If I remember correctly, I heard some people saying that the transistors on CPUs today are so small, that they have to use quantum physics to make CPUs. Is that correct?


Quantum mechanics is part and parcel of understanding the function of any transistor, and therefore any integrated circuit.

As component sizes have gotten smaller, tunneling has become an increasingly important limiting factor in the design and layout of chips.


Well, they are not using it, but taking into account.

Tunneling is one of the most important:

1) As gate oxide thickness is just 1-2nm, electrons can tunnel through it -> power consumption increases (or transistor might turn itself on if gate is not connected at the moment). So if you replace usual SiO2 with high-K dielectric (like HfO2) you would be able to increase gate oxide thickness (=dramatically reduce tunneling) but electromagnetic field will remain the same (i.e. transistor would work exactly as with thin oxide)

2) Flash memory directly rely on quantum tunneling effect - electrons into strong electromagnetic field tunnel right into middle of dielectric, and form 'trapped' charge, which may stay there for years.

  • $\begingroup$ I like how you said "they are not using it, but taking into account." I doubt CPU designers need to thoroughly understand the physics behind transistors and such, but they'd probably need to know how to put these pieces together into a functional CPU. $\endgroup$ – blah Jun 25 '11 at 23:55
  • 1
    $\begingroup$ CPU designers is a general term. Logic designers probably don’t even remember how a diode works (at least they don’t need to), on the other hand wafer designer have to know the implications on choosing different materials. $\endgroup$ – gurghet Jun 26 '11 at 0:14
  • 2
    $\begingroup$ @blah CPU design teams will be hundreds strong. There will be levels of CPU designer that are not much concerned with transistor details, but there will be other levels of CPU designers that are intimately concerned with quantum mechanics. $\endgroup$ – Peter Morgan Jun 26 '11 at 0:22
  • $\begingroup$ CPU designers surely does not care about it. But engineers working on tech.processes & libraries - they have to think about it. But these guys are really rare. $\endgroup$ – BarsMonster Jun 26 '11 at 7:23

There will be several teams of CPU designers - Logic Designers, Logic Implementation, Data Integrity, Data Storage, Data Transmission, materials research and then a management team. logic implementation will most likely have physics specialists who advise on a plethora of issues. They would be expected to look at new designs and decide whether the implementation would fail due to the physical constraints and effects of the materials they use. They have to have a firm grasp of all aspects of microelectronic physics - Thermodynamics, Quantum Physics, Electromagnet effects etc. things like alteration of electron speed as it passes through different materials has to be considered so data arrival times are consistent. It is very likely that that the CPU Caches use Quantum Tunnelling for data retention, and the Cache blocks are probably kept seperate from the Transistor arrays to minimise quantum interference.


Otherwise instead of "fight" these phenomena, like tunnelling, you can think a new solution :

From http://en.wikipedia.org/wiki/Quantum_computer

A quantum computer is a device for computation that makes direct use of quantum mechanical phenomena, such as superposition and entanglement, to perform operations on data.

On the Wikipedia page you can see the photograph of a chip constructed by D-Wave Systems Inc., designed to operate as a 128-qubit superconducting adiabatic quantum optimization processor.

I am not an expert, I understand that quantum computing can be explained as :

  1. put in the N input data as a physical quantity of the system, repeated N times
  2. let's the physical system evolves and finds its natural quiet state
  3. measure the N new values of the chosen physical quantity : this set is a possible solution of the problem ( and reading it is also the most challenging task )
  4. repeat from 1. until you find satisfactory output data sets

For certain classes of problems, quantum computing seems to be the best path to solve them.

An example, solving a classical search problem - see the demo on the marvellous Wolfram web site :



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.