Fluid-Cell Projection

• Fluid-cell projection is a numerical technique that separates a local update from a global correction to enforce constraints like incompressibility and mechanical equilibrium.
• It is applied across diverse domains such as incompressible flow, porous media, and fluid–structure interaction using elliptic or algebraic subproblems.
• The method enhances computational efficiency by enabling re-projection onto reduced-order or divergence-free spaces and facilitating accurate cut-cell reconstructions.

Searching arXiv for papers related to projection methods, fluid-cell/cut-cell projection, and fluid-structure interaction. Fluid-cell projection, understood here as an Editor’s term, denotes a family of numerical constructions in which a fluid or fluid-coupled field is first advanced in an unconstrained, local, or auxiliary form and is then projected, corrected, or reconstructed so that a global constraint, interface condition, or discretization-space compatibility condition is satisfied. In the cited literature, no single canonical definition is given; a plausible synthesis is that the term spans Chorin-type pressure correction for incompressibility, projection of trial stresses onto quasi-static equilibrium in solids, projection of coupled phase velocities in porous and multiphase media, divergence-free reduced spaces in immersed-boundary models, matrix-level reconstruction of cut-cell operators, and the projection of a physical flow field onto discrete space-time cells (Rycroft et al., 2014, Derr et al., 2022, Luo et al., 2021, Liu et al., 2018, Liu et al., 2017).

1. Algorithmic archetype

Across otherwise disparate formulations, the recurring architecture is a split between a local update and a global correction. In incompressible flow, one advances an intermediate velocity and then solves an elliptic problem for pressure so that the corrected velocity satisfies $\nabla\cdot \mathbf{v}=0$. In quasi-static elastoplasticity, one advances an intermediate stress and then solves an elliptic problem for velocity so that the corrected stress satisfies $\nabla\cdot\boldsymbol{\sigma}=\mathbf{0}$. In incompressible porous media, one advances provisional phase velocities and then projects the weighted mixture velocity onto the incompressibility constraint. In cut-cell finite elements, one integrates on integration cells and then projects the resulting matrices back into the cut-cell basis. In immersed-boundary reduction, one projects fluid variables onto a divergence-free structure-adapted reduced subspace (Rycroft et al., 2014, Derr et al., 2022, Liu et al., 2018, Luo et al., 2021).

Setting	Trial or local quantity	Projection or correction target
Incompressible flow	Intermediate velocity $\mathbf{v}^\ast$	Divergence-free velocity via pressure
Quasi-static elastoplasticity	Trial stress $\boldsymbol{\sigma}^\ast$	Equilibrium stress via velocity solve
Porous media flow	Provisional $v_s^\ast$, $v_f^\ast$	Divergence-free mixture velocity
Compressible two-fluid flow	Predicted $\alpha_k$, $\mathbf{u}_k$	Pressure-consistent mass/velocity state
Cut-cell FE integration	Integration-cell matrices	Cut-cell elemental matrices
IBM reduced models	Full fluid variables	Divergence-free reduced basis

The significance of this pattern is that the stiff global condition is isolated into an elliptic or algebraic subproblem. This is why projection methods recur in settings that otherwise differ radically in constitutive law, geometry handling, and discretization. What changes from one formulation to another is not the existence of projection, but the object being projected and the constraint manifold onto which it is projected.

2. Pressure correction and constraint enforcement in fluids and mixtures

The classical reference point is Chorin’s projection method: advance $\mathbf{v}^\ast$ without the pressure term, then impose incompressibility through

$$\mathbf{v}^{n+1}=\mathbf{v}^\ast-\frac{\Delta t}{\rho}\nabla p^{n+1},\qquad \nabla^2 p^{n+1}=\frac{\rho}{\Delta t}\nabla\cdot \mathbf{v}^\ast.$$

Several later formulations retain this structure while changing the continuum model and the meaning of the pressure variable (Rycroft et al., 2014).

For incompressible porous media with an incompressible fluid phase and an incompressible deformable solid phase, the pressure is a Lagrange multiplier enforcing

$\nabla\cdot\boldsymbol{\sigma}=\mathbf{0}$0

The projection step is written in terms of a pressure correction $\nabla\cdot\boldsymbol{\sigma}=\mathbf{0}$1 satisfying

$\nabla\cdot\boldsymbol{\sigma}=\mathbf{0}$2

with $\nabla\cdot\boldsymbol{\sigma}=\mathbf{0}$3. The method is fully Eulerian, uses a periodic Cartesian grid, stores $\nabla\cdot\boldsymbol{\sigma}=\mathbf{0}$4, $\nabla\cdot\boldsymbol{\sigma}=\mathbf{0}$5, and $\nabla\cdot\boldsymbol{\sigma}=\mathbf{0}$6 at cell centers, pressure at cell corners, and stress components on cell walls, employs second-order finite differences and second-order ENO for advection, advances explicit terms with Heun’s method and viscous stress terms with Crank–Nicolson, and demonstrates second-order convergence in space and time by a method of manufactured solutions. It is further coupled to large deformation through the reference map technique and demonstrated on phase separating neo-Hookean gels (Derr et al., 2022).

In the chemotaxis–Navier–Stokes setting, projection again separates the incompressibility constraint from the momentum solve. The fluid step computes an intermediate velocity $\nabla\cdot\boldsymbol{\sigma}=\mathbf{0}$7, the pressure is updated from

$\nabla\cdot\boldsymbol{\sigma}=\mathbf{0}$8

and the final velocity is

$\nabla\cdot\boldsymbol{\sigma}=\mathbf{0}$9

The fully discrete scheme uses backward Euler in time, a mixed finite element method in space with the Mini element for velocity-pressure, skew-symmetric trilinear forms satisfying $\mathbf{v}^\ast$0 and $\mathbf{v}^\ast$1, preserves the total mass of $\mathbf{v}^\ast$2, and admits the error estimate

$\mathbf{v}^\ast$3

The paper interprets this as first order in time and second order in space in the corresponding $\mathbf{v}^\ast$4 errors (Li, 7 Jun 2025).

For viscous compressible two-fluid flow, the projection structure is generalized to two transported phase masses

$\mathbf{v}^\ast$5

with one common pressure determined by pressure equilibrium and closure. The crucial differential relation is

$\mathbf{v}^\ast$6

The algorithm predicts masses, reconstructs intermediate $\mathbf{v}^\ast$7 and $\mathbf{v}^\ast$8, renormalizes intermediate pressure, solves intermediate velocities, performs a projection/correction step for pressure and mass, and finally renormalizes velocity. A positivity-preserving modification is introduced in the mass prediction step, and the semi-discrete analysis proves an unconditional discrete energy inequality. The numerical tests report first-order accuracy in time (Wu, 2023).

3. Projection beyond single-phase fluids

One of the most explicit generalizations of fluid-cell projection replaces the incompressibility constraint by quasi-static force balance in solids. The correspondence established for hypo-elastoplastic solids is

$\mathbf{v}^\ast$9

with pressure in the fluid role replaced by a velocity correction in the solid role. The solid algorithm computes a trial stress $\boldsymbol{\sigma}^\ast$0 from the constitutive update and then solves

$\boldsymbol{\sigma}^\ast$1

followed by

$\boldsymbol{\sigma}^\ast$2

The implementation is Eulerian finite difference on a fixed grid, the constitutive law is an elasto-viscoplastic bulk metallic glass model based on shear transformation zone theory, and the two-dimensional plane strain simple shear benchmark is reported to be in quantitative agreement with an explicit method. The paper also demonstrates extension to objects with evolving boundaries (Rycroft et al., 2014).

This solid analogue clarifies that projection is not intrinsically tied to fluid pressure. A plausible implication is that “fluid-cell projection” is best understood as an algorithmic architecture in which a constitutive or transport update is separated from a global constraint-enforcement solve. Under that interpretation, the incompressible velocity projection is only one member of a broader class.

4. Immersed boundaries, collocated grids, and fluid–structure interaction

In fluid–structure interaction, projection methods are often intertwined with boundary enforcement and pressure–velocity coupling on nonstandard grids. A hybrid Lagrangian/Eulerian collocated advection and projection method uses multiquadratic B-splines for velocity and multilinear basis functions for pressure, formulates the projection variationally rather than on a MAC staggered grid, and extends the weak formulation to cut cells for irregular domains with Dirichlet and free-surface boundary conditions. After eliminating velocity, the pressure/boundary-multiplier system is

$\boldsymbol{\sigma}^\ast$3

with a lumped mass matrix used to retain sparsity; the method also preserves a static pool of water exactly because hydrostatic pressure is representable in the multilinear pressure space (Gagniere et al., 2020).

A non-staggered Cartesian-grid phase-field/immersed-boundary solver advances the Cahn–Hilliard phase field, computes an intermediate velocity, applies an implicit immersed-boundary feedback correction, reconstructs face velocities with momentum-weighted interpolation, solves a pressure Poisson equation, and updates the final divergence-free velocity. The method is explicitly designed to be pressure-oscillation-free on collocated grids, with the correction aimed at suppressing odd-even decoupling and interface-pressure noise in large-density-ratio multiphase flow. The validation suite includes lid-driven cavity flow, droplet deformation, Rayleigh–Taylor instability, bubble coalescence, and a rising bubble bypassing an obstacle; the reported final droplet size error is 1.14%, compared with 8.42% for a referenced original phase-field method (Wang et al., 2024).

A target-fixed immersed-boundary projection method reformulates rigid-body FSI in a body-fixed frame to maximize the accuracy of surface stresses on the target body. The incompressible vorticity equations and Newton’s equations are coupled implicitly, fictitious fluid inside the rigid body is handled through effective mass and inertia,

$\boldsymbol{\sigma}^\ast$4

and spurious oscillations in surface stresses are filtered through the diagonal operator $\boldsymbol{\sigma}^\ast$5. The coupled discrete system is solved efficiently by block-LU decomposition and validated on a neutrally buoyant cylinder in planar Couette flow and a freely falling or rising cylindrical rigid body (Lin et al., 2020).

These formulations show that, in FSI, projection is frequently asked to do more than enforce $\boldsymbol{\sigma}^\ast$6: it may also stabilize body coupling, recover compact pressure–velocity coupling on a collocated grid, or enforce no-slip on immersed surfaces.

5. Projection as basis restriction and cut-cell reconstruction

A second major meaning of fluid-cell projection is basis-level or matrix-level projection rather than direct pressure correction. In reduced-order modeling for the immersed boundary method, pressure is eliminated by the discrete projection

$\boldsymbol{\sigma}^\ast$7

and the reduced trial and test spaces are chosen as

$\boldsymbol{\sigma}^\ast$8

This construction guarantees $\boldsymbol{\sigma}^\ast$9 for the reduced velocity approximation $v_s^\ast$0, preserves Lyapunov stability, and reduces online coefficient assembly through interpolation of time-dependent matrix entries. The reported speedups range from $v_s^\ast$1 to $v_s^\ast$2 for the oscillating elliptical membrane, $v_s^\ast$3 to $v_s^\ast$4 for rotation of an elliptical rigid particle in shear flow, and $v_s^\ast$5 to $v_s^\ast$6 for motion of two particles in laminar flow (Luo et al., 2021).

In unfitted finite elements, projection can occur entirely at the matrix level. Projection-based Gaussian quadrature subdivides a cut cell into integration cells, integrates each integration cell with standard Gaussian quadrature, and reconstructs the cut-cell matrices by

$v_s^\ast$7

with

$v_s^\ast$8

After summation over integration cells,

$v_s^\ast$9

The paper emphasizes that the FE formulation and quadrature rule are unchanged, that the method is consistent with the variational principle, and that it can be interpreted as a re-projection of residuals or a reduced-order modeling technique (Liu et al., 2018).

A related reduced-basis viewpoint appears in a semi-implicit partitioned reduced basis method for incompressible-fluid/linear-elastic FSI in an ALE framework. The fluid solver uses a Chorin–Temam pressure Poisson projection, the fluid velocity is homogenized by

$v_f^\ast$0

and POD–Galerkin projection is applied separately to fluid velocity, pressure, and solid displacement, with fluid mesh displacement recovered by harmonic extension of the solid modes. A Robin interface condition is introduced to improve stability of the partitioned coupling (Nonino et al., 2022).

Taken together, these works show that projection need not mean “solve a Poisson equation for pressure.” It may instead mean restriction to a divergence-free subspace, reconstruction of cut-cell operators in an original FE basis, or Galerkin projection of partitioned FSI subproblems onto reduced spaces.

6. Discrete-cell interpretation, scope, and related notions

A broader interpretation of fluid-cell projection appears in multiscale CFD, where the numerical flow field is described as the projection of the physical flow field onto discrete space and time. The governing quantities are the physical Knudsen number

$v_f^\ast$1

the cell Knudsen number

$v_f^\ast$2

and the numerical Knudsen number

$v_f^\ast$3

with the heuristic relation

$v_f^\ast$4

Under this viewpoint, mesh refinement is not merely convergence toward a fixed PDE solution; it may change the effective numerical regime itself, shifting the represented physics from continuum to near-continuum or non-equilibrium behavior (Liu et al., 2017).

This perspective is useful for delimiting the scope of the term. Not every model that combines “fluid” and “cell” is a projection method. The cell fluid model for phase transition divides the system volume into congruent cubic cells, develops an exact grand partition function in collective variables, and derives an analytic equation of state for a Morse-fluid-like system, but it is not a projection method in the Chorin or matrix-reconstruction sense (Kozlovskii et al., 2016). Likewise, the two-fluid continuum theory for lumen nucleation in cell assemblies couples cell stress, interstitial fluid flow, ion currents, fluid pumping, and active flexoelectricity, derives effective pressures and apparent surface tensions such as

$v_f^\ast$5

and predicts nucleation criteria and long-time states, but it does not formulate the dynamics as a projection algorithm (Duclut et al., 2019).

A common misconception is therefore to treat fluid-cell projection as a single standardized method. The literature instead supports a more precise taxonomy. In one class, projection enforces a differential constraint such as incompressibility or quasi-static equilibrium. In a second class, projection transfers dynamics to a reduced or divergence-free basis. In a third class, projection reconstructs cut-cell operators from subcell information. In a fourth class, projection denotes the resolution-dependent mapping from physical flow to numerical cell averages. The common denominator is not a single equation, but a strategy of separating a provisional representation from a globally admissible one.