Ultra-low-energy physical unclonable function enabled by trap-engineered ferroelectric tunnel junction crossbar

Abstract

Ferroelectric tunnel junctions (FTJs) have emerged as promising building blocks for energy-efficient and reconfigurable hardware security owing to their nonvolatility, scalability, and polarization-dependent tunneling transport. Here, we demonstrate an ultra-low-energy physical unclonable function (PUF) implemented using a 48 × 48 FTJ crossbar array based on a TiN/HZO/SiO2/n+-poly-Si metal–ferroelectric–insulator–semiconductor (MFIS) stack. All 2304 FTJ devices exhibit uniform and repeatable switching behavior with endurance (>103 cycles) and retention (>104 s), enabling reliable large-scale array operation. High-pressure annealing (HPA) is introduced to suppress defect activity in the HZO layer, which is experimentally verified by low-frequency noise (LFN) analysis. The LFN spectra reveal a clear transition from trap-induced 1/f noise to shot-noise-dominant behavior, indicating effective passivation of oxygen-vacancy-related traps and stabilization of direct tunneling in the high-resistance state (HRS). As a result, the HRS read current becomes nearly temperature independent at low bias, providing a robust entropy source for PUF operation. Based on a differential current-summation and comparison scheme, the FTJ crossbar PUF achieves a vast challenge–response pair (CRP) space of approximately 1027, enabling strong resistance to brute-force attacks. Owing to the HPA-induced tunneling stabilization, the proposed PUF exhibits a bit error rate (BER) below 1% even at 100 °C without any error-correction circuitry. Furthermore, the PUF operates with an ultralow bit-level energy consumption of 96 aJ, enabled by direct-tunneling-dominant transport and low-voltage readout. Statistical evaluations confirm nearly ideal uniformity, diffuseness, and uniqueness, while intrinsic cycle-to-cycle variation allows fully reconfigurable key generation without hardware modification. The generated responses successfully pass all NIST SP 800–22 randomness tests and show strong resilience against machine-learning-based modeling attacks. These results establish trap-engineered FTJ crossbars as a compelling platform for compact, thermally robust, and ultra-low-power hardware security in next-generation electronic systems.