IBM scientist Abu Sebastian develops future memory and computer paradigms with prestigious European grant

In 2007 the European Research Council (ERC) began funding frontier research projects through individual grants.

The grants are split into several categories including consolidator grants, which target world-class mid-career scientists. In 2015 the ERC awarded a total of €585 million, as part of the European Union Research and Innovation programme Horizon 2020 with grants worth millions of Euro each.

Nine IBM scientists in Zurich currently hold ERC grants and three have been recently awarded to Abu Sebastian, Kristen Moselund and Leo Gross.

Sebastian, who is based in Zurich, is focused on exploratory memory and cognitive technologies and his grant will focus on investigating resistive memory and its applications in non-von Neumann computing.

Before the hard work begins I asked him about the grant and his specific plans.

Resistive memory devices are being developed by several organizations around the world. Can you explain what makes your memory device different?

Abu Sebastian (AS): Resistive memory devices are nanoscale devices whose resistance depends on the history of the current that had previously flowed through it.

We can store information in these decides using the resistance as the storage variable, much like using the charge state of a capacitor to store information in DRAM and Flash.

These devices also possess other interesting attributes that go beyond information storage. It is believed that they could have a significant influence on future computing systems.

Therefore, it is not surprising that several research institutions are conducting research in this field. I should note that phase change memory is a type of resistive memory that we have been researching at IBM for nearly a decade.

In conventional resistive memory devices, the physical mechanism of resistance storage is coupled to the information-retrieval process. For example, in a phase change memory device, we use the same material for writing information, by making it undergo a phase transition, and for retrieving the information stored, by reading its low-field electrical resistance. In a projected memory device, this coupling is broken down by a careful design of the device geometry. This could lead to a dramatic improvement in its performance. This approach could also open up new applications.

In the ERC project, we will explore certain highly innovative projected memory designs and their applications in non-von Neumann computing where there is no physical separation between between the central processing unit (CPU) and the memory.

Your application is split into three parts. What are the time frames for each?

AS: Two of them focus more on the technology side of the project, where we investigate the design and fabrication of projected memory devices, and one on the algorithmic side, where we focus on applications. My guess is that these two aspects of the project will consume an equal amount of time and resources. I am hoping that, since it is an ERC project, there will be a certain amount of leeway in the execution of the project.

What impact will this technology have on society?

AS: As I mentioned earlier, resistive memory devices could have a dramatic impact on future computing systems, in particular on cognitive computing: the third era of computing where we derive insights and intelligence from the vast amount of data at our disposal. Resistive memory devices could play a critical role either as an ultra-fast, ultra-dense non-volatile cache memory or even as elements of non-von Neumann processers or co-processors. It is believed that the new concepts being pursued in the project could help make some key steps in this direction.

I believe you are looking to build up a team with your grant. What skills are you looking for?

AS: My emphasis will be on assembling a very competent research team with highly multi-disciplinary skills. The team as a whole will have competence all the way from physical sciences to the area of “mathematical engineering” – a term coined by the famous mathematician Norbert Wiener to cover the areas of signal processing, information theory, control, computation etc. Luckily at IBM we already have some very good resources to tap into. To fill any gaps, we will hire a few additional postdocs and students. If anyone is interested in applying please email me.

What advice do you have for future ERC applicants?

AS: The ERC application process was indeed a great learning experience. There are several pieces of advice I could give to potential applicants. But if I have to focus on 2 or 3, they would be the following ones:

• Be sure the proposed project falls into the high risk/high reward category
• Be fully committed to the effort. It is quite an ordeal and demands significant preparation
• Get as much help as possible. In my case. I got significant help from my colleagues, in particular Tomas Tuma and Wabe Koelmans. I also received advice and guidance from people outside my immediate research area who are also ERC grant winners including Govind Kaigala and Kirsten Moselund. Not to mention Catherine Trachsel, from our business development team at ZRL, Charlotte Bolliger from the publication team, and also Euresearch, the Swiss entity providing advice on European research programs.