UV exposure

UV exposure considerations

Given that the whole point of the machine is to pattern photoresist layers with UV light, this part of the optics is perhaps the most important. Luckily for us, the system is to run at a wavelength of 405nm. 405nm is a lot more forgiving in terms of optical glass so we can get away with lower cost optical components.

Intensity considerations

Let's take a nice and common easygoing photoresist as reference: Shipley S1813. It has a nominal exposure of 150mJ/cm^2 (at 435nm).

Let's compare this to the rating for the DLP300s (the lower power version of the chipset). Since Medjed is a 5um pixel size in practice, this means the DMD magnification is very close to 1. The intensity at the sample will therefore be near enough the same intensity at on the DMD. We can see from the table (taken from the DLP300s datasheet) that there is some headroom in terms of exposure. Assuming the worst case value of 450mW/cm^2 we could expose each FOV in 1/3 of a second. The more generous value of 3W/cm62 would allow exposure in 1/20th of a second.

Illumination wavelengths < 380 nm(10)	2	mW/cm2
Illumination wavelengths between 380 nm and 390 nm	55	mW/cm2
Illumination wavelengths between 390 nm and 400 nm	450	mW/cm2
Illumination wavelengths between 400 nm and 550 nm	3	W/cm2

Of course exposure time is only part of the total patterning time. Stage movement and settle times are the other big contributors.

Let's run an example

We run a superpixel strategy, meaning we get 4X the dose on substrate compared to the 1:1 strategy. That means our 'worst case' maximum doserate budget is now 450*4 = 1.8W/cm^2.

For S1813 that means we only need to expose for 150/1800 = 83ms. With the superpixel strategry we have an exposure field of 320x320 pixels of 5um each so we expose a 1.6x1.6mm area per exposure. If we want to write 10% of a 4-inch wafer surface we would need pi * (50)^2 * 0.1 / (1.6)^2 * 0.083 = 25s purely in terms of exposure time. That seems pretty practical. If we write a whole wafer surface (unrealistic) it would be just over 4 minutes for just the exposure time. Also not too significant.

Note that of course for thicker resist or just particularly insensitive ones the nominal dose may be factor 4 higher than S1813; but even in that case the exposure time and is not fundamentally prohibitive. And also note that in this example we have completely ignored intensity losses along the optical path.