Tag Archives: Yimei Zhu

An ‘artificial brain’ and life-long learning

Talk of artificial brains (also known as, brainlike computing or neuromorphic computing) usually turns to memory fairly quickly. This February 3, 2022 news item on ScienceDaily does too although the focus is on how memory and forgetting affect the ability to learn,

When the human brain learns something new, it adapts. But when artificial intelligence learns something new, it tends to forget information it already learned.

As companies use more and more data to improve how AI recognizes images, learns languages and carries out other complex tasks, a paper publishing in Science this week shows a way that computer chips could dynamically rewire themselves to take in new data like the brain does, helping AI to keep learning over time.

“The brains of living beings can continuously learn throughout their lifespan. We have now created an artificial platform for machines to learn throughout their lifespan,” said Shriram Ramanathan, a professor in Purdue University’s [Indiana, US] School of Materials Engineering who specializes in discovering how materials could mimic the brain to improve computing.

Unlike the brain, which constantly forms new connections between neurons to enable learning, the circuits on a computer chip don’t change. A circuit that a machine has been using for years isn’t any different than the circuit that was originally built for the machine in a factory.

This is a problem for making AI more portable, such as for autonomous vehicles or robots in space that would have to make decisions on their own in isolated environments. If AI could be embedded directly into hardware rather than just running on software as AI typically does, these machines would be able to operate more efficiently.

A February 3, 2022 Purdue University news release (also on EurekAlert), which originated the news item, provides more technical detail about the work (Note: Links have been removed),

In this study, Ramanathan and his team built a new piece of hardware that can be reprogrammed on demand through electrical pulses. Ramanathan believes that this adaptability would allow the device to take on all of the functions that are necessary to build a brain-inspired computer.

“If we want to build a computer or a machine that is inspired by the brain, then correspondingly, we want to have the ability to continuously program, reprogram and change the chip,” Ramanathan said.

Toward building a brain in chip form

The hardware is a small, rectangular device made of a material called perovskite nickelate,  which is very sensitive to hydrogen. Applying electrical pulses at different voltages allows the device to shuffle a concentration of hydrogen ions in a matter of nanoseconds, creating states that the researchers found could be mapped out to corresponding functions in the brain.

When the device has more hydrogen near its center, for example, it can act as a neuron, a single nerve cell. With less hydrogen at that location, the device serves as a synapse, a connection between neurons, which is what the brain uses to store memory in complex neural circuits.

Through simulations of the experimental data, the Purdue team’s collaborators at Santa Clara University and Portland State University showed that the internal physics of this device creates a dynamic structure for an artificial neural network that is able to more efficiently recognize electrocardiogram patterns and digits compared to static networks. This neural network uses “reservoir computing,” which explains how different parts of a brain communicate and transfer information.

Researchers from The Pennsylvania State University also demonstrated in this study that as new problems are presented, a dynamic network can “pick and choose” which circuits are the best fit for addressing those problems.

Since the team was able to build the device using standard semiconductor-compatible fabrication techniques and operate the device at room temperature, Ramanathan believes that this technique can be readily adopted by the semiconductor industry.

“We demonstrated that this device is very robust,” said Michael Park, a Purdue Ph.D. student in materials engineering. “After programming the device over a million cycles, the reconfiguration of all functions is remarkably reproducible.”

The researchers are working to demonstrate these concepts on large-scale test chips that would be used to build a brain-inspired computer.

Experiments at Purdue were conducted at the FLEX Lab and Birck Nanotechnology Center of Purdue’s Discovery Park. The team’s collaborators at Argonne National Laboratory, the University of Illinois, Brookhaven National Laboratory and the University of Georgia conducted measurements of the device’s properties.

Here’s a link to and a citation for the paper,

Reconfigurable perovskite nickelate electronics for artificial intelligence by Hai-Tian Zhang, Tae Joon Park, A. N. M. Nafiul Islam, Dat S. J. Tran, Sukriti Manna, Qi Wang, Sandip Mondal, Haoming Yu, Suvo Banik, Shaobo Cheng, Hua Zhou, Sampath Gamage, Sayantan Mahapatra, Yimei Zhu, Yohannes Abate, Nan Jiang, Subramanian K. R. S. Sankaranarayanan, Abhronil Sengupta, Christof Teuscher, Shriram Ramanathan. Science • 3 Feb 2022 • Vol 375, Issue 6580 • pp. 533-539 • DOI: 10.1126/science.abj7943

This paper is behind a paywall.

Skyrmions and ultra-thin multilayer film

The National University of Singapore (NUS) and skyrmions are featured in an April 10, 2017 news item on phys.org,

A team of scientists led by Associate Professor Yang Hyunsoo from the Department of Electrical and Computer Engineering at the National University of Singapore’s (NUS) Faculty of Engineering has invented a novel ultra-thin multilayer film which could harness the properties of tiny magnetic whirls, known as skyrmions, as information carriers for storing and processing data on magnetic media.

The nano-sized thin film, which was developed in collaboration with researchers from Brookhaven National Laboratory, Stony Brook University, and Louisiana State University, is a critical step towards the design of data storage devices that use less power and work faster than existing memory technologies. The invention was reported in prestigious scientific journal Nature Communications on 10 March 2017.

An April 10, 2017 NUS press release on EurekAlert, which originated the news item, describes the work in more detail,

Tiny magnetic whirls with huge potential as information carriers

The digital transformation has resulted in ever-increasing demands for better processing and storing of large amounts of data, as well as improvements in hard drive technology. Since their discovery in magnetic materials in 2009, skyrmions, which are tiny swirling magnetic textures only a few nanometres in size, have been extensively studied as possible information carriers in next-generation data storage and logic devices.

Skyrmions have been shown to exist in layered systems, with a heavy metal placed beneath a ferromagnetic material. Due to the interaction between the different materials, an interfacial symmetry breaking interaction, known as the Dzyaloshinskii-Moriya interaction (DMI), is formed, and this helps to stabilise a skyrmion. However, without an out-of-plane magnetic field present, the stability of the skyrmion is compromised. In addition, due to its tiny size, it is difficult to image the nano-sized materials.

To address these limitations, the researchers worked towards creating stable magnetic skyrmions at room temperature without the need for a biasing magnetic field.

Unique material for data storage

The NUS team, which also comprises Dr Shawn Pollard and Ms Yu Jiawei from the NUS Department of Electrical and Computer Engineering, found that a large DMI could be maintained in multilayer films composed of cobalt and palladium, and this is large enough to stabilise skyrmion spin textures.

In order to image the magnetic structure of these films, the NUS researchers, in collaboration with Brookhaven National Laboratory in the United States, employed Lorentz transmission electron microscopy (L-TEM). L-TEM has the ability to image magnetic structures below 10 nanometres, but it has not been used to observe skyrmions in multilayer geometries previously as it was predicted to exhibit zero signal. However, when conducting the experiments, the researchers found that by tilting the films with respect to the electron beam, they found that they could obtain clear contrast consistent with that expected for skyrmions, with sizes below 100 nanometres.

Dr Pollard explained, “It has long been assumed that there is no DMI in a symmetric structure like the one present in our work, hence, there will be no skyrmion. It is really unexpected for us to find both large DMI and skyrmions in the multilayer film we engineered. What’s more, these nanoscale skyrmions persisted even after the removal of an external biasing magnetic field, which are the first of their kind.”

Assoc Prof Yang added, “This experiment not only demonstrates the usefulness of L-TEM in studying these systems, but also opens up a completely new material in which skyrmions can be created. Without the need for a biasing field, the design and implementation of skyrmion based devices are significantly simplified. The small size of the skyrmions, combined with the incredible stability generated here, could be potentially useful for the design of next-generation spintronic devices that are energy efficient and can outperform current memory technologies.”

Next step

Assoc Prof Yang and his team are currently looking at how nanoscale skyrmions interact with each other and with electrical currents, to further the development of skyrmion based electronics.

Here’s a link to and a citation for the paper,

Observation of stable Néel skyrmions in cobalt/palladium multilayers with Lorentz transmission electron microscopy by Shawn D. Pollard, Joseph A. Garlow, Jiawei Yu, Zhen Wang, Yimei Zhu & Hyunsoo Yang. Nature Communications 8, Article number: 14761 (2017) doi:10.1038/ncomms14761 Published online: 10 March 2017

This is an open access paper.

Nanoscale imaging gets rough

Smooth is easier than rough when imaging at the nanoscale according to a June 17, 2015 Northwestern University news release by Megan Fellman (also on EurekAlert),

A multi-institutional team of scientists has taken an important step in understanding where atoms are located on the surfaces of rough materials, information that could be very useful in diverse commercial applications, such as developing green energy and understanding how materials rust.

Researchers from Northwestern University, Brookhaven National Laboratory, Lawrence Berkeley National Laboratory and the University of Melbourne, Australia, have developed a new imaging technique that uses atomic resolution secondary electron images in a quantitative way to determine the arrangement of atoms on the surface.

Many important processes take place at surfaces, ranging from the catalysis used to generate energy-dense fuels from sunlight and carbon dioxide to how bridges and airplanes corrode, or rust. Every material interacts with the world through its surface, which is often different in both structure and chemistry from the bulk of the material.

The real focus of the work is on corrosion, according to the news release,

“We are excited by the possibilities of applying our imaging technique to corrosion and catalysis problems,” said Laurence Marks, a co-author of the paper and a professor of materials science and engineering at Northwestern’s McCormick School of Engineering and Applied Science. “The cost of corrosion to industry and the military is enormous, and we do not understand everything that is taking place. We must learn more, so we can produce materials that will last longer.”

To understand these processes and improve material performance, it is vital to know how the atoms are arranged on surfaces. While there are many good methods for obtaining this information for rather flat surfaces, most currently available tools are limited in what they can reveal when the surfaces are rough.

Scanning electron microscopes are widely used to produce images of many different materials, and roughness of the surface is not that important. Until very recently, instruments could not obtain clear atomic images of surfaces until a group at Brookhaven managed in 2011 to get the first images that seemed to show the surfaces very clearly. However, it was not clear to what extent they really were able to image the surface, as there was no theory for the imaging and many uncertainties.

The new work has answered all these questions, Marks said, providing a definitive way of understanding the surfaces in detail. What was needed was to use a carefully controlled sample of strontium titanate and perform a large range of different types of imaging to unravel the precise details of how secondary electron images are produced.

“We started this work by investigating a well-studied material,” said Jim Ciston, a staff scientist at Lawrence Berkeley National Laboratory and the lead author of the paper, who obtained the experimental images. “This new technique is so powerful that we had to revise much of what was already thought to be well-known. This is an exciting prospect because the surface of every material can act as its own nanomaterial coating, which can greatly change the chemistry and behavior.”

“The beauty of the technique is that we can image surface atoms and bulk atoms simultaneously,” said Yimei Zhu, a scientist at Brookhaven National Laboratory. “Currently, no existing methods can achieve that.”

Les Allen, who led the theoretical and modeling aspects of the new imaging technique in Melbourne, said, “We now have a sophisticated understanding of what the images mean. It now will be full steam ahead to apply them to many different types of problems.”

Here’s a link to and citation for the paper,

Surface determination through atomically resolved secondary-electron imaging by J. Ciston, H. G. Brown, A. J. D’Alfonso, P. Koirala, C. Ophus, Y. Lin, Y. Suzuki, H. Inada, Y. Zhu, L. J. Allen, & L. D. Marks. Nature Communications 6, Article number: 7358 doi:10.1038/ncomms8358 Published 17 June 2015

This paper is open access.