Efficient Surface Plasmon Polariton Excitation and Control over Outcoupling Mechanisms in Metal-Insulator-Metal Tunneling Junctions.

Abstract Surface plasmon polaritons (SPPs) are viable candidates for integration into on‐chip nano‐circuitry that allow access to high data bandwidths and low energy consumption. Metal–insulator–metal tunneling junctions (MIM‐TJs) have recently been shown to excite and detect SPPs electrically; however, experimentally measured efficiencies and outcoupling mechanisms are not fully understood. It is shown that the MIM‐TJ cavity SPP mode (MIM‐SPP) can outcouple via three pathways to i) photons via scattering of MIM‐SPP at the MIM–TJ interfaces, ii) SPPs at the metal–dielectric interfaces (bound‐SPPs) by mode coupling through the electrodes, and iii) photons and bound‐SPP modes by mode coupling at the MIM‐TJ edges. It is also shown that, for Al‐AlOx‐Cr‐Au MIM‐TJs on glass, the MIM‐SPP mode outcouples efficiently to bound‐SPPs through either electrode (pathway 2); this outcoupling pathway can be selectively turned on and off by changing the respective electrode thickness. Outcoupling at the MIM‐TJ edges (pathway 3) is efficient and sensitive to the edge topography, whereas most light emission originates from roughness‐induced scattering of the MIM‐SPP mode (pathway 1). Using an arbitrary roughness profile, it is demonstrated that various roughness facets can raise MIM‐SPP outcoupling efficiencies to 0.62%. These results pave the way for understanding the topographical parameters needed to develop CMOS‐compatible plasmonic circuitry elements.


Device Fabrication.
The Al-AlO x -Cr-Au MIM-TJs were designed and fabricated in a similar method to our previous work. 1 Al was used because it natively oxidises forming a tunnelling barrier and Au was chosen due to wide availability, low plasmonic losses and it is well-established in laboratory fabrication techniques. All devices were fabricated on borosilicate coverslips (Paul Marienfeld GmbH, 22×22 mm, 0.16-0.19 mm thick) with a roughness σ of 0.17  0.01 nm measured over an area of 1 × 1 µm 2see Fig. S1 which is two orders of magnitude smaller than the roughness introduced by thermal metal deposition. A three-step fabrication process was performed. First, contact pads (100×80 μm) were patterned using a laserwriter (Microtech, LW405B), with bi-layer resist with a total resist thickness of ~1.3 μm: LOR 3A, (prebake at 170˚C for 5 min) and S1805 (prebake at 115˚C for 1 min) exposed with a 405nm beam at 320 mJ/cm 2 . This was followed by the thermal deposition (Kurt J. Lesker, NANO 36) of the bi-metal layer of Cr/Au (3/25 nm) where Cr serves as an adhesion layer. In the second step after lift-off, the electrodes were patterned again by laserwriter and the Al layer of different thicknesses (t Al =40, 80, 100 and 60 nm) were thermally deposited. Finally, the top electrodes were exposed using electron beam lithography (JEOL, JBX-6300FS), using PMMA 950 A4 resist (MicroChem, prebake at 180˚C for 2 min) exposed with 5 nA current. After resist development, samples were immersed into MF-319 optical developer, with active ingredient TMAH which is an Al etchant, to remove AlOx for 20s and rinsed with water for 1 min. Etching thins the Al electrodes in the region under the top electrode, so that the final thickness in the MIM-TJ area (5×5 µm 2 ) is t Al =20, 65, 90 and 40 nm. Fresh native AlO x was then grown in ambient conditions for 1.5 h, before the Au layer with 1 nm of Cr adhesion layer was deposited using thermal evaporator (t Au =155, 155, 155 and 40 nm). The first three samples all had top electrode fabricated at once to minimise substrate-to-substrate variation.

Electrical Measurements.
Prior to the light emission measurements, IV-characteristics were taken in order to confirm tunnelling behaviour of the MIMTJs. MIM-TJ voltage measurements were conducted using micromanipulators (Signatone) with Tungsten probes (ZN50R DC/RF) were performed using a source meter (Keithley 6430, Keithley Instruments) and controlled by homemade LabView programme. During all the experiments, the Au electrodes were grounded and the Al electrodes were biased. We have already shown previously 1 that Al-AlO x -Cr-Au junction have lower break down at positive bias than at negative bias. Therefore, we applied voltage between +1 and -1.5 V with a step of 50 mV. Obtained IV-curves do not depend on the thickness of the bottom electrode (Fig. 2c). The differential conductance (dI/dV) plot in Fig. 2d demonstrates the parabolic bias dependence of tunnelling, one of the Rowell criteria to identify tunnelling. 2 Our electrical measurements are representative of those that we have previously reported. 1 3. Optical Measurements.

Optical setup. An inverted optical microscope (Nikon Eclipse Ti-E) equipped with an
Andor spectrometer (Shamrock 122 303i) and an electron multiplying CCD (EMCCD, iXon Ultra 897) were used for the optical characterisation of the emitted light from the tunnelling junctions. Optical measurements were measured from the glass (back) side using a 100× oil objective (Numerical Aperture, NA=1.49). EMCCD images (both real and back focal planes) were taken with a 2 min integration time and 300 EM gain.

Al transmission measurements.
The optical contribution to the EMCCD images (both real and BFP) from the SPP Al-air mode decrease with thickness of the Al. We performed transmission measurements on the metal to quantify this reduction and found that the Al has a experimentally derived skin depth δ m = 23.2 nm at λ = 900 nm, and an optical penetration length of δ γ = 11.6 nm at λ = 900 nm (   (S1) where a m represents the total MIM-TJ area. All simulations were done over a 3D domain with volume of 16 × 16 × 2 μm 3 . These dimensions incorporate the entire 5 × 5 μm 2 MIM-TJ with 3 μm truncated waveguides extending from each side.

Angular light emission.
To understand the noise background in the BFP images ( Fig. 3ij) we simulated the angular distribution of light emission from the MIM-TJ (Fig. S3). Fig. S3 indicates that, for an MIM-TJ with a smooth interface (σ pv = 0 nm), no signal in the BFP would be expected for the partial solid angle corresponding to 240 o < ϕ < 300˚ and 0 o < θ < 30˚ for an MIM-TJ of any thickness combination. Therefore, the signal in Fig. 3e within   7 k xy /k 0 < 1 is direct evidence of photon scattering in the MIM-TJ from the surface roughness.
The light emission in Fig. S3 corresponds to a dipole at z5 from Fig. S2.

Mode profiles.
The mode profiles shown in Fig. 4 demonstrate the effects on outcoupling by σ pv (Fig. 4a,b) and t eff (Fig. 4c,d).  out of this transmission box P γ-spp = P γ + P spp is therefore the total power radiated into the forms of the photons and bound-SPPs. We can then estimate the power radiated into the form of the bound SPPs as P spp = P γ-spp -P γ . The coupling efficiency is subsequently calculated by integrating all contributions over MIM-TJ, and then normalising by MIM-TJ area, and are presented in Fig. 5 in the main text. The roughness profile is assumed to comprise Al and Au grains. These grains are modelled as half ellipsoids randomly generated around the dipole with the radii in the range of 25 to 50 nm and the heights in the range of 7.5 to 12.5 nm.