Image charge effect on the light emission of rutile TiO2(110) induced by a scanning tunneling microscope
Guo Chaoyu1, Meng Xiangzhi1, Wang Qin1, Jiang Ying1, 2, 3, †
International Center for Quantum Materials, School of Physics, Peking University, Beijing 100871, China
Collaborative Innovation Center of Quantum Matter, Beijing 100871, China
CAS Center for Excellence in Topological Quantum Computation, University of Chinese Academy of Sciences, Beijing 100049, China

 

† Corresponding author. E-mail: yjiang@pku.edu.cn

Project supported by the National Key R&D Program of China (Grant Nos. 2016YFA0300901 and 2017YFA0205003), the National Natural Science Foundation of China (Grant Nos. 11634001 and 21725302), and the Key Research Program of the Chinese Academy of Sciences (Grant No. XDPB08-1).

Abstract

The plasmon-enhanced light emission of rutile TiO2(110) surface has been investigated by a low-temperature scanning tunneling microscope (STM). We found that the photon emission arises from the inelastic electron tunneling between the STM tip and the conduction band or defect states of TiO2(110). In contrast to the Au(111) surface, the maximum photon energy as a function of the bias voltage clearly deviates from the linear scaling behavior, suggesting the non-negligible effect of the STM tip on the band structure of TiO2. By performing differential conductance (dI/dV) measurements, it was revealed that such a deviation is not related to the tip-induced band bending, but is attributed to the image charge effect of the metal tip, which significantly shifts the band edges of the TiO2(110) towards the Femi level (EF) during the tunneling process. This work not only sheds new lights onto the understanding of plasmon-enhanced light emission of semiconductor surfaces, but also opens up a new avenue for engineering the plasmon-mediated interfacial charge transfer in molecular and semiconducting materials.

1. Introduction

The light emission induced by the tunneling current of scanning tunneling microscope (STM) has received increasing attention recently, since it provides the unprecedented possibility of studying the electron-photon conversion at the atomic scale,[16] which is critical for understanding and controlling the properties of future nanoscale optoelectronic devices. Inelastic electron tunneling (IET) process can excite the localized surface plasmon (LSP) modes confined within the cavity between STM tip and the metal substrate. Those LSP modes decay into propagating photons in free space, which are detectable by far-field optical setups (Figs. 1(a) and 1(b)). The high spatial resolution of STM-induced light emission arises from the highly local charge injection and excitation through the tunneling current. The STM-induced light emission was first reported in 1989 by Gimzewski et al.[1] Ever since, the light emission of various systems has been extensively investigated by STM, focusing on metals surfaces,[7,8] adsorbed molecules,[911] quantum wells,[12] artificial nanostructures,[13,14] and so on. However, the light-emission experiments on semiconductor surfaces are very rare and our understanding of the microscopic mechanism is still far from complete.

Fig. 1. (color online) (a) Schematic of plasmon-enhanced light emission in the STM junction. (b) Energy diagram illustrating the mechanism for light emission of Au(111) sample. Elastic (ET) and inelastic tunneling (IET) processes are indicated by solid and dashed arrows, respectively. (c) Light emission spectra recorded on the bare Au(111) surface (constant-current mode, I = 2 nA, V = 1.5 V–3.5 V). The green arrows indicate the cut-off wavelengths (maximum photon energy, Emax) of the light emission spectra. (d) The maximum photon energy (Emax) as a function of bias voltage.

There are several difficulties for studying the plasmon-enhanced light emission on semiconducting materials by STM. First, the free-carrier density in semiconductors is much less than that in metals, leading to a much weaker plasmonic field.[15] Therefore, the plasmon enhancement of the tip-surface cavity is not as efficient as metal surfaces and the photon emission should be rather weak on semiconducting surfaces. Second, the band structure of semiconductors is more complicated than that of metals. The dopants and defects often induce prominent in-gap states,[1618] which greatly modulate the inelastic electronic transition and thus the light emission process. Third, the STM tip can induce significant changes in the electronic structure of semiconductors, such as tip-induced band bending[1921] and image charge effect.[2225] In this work, we choose to investigate the rutile TiO2(110), which is one of the most widely studied materials in photoconversion and photocatalysis. We are able to obtain high-quality spectroscopy of light emission using our homemade optical STM and the effect of STM tip on the light emission is unambiguously revealed.

2. Experimental details

The experiments were performed with a homemade ultrahigh vacuum (UHV) STM system, with a base pressure better than 2 × 10−10 mbar. A rutile TiO2(110) single crystal (MaTech) was cleaned with cycles of sputtering by Neon ions and annealing at about 1000 K. In order to increase the conductivity at low temperature, the TiO2 sample was prepared with a high concentration of surface and subsurface defects. After the surface preparation, the sample was transferred immediately to the cryogenic STM stage to avoid the contamination of the TiO2(110) surface.

The STM characterization and light emission experiments of the TiO2(110) surface were performed with electrochemically etched silver tips. The scanning tunneling spectroscopy (STS) was acquired using lock-in detection of the tunneling current by adding a 13 mVrms modulation at 287 Hz to the sample bias at 5 K. Bias voltage refers to the sample voltage with respect to the tip.

Photons emitted from the tunneling junction were collected by a UHV compatible aspheric lens (Edmund Optics, EFL = 20 mm, N.A. = 0.38), which was fixed to a high-precision nanopositioner (Attocube) and then guided through a mirror system into a grating spectrometer coupled to a liquid-nitrogen-cooled charge-coupled device (CCD) (Princeton Instrument, SP2300). All light emission experiments were carried out at a sample temperature of 77 K. All the emission spectra were acquired at constant-current mode.

3. Results and discussion

Owing to their characteristic dielectric function, noble metals support strong surface plasmon oscillations in the energy range corresponding to the visible lights.[26,27] The shape and intensity maxima of the light emission spectrum correspond to particular modes of plasmon resonance, determined by both the tip and the sample. For Au(111), it emits red lights when a bias of a few volts is applied. Figure 1(c) shows a series of emission spectra of the Au(111) surface at 77 K with varying sample bias from 1.5 to 3.5 V. The broad peak around 1.96 eV (630 nm) corresponds to the LSP resonance mode. The cut-off wavelength (maximum photon energy, Emax) of the spectral distribution (denoted by the arrows in Fig. 1(c)) exhibits a continuous blueshift when increasing the sample bias. In Fig. 1(d), we plot Emax as a function of the bias voltage, showing a perfect linear relation, accompanied with a level-off when the bias voltage is larger than 2.4 V. In the linear region, the maximum photon energies are identical to eV, consistent with the picture of IET-induced light emission (Fig. 1(b)). The maximum photon energies beyond the bias of 2.4 V are limited by the energy range of the LSP mode.

For semiconductor, such as TiO2, the mechanism of STM-induced light emission is similar with that of noble metal samples (Figs. 2(a) and 2(b)). However, the maximum photon energy Emax is smaller than eV due to the existence of the bandgap. For V > 0, the light emission is induced by the inelastic electronic transition between the tip and the conduction band, and Emax = eVECB, where ECB is the position of conduction band edge with respect to the Fermi level (EF) of TiO2 (Fig. 2(a)). For V < 0, the photons are generated by the electrons inelastically tunneling from the valence band to the tip and Emax = e|V| + EVB (Fig. 2(b)). It should be noted that surface/subsurface defects on TiO2 such as oxygen vacancies and Ti interstitials may lead to prominent in-gap states (Ed) below the EF.[28,29] In this case, the maximum photon energy , where is the upper energy bound of defect states with respect to EF.

Fig. 2. (color online) (a), (b) Energy diagrams illustrating the mechanism for light emission of TiO2(110) sample at positive and negative biases, respectively. The conduction band minimum (ECB), valence band maximum (EVB), Fermi level (EF), defect state (Ed) are indicated. The upper energy bound of defect state is denoted by . The double-headed arrows denote the sample bias (V) and the emitted photon energy (). The blue dashed line represents the inelastic tunneling processes. (c) Light emission spectra recorded at the defective TiO2(110) surface (constant-current mode, I = 3 nA, V = 3.0 V–4.0 V). The spectra are offset for clarity. The green arrows indicate the cut-off wavelengths (maximum photon energy, Emax). (d) The maximum photon energy (Emax) measured at positive (black) and negative (red) sample biases. The dashed lines mark the level-off bias voltage. (e) The energies of band edges (ECB and ) deduced from light emission spectra at positive (black) and negative (red) sample biases, respectively.

Figure 2(c) shows a series of emission spectra taken at a defective area of TiO2(110) with varying bias voltage from 3 V to 4 V, where the cut-off wavelengths are highlighted by arrows. The relationship between the maximum photon energy (Emax) and the sample bias is shown in Fig. 2(d), indicating a clear deviation from the linear behavior. Similar to the case of Au(111), Emax tends to level off at large biases (see the dashed lines). We then calculated ECB and by subtracting Emax from the energy of tunneling electrons (eV) before the level-off region. The results are shown in Fig. 2(e). It is interesting that both the band edges experience a noticeable shift towards EF when decreasing the bias voltage. It is worth noting that constant current mode was used during the light-emission measurements, so the tip is actually approaching the sample when the bias voltage is decreased.

In order to understand the origin of the band edge shift, we carried out systematic dI/dV measurements for TiO2 surface. A series of STS spectra were taken at the defective area with different tip heights (Fig. 3(a)), showing a distinct bandgap narrowing as the tip height decreases. We plot the energy of conduction and valence band edges as a function of the tip height in Fig. 3(b). It is striking that both edges shift towards EF when the tip is advanced to the surface, which is very similar to the band edge shift observed in Fig. 2(e). In addition, we found the defect states (Ed) also show considerable upward shift as decreasing the tip-surface separation (Figs. 3(c) and 3(d)), suggesting that the apparent shift of the valence band edge in Figs. 3(a) and 3(b) should correspond to the shift of upper energy bound of the defect states ( ).

Fig. 3. (color online) (a) dI/dV spectra taken at the defective TiO2(110) surface with different tip-sample distance in Logarithmic scale. The tip heights refer to the set point of 1.7 V/0.8 nA (0 nm). The spectrum taken at 0.097 nm and 0 nm were offset by 1.2 pA/V and 3.6 pA/V for clarity, respectively. The band edges are denoted by green arrows. (b) Band edge energies derived from the dI/dV spectra at different tip-sample distances. The green double-headed arrows show the energy gap between the conduction band and defect states. (c) dI/dV spectra taken at the same area in linear scale. The tip heights refer to the set point of 1.7 V/0.8 nA (0 nm). The spectrum taken at 0.028 nm and 0 nm were offset by 12 pA/V and 27 pA/V for clarity, respectively. The green arrows indicate the peak position of defect states (Ed). (d) The energy of defect states (Ed) as a function of tip-sample distance.

Such energetic shifts of band edge and defect states might be caused by the tip-induced band bending (TIBB).[20] In low-conductivity semiconductors, the potential between tip and sample drops over not only the vacuum gap but also an extended region in the semiconductor, which gives rise to TIBB. As a result, the dI/dV spectra may exhibit a considerable variation of the apparent bandgap. However, based on TIBB, the bandgap should appear substantially larger at smaller tip heights, which is contrary to our experimental results above. In addition, in our experiment, the EF is very close to ECB, which means that the sample is highly n-doped due to the high concentration of surface and subsurface defects. The high doping level of the sample may lead to efficient screening and thus negligible band bending.[30] Therefore, we can exclude TIBB as the origin of the energy shift.

The most possible origin of the bandgap narrowing is the image charge effect. When a point charge is placed before a metal tip, it induces a polarized cloud of opposite charges in the tip, which in turn lowers the energy of the point charge (Fig. 4(a)). In fact, the image charge effect can significantly renormalize the bandgap of semiconductor or molecular systems, leading to prominent band edge shift towards EF.[23,25] The mechanism is schematically summarized in Fig. 4(b).

Fig. 4. (color online) (a) A positive charge (green) in the sample at a distance z from a metal tip can induce negative image charge (red) in the tip. (b) Schematic of band edge shift in TiO2 as the metal tip approaches the surface arising from the image charge effect. The dash/solid arrow indicates an electron injected into/extracted from the conduction band/defect states of TiO2.

When an electron is injected into the conduction band of TiO2, the induced positive image charge in the tip will lower the electron energy, leading to the downward shift of the conduction band edge (ECB). When the electron tunnels from the defect state to the tip, the hole created in the defect state also feels an attractive interaction from the negative image charge in the tip, which pushes up the defect state (Ed) towards EF. Such a picture nicely explains the observed experimental results of light emission and dI/dV measurements.

4. Conclusion and perspectives

We have investigated the plasmon-enhanced light emission of rutile TiO2(110) surface by a low-temperature STM. The photon emission arises from the irradiative decay of LSP excited by the inelastic electronic transition between the STM tip and the conduction band or defect states of TiO2(110). In contrast to the Au(111) surface, the maximum photon energy is not linear with the sample bias voltage, suggesting that the band edges are shifted towards the EF when the tip height decreases. By performing differential conductance (dI/dV) measurements, we exclude the tip-induced band bending as the origin of band edge shifts. Instead, we propose that such band shifts arise from the image charge effect of the metal tip. This work not only sheds new lights onto the understanding of plasmon-enhanced light emission of semiconductor surfaces, but also opens up a new avenue for engineering the plasmon-mediated interfacial charge transfer in molecular and semiconducting materials.

Reference
[1] Gimzewski J K Sass J K Schlitter R R Schott J 1989 Europhys. Lett. 8 435
[2] Downes A Well M E 1998 Phys. Rev. Lett. 81 1857
[3] Aizpurua J Apell S P Berndt R 2000 Phys. Rev. 62 2065
[4] Qiu X H Nazin G V Ho W 2003 Science 299 542
[5] Dong Z C Zhang X L Gao H Y Luo Y Zhang C Chen L G Zhang R Tao X Zhang Y Yang J L Hou J G 2010 Nat. Photon. 4 50
[6] Krane N Lotze C Lager J M Reecht G Franke K J 2016 Nano Lett. 16 5163
[7] Berndt R Gimzewski J K Johansson P 1991 Phys. Rev. Lett. 67 3796
[8] Suzuki Y Minoda H Yamamoto N 1999 Surf. Sci. 438 297
[9] Dong Z C Guo X L Trifonov A S Dorozhkin P S Miki K Kimura K Yokoyama S Mashiko S 2004 Phys. Rev. Lett. 92 086801
[10] Wu S W Nazin G V Ho W 2008 Phys. Rev. 77 205430
[11] Cavar E Blum M C Pivetta M Patthey F Chergui M Schneider W D 2005 Phys. Rev. Lett. 95 196102
[12] Hoffmann G Kliewer J Berndt R 2001 Phys. Rev. Lett. 87 176803
[13] Nazin G V Qiu X H Ho W 2003 Phys. Rev. Lett. 90 216110
[14] Thirstrup C Sakurai M Stokbro K Aono M 1999 Phys. Rev. Lett. 82 1241
[15] Jackson J D 1975 Classical Electrodynamics 2 New York John Wiley & Sons 490
[16] Yim C M Pang C L Thornton G 2010 Phys. Rev. Lett. 104 036806
[17] Zuo F Wang L Wu T Zhang Z Y Borchardt D Feng P Y 2010 J. Am. Chem. Soc. 132 11856
[18] Naldoni A Allieta M Santangelo S Marelli M Fabbri F Cappelli S Bianchi C L Psaro R Dal Santo V 2012 J. Am. Chem. Soc. 134 7600
[19] Dombrowski R Steinebach C Wittneven C Morgenstern M Wiesendanger R 1999 Phys. Rev. 59 8043
[20] Berndt R Gimzewski J K 1992 Phys. Rev. 45 14095
[21] Mcellistrem M Haase G Chen D Hamers R J 1993 Phys. Rev. Lett. 70 2471
[22] Kaasbjerg K Flensberg K 2008 Nano Lett. 8 3809
[23] Thygesen K S Rubio A 2009 Phys. Rev. Lett. 102 046802
[24] Barr J D Stafford C A Bergfield J P 2012 Phys. Rev. 86 115403
[25] Perrin M L Verzijl C J O Martin C A Shaikh A J Eelkema R van Esch J H van Ruitenbeek J M Thijssen J M van der Zant H S J Dulic D 2013 Nat. Nanotechnol. 8 282
[26] Cooper B R Ehrenreich H Philipp H R 1965 Phys. Rev. 138 A494
[27] Ehrenreich H Philipp H R 1962 Phys. Rev. 128 1622
[28] Minato T Sainoo Y Kim Y Kato H S Aika K Kawai M Zhao J Petek H Huang T He W Wang B Wang Z Zhao Y Yang J L Hou J G 2009 J. Chem. Phys. 130 124502
[29] Wendt S Sprunger P T Lira E Madsen G K H Li Z S Hansen J O Matthiesen J Blekinge-Rasmussen A Laegsgaard E Hammer B Besenbacher F 2008 Science 320 1755
[30] Zhang Z Yates J T 2012 Chem. Rev. 112 5520