Template‑Assisted, BEOL‑Compatible Crystallization of Sub‑4 nm Ferroelectric HZO via Seed‑Layer Engineering for FeRAM

Background & Challenge: Next‑generation ferroelectric random‑access memory (FeRAM) requires ferroelectric Hf₁₋ₓZrₓO₂ (HZO) films scaled to ~4–5 nm. At these thicknesses, achieving robust ferroelectricity is exceptionally challenging under the strict back‑end‑of‑line (BEOL) thermal limit (~400 °C). Without assistance, ultrathin HZO crystallizes into non‑polar phases at low annealing temperatures, yielding weak remanent polarization. While selected oxide seeding can reduce crystallization temperature and improve BEOL compatibility, it often does so at the expense of polarization and reliability. A template‑assisted seeding strategy is therefore required to enable low‑temperature crystallization without sacrificing performance.

Objectives: Crystallize 4–5 nm HZO within BEOL limits while achieving high FeRAM performance: 2Pᵣ > 50 µC/cm², wake‑up‑free switching from the first cycle, operating voltage < 1.3 V, and endurance > 10¹² cycles. The working hypothesis is that engineered material seed layers can tailor nucleation and texture to preserve large, well‑oriented grains and strong ferroelectric response.

Approach: Apply template seed layers  developed by atomic layer deposition  (ALD) to assist HZO crystallization at low temperature while preserving performance. Two generic routes will be compared: (i) a seed promoting predominantly in‑plane grain orientations, and (ii) a seed favoring out‑of‑plane grain orientations through structural templating. ALD deposition followed by optimized annealing (time, pressure, atmosphere) will be used for each stack. Grain orientation distributions (e.g., in‑plane vs. out‑of‑plane textures) and ferroelectric properties will be systematically analyzed to identify the optimal BEOL‑compatible route.

Risks & Mitigation: A given seed may show limited influence or stability at BEOL annealing conditions; mitigation includes tuning anneal ambient and thickness, and evaluating alternative material seeds. If polarization is limited, co‑doping or multi‑layer seed concepts will be explored. Parallel evaluation ensures rapid convergence on the optimal strategy.

Expected Impact: Deliver ultra‑thin (4–5 nm) FeRAM capacitors that are BEOL‑compatible while providing high polarization, wake‑up‑free operation, and long endurance. This work will enable dense embedded non‑volatile memory in advanced chips without compromising performance.