Plasmonics Yields Efficient Electron Transport via Assembly of Shell-Insulated Au Nanoparticles.

Junctions built from metallic nanoparticles (NPs) can circumvent the diffraction limit and combine molecular/nanoelectronics with plasmonics. However, experimental advances in plasmon-assisted electron transport at the nanoscale have been limited. We construct junctions of a robust, molecule-free, suspended film, built solely from AuNPs, capped by SiO2 shells (Au@SiO2), which give insulating tunneling gaps up to 3.6 nm between the NPs. Current measured across monolayers of such AuNPs shows ultra-long-range, plasmon-enabled electron transport (P-transport), beyond the range of normal electron tunneling across insulators. This finding challenges the present understanding of electron transport in such systems and opens possibilities for future combinations of plasmonics and nanoelectronics.

Introduction

Molecular electronics devices, by using individual molecules as active electronic components (Nitzan and Ratner, 2003) to miniaturize conventional electronic components, hold great promise for electronics applications. However, such tiny devices, with current transport that can be described by accepted electron tunneling theory (Figure 1A), are mostly not robust for reliable practical uses as they often suffer from contact artifacts (Flood et al., 2004, Lau et al., 2004). Therefore, until it matures into a practical technology, alternative solutions to miniaturize devices and/or add functional logic beyond the existing binary ones are desired. Future practical devices may well consist of hybrid devices that combine molecules with nanoelectronics (Cui and Lieber, 2001) and require basic understanding of electron transport at the nanoscale.

Figure 1

Schematic Illustrations of the Classic Quantum Tunneling and Plasmon-Enabled Long-Range Electron Transport (P-transport) Regime

(A–C) In the P-transport regime (B and C), the inherent “built-in” strong plasmon field (in dark or bright plasmon modes) will facilitate electron transport over lengths beyond that of normal quantum tunneling across an insulating gap (A).

Plasmonics, defined as light-metal interactions via coupling with the conduction electrons at the metal-dielectric interface (Fang and Zhu, 2013, Jiang et al., 2014, Ozbay, 2006), converts light into propagating electrical signals and, therefore, merges photonics with electronics at the nanoscale and provides a possible path to next-generation (opto)electronic devices. In the past two decades, the plasmon resonances of metallic nanoparticles (NPs) have been extensively explored for sensitive biosensing and nanomedicine (Anker et al., 2008, Jin et al., 2010), extreme light concentration and manipulation in nanophotonics (Schuller et al., 2010), plasmonic nanochemistry (Baffou and Quidant, 2014), and solar energy harvesting and improved photovoltaic devices (Atwater and Polman, 2010). However, the development of plasmonics for nanoelectronics has been severely hindered by lack of fundamental understanding of plasmon-electron coupling; therefore current studies mainly focus on chemoelectronic circuits (Yan et al., 2016, Schlicke et al., 2015). In principle, plasmons in AuNPs generate an enhanced local electromagnetic field (when compared with Au atoms or bulk Au surfaces) in the “dark” mode; the collective oscillation of the conduction electrons of a AuNP will couple with incident light, upon illumination, to excite the plasmonic “light” mode and enhance further the field. As schematically illustrated in Figures 1B and 1C, when AuNPs, covered with a nanometer-scale insulating gap, are network linked, such inherent “built-in” strong plasmon field and the “dark/light” plasmon modes noted above may have distinct effects on electron transport, lowering barriers and facilitating electron transport across a gap that is larger than what is possible by accepted electron tunneling theory. This might be possible when electron transport meets with “plasmonics,” because the extra energy of “hot” electrons (compared with thermally equilibrated ones) makes them more mobile than expected (Najafi et al., 2017). Recently, tunneling plasmons have been probed optically (Savage et al., 2012) and direct observation of, and control over, quantum plasmon resonances at 0.4–1.3 nm length scales across molecular tunnel junctions made of two AgNPs has been achieved (Tan et al., 2014). Very recently, we revealed a long-range plasmon field and plasmoelectronic effect on catalysis (Li et al., 2017). However, reliable direct electrical probing of plasmonic effects on long-range electron transport in such nanosystems was not yet achieved and remains a great challenge in this field.

Previously we prepared a simple AuNP monolayer-based planar molecular tunnel junction and observed a plasmon effect, along with AuNP-enhanced current transport through nanometer-scale insulating layers (Jin et al., 2006, Jin and Friedman, 2005). However, such monolayer-based sandwich-type junction devices were not robust enough to rule out possible contact artifacts due to lightning rod effect (Shpaisman et al., 2012) or partially, shorting.

Herein, to eliminate contact artifacts and test the possible plasmonic effects on long-range electron transport, we construct a robust, molecule-free, suspended-film-type plasmonic circuit, based on a mechanically stable enough nanomembrane, made up of a monolayer of pure Au@SiO2 core-shell NPs (acting as an active component), by bridging it over two micrometer-gap electrodes. This device configuration provides multiple advantages. First, the suspended-film-type junction layout can not only effectively eliminate substrate effects and contact artifacts, rendering the current-carrying device robust to measure and uncover possible plasmonic effects, but also facilitates illumination of the device to study light-induced effects. Second, no additional molecules are introduced into the junction, which can rule out molecule effects and ensures that the observed current and light effects are indeed plasmonic in origin. Third, AuNPs isolated by a nanoscale insulating silica shell can effectively suppress the electronic excitation of plasmons by avoiding direct electric contact between electrodes and AuNPs, which is inevitable if using bare AuNPs in the current-carrying device, making the detection of “real” dark plasmon modes possible. This platform therefore allows us to explore P-transport systematically.

Results

Fabrication and Morphology Analysis of NOE

As depicted in Figures 2A and 2B, the junctions were fabricated by transferring a freshly prepared monolayered nanomembrane gently, via floating from the “soft” air-water interface onto micrometer-gapped Au trench electrodes, as described previously (Wu et al., 2016). The nanomembrane is made of uniform Au@SiO2 core-shell NPs with Au core diameter of ∼12 ± 1.2 nm and homogeneous silica shell thickness of ∼1.8 ± 0.5 nm (mean ± TEM, see Figure S1 for detailed characterizations) and was prepared by the method of liquid/liquid interface self-assembly (Shin et al., 2015, Gauvin et al., 2016, please see “Transparent Methods” in the Supplemental Information). Typically, 3 mL colloidal Au@SiO2 NPs with compact silica shell (prepared and well-characterized according to our previous report (Li et al., 2017, please see “Transparent Methods” in the Supplemental Information) were poured into a plastic container, and 460 μL hexane was added to the solution to form a liquid/liquid interface; then 3.7 mL methanol was poured into the mixture rapidly to capture the NPs at the hexane/water interface. After evaporation of the hexane, the NPs were simultaneously self-assembled into monolayer nanomembranes over a large area at the water/hexane interface and can be seen with the naked eyes (Figure S2A). The freshly prepared nanomembranes were then transferred onto transmission electron microscopic (TEM) grids and Au trench electrodes for TEM and current-voltage (I-V) measurements, respectively. As clearly seen from the gradually zoomed-in TEM images in Figures 2C–2E (Figure S2B shows a higher resolution TEM image of a nanomembrane), the resulting monolayer nanomembranes have AuNPs “jammed” in close contact, but nearly all are well-separated by a transparent gap of ∼3.0 ± 0.4 nm (Figure 2F, mean ± TEM). As this gap almost equals twice the silica shell thickness of the NPs, we identify this gap as silica. No “fatal” metal (Au) junction or filament formation between adjacent NPs was observed by direct TEM observation (Figure 2E); spherical aberration-corrected scanning TEM elemental line scan analysis (Figure 2G) indicates the stability and shell-isolated nature of the AuNPs. The optical properties of the NPs and the resulting nanomembrane were characterized by UV-visible spectra and microscopy-based selected-area bright field extinction spectra. The plasmon band of the nanomembrane red-shifted from ∼519 to 590 nm (Figure 2H) when compared with free AuNPs, which is explained by electromagnetic coupling of the NPs (Jain et al., 2007).

Figure 2

Experimental Device and Nanomembrane Characterizations

(A and B) (A) Schematics of the preparation of the shell-isolated AuNP-based nanomembrane-on-electrode (NOE) via a floating transfer process and (B) the resulting NOE junction configuration.

(C–F) (C–E) Zoomed-in TEM images and (F) gap distance analysis between AuNPs of the monolayered 12 nm Au@1.8 nm SiO2 nanomembrane, statistically obtained from more than 200 NP pairs. Data are represented as mean ± TEM.

(G) Spherical aberration-corrected scanning TEM image and corresponding (Au) elemental line scan analysis of three adjacent NPs in the 12 nm Au@1.8 nm SiO2 nanomembrane.

(H) UV-visible spectra of 12 nm AuNPs (black line) and 12 nm Au@1.8 nm SiO2NPs (red line) in solutions and microscopy-based selected-area bright field extinction spectra of the monolayered 12 nm Au@1.8 nm SiO2 nanomembrane (blue line).

(I) Typical scanning electron micrograph of the junction.

(J) AFM line scan height profile analysis of the NOE and device. The nanomembrane was maintained intact and suspended bridging over the trench electrodes.

See also Figures S1–S3.

Figure 2I (and Figure S3) shows a typical scanning electron microscopic image of the junction. Impressively, the ultrathin monolayered nanomembrane (∼15 nm thick, as measured by atomic force microscopy [AFM]) bridges over the 100-μm gap between the Au trench electrodes, denoted later as nanomembrane-on-electrode (NOE), remaining intact after the floating transfer to prepare the device. The mechanical stiffness and suspended-film geometry of the NOE were further confirmed by AFM line scan height profile analysis of the device (Figure 2J). The robustness of the junction allows making reliable electrical measurements systematically.

Electrical Characterization of NOE

Figure 3A shows typical room temperature current-voltage (I-V) curve of the junction, measured at ambient conditions over a ±1 V bias range with an NOE-electrode contact edge length of ∼2.1 mm (cf. Figure S4 allows contact to ∼1.4 × 105 Au@SiO2 NPs at the electrode edge). Remarkably, ∼10 pA current flows through the suspended NOE at 1.0 V, which is more than 2 orders of magnitude above our noise level (cf. Figure 3C). If we assume that current flows through all the NPs (i.e., 100% real contact), the junction current at 1 V is calculated to be ∼0.1 fA/(NP@contact). Even if this is what remains after passing through a long chain of NPs (∼6,500 NPs in series that span the 100-μm gap), this value is still amazing, because the AuNPs in the membrane are well isolated from each other by ∼ 3.6-nm insulating silica shells. Electron tunneling through such thick an insulator (with a total cumulative silica shell thickness of ∼15 μm) should not be measurable because the probability of electron tunneling (T) is an exponential function of the barrier length, L ( for a single 3.6-nm silica gap, where β is the tunneling decay parameter in units of (length)−1 and we used value of 13.3/nm for SiO2) (Müller et al., 2002, Salomon et al., 2003).

Figure 3

Room Temperature I-V Characteristics of the Junction

(A) Typical room temperature current-voltage (I-V) curve of the junction measured at ambient conditions in a range of ±1 V bias. Inset: microscope-based dark-field scattering image of the NOE nanomembrane.

(B) The corresponding differential conductance curve of the junction obtained numerically from the measured I-V characteristics.

(C) Photoconductance response curves when the light is switched on and off (532 nm, 35 mW).

(D) Typical I-V curve of a control junction fabricated from a plasmonically inactive 12 nm Pd@ 1.6 nm SiO2 nanomembrane. Inset: microscope-based dark-field scattering image of the NOE nanomembrane.

See also Figures S4–S7, Notes S1 and S2, Scheme S1.

Interestingly, the I-V curve (Figure 3A) shows nearly linear Ohmic responses, but with pronounced steps both in the dark and under illumination. Such features are unobservable at room temperature for (silica shell-free) AuNP array-based nanosheets or nanomembranes. Such steps may be caused by both low packing density of the networked nanomembrane (Wu et al., 2016) and the transport barrier, created by the thin silica shells between individual AuNPs. Such curves require kBT < charging energy; without the silica shell, kBT > charging energy of the NPs should hold (Liao et al., 2011). Although it is quite difficult to observe the current steps on an irregular thin nanofilm at room temperature, the Coulomb staircase can still be observed in some specific cases, for instance, a granular nanofilm system consists of several thousands of tunnel junctions (Imamura et al., 2000). In our case, the observed Coulomb staircases may be a reflection of (very) limited current transport paths (due to the low packing density) of the NOE NPs and a sign of unusual quantum transport of electrons (P-transport in this study) from grain (AuNP) to grain (AuNPs) through the intervening amorphous silica shells that act as potential barriers or weak links. The very presence of such current steps in the Au@SiO2 NPs system suggest that somewhere there is a weak link in what is otherwise a relatively well-conducting system. Also, the observed (overall) Ohmic-like response is quite different from the nonlinear response that typifies the quantum tunneling regime, implying a different electron transport regime. However, the origin for the current steps is still not clear and needs to be further studied.

Discussion

The underlying mechanism was primarily probed by measuring the temperature dependence of its electronic conductivity. From the results presented in Figures S5 and S6A–S6C, the conductance shows weak temperature dependence from room temperature till 200–170 K, below which it is roughly constant, which is consistent with electron tunneling below 170 K. The weak temperature dependence till 170 K, though, is much larger than what is expected from broadening of the Fermi-Dirac distribution of carriers in the electrodes with increasing temperature (Hrach, 1968, Vilan et al., 2017), implying that another mechanism dominates at these higher temperatures. Interestingly, as shown in Figure S7, the step-like current feature remained but became weaker if the I-V measurement was taken at higher temperature (∼333 K).

According to a rough estimate based on electron tunneling, using a simplified Wentzel–Kramers–Brillouin (WKB)-based model (Tomfohr and Sankey, 2002), the calculated current across a junction of even one row of Au@SiO2 NP-bridged NOE (the NOE-electrode contact edge length used in the calculation is the same as that used in the experimental junction, corresponding to Figure S4, i.e., ∼2.1 mm) is about 3×10−17 A at 1 V (see Note S1), which is far smaller than the detected junction current (∼10−11 A) for the whole NOE (for an electron to cross the gap between the electrodes ∼6,500 sequential tunneling events are needed!).

Considering the tunneling charge transfer plasmon mode reported previously (Tan et al., 2014), we propose that this huge enhancement of electron transport probability and ultra-long-range electron transport ability may be attributed to the plasmonic nature of the Au@SiO2 nanomembrane and the collective, wave-like charge density fluctuation (localized surface plasmon resonance [LSPR] coupling between adjacent AuNPs and its facilitation for ultra-long-range electron transport) of the 2D plasmonic nanomembranes (Warren et al., 2012).

The plasmonic nature of the NOE was manifested by in situ dark-field scattering imaging. As seen from Figure S8, only the NOE bridging over the trench is plasmonically active (survives plasmon damping caused by close electrode contact), and it persists after the successive I-V measurements. As current cannot flow through the blank Au-Au microtrench electrodes (Figure S9), and control devices using NOE made of bare AuNPs (without silica shell) displayed only Ohmic I-V response with current range in microamperes (Figure S10A) due to metallic Ohmic contact between adjacent AuNPs and hence partial plasmon damping of the NOE as revealed by TEM (Figure S10B) and dark-field image (Figure S10C), the observed I-V characteristics are therefore attributed to the physical properties of the NPs and NOE.

Figure 3B shows the differential conductance curve of the junction obtained by numerical differentiation of the measured I-V characteristics. The peaks are clearly seen with an average step width of about 120 mV, but they are not very periodic, possibly due to the complex 2D nanostructure and uncontrollable multiple electron tunneling paths of the NOE junction system. The value (120 mV) was then used to estimate the capacitance of the Au@SiO2 NPs (C ≈ e/ΔV = 1.3×10−18 F), which matches the calculated value (∼4×10−18 F) (see Note S2). Because in the present junction system

(1) the charging energy of the NPs (∼1.4×10−19 J) is much larger than the thermal energy kBT (4×10−21 J) and

(2) the resistance of the junction (∼1011Ω for NOE) is larger than the quantum resistance (ħ/e2 = 2.6 × 104 Ω) (Note S2),

we suggest that the ultra-long-distance electron transport through the NOE is due to the unique capacitive and plasmonic properties of the NPs. Then, the observed Coulomb staircases might be a reflection of Coulomb charging at room temperature. Measurements on the same devices through successive potential cycling showed very reproducible I-V characteristics and indicated reliable stable contact formation. Thanks to the solid contact of NPs (via assembly) and robustness of the NOE junctions (cf. Figure 2J), the I-V responses were quite reproducible from device to device with the same preparation (in the same order of magnitude, ∼ dozens of picoamperes, see Figure S11), further ruling out contact artifacts. Control experiments of the NOE junction made of big AuNPs (∼32 nm) with similar 1.6-nm silica shells showed similar I-V response but strongly weakened Coulomb staircases due to the size increase and the loss of NP homogeneity (Figure S12). We also examined the effect of the SiO2 shell thickness on the electron transport behavior of the assemblies. With an ultrathin silica shell coating of ∼1.1 ± 0.6 nm (Figure S13A), the I-V curves changed from linear, i.e., Ohmic (for the non-coated AuNPs, cf. Figure S10A) to non-linear, accompanied by a pronounced current drop from micro- to nanoamperes (at −1 V applied bias) and some step-like features (Figure S13B). With even a slight increase of silica shell thickness to the typical 1.8 ± 0.5 nm (Figure S1A, mean ± TEM), pronounced steps appear in the current curves (cf. Figure 3A). However, increasing the silica shell thickness further to ∼3.0 nm, no current was measured through the NOE junctions (Figure S14), as the ∼6-nm SiOx barrier likely decreases tunneling probability below our measurement limits (with nearly the same NOE-electrode contact edge length of about 2.5 mm).

Although efficient (measurable) electron tunneling through ∼3.6-nm insulating silica gap seems not possible, clearly electron transport occurs, not across one, but across >6,000 such junctions; this may be enabled by the strong “built-in” long-range plasmon field and mediated by “dark/light” plasmon modes of the NP system (cf. Figures 1A–1C). As shown in Figure 3A, in the dark the junction carries a current of approximately −12.3 pA at −1 V applied bias, uncovering for the first time (to the best of our knowledge) the “dark” plasmon mode and its contribution to current transport, which is hard to be revealed in a current-carrying device if using bare AuNPs since electronically excited plasmons (or hot electrons) will overlay onto it. Upon irradiation with light (35 mW), the conductance increases slightly to −13.5 pS (the relatively low laser intensity efficiently limits any possible direct photothermal heating effect on the electron transport), due to the excitation of additional plasmons optically. Upon turning off the light, the conductance decreased immediately to the original “dark” value. This can be cycled between these two states by alternating turn on/off light (Figure 3C). As closely seen from Figure 3C, under successive ∼1 min illumination the photoconductance fluctuation was maintained within the noise level (∼0.1 pS, cf. ∼ 1.2 pS for photoresponse in Figure 3A), i.e., no detectable drift in photoconductance occurred, which in fact can rule out the photothermal effect. More importantly, when we used both non-plasmonic metallic Pd NPs (∼12 nm) with similar 1.6-nm silica shells and insulating SiO2 NPs (with even larger NOE-electrode contact edge length of ∼2.9 mm, Figures S15B and S16D) to prepare NOE junctions, no electron transport was detected for these junctions as shown in Figures 3D and S15 (see Figure S16 for detailed characterizations). This is consistent with the plasmonic origin of the observed effects.

The plasmonic nature and “light” mode effect of the electron transport phenomenon was further confirmed by illumination-wavelength- and intensity-dependent I-V measurements of the NOE junctions. As shown in Figure 4A, illuminating the NOE with green (λ = 532 nm), blue (λ = 450 nm), and red (λ = 650 nm) lights with comparable light intensity (35 mW) causes an obvious increase in the junction conductance over that of the dark conductance, and the conductance with green light (more spectral overlap with monolayered nanomembrane's plasmon resonance spectrum) is obviously larger than that with other lights, which fits quite well with the LSPR band profile of the NPs (Figure 4B). The junction conductance also increases with the increase of illumination intensity (Figure 4C). This is consistent with the contribution of the photo-induced “hot electron,” whose effect on photocurrent is dependent on the irradiation intensity (Conklin et al., 2013). Since the temperature increase on NP surfaces is roughly estimated to be well below 1°C in our system (Qin and Bischof, 2012, Note S3), it cannot be the main driving force for the observed electron transport enhancement. In fact, the locally generated small amount of heat within the suspended ultrathin NOE nanomembrane can be further dissipated either laterally (via electron flow) or vertically (via ambient air). This has been further proved by photothermal imaging of the real temperature variation of the NOE junction during the light on/off I-V measurements (Figure S17). The observed electron transport (ET) enhancement is, therefore, primarily a (non-photothermal) plasmon effect.

Figure 4

Illumination Wavelength- and Intensity-Dependent Junction Conductance of the NOE Junctions

(A and B) (A) Illumination wavelength-dependent photoconductance of the NOE junctions, which fits well with (B) the extinction profile of the AuNPs.

(C) Illumination intensity-dependent photoconductance of the NOE junctions. Data are represented as mean ± photoconductance.

(D) FDTD simulation of the electric field distribution of three closely impacted 12 nm Au@1.8 nm SiO2 NPs.

See also Figures S8–S18, Note S3.

Figure 4D shows the result of finite difference time domain simulation (FDTD), displaying a remarkably enhanced electromagnetic field between adjacent NPs. Although the frequencies associated with the electron flow are much lower than the frequency variation of the enhanced electric field due to irradiation (∼1014 Hz), they merge at the nanoscale with “plasmonics,” and the plasmon-induced strong localized “built-in” electromagnetic near fields, along with collective wavelike charge density fluctuation at the nanoscale (Ambrosetti et al., 2016), are possibly supposed to facilitate electron tunneling between proximal AuNPs and current flow across the NOE. However, further investigations are needed to finally reveal the whole story of P-transport, including how it scales.

Conclusion

By constructing a robust suspended-film-type plasmoelectronic circuit, using shell-isolated Au@SiO2 NPs as building blocks, we have revealed a new electron transport regime—plasmon-enabled/plasmon-coupled long-range electron transport (P-transport) on the 2D metasurface. Unprecedented long-range electron transport across the monolayered jammed AuNPs with insulating tunneling gap up to 3.6 nm, which is impossible for the currently accepted electron tunneling theory, are observed. The crucial role of the LSPR coupling (collective wave-like charge density fluctuation and surfing of electrons) in P-transport was evidenced by photocurrent enhancement, wavelength-dependent current intensity, and FDTD simulation of the electric field distribution. The finding reshapes our thinking on fundamental electron transport regimes on the nanoscale, will open an interdisciplinary field of exploration, and will promote future development of plasmoelectronics, integrating plasmonics with nanoelectronics.

Limitations of Study

The origin for the current steps is not clear and needs to be further studied; issues with reproducibility and control of the structures also warrant further studies.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.