Self-powered dynamical systems

While developing low-power instrumentation, I started investigating the use of self-powered dynamical systems for battery-free recording of bio-signals. The AIMLab has developed scalable dynamical systems that run for a long duration and without any external power. The system consists of a floating-gate node (fabricated in standard CMOS processes) where electrons are allowed to slowly tunnel (via Fowler-Nordheim tunneling) into the floating-gate.

Once initialized, the system acts autonomously without any external power (similar to an hourglass). The graph on the right shows typical evolution of floating gate voltage as a function of time. There are two important takeaways here

FN leakage is slow – data are shown for a million seconds (~11 days)

FN leakage is reliable – >99% accuracy between the model and the data

These two features allow us to create reliable long term self-powered dynamical systems. These systems can be repurposed for a variety of real-world applications. Here are some of the applications that we have investigated

1. Self-powered time-keeping – At the core, the dynamical systems are very similar to hourglasses. Instead of sand, we have electrons. And, because we use electrons, we get immediate access to the immense technological progress we have made in electronics. We can fit 100s of our hourglasses on a tiny spec of dust. Because of the slow FN tunneling, our system can last for months, instead of hours. And we can measure charge down to the precision of few electrons.

2. Self-powered sensing and data logging – We have shown that we can interface the dynamical system with an external signal which would affect the dynamics. In this case, we would make the measurement at a known time and see how far we are away from the model. This deviation contains information about the signal history.

3. Dynamic authentication – The self-powered timer can be used to generate an authentication token. Briefly, each chip would have a set of timers with different initial conditions. These conditions are stored on a server. At any given time, the timer values are used to generate an authentication token on the chip. As the timers are dynamic, a different token will be generated after a pre-defined time interval. The server which has knowledge of the timer models generates its own tokens. The server and the chip tokens are compared. If they match, the chip is authenticated. The dynamical systems have several features that make them superior to static solutions. They are inherently securing against man in the middle attacks because same tokens are never used twice. They are tamper resistant, because any attempt at tampering would change the timer value which would render it unable to be authenticated. Moreover, it may be possible to estimate when tampering events occurred. By using industry-standard cryptographic methods, it would be impossible to generate the system model from the tokens.

4. Environmental signature – This is similar to the sensing application. It provides additional means of authenticating devices and ensuring compliance of the entire supply chain.

5. Time stamping of rare events – During occurrence of rare events, the time value of the system can be logged in.