Millipede memory
| Company type | Technology / Storage device |
|---|---|
| Industry | Computer storage |
| Founder | IBM |
| Products | Non-volatile computer storage devices |
Millipede memory represents a form of non-volatile computer storage designed to achieve data densities surpassing 1 terabit per square inch (equivalent to 1 gigabit per square millimeter), approaching the physical limits found in perpendicular recording hard disk technologies. Created by IBM, this technology was envisioned as a successor to magnetic hard drives, offering a drastically smaller footprint similar to that of flash storage.
IBM demonstrated a working prototype at CeBIT 2005, with plans for market introduction by 2007. Nevertheless, progress in alternative storage solutions hindered its commercial deployment.
Fundamental principle
Contemporary primary memory mainly relies on DRAM, which holds information as electrical charges within capacitors. This type of memory is volatile, losing stored data when power is interrupted. Conversely, hard disks utilize magnetic surfaces for persistent data retention but depend on mechanical components for reading and writing, which restricts speed despite offering higher capacity.
The Millipede approach merges these concepts by encoding bits as nanoscale impressions on a thermally sensitive polymer layer applied to a movable platform called the sled. Reading and writing are performed through arrays of numerous nanometer-scale atomic force probes, allowing parallel processing and throughput superior to that of conventional hard drives.
Each probe alters or detects bits by forming or erasing pits in the polymer via a heated tip. The sled is precisely maneuvered beneath the probes using electromechanical actuators following an x/y scanning pattern. Although each probe accesses a limited area, thousands operate concurrently—early models incorporated 1,024 probes arranged in a 32x32 matrix, which inspired the "millipede" designation. These cantilever arrays are entirely fabricated from silicon through surface micromachining techniques.
Data bits correspond to indentations generated when the heated probe tip softens the polymer above its glass transition temperature (approximately 120 °C for PMMA), enabling pits with lateral dimensions near 3 nm. Reheating an indentation melts and removes it, facilitating rewriting. The theoretical maximum storage density is around 0.9 Tb/in².
Data access mechanisms
Each cantilever probe manages data storage and retrieval using thermo-mechanical means, addressing one bit sequentially. For reading, the tip is heated to roughly 300 °C and brought close to the sled surface. When positioned over a pit, the cantilever deflects into it, enlarging the contact area and accelerating heat dissipation into the sled. In contrast, absence of a pit results in minimal contact and slower heat loss. Because the probe’s electrical resistance depends on temperature, cooling caused by contact over a pit reduces resistance, signaling a "1" bit; otherwise, it indicates "0". During reading operations, the tip scans continuously while monitoring resistance fluctuations.
Writing involves raising the probe tip temperature beyond the polymer’s glass transition temperature—about 400 °C for acrylic glass sleds. To record a "1," the softened polymer is indented by the tip; erasing to "0" occurs by lifting the tip and allowing surface tension to smooth out the indentation. Earlier versions utilized slower erasure processes with endurance near 100,000 cycles; subsequent enhancements have improved this aspect.
Power consumption for heating probes varies with data rates but remains significant. At speeds of several megabits per second, Millipede’s power draw is approximately 100 milliwatts—comparable to flash memory and lower than traditional hard drives. Its architecture supports extreme parallelism, enabling aggregate transfer rates in the GB/second range. While individual probes operate at kilobits per second rates—totaling several megabits per array—IBM’s experiments at their Almaden Research Center showed single tips achieving 1–2 megabits per second, indicating potential for gigabyte-per-second aggregate throughput.
Applications
Millipede memory was created as a form of non-volatile data storage designed to rival flash memory in terms of capacity, speed, and physical dimensions. Despite its initial promise, it has been overtaken by newer technologies and is no longer under active development.
Initial devices
The earliest millipede prototypes utilized probes measuring approximately 10 nm in diameter and 70 nm in length, producing indentations roughly 40 nm wide within 92 μm by 92 μm areas. A chip sized 3 mm by 3 mm incorporated a 32 by 32 probe matrix, yielding a storage capacity of 500 megabits (62.5 MB) and an areal density near 200 Gbit/in². IBM first demonstrated this concept in 2003, with intentions to launch commercially by 2005; however, hard drive densities were already approaching 150 Gbit/in² at that time and have continued to advance beyond that figure.
Planned commercial release
During the 2005 CeBIT Expo, IBM presented enhanced versions featuring 64 by 64 cantilever arrays on a 7 mm by 7 mm data platform, boosting storage density to approximately 800 Gbit/in² through reduced pit dimensions. Theoretical scaling suggested pit sizes could shrink to about 10 nm, potentially allowing densities slightly above 1 Tbit/in². IBM aimed to bring products with this level of density to market by 2007. In comparison, laptop hard drives had reached densities of 636 Gbit/in² by late 2011, while emerging methods such as heat-assisted magnetic recording and patterned media were predicted to achieve up to 10 Tbit/in². Flash memory densities had approached nearly 250 Gbit/in² by early 2010.
Present status
By 2015, no commercial millipede memory devices had been launched, largely due to swift progress in alternative storage technologies.
See also
- Nanoelectromechanical systems
- Nanotechnology
- Nanolithography
- Thermal scanning probe lithography
- Punched card