memories of future

8/4/2019 Memories of Future

1/10

Memories of the Future, Multimedia Systems Conference 2002.

Last edited: 05:28 PM 20/12/2001

1

Memories of the FutureEmerging replacements for semiconductor

memory, optical and magnetic disks

Paper by Martin J. Wickett

[email protected]

Multimedia Systems Coursework, Department of Electronics and Computer Science,University of Southampton, Southampton S017 1BJ, UK

ABSTRACT

Laws of physics will eventually force the computer industry into finding new storage solutions. In

the near future thresholds will be reached when semiconductor memory, magnetic disks and optical

disks cannot be made smaller or faster because it becomes physically impossible to pack data so

densely. This paper investigates the current technologies that could be candidates for their

replacements. I will look at volumetric solutions based on holographic and biological storage and

other contenders such as probe storage (storage at the atomic level) and non-volatile magnetic

RAM.

KEYWORDS

Memory, volumetric storage, biological storage, holographic memory, holostore, protein memory,

molecular memory, probe storage, atomic memory, non-volatile RAM

1. INTODUCTION

This paper is intended to give its reader an overview of the emerging storage technologies that are

currently in research and are likely to replace conventional memory and storage systems. This paperwill discuss why new approaches are needed; the main areas researchers are looking at, the

obstacles faced by each approach and the predicted time we will have to wait for each technology to

be viable for everyday use.

This paper looks in detail at four different technologies. Two which are called volumetric, because

they use an extra dimension to increase the storage capacity: holographic memory which stores

data on a hologram three dimensionally by reading and writing to a hologram at different angles and

protein memory which uses lasers and inert transparent gel with bacteria extracts to store

information. The other two are probe storage which stores data punch card style at the atomic

level and non-volatile magnetic RAM which is non-volatile because bits are stored magnetically

instead of as a charge like conventional RAM, hence bits can be more densely packed.


2/10


Last edited: 05:28 PM 20/12/2001

2

2. BACKGROUND

Over the years many different methods have been used to store information, from punch cards to

optical disks with gigabytes of space. This increase in speed and capacity of computer memory and

storage we have seen over the last decade is mainly due to the miniaturisation of components and

the density in which we can store data. It is obvious this reliance on miniaturisation to increase

performance must end somewhere.

As we try to pack more and more data into such small places, it is very difficult to get great

precision when we still use mechanical systems, which rely on moving parts. Also we are faced

with other problems when we deal with physics on very small scales. For example, when we try and

store bits on magnetic material at very high density, there becomes a point when the thermal

movement of electrons will cause the direction of a bits magnetisation to flip at random. In the case

of optical disks, the wavelength of the light used limits the distance between the bits.

It is estimated that in the next five or ten years we will reach the limiting density for storing data on

magnetic disks [PC PLUS 183]. There is currently much research into other methods of memory

and storage.

3. VOLUMETRIC SOLUTIONS

3.1 HOLOGRAPHIC MEMORY

Holographic memory (sometimes called holostore) is based on the same principles of the

photographic holograms that we all have seen. Holographic memory is very promising because it

not only offers greater capacity, but the access speeds are very fast because there are few moving

parts and no contact is required.

Photographic holograms are made by recording interference patterns of a reference beam of lightand a signal beam of light reflected off an object. Photosensitive material holds this interference

pattern, and the image can be reproduced by applying an identical beam of light to the reference

beam onto the photosensitive material. Many variations of the object can be recorded on a single

plate of material by changing the angle or the wavelength of the incident light. This is how it is

possible to produce animations on a hologram. Each frame of an animation is stored by varying the

angle of the incident light.

Holographic memory works exactly the same way except instead of reflecting the light off an

object, a small LCD display is used which shows a page of data. By varying the angle of the

incident light, it is possible to store many pages of data on one plate.

Figure 3.1.0: Diagrams from Holographic storage: are we there yet? by Glenn T. Sincerbox


3/10


Last edited: 05:28 PM 20/12/2001

3

The University of Oregon has conducted experiments with a frequency-selective crystal of

Tm3+

:YAG as the recording material. They successfully encoded a sequence of 1760 bits onto an

input beam, stored the sequence and then retrieved it. Their results show a density of about 8 Gbit/

in2

and a density-bandwidth product of 1.5x1017

bits/in2-sec [IEEE OCT 1995]. The limits of

spectral holography are far higher than this though, and the slowness of the experiments is due to

extra equipment that would not be needed if the technology were to be developed further. It is

possible that this crystal of Tm3+

:YAG could store densities up to 100 Gbit/in2

and bit rates up to 1

Gbit/sec. Other materials are thought to have potential for higher densities and bandwidths [IEEEOCT 1995].

Current challenges that need to be resolved before the technology is viable:

Currently it is difficult to read the data without causing some damage to the recorded data.

Also the same problem causes writes to degrade previous writes in the same region

[SINCERBOX].

It is difficult to grow large crystals of good quality [SINCERBOX].

The advantages of using holographic memory are:

Parallel access to different pages

Very fast for database searches and data mining functions

because data can be compared optically without having to retrieve it

(ideal for fast database searches, e.g.: fingerprint matching or photo recognition)

No moving parts, hence fast access speeds

Non-volatile, can be used for storage and memory

Storage Medium Access Time Data Transfer Rate Storage Capacity

Holographic Memory 2.4 s 10 GB/s 400 Mbits/cm2

Main Memory (RAM) 10 40 ns 5 MB/s 4.0 Mbits/cm2

Magnetic Disk 8.3 ms 5 20 MB/s 100 Mbits/cm2

Table 3.1.0: Comparison of Holographic memory, RAM and magnetic disks. Table from [BOYLES].

There are two main groups working on holographic memory technologies: the Holographic Data

Storage System (HDSS) consortium and the PhotoRefractive Information Storage Materials

(PRISM) consortium. Both consortiums combine companies (IBM is part of HDSS) and academic

researchers from institutions such as the California Institute of Technology, Stanford University, the

University of Arizona and Carnegie Mellon University.


4/10


Last edited: 05:28 PM 20/12/2001

4

3.2 PROTEIN MEMORY

Protein memory is based on bacteriorhodopsin that is extracted from bacteria. Bacteriorhodopsin is

an organic molecule that can exist in a variety of chemical states. It is relatively easy to detect

which state the molecules are in, because each state has different absorptions to light. By choosing

two of these states, one for binary zero and the other as binary one, it is possible to use this as a

memory device.

Bacteriorhodopsin is combined with inert transparent gel and stored in a cube. Two lasers are

positioned next to the cube, one looking vertically through the cube (red laser), and the other

looking horizontally down (green laser). Each laser has an LCD display between the laser and the

cube.

The green laser (paging LCD) illuminates a vertical slice of matter called page memory; the red

laser (write laser) illuminates the pattern displayed on the LCD (which is a binary representation of

the data) onto the matter on the cube (figure 3.2.0). The matter that is illuminated by the green laser

and also hit by the red laser shifts state. It requires both lasers to shift state, so the rest of the matter

that is illuminated by the green laser or the red laser only is not affected. The pattern that was

displayed on the LCD in front of the red laser (figure 3.2.1) has thus been transferred onto the

illuminated page of memory (figure 3.2.2).

Figure 3.2.1

Figure 3.2.0

Figure 3.2.2


5/10


Last edited: 05:28 PM 20/12/2001

5

On the opposite side of the cube, in front of the red laser there is a CCD (charge-coupled device)

detector that is used to read the data from the memory. The molecules do not have different

absorptions until the green laser illuminates them. When the green laser illuminates a page of

memory, the CCD detector can measure the absorption levels of the matter illuminated by the green

laser, and from that reconstruct the data pattern from that page of memory.

There is also a blue laser that erases a page of memory.

A prototype has been shown to work in 1997 that used a cube of 1x1x2 inches and was able to store100Mb. The developers of this prototype estimate that a two cubic inch device could eventually

store 125Gb [PC PLUS 183].

Issues are left to be resolved:

The polymer gel that the protein is put in breaks down faster than the protein itself. The

protein can withstand the laser light, but the gel breaks down after a while. This is a major

obstacle for protein memory [BROWN, CHUN & IKRAMULLA].

Mutations could affect the photochemical properties of the protein

[BROWN, CHUN & IKRAMULLA].

Advantages of using protein memory:

Because it is protein based it is inexpensive to produce in quantity

Can operate over a wider range of temperatures much larger than semiconductor memory

Parallel Access to different pages

Non-volatile, can be used for storage and memory

4 OTHER CONTENDERS

4.1 PROBE STORAGE

Probe storage is based on manipulations on an atomic scale. It is basically a punch card system, but

at the atomic level, where the properties of atoms at a particular point represents a binary zero or

binary one.

This system has very high potential for storage where capacity is the most important factor because

it uses the smallest area possible to store bits.

Figure 4.1.0 Electron beams writing data by heating atomic size cells

(Image from [SCIEN. AMER. 2000])


6/10


Last edited: 05:28 PM 20/12/2001

6

Researchers use advanced devices like the scanning tunnelling microscope (STM), the field

emission probe (FEB) and the atomic force microscope (AFM) to move around at the atomic level

and sense groups of atoms. These groups of atoms are usually changed from one state to another

(e.g.: from amorphous to crystalline) by applying heating spots of atoms. The probe tip emits a

beam of electrons onto a specific area when voltage is applied. This beam writes or erases a bit. A

weaker beam is used to read the data by detecting a phase dependant electrical property (e.g.: its

resistance).

Issues to be resolved:

States have to be stable at room temperature and across a specific range.

The material has to have 2 distinct states

The device has to be integrated with other electronics

The working elements have to be in a vacuum or controlled atmosphere to reduce scattering

of electrons and reduce heat flow between spots of data.

The current technology is very slow (slower than todays hard disks)

Advantages:

No power consumption when not performing operations (ideal for portable devices)

Very high capacity for very little space

4.2 NON-VOLATILE MAGNETIC RAM

This type of memory is also based on an old technology, magnetic core memory that was used in

early computers. Tiny blocks of magnetic material are magnetised north or south for binary zero or

one. It has also been on the cards for a long time and is only now becoming a viable technology.

IBM have been carrying out research into MRAM since the 1970s. Other large companies (Intel,HP and Siemens to name but a few) have taken interest in this technology because it is likely to

become the replacement for DRAM [BONSOR]. Motorola are researching this technology, hoping

to use it in portable devices [EBN].

In normal DRAM memory that we use today, bits are stored as a charge on a capacitor. This

memory has to be constantly energised to retain a charge. This system is volatile because the charge

is lost when there is no power and the data they contained is lost.

Magnetic RAM (MRAM) stores bits on small blocks of magnetic material by charging thempositively or negatively. This system is non-volatile because the magnetic charge (hence the data)

remains when the power is turned off.

These blocks are organised as a grid and each block on the intersection of two lines, a bit line and a

word line. By selecting a word line and a bit line, it is possible to reference a single block. Current

is passed through the word line and the bit line. The direction of the current along the bit line

determines the polarity that the block will switch to (figure 4.2.0).


7/10


Last edited: 05:28 PM 20/12/2001

7

Figure 4.2.0

Advantages of MRAM:

Non-volatile, ideal replacement for flash memory Very fast (estimated at 20 times as fast as DRAM [PC MAGAZINE])

Low power consumption and no power required when device is turned off to keep the data

Similar capacity per area than DRAM

IBM and Infineon are now working to bring this technology into the IT market [IBM THINK]. IBM

have been heavily involved in MRAM research over the past four years and have demonstrated

working prototypes. Companies like IBM, Infineon and Motorola are currently hoping to introduce

MRAM into their devices in the next couple of years.

5. PREDICTED TIME SCALE OF EMERGENCE

Magnetic RAM has now been proven to be technically viable and it is expected we should see it in

the next couple of years. Protein memory it the technology the next in line, but due to its

uniqueness, it remains to be seen if it would be a practicable solution and if the industry would

adopt it. Holographic memory has been stuck in the labs for many years, and people have started to

doubt weather it will ever be viable (even the American government have given up their funding

[EETIMES]). Probe storage is the furthest away, but the hurdles are mainly related to the way inwhich the technology will be adapted for commercial use, rather than uncertainties with the

technology itself.


8/10


Last edited: 05:28 PM 20/12/2001

8

6. CONCLUSIONS

Holographic memory is an old concept that people have been trying to get working since the early

70s, and as let, despite large investment into research by companies such as IBM, has not been

proven to be commercially feasible [EETIMES]. There are many hurdles left for this technology to

overcome. Protein memory seems a little more promising becauseit would be relatively cheap toproduce. It is interesting to note that because of the innovative nature of volumetric storage, which

have come from discoveries in physics, chemistry and biology, we may see new applications formemory appear that could not be done with two-dimensional memory. Probe storage shows

potential for handheld devices where access time is less of an issue than large capacity and (much

sought-after) low power consumption. Magnetic RAM is the most likely technology we should

expect to see appear first, as a replacement for DRAM. All these solutions are non-volatile so it

may be that the separation between memory and storage disappears in future computers and

electronic devices.

7. REFERENCES

General

Future of Computing (Optical & Biological Possibilities)

http://www.doc.ic.ac.uk/~nd/surprise_97/journal/vol1/ary/(accessed 16/11/2001)

[SCIEN. AMER. 2000] Avoiding a Data Crunch, Scientific American Feature Article, May 2000

Jon William http://www.sciam.com/2000/0500issue/0500toig.html(accessed 16/11/2001)

[PC PLUS 183] FrontDesk Article, PC Plus magazine, David Bedford, November 2001 Issue 183

Holographic Storage

[SINCERBOX] Holographic storage: are we there yet? Glenn T. Sincerbox (Professor of Optical

Science, Director of the Optical Data Storage Center)

http://w3.opt-sci.arizona.edu/Glenn/holograp1.htm(accessed 16/11/2001).

[IEEE OCT 1995] Spectral Holographic Memory at 8 Gbit/in2,

Hai Lin, Tsaipei Wang, and

Thomas W. Mossberg, Department of Physics, University of Oregon, IEEE Lasers and Electro-

Optics Society, Newsletter, Vol. 9, p.10 (Oct. 1995). (Accessed 16/11/2001).

http://opticb.uoregon.edu/~mosswww/memory/shm.html (accessed 16/11/2001)

Holographic data storage, IBM Research, J. Ashley, M.-P. Bernal, G. W. Burr, H. Coufal, H.Guenther, J. A. Hoffnagle, C. M. Jefferson, B. Marcus, R. M. Macfarlane, R. M. Shelby, and G. T.

Sincerbox, publication November 18, 1999

http://www.research.ibm.com/journal/rd/443/ashley.html (accessed 16/11/2001)

Holographic Memories Scientific American, Demetri Psaltis and Fai Mok, November 1995 VOL.

273, NO. 5 PP. 70-76

http://optics.caltech.edu/publications/SciAm-Nov1995/article.html(accessed 16/11/2001)

[BOYLES] Holographic Memory, Stephanie Boyles, CSI 3300, April 12, 2000

http://ucsu.colorado.edu/~stephanb/projects/CSI3300.htm (accessed 16/11/2001)

Holographic Storage Overview, IBM Almaden Research Center projects

http://www.almaden.ibm.com/st/projects/holography/(accessed 16/11/2001)


9/10


Last edited: 05:28 PM 20/12/2001

9

Say hello to hologram RAM, The Register, Tony Smith Posted: 08/02/2000

http://www.theregister.co.uk/content/archive/9128.html(accessed 16/11/2001)

[EETIMES] Final exams loom for holographic memory, eetimes.com, Margaret Quan, 31/08/1999

http://www.eetimes.com/story/OEG19990831S0001(accessed 17/12/2001)

Memory of the future: two directions, digit life, Maksim Lenhttp://www.digit-life.com/articles/memorytwodirections/(accessed 17/12/2001)

Holographic Memory, John Sand, University of Minnesota Morris

Biological Storage

[BROWN, CHUN & IKRAMULLA] Computer Memory Based on the Protein Bacteriorhodopsin

Utilizing the Two-Photon Method for Read/Write Procedures, Gregory Brown, Patrick Chun and

Faiz Ikramulla http://www.cem.msu.edu/~cem181h/projects/96/memory/(accessed 16/11/2001)

Probe Storage

Star Group Projects, University of California Santa-Cruz

http://www.cse.ucsc.edu/~tara/stargroup/Projects/Probe-Based_Storage/probe-based_storage.html

(accessed 16/11/2001)

High-density data storage using proximal probe techniques, by H. J. Mamin, B. D. Terris, L. S. Fan,

S. Hoen, R. C. Barrett, and D. Rugar, IBM Research, IBM Journal of Research and Developpement,

Volume 39, Number 6, 1995

Proximal probe microscopies , http://www.research.ibm.com/journal/rd/396/mamin.html

(accessed 16/11/2001)

Modeling Probe-Based Data Storage Devices., Katherine Pu Yang, University of California

Santa-Cruz, April 2000.

Physical Modeling of Probe-Based Storage, Madhyastha, T. M. and Yang, K. P, In Proceedings of

the Eighteenth IEEE Symposium on Mass Storage Systems (April 2001)

Workload Based Optimization of Probe-Based Storage, Sivan-Zimet, Miriam, Technical Report

01-06, University of California Santa-Cruz, July 2001.

Non-volatile Magnetic RAM

Non-Volatile Solid-State Memory using the Magnetic Spin-Dependent-Tunnelling Effect

Desmond J. Mapps, Frank (Z.) Wang & Lianna HE, University of Plymouth, UK

Infineon, IBM hone magnetic RAM strategy, Peter Clarke, EE Times

http://www.siliconstrategies.com/story/OEG20010417S0052 (accessed 16/11/2001)

[BONSOR] How Magnetic RAM Will Work, Kevin Bonsor, how stuff works

http://www.howstuffworks.com/mram2.htm and http://www.howstuffworks.com/mram1.htm

(accessed 17/12/2001)[IBM THINK] Computing Unplugged, IBM Think Research


10/10


Last edited: 05:28 PM 20/12/2001

10

http://www.research.ibm.com/thinkresearch/pages/2001/20010202_mram.shtml (accessed

17/12/2001)

[EBN] Motorola moves MRAM memory technology closer to market, EBNhttp://www.ebnews.com/ecomponents/semiconews/story/OEG20000510S0039

(accessed 17/12/2001)

Magnetic Random Access Memory (MRAM)http://www.imec.be/mcp/nmc/magneto/mram/mram.htm(accessed 17/12/2001)

[PC MAGAZINE]The Possibility of Commercial MRAM, PC Magazine,John C. Dvorak ,

February 6, 2001

http://www.pcmag.com/article/0,2997,s=1501&a=4540,00.asp (accessed 17/12/2001)

A Study of Magneto resistance Random-Access Memory, C.L. Lee,www.andrew.cmu.edu/~zlee/mram.pdf(accessed 17/12/2001)

Magneto resistive Random Access Memory (MRAM), James Daughton, 02/04/2000

8. FURTHER READING

Brief Articles and Information

Press Release 2000/09/21 Development of High-performance Material for Holographic Memory

and Small Recording / Playback System

http://www.mext.go.jp/english/news/2000/09/000958.htm(accessed 16/11/2001)

Glass Act, Toby Howard, University of Manchester

http://www.cs.man.ac.uk/aig/staff/toby/writing/PCW/holo.htm(accessed 16/11/2001)

Terabytes, shrink-wrapped : Is Organic Mass Memory Ready for Series Production? c't 3/98, page

18 http://www.heise.de/ct/english/98/03/018/(accessed 16/11/2001)

Making Room for Digital Data, Jamie Beckett 4 November 1999 HP Labs

http://www.hpl.hp.com/news/storage.html (accessed 16/11/2001)

High-Protein Computers, PC Magazine online.

http://www.zdnet.com/pcmag/issues/1410/pcm00008.htm(accessed 16/11/2001)

Instant Access Memory, Wired, David Voss, April 2000http://www.wired.com/wired/archive/8.04/mram.html

Hardware News: Faster, driver, there's a deluge of data coming, Jeremy Torr

http://www.it.mycareer.com.au/hardware/20000620/A16808-2000Jun19.html

(accessed 16/11/2001)

Flash to the Future, David Essex, Technology Review

http://www.techreview.com/web/essex/essex072601.asp (accessed 16/11/2001)

IBM, Infineon to commercialize MRAM in 2004 - Interview with Infineon Executive

VP,08/12/2000http://ne.nikkeibp.co.jp/english/2000/12/1207ibm_devst.html(accessed 17/12/2001)

memories of future

Documents