Like… I can’t even imagine something this small. How can you produce a transistor smaller than 10nm, especially billions of them all working flawlessly?
I understand that CPUs have cores disabled due to manufacturing defects but… even so we’re talking about what feels like a mind-boggling level of precision.
How is it done?
In: Engineering
I had an engineer explain but basically very very good lenses and materials that react to light or electrons that essentially make a very small layer of material on a wafer.
Also there is a significant error rate.
The first thing to understand is that the name of the process is all marketing. There’s nothing that actually measures 3nm or 5nm in any of the transistors. It used to be a physical feature that you could measure but it stopped several years ago
The second is that it generally works by etching patterns into the silicon. The lasers don’t actually have the precision to move from transistor to transistor anymore cus they are so close and the distances are so small. So instead, they etch the first line, move over, etch the first line of another one, etc. then wrap back around to the beginning and etc the second line.
| _____ | _____ | _____ | _____ | _____
|| ____ || ____ || ____ || ____ || ____
If you get what I mean
The smallest features are the gate of the transistor, the most critical part. Open the gate, the electrons can go through. The gate can be very small in one direction and long in the other direction. So instead of a swing gate, it is more like a garage door with the voltage raising or lowering the door to let the electrons through.
The surrounding parts of the transistor are quite a bit larger to allow connections to be made and make sure the voltage is applied uniformly.
Very short wavelengths of light help define the very small dimensions, (sometimes electrons instead of light) and basically a lot of different engineering tricks have to be played to have the necessary precision to add energy to only the very small regions you want to change. Once you have changed that region, you can either remove that small region, or sometimes you want to leave that region and remove the parts you didn’t change. If the regions you are changing are connecting molecules or atoms in ways they won’t be removed by the next chemistry step, since atoms and molecules can be small the regions left behind can be small.
Other tricks including having ways were only one layer of atoms can be deposited at a time. Atomic layer deposition.
What is also neat is how all the transistors have to be connected together in the right ways with little strips of metal, and how you keep all the little metal lines from touching each other with insulating materials called dielectrics. It you dissolve the dielectrics and look at it under a scanning electron microscope it looks like a three dimensional maze.
Other things to think about is how you have to go from very tiny wires to bigger wires until you leave the chip to connect to the much larger world we live in.
First you start with a base. Then you put down a layer of stuff you want to draw with, call it icing. Then you put down a layer of stuff that is hardened by light, call it ink. Then on a separate mask that is much larger than your base, you create the pattern of things you want to remove. This mask can be big so it is easier to make. Then you put that close to the base, and focus a very bright light on the base through the mask. Think of the light as the bottom of a triangle, the base is the tip, and the mask is in the middle. This lets you draw the features on the mask much larger than they will be on the base. As long as the light is very small wavelength, you can etch small features onto the ink.
Once you etch the features on the ink you can wash off the remainder with a chemical that can remove the non-hardened ink. Now the you can put down a chemical that dissolves the parts of the icing that is not covered in ink. Then you put down a chemical that dissolves the remaining ink. Now you are left with a layer of icing that has the opposite of the pattern that you made with the mask.
You can also do the opposite of this, you can make the mask the same as the pattern you want in the icing, and put down ink that is damaged by light instead of hardened.
So a key to this process is having very bright light that has very small wavelength, and a lens that can let you focus the light inwards in a way that lets you project the pattern of the mask onto the ink on the base. This is like the opposite of a shadow puppet where you make your finger look enormous on the wall by putting a flashlight close to your finger.
Latest Answers