Memristor Technology Fabrication
In the journal Nano Letters, researchers have published a paper that details a novel way to manufacture memristor integrated circuits. Memristors are a recently developed technology that has the potential to power future electronic devices and create neuromorphic hardware. Last year HP scientists demonstrated another technique that had a 20% yield for functional memristors on a circuit. The newer method reduces the complexity of the manufacturing process and facilitates a higher throughput. These types of innovations are critical in order to enable the production of microchips that can be competitive with currently existing technology.
We demonstrate a technique to fabricate memristor cross-point arrays using a self-aligned, one step nanoimprint lithography process that simultaneously patterns the bottom electrode, switching material film and the top electrode.
The critical interfaces are exposed to much less contamination and thus under better chemical control. With this technique, we fabricated arrays of TiO2-based memristive devices (junction area 100 nm by 100 nm) that did not require electrical forming and were operated with nanoampere currents.
Neuromorphic tech is one of the more intriguing possibilities that may eventually come to fruition based on recent work in this field. Simulating a brain on a supercomputer via software is currently not a cost effective way of creating an artificial intelligence. A large amount of computational strength may be needed to perform such a task (perhaps 100′s of petaflops). Currently the top supercomputer in the world is only in the 8 petaflop range, so there may be a long way to go before it would even become possible. Even when that needed number crunching milestone is reached, the machine will require an enormous amount of electricity to maintain the model. It is also not clear that von Neumann-based computers could sustain consciousness in the way biological tissue does. The manner in which a CPU processes information from random access memory is very different than the electrochemical signaling of the mind. This kind of model may merely interpret what a brain does instead of mimicking it exactly. Memristive nanodevices on the other hand could potentially overcome the obstacles to synthesizing an artificial mind.
In a biological brain, neurotransmitters allow neurons to communicate with one another via synaptic connections. Synaptic inputs can increase or decrease the likelihood that a neuron will fire an action potential. The signaling strength of a synapse can change over time by method of LTP or LTD. The properties of memory resistance are suited to copying the functioning of these connections. They have the capacity to alter and “remember” their resistance as a result of electrical inputs. Thus they can mimic to a certain extent neuronal processes. Previous neuromorphic chips have been been shown to decrease the necessary power consumption considerably, enabling them to overcome CPU or GPU type bottlenecks to neural simulations. The analogue nature of memristors may support qualia on an alternate substrate.
Neuromorphic circuits might need to have fundamental features around one tenth the size of today’s average transistor in order to compete with organic neural configurations. So there’s still a long way to go before we see their true potential. There is certainly a lot of hype that may never live up to sky high expectations. However many researchers are focusing intensely on advancing this radical science, so there may be many more breakthroughs yet to come.
Nanoscale resistive switches: devices, fabrication and integration
Self-Aligned Memristor Cross-Point Arrays Fabricated with One Nanoimprint Lithography Step
