Raisul Islam
Purdue University, IN·Engineering
Today's artificial intelligence takes inspiration from the brain but works very differently, and it burns enormous amounts of energy to train and run. Industry projections warn that if models keep growing while chips keep scaling the same way, machine learning could demand more electricity than the world produces.
- Fields
Engineering
Materials Science
Computer Science
- Topics
neuromorphic computing
memtransistor
ferroelectric field-effect transistor
- Grants
- 1 known grant, through Sep 2028
What they work on
- Fabricate and characterize the Diffusive Ferroelectric Field-Effect Transistor (DFeFET), combining engineered drain contact metals with amorphous oxide semiconductor channels in a ferroelectric-gated transistor to obtain controllable volatile hysteresis in drain current-voltage curves.
- Test whether the device's short-term memory arises from reversible ion or vacancy exchange at the drain/channel interface, modulated by gate voltage and gradual switching of the ferroelectric gate polarization.
- Use the DFeFET to implement three biologically inspired spike frequency adaptation mechanisms, in which a spiking neuron self-regulates its firing, with lower energy and smaller area than conventional circuits.
- Build a CMOS-compatible physical reservoir computing architecture with task-specific timescale adaptability and coupled higher-order nonlinear dynamics, using a single device type for both reservoir and readout layers.
- Develop a device-algorithm co-optimization framework that tunes accuracy, latency, area, and energy efficiency for these analog neuromorphic implementations.
More about the research
The brain, by contrast, does complex computation on very little power. One feature it uses that computer hardware largely ignores is having short-term and long-term memory that are physically intertwined in the same elements. This project builds a transistor that does the same thing.
Why it matters
If machine learning keeps growing on conventional hardware, its energy demand could become unsustainable. A single transistor that holds both short- and long-term memory could cut the energy and chip area needed for neuromorphic processors, simplify their architectures, and support efficient processing of time-varying signals. Such devices may benefit designers of AI hardware and of post-silicon technologies, and the project also offers interdisciplinary training across engineering, physics, chemistry, neuroscience, and AI.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Shantanu Chakrabartty
Washington University in St Louis, MO·Engineering
Model single-electron transport across a Fowler-Nordheim quantum tunneling barrier as the equivalent of a neural spike in NeuroSA, a spiking architecture where neuron firing thresholds follow tunneling physics.
Jonathan Bird
University at Buffalo, NY·Engineering
Build non-volatile transistors that combine a voltage-switched magneto-electric gate with a channel made of transition-metal dichalcogenides, two-dimensional semiconductors only atoms thick, and show they retain their programmed state without power.
Rashmi Jha
University of Cincinnati, OH·Engineering
Model, design and synthesize specific synthetic DNA sequences, varying base pairs, sequence, length and crystallinity, for integration with organometallic halide perovskite semiconductors.
The school
Purdue University: 43 labs on PI Pages, $932M a year on research, 833 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.