Applied Physics A

, Volume 110, Issue 1, pp 129–135 | Cite as

Unusual photoelectric behaviors of Mo-doped TiO2 multilayer thin films prepared by RF magnetron co-sputtering: effect of barrier tunneling on internal charge transfer

  • B. X. Yan
  • S. Y. Luo
  • X. G. Mao
  • J. Shen
  • Q. F. Zhou


Mo-doped TiO2 multilayer thin films were prepared by RF magnetron co-sputtering. Microstructures, crystallite parameters and the absorption band were investigated with atomic force microscopy, X-ray diffraction and ultraviolet-visible spectroscopy. Internal carrier transport characteristics and the photoelectric property of different layer-assemble modes were examined on an electrochemical workstation under visible light. The result indicates that the double-layer structure with an undoped surface layer demonstrated a red-shifted absorption edge and a much stronger photocurrent compared to the uniformly doped sample, signifying that the electric field implanted at the interface between particles in different layers accelerated internal charge transfer effectively. However, a heavily doped layer implanted at the bottom of the three-layer film merely brought about negative effects on the photoelectric property, mainly because of the Schottky junction existing above the substrate. Nevertheless, this obstacle was successfully eliminated by raising the Mo concentration to 1020 cm−3, where the thickness of the depletion layer fell into the order of angstroms and the tunneling coefficient manifested a dramatic increase. Under this circumstance, the Schottky junction disappeared and the strongest photocurrent was observed in the three-layer film.

1 Introduction

Cationic doping is an effective approach to enhance the photoelectric property of TiO2 films [1, 2, 3, 4, 5, 6], which otherwise usually demonstrate poor response under visible light due to a wide band gap (E g=3.20 eV). Almost all these researches pointed out that there is a best content of metal ions in TiO2, although the concrete value varies with different doping methods adopted. Larger doping amount beyond the optimized value is found to inhibit internal charge separation much more severely, although in most cases it could bring about higher absorption of visible light. For that reason, quite often we have to abandon the content with the best absorption property resulting from a relatively heavier doping, and assign a smaller doping amount as the optimized ion content considering the fastest carrier transfer it begets [7, 8, 9].

Theoretically, if we could speed up the separation rate of photogenerated carriers in a heavily doped film, a stronger photocurrent density is to be expected under visible light. There have been some attempts aimed at this objective and the approach was to cover the ion-doped layer with another undoped TiO2 layer, whereby a self-built electric field at the interface was created [10, 11, 12]. Theoretically, another heavily doped layer implanted under the lightly doped layer is assumed to: (1) introduce an upward self-built electric field which could promote charge separation; (2) yield more photogenerated carriers under visible light as the energy gap would be curtailed; (3) inhibit charge separation because of a large number of defects; (4) intensify the Schottky barrier existing between it and the metal substrate, and thus hinder charge transfer. Obviously, assumptions (1) and (2) are desirable attainments while (3) and (4) are unfavorable effects; thereby, the overall result depends on which of them plays a leading role in fact. If (3) or (4) could be inhibited or even utterly eliminated under certain circumstances, especially when these circumstances could be manipulated, then probably a promising photoelectric performance of multilayer films could be observed.

Based on radio-frequency (RF) magnetron co-sputtering, we managed to take accurate control of the doping content as well as the thickness of each layer, making it possible and convenient to implant several upward electric fields, which could speed up internal charge transfer, during a single sputtering process. Since we have investigated the photoelectric property of uniformly Mo-doped TiO2 film [9], we continue to choose Mo as dopant in this study to investigate the microstructure and photoelectric properties of multilayer samples.

2 Experimental

All samples were deposited on titanium substrates with a RF magnetron sputtering machine (SY-300, Chinese Academy of Sciences) at room temperature and the background pressure was 2.0×10−3 Pa. Each sputtering process lasted for 120 min at a pressure of 0.5 Pa (Ar) that contained a little O2 (pressure ratio of O2/Ar was 1/143) to inhibit the decomposition of TiO2. Samples obtained were annealed in a furnace (SX2-8-10, Tianye) under 550 °C for 2 h and with an Alpha-step device (ET2000, Hitachi) we confirmed that the thicknesses of all films are around 300 nm.

Two targets were utilized for co-sputtering. One is a TiO2 ceramic target and the other is an embedded target, which was prepared by attaching a tiny Mo tablet on the sputtering ring of a TiO2 ceramic target. Besides the area ratio, the sputtering yield (Y), which indicates how many atoms could be sputtered out of the target surface by each ion (Ar+), should and has been considered for the accurate rectification of Mo content (Y Mo/Y Ti is 8/5) and all contents given below are rectified values. For a co-sputtering case, the power of each target would affect the overall composition of the film, so samples containing different contents of Mo were prepared by (1) fixing the power of the TiO2 (P T) target on 300 W; (2) modulating the power of the Mo-embedded target (P E) as 30, 60, 100, 200 and 300 W separately. In this way, we managed to prepare a multilayer film in one step. To give an appropriate judgment on the photoelectric performance of multilayer films, the best doping content of uniform doping was confirmed at first.

With respect to multilayer samples, the heavily doped, lightly doped and undoped layers were labeled as H, L and T (TiO2), respectively. Figure 1 gives a brief illustration of samples T, TL, LH and TLH. The thickness of each layer was optimized at first.
Fig. 1

Graphic illustrations of films with different layer structures

Structural characteristics of T, L and H were determined by X-ray diffraction (XRD, pw1710X, Philips) at a scanning speed of 4/min (Cu target, K α =0.1542 nm). An atomic force microscope was used in tapping mode (AFM, Park Scientific Instruments). The tip used here is an NCHV-A (42 N/m, 320 kHz, Al reflective coating), with a radius of 10 nm. The absorption edge was studied by an ultraviolet-visible spectrometer (UV2300, Hitachi) at a speed of 0.5 nm/s. To investigate the photoelectric property of the titanium substrate/multilayer Mo–TiO2 electrode, the photocurrent density was collected on an electrochemical workstation (CHI660B, CH Instruments) under visible light, with a three-electrode system—a saturated calomel electrode (SCE) used as standard cell electrode, a Pt electrode chosen as the counter electrode (CE) and the electrolyte was 0.5 mol/L Na2SO4. The illuminant was a xenon lamp (CHF-XM35-500) with light below 380 nm eliminated by a light filter and the light density was 100 mW cm−2.

3 Results and discussion

The photocurrent density of uniformly doped samples under visible light is presented in Fig. 2, from which we could infer that the power of 300 W (TiO2) plus 60 W (Mo-embedded target) yields the best doping content. A further rise of doping content reduced the current density because of increasing concentration of lattice defects, which begot a lower efficiency of carrier separation. Mo ions introduced into the crystal lattice could curtail the energy gap, but at the same time these ions also act as shallow traps for photogenerated carriers. In general, the recombination rate K recomb is affected by the average distance R between traps as K recomb∝exp(−2R/a 0), where a 0 is the hydrogenic radius of the wave function in the shallow trap [13]. Therefore, if lattice defects huddle closer as the doping content increases, the recombination rate K recomb will undergo an exponential growth due to a smaller R. This is the reason why heavy doping usually yields poor photocurrent.
Fig. 2

Photocurrent density of TiO2 thin films uniformly doped with different Mo contents under visible light

Mo content at the film surface in Table 1 was identified by X-ray Photoelectron Spectroscopy (XPS). Note that the content identified by XPS is usually much higher than the value within the film, mainly because XPS merely reveals the film composition near the surface (below 10 nm), where factors like surface contamination or surface enrichment could cause the XPS result to differ from its real value remarkably [14]. Therefore, contents of L (0.6 at.% Mo), H (1.2 at.% Mo) and H+ (3.6 at.% Mo) were calculated by target area ratio and the sputtering yields of Mo and Ti and details for these have been described in the Experimental part.
Table 1

Crystallite parameters of TiO2 thin films with different contents of Mo

Sample no.


w (Mo)/% at surface



Lattice parameters






300 + 0








300 + 60








300 + 300








0 + 300







D: crystallite size, Δd/d: lattice distortion, P T: sputtering power of TiO2 target, P E: sputtering power of Mo-embedded TiO2 target, w (Mo) was identified by XPS

Figure 3 is the XRD pattern of T, L, H and H+. For L and T only features of anatase are observed while for H and H+ all characteristic peaks disappeared because of a high concentration of dopant. Crystallite size and distortion could be calculated from d=/βcosθ and \(\varepsilon = \beta /4\operatorname {tg}\theta\), respectively [15], where θ is the diffraction angle belonging to the main peak, β is the full width at half maximum (FWHM) of the peak, λ is the wavelength of X-rays applied and K here is 0.89; the results attained are listed in Table 1.
Fig. 3

XRD patterns of TiO2 thin films doped with different contents of Mo

The radius of Mo6+ (r=0.062 nm) is similar to Ti4+ (r=0.068 nm), which makes it easy for Mo6+ to replace Ti4+ when the doping content was kept on a relatively low level. The substitution for Ti4+ would result in smaller particles and narrower energy gaps, as confirmed in the following UV-visible spectroscopy. For a heavy doping case like H or H+, most Mo would squeeze into the lattice, boost lattice distortions and thus reduce crystallite size [13, 14, 15]. The smallest crystallite is expected to occur in H+, which contains the largest number of Mo. For H or H+, neither crystal size nor distortion could be calculated accurately since all peaks have disappeared. Figure 4 is the AFM pattern of T, L, H and H+, from which the microstructure of each layer is manifested. Particles of a diameter of about 100 nm were observed at a rough surface of T, which is in good accordance with XRD analysis. As the doping content of Mo grows, particles begin to decrease in size. In H and H+, diameters are around 70 nm and 40 nm, respectively.
Fig. 4

AFM of layers with different contents of Mo

In Fig. 5a we can easily find a notable red shift of the absorption edge after doping. Usually, a longer edge signifies a larger number of carriers generated under visible light. For the n-type case, metal doping is proposed to introduce new donor levels under the conduction band of TiO2 by rendering excessive valence electrons that could be incited easily into free carriers. The right-most absorption edge showed up when the power of the embedded target P E reached 200 W. When P E reached 300 W, however, the absorption edge moved inversely and even utterly covered the edge of 100 W. This could be understood as a typical consequence of the well-known quantum effect [16]. As the particle size keeps decreasing, the movement of electrons would be confined more intensively, which causes the differentiation of energy levels and makes the band gap grow wider. For the same reason, compared to bulk materials, thin films usually have wider energy gaps, which is attested by our observation that the undoped film has a gap around 3.70 eV (for TiO2 bulk material it is 3.20 eV). In Fig. 5b, TL and LH both present a longer absorption edge compared to L, but instead of curtailing the energy gap as H did, the mechanism of TL is more similar to the window effect, which has been thoroughly investigated in thin-film solar cells [17]. Although LH presented better absorption than L, the absorption edge of TLH is shorter than TL. Therefore, it is quite difficult to judge whether the influence of H on the film’s absorption is beneficial or not. The edge of TLH+ moved remarkably toward a contrasting direction due to the quantum effect since the crystallite of H+ is the smallest, as we have mentioned above.
Fig. 5

UV-visible absorbance spectra of (a) samples deposited at various powers of the Mo-TiO2 target plus 300 W (TiO2 target); (b) of different layer structures

The photocurrent density of each sample is exhibited in Fig. 6. When we take a comparison between L and TL, or between LH and TLH, we can easily find that the surface layer T brought a notable rise of photocurrent. At the same time, comparison between L and LH, or between TL and TLH, suggests that the bottom layer H right did the contrast. Cen et al. also observed that the photoelectric property enhanced after covering the La-doped layer with another layer of pure TiO2 [10], while similar results were confirmed by Zhang et al. [12]. In these previous researches, the better photoelectric property obtained was ascribed to the p–n junction formed at the interface of two ideal layers, while the influence of particle size on energy levels was not discussed. Though the mechanism is similar to some extent, we have to make it clear that the junction discussed here is a homojunction rather than a p–n junction. It should be noted especially that the interface here does not refer to a substantive one between ideal layers; instead, it refers to numerous tiny interfaces between very small particles that belong to L and T separately. Thereby, the particle size must be taken into consideration since the existence of an electric field has long been controversial under small scales [18, 19, 20, 21].
Fig. 6

Photocurrent density of films with different layer structures under visible light

For the n-type case, according to basic knowledge in solid-state physics, the donor density N d is connected to the Fermi level E F:
$$ E_{\mathrm{c}} - E_{\mathrm{F}} = k_{\mathrm{B}}T\ln\biggl(\frac{N_{\mathrm{c}}}{N_{\mathrm{d}}} \biggr), $$
where E c is the lowest energy level of the conduction band, N c is the effective density of states of the conduction band, N d is the atomic density of donors and k B is the Boltzmann constant. Consequently, increasing doping amount will result in higher Fermi level; thus, electrons and holes generated by light would separate at the L/T interface, driven by an enlarged drop of Fermi levels. In such cases, electrons in L intend to keep diffusing into T because of the higher Fermi level, until an upward electric field retarding this diffusion grows sufficiently strong so as to maintain a dynamic balance at the interface. Based on a spherical crystallite model with a relative dielectric constant ε, Curran and Lamouche attempted to calculate the minimum particle radius (r 0) required to sustain a depletion layer that could exert a notable influence on carrier transport [22]. Suppose all the donors within the sphere (for the n-type case) are ionized and no electrons left in the conduction band; the maximum potential drop from the surface to the particle center is then
$$ \Delta V =\frac{Nr^2_0 q}{6\varepsilon\varepsilon_0}, $$
where N is the donor density of the crystallite, q is 1.6×10−19 C and ε 0 is 8.85×10−14 F cm−1. In general, if ΔV≤2k B T/q, its influence over carrier transport could be neglected. Hence, for a particle with radius r 0 satisfying the following inequality:
$$ r_0 \biggl(\frac{N}{\varepsilon}\biggr)^{1/2}\leq \frac{(12k_{\mathrm{B}}T\varepsilon_0)^{1/2}}{q} =407, $$
its thickness is supposed to be unable to sustain a potential drop that could influence the carrier transport remarkably. Here we have to point out a small mistake in Curran and Lamouche’s paper that the denominator in the inequality above should be q rather than q 2 and Eq. (3) is the corrected form. Fortunately, the final result of their calculation seems not to be affected by this. For this reason, there is supposed to be a critical size, which determines whether the electric field of a depletion layer should be considered or not.

For anatase, ε is 48; thereby, for Mo-doped samples such as L when N is 2.0×1018 cm−3 [18], the critical radius approaches 30 nm and would get much smaller with more Mo ions introduced in H and H+. It should be noted here that many researches have testified the existence of an internal electric field at the interface between particles on much smaller scales [10, 11, 12, 21]. Therefore, we suspect that Curran and Lamouche’s model cannot yield an appropriate estimation of the critical size because it considered neither the quantum effect nor the interface effect, both of which were later proved to exert great influences on the differentiation of energy levels on small scales (below 10 nm). Since particles obtained in T and L are larger than the critical size in this condition (30 nm) and, therefore, thick enough to sustain a considerable electric field, the existence of E as well as its acceleration on internal charge transfer was ensured, as confirmed in Fig. 6.

Nevertheless, from LH and TLH we can see that layer H did not improve the photocurrent a step further, as shown in Fig. 6, due to fast carrier recombination. Another significant cause of the photocurrent declining is the Schottky junction at the interface between the bottom layer and the substrate. Compared to L, as the Fermi level is higher and particles get smaller in H, this junction was intensified and would retard the flowing of electrons to the substrate much more significantly [23, 24, 25, 26]. Surprisingly, as Mo content in H kept rising, the photocurrent of TLH+ turned out to be stronger than TL, as shown in Fig. 7. This seems at odds with the general belief that there is an optimized doping content yielding the highest carrier density, while a larger doping amount over that would definitely slow the separation rate of carriers, and consequently result in declined photocatalytic and photoelectric properties [27, 28]. Recall that TLH+ did not present a further red-shifted absorption edge compared to TL; thus, this increase of photocurrent could only be ascribed to faster charge transfer. To give a further confirmation of this unusual behavior of TLH+, we raised the Mo concentration of L to 0.8 at.% Mo (L′, P E is 100 W) and then examined the function of H+ in this new case. As shown in the inset, similar to TLH+, TL′H+ presented a much stronger photocurrent than TL′ as well.
Fig. 7

Photocurrent density of films based on H (1.2 at.% Mo) and H+ (3.6 at.% Mo); the inset presents photocurrent changes after replacing L by L′ (0.8 at.% Mo)

It is possible that MoO3 may emerge and then together with TiO2 form a mixed semiconductor at such a high doping concentration. To examine the validity of this assumption, we prepared another single-layer sample named H+, which is of the same thickness (300 nm) as other samples. In contrast to expectation, its photocurrent came out to be extremely small, as shown in Fig. 7. What is more, as there was no sign of MoO3 in the XRD patterns, the stronger photocurrent should not be ascribed to the supposed mixed semiconductor.

Under a very high doping amount, another possible reason, barrier tunneling, caused us to pay more attention to the depletion layer of the Schottky junction, which might have been compressed much more thinly by a high content of Mo as
$$ d = \biggl(\frac{2\varepsilon_{\mathrm{s}}V_{\mathrm{bi}}}{eN_{\mathrm{d}}}\biggr)^{1/2}, $$
where ε s is the relative dielectric constant of the semiconductor. V bi is the barrier height, whose concrete value could be calculated by
$$ V_{\mathrm{bi}} = \varPhi_{\mathrm{m}} - \chi - \varPhi_{\mathrm{n}}, $$
where Φ m is the work function of the metal, χ is the electron affinity of the semiconductor and Φ n is the energy drop from the bottom of the conduction band to the Fermi level. Note that the thickness of the depletion layer was mainly determined by the donor density rather than the thickness of H, which was set at 20 nm because we found that a thicker bottom layer would drop the photocurrent drastically due to fast recombination. The atomic density of Ti is on the order of 1×1022 cm−3. Since the area ratio (Mo/Ti) of the embedded target used for H+ was 1/18, taking the different sputtering yield into consideration the actual atomic density of Mo in H+ was conservatively estimated as 3.6 at.%, i.e. above 9×1020 cm−3. For heavily doped TiO2, Φ m for titanium is 4.33 V, χ is 4.21 V, ε s for anatase is 48 and, if we neglect the extremely small Φ n in this case, then with Eqs. (4) and (5) the thickness of the depletion layer (d) could be calculated as 8.4 Å.
Figure 8 presents how a photogenerated electron tunnels through the ultra-thin barrier above the titanium substrate. Regions I and II belong to the heavily doped bottom layer (H+). It is well known that the tunneling coefficient largely depends on barrier width as well as barrier height. A high donor density in the heavily doped bottom layer would not only contribute to a thinner depletion layer, but also result in a smaller barrier height. On such a scale, the tunneling transmission coefficient follows [29]
$$ T \propto\exp\biggl( - \sqrt{\frac{2\varepsilon_{\mathrm{s}}V_{\mathrm{bi}}}{eN_{\mathrm{d}}}} \biggr). $$
Fig. 8

Tunneling of the ultra-thin Schottky barrier exists between the Ti substrate and the heavily doped bottom layer

For that reason, if the donor density N d reaches a higher level, it is physically valid to suppose a more obvious tunneling transmission. The specific value of T requires more discussion of the energy of electrons E outside the barrier as well as V bi, both of which differ a lot from bulk materials as the particles in H+ are much smaller than those in T and L. In such cases, the increasing of the energy level caused by the quantum effect should be taken into consideration for the sake of deciding the changing trend of V bi as well as E. Brus found that for small spherical particles with a radius below 10 nm, the value of E mainly depends on three terms [30]:
$$ E = \frac{\hbar^{2}\pi^{2}}{2R^{2}} \biggl[ \frac{1}{m_{\mathrm{e}}} + \frac{1}{m_{\mathrm{h}}} \biggr] - \frac{1.8e^{2}}{\varepsilon_{2}R} + \frac{e^{2}}{R}\overline{\sum _{n = 1}^{\infty } \alpha_{n} \biggl( \frac{S}{R}\biggr)^{2n}}. $$
The first term is the quantum energy of localization, the second term is the Coulomb attraction and the third term, in a formal sense, means the salvation energy loss. As the radius of the particle falls, the lowest energy E that is required for exciting an electron to the least-excited state will increase notably. Under this circumstance, V bi is also expected to decline with particle shrinking. Therefore, if we considered this variation trend of E and V bi caused by particle shrinking, the transmission coefficient T would grow larger. Theoretically, if tunneling indeed plays a dominant role in charge transfer, the Schottky barrier would be largely inhibited and thus faster charge transfer is to be expected. On this point, the unusually high photocurrent of TLH+ could be taken as evidence, while more proof for this could be found in Fig. 9.
Fig. 9

Photocurrent versus bias potential of samples with different layer structures

To apply a positive bias potential on the substrate is to introduce an upward electric field attracting the down-flowing photogenerated electrons. For undoped TiO2 and sample TL, there is no obvious sign of a Schottky junction. With respect to all samples based on a heavily doped bottom layer (LH or TLH), however, the current increases dramatically when a positive voltage over 0.2 V was exerted, which is a typical characteristic of Schottky junctions. Theoretically, if this junction was strengthened by increasing the doping amount in the bottom layer, its inhabitation on electron transfer would turn more obvious, which has been confirmed by the comparison between TL and TLH in Fig. 6. As the consequence of barrier tunneling, this junction, instead of getting stronger, totally disappeared for TLH+ in Fig. 9, offering a crucial clue to understand why the strongest photocurrent was observed in such a three-layer structure.

4 Conclusions

Mo-doped thin films deposited by RF co-sputtering and annealed at 500 °C in air manifest anatase phase and a porous surface that consists of many tiny particles. The particle diameter decreases from 100 nm to 40 nm with Mo concentration growing to 3.6 at.% Mo. Compared to uniform Mo doping, the surface layer of TiO2 proved effective to enhance the photoelectric property of the multilayer samples as well as to shift the absorption edge toward the visible region notably. In contrast, films implanted with a heavily doped bottom layer H (1.2 at.% Mo) did not yield a stronger photocurrent, caused by a too fast recombination of carriers. Neither did this extend the absorption edge toward the visible region further due to a stronger quantum effect intensified by shrinking particle size.

The three-layered sample based on a heavier doped bottom layer (about 3.6 at.% Mo) presented the strongest photocurrent because higher donor density and stronger quantum effect together resulted in an ultra-thin depletion layer, which remarkably boosted the tunneling coefficient of the Schottky barrier at the interface between film and substrate. Tunneling is proved an effective approach to eliminate the side effect of such barriers, which otherwise will retard the internal carrier transfer within films.



This work was financially supported by the National Key Basic Research Program of China (973) (2012CB934303), the Joint Fund launched by the Department of Science and Technology of Guizhou province and Guizhou Minzu University (LKM [2012]24), and the Guiyang Science & Technology Department ([2012205]6-12).


  1. 1.
    J.F. Zhu, F. Chen, J.L. Zhu, H.J. Chen, M. Anpo, J. Photochem. Photobiol. A, Chem. 180, 196 (2006) CrossRefGoogle Scholar
  2. 2.
    A. Kubacka, G. Colón, M. Fernández-García, Catal. Today 143, 286 (2009) CrossRefGoogle Scholar
  3. 3.
    J.J. Sene, W.A. Zeltner, M.A. Anderson, J. Phys. Chem. B 107, 1597 (2003) CrossRefGoogle Scholar
  4. 4.
    Y. Cao, W.S. Yang, W.F. Zhang, G.Z. Liu, P. Yue, New J. Chem. 28, 218 (2004) CrossRefGoogle Scholar
  5. 5.
    C.C. Chang, C.K. Lin, C.C. Chan, C.S. Hsu, C.Y. Chen, Thin Solid Films 494, 274 (2006) CrossRefADSGoogle Scholar
  6. 6.
    X. Zhang, F. Zhang, K.Y. Chan, Mater. Chem. Phys. 97, 384 (2006) CrossRefGoogle Scholar
  7. 7.
    L.J. Wang, J.X. Yuan, P.Y. Peng, T.H. Lu, Chin. J. Appl. Chem. 24, 53 (2007) Google Scholar
  8. 8.
    V.C. Štengl, S. Bakardjieva, J. Phys. Chem. C 114, 19308 (2010) CrossRefGoogle Scholar
  9. 9.
    B.X. Yan, S.Y. Luo, J. Shen, Acta Phys. Chim. Sin. 28, 381 (2012) Google Scholar
  10. 10.
    J.W. Cen, X.J. Li, M.X. He, S.J. Zheng, M.Z. Feng, J. Chin. Rare Earth Soc. 23, 668 (2005) Google Scholar
  11. 11.
    H. Yu, X.J. Li, S.J. Zheng, G. Xu, Chin. J. Inorg. Chem. 22, 978 (2006) Google Scholar
  12. 12.
    K.J. Zhang, W. Xu, X.J. Li, S.J. Zheng, G. Xu, J.H. Wang, Trans. Nonferr. Met. Soc. China 16, 1069 (2006) CrossRefGoogle Scholar
  13. 13.
    P. Ho, M.E. Coltrin, J.S. Binkley, C.F. Melius, J. Phys. Chem. 90, 3399 (1986) CrossRefGoogle Scholar
  14. 14.
    C.X. Li, D. Zhang, Z.H. Jiang, Z.P. Yao, F.Z. Jia, New J. Chem. 35, 423 (2011) CrossRefGoogle Scholar
  15. 15.
    L.S. Yin, Q.F. Zhou, X.G. Tang, J. Funct. Mater. 30, 498 (1999) Google Scholar
  16. 16.
    J.C. Yu, J. Yu, W. Ho, Z. Jiang, L. Zhang, J. Photochem. Photobiol. A, Chem. 14, 3808 (2002) Google Scholar
  17. 17.
    Z.M. Chen, J.N. Wang, X.Y. Mei, G.L. Kong, Solid State Commun. 58, 379 (1986) CrossRefADSGoogle Scholar
  18. 18.
    S.E. Lindquist, B. Finnstrom, L. Tegner, J. Electrochem. Soc. 130, 351 (1983) CrossRefADSGoogle Scholar
  19. 19.
    K. Schwarzburg, F. Willig, J. Phys. Chem. B 103, 5743 (1999) CrossRefGoogle Scholar
  20. 20.
    A. Solbrand, H. Lindström, H. Rensmo, A. Hagfeldt, S.E. Lindquist, S. Södergren, J. Phys. Chem. 101, 2514 (1997) Google Scholar
  21. 21.
    K.Q. Tan, H.R. Zhang, C.F. Xie, Y.Z. Gu, W.F. Zhang, Int. J. Nanotechnol. 11, 331 (2010) Google Scholar
  22. 22.
    J.S. Curran, D. Lamouche, J. Am. Chem. Soc. 87, 5405 (1983) Google Scholar
  23. 23.
    F. Ruffino, M.G. Grimaldi, F. Giannazzo, F. Roccaforte, V. Raineri, Appl. Phys. Lett. 89, 243113 (2006) CrossRefADSGoogle Scholar
  24. 24.
    I.B. Altfeder, J.A. Golovchenko, V. Narayanamurti, Phys. Rev. Lett. 87, 056801 (2001) CrossRefADSGoogle Scholar
  25. 25.
    G.D.J. Smit, S. Rogge, T.M. Klapwijk, Appl. Phys. Lett. 81, 3852 (2002) CrossRefADSGoogle Scholar
  26. 26.
    F. Ruffino, A. Canino, M.G. Grimaldi, F. Giannazzo, F. Roccaforte, V. Raineri, J. Nanomater. 2008 243792 (2008) CrossRefGoogle Scholar
  27. 27.
    B.X. Lin, Z.X. Fu, C.H. Liu, G.H. Liao, J.J. Zhu, L. Duan, Res. Prog. Solid State Electron. 22, 417 (2002) Google Scholar
  28. 28.
    L.G. Devi, B.N. Murthy, Catal. Lett. 125, 320 (2008) CrossRefGoogle Scholar
  29. 29.
    H.M. Wang, Y.F. Zhang, Acta Phys. Sin. 54, 2226 (2005) Google Scholar
  30. 30.
    L.E. Brus, J. Chem. Phys. 80, 4403 (1984) CrossRefADSGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2012

Authors and Affiliations

  • B. X. Yan
    • 1
  • S. Y. Luo
    • 1
    • 2
  • X. G. Mao
    • 1
  • J. Shen
    • 1
  • Q. F. Zhou
    • 1
  1. 1.Department of Materials ScienceFudan UniversityShanghaiP.R. China
  2. 2.College of ScienceGuizhou University for NationalitiesGuizhouP.R. China

Personalised recommendations