[math-fun] New (?) idea for no-semiconductor "transistor": MFIM
New (?) idea for no-semiconductor transistor: MFIM =======Warren D. Smith===16 October 2013============ Here's a design for a "transistor" which will switch current using 100% solid state, no moving parts, cold device, and with no semiconductors. It initially looks to me like for most purposes it will not be as desirable as more conventional transistors, but perhaps there are some purposes for which it would be superior. xxxxxxxxxxxxxxxxxxxx ++++++++=====+++++++ ++++++++=====+++++++ ++++A+++==B==+++C+++ ++++++++=====+++++++ ++++++++=====+++++++ yyyyyyyyyyyyyyyyyyyy A and C are metal conductors forming the two endpoints of the "switch." B is a thin soft-ferromagnetic insulator, such as Fe2O3 or Gd3Fe5O12 or Y3Fe5O12. Acronym "MFIM"=Metal-Ferromagnetic Insulator-Metal. The idea of this transistor is simple. It is quantum "tunneling" by electrons through a potential barrier. The conduction band in B is unoccupied and at a high potential energy (that is what being an "insulator" means), thus forming a barrier. Because B is thin, tunneling through the barrier is possible, and the tunneling rate will be tremendously modifiable by altering the barrier height. [WKB approximation: Tunneling rate depends on exponential function of W*squareroot(H) where W=width and H=height of barrier.] To switch the transistor, we apply a magnetic field between x and y. (Applying magnetic field along the A-B direction could be another, probably not equivalent, strategy.) This makes all the unpaired electron spins in the ferromagnet align, changing the character of the insulator material B, and presumably altering its band diagram and thus the potential-barrier height it provides. As a result, the tunneling rate is greatly increased, or decreased, depending on the sign of the effect. So this is a magnetically-switchable switch. Now for most purposes, I would think that magnetic control is less desirable than electrostatic control like in a MOSFET. Also, it would be crucial to precisely control the barrier width, presumably making manufacturing difficult. So this (at least naively) seems less desirable for most purposes (e.g. building computers) than more conventional transistors. Perhaps there are ways to overcome that naive impression. For example, a pulse traveling along a transmission line carries with it a small blob of electric and magnetic field, which perhaps could be used via an MFIM to amplify/create other such pulses, and some kind of pulse-based (rather than voltage-level-based) logic system could be employed. Or perhaps this kind of switch would work well in combination with superconductors. (All that is a fantasy. I'd like to know if anybody can think of reasons the MFIM should be more useful than it naively appears. I suspect is it not very useful, but hopefully that suspicion is naive.) On the other hand, MFIMs have a few virtues that conventional transistors cannot match. They will work at arbitrarily cold temperatures, in contrast to doped semiconductors, which I believe depend on thermal ionization of the dopants. Indeed MFIMs should be extremely immune to temperature changes (although exceeding the Curie temperature of the ferromagnet would not be allowed; the Curie temperature is likely to be 300-800 kelvin depending which material). This would also be a thermally-switchable switch (Curie temp.) which with correct design could always switch "off" at high temperature, unlike semiconductors which always become more conductive at high temperature. This feature perhaps could protect against overheating-caused failure. It does not need high purity materials like 99.999999% silicon. Indeed it uses only metal and insulator. It might be designable to work well at either much lower or much greater voltages than conventional transistors. Perhaps it can be shrunk smaller than conventional transistors (?) since one factor limiting conventional transistors is dopant control and the fact tiny gate regions literally have a small number of dopant atoms inside, whose number varies randomly. Another factor limiting conventional MOSFETs is insulator breakdown caused by high electric fields caused by combination of gate plus thin insulator. These seem inescapable at small sizes since high electric fields are needed for switching effect at small sizes. With the MFIM, switching might still be possible even at arbitrarily low electric fields. Conclusion: at present I'm dubious the MFIM is very useful, but I'm unsure. I'm about 99% sure it will work (the physics involved seems very well understood and the assumptions needed are very mild). The fact it should be very immune to temperature changes and mechanically robust (since solid state) is one respect in which it seems superior to all competition I can think of. -- Warren D. Smith http://RangeVoting.org <-- add your endorsement (by clicking "endorse" as 1st step)
participants (1)
-
Warren D Smith