Thallium substitution effects on Co valence states and room-temperature ferromagnetism in co-doped In2O3: Experimental and machine-learning insights

• Tl substitution in In2O3:Co increases the Co3+ fraction before annealing. • Vacuum annealing partly reduces Co3+ to Co2+, but higher Tl weakens ferromagnetism. • RT ferromagnetism appears after annealing for x=0.0 and diminishes with Tl content. • Ridge‑regression model: larger ionic radius favors Co3+, high ionization energy suppresses RTF. • ML suggests Pr, Lu, Ce, Nd as promising trivalent dopants for stronger room‑temperature magnetism. We systematically investigate the effects of thallium (Tl) substitution in Co-doped indium oxide ( I n 1.9 - x T l x C o 0.1 O 3 ), focusing on its influence on Co valence states and magnetic properties. X-ray diffraction (XRD) confirmed that the bixbyite structure was preserved across a wide Tl composition range ( x = 0.0–0.6). X-ray photoelectron spectroscopy (XPS) revealed that Tl incorporation increased the C o 3 + / C o 2 + ratio before annealing. Vacuum annealing induced partial reduction of C o 3 + to C o 2 + , but higher Tl substitution led to weaker enhancement of ferromagnetic order. Magnetization measurements showed that the pristine ( x = 0.0) composition transformed from a paramagnetic to a room-temperature ferromagnetic state upon annealing, while higher Tl content systematically weakened ferromagnetism. These results suggest that the larger ionic radius of T l 3 + modifies crystal-field environments and magnetic exchange interactions, promoting C o 3 + formation yet suppressing long-range ferromagnetism. A machine-learning model using ridge regression predicted the C o 3 + ratio and room-temperature ferromagnetism (RTF) based on ionic radius and ionization energy with high accuracy (R 2 ≥ 0.85). Predictions for hypothetical dopants in I n 1.6 M 0.3 C o 0.1 O 3 suggested rare-earth elements such as Pr, Lu, Ce and Nd could enhance C o 3 + content and ferromagnetic properties. This study highlights how ionic size and ionization energy influence valence states and magnetism in I n 2 O 3 -based diluted magnetic semiconductors, demonstrating the synergy of experiments and machine learning in spintronics material design.