When an IC photomask is being fabricated, a laser carefully dissolves specific portions of the photo resist to develop the precise pattern. But when we make ICs, we use a photomask to place over the silicon and shine coherent light over it at once.

Since most of the engineering problems arise from this photolithographic process using a photomask, why do we not just laser etch the silicon wafers themselves? Would it take too long?

 

We press records in the factory, we cut them in the studio...

 

I have a few hundred photo copies of a small circuit board on a clear sheet, 8&1/2 by 11 inches.
Is it faster to program a machine to cut the resist layer for 200 circuits on one board or make a clear sheet?
Which will wear out faster? The laser or the clear sheet?

 

One of the first semi projects I worked on was a direct-write electron-beam machine from Fujitsu called the DAX. While it worked well for prototype ASIC masks with 5 or 6 inch wafers it was a major pain to keep the precision stage and beam alignment from drifting. Yes, it would take too long to write the same patterns over and over again for each die. With electrons it was possible to deflect the beam over a sizable area without moving the stage with electrical deflector plates, with a laser that's not possible so a system of optical write rotational scanners would have to be used instead.

Slow, expensive, complicated and did I say slow.