Digital Infrastructure for Materials Science: A Flexible Research Data Management System with User-Defined Types

Digitalisation in materials science is an essential prerequisite for transparent and efficient research practices. Every laboratory engaged in materials research benefits from maintaining a digital twin of its experimental processes, enabling not only efficient data collection and processing, but also long-term planning and reuse of accumulated data. As research projects in materials science grow in scale and complexity - often involving multiple institutions and interdisciplinary teams - there is an increasing demand for adaptable platforms that can handle heterogeneous data types from both experimental and computational sources, and that can integrate laboratory data with theoretical models. To address this challenge, we have developed and continuously improved MatInf, a highly customizable Research Data Management System (RDMS) tailored for the needs of materials science. MatInf is designed to support a broad range of data management tasks, offering an extensible system of data types built upon a few fundamental ones. The core type is a simple object, that encapsulates basic metadata common across all entities in the system. Specialized types such as chemical system (for qualitative composition), chemical compound (for quantitative composition, inheriting from chemical system), literature publication, and handover (for tracking physical sample transfers between users) are derived from this core type. On top of this foundation, administrators can define domain-specific data types to match their research needs, such as experiment plan, nano-powder sample, materials library, volume composition, surface composition, composition measurement, measurement device, experiment report, and many others. Each custom type can be configured using flexible templates that define how data and metadata are structured, either as ordered key-value pairs with strict typing or as tables with user-defined column structure. This approach enables accurate modeling of real-world entities in a semantically rich and reusable manner. Each object in the system can also store an associated document, such as a data file. A key feature of MatInf is its capability to interface with these documents through a customizable API that supports document validation, metadata/data extraction, and visualization. These actions are handled by external web services, allowing seamless integration of new document formats without modifying the RDMS core. As a result, documents can be checked for validity before upload (ensuring uniqueness and correctness), parsed for relevant data during upload (minimizing manual entry errors), and rendered interactively using third-party applications configured for the document type. For user-friendly organization, all data objects are stored in a hierarchical tree of projects, which supports unlimited nesting. More importantly, users can define directed, typed relationships between objects, forming a Directed Weighted Multigraph that enables the construction of a domain-specific Knowledge Graph. This graph structure enhances semantic navigation and enables intelligent querying across complex data relations. MatInf provides a web-based user interface for efficient and secure data access, with search capabilities across both qualitative and quantitative chemical compositions, as well as any custom-defined properties. Additionally, it exposes a powerful API that supports arbitrary SQL-based queries for advanced use cases. The system is fully multi-tenant, highly scalable, and has been successfully deployed in several major materials science research initiatives, including CRC 247, CRC 1625, and ERC DEMI.