Uvula Antennas and Plasmonic Slot Waveguides

Brian Edwards, Noah Rubin, Nader Engheta

In this largely unpublished work, we fabricated plasmonic "slot" (aka MIM) waveguides.  Our goal was to measure the complex transmission through these waveguides with an NSOM.  These results were to serve a greater purpose that was never realized, however, I believe the part of the road that was traveled was interesting enough to warrant discussion here.

In a plasmonic slot waveguide, the waveguide mode is confined between two plasmonic mediums (metals).  This is very different from a traditional optical waveguide which is made out of various glasses (dielectrics).  In this way, it has more in common with the power cord on your bedside lamp than it does an optical fiber.  Similar to a power cord, the plasmonic waveguide is more tolerant of bends than a traditional optical waveguide. 

Just as certain metals are better for conducting electricity (ie copper and aluminum), certain metals are better at conducting light.  Gold and silver are the most popular choices.  Light, apparently, has expensive taste.  However, the fundamental challenge of plasmonics is that the materials are "lossy," meaning the light is quickly turning into heat as it travels down the length of the waveguide.  It is not uncommon for the power to decrease by half for every several tens of microns traveled.  This creates very challenging experimental constraints.

Given that our goal was to probe these waveguides with an NSOM, the waveguides could not be encapsulated with a dielectric.  Modal simulations indicated that if they were fabricated on glass, the majority of the power would be contained within the glass and also be inaccessible.  Therefore, we built the waveguides such that they were free-standing.  This was achieved through the following fabrication steps:

This created free-standing gold slot waveguides with clear optical access from both above and below.  We patterned an input nano-antenna at one end of the waveguide and an output nano-antenna at the other.  Given the heavy losses incurred within the slot waveguide, the primary challenge of the experiment was to see the light coming from the exit antenna over the glare from the input antenna.  We employed several strategies toward to realize this goal.

1. The input antenna was illuminated from below and the output antenna was observed from above.  The majority of the uncoupled incident optical energy simply reflected off of the underside gold and didn't make it into the upper half-space.
2. By incorporating a 90 degree bend in the structure, we were able illuminate the input structure with one polarization (ie linear in the x-direction).  The output antenna was observed through a linear polarization filter oriented in the y-direction.  This filtered out the vast majority of the upwards scattered light from the input antenna.
3. The input antenna was not optimized for maximum coupling, but rather to maximize the ratio of coupled light to upward scattered light that would pollute the measurement.
4. The output antenna was optimized to radiate the most amount of light within the numerical aperture of the objective lens above the structure.

I'm a firm believer in simulating an experiment as close to reality as possible all the way to observables.  In order to understand exactly what we would see, the experiment needed to be modeled very carefully.  The images above show the steps involved.  The input antenna at (0,0,0) was illuminated with a focused Gaussian beam limited by the numerical aperture of the objective lens under the sample.  The fields evolved through the system and ultimately were radiated out of the output antenna at (4,4,0).  Within the simulation software, we captured the complete fields (complex E and H) just above the sample at the z=0.6um plane.  This is shown in the first panel above.  It is clear that the majority of the light was scattered by the input antenna.  We can calculate the intensity image that the microscope would see through the upper objective.  This is done by performing a plane wave expansion on the fields, filtering for the light which is within the NA of the upper objective lens, and then integrating the field again.  The image shown in the second panel is representative.  Finally, that image can be spatially windowed for some distance around the output antenna.  The transmission coefficient for the device can be calculated by comparing the total power emanating from the output antenna to the total power incident on the input antenna.

As a quick test, we built two waveguide systems which were identical except that one had a blockage midway along its length.  Both waveguides were illuminated.  The results were very promising.

The experiment was carried out for many different lengths.  In each image, the camera exposure time was adjusted so that the output intensity was approximately the same value, thus avoiding sensor saturation.  The normal microscopy image was captured also and superimposed.  The best of the samples performed with very near the expected performance calculated in simulation.  The typical exponential decay typical of waveguide transmission is apparent in this image.

Unfortunately, many of them fell short and the majority showed no measurable response.  This occurred across many trials over the coarse of many months.  We determined that this was not reliable enough to be a test-bed on which we could base other experiments.  We might have had better luck of we had moved to wavelengths greater than 1µm where plasmonic losses were lower, but this would have required us to retool the lab in expensive GaAs detectors.  I don't give up easily, but there was no way to trouble-shoot this besides throwing far too money at it with little guarantee that it would work.  Ultimately we cut our losses and abandoned this venture.