Computational filler screening method integrated with thermal interface material manufacturing
A multi-potential phonon screening workflow that gates filler selection on cross-engine harmonic stability, then feeds directly into composite preparation and package assembly, creating a defensible materials-selection record.
The opportunity
Family L lead claim recast as method-of-manufacture: compute harmonic phonons under >=2 architecturally-distinct MLIPs, designate a controlling potential per chemistry class, select a filler only if the controlling engine reports min phonon freq >=-0.05 THz, then actually prepare the composite, deposit it as a TIM layer, and assemble the package; if unstable, iterate to an alternative before preparing. The 101-mitigation form vs the original method-of-evaluation (preserved for written-description only). Supports the whole portfolio's stability/candor substrate.
Investment thesis
This asset protects a computational materials-screening workflow that has been deliberately recast as a method of manufacture, embedding multi-engine machine-learning interatomic potential (MLIP) phonon validation directly into the sequence of steps by which a thermal interface material (TIM) composite is selected, prepared, deposited, and assembled into a semiconductor package. The core insight is structural: by requiring that a filler candidate clear a cross-engine phonon stability threshold before any physical preparation begins — and by tying the computational gate explicitly to downstream manufacturing acts — the claim resists abstract-idea invalidity challenges that would otherwise threaten a pure screening-method claim. The result is a process patent that describes something a fabrication line actually does, not merely something a computer calculates. Within the high-power thermal-interface materials portfolio, this is a foundational defensive asset rather than a standalone commercial product. Its primary job is to anchor the broader patent family to a concrete manufacturing transformation, providing a written-description substrate from which narrower product and composition claims can draw support. Every other asset in the portfolio that relies on cross-engine MLIP stability screening benefits from the disclosure made here — including the negative-control results and the cross-engine disagreement data that the portfolio commits to disclosing candidly rather than suppressing. That transparency posture is itself a differentiating legal and commercial feature: a buyer acquires not only the claim coverage but an evidentiary record demonstrating that the screening process was rigorous, reproducible, and architecturally diverse. The timing logic for this filing category is straightforward. Computational screening of inorganic fillers using modern MLIP architectures is an accelerating practice, and the window for establishing defensible process-claim coverage over multi-engine consensus workflows is open now, before the methodology becomes so widely published that prior-art density closes off differentiated claim scope. A strategic acquirer — most plausibly a Tier-1 semiconductor packaging house, a TIM formulator, or an electronic materials IP licensor — would value this asset as infrastructure that supports the entire portfolio's validity and enforceability, not merely as a standalone revenue line.
Asset rating
Specification
- stability threshold
- >=-0.05 (controlling-engine min phonon freq) THz
Computational validation
How this system was validated in silico — targeted molecular-dynamics and property simulations
Phonon-stability consensus applies to crystalline solids; this is a process-level claim, so it is validated through 4 targeted simulations of the candidate chemistry rather than lattice-dynamics screening.
Technical deep-dive
The method operates by enforcing a minimum phonon frequency threshold — specifically, the lowest-frequency phonon mode of a candidate filler material must be no worse than -0.05 THz as evaluated by a designated "controlling" machine-learning interatomic potential before any physical processing begins. The -0.05 THz tolerance is a calibrated engineering choice, not an arbitrary cutoff: it accommodates small numerical artifacts in harmonic force-constant calculations while meaningfully excluding structures with genuine soft modes that signal structural instability or phase tendency. The controlling potential is not fixed across all chemistries; the method specifies that the appropriate potential is designated per chemistry class, which reflects real practice — MACE, CHGNet, MatterSim, and ORB each show different accuracy profiles across oxide, nitride, and intermetallic chemical spaces, so no single potential is universally most reliable. What distinguishes this workflow computationally is the cross-engine architecture. The requirement for at least two architecturally distinct MLIPs is not cosmetic redundancy. Graph neural network potentials trained on different datasets and with different message-passing depths (MACE uses equivariant convolutions; CHGNet incorporates magnetic moment supervision; ORB uses a distinct graph transformer backbone; MatterSim targets multi-scale transferability) can and do produce conflicting phonon assessments for borderline candidates. The workflow's design explicitly handles disagreement: when the controlling engine and a secondary engine diverge, that disagreement is recorded and disclosed rather than resolved by selecting the more favorable answer. This is the "candor framework" dimension — the process generates a transparent evidentiary record that can be produced in prosecution, litigation, or a due-diligence review to demonstrate that no cherry-picking of favorable computational results occurred. The simulations supporting this method's real-world application include a live candidate-gate batch that processed at least 32 filler candidates using cloud-GPU MACE dispatch (completed in approximately 20 minutes), producing gating decisions that retained Ge2N2O and MgSnN2 as stable candidates while excluding CaHN and AlAsO4 as unstable. A five-criterion intersection discovery run and an evidence-graph integrity audit further validate that the screening pipeline produces internally consistent, reproducible candidate sets. These are not toy demonstrations — they represent the same operational pipeline that a TIM manufacturer would need to run during formulation development, which is precisely why the claims integrate computation and fabrication into a single method rather than treating them as separate steps. The key physical properties being screened for are phonon dynamic stability (no imaginary modes beyond the tolerance band), which is a necessary but not sufficient condition for a filler material to function reliably in a TIM composite under thermal cycling. Stable phonon structure correlates with resistance to thermal decomposition and phase transformation during package operation — critical reliability requirements for high-power density applications where junction temperatures routinely exceed 100 degrees C and thermal resistance budgets are measured in single-digit K/W. The method as claimed does not itself specify what thermal conductivity the selected fillers must achieve, but the portfolio context makes clear that the stability screen is the gate that precedes targeted simulation of dielectric, interface, and migration properties.
Market & opportunity sizing
This asset does not map to a standalone commercial product with a discrete TAM, and stating otherwise would misrepresent its nature. It is a defensive process patent whose value is portfolio-level: it supports the validity, enforceability, and breadth of every other asset in the high-power thermal-interface materials portfolio. The appropriate way to think about its market exposure is therefore as a fraction of the licensing value of the portfolio as a whole, rather than as an independent revenue stream. The addressable market context is the thermal interface materials industry, which services semiconductor packaging, power electronics, and high-density computing infrastructure. This is a multi-billion-dollar annual market driven by the density scaling of AI accelerators, power modules, and advanced packaging formats (chiplets, 3D-IC, co-packaged optics) that demand lower thermal resistance and greater reliability than conventional silicone-grease or phase-change TIM solutions. Major buyers of TIM IP include IDMs, OSAT providers, TIM formulators, and electronics materials suppliers who need freedom to operate and/or offensive claim coverage in filler screening and formulation. The licensing logic for a process patent of this type is typically one of two forms: a portfolio license that bundles this defensive substrate with the composition and product claims it supports, or a defensive cross-license in which the acquirer values the asset primarily for its ability to defend the rest of the portfolio against validity challenges. A royalty-per-unit model is conceivable but less natural given that the claim covers a manufacturing method rather than a product composition — more likely, this asset contributes to a lump-sum or running-royalty portfolio license where its presence increases the package's robustness and therefore the achievable license rate for the broader family.
Market & competitive position
transparent, defensible cross-engine candor framework; discloses disagreement as candor rather than relying on a single computational source
No incumbent TIM manufacturer currently publishes — let alone claims — a multi-engine MLIP screening protocol as a required gate in their manufacturing process. The dominant practice in filler development remains empirical: candidate materials are synthesized or procured, formulated into test composites, and characterized physically. Computational pre-screening, where it occurs, is typically done with a single DFT calculation or a single MLIP model, and the output is not formally integrated into a manufacturing method for patent purposes. This leaves a clear whitespace in which a multi-engine consensus requirement is novel and, as of this filing's effective date, not anticipated by published prior art in the TIM domain. The competitive differentiation is also methodological. Single-potential stability assessments are vulnerable to the criticism that the potential chosen was the one most likely to produce a favorable result. The multi-engine cross-disclosure architecture — requiring agreement between architecturally distinct potentials, recording disagreements, and advancing candidates only when the controlling potential clears the threshold — provides a more defensible scientific and legal posture. If a competitor attempts to design around this claim by using a single MLIP, they lose the cross-validation benefit and expose their composition claims to the charge that their stability assessments were insufficiently rigorous. The existence of this method patent in the portfolio thus creates a reputational and evidentiary moat around the portfolio's stability data, regardless of whether the method claim is directly asserted.
Claims & IP position
What's claimed, the protected family, and the freedom-to-operate read
The primary claim family (spanning Claims 37, 64, 80, 86, 130, 149, 150, 159, 175, 176, 177, 180, 189, 192, 194, 199, 208, and 67A) covers the integrated method in its method-of-manufacture form. The lead structure requires: (1) computing harmonic phonons for a filler candidate using at least two architecturally distinct machine-learning interatomic potentials; (2) designating a controlling potential specific to the candidate's chemistry class; (3) applying the -0.05 THz minimum phonon frequency gate using the controlling potential's output; (4) preparing a TIM composite incorporating the filler only if the gate is cleared; (5) depositing the composite as a TIM layer; and (6) assembling the resulting structure into a semiconductor package. If the gate is not cleared, the method requires iterating to an alternative filler before preparation — making the screening integral to, not merely advisory of, the manufacturing sequence. The deliberate architectural choice in this claim family is the integration of computation and physical manufacturing. The original form of this disclosure was a pure method-of-evaluation — describing the phonon screening steps without connecting them to downstream fabrication. That pure-evaluation form is preserved in the written description for disclosure credit, but it is explicitly not the preferred claim form, because a pure computational method claim faces heightened abstractness risk under Section 101 jurisprudence. By requiring that the computation gate an actual physical transformation — preparing a composite, depositing a layer, assembling a package — the method-of-manufacture recasting grounds the claim in patent-eligible subject matter. The claim family is broad enough to cover the controlling-potential designation as a per-chemistry-class decision, which captures the real-world engineering practice of matching MACE, CHGNet, ORB, or MatterSim to the chemistry where each is most accurate, rather than mandating a single universal potential.
- Claim type
- Method_of_use
- Drafted claims
- 18 claims
- Freedom to operate
- Defensive position
- Blocking patents
- None found — white space
| 1 | Claim 37 |
| 2 | Claim 64 |
| 3 | Claim 80 |
| 4 | Claim 86 |
| 5 | Claim 130 |
| 6 | Claim 149 |
| 7 | Claim 150 |
| 8 | Claim 159 |
| 9 | Claim 175 |
| 10 | Claim 176 |
| 11 | Claim 177 |
| 12 | Claim 180 |
| 13 | Claim 189 |
| 14 | Claim 192 |
| 15 | Claim 194 |
| 16 | Claim 199 |
| 17 | Claim 208 |
| 18 | Claim 67A |
integration with actual TIM prepare/deposit/assemble grounds the method in a manufacturing transformation (101 mitigation)
The freedom-to-operate posture for this asset is primarily defensive: the portfolio has screened over 300,000 materials patents, and the specific integration of multi-engine MLIP phonon gating into a TIM manufacturing method sequence does not appear in that corpus. The whitespace is credible because this claim structure — requiring computational consensus across architecturally distinct potentials as a manufacturing gate, with a specific frequency tolerance, tied to actual composite preparation and package assembly — is sufficiently novel in its combination that it is unlikely to be anticipated by any single prior-art reference. The 101-mitigation strategy does, however, create a specific claim boundary that defines where the coverage ends: a pure computational screening method that is not tied to any physical manufacturing act falls outside this claim family's scope by design (and is also the form most vulnerable to abstract-idea challenges). This means a competitor who runs multi-engine MLIP phonon screening purely as a research tool — without integrating it into a manufacturing flow for which they hold the liability — would not be within the literal scope of these claims. That is an acceptable tradeoff: the coverage is concentrated where it matters commercially, at the point where a manufacturing organization makes a filler-selection decision that is locked into a physical production run. A buyer should understand this boundary clearly and plan their prosecution strategy for any continuation claims accordingly if broader computational-method coverage is desired.
Validation roadmap
What's proven so far, and what a buyer would fund next
The computational validation supporting this method is operational and documented. A live candidate-gate batch processed 32 filler candidates through cloud-GPU MACE dispatch in approximately 20 minutes, producing clear gate decisions: Ge2N2O and MgSnN2 were retained as dynamically stable, while CaHN and AlAsO4 were excluded as failing the phonon stability threshold. A five-criterion intersection discovery run identified candidates that simultaneously satisfy multiple physics-based filters, and an evidence-graph integrity audit confirmed that the screening pipeline produces internally consistent results across successive runs. These demonstrations validate that the method as claimed is not merely a theoretical construct — it is an operational pipeline that has produced scientifically reproducible screening decisions at meaningful candidate throughput. The primary validation gate that remains open is a retrospective DFPT (density functional perturbation theory) benchmark comparison against the MLIP phonon results. DFPT provides the gold-standard harmonic phonon calculation from first principles, and completing this benchmark would establish the quantitative accuracy of the controlling-potential designation strategy — specifically, how reliably the per-chemistry-class controlling potential identifies true stability versus true instability relative to DFT ground truth. Until this benchmark is completed, the claim that the controlling-engine approach is sufficiently accurate to serve as a manufacturing gate rests on the internal consistency of the MLIP cross-checks rather than on validated agreement with DFT. This is a meaningful but surmountable gap: the benchmark is straightforward to execute for the specific filler candidates already in the dataset, and positive results would substantially strengthen both the technical disclosure and the prosecution record.
- Evidence receipts
- 11
Applications
Strategic fit & buyers
The most natural acquirers of this asset are organizations that hold or are building a manufacturing position in high-performance TIM formulation and who need defensible intellectual property covering the computational-to-physical screening workflow that increasingly governs how advanced fillers are selected. Tier-1 semiconductor packaging companies (including OSAT providers building TIM integration capabilities for AI accelerator and HPC packages), advanced materials companies with TIM product lines, and electronic materials IP licensing entities are the primary strategic fits. For any of these buyers, the asset's value is multiplicative: it strengthens the validity and enforceability of every composition or product claim in the portfolio by establishing a rigorous, documented, and cross-validated screening methodology as the basis for material selection. Secondary interest could come from IP assertion entities that specialize in computational materials and manufacturing process claims, and from litigation defense counsel representing TIM manufacturers who need a counter-assertion portfolio. The asset's candor-framework dimension — the explicit disclosure of cross-engine disagreements rather than selective reporting of favorable results — may also be of specific interest to organizations facing regulatory or contractual obligations around computational validation transparency, including defense contractors subject to materials qualification standards and medical device manufacturers subject to materials change-control requirements. In either case, the asset transfers cleanly as part of a portfolio transaction and would require very limited technical diligence beyond the existing computational records.
Risks & roadmap
The principal risk is claim scope under Section 101 if a court or examiner concludes that the computational steps remain the dominant character of the claim despite the manufacturing integration. The method-of-manufacture recasting is designed to address this risk, but it is not a guarantee: courts applying the Alice/Mayo framework have at times found that computationally-heavy manufacturing methods remain abstract when the physical steps are generic or routine. The mitigation is the specificity of the physical integration — not merely "prepare a composite" in the abstract, but preparing a TIM composite where the specific filler identity is gated by a defined phonon stability criterion applied by a per-chemistry-class controlling potential — which should distinguish the claim from purely abstract computational methods. A buyer should budget for a prosecution response to a 101 rejection and ensure continuation claims are drafted to allow flexibility in how the physical manufacturing steps are characterized. The secondary risk is the open DFPT benchmark validation. Until retrospective DFPT calculations confirm that the controlling-potential designation accurately predicts stability for the specific chemistry classes covered by the claims, there is a residual uncertainty about whether the method's screening decisions are reliably accurate or whether false positives (unstable candidates cleared by the MLIP) occur at a meaningful rate. A false-positive rate would not necessarily invalidate the claims, but it would complicate commercial adoption arguments in licensing negotiations. Completing the DFPT benchmark is the single highest-priority de-risking action for this asset, and it is straightforward to execute given that the candidate structures and DFT infrastructure already exist within the portfolio's computational pipeline.
More in Thermal-interface materials
Related assets in the same portfolio — each a separately filed position
License or acquire Computational filler screening method integrated with thermal interface material manufacturing
Request the full data room: complete claim set, proof packet, FTO memo, and licensing / acquisition terms.