New molecular device has unprecedented reconfigurability reminiscent of brain plasticity

Credit: Pixabay/CC0 public domain
A team of international researchers discovered a new molecular device that can perform exceptional computing power in a publication published in Nature.

The device is similar to the plasticity of human brain connections. It can be reconfigured quickly for different computational tasks by changing the applied voltages. The same way that nerve cells store memories, the same device also has the ability to retain information for future retrieval or processing.

"The brain is capable of changing its wiring by connecting and disconnecting nerve cells. It was extremely difficult to achieve something similar in a physical system," stated Dr. R. Stanley Williams from Texas A&M University's Department of Electrical and Computer Engineering. "We now have a molecular device that can be reconfigured in a dramatic way. This is not possible by altering physical connections, as it happens in the brain. It's done by reprogramming its logic."

T. Venkatesan is the director of the Center for Quantum Research and Technology at the University of Oklahoma. He is also a Scientific Affiliate at the National Institute of Standards and Technology in Gaithersburg. Additionally, he said that their molecular device could be used to design next-generation processing chip designs with increased computational power and speed while consuming significantly less energy.

Digital technologies, whether they are the common laptop or the supercomputer, face the von Neumann bottleneck. The current computer architectures are such that the memory, which contains data and programs, is physically isolated from the processor, this causes delays in computational processing. Computers spend a lot of time switching information between them, which causes the bottleneck. These units can also be idled for long periods of information exchange, despite their extremely high processor speeds.

Memristors are an alternative to traditional electronic parts that are used to design memory units and processors. At a certain temperature, memristors such as those made from vanadium dioxide or niobium dioxide, can transition from being an insulation to a conductor. These memristors can perform computations and store information.

These metal oxide memristors have many benefits, but they are limited in their ability to operate at restricted temperatures. Williams said that there is a constant search for organic molecules capable of performing a similar memristive function.

The material was designed by Dr. Sreebrata Ghoswami, an Indian Association for the Cultivation of Science professor. The central metal atom of the compound (iron) is bound to three phenylazo pyridine organic molecules known as ligands.

Sreebrata stated that this behaves like an electron sponge and can absorb six electrons reversibly. This results in seven different redox state. This work demonstrates reconfigurability through the interconnectivity of these states.

This project was created by Dr. SreetoshGoswami, a researcher from the National University of Singapore. It consists of a 40-nanometer thick layer of molecular films sandwiched between layers of gold and nanodiscs and indium tin dioxide at the bottom.

Sreetosh noticed a profile of current-voltage that was unlike anything he had ever seen. The organic molecular devices can switch between insulator and conductor at multiple discrete voltages, unlike metal-oxide memristors which can only switch from metal to an insulator at one fixed voltage.

"If you see the device as an off-off switch, it will first turn from on to off, then to on and back to on. As we were drawing the voltage more negative, the device switched from on on to off, then to off to on. Venkatesan said, "We were just blown to our seats." "We had to convince our selves that what we saw was real."

Sreetosh, Sreebrata studied the molecular mechanisms that underlie this curious switching behavior. They used an imaging technique called Raman spectrumcopy. They were particularly interested in spectral signatures that could explain multiple transitions. They discovered that the voltage was swept to the negative, causing the ligands to undergo a series electron-gaining events, which caused the molecule's transition from off to on state.

Williams then used a mathematical approach to describe the complex current-voltage profile for the molecular device. This was a departure from the traditional approach of using basic physics-based equations. He instead used a decision tree algorithm that included "if-then" statements to describe the molecular behavior. This is a common line of code found in many computer programs, especially digital games.

Video games use a structure that allows you to have a character do something and then have something happen. Williams explained that if you put it in a computer algorithm they will be if-then else statements. "Here, the molecular is switching on and off as a result of applied voltage. That's when I had my eureka moment and decided to use decision trees to explain these devices. It worked very well."

Researchers went one step further and used these molecular devices for running programs for real-world computation tasks. Sreetosh demonstrated experimentally that the devices could do complex computations in one time step, and then be reprogrammed to complete another task in the next moment.

Sreetosh said, "It was quite remarkable; our device was doing something similar to what the brain does but in a very unique way." The brain can reconfigure or change the physical wiring when it's learning new things or when it's deciding. We can also logically reprogram our devices or reconfigure them by giving them a voltage pulse that is different from what they have seen before.

Venkatesan pointed out that thousands of transistors would be required to perform the same computation functions as a molecular device with different decision trees. Venkatesan suggested that their technology could be first used in handheld devices such as cell phones, sensors, and other applications with limited power.

The research was also supported by Dr. Abhijeet and Dr. Ariando of the National University of Singapore, Dr. Rajib Pramanick, and Dr. Santiprasad Rath, both from the Indian Association for the Cultivation of Science. Dr. Martin Foltin, Hewlett Packard Enterprise in Colorado, and Dr. Damien Thompson, University of Limerick, Ireland.

Venkatesan stated that this research was indicative of future discoveries by this collaborative team. This will include the center of Nanoscience and Engineering at the Indian Institute of Science and at the Microsystems and Nanotechnology Division of the NIST.

More information: Sreetosh Ghoswami et. al. Decision trees within the molecular memristor. Nature (2021). Information from the Journal: Nature Sreetosh Goswami and colleagues, Decision trees within an molecular memristor (2021). DOI: 10.1038/s41586-021-03748-0