Report

Quantum Plasmon Resonances Controlled by Molecular Tunnel Junctions

See allHide authors and affiliations

Science  28 Mar 2014:
Vol. 343, Issue 6178, pp. 1496-1499
DOI: 10.1126/science.1248797

Controlling Quantum Plasmonics

Electron tunneling across cavities could potentially induce a quantum mechanical plasmon mode that would be important in nano-electronics, catalysis, nonlinear optics, or single-molecule sensing, but has been expected to occur only at length scales beyond the reach of current state-of-the-art technology. Using a system of plasmonic dimers comprising silver nanocubes bridged by a molecular self-assembled monolayer, Tan et al. (p. 1496; see the Perspective by Nordlander) observed quantum plasmonic tunneling between the resonators and were able to tune the frequency of this tunneling plasmon resonance via selection of the molecular tunnel junctions. Moreover, the effects were observed at length scales that are technologically accessible.

Abstract

Quantum tunneling between two plasmonic resonators links nonlinear quantum optics with terahertz nanoelectronics. We describe the direct observation of and control over quantum plasmon resonances at length scales in the range 0.4 to 1.3 nanometers across molecular tunnel junctions made of two plasmonic resonators bridged by self-assembled monolayers (SAMs). The tunnel barrier width and height are controlled by the properties of the molecules. Using electron energy-loss spectroscopy, we directly observe a plasmon mode, the tunneling charge transfer plasmon, whose frequency (ranging from 140 to 245 terahertz) is dependent on the molecules bridging the gaps.

Quantum mechanical effects in plasmonic structures are believed to become important when two plasmonic resonators are placed so closely that electrons can tunnel across the gap (111). Direct experimental access to the resulting tunneling charge transfer plasmon (tCTP) mode is expected to open up new opportunities in, for instance, nanoscale opto-electronics, single-molecule sensing, and nonlinear optics (1). Experimental and theoretical studies so far have concluded that quantum mechanical effects are important only at length scales below 0.3 to 0.5 nm, close to the bond length of gold and silver (811). Such structures are technologically inaccessible; therefore, it is important to demonstrate the tCTP mode across gaps larger than a nanometer that can be fabricated by state-of-the-art fabrication techniques (10). Unlike past works that investigated tunneling through a vacuum (12), we placed molecules in the gap because tunneling rates across molecules depend on the molecular structure and are much higher than across a vacuum. This approach made it possible to directly observe and control tCTPs experimentally in tunneling gaps up to at least 1.3 nm, depending on the type of molecules bridging the gap, and to move quantum plasmonics into the size domain that is accessible via bottom-up or top-down fabrication methods (10).

Quantum effects have been observed only indirectly as shifts in the bonding dipolar resonance plasmon mode (1, 9, 11). Our aim was to perform an experiment in which the presence of a tunneling barrier can be directly imaged while the tCTP mode is simultaneously measured spectroscopically by introducing two experimental innovations: The cross-sectional area of the tunnel junction was increased from a few nm2 to roughly 103 nm2, and the tunneling rate across the nanogaps was increased by tunneling through molecules rather than vacuum.

Cuboidal silver nanoparticles were used (13), separated by SAMs with thicknesses of 0.5 to 0.6 nm forming metal-SAM-metal junctions through self-assembly (Fig. 1). The facets of the nanoparticles are atomically flat, which results in a very large cross-sectional area of around 103 nm2, maximizing the number of tunneling events across the junctions. The silver nanoparticles were functionalized with either saturated, aliphatic 1,2-ethanedithiolates (EDT) or aromatic 1,4-benzenedithiolates (BDT) (14). The lengths of EDT and BDT are similar, but they have very different HOMO (highest occupied molecular orbital)–LUMO (lowest unoccupied molecular orbital) gaps of 8 and 5 eV, respectively (1517). Therefore, the tunneling rates across junctions with BDT molecules are higher than those junctions with EDT. The interaction betwfseen the two nanoparticles was optimized to avoid aggregation or misalignment by diluting the dithiols with 1-propanethiol (PT) (14). After self-assembly of the dimeric structures, they were deposited on a 30-nm-thick, electron-transparent silicon nitride membrane.

Fig. 1 Quantum plasmonic tunnel junctions.

(A) Schematic illustration of the molecular tunnel junctions made of two silver nanoparticles bridged by a SAM on an electron-transparent silicon nitride membrane. The contactless electron nanoprobe was placed near the functionalized silver nanoparticles to excite and measure the surface plasmons of individual dimers. (B) The distance between two adjacent nanoparticles is determined by the thickness of the SAMs of EDT or BDT. (C) A schematic energy-level diagram of the junctions.

A simplified form of the Simmons equation (Eq. 1) is commonly used to approximate molecular tunnel junctions (18, 19)

Embedded ImageEmbedded Imagewith(1)

where β (Å−1) is the tunneling decay coefficient, d (nm) is the width of the tunneling barrier, and the pre-exponential factor J0 (A/cm2) is the hypothetical current when d = 0; m is the mass of the charge carrier (kg), and ħ is the reduced Plank’s constant. The value of β depends on the barrier height ϕ (eV). Tunneling rates through molecular bonds, so-called through-bond tunneling, are much higher (β ≈ 0.8 to 0.9 Å−1 for saturated molecules and 0.1 to 0.3 Å−1 for unsaturated molecules) than through vacuum (β = 2.9 Å−1) (20). Figure 1C shows the energy level diagram of the metal-SAM-metal junctions schematically. As indicated in Fig. 1C, in molecular electronics d is defined by the length of the molecule dl (nm), and ϕ by the offset between the Fermi levels of the metal and the energy level of the molecular frontier orbitals. In contrast, when tunneling through a vacuum is the dominant mechanism of charge transport, the barrier height equals the work function of the electrode materials and d equals the gap, dg (nm), between the two electrodes— that is, the distance between the nanoparticles (21).

Instead of bringing the dimer particles in contact with electrical probes that will perturb the plasmon resonances, we used in our experiment a focused beam of energetic electrons in a scanning transmission electron microscope (STEM) as a contactless nanoprobe to excite and analyze the surface plasmon resonances in individual dimers. We positioned the electron probe opposite the gap at the long axis of a silver particle dimer, as illustrated in Fig. 1A, to excite plasmon modes—similar to lateral plane wave illumination of the dimer (14). During the plasmon excitation, the tunnel junction was therefore not exposed to the electron beam to minimize irradiation damage (14). The excitation of plasmon resonances results in an energy transfer from the electron beam to the particle system, which we analyzed with monochromated electron energy-loss spectroscopy (EELS) (2224). The junctions were imaged before and after acquiring the EELS spectra to ensure that none of these junctions formed conductive metal filaments during the experiment (14). Thus, in all of our experiments, we could discriminate between the tunneling and conduction through metal filaments—CTP modes conclusively.

Figure 2A shows atomic-resolution TEM images of a silver nanoparticle dimer. The first peak in the histograms of values of dg estimated from TEM images on a series of dimers is centered at 0.55 ± 0.08 nm for EDT and 0.67 ± 0.12 nm for BDT, which are close to dl as expected for Ag-SAM-Ag structures. The second peak in these histograms is attributed to dimers with SAMs on both nanoparticles (Ag-SAM//SAM-Ag structures, where “//” indicates a noncovalent contact). Intercalating SAMs or incompletely removed polymer that was used in the nanoparticle synthesis may result in smaller and larger gap sizes than expected from the molecular lengths, as indicated schematically in Fig. 3.

Fig. 2 Direct observation of quantum tunneling between plasmon resonators.

(A) A high-resolution TEM image of the junctions and histograms of the gap-sizes (14). (B) Two examples of measured EELS spectra with the occurrence of quantum tunneling directly observed by the tCTP peak and quantum-corrected simulations of the extinction spectra, confirming the identification of the peaks. (C) Experimentally measured plasmon energy as a function of gap size for dimers functionalized with monolayers of BDT (blue circles) and EDT (red triangles). Theoretical calculations for through-space and through-bond tunneling are shown as dotted lines and solid lines, respectively, for the two SAMs. (D) Simulated maps of the electrical-field distributions for the plasmon modes I to IV, corresponding with the spectral peaks.

Fig. 3 Quantum plasmon resonances as a function of tunneling distance.

(A) Experimentally measured plasmon energy as a function of gap size for BDT- (blue circles) functionalized dimers. The gap size varies between individual dimers because of structural disorder in the SAMs. (B) Measured EELS spectra for double SAMs of EDT (red) and BDT (blue). Tunneling was observed for the double-layer BDT but not in the double-layer EDT junctions.

Figure 2B shows EELS spectra recorded from junctions with SAMs of EDT and BDT. Three main plasmon peaks were observed around 2.2 eV, 3.2 eV, and 3.6 eV, which we assigned to the bonding dipolar plasmon mode (II), the transverse corner mode (III), and the transverse edge mode (IV), respectively, in agreement with the finite-element-model (FEM) simulations (Fig. 2D) (14, 21). A new low-energy plasmon mode is observed at 0.60 ± 0.04 eV for EDT and 1.01 ± 0.01 eV for BDT.

We assigned this plasmon mode to the tCTP based on our calculations that show the transfer of net charge between the cuboids (Fig. 2D, mode I). The plasmon resonances of the Ag-SAM-Ag system was simulated using a quantum-corrected FEM optical simulation model (14, 21). Briefly, the optical properties of the junctions are calculated through a quantum mechanical approach and then used to simulate the plasmon resonances of the Ag-SAM-Ag system. The model predicts that the tCTP mode strongly depends on ϕ, d, and gap field, Egap (V/m). The value of ϕ is modeled analytically as ϕ = α EHL, where 0 < α < 1 relates to the energy-level alignment, and EHL is the HOMO-LUMO gap, which was obtained from single-molecule experiments (1517). We assume Egap = 7 ×108 V/m throughout the calculations based on previously reported work (14, 21).

The nature of the charge transport—through-space or through-bond tunneling—was determined from EELS measurements on 32 junctions, for which the energy of the tCTP mode was plotted as a function of dg (Fig. 2C). This graph shows that the tCTP depends only weakly on the value of dg. The dotted lines in Fig. 2C show simulations of the tCTP energy shifts if through-space tunneling dominated, with d = dg (14). The solid lines are simulations for through-bond tunneling with d = dl, where the through-bond tunneling distance depends on the length of the molecule and can be different from dg when the molecules are not perfectly aligned in the gap. Figure 2C shows that through-bond tunneling has a much weaker dependence on gap size than through-space tunneling. The good agreement with the experimental results indicates that coherent through-bond tunneling is the dominant mechanism of charge transport (14).

Through-bond tunneling allows us to explore the tCTP mode across a large gap because the tunneling is less dependent on the gap size. EELS spectra were recorded on Ag-SAM//SAM-Ag junctions to study whether the tCTP mode could be observed over larger length scales up to 1.3 nm. Figure 3 shows a tCTP mode for the structures with BDT at 0.975 to 1.015 eV. The tCTP peak energy only weakly depends on dg, which confirms that though-bond tunneling is the dominant mechanism of charge transport. The marginal difference in energy of the tCTP mode for single and double SAMs is likely due to strong π–π coupling between the BDT SAMs (25). For Ag-SAM//SAM-Ag structures with EDT, the tCTP mode was not observed because of the low β value, and because no π–π coupling occurs between aliphatic molecules, we expected its peak—if any—to be at very low energies below the detection limit of our instrument.

By combining atomic-resolution imaging, single-particle spectroscopy, and monolayer molecular control, we have demonstrated quantum-mechanical electron tunneling at optical frequencies between plasmon resonators. By varying the self-assembled molecular monolayers in the junctions, we found that the plasmon-induced tunneling frequencies could be controlled from 1.01 ± 0.01 eV, or 244 ± 3 THz, for a monolayer of BDT molecules, to 0.60 ± 0.04 eV, or 145 ± 10 THz, for a monolayer of EDT molecules. The mechanism of charge transport was coherent through-bond tunneling, which is only weakly dependent on the gap size. The relatively large distance of up to 1.3 nm over which the tunneling takes place in Ag-BDT//BDT-Ag junctions may provide potential for molecular control over quantum plasmonic systems through longer molecules to perhaps 4 to 5 nm (26)—that is, gap sizes that are currently accessible by top-down fabrication techniques. Our results show that tunneling can reconcile molecular electronics with plasmonics, opening up a whole new interdisciplinary field of exploration.

Supplementary Materials

www.sciencemag.org/content/343/6178/1496/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S11

References (2740)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: We acknowledge the National Research Foundation (NRF) for supporting this research under the Competitive Research Programme (CRP) program (award NRF-CRP 8-2011-07). J.K.W.Y, P.B., and L.W. acknowledge the Agency for Science, Technology and Research (A*STAR) for the A*STAR Investigatorship Grant, and TSRP grant 1021520014.
View Abstract

Navigate This Article