Industrialization · The invention engine Planned

From industrial problem to chemical asset.

Discovery programs are designed to run steps 01–05 on a partner’s problem. Avestiq-owned candidates are intended to continue through steps 06–08. Results are designed to return to the system.

The nine-step invention engineA loop of nine steps: define, design, qualify, synthesize, validate, engineer, protect, commercialize and learn, which feeds results back to define.

    The objective is not a predicted molecule.
    It is a deployable chemical technology.

    FTO: freedom to operate. Steps 06–08 are designed to be capital-light: pilot work at external facilities and toll manufacturing, with in-house capacity only where strategic economics justify it.

    Process consequences Illustrative schematic

    A molecular property becomes a process cost.

    In a solvent loop, every property of the ionic liquid shows up somewhere in the plant: selectivity in the column, viscosity in the pumps and mass transfer, thermal stability in the regenerator, losses in the make-up stream. LILM’s process layer is designed to carry those consequences back into ranking.

    Extraction loop for an aromatic / aliphatic separation

    Illustrative · not an Avestiq process
    Illustrative ionic-liquid extraction loopA toluene and n-heptane feed enters the bottom of an extraction column while lean ionic liquid enters the top. Heptane-rich raffinate leaves the top. The toluene-loaded extract leaves the bottom, is heated and enters a regenerator, where toluene is removed overhead, condensed and recovered. Lean ionic liquid is pumped, cooled and recycled to the column, with a make-up stream and a purge for water and impurities. Labels mark where selectivity, capacity, viscosity, thermal stability, losses, toxicity and cost enter the design.
      Generic configuration of the kind studied for ionic-liquid aromatic extraction, drawn to show where properties enter process design. It is not a design, a mass balance or an Avestiq process; the 2013 design case on the Science page addressed the same toluene/n-heptane separation.

      Solvent-loop model Illustrative model

      Turn molecular properties into operating numbers.

      A one-page model of the loop above, per tonne of solute recovered. Move a property and follow it through the mass balance, the regeneration duty and an operating-cost index. Every input is an assumption or a cited value, every equation is shown, and nothing here describes an Avestiq process or a measured cost.

      Inputs

      Equationsmake-up m = S(1 − R) · regeneration duty Q = S·cp·ΔT·(1 − ε) + h · minimum equilibrium stages Nmin = ln[(x/(1 − x))2] / ln β (Fenske, total reflux, constant β, purities x) · pressure drop ΔP = ΔP0·μ/μ0 (laminar, fixed line size and velocity) · pumping E = (1000·S/ρ)·ΔP/ηp · equipment burden ∝ Nmin·S·(μ/μ0)1/2 at fixed contact time (Stokes–Einstein D ∝ 1/μ with penetration-theory k ∝ D1/2) · operating index = 1000·m·price + Q·heat price + E·electricity price

      R: solvent recovered per pass. S: tonnes of solvent circulated per tonne of solute. The solute is taken as toluene for h; cp comes from the CO2-screen candidates. The operating index leaves out capital, labor and solvent degradation; the burden index is relative to β = 8, S = 6, 30 mPa s. Measured gas diffusivities in ionic liquids fall more weakly than 1/μ (Morgan, Ferguson & Scovazzo, Ind. Eng. Chem. Res. 2005, 44, 4815–4823, doi:10.1021/ie048825v), so the burden index overstates the viscosity penalty.

      Process flow, per tonne of solute

      Stream values update with the inputs

      Illustrative model

        Where the regenerator outflow goes

        Illustrative model
        Band widths are proportional to mass, with a minimum width so that small losses stay visible.

        Operating-cost index

        Illustrative model
        Per tonne of solute. An index for comparing inputs, not a cost estimate.

        Which input moves the index most?

        One input at a time over the stated range, all others held at the current settings

        Illustrative model

        Fixed assumptions

        Fixed assumptions of the solvent-loop model

        Scale-up exposure Illustrative · editorial assessment

        Some risks cannot be seen until the scale is large enough.

        Where each class of risk becomes measurable and decisive on the path from grams to a commercial plant. This is why a prediction that wins in silico still has to earn its place at each scale.

        Risk exposure by scale

        Select a cell for the reasoning

        Illustrative

          Levels are Avestiq’s editorial reading of where these risks usually become decisive in chemical scale-up; they are not statistics. No Avestiq candidate has passed through any of these stages. Regulatory requirements depend on jurisdiction, volume and use, and are summarized here only to show when they apply.

          Molecule and process IP Planned stage gates

          A candidate advances only on measured evidence.

          Avestiq-owned candidates are intended to move through explicit gates. Each gate can stop a candidate, and a stopped candidate still returns data to the model.

          From design to license

          Planned · schematic
          Planned stage gates for Avestiq-owned candidatesEight stages in sequence: design and rank, freedom-to-operate review, synthesis, experimental proof, benchmark against the incumbent, toxicity, life-cycle and techno-economic evaluation, and patent protection, followed by three routes: license, co-develop or spin out. A gate between stages can stop a candidate; stopped candidates send their data back to the model.
          Describes intended practice, not program status. Spin-outs or joint ventures are considered only where a specific asset justifies one.

          Defensibility Planned · none of these layers exists yet as an Avestiq asset

          The moat moves from algorithm to asset.

          Each layer outward is harder to reproduce. Data rights are to be set by contract: customer-private data are to be kept isolated and to train shared models only where an agreement grants that right.

          Seven concentric layers of defensibilitySeven concentric rings from the innermost model layer to the outermost customer-qualification layer.

            An algorithm can be reproduced.
            A qualified chemical technology embedded in a customer process is substantially harder to reproduce.

            Working with Avestiq

            Work with the engine—or with the chemistry it creates.

            Engagement models Avestiq intends to offer. Terms are set case by case.

            Discuss an industrial problem

            Privacy

            Terms