A 17-year-old from Andover, Massachusetts has won the Craig R. Barrett Award for Innovation at the Regeneron International Science and Engineering Fair for a chipless RFID tag that can be printed straight onto plastic packaging, read to identify the polymer, and then dissolved away during the wash stage of recycling.
Kevin Sun, a student at Andover High School, received the $10,000 award, which the Society for Science gives to the finalist whose work best demonstrates innovation across the STEM disciplines. The prize money is designated for post-secondary education. His project, titled “Printing 2D MXene-Based Chipless Radio Frequency Identification Tags to Enable Plastic Waste Sorting for Recycling”, reached the international fair by way of the Massachusetts State Science and Engineering Fair. Regeneron ISEF 2026 ran from 9 to 15 May in Phoenix, Arizona, bringing together more than 1,700 finalists from around 60 countries and regions.
Chipless RFID, and why it has never quite arrived
A conventional RAIN RFID tag is an antenna bonded to a silicon ASIC. The identity lives in the chip’s memory, the chip modulates its backscatter to send that identity, and the whole exchange follows a protocol with anti-collision, so a reader can single out one tag among hundreds.
A chipless tag has no silicon at all. It is a printed resonant structure, and its identity is encoded in its own electromagnetic signature: where its resonances sit in frequency, how deep they are, how the response behaves in the time domain. The reader is effectively taking a miniature radar cross-section measurement and matching it to a codebook.
The appeal is cost. Strip out the ASIC and the assembly step that attaches it, and the tag becomes ink on a substrate. The penalties are just as clear. Data capacity is measured in a handful of bits rather than the 96 or more of an EPC, so chipless tags identify a class of item far more readily than an individual one. There is no protocol layer, so there is no anti-collision and no way to selectively silence a tag. Read reliability suffers because the signature depends on the geometry of the printed structure and on whatever the tag is sitting on, and a printed resonator on a curved, damp or metallised surface is not the resonator that was characterised in the lab. That combination is why chipless RFID has stayed in the research literature for two decades while RAIN RFID took the market.
Item-level identification of a specific unit is the wrong job for a chipless tag. Telling a machine which of seven polymer types it is looking at is very nearly the right one.
Why MXene
MXenes are a family of two-dimensional transition metal carbides and nitrides, produced by selectively etching the A layer out of a MAX phase to leave atomically thin flakes with metal-like conductivity and hydrophilic surface chemistry. That last property matters as much as the first: MXene flakes disperse in water to give a stable, additive-free ink, so a conductive pattern can be printed without the high-temperature sintering that metal nanoparticle inks demand. For a printed RF structure you want high conductivity for a sharp, deep resonance, and you want deposition compatible with a plastic substrate. MXene offers both, which is why it has drawn steady interest for printed antennas and shielding.
The sorting problem it addresses
Materials recovery facilities separate plastics mostly by near-infrared spectroscopy. A NIR camera looks at light reflected off material on a fast-moving belt, matches the absorption fingerprint to a polymer, and an air jet ejects it. It works well, and it fails in well-known ways. Carbon black pigment absorbs across the NIR band, so black plastics return no usable spectrum and drop out as residue. Multilayer packaging presents a surface layer that says nothing about what is underneath. Labels, sleeves, fillers and additives all skew the reading. The sorted output is contaminated, and contaminated bales fetch less or get rejected.
Putting a tag on packaging to solve this is not a new idea, and it usually dies on the same objection: the tag is itself a contaminant. A silicon die, a metal antenna and an adhesive all end up in the recyclate. Sun’s approach is to make the identifier temporary by design. The printed MXene structure carries a thin protective coating that dissolves in alkaline water, which is exactly what the hot caustic wash in a plastics recycling plant already provides. The tag survives long enough to be useful and then goes away as part of a process step that is already in the flowsheet. Testing showed the printed tag stayed readable for at least six weeks, which is the more important number: a tag that dissolves in the wash is no use if it also degrades on a shelf.
A proof of concept, and what would have to follow
This is a science fair project, and it should be read as one. It is a well-posed idea supported by benchtop results, not a product with a route to market. Several hard questions sit between here and a working system.
Read performance is the first. A resonant signature measured on flat test coupons is not the same as one read off a crushed bottle tumbling on a belt at two or three metres per second, at varying range and orientation, among other tagged and untagged items. Reader hardware is the second: chipless interrogation typically wants a broadband frequency-swept or UWB front end rather than the standard RAIN readers a facility might already own. Then there is cost per unit at packaging volumes, MXene’s own oxidative stability over a real shelf life, food-contact and regulatory approval for a printed material on packaging, and the plain fact that nobody prints anything on billions of bottles unless brand owners, converters and reprocessors all move together.
None of that diminishes the work. Digital watermarking and other marker-based sorting schemes are already being pushed for the same reason, that NIR alone cannot see everything it needs to see, and a cheap identifier that removes itself during the wash is a genuinely elegant answer to the contamination objection those schemes keep running into. Whether it can be made to read reliably at line speed is the question worth watching.
