image sensors

Plasmonic filters — the deus ex machina of the century

Today’s semiconductor news on Spectrum present an article about plasmonic color filters “Flexible and Colorful Electronic Paper Promises a New Look for eBooks” by Dexter Johnson. Coincidentally, last week I had a hot discussion with a few physicists on the same topic, so I thought I’d introduce this animal here and also state my personal engineering view on the topic.

Act one: introduction

In a very broad sense, surface plasmons are free electron oscillations occurring between a metal and a dielectric, which are excited by a light-metal-dielectric interaction. The unbonded to any atom oscillating electrons can be created by incident photons or electrons falling on the junction between the two materials. The frequency of their oscillations depend on the junction thickness itself, as well as the distance to neighboring pairs of oscillating electrons. When light strikes a plasmonic material, apart from exciting free electrons, it also couples to these (which actually form a kind of surface electromagnetic field), and thus creates a self-sustaining interference phenomenon. The key feature of this concept is that only photons with specific energy can couple with the oscillating electrons, while the rest pass though, hence, this process can be naturally used as a color filter.

Such metamaterials have been defined theoretically in the mid-50s but they only become very popular during the past ten years due to the rapid improvement in lithographic techniques. They could be used in applications ranging from display technology and image sensors, which is why it has just recently been approached by the big players in the chip fabrication field. Both of these technologies need some kind of light filtering element, and both of them now are using organic color filters to achieve their goals. The problem with organic color filters is that they are complicated to produce and quickly degrade with time, especially when UV and high temperatures are involved, as is the case with the die of an image sensor.

Act two: the complication

Plasmonic color filters are physically formed by a sandwitch made of metallic bread (usually Tungsten or any of the noble metals) and a dielectric butter (SiO2 or eqv…). This sandwitch is also accurately bitten by rats creating a superfine nanohole array which looks like this:

Basic structure of a generic plasmonic color filter

Basic structure of a generic plasmonic color filter

At first glance such a technology looks very silicon friendly as all we need to create our filter structures is the addition of two extra metal layers to the CMOS process, sounds like a piece of cake. But why did the big semiconductor players decide to abandon this scheme as soon as they had a sniff at its surface? Plasmonic color filters are still very experimental and I think we can not identify them as, even an immature technology yet. Although there are quite a few academic groups working on the problem, the prospects for production on a mass scale currently seem miraculous. But hey, I am very happy when I see progress on the topic folks! We’ve seen it many times, many advancements in history have been the outcome of scientific mambo-jumbo once labeled as absurd or strange.

It has been identified that plasmonic filter structures have an excellent bandpass quality factor dwarfing out even the best organic compound color structures ever reported. The Q-factor, however, is not the only element in the picture. The transmission coefficient of the best reported plasmonics is in the order of 0.2-0.3 which is very disappointing. The filter’s response is also not very steady — towards UV and deep UV full transmission is usually observed. Nevertheless, perhaps this could be solved with an extra glass UV filter. But still, can we not use them for accurate light spectrometers, where light is of abundance and low transmission coefficients are affordable?

Act three: climax

Well, here is where engineering comes into play and destroys everything, for the time being… Plasmonic filters currently rely on an extremely accurate lithographic process called electron beam lithography which, combined with dry plasma etching has an accuracy in the order of few nm. Apparently, even that is not enough to create a filter with a good yield. All reported plasmonic filters, including this morning’s popular science article in Spectrum, are manufactured on a scale of a few micrometers. I.e. as the authors of the paper suggest, just a pixel of the retina display of an iphone. Using e-beam lithography in mass production is a fairytale. Gold is a forbidden word in semiconductor fabs, so this material as a filter falls out as well. Solving optical crosstalk problems and alignment in adjacent filters for RGB arrays seems like another wonder to me. How about the fact that the microlenses deposited on top of the filters are polymer which is another dielectric and as the theory suggests — will cause a modulation of the surface plasmon resonance. Or the non-uniformity of the dielectric’s thickness? All these issues create a highly non-linear outcome. Engineers don’t like non-linearities… neither does large scale production.

Act four: resolution

The outcome of this drama is still very unforseeable and here I let you cast the die. One thing is certain — no matter the aftermath — a generation of scientists is gaining momentum along with the new era of metamaterial sciences. However, for the time being, whenever somebody starts talking about how plasmonics will change the world in a year time, I just smile and listen carefully.