Researchers boost reliability of low-power alternative to flash memory

Ferroelectric random access memory, or FeRAM, has long promised to offer a low-power alternative to flash computer memory, but its limited endurance cannot offset its deficiency in density such that this option is stuck as a small part of the wider semiconductor market. Moreover, FeRAM suffers from reliability issues such as fatigue and imprint. A fresh investigation of the crystal structure and device characteristic of a hafnium-based FeRAM however has allowed researchers to understand the root cause of the problem and propose ways to overcome it.

 

A paper describing the investigation appeared in the journal Nano Research on January 17.

 

Random access memory, or RAM, on your computer is where data from applications you are actively using are stored temporarily and kept rapidly accessible. This short-term storage is described as ‘volatile’, meaning it is stored electrically via transistors and so disappears the minute that you shut off your computer (because there is no longer applied voltage). Meanwhile read-only memory, or ROM, is described as non-volatile, as it is written permanently on a chip in individual cells in binary code. As a result, it can still hold on to data when the computer is off—but at the cost of much slower performance.

 

Flash memory, common in computer hardware for some two decades, offers the benefits of both RAM and ROM: rapid data retrieval with no data loss accompanying power loss. Flash works by applying a high voltage across a transistor so that an electron gets stuck at one end—permitting permanent instead of temporary storage of binary 0’s and 1’s: when the computer is turned off, that electron is still there.

 

In recent years however, manufacturers of integrated circuits (microchips) have begun to take another look at ferroelectric random access memory or FeRAM as a superior alternative to flash memory. When an electric field is applied to the ferroelectric material, small alterations to the positions of its atoms change the electronic charge or ‘polarity’ of the material’s crystal structure. When the electric field is removed, that polarized state remains, again permitting permanent storage of information in binary form.

 

But unlike flash, FeRAM does not depend on high voltages, meaning it enjoys far lower power use. Flash memory also takes at least a millisecond to complete a ‘write’ of memory, while FeRAM performs this in mere nanoseconds.

 

Despite this faster performance, lower power use, and expanded endurance of read and write capability, FeRAM faces reliability challenges from phenomena called ‘imprint’ and ‘fatigue’. Together with the lower memory density in traditional 2T2C architecture, while FeRAM was commercialized in the 1990s, to this day, it enjoys a relatively small market share of the semiconductor market. The discovery of ferroelectricity in Hf_0.5Zr_0.5O₂ (or more simply, HZO) promises a much greater memory density for FeRAM as 3D integration becomes possible. Unfortunately, HZO suffers from more severe imprint problem compared with traditional perovskite ferroelectrics.

 

 

Imprint describes the tendency of the crystal structure of the ferroelectric material to prefer a particular polarization state as a result of previous writes to that state. Fatigue is a related phenomenon where after multiple cycles of writing, there is a need for ever greater voltages to achieve the same crystal polarization, diminishing FeRAM’s power advantages over flash.

 

“You can think of ‘imprint’ as similar to how repeated brushing of one’s hair in a particular direction makes your hair resistant to being brushed in a different direction,” said Kan-Hao Xue, a professor with the School of Integrated Circuits at Huazhong University of Science and Technology and a co-author of the article. “And ‘fatigue’ is similar to how an alcoholic requires ever greater amounts of vodka to achieve the same ‘high’ that just one martini used to produce when they first started to drink.”

 

Engineers are familiar with the broad strokes of the imprint phenomenon, but not its deeper, underlying mechanisms. The researchers wanted to understand its root cause in order to possibly overcome it, and so they systematically investigated a novel ferroelectric material that makes use of the metal hafnium, Hf0.5Zr0.5O2, (or more simply, HZO).

 

Using a transmission electron microscopy image of a thin film of HZO to consider the structure of the material, the researchers saw that the ‘injection’ of electric charge across it results in an ‘unsticking’ of the ‘stuck’ electrons from oxygen atom vacancies within the molecule. These were the locations where they had previously been trapped. It is interesting that at least three distinct imprint mechanisms are prevailing for traditional ferroelectrics, but now it is shown that for an ultra-thin ferroelectric HZO with low permittivity, the impact of charge injection overwhelms others, and the imprint in HZO is actually much simpler.

 

Then, using this new model of how imprint works, the researchers used a high unipolar voltage stimulus (that is still lower than the high voltages required by flash) to cause electron trapping on the surface or near the surface of these oxygen vacancies. This reduced the problem by 90 percent compared to normal. In addition, the researcher found that by baking the capacitors in the memory architecture at a high temperature, this achieved still better performance.

 

Such techniques will improve device reliability of hafnium-based FeRAMs, but the researchers now want to go further, and investigate and overcome fatigue using HZO FeRAMs.

 

 

###

 

About Nano Research 

 

Nano Research is a peer-reviewed, international and interdisciplinary research journal, sponsored by Tsinghua University and the Chinese Chemical Society. It offers readers an attractive mix of authoritative and comprehensive reviews and original cutting-edge research papers. After more than 10 years of development, it has become one of the most influential academic journals in the nano field. Rapid review to ensure quick publication is a key feature of Nano Research. In 2020 InCites Journal Citation Reports, Nano Research has an Impact Factor of 8.897 (8.696, 5 years), the total cites reached 23150, and the number of highly cited papers reached 129, ranked among the top 2.5% of over 9000 academic journals, ranking first in China's international academic journals.

 

About SciOpen 

 

SciOpen is a professional open access resource for discovery of scientific and technical content published by the Tsinghua University Press and its publishing partners, providing the scholarly publishing community with innovative technology and market-leading capabilities. SciOpen provides end-to-end services across manuscript submission, peer review, content hosting, analytics, and identity management and expert advice to ensure each journal’s development by offering a range of options across all functions as Journal Layout, Production Services, Editorial Services, Marketing and Promotions, Online Functionality, etc. By digitalizing the publishing process, SciOpen widens the reach, deepens the impact, and accelerates the exchange of ideas.

 

 

---

---