The core process for manufacturing 1/100,000,000 nano semiconductors.jpg

Currently, Tesla in the US is working with xAI and Intel on the TeraFab project,
where they have decided to use FEL-EUV instead of standard EUV.
This involves using a particle accelerator to reduce the wavelength from 14nm down to 0.1nm.
Actually, the wavelength size itself does not determine the resolution of the particles.
In semiconductor manufacturing, light must be shone onto a mask before the wafer is patterned via the photolithography process;
during the process where light hits the mask and reflects back, a shorter wavelength results in higher resolution.
This is because the light must be refocused through lenses and mirrors to be patterned onto the wafer.
This happens because the overlapping of light reflected from the mask is reduced;
it is true that a shorter wavelength allows for sharper patterning of smaller masks.
For example, if the ideal mask size for 14nm light is 10cm x 10cm with a 2nm mask,
then 0.1nm wavelength light can achieve high resolution up to 0.01nm on a 10cm x 10cm area.
However, the real core process is reducing the actual size of the light itself.
For instance, if you shine 14nm EUV light onto a 10cm x 10cm 2nm mask,
and pass the reflected light through a lens filter that reduces light density to 1/1,000,000,
then pass it through a reduction lens to develop the mask, if you push the light back by 0.01cm,
the size patterned on the wafer becomes 1/1,000,000th smaller.
Since there is a limit to reducing the wavelength of light,
instead of just reducing the wavelength, the method is to reduce the light density and pass it through a high-reduction lens to shrink the light size.
As the wavelength of light gets smaller, the light hitting the mask can be focused with higher resolution,
but when patterning on the wafer, you must be able to reduce the size of the light itself to be smaller than the wavelength.
Currently, ASML's processes do not include a method to reduce the size of the light; instead, they are designed to reduce the final semiconductor size
by reducing the mask size.
The truly core technology is not reducing the wavelength, but reducing the size of the light.
When comparing a light size of 0.0001nm with a 14nm wavelength versus a light size of 0.1nm with a 0.1nm wavelength,
the size patterned on the wafer is superior the smaller the light size becomes.
There is no limit to shrinking light, but there is a limit to expanding it.
So, what are the benefits of reducing the wavelength?
For example, if we increase the integration density of the mask to create a mask with 100x higher density at 0.02nm instead of 2nm,
shining 14nm EUV on this would result in poor resolution, but using 0.1nm particle accelerator EUV would yield high resolution.
Is there no way to increase the integration density of a 0.02nm mask using 14nm EUV? It's not impossible;
the method is to make the mask in a 3D format rather than a cross-section, so that when the 14nm EUV light hits and reflects/spreads, it spreads without interference,
and then the light is gathered at a specific point using mirrors to create a single beam for high-resolution patterning.
This requires a highly advanced process where light is reflected off hundreds or thousands of micro-mirrors to be gathered back into one.
Later, when mask integration reaches ultra-high levels like 1/100,000,000nm, the mask will move from a cross-section to a 3D format,
where light is shone onto a spherical mask and the scattered light is gathered to pattern it.
Interestingly, when using 14nm EUV to pattern a 2nm semiconductor, the resolution is 80%,
but when using 14nm EUV to pattern a 0.1nm semiconductor, the resolution drops to 30%.
However, as we move toward 0.1nm semiconductors, the resolution could potentially rise to 50%.
This is because during the process of light passing through the lens filter and being reduced, light that did not reflect off the mask fills the gaps left by other light to extract resolution.
Then, by etching, refilling that space, etching again, and cutting to a constant size,
you can make it appear as if it was etched to the same depth even if the light etched 30% at point A and 70% at point B.
Then, these semiconductors are made independently in layers and stacked like films to complete the product.
Furthermore, if the mask is made 10 times larger, at 100cm x 100cm instead of 10cm x 10cm, the integration density can be increased by another 100 times.
That light is gathered, passed through a lens filter to reduce density, and then passed through a reduction lens for development.
This allows the light to be patterned even smaller.
However, to manufacture 0.1nm scale semiconductors, specific semiconductor materials are required;
if the light size is too small, like a neutrino, the light simply passes through.
Therefore, it is crucial to spray neutrino-sized light onto the photoresist to form the shape.
The wavelength of a neutrino is 0.0012 nm, which is 100 times shorter than FEL-EUV.
This means it has much higher penetration power.
Even if only 2% of the semiconductor elements on a mask developed with light reduced to 1/1,000,000th the size actually function (while 98% do not),
the integration density per unit area would be thousands of times higher than a 2nm semiconductor, and energy efficiency could improve by hundreds of times.
It can be said that this is a technology with massive potential.