New Transistors Mimic Human Brain's Synapses
A new transistor designed to mimic structures in the human brain could pave the way for increasingly efficient computer systems that "think" like humans, scientists say.
The transistor is the first to mimic a crucial process used by brain cells, or neurons, when the cells signal one another.
The goal is to build nanometer-scale circuit components that can be used in neuron-inspired computers, said physicist and study author Dominique Vuillaume of the Institute of Electronics, Microelectronics and Nanotechnology in France.
Such computers would be useful for tasks that traditional computers aren't very good at, especially image processing and recognition, Vuillaume said.
Transistors are the building block of electronics . They allow control of the electrical current running through a circuit by amplifying or switching the current on and off.
Similarly, the synapse, a tiny gap between neighboring neurons, is a crucial component of the brain. The neuron transmits a small electric pulse along its length, triggering the release of chemicals called neurotransmitters into the synapse. The neurotransmitters traverse the synaptic gap and trigger a response in the neighboring neuron.
The timing of the electrical pulses helps determine how large of a chemical signal gets sent. In some neurons, repeated stimulations yields stronger, or facilitated, firings.
In others, multiple stimulations elicit weaker, or depressed, responses. These adaptations, known as short-term plasticity, happen within milliseconds.
Previous mock-neural networks required at least seven transistors to replicate short-term plasticity. The new transistor, called the nanoparticle organic memory field-effect transistor, or NOMFET for short, does it with just one.
That's important, because the smaller and more adaptable the transistors, the cheaper and easier it will be to scale from a few artificial synapses to thousands, Vuillaume said.
To build NOMFET, Vuillaume and his team placed gold nanoparticles in a trough between two electrodes. The particles, just five to 20 nanometers across, were covered with a very thin layer of a substance called pentacene, which conducts electricity.
Positive charges called "holes," which are created by missing electrons in the pentacene, transmit the current across this valley of scattered gold.
At each voltage input, some holes are temporarily trapped by the gold, and this changes the electrical output of the transistor. Depending on the voltages used, NOMFET can produce either weaker or stronger outputs – just like human neurons undergoing short-term plasticity.
Because of this adaptability, NOMFET is more flexible than traditional transistors, the researchers say.
The research "is definitely an interesting and well-conceived work," said physicist Massimiliano Di Ventra of the University of California, San Diego, who was not involved in the study.
The next step, Vuillaume said, is to combine several NOMFET transistors together to see how closely they approximate real neural circuits.
The research is detailed in a recent issue of the journal Advanced Functional Materials.
- 10 Profound Innovations Ahead
- The End of Silicon May Be in Sight
- Engineering the Computer of the Future, One Atom at a Time