Future Technology Trends

Computers of the future may route data at an ultra-fast rate. Plasmonics is a way to create optical interconnects smaller than 100 nanometers wide. There are drawbacks to other methods. An electrical signal begins to degrade as the frequency goes above a certain amount of gigahertz. This limits the number of bytes that transfer across a wire at any given time. Conventional photonics, on the other hand, cannot shrink down past a specific size (hundreds of nanometers). A new European project called NAVOLCHI is seeking to develop nanometer scale chip-to-chip connections that utilize plasmon polaritons as a carrier of information. There is hope that this channel will overcome a few of the restrictions of other types of technology. Electronic, photonic and plasmonic combinations may become easier to carry out. 

The researchers are building small plasmonic transceivers. A chip will have to accommodate integrated lasers, amplifiers, detectors and modulators. The components can allow for a massively parallel inter-chip communication. This has to function with existing complementary metal-oxide semiconductor fabrication. Being able to build it with these sorts of techniques will ensure that the scientists can produce the processor cheaply. A novel CPU needs many of these links to receive bits from another chip or random access memory. Upgraded plasmonics could require less juice to transmit data over short ranges than alternate options. The propagation distance for the quasiparticles is rather short and they decay rapidly. There has been an intense focus on improving that situation. In the past, workers have said that pure plasmonic links do not provide an energy efficient solution over small lengths. It seems, though, that a lot of progress has been made in reducing the power requirements. Recently, academics at Stanford University have created a tiny LED that uses 0.25 femto-joules per bit, as an example.

The NAVOLCHI academics have already developed a nano-scale plasmonic pillar laser. It is difficult to say how far they can push the tech. There will likely be frequency or power limitations as to what is practical. This science may eventually find its way into high performance computing applications. Consumer devices could one day contain the parts as well. That outcome lies further down the road. This circuitry might be necessary to enable exaflop supercomputers or beyond. Obtaining zettaflop mainframes is only possible with some kind of breakthrough in logic gates. The concept of that much calculating muscle is speculative. It might not actually come to fruition.