NIST has issued Research Grade Test Material (RGTM) 10279, Textiles for Feedstock Identification: five fabric squares, 4 inches (10.2 cm) on a side, made from different fibers, dyed and undyed, with an undisclosed fiber composition. The material is available free through the NIST Store until July 30, 2026, in exchange for feedback and participation in a interlab study that closes Sept. 30, 2026. Interested labs can contact [email protected].

The artifact addresses a concrete measurement problem. More than half of all clothing and textiles are suitable for recycling, yet most are not repurposed, partly because manual sorting is slow and labor-intensive. The industry is moving to AI-assisted sorting, but the identification step rests on spectroscopy and algorithms whose accuracy has not been exhaustively tested.

"We've identified an industrywide measurement challenge," said NIST researcher Michelle Seitz. "Standards like this RGTM help improve textile identification and sorting, which supports advances in AI-enabled sorting of textiles and U.S manufacturing and industry."

The dominant identification method is near-infrared spectroscopy. Handheld scanners shine light on a garment; the device measures how much light passes through or scatters from the fabric, yielding a spectral fingerprint that identifies the fiber. Workers then sort by hand into bins. Automated lines feed garments on a conveyor past cameras and sensors, where algorithms perform the identification and sorting. Two competing approaches coexist: computer vision sorting by color and appearance, and hyperspectral imaging, which combines NIR and camera sensors.

Each facility runs its own variant of these technologies. Until now, nothing existed to let one sorting center verify its fiber identifications against another's. "This textile material will help validate sorting methods and make textile sorters' measurements comparable from one center to another," said NIST materials research engineer Amanda Forster. "This lays the foundation for expanding supply chains and increasing the recovery of the economic value from textiles and clothing in the U.S."

RGTMs are a relatively new category at NIST, produced on a shorter timeline than the institute's standard reference materials. NIST distributes them to laboratories that agree to measure them and share results, so the community itself determines whether the material suits its intended purpose. In this case that purpose is threefold: assessing the accuracy of sorting methods, validating the algorithms that identify fibers, and providing a physical benchmark for labs comparing methods or developing new sorting technologies.

The blend problem gives the material practical weight on the production floor. Many new textiles are blends of different fibers that are hard to identify, and the RGTM gives sorting facilities a route to production quality control. "This material also provides a way to detect things that aren't reported on the label, which is important for recycling," Forster said.

Incoming inspection is a second application. Much current research focuses on used or heavily worn clothing, but the RGTM could also serve before a garment is designed. "For example, if a brand is buying a fabric that is 100% cotton, but it ends up being a cotton-polyester blend, then they would like to know that difference," said NIST guest researcher Katarina Goodge. The material could verify composition at receiving. Fashion authentication — checking whether luxury goods are fake — is a further potential use, though NIST researchers are not currently working on it.

The interlab study now under way will decide the material's fate. Labs, manufacturers, and other organizations will apply their own fiber identification methods to the RGTM, whose composition NIST deliberately withholds, and report whether they can analyze the fibers accurately. Feedback remains anonymous. Researchers will fold the results into a more thoroughly analyzed reference material that meets industry needs — the conventional route to a certified SRM.

The open question for instrument makers and recycling operators alike is adoption: will NIR and hyperspectral vendors, and the AI sorting-algorithm developers behind them, accept an undisclosed-composition blind standard as the basis for comparable accuracy statements — and will enough labs enroll before the September 2026 deadline to make the resulting dataset statistically meaningful?