Monday, May 21, 2018

Physics around you: the field-effect transistor

While dark matter and quantum gravity routinely get enormous play in the media, you are surrounded every day by physics that enables near miraculous technology.  Paramount among these is the field-effect transistor (FET).   That wikipedia link is actually pretty good, btw.  While I've written before about specific issues regarding FETs (here, here, here), I hadn't said much about the general device.

The idea of the FET is to use a third electrode, a gate, to control the flow of current through a channel between two other electrodes, the source and drain.  The electric field from the gate controls the mobile charge in the channel - this is the field effect.   You can imagine doing this in vacuum, with a hot filament to be a source of electrons, a second electrode (at a positive voltage relative to the source) to collect the electrons, and an intervening grid as the gate.  Implementing this in the solid state was proposed more than once (LilienfeldHeil) before it was done successfully. 

Where is the physics?  There is a ton of physics involved in how these systems actually work.  For example, it's all well and good to talk about "free" electrons moving around in solids in analogy to electrons flying in space in a vacuum tube, but it's far from obvious that you should be able to do this.   Solids are built out of atoms and are inherently quantum mechanical, with particular allowed energies and electronic states picked out by quantum mechanics and symmetries.  The fact that allowed electronic states in periodic solids ("Bloch waves") resemble "free" electron states (plane waves, in the quantum context) is very deep and comes from the underlying symmetry of the material.  [Note that you can have transistors even when the charge carriers should be treated as hopping from site to site - that's how many organic FETs work.]  It's the Pauli principle that allows us to worry only about the highest energy electronic states and not have to worry about, e.g., the electrons deep down in the ion cores of the atoms in the material.  Still, you do have to make sure there aren't a bunch of electronic states at energies where you don't want them - these the are traps and surface states that made FETs hard to get working.  The combo of the Pauli principle and electrostatic screening is why we can largely ignore the electron-electron repulsion in the materials, but still use the gate electrode's electric field to affect the channel.  FETs have also been great tools for learning new physics, as in the quantum Hall effect

What's the big deal?  When you have a switch that is either open or closed, it's easy to realize that you can do binary-based computing with a bunch of them.  The integrated manufacturing of the FET has changed the world.  It's one of the few examples of a truly disruptive technology in the last 100 years.  The device you're using to read this probably contains several billion (!) transistors, and they pretty much all work, for years at a time.  FETs are the underlying technology for both regular and flash memory.  FETs are what drive the pixels in the flat panel display you're viewing.  Truly, they are so ubiquitous that they've become invisible.

4 comments:

Anonymous said...

For better or for worse, transistor scaling is slowly coming to an end. Doug, we need smart physicists like you to find us a new device that will replace FinFETs! None of the devices currently on the "post-CMOS" roadmap are actually satisfying to circuit engineers.

Douglas Natelson said...

Anon, I think it's Si all the way down to a few nm, and then 3d integration of some form may be the way to go rather than a wholesale switch to new materials. I've always liked surround-gate designs (http://nanobio.umn.edu/Library/Hergenrother_VRG_Elsevier02.pdf and nanowire-based versions). I agree that scaling is looking very tough. Reading the Roadmap discussion of "Beyond CMOS" technologies (https://irds.ieee.org/images/files/pdf/2017/2017IRDS_BC.pdf) drives home to me how far we are from viable, mass-produceable alternatives to CMOS.

David Brown said...

"... a wholesale shift to new materials ..." Such a shift might work, but the problems are severe.
Carbon nanotube field-effect transistors#Disadvantages, Wikipedia

Anonymous said...

Well-timed comment about gate-all-around transistors Doug... Samsung just announced that these were their transistors of choice for the 3nm node. But GAA FETs also tend to deliver less current than FinFETs, so they are not an ideal replacement.