Steric‐Pressure‐Driven Topological Reconstruction for High‐Performance Zintl Phase Thermoelectrics with Machine‐Learning Validation

ABSTRACT

The Zintl phase A ₉ M _{4+ x} Pn ₉ (A = Ca, Sr, Eu, Yb; M = transition metal; Pn = pnictogen) family is a promising class of thermoelectric (TE) materials, but conventional optimization has relied on interstitial site tuning rather than structural reconstruction. In this work, we achieve a steric‐pressure‐driven topological transformation in the quinary Ca _{9− x} Sr _x Zn _{4+ y − z} Ag _z Sb ₉ system through synergistic Sr and Ag co‐doping. Single‐crystal X‐ray diffraction reveals that substituting larger Sr ²⁺ for Ca ²⁺ progressively drives Zn migration, transforming the parent 2‐dimensional (2D) zigzag‐shaped anionic layers into a cross‐linked 3D polyanionic framework. Density functional theory calculations indicate that Ag incorporation preserves the pseudogap near the Fermi level and reduces bond polarity, optimizing the electronic structure for hole transport. Together with the reconstructed 3D framework, these electronic effects significantly enhance carrier mobility and electrical conductivity without sacrificing a large Seebeck coefficient. As a result, the fully transformed quinary compound Ca _7.59(2) Sr _1.41 Zn _4.37(2) Ag _0.19 Sb ₉ achieves a maximum ZT of 0.73 at 873 K, doubling the performance of the parent ternary Ca ₉ Zn _4.5 Sb ₉ . A hybrid transfer machine‐learning approach (CrabNet and XGBoost) validates compositional trends and identifies the high‐performance region. These findings demonstrate that steric‐pressure‐induced 2D‐to‐3D framework reconstruction coupled with electronic optimization offers an effective design principle for high‐performance Zintl TE materials.