A multiscale treatment of the moving-contact-line singularity. The Generalized
Lubrication Equation (GLE) resolves the interface from the Navier-slip scale
(
The steady dip-coating branch reproduced through its entrainment fold: the
bifurcation diagram
Steady contact-line height
The legacy effective slip was calibrated to the digitised fold at its reported
resolution. The Chan--Scott--Hocking and direct Luo--Gao slips were then fitted
to that legacy branch's computed fold,
The separate equal-input comparison fixes
The full methods document — GLE derivation, the two solvers, the calibration, and the component-by-component validation — is docs/gle-theory-and-implementation.pdf (source docs/gle-theory-and-implementation.tex).
State vector
Here
The mobility
$$ M(\theta,\mu_r) = \frac{2\sin^3\theta,\bigl[\mu_r^2 f_1(\theta)
- 2\mu_r f_3(\theta) + f_1(\pi-\theta)\bigr]} {3,\bigl[\mu_r f_1(\theta) f_2(\pi-\theta) ;\boldsymbol{+}; f_1(\pi-\theta) f_2(\theta)\bigr]}, $$
with f: in Basilisk that name is the VoF volume
fraction (issue #4).
The Chan cutoff is represented by the Cox matching constant
The numerical authorities for gle-ode/reference-generator. One solves the
Scott--Hocking
The dependency-free runtime consumes that frozen table and records which
closure supplied each result. Its automatic order is: the Scott--Hocking
The direct Luo--Gao model is not the Chan equation with an explicit value of
Use Chan with Scott--Hocking, the right-angle branch, or the numerical table
when reference inner-Stokes matching covers the case. Use direct Luo--Gao when
an explicit local closure is preferred or no table value is available, while
retaining its approximation status and, at finite
At the 270 certified table nodes,
Two complementary solvers share the same right-hand side; both are header-only C99 with no external dependencies.
Shooting (src-local/gle-shoot.h) — adaptive
embedded Cash–Karp RK5(4) integration
(src-local/gle-integrate.h) with a
static-meniscus-manifold outer boundary condition, reducing the BVP to a single
scalar equation in the contact-line curvature
Collocation (src-local/gle-collocate.h) — a
fixed hybrid-mesh implicit-midpoint discretisation with a bordered banded-LU
Newton solve, marching the meniscus rise solve_bvp inside a continuation loop could not do — the failure mode
of the historical Python continuation (issues #14 and #16). The older
shooting-based continuation is kept in
src-local/gle-continuation.h for provenance and
lower-branch scans only; it is not safe through the fold.
Verified in the solver stack: the Cash–Karp tableau against its published
coefficients; the banded-LU border solve against a dense reference to
Cross-validated against the historical Python solve_bvp reference
(python/GLE_solver.py), which solves the same BVP by
SciPy collocation. Driver:
postProcess/compare-c-python.py; figure:
img/c-vs-python.png.
| Check | Result |
|---|---|
| Huh–Scriven mobility, C vs Python | machine precision |
| Inner-Stokes |
270 converged nodes; max node sensitivity |
| Frozen |
234 independent cell centres; max $ |
| Corrected right-angle cutoff | one-phase analytic anchor and viscosity-inversion symmetry |
| Interface angle |
max rel. error |
| Interface thickness |
max rel. error |
Apparent-angle difference at --quick) |
|
| Gravity-rescaling invariance ( |
|
| Static ( |
finite-$s_0$ value |
| Fig. 4b external theory curve | max $ |
The Fig. 4b study separates calibration from comparison. The legacy effective
The far-field angle is a soft condition (its gain runs to
make # build the standalone C solvers (any C99 compiler + libm)
make test # regressions; runs the bounded Basilisk seam test when qcc is present
./reproduce-fig4b.sh # trace the branch and regenerate img/fig4b-reproduction.pngSingle solve at one capillary number:
cd gle-ode && ./gle-solve fig4b.params Ca=5e-3Select the case-level Chan closure or the direct Luo--Gao model at runtime:
cd gle-ode
./gle-solve fig4b.params c_method=auto
./gle-solve fig4b.params gle_model=luo_gaoEvaluate the Chan matching constant for a prescribed microscopic angle and viscosity ratio without solving a GLE profile:
cd gle-ode
./gle-cutoff theta_mic_deg=60 mu_r=0.1 c_method=autoThe output names the resolved authority (scott_hocking,
corrected_right_angle, reference_table, or the explicitly approximate
luo_gao_approx) together with luo_gao_approximation flag says only whether that fallback supplied the
value; the named method carries the reference or interpolation provenance.
Trace the bifurcation branch through the fold:
cd gle-ode && ./gle-continuation fig4b.params branch_out=output/branch.csvThe GLE-coupled DNS case (serial Basilisk):
cd simulationCases
qcc -O2 -disable-dimensions -I../src-local contactline-gle.c -o run -lmRequirements: any C99 compiler and libm for the solvers — no GSL, no
SUNDIALS. Basilisk qcc is needed only for the DNS case. uv/Python is
confined to development-time plots, cross-validation, and the open-source inner-Stokes
reference generator; none enters the production C solver.
Contact-line-subgrid-modeling
├── src-local - header-only C99 solver stack (no external dependencies)
│ ├── gle-model.h - common parameters, corrected Huh–Scriven mobility and model dispatcher
│ ├── gle-model-chan.h - Chan constant-cutoff GLE
│ ├── gle-model-luo-gao.h - direct Luo–Gao slippery-wedge GLE
│ ├── gle-slip-closure.h - named case-level Q and c closure policy
│ ├── gle-slip-reference.h - Scott–Hocking and frozen-table runtime data
│ ├── gle-slip-table-data.h - generated dependency-free Q table
│ ├── gle-integrate.h - adaptive Cash–Karp RK5(4) integrator with thickness-event stopping
│ ├── gle-shoot.h - single-shooting BVP solve on the contact-line curvature
│ ├── gle-collocate.h - fixed-mesh implicit-midpoint collocation; fold-free branch tracer
│ ├── gle-continuation.h - legacy shooting-based arclength continuation (lower branch only)
│ ├── gle-basilisk.h - the GLE ↔ DNS coupling seam
│ └── gle-params.h - key=value runtime parameter loader with CLI overrides
├── gle-ode - standalone drivers and parameter files
│ ├── gle-cutoff.c - evaluates c(theta_e,M) without solving a GLE profile
│ ├── gle-solve.c - single GLE solve, writes the interface profile
│ ├── gle-continuation.c - traces the dip-coating branch through the fold
│ ├── reference-generator - open-source two-phase Stokes-wedge Q generator
│ ├── fig4b.params - calibrated parameters reproducing the Fig. 4b theory curve
│ └── Makefile - builds all three drivers with cc -lm
├── simulationCases - Basilisk two-phase DNS cases
│ ├── contactline-gle.c - plate coating with the per-timestep GLE subgrid coupling
│ └── contactline.c - baseline case with a fixed contact angle (no GLE coupling)
├── postProcess - uv-runnable plotting and validation
│ ├── plot-fig4b.py - overlays the traced branch on the digitized reference
│ ├── plot-finite-m-closure-comparison.py - compares finite-M FEM and Luo–Gao matching
│ └── compare-c-python.py - cross-validates the C solver against the Python reference
├── python - historical reference implementation and linearized fixture
│ ├── GLE_solver.py - SciPy solve_bvp reference (shares the C physics exactly)
│ ├── validate-linearized-gle.py - checks the linearized GLE against Kansal et al. (2024)
│ ├── reference-data/kansal-minkush-linearized.csv - digitized linearized-GLE reference
│ └── requirements.txt - Python dependencies for the reference scripts
├── data/fig4b-digitized - vector-digitized reference data (do not modify)
│ ├── CALIBRATION.md - axis calibration and extraction provenance
│ ├── theory_curve.csv - the review's multiscale-lubrication theory curve
│ └── symbols_*.csv - the five experimental series of Delon et al. (2008)
├── img - generated figures committed for the README and docs
│ ├── fig4b-reproduction.png - the headline reproduction figure (also .pdf)
│ ├── fig4b-model-comparison.png - equal-slip Chan/Luo–Gao comparison
│ ├── finite-m-closure-comparison.png - finite-M FEM/Luo–Gao matching comparison
│ └── c-vs-python.png - the C-vs-Python cross-validation figure
├── docs - LaTeX methods document
│ ├── gle-theory-and-implementation.tex - theory, numerics, and validation write-up
│ └── gle-theory-and-implementation.pdf - the compiled methods document
├── _Archive - superseded code kept for provenance (do not resurrect)
├── .github - documentation pipeline and website generator
├── Makefile - top-level build delegating to gle-ode
├── reproduce-fig4b.sh - one-shot branch trace + figure regeneration
└── LICENSE
The coupling seam is implemented and runtime-verified.
simulationCases/contactline-gle.c runs a
per-step event: it samples the DNS interface curvature at the grid scale,
converts its sign to the GLE orientation, solves the subgrid GLE at the actual
local cell size, and refreshes the associated height-function contact angle.
The signed capillary number uses the plate speed relative to the measured
contact-line speed, while a dumped scalar preserves the dynamic angle across
restart. The remaining production qualification is a grid-convergence study
of the coupled curvature and contact-line-position samples.
- Snoeijer, J. H. (2006). Free-surface flows with large slopes: beyond lubrication theory. Phys. Fluids 18, 021701. doi:10.1063/1.2171190
- Snoeijer, J. H. & Andreotti, B. (2013). Moving contact lines: scales, regimes, and dynamical transitions. Annu. Rev. Fluid Mech. 45, 269–292. doi:10.1146/annurev-fluid-011212-140734
- Delon, G., Fermigier, M., Snoeijer, J. H. & Andreotti, B. (2008). Relaxation of a dewetting contact line. J. Fluid Mech. 604 (source of the experimental series in Fig. 4b).
- Huh, C. & Scriven, L. E. (1971). Hydrodynamic model of steady movement of a solid/liquid/fluid contact line. J. Colloid Interface Sci. 35, 85–101 (the wedge-flow mobility).
- Chan, T. S., Snoeijer, J. H. & Eggers, J. (2012). Theory of the forced wetting transition. Phys. Fluids 24, 072104 (two-fluid mobility).
- Chan, T. S., Kamal, C., Snoeijer, J. H., Sprittles, J. E. & Eggers, J. (2020). Cox--Voinov theory with slip. J. Fluid Mech. 900, A8. doi:10.1017/jfm.2020.499
- Hocking, L. M. (1977). A moving fluid interface. Part 2. The removal of the force singularity by a slip flow. J. Fluid Mech. 79, 209--229. doi:10.1017/S0022112077000123
- Scott, J. F. (2020). Calculation of a key function in the asymptotic description of moving contact lines. Q. J. Mech. Appl. Math. 73, 279--291. doi:10.1093/qjmam/hbaa012
- Kansal, M. et al. (2024). Eur. Phys. J. Spec. Top. doi:10.1140/epjs/s11734-024-01443-5 (linearized-GLE reference data).
- Luo, J. & Gao, P. (2025). Explicit theory of moving contact lines. J. Fluid Mech. 1019, A52. doi:10.1017/jfm.2025.10587
- Afkhami, S., Zaleski, S. & Bussmann, M. (2009). A mesh-dependent model for applying dynamic contact angles to VOF simulations. J. Comput. Phys. 228, 5370–5389 doi:10.1016/j.jcp.2009.04.027.
Vatsal Sanjay · CoMPhy Lab, Department of Physics, Durham University ·
vatsal.sanjay@comphy-lab.org


