IMEC proposes liquid memories for higher density storage
European research institute IMEC (Leuven, Belgium) has proposed a couple of liquid memory systems to be used for storage instead of NAND flash solid-state memory.
One or other of these systems could provide an alternative to solid-state drives (SSDs) that will slow-down in their ability to scale density over time by the later part of this decade, IMEC said.
The two systems are called colloidal and electrolithic memory. Both are based on the ability to use nanometer-scale particles as data symbols. These symbols can be moved around in a liquid medium and, theoretically, accessed by a high-density array.
Colloidal memory
The first of these is the so-called colloidal memory. This requires a liquid-filled reservoir in which at least two types of nanoparticles dissolved. These could be ions or molecules or other nanoparticles, according to Maarten Rosmeulen, program director of storage memory at IMEC, writing on the institute’s website.
The reservoir feeds an array of capillaries into which the nanoparticles can be introduced.
Schematic of colloidal memory cross-section. Source: IMEC
The key thing is that the nanoparticles are slightly smaller than the capilliary tubes and so the order of their insertion is preserved thus producing an extended bit sequence. The nanoparticles can be selectively induced (and sensed) by electrodes positioned at the entrance of each capillary tube. A CMOS peripheral circuit controls the array of electrodes.
The density of encoded data can be increased by using more than two types of particle but at the expense of more complex control and reading circuitry.
One of the control methods IMEC researchers are investigating is frequency-dependent dielectrophoresis. This is the force that is exerted on a dielectric particle when it is subjected to a non-uniform electric field. The force can be attractive or repulsive depending on the particle and its frequency dependence providing a method for sorting the particles as they enter the capillary tube. Reading is the reverse process and destructive although it is possible to rewrite the data.
IMEC claims to have demonstrated the selective writing process using micron-dimension electrodes to extract polystyrene nanoparticles from a mixed solution. The next step is to refine the process at the nanometer scale.
Electrolithic memory
The electrolithic process is similar to colloidal memory in that it also uses a fluid reservoir and an array of capillary tubes. In this case two or more metal ions are dissolved in a liquid carrier and writing and reading are achieved by conventional electrodeposition and electrodissolution.
An electrode made of an inert metal such as ruthenium is at the bottom of each capillary. The reservoir is also in contact with a single counter electrode. This forms an electrochemical cell for each capillary. The array of capillary tubes is addressed using conventional CMOS control circuitry and then by applying appropriate voltages a series of ones and noughts can be written in layers of different metals. The two or more metals are selected to have different onset voltages.
The data is detected by measuring the voltage signal as layers of metal are dissolved.
IMEC researchers have fabricated two generations of electrolithic memories. The second has nanowells of 80 to 150nm diameter that are 300nm deep and used copper and cobalt-nickel as the metals.
End of the decade
Rosmeulen said that IMEC thinks that this type of liquid memory could introduced from 2030 onwards when the bit density of 3D-NAND flash will hit limits.
The bit density of liquid memory could reach 1Tbit per square millimeter at a lower cost per square millimeter than 3D-NAND flash. To achieve this IMEC calculates that capillaries will need to be made on a 40nm pitch and achieve aspect ratios of 165:1 for electrolithic memory and 400:1 for colloidal memory. This is similar to the aspect ratios for 3D-NAND flash memory strings and is therefore considered a realistic target.
Work continues at IMEC investigating possible response time, bandwidth (of the order 20Gb/s), endurance (10^3 cycles or better), energy consumption (picojoules to write a bit) and retention (at least 10 years).
Related links and articles:
News articles:
CeRAM paper broadens material scope, highlights role of carbon
ARM spin-off to develop CeRAM memory
MCUs with embedded PCM meet automotive needs
Weebit moves ReRAM on to ‘secret-sauce’ materials
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
