[This is largely a 'kragen-fw' post, but there's enough new material I thought I should just send it to kragen-tol. 3300 words.]
The first digital logic was mechanical; the Differential and Analytical Engines were purely mechanical. The next digital logic machines were electromechanical, built of electromagnetic relays; computing machines, including the first stored-program computers, up through the 1940s worked this way. The next digital logic machines --- the so-called "first generation of computers" --- were built of vacuum tubes. Vacuum tubes were orders of magnitude faster than relays, and they weren't any more unreliable or power-hungry. Still, in large machines, they failed frequently and used a lot of power. The so-called "second generation" of computers were built of transistors, solid-state devices similar to vacuum tubes. They were much more reliable and efficient than vacuum tubes, and also much cheaper. The "third generation" of computers, around 1960, was built of integrated circuits or "chips", groups of transistors printed on a single piece of silicon by photolithography --- far cheaper, and eventually much faster, than separate transistors. There was much speculation about what the "fourth generation" would be like; but gradual improvements in integrated-circuit technology have led to something called "Moore's Law", cutting the width of chip features (such as wires and transistors) by half every three years. These improvements have so far preempted any potential alternative "fourth generation" technologies. Moore's Law seems to have another 21 years left before hitting any fundamental barriers --- we have 130-nm-wide features on chips now, and the atoms involved are perhaps .1-nm wide; researchers at several labs have built one-atom-wide transistors. Still, the cost of building smaller chips is getting larger, approximately doubling every year and a half. So the number of fabrication plants is dwindling, centralizing control over cutting-edge chip manufacturing in the hands of a few large vendors; this is already drastically reducing choice in the CPU and RAM markets, for example. More seriously, the cost of an individual fab run is high and getting higher. On the other hand, gate arrays --- programmable chips --- provide an excellent alternative for small-run circuits, but at much higher cost and significantly lower performance than non-gate-array chips. Integrated circuits are also heavy, fragile, and inflexible, at least compared to materials such as paper or cloth; this limits their integration into our everyday lives. So there are several potential markets for alternative digital logic hardware technologies: small-run circuits for which FPGAs are too expensive; circuits that need to be flexible, light, or sturdy; and circuits for which current digital logic just isn't fast, small, or cheap enough. There was a "First Workshop on Non-Silicon Computing" held this year, covering many of these topics. [24] Inkjet or Laser-Printed Semiconductors -------------------------------------- Starting around 1994, several university and industry labs [13] [1] [10] [11] [12] [14] [48] have printed circuits made of metal, amorphous silicon, polycrystalline silicon, or organic semiconductors on glass or plastic substrates using off-the-shelf or specialized laser and inkjet printers and photolithographed rubber stamps. Plastic Logic, Inc.[0], is a company founded out of the University of Cambridge to commercialize these circuits. They hope to find a market making active-matrix LCD screens. Rod Logic --------- Nanotechnology advocates suspect that mechanical logic will become viable again once we can make it small enough --- atoms are more massive than electrons, so their quantum positional uncertainty is much smaller, so you can do logical things with them at much smaller scales. Sliding-rod logic is one theoretical mechanical logic [9]; rods with knobs on the side slide back and forth, driven by springs. Depending on the position of a rod, its knobs might or might not block the motion of knobs on another rod sliding at right angles to it; various arrangements of these knobs can be used to construct AND, OR, and NOT gates. Nanometer-scale rod logic is not close to becoming practical as an alternative digital logic technology. Buckling-Spring Logic --------------------- Merkle was unhappy with rod logic's imperfections; in particular, rods dissipate energy when their knobs collide, and may also dissipate energy when they are sliding. In response, he invented buckling-spring logic [7], which performs reversible digital logic with arbitrarily little energy dissipation by the oscillations of a single solid object. (There are fundamental physical reasons that irreversible digital logic must dissipate a certain minimum amount of energy per bit erased. [8]) Buckling-spring logic is not close to becoming practical as an alternative digital logic technology. Optical Computing ----------------- For decades, many research groups [5] [2] [15] [16] [17] [18] have been working on optical digital logic devices [6], using optical fibers or free space to carry optical signals from one part of the device to another, sometimes using optical delay lines for storage to reduce the logic hardware needed. A stored-program optoelectronic computer, the "SPOC", was built in 1994 [2] [3] [4] by Harry F. Jordan, Carl E. Love, Dave Sarrazin, Doug Straub, Robert F. Feuerstein, Jay Desai, and John Feehrer, and Vincent P. Heuring, at the Optoelectronic Computing Systems Center at the University of Colorado; it was clocked at 50MHz, had a word size of 16 bits, and a memory of 64 words. Logic in this machine was performed purely electronically, while storage and communication was purely optical. This is a very common combination in current research. Analog computation with optical devices is common --- in particular, lenses compute the two-dimensional Fourier transform of their input light beam, which is useful for some kinds of pattern recognition [19]. There are a variety of devices that can allow light beams to interact in the nonlinear fashion required for general digital computation. LCD panels controlled by photocells are one example; materials with nonlinear refractive properties can be employed to build a "transphasor", which unfortunately only functions under high laser power. Crystalline lithium niobate (LiNbO_3) is the most commonly-used nonlinearly-optical material; it is also photoelastic, piezoelectric, pyroelectric, and ferroelectric. Optical holographic memory is a promising application of photonics, promising much larger (probably WORM) and faster computer memory systems than we currently have; but it is not yet available in the mainstream. Carbon Nanotubes ---------------- Several research groups [20] [21] [22] have been investigating using nanotubes and buckminsterfullerenes as semiconductors and conductors, building transistors and wires from them. It isn't yet clear how to build large-scale circuits from nanotubes. Spherical Integrated Circuits ----------------------------- A Japanese company called "Ball Semiconductor" --- presently at http://www.ballsemi.com/ and previously at http://www.ball.co.jp/ --- was developing methods for lithographing integrated circuits onto the surfaces of one-millimeter-diameter silicon spheres. It appears to have gone out of business around 1998. Fluidics -------- Some digital logic circuits work by flows of fluids, typically air, flowing through shaped channels and cavities; this is called "fluidic logic". Intersecting flows of fluid have nonlinear properties, and fluid flows interacting with chambers can be bistable. Integrated fluidic logic circuits with around 100 components have been built with submillimeter line widths and can run at kilohertz frequencies. [30] [32] [23] [28] [29] [31] I think that fluidic digital logic used to be more popular than it is today, but it is still used in some cases where electronic digital logic is not yet applicable; it is more resistant to ionizing radiation, electromagnetic interference, electrostatic discharge, intense vibration, high temperature, harsh chemicals, and lack of electrical power than semiconductor logic. So it is used in many industrial control applications. Research efforts in microfluidics are currently quite hot --- not for computation, but in order to integrate huge numbers of chemical sensors into integrated circuits. The same systems used in fluidic computers are also in use in pumping applications where reliability and lack of moving parts are paramount, such as in moving radioactive waste. [27] It is plausible that miniaturization of fluidic circuits may make them useful for computation again, as their speed should increase as they get smaller; there also seems to be some amount of excitement that they may be easier to build from living cells than electrical systems would be. Quantum computation ------------------- While essentially all physical computation devices are "quantum computation" devices in the sense that they compute with quantum physics, "quantum computation" usually refers to computing with superimposed quantum states --- promising the possibility of doing many simultaneous possible computations at once. Current practical quantum computers are molecules, with nuclear magnetic resonance coupling the spins of their nuclei; in order to detect the output of the computer, test tubes full of these molecules perform the same computation in parallel. [38] [39] [41] The present state of the art is that a quantum computer consisting of a vial of pentachloroethane has factored the composite number 15 into the primes 3 and 5 [40], so quantum computation is not yet a viable alternative to integrated circuits. It is, however, the only alternative that is inherently more powerful. Josephson junctions ------------------- Josephson junctions [34] are a possible alternative to transistors, consisting of a thin layer of a nonsuperconducting material between two layers of superconductor, which develop AC voltages at hundreds of GHz when the DC current through them exceeds some critical threshold. Several families of digital logic devices based on Josephson junctions have been proposed, of which the fastest is "(Rapid) Single Flux Quantum" (usually abbreviated "SFQ" or "RSFQ") logic, which has been tested successfully at clock rates in the hundreds of GHz. [35] [36] Being superconducting devices, they unfortunately require heroic cooling, outer-space operation, or room-temperature superconductors, and therefore are not presently a viable alternative to semiconductor ICs. However, being superconducting devices, they use three orders of magnitude less power than equivalent semiconductor devices. [37] Molecular Computation --------------------- In 1994, Leonard Adleman coaxed DNA to do graph-theory problems in a test tube; in a test tube, he assembled 10^23 strands of random DNA from DNA fragments representing edges in the graph, and then filtered out the strands that didn't represent the correct solution. In theory, this could be faster than trying the results in sequence on a normal computer; the particular problem Adleman chose, however, was small enough to be solved by visual inspection. [42] [43] [45] [46] Dan Boneh has theoretically extended these methods to more general problems. [44] At present, molecular computation (also known as chemical computing, DNA computing, and biomolecular computing) has not been demonstrated to be competitive with semiconductor computation; there aren't yet many experimental results, and they are still only solving problems that are easy to solve by hand and that are not practically useful. [47] Other ----- Due to lack of time, I haven't surveyed resonant tunnelling devices (RTDs), other kinds of quantum dots, molecular-scale electronics (other than nanotube transistors), atom relays, neurons grown for bioelectronic computers, colliding solitons, or the different kinds of traditional semiconductor logic (RTL, DTL, TTL, NMOS, CMOS). References ---------- Unlike the bibliography of a good research paper, these references are not intended to identify the originators of particular concepts or to survey the field; they are included simply because they provide more information or context for particular statements above. [0] http://www.plasticlogic.com/ [1] http://www.physicstoday.org/pt/vol-54/iss-2/p20.html --- "New Printing Technologies Raise Hopes for Cheap Plastic Electronics", Physics Today, February 2001, describing work by John Rogers at Bell Labs, Richard Friend at the University of Cambridge, George Whitesides at Harvard, and Dago de Leeuw and Bart-Hendrik Huisman at Philips Research Labs in Eindhoven. BE WARNED: this article renders incorrectly if your browser doesn't respect <font face=""> or if you don't have a "Symbol" font installed; specifically, it reads "mm" where it means "micron". [2] http://ece-www.colorado.edu/~harry/spoc/spoc.html [3] http://ece-www.colorado.edu/~harry/spoc/procieee.ps --- Harry F. Jordan, Vincent P. Heuring, and Robert F. Feuerstein, "Optoelectronic time-of-flight design and the demonstration of an all-optical, stored program, digital computer," Proc. IEEE, Special Issue on Optical Computing, Vol. 82, No. 11 (Nov. 1994). [4] http://ece-www.colorado.edu/~harry/spoc/potentials.ps --- Carl E. Love and Harry F. Jordan, "SPOC," IEEE Potentials, Vol. 13, No. 4 (Oct./Nov. 1994). [5] http://web.archive.org/web/20001019085433/http://www-phys.llnl.gov/H_Div/photonics/PhotonicsHomePage.html --- the Lawrence Livermore Photonics home page used to have lots of information about the optoelectronic work they were doing there. [6] http://members.tripod.com/Ranjan_Grover/opticalcomputers.pdf --- "Sources of Light", or "Optical Computers", a 25-page paper on optical computing by Ranjan Grover. [7] http://www.zyvex.com/nanotech/mechano.html --- "Two Types of Mechanical Reversible Logic", 1990, by Ralph Merkle, then of Xerox PARC; Nanotechnology; Volume 4, 1993, pp. 114-131. [8] This is known as Landauer's principle, and is discussed in http://citeseer.nj.nec.com/smith95fundamental.html or http://neci.nj.nec.com/homepages/wds/fundphys.ps --- "Fundamental physical limits on computation", Warren D. Smith, May 1995. Considers limitations on computing performance arising from thermodynamics. [9] See chapter 12 of _Nanosystems_, by Drexler, published by Wiley-Interscience, 1992. Some other parts of the book are on the Web at http://www.foresight.org/Nanosystems/toc.html. [10] "The Promise of Plastic Transistors", a Business 2.0 article by Erick Schonfeld, April 13, 2001. It mentions work by Raj Apte at Xerox PARC, and work by Alan Heeger at a company called Uniax. [11] "IBM Scientists Take Significant Step Toward Production of Flexible Electronics", IBM Research News press release, 2001 August 2; mentions an article published in Nature, Volume 412, on the same day, "Growth Dynamics of Pentacene Thin Films", by Frank-J. Meyer zu Heringdorf, Mark C. Reuter, and Rudolf M. Tromp of IBM's T.J. Watson Research Laboratory. [12] "Flexible Electronics: On the Brink of a Tech Revolution", a Semiconductor Magazine (Vol. 2, No. 11, 2001 November) article by Jeff Dorsch. Mentions IBM's program, Alien Technology Inc., E Ink Corp., FlexICs Inc., Rolltronics Corp., and Iowa Thin Film Technologies, the manufacturer of thin-film amorphous silicon solar cells. Partly plagiarized from press release [11]. E Ink is famous for its "electronic paper" display technology, but its VP of R&D says they want to be first to market with a high-volume flexible-transistor production process. [13] "'Giant-electronic' effort tries OLED, TFT Mix", an EE Times article by Gail Robinson from 1997. Describes work done by Sigurd Wagner and others at Princeton's Center for Photonic and Optoelectronic Materials, supported by DARPA and the New Jersey Advanced Technology Center on Photonics and Opto-Electronic Materials, and presented at a Materials Research Society meeting; they coated steel foil with an insulating glass film and printed organic LEDs onto the glass film. In a demonstration project, an unmodified laser printer with normal toner was used to print circuit traces; the toner worked as "photoresist". [14] http://www.ee.princeton.edu/~asg/ --- originally "Amorphous Silicon Group", now "Macroelectronics Group". "Macroelectronics are integrated circuits bigger than semiconductor wafers." Work is being done by Sigurd Wagner (mentioned in [13]), Helena Gleskova, Ming Wu, Bob Min, I-Chun Cheng, Bill Jordan, Rabin Bhattacharya, Eitan Bonderover, Samir Succar, Joe Valentino, Becca Jones, Todd Johnson, Joe Barillari, Steve Saar, Amy Buerkle, and Darci Taylor, and collaboration with the following people is mentioned: James Sturm; Stephen Fonash of Penn State; Zhigang Suo; Sandra M. Troian; Tony Evans; Bruce Gnade of the University of North Texas; Yu Chen of E Ink; Stephen Y. Chou; Tanja Cuk of Stanford; Anton Darhuber; Anthony Evans; Alex Gelbman of VisibleTechKnowledgy; Pai-Hui Iris Hsu; Rolf Koenenkamp of the Hahn-Meitner-Institut; Paul Kydd of Parelec Corp.; Eugene Ma and Adam Payne of Aegis Semiconductor; Scott Miller; Marcelo Mulato of the Universidade de Sao Paulo; Ruud Schropp of the Universiteit te Utrecht; Wole Soboyejo; and Valerie M. Thomas. [15] http://dmoz.org/Science/Physics/Optics/Optical_Computing/ --- Open Directory Project category for optical computing, listing 1 [16] http://www.csm.ornl.gov/laser1.html --- the CESAR Optical Computing and Quantum Communications Lab at Oak Ridge National Labs. Mostly working on quantum computation now. [17] http://optlab2.bk.tsukuba.ac.jp/ocg/ --- the Optical Computing Research Group of the Optical Society of Japan. Mostly in Japanese. [18] http://sipi.usc.edu/info/OpticalComp.html --- SIPI Optical Computing Laboratories. Mostly working on optoelectronics. [19] http://www.spie.org/web/oer/february/feb98/ltconstr.html --- "Light Constructions: Optical Correlators Escape From the Lab", by Sunny Bains, OE Reports 170, February 1998. Describes Fourier pattern recognition by "joint transform correlators" and "Vander Lugt correlators" using lenses. [20] http://www.aip.org/enews/physnews/1998/split/pnu371-3.htm --- "A Carbon Nanotube Transistor", by Phillip F. Schewe and Ben Stein, American Institute of Physics's Physics News Update, Number 371, story #3, May 13, 1998. Describes work done by Cees Dekker at the Delft Institute of Technology and published in "S.J. Tans et al., Nature, 7 May 1998".) [21] http://www.ibm.com/news/2001/04/27.phtml --- "IBM Scientists Develop Carbon Nanotube Transistor Technology", 2001 April 27 press release from IBM describing research done at IBM's Nanoscale Science Research Department, describing a report entitled "Engineering Carbon Nanotubes and Nanotube Circuits Using Electrical Breakdown" by Philip G. Collins and Michael S. Arnold, published in Science, vol. 292, issue 5517, April 27, 2001. [22] http://qt.tn.tudelft.nl/~hadley/tube_logic/tubelogic.pdf --- preprint of "Logic Circuits With Carbon Nanotube Transistors", by Adrian Bachtold, Peter Hadley, Takeshi Nakanishi, and Cees Dekker, Department of Applied Physics and DIMES, Delft University of Technology, 2001 October 4. Describes simple RTL logic circuits the researchers built from nanotube FETs. [23] http://sas.org/E-Bulletin/labNotes/2001-08-17/2001-08-17LabNotes.html --- "A Fluidic Flip-Flop", by C.L. Stong and Shawn Carlson, an article claimed to be from the May 1966 (not 1996, 1966) issue of "The Amateur Scientist" column of Scientific American. [30] claims this article to be by James Sharpsteen and on pages 128-135. [25] http://www.crhc.uiuc.edu/nsc/ --- "1st Workshop on Non-Silicon Computation (NSC-1)", 2002 February 3. [26] http://www-2.cs.cmu.edu/~phoenix/nsc1/ --- "NSC-1 Final Program" --- program for [25]. [27] http://www.em.doe.gov/tie/spr9713.html --- "Power Fluidic Devices Help Solve Operational Problems at ORNL and SRS", describing the use of fluidic control systems to move radioactive waste without moving parts. [28] See the "Proceedings of the IFAC Symposium on Fluidics", which apparently ended in 1971. [29] See "Fluidics Quarterly", which apparently renamed itself to "Journal of Fluid Control" in 1973. [30] http://www.neci.nec.com/homepages/wds/navstokes.ps --- "navstokes.dvi", "On the Uncomputability of Hydrodynamics", or "Hydrodynamics is Unsimulable" by Warren D. Smith of NEC, section 7: "Nikola Tesla, 'Fluidics', and hydraulic logic components," pp. 7-9; see also footnotes 1-4, 9, 11, 16, 19, 28, 29, 31, 45, 55. This was my best source on fluidics; although the title makes the guy sound like a raving loon, he explains on page 17 that his point is not that you could actually solve uncomputable problems with hydrodynamic systems, but that the standard hydrodynamic equations become pathological in surprising conditions. [31] http://www.air-logic.com/Summary%20Cat/sum14and15.html --- Air Logic is a vendor of pneumatic and vacuum control equipment, including the fluidic logic NOR components on this catalog page. [32] http://www.iitb.ernet.in/~insight/issues/new/vol4iss2/fluidics.htm --- "Drop Those Wires", an article by Amol S. Gogate in "Insight" magazine (volume 4, issue 2, 2001 September 10) on fluidics. [33] http://www.ess.nthu.edu.tw/~fangang/ess5841/lect/doc/ess5841-Lec4-1.doc --- "Lecture 4-1, Microfluidic Devices". [34] http://www.sciam.com/askexpert/physics/physics37/physics37.html --- "What are Josephson junctions? How do they work?", a 1997 November 24 Scientific American "Ask the Experts" article on Josephson junctions. [35] http://www.ifp.fzk.de/hotline/statusreport/sr98/dcdigital.html --- "Applied Superconductivity Status Report '98", section 3 "Digital Circuits", subsection C "Digital Logic". Describes 1998 Josephson junction technology. [36] http://www.physics.georgetown.edu/~jkf/northwestern/tsld006.htm --- "Digital Electronics and RSFQ logic", a slide in a presentation by J.K. Freericks of Georgetown University entitled "Ultrafast Digital Electronics: Optimizing the Speed of a Josephson Junction". [37] http://www.cs.sunysb.edu/~ors/Fall-1998/lw --- abstract from "Design of a Petaflops Network Computer", by Larry Wittie. [38] http://www.qubit.org/ --- the Centre for Quantum Computation of the University of Oxford. [39] http://www.sciam.com/1998/0698issue/0698gershenfeld.html --- "Quantum Computing with Molecules", a Scientific American article by Neil Gershenfeld and Isaac L. Chuang. [40] http://www.research.ibm.com/resources/news/20011219_quantum.shtml --- "IBM's Test-Tube Quantum Computer Makes History", an IBM press release from 2001 December 19, announcing the use of Shor's algorithm to factor the number 15. [41] http://dmoz.org/Computers/Computer_Science/Theoretical/Quantum_Computing/ --- the Open Directory Project's category for Quantum Computing. [42] ftp://ftp.krl.caltech.edu/pub/users/brown/adleman.ps.gz or http://citeseer.nj.nec.com/adleman94molecular.html --- "Molecular Computation of Solutions to Combinatorial Problems", by Len Adleman, published in Science, Volume 266, 1994 November 11, pp. 1021-1023; this is the paper in which Adleman announced the invention of DNA computation. [43] http://www.mitre.org/research/nanotech/hapgood_on_dna.html --- "Explanation of Molecular Computing with DNA", by Fred Hapgood, April 1995; describes Adleman's experiment. [44] http://crypto.stanford.edu/~dabao/biocircuit.ps.gz --- "On the Computational Power of DNA", apparently written in 1995, by Dan Boneh, Richard Lipton, Chris Dunworth, and Jiri Sgall. Extends DNA computation to SAT and some optimization problems. [45] http://www.liacs.nl/~pier/dna.html --- "A Bibliography of Molecular Computation and Splicing Systems", a half-megabyte bibliography on molecular computation, presently maintained by Pierluigi Frisco. [46] http://hagi.is.s.u-tokyo.ac.jp/dna8/ --- "Eighth International Meeting on DNA Based Computers". [47] http://www.house.gov/science/landweber_091200.htm --- "Beyond Silicon Computing: DNA Computers", Congressional testimony by Laura Landweber, 2000 September 12. [48] http://www.eet.com/at/news/OEG20020410S0013 --- "Scientist makes photonic circuits with inkjet printer", by R. Colin Johnson, an article for the EE Times (2002 April 10 9:37 AM EST). Describes work done by Ghassan Jabbour and Yuka Yoshioka at the University of Arizona printing photonic circuits (solar cells and computer displays) onto flexible surfaces with standard inkjet printers. -- <[EMAIL PROTECTED]> Kragen Sitaker <http://www.pobox.com/~kragen/> Irony and sarcasm deflate seriousness, and when your seriousness becomes detum- escent, you're not held responsible for your thoughts. Irony beats thinking like rock beats scissors. -- http://www.hyperorg.com/backissues/joho-june2-98.html
