about 0.13 eV. This result is different with the N-doped oxide semiconductor such as TiO2 and ZnO, which gives significant isolated N 2p impurity states in the gap.33,34 We have contributed the difference to that the N 2p orbital energy (-6.796 eV from GGA calculation) is very close to the S 3p orbital energy (-6.817 eV) but is much larger than O 2p orbital energy (-8.831 eV). Figure 3c gives the band gap of the substituted C-doped ZnS narrowing to about 1.95 eV. Likewise, the gap narrowing is attributed to the mixing of C 2p states with little S 3p states in the top of the VBM. The extension of valence bandwidth is larger than that in the N-doped structure, which should be attributed to
the higher C 2p orbital energy (-5.111 eV) inducing more obvious dispersion of C 2p states than N doping. For the Ni-doped ZnS, two impurity levels appear in the forbidden gap; one passes through the Fermi level, and the other locates below the Fermi level showing a donor character, as shown in Figure 3d. Although the host band gap has a slight increase to about 2.19 eV, the photo transition energy has a larger red shift to about 1.39 eV measured from the impurity levels in the gap. Moreover, the photo transition energy from VBM to the partially occupied level has a larger red shift to about 1.04 eV. The DOS and PDOS in parts d and d0 of Figure 4 show that the two impurity levels consist of the mixing of N 2p and S 3p states. For the interstitial C-doped ZnS (see Figure 3e), two impurity levels are also lying in the gap, and one is above the Fermi level showing unoccupied character, and the other lies below the Fermi level acting as a donor level. Also the electron transition energy from the occupied impurity states to CBM has a larger red shift to about 1.16 eV. As shown in the parts e and e0 of Figure 4, it is expected that the impurity levels are mainly originated from the C 2p and S 3p hybrid states. 2222
dx.doi.org/10.1021/jp110263e |J. Phys. Chem. C 2011, 115, 2218–2227
The Journal of Physical Chemistry C
ARTICLE
Figure 4. The total density of states (left) and the project density of states (right) for (a), (a0 ) pure, (b), (b0 ) Ns-doped, (c), (c0 ) Cs-doped, (d), (d0 ) Nidoped, and (e), (e0 ) Ci-doped ZnS. The Fermi level is set at zero energy.
Table 3. Values of the N-C Distance after Relaxation and the Relative Stable Energies (ΔE) Calculated for NsþCs-Doped Configurations |N-C|(Å) 4.301
ΔE (meV) 0
4.336
4
5.499
335
6.671
775
7.551
617
On the basis of the above calculations and analyses, several conclusions have been acquired. (1) The band gap narrowing of mono N- or C-substituted doping in ZnS systems is slight; thus the optical absorption of ZnS is not significantly improved and still in the UV region. (2) The interstitial doping types will induce some donor states in the gap, which are a benefit for improving the visible-light absorption, yet they are very difficult to form in the ZnS due to their larger formation energy. (3) For the N- and C-doped ZnS configurations, some partially filled states are formed in the band, which may act as a trap of the photoexcited electron and thus reduce the photogenerated current. 3.2. NþC-Codoped ZnS. As in the discussions in the above section, the experimentally observed improved photocatalytic activity of (N, C)-codoped ZnS is impossible originated from the mono N or C impurity. Therefore, in this section, to study the origin of the enhanced visible-light activity, we have implemented an accurate DFT calculation for the electronic structures of NþC-codoped configurations and given detailed discussions for the synergistic effects of N- and C-doping in this section. Three codoped configurations have been considered as NsþCs, NsþCi, and NiþCs. Because of the higher formation of interstitial doping, the two interstitial atoms in the lattice are not considered in the present work. A. Geometric Structures. To investigate the most stable configuration of NsþCs-codoped ZnS, several distinct supercell structures
with two substitutions placed at several possible sites are examined. Calculated results are tabulated in Table 3 according to the N-C separation. It is found that the structure for the nearest N-C separation is the most stable configuration. Figure 5a gives the most stable NsþCs structures after geometry optimization. There are four N-Zn bonds with 1.931, 2.001, 2.012, and 1.985 Å lengths, yet only three C-Zn bonds are formed with lengths 2.000, 2.007, and 1.982 Å. The reduced coordination of C impurity atom should be due to the stronger electronegativity of N atom, which induces the adjacent Zn more close to the N atom yet away from the C atom. The calculated formation energy of the NsþCs-doped configuration is larger than other doping types, which should be attributed to the significant local structure changes. When the Ns along with the Ci incorporated into ZnS, after geometry optimization, the C atom is bonded to the substituted N atom with bond length of about 1.248 Å (see Figure 5b). The formation energy for the NsþCicodoped configuration is also listed in Table 1, which is smaller than that of mono interstitial carbon-doped ZnS. This result suggests that the substituted impurity will promote the introduction of the interstitial doping, and the NC species may be present in the higher dopant concentration. The optimized structures for NsþCi and NiþCs codoping are almost the same, and the calculated electronic properties are similar. So, only NsþCi-codoped configuration is discussed in the present work (the optimized structure of NiþCsdoped ZnS is shown in the Supporting Information). The electron density and electron density difference maps give more information about the changes of chemical bonds due to the doping. After electronic redistribution, there are some significant changes of the chemical bonds in the NsþCs-codoped ZnS (parts a and a0 of Figure 6). The original mutual Zn of two adjacent tetrahedrons is drawn close to the N atom through stronger coulomb attraction. Thus, there are only three C-Zn bonds formed by the common electron cloud. The calculated charge population is also in agreement with this result, which shows more charges on the N ion (-0.94 e) than the C ion (-0.64 e). The changes of C-Zn chemical bonds may induce a significant change of C impurity states in the band structure and 2223
dx.doi.org/10.1021/jp110263e |J. Phys. Chem. C 2011, 115, 2218–2227
The Journal of Physical Chemistry C
ARTICLE
Figure 5. The optimized structures of (a) NsþCs-codoped, (b) NsþCi-codoped, (c) 2NsþCi-codoped, and (d) 2CsþNi-codoped ZnS.
further influence the electronic structures of NsþCs-codoped ZnS. For the NsþCi codoping, the electron density map in Figure 6b illustrates the substituted N atom attracts the interstitial C by Coulomb force, and the N-C bond is formed by the common electron cloud showing a covalent character. The total charges on the N and C are about -0.83 e capturing from the adjacent Zn ions, indicating a NC- species in the lattice. B. Electronic Structures. Figure 7a illustrates the band structure of the NsþCs-codoped ZnS, showing that one isolated level is localized in the band gap above the Fermi level about 0.52 eV. Thus, the photo excitation energy from the occupied level to the isolated level has induced a larger red shift for the optical absorption edge compared with the pure phase. The calculations of DOS and PDOS in parts a and a0 of Figure 8 indicate some of the N 2p and C 2p hybrid states with the valence band and thus increase the width of valence band. Hence, the host band gap has a slight narrowing about 0.10 eV. The impurity level in the gap mainly originates from the C 2p states with little input from the N 2p and S 3p states. This electronic structure characteristic is different with the only Ns- or Cs-doped case. In the only N- or C-doped ZnS, the local structures of doped atoms are similar to the original S atoms in the tetrahedron coordination; thus the N or C 2p states just form the continuous states at the valence band topmost. However, in the NsþCs-codoped ZnS, the coordination number of C atom decreases to three, inducing a significant local structural distortion. Therefore, the energy of C 2p orbital has a larger change so as to form impurity states in the gap. In addition, the Fermi level is pinned in the top of the valence band, which indicates the partially filled states are still in the band. For
the NsþCi-codoped ZnS in Figure 7b, two impurity levels are located below the bottom-most CBM, and the lower level is pinned through the Fermi level showing a half-filled state. Although the host band gap has a larger increase to about 2.61 eV, the electronic transition energy from the half occupied impurity level to the CBM is only about 0.4 eV, showing a larger red shift than pure phase and other doping. From the DOS and the PDOS in parts b and b0 of Figure 8, we can see the two impurity levels are predominantly originated from the mixing of the C 2p and N 2p states. Combined with the analyses of the geometrical structures and charge distribution, we assign the impurity states to the contribution of the NC π antibonding states. And thus we conclude the forming of NC species may be an important influencing factor for improving the ZnS photocatalytic activity in the visible light. In this section, the geometrical and electronic structures have been discussed, and several important viewpoints are given. (1) The substituted doping is bonded with the interstitial doping and thus steady the interstitial doping in the ZnS lattice, which suggests a possibility of both substituted and interstitial doping under higher impurity concentrations. (2) The mixing states of N 2p and C 2p states originated from the NC species should be responsible for the visible-light absorption in the experiment. (3) The half-occupied states are still formed in the both NþCcodoped models, which may be disadvantageous for improving the photocurrent of ZnS catalysis. 3.3. 2NsþCi-Codoped ZnS. The above-mentioned monodoped or lower concentration ratios of NþC-codoped systems have some difficulties on improving the visible photocatalytic 2224
dx.doi.org/10.1021/jp110263e |J. Phys. Chem. C 2011, 115, 2218–2227
The Journal of Physical Chemistry C
ARTICLE
Figure 6. The total electron density maps (left) and the electron difference density maps (right) for (a), (a0 ) NsþCs-codoped, (b), (b0 ) NsþCi-codoped, (c), (c0 ) 2NsþCi-codoped, and (d), (d0 ) 2CsþNicodoped ZnS. The cross section is the (001) plane.
activity. To enhance the photocatalytic efficiency of ZnS and red shift the absorption edge to visible-light region, several items have been cautiously considered. To achieve a better visible-light response, the interstitial doping should be employed. And the calculations indicate the substituted doping could promote the incorporation of interstitial impurity in the lattice by a stronger bonding. Therefore, the main point is how to avoid creating the partially occupied states. In the present work, the donor-acceptor pairs are suggested to adjust the electronic structure of ZnS with the charge compensated effect. As shown in the above section, NsþCi codoping gets a larger red shift yet induces one partially filled level in the band. The experiment by Muruganandham et al. shows the concentration ratio of N/C is about 2:1, which noted a Zn-N-C-N species may be in the lattice.7 Thus, on the basis of the above discussions and the electronic configurations of N 2s22p3 and C 2s22p2, we think that 2NsþCi codoping as a codopants pair probably get a desirable visible photocatalytic activity. In this case, the electrons on the donor levels passivate the same amount of holes on the acceptor levels, so the systems still keep semiconductor character. In this section, the synergistic effects of 2NsþCi codoping have been discussed in detail for the optimized structure. For comparisons, with the same total impurity concentration, the model of 2CsþNi codoping is also calculated and discussed. A. Geometric Structures. Figure 5c gives the whole structure of the 2NsþCi-codoped ZnS, which shows the two N and one C atoms tend to form N-C-N local trimers structure. The
Figure 7. The band structures of (a) NsþCs-codoped, (b) NsþCicodoped, (c) 2NsþCi-codoped, and (d) 2CsþNi-codoped ZnS. The Fermi level is set at zero energy.
N-C-N is in one line, and the lengths of the two N-C bonds are 1.242 and 1.259 Å. The trimer structure is similar to the NsþCi codoping, which gives N-C species in the lattice. The calculated formation energy indicates the N-C-N structure is easily formed and may be common in the ZnS host with higher (N, C)-codoping concentration. To see if the formation of the N-C-N defect pair is stable, we calculated the defect pair binding energy Eb = E(D1) þ E(D2) - E(D1þD2) - E(ZnS),23 where E(D) is the total energy of the doped system calculated with the same supercell. Positive Eb indicates that the defect pairs tend to bind to each other when both are present in the sample. The calculated binding energies Eb for the (NsþCs), (NsþCi), and (2NsþCi) pairs are 4.24, 7.92, and 9.68 eV, respectively, indicating that the N-C-N trimers structure is the most stable 2225
dx.doi.org/10.1021/jp110263e |J. Phys. Chem. C 2011, 115, 2218–2227
The Journal of Physical Chemistry C
ARTICLE
Figure 8. The total DOS (left) and the PDOS (right) for (a), (a0 ) NsþCs-codoped, (b), (b0 ) NsþCi-codoped, (c), (c0 ) 2NsþCi-codoped, and (d), (d0 ) 2CsþNi-codoped ZnS. The Fermi level is set at zero energy.
configuration for codoped systems. The results are in good agreement with experimental XPS measuring with a N/C ratio about 2:1, which suggests the possibility of Zn-N-C-N- type linkage may be in the lattice.7 For the 2CsþNi-codoped ZnS supercell, after geometric relaxing, the interstitial N atom and two substituted C atoms also incline to form C-N-C trimers structure (Figure 5d). The two C-N bond lengths are 1.243 and 1.301 Å. The formation energy of 2CsþNi codoping is larger than that of 2NsþCi codoping, which indicates the N-C-N structure may be more common in the (N, C)-codoped ZnS. However, the binding energy of C-N-C impurity is 4.83 eV, indicating that the trimer structure is also stable in the lattice once it is formed. The electron density map in Figure 6c shows some common electron clouds are around the whole N-C-N trimers, representing a covalent characteristic for N-C bonds. The electron density difference analysis in Figure 6c0 notes some charges transfer from the C atom to the adjacent two N atoms. This conclusion is in good agreement with the results obtained from the Mulliken charge population in Table 2, showing the charges on the C ion are 0.16 e with a positive ionic state. The results confirm the charge compensation between the nitrogen and the carbon impurities in the N-C-N structure. Thus, we conclude the large binding energy should result from charge transfer from donor (Ci) to acceptors (Ns) and the associated strong Coulomb interaction between positively charged donor and negatively acceptors. For the 2CsþNi codoping, the similar results are achieved and given in parts d and d0 of Figure 6. However, because of the stronger electronegativity of the N atom than the C atom, there is still some charge transfer from the adjacent C toms to the interstitial N atom (-0.28 e), as shown in Table 2. B. Electronic Structures. From the band structure of 2NsþCicodoped ZnS in Figure 7c, we can see that host band gap has a slight increase to about 2.29 eV and that one impurity level is localized in the band gap. The Fermi level is pinned above the impurity level, indicating no partially occupied states induced by 2NsþCi codoping, showing a passivated characteristic. The DOS
and PDOS in parts c and c0 of Figure 8 show N 2p states mix with S 3p states in the whole valence band, while the C 2p states are mainly contributing to the conduction band, which should be attributed to the loss of electrons on C impurity. The impurity level in the gap predominantly originates from the N 2p states with little input from the C 2p states, and the electron excitation energy from the N 2p gap states to the CBM is about 1.13 eV. By consideration of a scissor operator of 1.6 eV to correct the calculated gap of pure phase to the experimental value of 3.7 eV, the transition energy from the gap states to CBM should be about 2.7 eV, corresponding to a 460-nm visible-light absorption. In addition, the top of the valence band has a larger shift about 1.30 eV to the lower energy, suggesting the oxidation ability of the valence band hole carriers induced by the photoexcited electrons may be enhanced, thus further improving the photocatalytic activity of the ZnS. For 2CsþNi codoping in Figure 7d, the band structure shows the host band gap has no significant change and also one impurity level localized in the gap. The calculated DOS and PDOS in parts d and d0 of Figure 8 indicate the level in the gap mostly originates from the C 2p states with little input from the N 2p states, which extends beyond the Fermi level showing a partially filled character. The transition energy from the impurity states to the CBM is about 0.49 eV, which is a benefit to the visible-light absorption. However, due to the halffilled states in the gap, 2CsþNi codoping for ZnS still has some defaults on improving the photocurrent density. Therefore, on the basis of the above discussions, we suggest that 2NsþCi codoping may be the more favorable “good pair” to improve the visible photocatalytic activity for ZnS.
4. CONCLUSIONS On the basis of the calculated results, several appropriate viewpoints have been proposed for the exploiting of the visiblelight photocatalytic activity on the (N, C)-codoped ZnS. For both substituted N and C doped configurations, the band gap narrowing is slight and thus not enough to improve the visible 2226
dx.doi.org/10.1021/jp110263e |J. Phys. Chem. C 2011, 115, 2218–2227
The Journal of Physical Chemistry C optical absorption. Although the interstitial doping will induce a considerable red shift for ZnS, the calculated energy indicates the formation of the doping is difficult with the lower impurity concentration. In the NsþCs-codoped ZnS, due to the stronger electronegative of N, the coordination number of C is lowered. Thus, the electron excitation energy from C 2p impurity states these states to conduction band minimum decreases by about 0.74 eV compared to the pure phase. In the NsþCi-codoped ZnS, the transition energy of the photoexcitated electron has a large red shift, suggesting a visible-light absorption. However, in both NþC-codoped models, some half-occupied states are located in the gap, which is disadvantageous to improve the photoinduced current. Thus, in the present work, the passivated 2NsþCi-codoping pairs have been investigated as a candidate to improve the photocatalytic activity. Our calculations indicate that the N-C-N trimer structure is formed in the lattice and induces an impurity state in the gap, which reduces the electron transition energy to the ideal visible-light region. In addition, no half-filled states are formed in this configuration because of the charge compensation effect in the acceptor-donor-acceptor species. Although the compensation effect also exists in the 2CsþNi-codoped ZnS, there are still some partially occupied states in the gap, which we contribute to the mismatching of acceptor and donor. Our work provides a solid basis for the rationalization of experimentally observed red shift of optical absorption in wurtzite ZnS as a consequence of (N, C) codoping and shows that codoping with 2NþC will be a promising way for improving the visible-light activity of semiconductor photocatalysts.
’ ASSOCIATED CONTENT
bS
Supporting Information. The original structures of the doped ZnS and the optimized structure of NiþCs-codoped ZnS are shown in the Supporting Information. This material is available free of charge via the Internet at http://pubs.acs.org/
’ AUTHOR INFORMATION Corresponding Author
*Phone: 86-531-88366330. Fax: 86-531-88364864. E-mail: fwl@ sdu.edu.cn.
’ ACKNOWLEDGMENT This work is supported by the National Natural Science Foundation of China (Grant No. 50802056 and 91022034), 973 Program of China (Grant No. 2009CB930103), Excellent Youth Foundation of Shandong Scientific Committee (Grant No. JQ201015), Youth Scientist (Doctoral) Foundation of Shandong Province of China (Grant No. BS2009CL038), and Independent Innovation Foundation of Shandong University (Grant No. 2009TS016).
ARTICLE
(7) Kudo, A.; Sekizawa, M. Cat. Lett. 1999, 58, 241. (8) Kudo, A.; Sekizawa, M. Chem. Commun. 2000, 15, 1371. (9) Tsuji, I.; Kudo, A. J. Photochem. Photobiol. A: Chem 2003, 156, 249. (10) Arai, T.; Senda, S.; Sato, Y.; Takahashi, H.; Shinoda, K.; Jeyadevan, B.; Tohji, K. Chem. Mater. 2008, 20, 1997. (11) Muruganandham, M.; Kusumoto, Y. J. Phys. Chem. C 2009, 113, 16144. (12) Jang, J. W.; Choi, S. H.; Jang, J. S.; Lee, J. S.; Cho, S.; Lee, K.-H. J. Phys. Chem. C 2009, 113, 20445. (13) Hu, J. S.; Ren, L. L.; Guo, Y. G.; Liang, H. P.; Cao, A. M.; Wan, L. J.; Bai, C. L. Angew. Chem., Int. Ed. 2005, 44, 1269. (14) Lu, M. Y.; Lu, M. P.; Chung, Y. A.; Chen, M. J.; Wang, Z. L.; Chen, L. J. J. Phys. Chem. C 2009, 113, 12878. (15) Jun, J.; Jin, C.; Kim, H.; Kang, J.; Lee, C. Appl. Phys. A: Mater. Sci. Process. 2009, 96, 813. (16) Di Valentin, C.; Finazzi, E.; Pacchioni, G.; Selloni, A.; Livraghi, S.; Czoska, A. M.; Paganini, M. C.; Giamello, E. Chem. Mater. 2008, 20, 3706. (17) Gai, Y. Q.; Li, J. B.; Li, S. S.; Xia, J. B.; Wei, S. H. Phys. Rev. Lett. 2009, 102, 036402. (18) Du, X.; He, J. H.; Zhao, Y. Q. J. Phys. Chem. C 2009, 113, 14151. (19) Park, H.; Choi, W. J. Phys. Chem. B 2004, 108, 4086. (20) Park, J. S.; Choi, W. Langmuir 2004, 20, 11523. (21) Mrowetz, M.; Selli, E. Phys. Chem. Chem. Phys. 2005, 7, 1100. (22) Reyes-Garcia, E. A.; Sun, Y. P.; Raftery, D. J. Phys. Chem. C 2007, 111, 17146. (23) Zhang, J.; Pan, C. X.; Fang, P. F.; Wei, J. H.; Xiong, R. Appl. Mater. Interfaces 2010, 2, 1173. (24) Chen, D. M.; Jiang, Z. Y.; Geng, J. Q.; Wang, Q.; Yang, D. Ind. Eng. Chem. Res. 2007, 46, 2741. (25) Cong, Y.; Chen, F.; Zhang, J.; Anpo, M. Chem. Lett. 2006, 35, 800. (26) Yamamoto, T. Jpn. J. Appl. Phys. 2003, 42, L514. (27) Yang, K. S.; Wu, R. Q.; Shen, L.; Feng, Y. P.; Dai, Y.; Huang, B. B. Phys. Rev. B 2010, 81, 125211. (28) Segall, M. D.; Lindan, P. J. D.; Probert, M. J.; Pickard, C. J.; Hasnip, P. J.; Clark, S. J.; Payne, M. C. J. Phys.: Condens. Matte. 2002, 14, 2717. (29) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865. (30) Kisi, E. H.; Elcombe, M. M. Acta. Crystallogr. C 1989, 45, 1867. (31) Monkhorst, H. J.; Pack, J. D. Phys. Rev. B 1976, 13, 5188. (32) Martin, R. M. Electronic Structure: Basic Theory and Practical Methods; Cambridge University Press: Cambridge, England, 2004. (33) Di Valentin, C.; Pacchioni, G.; Selloni, A. Phys. Rev. B 2004, 70, 085116. (34) Li, J. B.; Wei, S. H.; Li, S. S.; Xia, J. B. Phys. Rev. B 2006, 74, 081201.
’ REFERENCES (1) Fujshima, A.; Honda, K. Nature 1972, 238, 37. (2) Tryk, D. A.; Fujishima, A.; Honda, K. Electrochim. Acta 2000, 45, 2363. (3) Hashimoto, K.; Irie, H.; Fujishima, A. Jpn. J. Appl. Phys. 2005, 44, 8269. (4) Kudo, A.; Miseki, Y. Chem. Soc. Rev. 2009, 38, 253. (5) Zhou, H.; Li, X. F.; Fan, T. X.; Osterloh, F. E.; Ding, J.; Sabio, E. M.; Zhang, D.; Guo, Q. X. Adv. Mater. 2010, 22, 951. (6) Reber, J. F.; Meier, K. J. Phys. Chem. 1984, 88, 5903. 2227
dx.doi.org/10.1021/jp110263e |J. Phys. Chem. C 2011, 115, 2218–2227