"Process Science Platform for Innovation in Materials Creation Technology -Process Informatics-" is a research and development strategy to establish a scientific framework for the efficient and integrated exploration of synthesis processes of target materials. Here, Process Informatics (PI) is defined as "a method for efficient and integrated exploration of synthesis processes of target materials through the integrated use of experimental, theoretical, and computational sciences, as well as data science, which has made remarkable progress in recent years".

By combining PI with materials informatics (MI), which efficiently searches for materials with desired functions, and measurement informatics, which observes the internal states and products of the process in real time, it is possible to advance the materials creation technology to the next stage. In addition, we aim to expand the process science platform by reforming the way of thinking about synthetic processes, which have tended to be discussed locally and individually, through the incorporation with tacit knowledge (intuition and tips) of skilled people and the combination of physicochemical analysis and data science analysis.

Today, expectations for "materials" are rising in all sorts of situations: the achievement of the SDGs, the establishment of a carbon-neutral society, the realization of resources and materials recycling, the realization of the Society 5.0 society that Japan is aiming for, and the trend toward digitalization accelerated by the onslaught of the new coronavirus infection (COVID-19). The creation of new materials has become critically important in all these areas.

Various examples have demonstrated that MI can be a powerful tool in new materials creation, but although the compositions and structures of candidate new materials are predicted, not all of them are shown to be "actually possible to make" or "how to make".

Material synthesis processes are difficult to handle in a unified manner because there are many variations of methods for each material and the parameters that control them are complex. For this reason, the improvement and optimization of individual processes have been the main focus, and scientific approach has hardly been taken for the process optimization. However, with the recent evolution of various technologies (e.g., data science, simulation technology, high-throughput experimental technology, operando measurement technology for real-time observation of processes, etc.) and the use of a supercomputer "Fugaku", which is compatible with data science, the environment for working on PI is being prepared.

The R&D issues to be addressed in this proposal include (1) development of PI methodologies in each material area, (2) construction of a common infrastructure for PI, and (3) creation of new guidelines and concepts that will enable expansion of the process science platform.

• (1) Development of PI methodologies in each material area

  We will select core processes from each area, such as organic materials, inorganic materials, and composite materials with modular structures, and start development of PI methodologies in each material area.
  For organic materials synthesis processes, the micro-scale continuous flow reactor system is a candidate example. R&D will be conducted to efficiently establish synthetic pathways for target compounds by combining synthetic experiments in a reaction space close to an ideal system, such as micro-flow chemistry, with synthetic pathfinding methods that utilize computational science and data science.
  For inorganic materials synthesis processes, the crystal growth process is a candidate example. Although it is possible to conduct precise simulations, it is difficult to use for process design because of the long calculation time required at present. R&D will be conducted to enable fast prediction of process states by building machine learning models using actual measurable data, and to utilize them in process design.
  The synthesis processes of composite materials with modular structures are complex, both in terms of the materials themselves and their synthesis processes, and the number of required parameters is very large. To optimize the entire complex process, R&D will be conducted to develop new methods such as using physically meaningful values as intermediate data in the dimensionality reduction operation known as auto-encoder in the field of machine learning.

• (2) Construction of a common infrastructure for PI

  Fundamental research that not only meets the requirements of individual processes, but also accelerates the entire PI, is important. For this purpose, it is necessary to develop machine learning algorithms optimized for PI that can deal with the characteristics of materials synthesis processes, such as the vast parameter space of the process and the influence of the multi-step processes effected each other.
  Simulation and modeling of material synthesis processes includes a) modeling using empirical parameters based on experimental data, and b) simulation using first-principles calculations and molecular dynamics. In addition to these methods, multi-physics simulation is also important as a method to connect them. Using these methods, it is also important to integrate first-principles calculations and statistical thermodynamics to construct models that can analyze the contribution of each parameter.
  The standardizations of experimental data collection methods, high-throughput experimental methods, and unit operations of robots are also important.

• (3) Enhancement of process science platform

  By designing descriptors that appropriately represent process characteristics using data science, it will be possible to classify processes that have been handled individually into new categories and discuss them. This will enable us to understand important common factors that control processes, which could not be understood through individual process analysis, and may enable us to utilize knowledge and data of each process among different processes. Such approaches will renew the discussion of synthetic processes, which has been held individually and locally, and lead to the enhancement of the process science platform.

In order to promote the R&D issues described above, it is desirable to establish a process center for each target process, and to equip it with synthetic process equipments, evaluation and measurement equipments, etc. The common infrastructure for PI requires the participation of a wide range of experts in theoretical science, computational science, data science, process technology, measurement technology, and so on, but it can function even if it is a virtual center operating in separate locations. However, they need to be established as a center for continuous powerful collaborations, rather than temporal collaboration for each project. In such a situation, the expansion of the process science platform requires a strong linkage between the process centers and the center for common infrastructure. In addition, a mechanism that unites the entire group (governing board, etc.) is necessary to make synergistic progress by utilizing the results of the R&D projects. As for human resource development, it will be important to integrate the knowledge and experience of materials researchers and data scientists. It is also important to establish a mechanism for the participation of industry and to set rules for data handling.