Sulfonated polyaniline (SPAN) is a conducting polymer that is self-doped. It has a high water solubility and a novel pH-dependent DC conductivity, which is interesting for fundamental science as well as applications in areas such as rechargeable batteries and pH control technologies.
A group of scientists built a sulfonated polyaniline (SPAN) organic electrochemical network device (OEND) for reservoir computing. SPAN was deposited on gold electrodes, resulting in the formation of a disordered network with humidity-dependent electrical properties. The SPAN OEND was tested for reservoir computing using benchmark tasks and spoken-digit classification, and it was performed with 70% accuracy. The device could be used for a variety of artificial intelligence tasks, including speech recognition.
Reservoir computing is a world-class machine learning algorithm for processing information generated by dynamical systems based on observed time-series data. Notably, it necessitates very small training data sets, employs linear optimization, and thus necessitates minimal computing resources.
Researchers assembled a sulfonated polyaniline (SPAN) organic electrochemical network device (OEND) for use in reservoir computing. SPAN was deposited on gold electrodes which formed a disordered network providing humidity-dependent electrical properties.
Reservoir computing (RC) addresses complex problems by simulating how information is processed in animal brains. It is based on a randomly connected network that serves as an information reservoir and ultimately leads to more efficient outputs. Numerous reservoir materials have been investigated to date in order to realize RC directly in matter (rather than simulating it in a digital computer). A team led by Osaka University researchers has developed a sulfonated polyaniline network for RC.
Electrochemical signals carried by ions are used by neural networks in the brain. As a result, when selecting a material system for RC, an electrochemical approach is a logical choice. Organic electrochemical field-effect transistors (OECFETs) are widely used in bioelectronics, but they have yet to be widely used in RC.
The reservoir material’s key feature is that it has rich (time-dependent) behavior and is disordered, which makes polymer materials an excellent choice because they form random networks on their own.
Polyaniline is a promising polymer for RC applications because it is simple to polymerize, has good atmospheric stability, and has reversible doping/de-doping behavior, which means its conduction can be changed.
The researchers looked into sulfonated polyaniline (SPAN), which has high water solubility and self-doping behavior in addition to the benefits of polyaniline. These improvements make SPAN easier to work with and the doping more consistent.
“Atmospheric protons are injected directly into the polymer chain of SPAN, causing it to conduct,” study lead author Yuki Usami explains. “The humidity can then be adjusted to control this conduction.”
The SPAN was assembled on gold electrodes using a simple drop-casting method, yielding an organic electrochemical network device (OEND). The SPAN OEND was evaluated for RC by examining the waveform and evaluating its performance in short-term memory tasks. The results of a test to see how well speech recognition could be achieved 70 percent accuracy. This capability of SPAN OEND was comparable to an RC software simulation.
“We demonstrated that our SPAN OEND system can be used in RC,” says study co-author Takuya Matsumoto. “Future efforts to develop systems that do not rely on humidity will provide more practical options; however, the success of our SPAN-based system is a positive step forward for material-based reservoir computing, which is expected to have a significant impact on the next generation of artificial intelligence devices.”
