BY 4.0 license Open Access Published by De Gruyter September 3, 2021

Electroluminescence in plasmonic actuator based on Au/SiO2/n-Si tunnel junction

Yan Liu ORCID logo, Soo-Jin Chua, Siping Gao, Wenrui Hu and Yongxin Guo
From the journal Nanophotonics

Abstract

A compact electrical source capable of generating surface plasmon polaritons would represent a crucial step for on-chip plasmonic circuitry. The device fabrication of plasmonic actuator based on Au/SiO2/n++Si tunnel junction and performance have been reported in [ACS photonics, 2021, 8, 7, 1951–1960]. This work focuses on the underlying mechanisms of electroluminescence. The n-type Si samples were doped with concentrations ranging from 1.6 × 1015 cm−3 to 1.0 × 1020 cm−3. A low voltage of 1.4 V for intense light emission was achieved at the highest concentration. The electrical/spectral characteristics and energy band diagrams calculation show two distinct behaviors indicating two distinct mechanisms of light emission are at work in the heavily doped versus the lightly doped Si. In the heavily doped case, the light output is correlated to tunneling current and the subsequent conversion of surface plasmons to photons, while that for the lightly doped case is due to indirect band-to-band recombination in silicon. The results are validated by numerical simulation which indicates that the heavy doping of the n++-Si is necessary to achieve surface plasmon generation via electron tunneling due to the presence of band tail states and their effect on lowering the barrier height.

1 Introduction

The past two decades has seen an unprecedented increase in the amount of digital information transmitted all over the world. The rapid growth of computing power and the data transmission speed have changed our way of life and work. An important invention in the information age, the complementary metal-oxide semiconductor (CMOS), is now approaching its fundamental limit on its feature size and interconnect dimension hindering further progress in device integration predicted by Moore’s law. Integration of electronics with photonics is one way forward to achieve more than Moore’s. Plasmonic, as an integration of electronic and photonic device, is considered a very promising technology platform for the development of next generation of chip in the more than Moore’s era [1]. Plasmonic devices are the building blocks for a new generation of ultrafast and ultracompact nanophotonics chip, including active components (plasmonic actuator [2], [3], [4]), plasmonic detector [5], plasmonic modulator [6]) and passive components (plasmonic waveguide [7]). Among these, the plasmonic actuator is one of the most important components in plasmonic circuits serving as the plasmon source interfacing between electronics and photonics. Recent achievements have resulted in the surface plasmon polaritons (SPP) excitation efficiency in metal–insulator–metal (MIM) tunnel junction(TJ) exceeding 1% as reported by W. Du et al. and H. L. Qian et al. [2, 3] and in modulating the emission spectrum as demonstrated by X. B. He [4].

As an active component, the silicon-based plasmonic actuator consisting of a metal–insulator–silicon (MIS) tunnel junction can fulfill the role of current-plasmon conversion and be integrated in Si integrated circuit (IC) mass production. The plasmonic waveguide consisting of metal lines embedded in dielectric films in back-end of line in CMOS technology are suitable for SPPs transmission [8]. Therefore, both active and passive plasmonic components integrated on a silicon chip have received significant attention recently [9]. A silicon-based plasmonic device using SPPs for signal transmission working at optical frequencies has the advantage of small feature size not limited by the diffraction limit of light. The first observation of surface plasmon excitation in degenerate semiconductor [10] by tunneling electrons was traced back to 1969. In 1993, J. Watanabe et al. [11] observed visible light from Si–metal–oxide–semiconductor tunnel junction due to SPP excitation. They claimed that the tunneling current radiated light by scattering from residual surface roughness. A similar piece of work was reported by M. X. Wang in 2000 [12] who applied a very large DC voltage bias (5–7 V) on MIS tunnel junctions to generate surface plasmon and observed light emission. In 2018, H. Goktas [13] observed light emission from a p-type silicon doped with 2 × 1017 cm−3 and proposed that the emission is due to tunneling of accumulated holes generating a combination of hybrid-photon–plasmon mode and a surface plasmon mode. However, the reports mentioned earlier do not have enough data to conclusively deduce that the light generation is due to plasmon–photon outcoupling. For practical application in nanophotonics, there is a need to generate and read out the plasmonic signals by electrical means only [2], without involving any intermediate generation of excitons or photons. Thus, any forms of intermediate process such as generation of hybrid plasmon–photon modes may reduce efficiency of plasmonic devices and increase transmission loss. Here we propose the model for an electrically driven SPP in an Au/SiO2/n++Si tunnel junction which showed plasmon excitation at a low voltage of 1.4 V. The device fabrication of the Au/SiO2/n++Si tunnel junctions and performance have been reported in [ACS photonics, 2021, 8, 7, 1951–1960]. This work focuses on the theoretical analysis of the electroluminescence in Au/SiO2/n++Si junctions with different doping concentrations. The C–V behaviors of metal–insulator–Si (MIS) TJs with different levels of doping concentrations have been presented to help interpret the capacitance contributed by band tail states and highlight the crucial role they play in the electron tunneling process. The computed energy diagrams have been provided to explain the underlying mechanism of electroluminescence in MIS TJ with different levels of doping concentrations. We have also included the simulation of the electroluminescence spectra as a function of applied voltage for different doped samples to compare with the experimental results reported in the earlier ACS photonics paper.

2 Device design and characterizations

A desirable option is to use purely quantum tunneling of electrons across an electrically biased barrier between the n-type silicon and a metal contact (Au) to excite surface plasmon at the gold/dielectric interface. In the process of electron transmission to the metal from silicon, the electrons can either (1) tunnel elastically without any energy loss, or (2) tunnel inelastically imparting the energy to the free electrons in the gold electrode and generating surface plasmons [14], or (3) overshoot the barrier under a high voltage bias i.e. via thermionic emission [15]. Here, we report on the demonstration of an electrically driven, Au/SiO2/n-Si plasmonic actuator based on the inelastic electron tunneling process at room temperature and study the mechanism of light emission from metal–oxide–silicon junctions with n-type doping concentrations varying by four orders of magnitude in silicon substrates. Figure 1a shows the perspective view of the Au/SiO2/n-Si junction, consisting of a heavily doped n-type silicon substrate with a thermally grown SiO2 (th-SiO2) layer of ∼2 nm as a tunneling barrier, a 40-nm-thick gold film on the top of junction area, and a 200-nm-thick gold as a top electrode outside of cross-point junction, as demonstrated in our previous work [16]. The 100-nm-thick oxide layer formed by sputtering (sp-SiO2) covers an area wider than the junction area of ∼5 × 5 μm to ensure that the current detected is obtained via tunneling through the 2 nm SiO2 barrier. The top view of Au/SiO2/n-Si junction was shown in Figure 1b. To facilitate electrical and optical measurements, the whole silicon chip comprising of an array of 6 × 6 tunnel junctions is housed in a ceramic butterfly package with wires bonded for electrical contact. This large number of samples is to ensure the reproducibility of the events observed. Here, the forward current is defined corresponding to majority carriers (electrons) flow from semiconductor to metal, while the reverse current flow is from metal to semiconductor. It is found that the forward current flow is strongly correlated with doping concentration of n-type Si. In the case of heavy doping in n++-Si, the forward electron current flow from n++-Si to Au is dominated by quantum tunneling, accompanied by surface plasmons excitation, as shown in Figure 1c. In the case of lightly doped n-Si (as shown in Figure 1d), the small amount of electron current flow is mainly contributed by majority carriers (electrons) in n-Si tunneling with a low probability through a thin SiO2 barrier to Au, when a large positive voltage (>5 V) is applied on the Au electrode. Meanwhile, a large quantity of electrons accumulates at the n-Si/SiO2 interface resulting in electron–hole recombination in Si via indirect band-to-band transition assisted by phonons. This inefficient recombination process results in few photons being emitted. The heavily doped n++-Si is employed to ensure sufficient band tail states in the band gap of n++-Si to aid the tunneling process through lowering of the oxide barrier. In this work, a positive DC bias, V b is applied on the gold electrode of the Au/SiO2/n++-Si tunnel junction to inject electrons into the metal and voiding any photons generation by electron–hole indirect phonon-assisted recombination in the silicon. The calculated band diagram for this situation is shown in Figure 1e. In quantum tunneling, an electron with a wave function of ψ Si(E) in the conduction band (E fs ≥ E c) with energy of E (≥E fs − E fAu) tunneling through the SiO2 barrier to the gold contact under the positive voltage may result in two possible pathways: (1) elastic tunneling without any energy loss, arriving at the gold contact with the wave function of ψ Au(E). (2) inelastic tunneling losing an energy of ω ( e V b ω ), arriving at gold contact with a wave function of ψ Au ( E ω ) , where is the reduced Planck constant and ω is the angular frequency of the optical mode. In the inelastic tunneling pathway (2), the tunneling electrons may impart the energy ( ω ) to the free electrons in the gold film, resulting in free electrons oscillation, and generating a surface plasmon mode at the Au/SiO2 interface. The light emission originates from the surface plasmon mode outcoupling via a very thin and narrow Au strip with rough edges. The 100-nm-thick oxide layer formed by sputtering (sp-SiO2) covers an area wider than the junction area of ∼5 × 5 μm to ensure that the current detected is obtained via tunneling through the 2 nm SiO2 barrier. In the calculated energy band diagram of Figure 1f, with a low doped n-Si, a very large positive voltage bias (V b = 7.5 V) is needed to be applied on the Au contact to achieve a faint light emission, resulting in the majority carriers (electrons) in conduction band accumulating at n-Si/SiO2 interface. Some of electrons tunnel through with a very low tunneling probability due to a triangle shape SiO2 barrier height of ∼4.25 eV.

Figure 1: 
(a) Schematic of Au/SiO2/n-Si tunnel junction consisting of gold strip electrode (thickness 40 nm and width 5 μm) on the n++-Si substrate (thickness ∼ 500 μm) within a 2-nm-thick thermal SiO2 (“th-SiO2”) in between. Au/SiO2/n-Si plasmons are excited in junction area. “sp-SiO2” stands for sputtered oxide. (b) The top view of MIS tunnel junction (in red frame) with an area of 5 × 5 μm. (c) A schematic of energy diagram at low forward bias of Au/SiO2/n++-Si showing quantum tunneling mechanism dominating the forward current flow and plasmons generation originating from inelastic tunneling. The short horizontal black lines at n++-Si/SiO2 interface stand for band tail states in the silicon band gap. (d) A schematic of Au/SiO2/n-Si to show the mechanism of light emission caused by electron–hole recombination in silicon when a large positive voltage is applied on Au. (e) Calculated energy band diagram of the MIS tunnel junction with a positive voltage bias V
b = 1.4 V on top electrode Au for case of n++Si. The red arrow indicates an inelastic electron tunneling process resulting in surface plasmon excitation with the energy of 



ℏ
ω



$\hslash \omega $



 (wiggly arrow). The tunneling barrier height is φ
B,eff. (f). Calculated energy band diagram of MIS junction with a large positive voltage bias V
b = 7.5 V on top electrode Au. A thin green horizontal arrow represents a small number of electrons tunneling through SiO2 barrier, and the thick green arrow indicates the accumulated electrons at conduction band at the n-Si/SiO2 interface recombining with the holes at the valence band.

Figure 1:

(a) Schematic of Au/SiO2/n-Si tunnel junction consisting of gold strip electrode (thickness 40 nm and width 5 μm) on the n++-Si substrate (thickness ∼ 500 μm) within a 2-nm-thick thermal SiO2 (“th-SiO2”) in between. Au/SiO2/n-Si plasmons are excited in junction area. “sp-SiO2” stands for sputtered oxide. (b) The top view of MIS tunnel junction (in red frame) with an area of 5 × 5 μm. (c) A schematic of energy diagram at low forward bias of Au/SiO2/n++-Si showing quantum tunneling mechanism dominating the forward current flow and plasmons generation originating from inelastic tunneling. The short horizontal black lines at n++-Si/SiO2 interface stand for band tail states in the silicon band gap. (d) A schematic of Au/SiO2/n-Si to show the mechanism of light emission caused by electron–hole recombination in silicon when a large positive voltage is applied on Au. (e) Calculated energy band diagram of the MIS tunnel junction with a positive voltage bias V b = 1.4 V on top electrode Au for case of n++Si. The red arrow indicates an inelastic electron tunneling process resulting in surface plasmon excitation with the energy of ω (wiggly arrow). The tunneling barrier height is φ B,eff. (f). Calculated energy band diagram of MIS junction with a large positive voltage bias V b = 7.5 V on top electrode Au. A thin green horizontal arrow represents a small number of electrons tunneling through SiO2 barrier, and the thick green arrow indicates the accumulated electrons at conduction band at the n-Si/SiO2 interface recombining with the holes at the valence band.

To prove that a surface plasmon mode is indeed excited, the IV characteristics of four types of MIS junctions with different n-doping concentrations (1.0 × 1020 cm−3, 3.6 × 1018 cm−3, 1.9 × 1016 cm−3, and 1.6 × 1015 cm−3) were measured. By applying a positive DC bias (positive voltage on the Au electrode as shown in Figure 1a) across the MIS junction, electrons tunnel through the oxide barrier. Simultaneously, the light emitted from the junction is captured by an inverted microscope for spectral analysis [2]. The intensity of the light output is determined by the external quantum efficiency (η MIS) comprising of the product of internal quantum efficiency (conversion of electrons to plasmons) and the radiation efficiency (conversion of plasmons to photons). Generally, the internal quantum efficiency is strongly correlated to the probability of inelastic tunneling events [17]. With heavy doping concentration in n++-Si, more electrons are available to participate in the tunneling process because of a large concentration of band tail states of n++-Si. At the same time, these charged band tail states at the n++-Si/SiO2 interface act to lower the barrier height through image force bearing in mind that the oxide layer is only 2 nm thick. The rough edges of the tunnel junction may provide a spectrum of k-vectors to facilitate the conversion of surface plasmons to photons. The radiation efficiency improvement has been investigated by K. S. Makarenko et al., and it shows that the roughness of electrode is very important for surface plasmon out coupling into photons [18]. Previous work has shown that quantum efficiency is prohibitively low (10−4–10−7) [1923]. Therefore, improving quantum efficiency in plasmonic device is fundamental to engineering an efficient plasmonic actuator.

3 Results and discussions

The underlying physical mechanism of surface plasmon mode excitation is based on quantum-mechanical inelastic electron tunneling across a barrier. To identify the effect of the density of free carriers on tunneling MIS junctions, on the magnitude of the tunneling current, four different doping concentrations (N d) in n-type Si substrates are used viz. 1.0 × 1020 cm−3, 3.6 × 1018 cm−3, 1.9 × 1016 cm−3, and 1.6 × 1015 cm−3. The fabrication process of the MIS junction is reported in our previous work [16]. IV and CV characterization of Au/SiO2/n-Si junctions were performed using a Keithley 4200A-SCS semiconductor parameters analyzer with multichannel automatic switching modules. Four chips, each with a different doping level, are packaged for the characterization. Each chip is comprised of an array of 6 × 6 junctions. Figure 2a shows the typical IV characteristics over the voltage range from −2 to 2 V, of four MIS junctions with different orders of magnitude in doping concentrations. The IV curve for each doping concentration is very repeatable among the 6 × 6 array of devices in each chip. The approximate symmetry of the IV curves for positive and negative bias voltages of the two samples with N d = 1.0 × 1020 cm−3 and 3.6 × 1018 cm−3 indicates that the free electrons tunneling occurs through the 2-nm SiO2 barrier in both directions with the same ease. In comparison, the asymmetry of the IV curves with respect to voltages of samples with N d = 1.9 × 1016 cm−3 and 1.6 × 1015 cm−3 implies that the accumulation layer and the Schottky barrier mediate differently the electrical conduction path in the positive and negative bias, respectively. Dahlke [24, 25] had investigated an MIS TJ formed by dry oxidation of Si and showed that it has a high density of band tail states. In heavily doped n++-Si, there is a narrowing of bandgap due to band tails. Thus, it appears that band tail states and the narrowing of bandgaps make the IV curves of an MIS tunnel structure more like those of an MIM tunnel structures. The high density of band tail states at the SiO2/n++-Si aids the electrons tunneling from the n++-Si into the conduction band of gold. Here, the band tails smear out the effect of the energy band gap of silicon, giving credence to the use of MIM tunneling theory to analyze MIS tunnel structures with Si heavily doped. In this work, quantum tunneling theory was used to estimate the Au-SiO2 (φ B,eff) barrier height in an MIS tunnel junction with degenerate n-type silicon [26] (Figure S1(a) and (b)). It must be noted that the abundance of band tail states at the SiO2/Au and at SiO2/n++-Si coupled with the small width of the oxide layer results in the lowering (by image force) of the actual barrier height in the oxide to the effective barrier height, φ B,eff. By extracting the tunneling barrier (φ B,eff) in the tunnel MIS junctions, it is found that the higher doping concentration of N d = 1.0 × 1020 cm−3 lowers the tunneling barrier height to φ B,eff = 1.69 eV from the theoretical value of 4.25 eV. This lower barrier height results in a larger tunneling current flow at the positive voltage bias. It is therefore expected that the surface plasmon mode would be excited at a lower electrical bias in highly doped MIS tunnel junction. In the other group of MIS junctions with much lower doping concentrations of N d = 1.9 × 1016 cm−3 and 1.6 × 1015 cm−3, the asymmetrical IV curves with respect to the voltage indicate that the forward current flow is dominated by a low probability tunneling due to SiO2 barrier, whereas the reverse current flow is even lower because of a thicker barrier width comprising of SiO2 layer and Schottky barrier. At 2 V bias, this tunneling current is two orders of magnitudes lower than that in heavily doped samples. At a very high voltage bias (V b ≥ 5.5 V), the accumulated electrons at the conduction band of n-Si recombine with the holes in valence band electrons. This is an indirect radiative recombination in n-Si contributing to the light emission of MIS junctions, as shown in Figure 1d and f. The lower doped n-Si does not have the charged band tail states to lower the barrier height in positive bias and thus impede the electron tunneling from n-Si to Au, resulting in a reduced forward current. More details of electrical characteristics are shown in Supplementary Note 1.

Figure 2: 
Electrical characteristics of MIS junctions in four doping levels in n-Si. (a) I–V curves of MIS junctions with the area of 5 × 5 μm2, (b) C–V curves of MIS junctions with the area of 500 × 500 μm2 at a low frequency of 10 kHz.

Figure 2:

Electrical characteristics of MIS junctions in four doping levels in n-Si. (a) IV curves of MIS junctions with the area of 5 × 5 μm2, (b) CV curves of MIS junctions with the area of 500 × 500 μm2 at a low frequency of 10 kHz.

The quantum-mechanical electrons tunneling in MIS junctions should also be verified by electrical AC measurements [27]. Four capacitances are involved, C ox, C BT, C Dep, and C Acm and they are, respectively, the oxide capacitance, capacitance due to band tail states, capacitance to the depletion layer, and the capacitance due to the accumulation layer on the Si surface. The band tail states capacitance and that at the Si interface (due to either the accumulation layer or the depletion layer) are connected in parallel. The effective capacitances are then connected in series with the oxide capacitance. The CV characteristics of the four MIS junctions with different doping concentrations were measured in the frequency range of 10–100 kHz. We now focus on 10 kHz and explain the low frequency behavior of the CV characteristics in Figure 2b of the MIS junctions measured with the voltage bias sweeping from −2 to 2 V at an AC drive signal of 20 mV. To minimize measurement error, a larger device area of 500 × 500 μm2 was used for the four types of MIS junctions (n-type Si: 1.0 × 1020 cm−3, 3.6 × 1018 cm−3, 1.9 × 1016 cm−3, and 1.6 × 1015 cm−3). The equivalent capacitances at positive voltage bias (C eq+) (Eq. (1)) and negative voltage bias (C eq−) (Eq. (2)) in MIS junctions can be expressed, respectively, as follows,

(1) C eq + = ( C BT + C ACM ) × C ox C BT + C ACM + C ox

(2) C eq = ( C BT + C DEP ) × C ox C B T + C DEP + C ox

where C BT is the capacitance contributed by band tail states when they are fully occupied by electrons at the interface of SiO2/n++-Si under a positive voltage bias, C BT is the capacitance contributed by the band tail states when they are partially occupied by electrons under a negative voltage bias, and both capacitances are only present in the heavily doped n++-Si. The other capacitances which exist in both the heavily and low doped samples are, viz. C ox, contributed by the 2-nm SiO2 capacitance (∼1.73 × 10−2 pF/μm2); C ACM, contributed by the electrons in the accumulation layer at the n-Si surface under a positive voltage bias, and C DEP, the depletion layer capacitance in the n-Si region under a negative voltage bias. The highly doped MIS junction (N d = 1.0 × 1020 cm−3) is measured to have a higher capacitance (7.89 × 10−3 pF/μm2) compared to that of the lower doped level of N d = 3.6 × 1018 cm−3 at 2 V bias (6.20 × 10−3 pF/μm2), as shown in Figure 2b. This is in agreement with the higher density of band tail states of 2.78 × 1012 cm−2 eV−1 for the highly doped sample compared to 1.76 × 1011 cm−2 eV−1 for the less heavily doped sample extracted from AC measurements [28]. This also applies for the measurements conducted at −2 V (The equivalent capacitances are summarized in the Supplementary Note Table S1). The equivalent circuit of Au/SiO2/n++-Si capacitance is demonstrated in Figure S2(a). So, the sum of the two capacitances (∼7.89 × 10−3 pF/μm2), i.e. band tail capacitance (C BT) and accumulation capacitance C ACM dominate at 2 V bias. Thus, (C BT + C ACM) is smaller than C ox, in Eq. (1). And at −2 V bias, (C BT’ + C DEP) is even smaller than C ox in Eq. (2)). C BT is larger than C BT due to the higher charge occupancy. The capacitance contributed by band tail states play a crucial role in electron tunneling process. In the lightly doped MIS junctions with the concentrations of 1.9 × 1016 cm−3 and 1.6 × 1015 cm−3, there are no band tail capacitances (C BT = 0). The equivalent circuit of Au/SiO2/n-Si capacitance is demonstrated in Figure S2(b). Under negative bias (negative voltage on the Au), the equivalent capacitance (C eq−) varies slightly because the capacitance is determined by a large width of depletion layer (C ox > C DEP in Eq. (2)). Under positive voltage bias, the equivalent capacitance (C eq+) increases with bias because the total charge of accumulated electron (C ACM) increases ( C eq + = C ACM × C ox C ACM + C ox in Eq. (1)), as shown in Figure 2b. The equivalent capacitance in heavily doped MIS tunnel junction is approximately 50 times larger than that in lightly doped MIS junction at 2 V bias. The largest equivalent capacitance means the smallest impedance. From the above discussion, it is seen that the band tail states formed in heavily doped MIS junction facilitates for electrons tunneling. A higher doping level, besides generating a larger concentration of free electrons, also creates band tail states which lower the tunneling barrier height. Experimental evidence of the lowering of the tunneling barrier with increase in doping concentration is provided in Supplementary Note 1. The measured values φ B,eff decreases with the heavy doping concentration.

The relationship between the electrical characteristics and optical characteristics in MIS tunnel junction has been a matter of continuous research for half a century [29]. In this work, the electroluminescence (EL) of MIS junctions with different doping concentrations was investigated under positive voltage bias. To interpret the experimental results, an FEM model is provided to explain the spectral characteristics with respect to voltage bias as a function of doping concentration (Supplementary Note 2). Experimentally, the optical images of the emission region of the MIS junctions were captured by an inverted optical microscope (Nikon Eclipse Ti-E) equipped with a spectrometer (Shamrock 303i) and an electron multiplying CCD (EMCCD, iXon Ultra 897). The recorded spectra of plasmon–photon conversion show the same blueshift characteristics as reported in our previous work in [2]. To increase the numerical objective, an oil immersion objective was used (NA = 1.49). The images of light emission regions (Figure 3a insets) are captured by an EMCCD with 2 min integration time and an EM gain of 300. To eliminate the emission due to interband transition in gold film, the voltage bias, V b, is limited to below 2.4 V [30]. Here, we only discuss the results of positive biases on the MIS TJs to excite SPP on Au film. Negative biases did not result in light emission as the electrons tunneling into the n++-Si is ineffective to excite bulk plasmons due to the large thickness of the substrate. The spectra are recorded for samples with different doping concentrations viz. (a) 1.0 × 1020 cm−3, (b) 3.6 × 1018 cm−3, and (c) 1.9 × 1016 cm−3 in our previous work [16]. The spectra of sample (a) show a voltage dependent peak blueshifts (voltages ranging from 1.4 to 1.9 V in Figure 3b) like the spectra reported by W. Du [2]. There is an insignificant spectral peak blueshifts (voltages ranging from 1.5 to 2.5 V) in sample (b), while no peak blueshifts are observed in sample (c). The light emission from the sample (d) with the lowest concentration of 1.6 × 1015 cm−3 was very weak and cannot be captured reliably by the instrument.

Figure 3: 
(a) Identification of surface plasmons conversion to photon as observed in electroluminescence experiment of the MIS tunnel junction with the highest doping concentration of 1.0 × 1020 cm−3. The insets show the EMCCD images in real plane corresponding to different voltage bias (V
b) [16]. λ
cut-off is the shortest wavelength emitted. (b) The peaks of spectrum blueshift from 985.5 to 883.4 nm, while the intensities of the peaks increase from 6.14 to 262.40 (a.u) with the voltage bias increasing from 1.4 to 1.9 V.

Figure 3:

(a) Identification of surface plasmons conversion to photon as observed in electroluminescence experiment of the MIS tunnel junction with the highest doping concentration of 1.0 × 1020 cm−3. The insets show the EMCCD images in real plane corresponding to different voltage bias (V b) [16]. λ cut-off is the shortest wavelength emitted. (b) The peaks of spectrum blueshift from 985.5 to 883.4 nm, while the intensities of the peaks increase from 6.14 to 262.40 (a.u) with the voltage bias increasing from 1.4 to 1.9 V.

For an MIS tunnel junction with the highest doping concentration of N d = 1.0 × 1020 cm−3, the round dots in Figure 3a give the shortest wavelength or cut-off wavelength (λ cut-off) in nm of the emission (corresponding to the highest photon energy). The square dots give the corresponding cut-off energy in eV. The cut-off energy of the light emission spectrum in an Au/SiO2/n++-Si tunnel junction increases linearly with the voltage bias [16]. Here we have provided the additional information of light emission intensity in six insets as well and illustrated the electroluminescence intensity increases with the voltage bias intuitively. The main features in Figure 3a and b are: with increase in voltage (1) the cut-off wavelength (λ cut-off) decreases from 885.7 nm (V b = 1.4 V) to 652.6 nm (V b = 1.9 V), (2) the intensity of light emission at each voltage bias increases from 6.14 a.u (V b = 1.4 V) to 262.40 a.u (V b = 1.9 V) ,and (3) the peak of the spectrum blue shifts from 985.5 nm (V b = 1.4 V) to 883.4 nm (V b = 1.9 V). Feature (1) is an expression of the conservation of energy i.e. the plasmon energy ( ω ) is less than or equal to the electron excitation energy (eV b), while feature (2) indicates that the higher plasmon intensity is obtained at a higher tunneling current corresponding to a higher voltage bias. The blue shift in the emission spectrum seen in feature (3) with increasing bias is an important feature of surface plasmon excitation which will be explained in the theoretical model to be presented later. From the six EMCCD images shown in the insets of Figure 3a, the emission from the surface of the Au film is relatively weak with high intensities distributed at the edges of the junction area in the lower voltage range from 1.4 to 1.7 V. This is due to the higher photon out-coupling from the localized surface plasmons via the rough edges of the Au strip [31]. At 1.9 V, the whole junction area is lighted up brightly due to the surface plasmon outcoupling through the 40-nm thin Au film as well [18]. It is worthy to note that at the highest doping concentration of 1.0 × 1020 cm−3, plasmons can be excited at a low voltage bias of 1.4 V in the MIS tunnel junction, which is much lower than the higher voltage bias ranging from 2 to 8 V reported by other groups [12, 13, 3241].

Figure 4a and b shows the plots of the η MIS, defined as the quantity of emitted surface plasmon polaritons per tunneling electron and per unit spectral width (Supplementary Note 2), which correspond to samples (a) and (b). At a given bias, η MIS is proportional to the plasmon intensity and in turn gives rise to the light intensity assuming the out-coupling efficiency is independent of wavelength. The method of calculating η MIS is as per reference [42] but by replacing the MIM junction with an MIS junction. This requires changing the optical parameters such as refractive indices of the metal with that of the n-type Si which is doping concentration dependent. It is seen that the simulated results of the spectra at different bias match the experimental results qualitatively. The detail of the simulation is shown in Supplementary Note 2. They show the same features as described earlier viz. the peak wavelength blue shifts with bias for the heavily doped sample while the low doped sample has a constant peak wavelength corresponding to the bandgap energy. For the lightly doped MIS junction (Sample (c)), the light emission is based on phonon-assisted indirect radiative recombination in n-Si. Hence, the normalized electroluminescence efficiency of MIS junction with a doping concentration of 1.9 × 1016 cm−3 is calculated by FEM model as shown in Figure 4c.

Figure 4: 
The external quantum efficiency, η
MIS, of (a) N
d = 1.0 × 1020 cm−3, (b) N
d = 3.6 × 1018 cm−3 and the electroluminescence efficiency of (c) N
d = 1.9 × 1016 cm−3 from FEM models. η
MIS can be interpreted to be proportional to the emitted photon flux.

Figure 4:

The external quantum efficiency, η MIS, of (a) N d = 1.0 × 1020 cm−3, (b) N d = 3.6 × 1018 cm−3 and the electroluminescence efficiency of (c) N d = 1.9 × 1016 cm−3 from FEM models. η MIS can be interpreted to be proportional to the emitted photon flux.

To analyze the mechanisms of electrical conduction and the phenomenon of light emission in the MIS junctions, the energy band diagram calculations (TCAD_Sentaurus_2014.09-SP1) for Au/SiO2/n-Si structures are indispensable. Figure 5a–d shows the simulation results of the energy band (with energy levels E 0E c E f E v ) in Au/SiO2/n-Si structure under zero and positive bias. To understand the blueshift in the spectra with voltage bias, in the samples with high doping concentrations, the band diagrams shown in Figure 5a and b are used. It is seen in both Figure 5a and b that the emission energy (E cut-off) increases with bias as indicated by the longer vertical blue arrows with solid lines on the metal side. And all the electrons located at band tail states (short horizontal black lines) in the band gap of n++-Si can tunnel through the barrier resulting in plasmons with energy as indicated by the vertical dotted blue arrow. The band tail states provide tunneling electrons at the energy less than the energy provided by the bias voltage. For the case of low doping concentration in n-Si as seen in Figure 5c and d, the emission comes mainly from the phonon-assisted indirect radiative recombination between electrons in the accumulation layer and holes in the valence band at the n-Si surface [43, 44]. Thus, the emission energy does not change with voltage bias, which is confirmed by the fixed value of the peaks (λ = 1100 nm) in sample (c). The blue-shift effect in the light emission spectra originating from surface-plasmon excitation via band tail states tunneling in Au/SiO2/n++-Si junction with a high doping concentration (N d ≥ 1018 cm−3), are reflected in Figure 5a and b. However, the light emissions coming from the MIS junctions with a low doping level (below 1018 cm−3) is due to phonon-assisted indirect band-to-band recombination, as indicated in the insets of Figure 5c and d. The weak emission from such recombination does not result in any shift in the peak of the spectra with voltage bias.

Figure 5: 
Energy band diagrams of Au/SiO2/n-Si junctions at different voltage bias (faint line for the vacuum level is for V
b = 0, red line is for the higher voltage) (a) 1.0 × 1020 cm−3, (b) 3.6 × 1018 cm−3, (c) 1.9 × 1016 cm−3, (d) 1.6 × 1015 cm−3 in n-Si. The insets of c and d are the zoom-in at the interface of SiO2/n-Si.

Figure 5:

Energy band diagrams of Au/SiO2/n-Si junctions at different voltage bias (faint line for the vacuum level is for V b = 0, red line is for the higher voltage) (a) 1.0 × 1020 cm−3, (b) 3.6 × 1018 cm−3, (c) 1.9 × 1016 cm−3, (d) 1.6 × 1015 cm−3 in n-Si. The insets of c and d are the zoom-in at the interface of SiO2/n-Si.

4 Conclusions

In conclusion, we have theoretically demonstrated in this paper that surface plasmons can be excited at a low voltage bias by electron tunneling in an MIS structure Au/SiO2/n++-Si is attributed to the high density of band tail states in degenerated Si. The high density of band tail states in heavily doped Si (1.0 × 1020 cm−3) aids the electron tunneling by lowering the barrier height, results in a lower voltage for tunneling and a larger tunneling current at a given voltage bias. Evidence of band tail states are obtained by CV measurement. For samples with low doping concentration, there are no band tail states to help lower the tunneling barrier height. At a voltage bias of 2 V, the low doped samples give a tunneling current two orders of magnitude lower than that in the heavily doped sample. With low tunneling current at forward bias, phonon assisted indirect band-to-band recombination takes place between the accumulation layer in conduction band and valence band of n-Si. Therefore, the emitted light originating from the semiconductor is weak and the emitted wavelength does not shift with voltage bias being pegged to the bandgap energy of n-Si. This work paves the way for using MIS junction as a plasmonic actuator in integrated circuits. Given the small size of the plasmon actuator, it can be incorporated in front-end of line (FEOL). The electronic wires in Metal 1 layer can then be replaced with denser plasmonic waveguides thus squeezing more out of Moore’s law.


Corresponding author: Soo-Jin Chua, Department of Electrical and Computer Engineering, 4 Engineering Drive 3, National University of Singapore, Singapore 117576, Singapore; and LEES Program, Singapore-MIT Alliance for Research & Technology (SMART), Singapore 138602, Singapore, E-mail:

Funding source: National Research Foundation SingaporeNational Research Foundation Singapore

Award Identifier / Grant number: NRF-CRP17-2017-08

Award Identifier / Grant number: Unassigned

Acknowledgments

We thank Prof Christian A. Nijhuis, Mr. Fangwei Wang (Ph. D student in Department of Chemistry), Dr. Lu Ding (IMRE, A-star), Dr. Dongyang Wan (Department of Physics, NUS) and Dr. Yu Cao (Department of ECE, NUS) who have kindly provided the constructive suggestions in devices fabrication and improvement.

  1. Author contribution: Y. Liu conceived the experiments and simulations, analyzed the data, the organized the manuscript. S. P. Gao and W. R. Hu assisted in electrical characteristics, Y. X. Guo initiate the research and S. J. Chua verified the experimental data and simulation results and assisted in their interpretation and in the manuscript preparation.

  2. Research funding: We acknowledge the support from the National Research Foundation Singapore project of Integration of Electrically Driven Plasmonic Components in High Speed (NRF-CRP17-2017-08).

  3. Conflict of interest statement: The authors declare no conflicts of interest regarding this article.

References

[1] R. J. Martin-Palma and J. M. Martinez-Duart, Nanotechnology for Microelectronics and Photonics, 2nd ed., Kidlington, UK, Elsevier, 2017, ISBN 978-0-323-46176-4.Search in Google Scholar

[2] W. Du, T. Wang, H.-S. Chu, and C. A. Nijhuis, “Highly efficient on-chip direct electronic–plasmonic transducers,” Nat. Photonics, vol. 11, no. 10, pp. 623–627, 2017. https://doi.org/10.1038/s41566-017-0003-5.Search in Google Scholar

[3] H. Qian, S.-W. Hsu, K. Gurunatha, et al.., “Efficient light generation from enhanced inelastic electron tunneling,” Nat. Photonics, vol. 12, pp. 485–488, 2018. https://doi.org/10.1038/s41566-018-0216-2.Search in Google Scholar

[4] X. B. He, J. B. Tang, H. T. Hu, et al.., “Electrically driven highly tunable cavity plasmons,” ACS Photonics, vol. 6, no. 4, pp. 823–829, 2019. https://doi.org/10.1021/acsphotonics.8b01620.Search in Google Scholar

[5] E. Ho-Seok, Y.-S. No, J. Kim, H.-G. Park, and M.-K. Seo, “Long-range surface plasmon polariton detection with a graphene photodetector,” Opt. Lett., vol. 43, no. 12, pp. 2889–2892, 2018. https://doi.org/10.1364/OL.43.002889.Search in Google Scholar

[6] A. Masafumi, F. Yuriy, W. Heni, et al.., “A high-speed, small-footprint plasmonic modulator is fabricated from a single layer of gold,” Science, vol. 358, no. 6363, pp. 630–632, 2017.Search in Google Scholar

[7] Y. Fang and M. Sun, “Nanoplasmonic waveguides: towards application in integration nanophotonic circuits,” Light Sci. Appl., vol. 4, no. e294, pp. 1–11, 2015. https://doi.org/10.1038/lsa.2015.67.Search in Google Scholar

[8] Y. Liu, L. Ding, Y. Cao, et al.., “The deisgn of CMOS-compatible plasmonic waveguide for intra-chip communication,” IEEE Photonics J., vol. 12, no. 5, pp. 1–11, 2020, 4800810. https://doi.org/10.1109/jphot.2020.3024119.Search in Google Scholar

[9] K. Roy, B. Jung, D. Peroulis, and A. Raghunathan, “Integrated systems in the more-than-moore era: designing low-cost energy-efficient systems using heterogeneous components,” IEEE Des. Test, vol. 33, no. 3, pp. 56–65, 2016. https://doi.org/10.1109/mdt.2011.49.Search in Google Scholar

[10] D. C. Tsui, “Observation of surface plasmon excitation by tunneling electrons in GaAs-Pd tunnel junctions,” Phys. Rev. Lett., vol. 22, no. 7, pp. 293–295, 1969. https://doi.org/10.1103/physrevlett.22.293.Search in Google Scholar

[11] J. Watanabe, Y. Uehara, J. Murota, and S. Ushioda, “Light emission from Si-Metal-Oxide-Semiconductor tunnel junctions,” Jpn. J. Appl. Phys., vol. 32, no. 1A, pp. 99–104, 1993. https://doi.org/10.1143/jjap.32.99.Search in Google Scholar

[12] M. X. Wang, J. H. Yu, and C. X. Sun, “Light emission characteristics and negative resistance phenomenon of Si-based metal/insulator/semiconductor tunnel junction,” Appl. Surf. Sci., vol. 161, pp. 9–13, 2000. https://doi.org/10.1016/s0169-4332(00)00018-0.Search in Google Scholar

[13] H. Goktas, F. S. Gokhan, and V. J. Sorger, “Electrical-driven plasmon source of silicon based on quantum tunneling,” ACS Photonics, vol. 5, no. 12, pp. 4928–4936, 2018.Search in Google Scholar

[14] K. H. Gundlach, “Theory of metal-insulator-metal tunneling for a simple two-band model,” J. Appl. Phys., vol. 44, no. 11, pp. 5005–5010, 1973. https://doi.org/10.1063/1.1662078.Search in Google Scholar

[15] G. T. Nicholas, Metal-Insulator-Semiconductor Tunnel Junctions and Their Application to Photovoltaic Energy Conversion, Ph. D Thesis, The University of British Columbia, 1981, pp. 64–98.Search in Google Scholar

[16] F. Wang, Y. Liu, T. J. Duffin, et al.., “A silicon-based quantum mechanical tunnel junction for plasmon excitation from low-energy tunneling electrons,” ACS Photonics, vol. 8, no. 7, pp. 1951–1960, 2021. https://doi.org/10.1021/acsphotonics.0c01913.Search in Google Scholar

[17] S. L. McCarthy and J. Lambe, “Enhancement of light emission from metal– insulator–metal tunnel junctions,” Appl. Phys. Lett., vol. 30, no. 8, pp. 427–429, 1977. https://doi.org/10.1063/1.89434.Search in Google Scholar

[18] K. S. Makarenko, T. J. D. Thanh Xuan Hoang, A. Radulescu, et al.., “Efficient surface plasmon polariton excitation and control over outcoupling mechanism in metal-insulator-metal tunneling junctions,” Adv. Sci., vol. 7, no. 1900291, pp. 1–9, 2020. https://doi.org/10.1002/advs.201900291.Search in Google Scholar

[19] M. Parzefall, P. Bharadwaj, A. Jain, T. Taniguchi, K. Watanabe, and L. Novotny, “Antenna-coupled photon emission from hexagonal boron nitride tunnel junctions,” Nat. Nanotechnol., vol. 10, pp. 1058–1064, 2015. https://doi.org/10.1038/nnano.2015.203.Search in Google Scholar

[20] J. Lambe and S. L. McCarthy, “Light emission from inelastic electron tunneling,” Phys. Rev. Lett., vol. 37, pp. 923–925, 1976. https://doi.org/10.1103/physrevlett.37.923.Search in Google Scholar

[21] P. Dawson, D. G. Walmsley, H. A. Quinn, and A. J. L. Ferguson, “Observation and explanation of light-emission spectra from statistically rough Cu, Ag, and Au tunnel junctions,” Phys. Rev. B, vol. 30, pp. 3164–3178, 1984. https://doi.org/10.1103/physrevb.30.3164.Search in Google Scholar

[22] J. Kirtley, T. N. Theis, and J. C. Tsang, “Light emission from tunnel junctions on gratings,” Phys. Rev. B, vol. 24, pp. 5650–5663, 1981. https://doi.org/10.1103/physrevb.24.5650.Search in Google Scholar

[23] M. Buret, A. V. Uskov, J. Dellinger, et al.., “Spontaneous hot-electron light emission from electron-fed optical antennas,” Nano Lett, vol. 15, pp. 5811–5818, 2015. https://doi.org/10.1021/acs.nanolett.5b01861.Search in Google Scholar

[24] W. E. Dahlke, “Tunneling into interface states of MOS structures,” Appl. Phys. Lett., vol. 10, no. 10, pp. 261–262, 1967. https://doi.org/10.1063/1.1754801.Search in Google Scholar

[25] W. E. Dahlke and S. M. Sze, “Tunneling in metal-oxide-silicon structures,” Solid State Electron., vol. 10, no. 8, pp. 865–873, 1967. https://doi.org/10.1016/0038-1101(67)90169-4.Search in Google Scholar

[26] L. A. Kasprzak, “Dependence of the Si-SiO2 barrier height on SiO2 thickness in MOS tunnel structure,” J. Appl. Phys., vol. 48, no. 10, pp. 4281–4286, 1977. https://doi.org/10.1063/1.323415.Search in Google Scholar

[27] S. Hlali, N. Hizem, and A. Kalboussi, “Investigation of capacitance characteristics in metal/high-k semiconductor devices at different parameters and with and without interface state density (traps),” Bull. Mater. Sci., vol. 40, no. 5, pp. 1035–1041, 2017. https://doi.org/10.1007/s12034-017-1443-8.Search in Google Scholar

[28] H. Doğan, N. Yıldırım, İ. Orak, S. Elagöz, and A. Turut, “Capacitance-conductance-frequency characteristics of Au/Ni/n-GaN/undoped GaN Structures,” Phys. B Condens. Matter, vol. 457, pp. 48–53, 2015.Search in Google Scholar

[29] W. A. Harrison, “Tunneling from an independent-particle point of view,” Phys. Rev., vol. 123, no. 1, pp. 85–89, 1961. https://doi.org/10.1103/physrev.123.85.Search in Google Scholar

[30] G. V. Naik, M. S. Vladimir, and A. Boltasseva, “Alternative plasmonic materials: beyond gold and silver,” Adv. Mater., vol. 25, pp. 3264–3294, 2013. https://doi.org/10.1002/adma.201205076.Search in Google Scholar

[31] D. Lu, J. J. Kan, E. E. Fullerton, and Z. Liu, “Enhancing spontaneous emission rates of molecules using nanopatterned multilayer hyperbolic metamaterials,” Nat. Nanotechnol., vol. 9, pp. 48–53, 2014. https://doi.org/10.1038/nnano.2013.276.Search in Google Scholar

[32] S. T. Chang, K. F. Chen, C. R. Shie, C. W. Liu, M.-J. Chen, and C.-F. Liu, “The band-edge light emission from the metal-oxide-silicon tunneling diode on (110) substrates,” Solid State Electron., vol. 46, pp. 1113–1116, 2002. https://doi.org/10.1016/s0038-1101(02)00051-5.Search in Google Scholar

[33] I. V. Gekhov, A. F. Shulekin, and M. I. Vexler, “In tunnel MIS junction,” in MIEL conf. Proc, 1997, pp. 165–168. 21st.Search in Google Scholar

[34] C. W. Liu, M. H. Lee, C. F. Lin, W. T. Liu, and H. H. Liu, “Light emission and detection by metal oxide silicon tunneling diodes,” IEDM Conf. Proc., pp. 749–752, 1999.Search in Google Scholar

[35] C. W. Liu, W. T. Liu, M. H. Lee, W. S. Kuo, and B. C. Hsu, “A novel photodetector using MOS tunneling structures,” IEEE Electron. Device Lett., vol. 21, no. 6, pp. 307–309, 2000. https://doi.org/10.1109/55.843159.Search in Google Scholar

[36] C. W. Liu, M. H. Lee, M.-J. Chen, I. C. Lin, and C.-F. Liu, “Room-temperature electroluminescence from electron-hole plasmas in the metal–oxide–silicon tunneling diodes,” Appl. Phys. Lett., vol. 76, no. 12, pp. 1516–1518, 2000. https://doi.org/10.1063/1.126081.Search in Google Scholar

[37] C. W. Liu, S. T. Chang, W. T. Liu, M.-J. Chen, and C.-F. Lin, “Hot carrier recombination model of visible electroluminescence from metal–oxide–silicon tunneling diodes,” Appl. Phys. Lett., vol. 77, no. 26, pp. 4347–4349, 2000. https://doi.org/10.1063/1.1332825.Search in Google Scholar

[38] T. Matsuda, M. Nishio, T. Ohzone, and H. Hori, “Electroluminescence of MOS capacitors with Si-implanted SiO2,” Solid-State Electron., vol. 41, no. 6, pp. 887–893, 1997. https://doi.org/10.1016/s0038-1101(97)00025-7.Search in Google Scholar

[39] G. Z. Ran, J. S. Fu, W. C. Qin, B. R. Zhang, Y. P. Qiao, and G. G. Qin, “Blue to red electroluminescence from Au/native silicon oxide/p-Si structure subjected to rapid thermal annealing,” Thin Solid Films, vol. 388, nos 1-2, pp. 213–216, 2001. https://doi.org/10.1016/s0040-6090(00)01921-0.Search in Google Scholar

[40] J. Seidel, T. Gohler, S. Grafstrom, and L. M. Eng, “Near-field optical characterization of surface-plasmon-mediated light emission from electrically biased metal-insulator-semiconductor tunnel junctions,” Appl. Phys. Lett., vol. 92, no. 103123, pp. 1–3, 2008. https://doi.org/10.1063/1.2898201.Search in Google Scholar

[41] M. X. Wang, J. H. Yu, and C. X Sun, “Light emission characteristics and negative resistance phenomenon of Si-based metal/insulator/semiconductor tunnel junction,” Appl. Surf. Sci., vol. 161, nos 1–2, pp. 9–13, 2000. https://doi.org/10.1016/s0169-4332(00)00018-0.Search in Google Scholar

[42] M. Parzefall and L. Novotny, “Light at the end of tunnel,” ACS Photonics, vol. 5, no. 11, pp. 4195–4202, 2018. https://doi.org/10.1021/acsphotonics.8b00726.Search in Google Scholar

[43] S. Imhof and A. Thranhardt, “Phonon-assisted transitions and optical gain in indirect semiconductors,” Phys. Rev. B, vol. 82, no. 085303, pp. 1–6, 2010. https://doi.org/10.1103/physrevb.82.085303.Search in Google Scholar

[44] L. Pavesi, “Routes toward silicon-based lasers,” Mater. Today, vol. 8, no. 1, pp. 18–25, 2005. https://doi.org/10.1016/s1369-7021(04)00675-3.Search in Google Scholar

Supplementary Material

The online version of this article offers supplementary material (https://doi.org/10.1515/nanoph-2021-0287).

Received: 2021-06-07
Accepted: 2021-08-20
Published Online: 2021-09-03

© 2021 Yan Liu et al., published by De Gruyter, Berlin/Boston

This work is licensed under the Creative Commons Attribution 4.0 International License.