Skip to content

README.md

Latest commit

 

History

History
84 lines (56 loc) · 5.1 KB

README.md

File metadata and controls

84 lines (56 loc) · 5.1 KB

CIF Reference Implementation

License: MIT

DOI

Open-source Python pipeline for designing Controlled Incremental Filtration (CIF) microfluidic devices, with COMSOL 3-D Stokes validation of the Dinh 2024 V1 and V3 outlet geometries.

Repository contents

.
├── CIF_Generator_v3_20.ipynb       Main Jupyter notebook
├── COMSOL/                         3-D Stokes outlet-cell artifacts
│   ├── v1 5gap coarse.docx            COMSOL report — Dinh V1, coarse mesh
│   ├── v1 5gap extra coarse.docx      COMSOL report — Dinh V1, extra-coarse mesh
│   ├── v3_5gap_coarse.docx            COMSOL report — Dinh V3, coarse mesh
│   ├── v3_5gap_extra_coarse.docx      COMSOL report — Dinh V3, extra-coarse mesh
│   └── *.mph, *.zip                   Source model files + archives
├── outputs/                        Generated figures + provenance_log.md
├── LICENSE
└── README.md

The COMSOL reports tabulate Q-fluxes and pressure samples from V1/V3 × coarse/extra-coarse mesh runs; Block 19 of the notebook transcribes those tables for post-processing. The .mph files reproduce the simulations from scratch in COMSOL Multiphysics 6.3 (CFD Module).

Headline result

Under the 1-D pressure-equilibration assumption and the inlet condition $Q_s(0) = 0$, the volumetric filtrate fraction at a CIF outlet is given exactly by

$$ F ;=; \frac{2,R_c}{R_s + 2,R_c} $$

where $R_c$ and $R_s$ are the per-pitch hydraulic resistances of the center and side channels at the outlet.

  • Closed-form vs. iterative recursion. Matches the Gifford recursion across five reference devices in three architectural regimes (Gifford 2014 bead; Dinh 2024 V1/V2/V3; Xia 2016 RS-7) to $|\Delta F| \leq 2.35 \times 10^{-4}$.
  • 3-D COMSOL Stokes at the Dinh outlets (Block 19, computed from the transcribed Q-flux tables): matches published measurements within $1\sigma$:
    • V1: $F = 0.5070$ vs. paper $0.505 \pm 0.033$ $(+0.06\sigma)$
    • V3: $F = 0.6281$ vs. paper $0.649 \pm 0.036$ $(-0.58\sigma)$

Quick start

Tested with Python 3.8.20, NumPy 1.24.4, SciPy 1.10+, Matplotlib 3.7.5, plus ezdxf (DXF photomask export), sympy (closed-form identity proof), and jupyter.

git clone https://github.com/Austin-Routt/cif-reference-implementation.git
cd cif-reference-implementation
pip install numpy scipy matplotlib ezdxf sympy jupyter
jupyter notebook CIF_Generator_v3_20.ipynb

Run the cells top-to-bottom. End-to-end execution produces the Gifford 2014 photomask (gifford_2014_bead_device.dxf) and the diagnostic figures already mirrored in outputs/. COMSOL is only needed if you want to re-run the .mph studies; the notebook itself reads the transcribed tables, not a live COMSOL solution.

Validation

Validators are implemented as inline notebook checks (helper-function calls that print per-block pass/fail summaries — not a separate pytest suite). After a full top-to-bottom run, the §8.8 epilogue ("v3.20 POST-PRIORITY-B SUPPLEMENTARY AUDIT") reports the final state:

  • 218 / 218 core validators across 19 sub-blocks (geometry, manufacturability, PE convergence, numerical stability, dimensional consistency, mask integrity, COMSOL post-processing).
  • 19 / 19 supplementary validators (Priority-B fabrication-tolerance sweep, Xia RS-7 closed-form regression, channel-depth sensitivity, PE half-life cross-validation, effective $R_c / R_s$ from COMSOL).
  • 19 regression stamps (15 core + 4 supplementary). 13 are bit-identical across environments; 2 (Blocks 12 and 15) hash scipy.sparse.linalg.spsolve output and reproduce the v3.1 baseline to 12+ decimal places under documented BLAS/LAPACK portability.

An earlier audit cell (§8.5) prints an interim 9 supplementary / 16 stamp count from before the Priority-B blocks ran; the 19 / 19 / 19 figures in the §8.8 epilogue are the final state.

Scope

The closed-form identity addresses the volumetric throughput (filtrate-fraction) design question only. The size-cutoff question is still handled by the iterative $f_{\text{gap}}$ calibration described in the source papers. The mask-layout (DXF) module currently produces the Gifford 2014 photomask only; generalized DXF for the remaining reference devices is a documented roadmap item.

Source papers

Citation

The companion manuscript is in preparation. Until then, please cite this repository as:

Routt, A. H. (2026). CIF Reference Implementation v3.20.
https://github.com/Austin-Routt/cif-reference-implementation

License

MIT — see LICENSE.