University of Michigan and Imec launch Common Earth project to cut rare elements and PFAS from chipmaking
Why one obscure element and one class of chemicals sit on the critical path for every transistor is a lesson in supply-chain thinking.
IEEE Spectrum published on 15 August an interview with two University of Michigan researchers leading a project called Common Earth, a collaboration with the nanoelectronics research centre Imec. The project's goal is to remove supply-chain bottlenecks in silicon chip manufacturing by finding substitutes for critical elements and for persistent "forever chemicals" used in fabrication.
Valeria Bertacco, a professor of computer engineering and vice provost, and John Heron, an associate professor of materials science and engineering, describe the problem as structural. Many chip inputs are location-constrained, can be throttled by geopolitical events, and carry social and environmental costs. The name Common Earth is a deliberate contrast with the industry's reliance on rare earths.
Hafnium is one concrete target. Heron explains that it serves as the gate dielectric in CMOS transistors, yet it is produced as a byproduct of zirconium mining, and zirconium is mined for nuclear development. Chip supply of hafnium therefore depends on activity in an unrelated industry. The team is investigating elemental alternatives, with salt-based approaches among those under study.
The second target is PFAS, the per- and polyfluoroalkyl substances that Bertacco notes can never be eliminated once released into the environment. Heron's group is working on a nitrogen-based precursor, a chemical layer deposited during manufacturing, to remove fluorine from a process step, and separately on filtration materials that capture PFAS from fab waste streams. On the design side, Bertacco says the project intends to compensate for any performance cost of alternative materials through circuit design, power management and frequency throttling, and to use chiplets so a reusable slice of a design can scale from low-end to high-end CPUs and GPUs. The article gives no funding figures or timelines.
Why it matters
Semiconductor policy has focused heavily on where fabs are built, but this project addresses what goes into them. Substituting a gate dielectric or a fluorinated precursor requires materials science and circuit design to move together, which is why the team pairs a hardware architect with a materials researcher. Regulatory pressure on PFAS is rising in several jurisdictions, so fabrication chemistry may change regardless of geopolitics. Chiplets are increasingly the mechanism by which the industry absorbs such changes without redesigning whole products.
Semiconductor policy has focused heavily on where fabs are built, but this project addresses what goes into them.
Zircon mining
Zirconium ore is mined for uses outside the chip industry
Zirconium refining for nuclear use
Volume is set by demand for nuclear-grade zirconium, not by chipmakers
Hafnium separated as a byproduct
Hafnium is recovered during zirconium refining, so its supply tracks that step
Gate dielectric in CMOS transistors
Chipmakers deposit hafnium-based films as the insulating layer under the gate
What you can learn from this
- A gate dielectric is the insulator that lets a transistor switch. In a MOSFET the gate electrode sits on a thin insulating layer above the channel; voltage on the gate creates a field that allows current to flow. As transistors shrank, silicon dioxide became so thin that electrons tunnelled through it and leaked power. Hafnium-based oxides have a high dielectric constant, so they can be physically thicker while giving the same electrical control, which is why the industry adopted them. Replacing hafnium means finding another material with a high dielectric constant, a clean interface with silicon and stability through later high-temperature steps.
- Byproduct elements have supply curves that ignore demand. Hafnium is chemically similar to zirconium and is separated from it during refining. Its output is set by how much zirconium is refined, so rising chip demand cannot on its own increase supply. Several other chip inputs, such as gallium and germanium, are also byproducts of other metals' production, which produces the same inelastic behaviour and the same exposure to decisions made elsewhere.
- PFAS are used because fluorine bonds are stable. Carbon-fluorine bonds are among the strongest in organic chemistry, which makes fluorinated compounds well suited to etching, photoresists, coatings and heat-transfer fluids in fabs. The same stability means they do not break down in soil or water, hence the "forever" label. Swapping a fluorinated precursor for a nitrogen-based one changes the surface chemistry of the film being deposited, so both the process recipe and the device characteristics have to be re-qualified.
- Design can trade performance for material freedom. If an alternative dielectric leaks more or switches more slowly, circuit-level techniques can claw some of that back: dynamic voltage and frequency scaling, power gating of idle blocks, and tighter timing margins. This is the same toolkit used to manage thermal limits today. The point is that material and architecture form one optimisation problem rather than two separate departments.
- Chiplets isolate change to a small piece of silicon. A chiplet is a functional block manufactured as its own die and joined to others inside a package. If one block needs a new process because of a materials change, only that die is redesigned and re-qualified. The reusable slice Bertacco describes lets one validated design scale from a small product to a large one by adding copies rather than starting again.
We teach this
How to use this in practice
- Draw the transistor cross-section. Sketch a planar MOSFET: substrate, source and drain, channel, gate dielectric and gate electrode, and label where the hafnium oxide sits. Add a second sketch of a FinFET or gate-all-around structure showing the dielectric wrapping the channel. Label the physical thickness of the dielectric in nanometres from any reputable process reference, so the scale of the leakage problem is visible. Done is two labelled diagrams you could explain to a colleague in under three minutes.
- Map the elements in one device you own. Pick a laptop or phone, list its main chips such as CPU, memory, radio and power management, and for each write down two or three elements it depends on, using manufacturer documentation or reputable materials references. Mark which of those are byproducts of another metal's production. Keep the table to what you can source, and record the reference beside each entry. Done is a one-page table with a column for primary source country where you can find it, and honest blanks where you cannot.
- Trace hafnium's supply chain as a flow. Boxes: zircon sand mining, zirconium refining, hafnium separation, precursor synthesis, and atomic-layer deposition in the fab. Note on each arrow what determines volume at that step. Add a parallel row for a second byproduct element such as gallium, and note where the two chains diverge. Done is a diagram that makes visible why chip demand alone cannot pull more hafnium into the market.
- Read one PFAS regulation summary. Find the current PFAS restriction proposal from the European Chemicals Agency or from your national regulator and note which industrial uses are covered, what exemptions are proposed for semiconductors, and the timeline. Compare the transition period against a typical fab process-qualification cycle described in the same references. Done is a half-page note listing the compounds named, the transition period, and one fab process step that would be affected if the exemption lapsed.
Sources
- Common Earth Project Aims to End Chip Supply Chain Bottlenecks — IEEE Spectrum
Our reporting is an original summary; full coverage is at the links above.
Don't just read about it — build it.
Square 1 teaches the skills behind the headlines, with every line of your work graded by AI. Find your starting point in 3 minutes.
Get your free skill report