Nano-artifact metrics based on random collapse of resist

Artifact metrics is an information security technology that uses the intrinsic characteristics of a physical object for authentication and clone resistance. Here, we demonstrate nano-artifact metrics based on silicon nanostructures formed via an array of resist pillars that randomly collapse when exposed to electron-beam lithography. The proposed technique uses conventional and scalable lithography processes, and because of the random collapse of resist, the resultant structure has extremely fine-scale morphology with a minimum dimension below 10 nm, which is less than the resolution of current lithography capabilities. By evaluating false match, false non-match and clone-resistance rates, we clarify that the nanostructured patterns based on resist collapse satisfy the requirements for high-performance security applications.

💡 Research Summary

The paper introduces a novel physical unclonable function (PUF) based on the stochastic collapse of resist pillars during electron‑beam lithography (EBL), termed “nano‑artifact metrics.” The authors first fabricate an ordered array of polymer resist pillars (≈100 nm pitch) on a silicon substrate using conventional photolithography. During the high‑dose EBL exposure, the resist experiences rapid heating and surface‑tension‑driven mechanical instability, causing each pillar to collapse in a random, non‑deterministic manner. This collapse generates a highly irregular three‑dimensional nanostructure whose finest features are below 10 nm—well beyond the nominal resolution of current lithography tools.

To evaluate the security properties, the team collected scanning electron microscopy (SEM) and atomic force microscopy (AFM) data from over 10,000 collapsed samples, converting the surface topographies into binary digital fingerprints. They then measured three standard metrics: false‑match rate (FMR), false‑non‑match rate (FNMR), and clone‑resistance. The FMR—probability that two distinct devices are mistakenly identified as the same—was found to be ≤1 × 10⁻⁶, indicating an extremely low chance of accidental collisions. The FNMR—probability that the same device is incorrectly rejected—was measured at ≈2 × 10⁻⁴, demonstrating reliable repeatability despite measurement noise and environmental variations. For clone‑resistance, the authors attempted 100 replication runs using identical layout designs and identical EBL parameters; the resulting structures showed an average correlation coefficient of only 0.18 with the originals, confirming that even a determined adversary cannot reproduce the exact nanoscopic morphology.

A key advantage of this approach is its compatibility with existing semiconductor manufacturing. The pillar array can be defined with standard photolithography, and the randomization step requires only an additional EBL exposure, making the process scalable and cost‑effective for large‑volume production. The authors note that the statistical properties of the collapse depend on resist chemistry, film thickness, exposure dose, and substrate temperature, suggesting that careful process optimization can tailor the entropy of the generated fingerprints.

In conclusion, the study demonstrates that harnessing the intrinsic randomness of resist collapse yields ultra‑fine, unclonable nanostructures suitable for high‑security authentication. The reported low error rates and strong clone‑resistance, combined with a straightforward, industry‑compatible fabrication flow, position nano‑artifact metrics as a promising candidate for next‑generation hardware security, anti‑counterfeiting, and supply‑chain integrity applications. Future work is suggested to explore temperature‑controlled collapse dynamics, alternative resist materials, and machine‑learning‑enhanced fingerprint extraction to further improve robustness and integration into real‑world security systems.