Lattice Graph × Ucore Rare Metals

Lattice Graph operates a computational materials-discovery platform built around a knowledge graph that spans millions of compositions and connects structure, thermodynamic properties, synthesis recipes, and patent claims into a single queryable fabric. For Ucore, the most relevant capability is what the platform does with separation and recovery chemistry: it generates candidate sorbent architectures and separation routes computationally, then subjects each candidate to consensus validation across multiple independent physics engines — machine-learning interatomic potentials including MACE and CHGNet, alongside density functional theory — before a result is promoted. A chemistry that clears only one engine is flagged; one that survives the full ensemble carries a documented confidence level that can be handed to process engineers, diligence teams, and government counterparties as evidence rather than assertion. That validation layer matters to Ucore because the Strategic Metals Complex is being built against real offtake and financing conversations, not a research program. Every separation or recovery route that Lattice Graph proposes for rare-earth oxide production, heavy-rare-earth finishing, or allied critical-mineral recovery arrives with thermodynamic stability data, a predicted selectivity range grounded in geometry-optimized binding energetics, and a cross-source agreement score that lets a process chemist know how much to trust the number before committing it to a lab trial. The phonon and free-energy checks that the platform runs as part of its standard validation pipeline are the same computations that would otherwise cost months of pilot time. The third pillar is screening at the IP boundary. Lattice Graph holds a continuously updated patent knowledge graph spanning more than 300,000 materials patents, and every candidate that clears computational validation is automatically screened for freedom to operate and whitespace positioning before it reaches Ucore's desk. That means the separation and recovery assets the platform surfaces are not just computationally credible — they have already been assessed against the existing patent landscape so that licensing and deployment decisions can be made with clarity on the IP path. Combined with a proprietary atlas of labeled negative results from failed experiments that most models and competitors never see, the platform can tell Ucore which candidate routes are dead ends before they are run on plant time rather than after.

Ucore Rare Metals is executing a focused mandate: stand up the first fully domestic rare-earth separation facility under the Strategic Metals Complex in Louisiana, using RapidSX technology to convert allied-origin mixed rare-earth feedstock into separated oxides for defense and commercial magnet supply chains. The gating constraint is not the separation technology itself — RapidSX has been demonstrated — it is the feed. Locking recurring, traceable, allied-origin feedstock that the SMC can qualify and convert is the problem every financing and offtake conversation circles back to, and it is the problem Lattice Graph is directly positioned to help solve. Lattice Graph fits Ucore on two axes that are both essential right now. The supply intelligence and mineral-deposit data layers give Ucore's commercial and technical teams a sourcing map and an element-level supply-risk read that can be turned into a defensible feedstock thesis — one with documented provenance, concentration-risk quantification, and conversion-route intelligence that tells Ucore which allied deposits produce which element co-products and how they travel through existing midstream capacity. That kind of structured, auditable sourcing narrative is what counterparties are demanding before they sign offtake or commit financing, and it is very difficult to construct from public sources alone. The second axis is the separation and recovery IP portfolio. Ucore's competitive position depends on its chemistry being both differentiated and free to operate, and on the SMC being able to qualify a wider range of feeds than a single-chemistry plant would accept. The critical-mineral recovery and recycling separations portfolio that Lattice Graph has built computationally gives Ucore a set of freedom-to-operate-clean separation and recovery assets that are directly on-thesis: heavy-rare-earth selective sorbents for the Dy/Tb splits that magnet customers pay for, broad-spectrum chelating resin platforms that extend the feedstock envelope, and integrated flowsheet claims that frame the SMC as the multi-metal recovery node that allied defense policy increasingly expects it to be.

Ucore Rare Metals is executing a focused mandate: stand up the first fully domestic rare-earth separation facility under the Strategic Metals Complex in Louisiana, using RapidSX technology to convert allied-origin mixed rare-earth feedstock into separated oxides for defense and commercial magnet supply chains. The gating constraint is not the separation technology itself — RapidSX has been demonstrated — it is the feed. Locking recurring, traceable, allied-origin feedstock that the SMC can qualify and convert is the problem every financing and offtake conversation circles back to, and it is the problem Lattice Graph is directly positioned to help solve. Lattice Graph fits Ucore on two axes that are both essential right now. The supply intelligence and mineral-deposit data layers give Ucore's commercial and technical teams a sourcing map and an element-level supply-risk read that can be turned into a defensible feedstock thesis — one with documented provenance, concentration-risk quantification, and conversion-route intelligence that tells Ucore which allied deposits produce which element co-products and how they travel through existing midstream capacity. That kind of structured, auditable sourcing narrative is what counterparties are demanding before they sign offtake or commit financing, and it is very difficult to construct from public sources alone. The second axis is the separation and recovery IP portfolio. Ucore's competitive position depends on its chemistry being both differentiated and free to operate, and on the SMC being able to qualify a wider range of feeds than a single-chemistry plant would accept. The critical-mineral recovery and recycling separations portfolio that Lattice Graph has built computationally gives Ucore a set of freedom-to-operate-clean separation and recovery assets that are directly on-thesis: heavy-rare-earth selective sorbents for the Dy/Tb splits that magnet customers pay for, broad-spectrum chelating resin platforms that extend the feedstock envelope, and integrated flowsheet claims that frame the SMC as the multi-metal recovery node that allied defense policy increasingly expects it to be.

The critical-mineral recovery and recycling separations portfolio contains the assets most directly relevant to what Ucore is building at the SMC. The anchor for the rare-earth core business is the ion-imprinted phosphonate-bis-picolinamide resin for dysprosium and terbium separation from magnet leachate. Dy and Tb are the heavy rare earths that defense and premium magnet customers actually pay a premium for, and they are the hardest pairs to resolve in any separation train. This asset achieves predicted single-pass Dy/Tb separation factors of four to eight and Dy/Nd factors of thirty to one hundred, targeting the exact split that RapidSX needs to deliver at the SMC finishing step, with a freedom-to-operate carve-out against the conventional solvent-extraction reagent art. For Ucore it is both a differentiation story and a capital-efficiency argument: a resin route that compresses the solvent inventory and physical footprint on the step that matters most. The second cluster in the portfolio widens the feed envelope that the SMC can credibly accept and monetize beyond the rare-earth core. The universal chelating-resin platform configures a single crosslinked support across twelve critical-mineral oxocations — germanium, antimony, tin, vanadium, molybdenum, tungsten, and others — from named zinc, copper, and Bayer process streams, all on a clean freedom-to-operate position. The sterically hindered catecholate resin for germanium recovery from zinc-refinery residue operates at pH one to three with confirmed selectivity against a zinc matrix, and is the platform's highest-confidence single-element recovery asset. The gallium recovery ladder and the thioglycolate antimony process round out coverage of the export-controlled critical minerals that allied defense and government counterparties are increasingly asking strategic facilities like the SMC to address. Tying the individual assets together, the integrated critical-mineral recovery, battery recycling, and advanced packaging flowsheet platform is a system-level claim that describes exactly what the SMC physically is: a multi-stage, multi-element complex rather than a single column. It chains the recovery cascade, the magnet-recycling separation train, and the battery-recycling closed loop into a unified IP architecture. As Ucore moves from RapidSX demonstration to a financed, multi-line facility under offtake and government contracts, having a system-grade IP wrapper around the full flowsheet positions the SMC as defensible integrated infrastructure rather than a collection of individually licensed point solutions.

For Ucore's commercial and technical teams, the most-used surfaces in the Lattice Graph platform are the supply intelligence dashboards and the IP screening workflows. The conversion-route and deposit-query interfaces let the team build and iterate on the feedstock thesis interactively — running element-level supply-risk views, ranking deposit candidates by criticality tier and concentration risk, and exporting structured sourcing maps that go directly into offtake decks and government program briefings. The freedom-to-operate and patent-whitespace screening dashboard sits alongside this, letting chemists and process engineers vet candidate separation and recovery assets against the materials patent corpus before any licensing or design-around decision is made. At the chemistry level, the knowledge-graph explorer and composition-intelligence views give Ucore's technical staff a single place to trace a candidate sorbent or recovery route from composition through thermodynamic properties to the patents and published recipes in its neighborhood, with cross-source agreement scores and provenance signals attached at every step. In practice the team would run a recurring loop: use the mineral-deposit and supply-risk dashboards to scout and rank feed candidates, use the whitespace and freedom-to-operate screening to vet which separation and recovery assets to license or design around, and use the composition-intelligence and batch-screening tools to pressure-test feed-specific chemistries against the platform's negative-result atlas before committing them to RapidSX line trials. That keeps feed-qualification and IP diligence auditable, repeatable, and documented rather than ad hoc.

The most natural entry point is a scoped, paid discovery engagement lasting six to eight weeks, delivering two concrete outputs: a feedstock and supply-risk read for the SMC built from the supply intelligence and mineral-deposit layers, and a freedom-to-operate and whitespace screen of the candidate separation and recovery assets across the critical-mineral recovery and recycling separations portfolio. That engagement is designed to produce material Ucore can use in active offtake and financing conversations — a sourcing map with provenance, a supply-risk quantification, and a documented IP-position assessment — rather than a research report. At the end of the discovery engagement, the two follow-on tracks are a recurring data subscription to the supply intelligence and mineral-deposit products and a license or co-development agreement on the specific assets that prove most fit-for-purpose for the SMC, with field-of-use terms scoped to rare-earth separation and allied critical-mineral recovery. The data subscription is the natural recurring layer and is scoped to seat count and API call volume; a mid-five to low-six-figure annual range is a reasonable framing estimate for a team of this size and use pattern, to be confirmed in the paid pilot. Asset licensing terms, royalty structures, and field-of-use carve-outs are set asset by asset and would be established in the engagement based on which picks Ucore elects to build around. All commercial figures here are indicative framing only; no asset claims, predicted lab values, or commitments beyond what the platform's documented computational outputs and the portfolio actually support are implied or warranted.