In recognition of National Inventors’ Day, IBM researcher, Abu Sebastian, writes about his life as an inventor and his research on phase-change memory devices.
Abu Sebastian was named Master Inventor in 2016 and holds around 40 granted patents.
Inventions come in different forms. Often, they are specific tricks or improvements that help us overcome a roadblock that presents itself when developing a specific product or prototype. But what really excites me is when we create something completely out of the box and not immediately obvious. Those inventions have the potential to open up entirely new research areas.
One specific example is how we countered the problem of resistance drift in phase-change memory devices. Phase-change memory (PCM) is arguably the most advanced emerging resistive memory technology and is finding widespread application in areas such as storage-class memory and non-von Neumann computing. The essential idea of PCM is to store information in terms of the atomic arrangement of certain materials such as Ge2Sb2Te5.
Typically, in the ordered crystalline phase, these materials have a low resistivity, whereas in the disordered amorphous phase, they exhibit high resistivity. If a tiny volume of these materials is sandwiched between two metal electrodes, we can reversibly change the phase configuration of these materials in a continuous and reversible manner. The phase-configurational information can be deciphered by measuring the resistance of these devices.
A key challenge associated with PCM is that the resistance values of these PCM devices increase with time, a phenomenon typically referred to as resistance drift. Most approaches to this problem revolved around using reference devices to keep track of the drift, better detection schemes etc. One of my key inventions was to question whether using the low-field resistance is the right way to decipher the different states.
We subsequently conceived a non-resistance-based metric (US Patent 8,755,214) that is a much better representative of the phase configuration. This invention has led to several other inventions involving clever read-out circuits to implement this scheme as well as new drift-resilient PCM cell designs (US Patent 9,530,493).
Another example of out-of-the-box thinking is when I was trying hard to come up with a solution in which a high-level computational task can be performed in the memory itself without having to shuttle data back and forth between the memory and a separate processing unit. One of my colleagues and I were having one of our routine white-board discussions, during which we hit upon a clever idea, namely, to exploit the crystallization dynamics of PCM devices to solve an unsupervised learning algorithm entirely in-memory (US Patent 9,466,364).
Some of the inventions, however, arise purely by luck.
Once, we were characterizing the switching behavior of various types of phase-change materials using a conductive-mode atomic force microscope. One of our collaborators in the UK was depositing the material stack for us. To protect the phase-change materials from oxidation, he used a capping material.
In one sample, he forgot to deposit the phase-change material and only had the capping material. Without knowing this, we performed our routine switching studies on his sample, and to our surprise, found that the capping material was also switching similarly to the phase-change material. This led to further research and inventions that culminated in a new type of resistive memory technology (US Patent 9,105,842 and US Patent 9,240,550).
What I’ve learned from all these experiences is that inventions rarely occur in a vacuum. Often, I need to immerse myself into a stream of different ideas. White-board discussions, chats in the cafeteria, group meetings, workshops, conferences — they all help in this direction. They are all like the inputs to an integrate-and-fire biological neuron: The membrane potential reaches a certain threshold and typically results in an inventive spike. It is also essential to have some quiet time, free of disturbance, to contemplate various ideas at leisure. My personal day of choice is Sunday.
In a new paper published in Nature Electronics, IBM researchers demonstrate the smallest ever built DRAM memory cell, fifty years after its invention. The new DRAM cells feature potentially low power consumption and an unprecedented small footprint. They could be therefore particularly appealing for implementation in mobile devices or as cache memory.