Towards a sustainable high-quality materials data ecosystem: frameworks and strategies.

High-quality data have increasingly been recognized as a critical catalyst in advancing artificial intelligence-driven materials discovery and design. However, the current data generation and management systems in materials science are often deficient, resulting in an abundance of low-quality data characterized by heterogeneous formats, incomplete entries and semantic inconsistencies. Such a fragmented data landscape limits direct access to reliable datasets, making it inevitable to allocate substantial resources toward data consolidation and governing existing data. Here, we propose a comprehensive quality-controlled framework encompassing all three principal stages of the research lifecycle-acquisition, management and utilization. By establishing stage-specific requirements, materials data quality can be concurrently optimized and enhanced throughout the materials research and development process. Furthermore, we propose five strategic initiatives to operationalize this framework, including an integrated computational/experimental 'data factory' for data acquisition and an automated toolkit for data quality assessment. This work aims to embed quality-governed production directly into scientific inquiry, fostering a sustainable and robust materials data ecosystem.