In-House Alloy MI Search System Selects Only “Makeable” Material Candidates, Verified Across Multiple Material Systems
Space Seed Holdings Inc. (Headquarters: Minato-ku, Tokyo; CEO: Kengo Suzuki, “the Company”) announces that it has built in-house an alloy MI (Materials Informatics) search system that narrows AI-generated material candidates down—at the very top of the workflow—to only those “actually makeable on the Company’s own manufacturing equipment,” and has verified end-to-end operation, from search to the output of promising candidates, across multiple material systems, including for space applications.
The defining feature of the system is that, immediately after AI generates a large number of material candidates, it inserts a “can this actually be made?” gate that removes from the outset any candidate that looks excellent on paper but cannot in fact be produced.
Background: Manufacturing in Space and on the Moon, and a Focus on Sintering
The Company is developing materials with an eye toward manufacturing in space, materials that protect habitats, and manufacturing on the Moon. In space- and lunar-based manufacturing, where materials must be formed using limited resources and energy, sintering (SPS: Spark Plasma Sintering)—which consolidates powder into a dense solid in a short time—has drawn attention as a highly compatible forming technology.
At the same time, identifying promising options from the vast number of possible material combinations is an enormous task. The Company therefore focused on MI (Materials Informatics), a data-driven materials-design approach that uses AI and related methods to efficiently narrow a huge space of material combinations down to promising candidates.
The Achievement: An End-to-End MI Search System from Search to Selection
The Company has built an MI search system (software) that performs alloy-material exploration end to end. The system automatically executes a sequence in which (1) AI generates many material candidates, (2) only those “actually makeable on the Company’s own manufacturing equipment” are filtered at the very top of the workflow, and (3) promising candidates are output.
Excluding from the outset “candidates that are excellent in calculation but cannot actually be made” is the system’s defining feature. Rather than leaving this as a desk-level design, the Company built it to a state in which it actually runs and implemented it internally. The Company has confirmed end-to-end operation—from search to filtering to the output of promising candidates—and has also confirmed that candidate filtering can be performed across multiple material systems, including for space applications.
Connecting Software (Design) and Hardware (Verification)
Materials development gains substantially in value when design (software) and verification (hardware) mesh together. The Company is advancing, together with collaborators, a framework for material verification using sintering (SPS).
By placing the “makeable candidates” output by MI onto the sintering process, the Company connects design through to real-equipment verification as a high-throughput flow that tests many candidates at high speed. Design intelligently with AI, verify quickly with sintering: the Company aims to build a foundation that turns this round trip quickly.
Anticipated Use Cases
・Search from the outset only for candidates the equipment “can make”: By placing a “gate of equipment-makeable conditions” at the very top of the search, rework in later stages is reduced and effort can concentrate on candidates that can proceed to verification.
・Design materials for space that are “light, strong, and heat-resistant”: Envisioned applications include structural materials, exterior wall materials for habitation modules, and lunar manufacturing. Light weight tends to conflict with load-bearing capacity and heat resistance, and as the number of conditions to satisfy grows, manual trial and error is prone to losing sight of the optimum. The system supports such multi-condition exploration.
Outlook
The Company aims to build a structure that can make proposals from a materials-development standpoint across both software (design) and hardware (verification), targeting new-materials fields beginning with semiconductors. The alloy MI search system built this time is positioned as a first step toward the software foundation for that structure.
Comment from the CEO
“Excellent in calculation, yet impossible to actually make”—to overcome this wall that the materials-development field faces, we built with our own hands a search system that narrows AI’s proposals down from the outset to only “makeable candidates,” and we confirmed it to the point where it actually runs. What we are looking toward is an era in which we manufacture in space and people can live in space. The materials needed for that era—we design and verify them here on the ground, now. Design intelligently with AI; verify quickly with sintering. Starting from a foundation that can turn this round trip quickly, we will dramatically raise the speed of new-materials development and bring our mission, “Turning Science Fiction into Nonfiction,” into reality one step at a time.
—Kengo Suzuki, Representative Director and CEO, Space Seed Holdings Inc.
About Space Seed Holdings Inc.
Space Seed Holdings Inc. is a space-focused deep-tech venture builder with the mission of “Turning Science Fiction into Nonfiction.” The company engages in investment, research, and venture creation, centered on operating the “Fermentation and Longevity Fund,” which supports the social implementation of fermentation and longevity technologies to address societal challenges. In collaboration with diverse stakeholders, Space Seed Holdings aims by 2040 to assemble the technologies necessary for sustained human habitation in space.
https://ss-hd.co.jp