In this project, we'll create a 70s-style microfeature using photolithography. Before we go any further, we should establish that this DIY setup isn't meant to produce a device suitable for practical use in modern applications. We're building this to help us understand a process involved in old-school IC fabrication—and to marvel at how much technology has advanced over the last 50 years.
History of Silicon and DIY Transistors
Integrated circuits are at the heart of modern electronics and have been for decades. It is amazing to think that billions of calculations can be done per second on a device that fits on your fingernail where the smallest details are only nanometres in width.
Modern devices use such precise equipment that the air in the fabrication facilities has to contain less than 100 dust particles per cubic meter. To give you an idea of how clean that is, it is significantly cleaner than an operating room might be.
In no small part because of requirements like these, making modern ICs at home is impossible. You can, however, have the experience of fabricating an IC at home that would have been produced in the 1960s or 70s. After all, if you look at photos from that era, you will find that PCB fabrication took place in environments far less clean than any operating room.
Photo-engraving on silicon wafers in 1967. Note the lack of high-tech cleaning equipment. Image courtesy of the IET Archives Blog.
In this project, we will begin with the first step in micro designs, micro-lithography.
Before we continue, it is worth noting a pioneer of home integration, Jeri Ellsworth. Ellsworth was one of the first (if not the first) hobbyist to actually create transistors at home. Her method involved taking silicon slices, growing oxides, and adding contacts to create a successfully working FET. But only one transistor could be made at a time using this method and the transistors were fairly large (measured in inches).
What if these transistors could be shrunk down in size? What if layers could be added?
The aim of this project is to take a fairly large image (2cm x 2cm) and shrink it down to at least 2mm x 2mm. This means that individual features are smaller than the width of a hair and will provide the foundation for integrated circuits at home.
I would like to give special thanks to my brother, Toby Mitchell, who is co-running this project with me. I would have never gotten any of this to work without his input and aptitude for science.
Image reduction is a fairly easy process which involves using a series of lenses to take a point source of UV light (UV LED, white LED), collimate the light, condense the light over a mask, and then shrink the mask image down to the workpiece.
A diagram is shown below of the setup of the lenses. In the experiment, a microscope was used to shrink the projection of the mask as opposed to using custom lenses. The microscope gives the ability to shrink the image down by a factor of 400, but a magnification of 40 was used for testing purposes.
Optical set-up of the image reduction stage
The aspheric lens used is a generic 44mm LED collimator found on eBay for a few dollars.
44mm LED collimator. Image courtesy of eBay.
Collimator lens, artwork, and microscope entrance side view
The setup for initial experimentation is shown below. Note that the microscope is on its side for ease of use. The microscope draw tube (the section you look through) was at 45 degrees to its normal position, which made it difficult to align the lenses. The workpiece holder is mounted onto an adjustable table to enable focusing of the projection. But over time (due to gravity) the table would slowly lower due to its own weight. Therefore, it was easier to rest the scope on its side and use a flat table to align the lenses.
Microscope and lens setup. LED is off.
Microscope and lens setup. LED is on and the projection can be seen just above the 40x magnifier.
Microscope and lens setup. Test projection onto black plastic.
Photoresist Film on Substrate
The industry uses spin coated materials for the photo-resistive layer but, as we do not have access to such chemicals, we used dry film resist instead.
The substrate in this experiment is FR2 copper laminate because obtaining silicon pieces was outside our budget and we can etch the copper using ferric chloride (which is abundantly available).
Here is how to apply the film:
- Pre-heat a laminator
- Clean the copper to a high degree via steel wool and then use isopropyl alcohol (IPA) to remove grease
- Etch the copper for 10 seconds in ferric chloride. This does several things:
- Removes scratches
- Smooths the copper
- Cleans the copper
- Makes the copper dark and easier to see under a microscope
- Clean the copper a second time but ONLY using IPA
- Laminate the dry film onto the copper
Remember to perform the lamination process in safe light (or low light conditions) as you can potentially expose the resist and ruin the substrate.
With the substrate ready, it is time to prepare your setup for exposure!
The mask holds the image that you wish to expose onto the workpiece. But be warned—you will need to use a negative of the image that you intend to put into the copper.
Areas on the mask that are dark will be etched away and areas that are light will be safe from the etchant. So if you wanted to expose “hello” onto the workpiece, the image will have to look like this:
Note that your image will be flipped depending on the number of focusing lenses you use.
The mask can be printed using either an inkjet or laser transparency. What matters is how opaque the image is because the higher the contrast, the better the image will turn out. The less opaque the dark areas are, the more light will pass through which will result in a poorer image.
High DPI is recommended and, if necessary, you can double or even triple up multiple masks to create a very brilliant image.
Microscope Prep and Substrate Exposure
The microscope needs to be configured before you can expose the substrate. This is because the resist will begin to react to light and potentially create ghosting on the final workpiece.
So to set up, get some scrap copper clad and place it under the microscope where the workpiece will sit. You can either use a USB microscope or your eye (if you can see small detail) to correctly focus the image on the workpiece.
Once you are happy with the setup, place the substrate into the setup and expose the workpiece.
You will need to determine exposure time through experimentation but, for a typical 5mm UV LED with a magnification of 4x, the resist will be well exposed after 5 min.
A white LED can be substituted for a UV LED but the exposure time can end up longer, possibly hours. This does, however, give the option of focusing the image directly onto the substrate instead of using a scrap piece.
Projected image onto photoresist and copper. Taken with a digital microscope (image size approx 5mm x 5mm).
Image after one hour of exposure to white light (LED 3mm).
Carefully develop the piece in the 1% sodium carbonate solution until the unwanted resist is gone. This will be difficult to determine but do not be tempted to rub the copper as you may damage the small features (which are easily scratched). At the same time, you will also be surprised at how strong the resist is (considering its size).
With the workpiece developed, you should examine the piece under a microscope to check that the unwanted areas have been removed. It is at this stage that you can troubleshoot your system to improve the resolution of the image as well as the sharpness.
Design approx 3mm across. Wire widths approx 62.5um.
Test with 10x magnifier. Image is approx 2mm x 2mm. (These features are smaller than the width of a hair!)
Etching will be difficult because ferric chloride has a bad habit of not tidying up after itself. When the iron displaces the copper, it leaves an iron oxide residue which can sit on top of the copper and prevent further etching of the copper. This is why agitation is imperative in etching such small details but you must be careful not to disturb the film.
One promising etchant is Copper Chloride in Aqueous Hydrochloric Acid. This is closer to ideal as it does not leave a residue behind but it takes much longer to etch than ferric chloride.
First copper etch test. Note that the device was not fully developed during the sodium carbonate submersion (device approx 3mm x 3mm).
Integrated Circuits of the Past and Future
In this project, we completed a microfeature using photolithography without needing any facilities outside of our own home. This is the first step towards old school micro-design. Again, any device derived from this project clearly can't stack up to modern ICs, which are far more delicate and complex.
Remember the image at the beginning of the article of an IC fabrication facility from the 60s? Compare that to this clean room at the Rochester Institute of Technology:
Researchers in the Semiconductor and Microsystems Fabrication Laboratory at the Rochester Institute of Technology. Image courtesy of RIT.
This juxtaposition serves to demonstrate how IC fabrication has changed. ICs are now so sensitive that dust can compromise the manufacturing process.
Cutting-edge ICs can't be created at home, and there is no practical reason to try when so many reasonably priced professional-grade components are readily available. Nevertheless, Jeri Ellsworth and other pioneers like her have shown that home transistor-creation processes are possible—perhaps one day we could see home chip fabrication.
Regardless of the future, the development of this technology from half a century ago still has relevant lessons to teach the engineers of today.