How do they make transistors so small that you can fit billions within a graphics card or computer components?
In: 3
A transistor is just conductor, semiconductor, and insulators laid out in a certain pattern, so that applying current to one wire affects the current going through the other two wires.
Since it’s so simple, you can literally “print” atoms of the materials onto a substrate to get the same effect. That’s how they get so small.
[Optical lithography](https://en.wikipedia.org/wiki/Photolithography).
Basically you can think of it as a kind of (very high quality) microscope objective used in reverse to image a strongly reduced map of the circuit on the silicon wafer.
The light (often ultraviolet) hardens a resin, similar to what dentists do, which will protect some parts of the silicon from the current processing step. Then the resin gets washed away, a new coat layered on, a new, complementary image used for the next step, and so on. It’s a very complex process.
Others have already explained about litography. I just would like to share this as an illustration.
In this video you can see how the transistor looks like. It even shows extreme zoom in where you can count individual atoms.
Do you mean how as in what fabrication methods or how as in how is it physically possible. I’m not as well-versed in the methods, but the theory behind it goes like this: Silicon is a semiconductor because it has a half full outer electron shell. If it were mostly empty, the few free electrons would be loosely bound and free to move about (basically, you know that opposites attract. The nucleus has protons which are positively charged and attract the negative electrons. Fewer electrons means fewer negative charges, meaning less attraction. It can’t be completely empty or else that isn’t the outer shell, the shell underneath would be)
Meanwhile, if it were mostly full, the atom hugs those electrons very tightly. So tightly that it can randomly pull electrons away from their neighbor (unless it’s completely full already… Also, this is a very simplified model for ELI5, it doesn’t actually happen exactly like this) In both cases, electrons can flow around which causes the semiconductors to act more like a conductor. Normally, the charges move randomly, but if we put a voltage across conductors, we can bias the direction of motion in one direction to get some current. As it stands, however, Silicon is pretty stable halfway between. It finds a nice little stable equilibrium.
So why are semiconductors so useful? We can introduce some imperfections in the crystal structure. Phosphorus has one more electron in the outer shell than Silicon, so if we replace a few atoms here and there, well end up with a few free electrons that can bounce around. As I said, Silicon is pretty happy with it’s half-full she’ll, so if the electron comes to it, it has no problem letting it go to the next one. This adds an extra free-moving electron or negative charge carrier so it is called n-type doping.
Boron has one fewer electron in the outer shell than Silicon, so if we replace a few atoms here and there, well end up with a few extra holes that can accept nearby electrons. As I said, funnily enough, we can actually track the hole as if it were a positively charged electron. Basically, if one atom steals a neighbor’s electron to fill the hole, it’s the same as giving the hole to its neighbor. This adds an extra free-moving hole or positive charge carrier so it is called p-type doping.
A transistor is a chunk of three doped semiconductors, either NPN or PNP and the way that turns them into magic switches is outside the scope of this answer. However, the typical concentration of these impurities is so insanely small. On the order of 1 P or B atom for every 1 billion Si atoms. That sounds like a lot. However, you can fit about 50 billion Si atoms in a cube 1 micrometer on every side. That means you can control the doping percentage to 2% and still fit 1 trillion transistors in a CPU that’s only 60 mm³ (1% of a teaspoon) but with the wiring, you’re going to have to bump that up a bit. Even if the wiring takes up 99% of the space of the CPU, that’s still only 1 tsp of space required for 1 trillion transistors.
Latest Answers