Skip to content

Capacitance Matrix Method

Extends fast spectral solvers to irregular / masked domains using the Sherman-Morrison capacitance correction (Buzbee, Golub & Nielson 1970).

Solver

finitevolx.CapacitanceSolver

Bases: Module

Spectral Poisson/Helmholtz solver for masked irregular domains.

Uses the capacitance matrix method (Buzbee, Golub & Nielson 1970) to extend a fast rectangular spectral solver to a domain defined by a binary mask.

The algorithm (Buzbee, Golub & Nielson 1970):

  1. Solve the PDE on the full rectangle using a fast spectral solver (DST/DCT/FFT), ignoring the mask. Call this u.
  2. u generally violates ψ = 0 at inner-boundary points. Correct it: ψ = u − Σ_k α_k g_k, where g_k are precomputed Green's functions (rectangular-domain response to δ-sources at each boundary point b_k).
  3. The coefficients α are found by requiring ψ(b_k) = 0 at all N_b boundary points, giving the linear system C α = u[B] where C[k,l] = g_l(b_k) is the capacitance matrix.

The capacitance correction is delegated to :class:gaussx.CapacitanceSolver; this class adds the field reshaping and the exterior masking. Construct with :func:build_capacitance_solver.

Attributes:

Name Type Description
solver CapacitanceSolver

The generic capacitance solver operating on flat vectors.
mask Float[Array, 'Ny Nx']

Domain mask (1.0 = interior, 0.0 = exterior). Applied to the output so that values outside the physical domain are exactly zero.
shape tuple[int, int]

Grid shape (Ny, Nx).
dx float

Grid spacing in x.
dy float

Grid spacing in y.
lambda_ float

Helmholtz parameter.
base_bc str

Spectral solver used as the rectangular base.
Source code in .venv/lib/python3.12/site-packages/spectraldiffx/_src/fourier/capacitance.py 
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
class CapacitanceSolver(eqx.Module):
    """Spectral Poisson/Helmholtz solver for masked irregular domains.

    Uses the **capacitance matrix method** (Buzbee, Golub & Nielson 1970) to
    extend a fast rectangular spectral solver to a domain defined by a binary
    mask.

    The algorithm (Buzbee, Golub & Nielson 1970):

    1. Solve the PDE on the **full rectangle** using a fast spectral solver
       (DST/DCT/FFT), ignoring the mask.  Call this ``u``.
    2. ``u`` generally violates ψ = 0 at inner-boundary points.  Correct it:
       ``ψ = u − Σ_k α_k g_k``, where ``g_k`` are precomputed Green's
       functions (rectangular-domain response to δ-sources at each
       boundary point b_k).
    3. The coefficients α are found by requiring ψ(b_k) = 0 at all
       N_b boundary points, giving the linear system ``C α = u[B]``
       where ``C[k,l] = g_l(b_k)`` is the **capacitance matrix**.

    The capacitance correction is delegated to :class:`gaussx.CapacitanceSolver`;
    this class adds the field reshaping and the exterior masking. Construct with
    :func:`build_capacitance_solver`.

    Attributes
    ----------
    solver : gaussx.CapacitanceSolver
        The generic capacitance solver operating on flat vectors.
    mask : Float[Array, "Ny Nx"]
        Domain mask (1.0 = interior, 0.0 = exterior).  Applied to the output
        so that values outside the physical domain are exactly zero.
    shape : tuple[int, int]
        Grid shape ``(Ny, Nx)``.
    dx : float
        Grid spacing in x.
    dy : float
        Grid spacing in y.
    lambda_ : float
        Helmholtz parameter.
    base_bc : str
        Spectral solver used as the rectangular base.
    """

    solver: _GaussxCapacitanceSolver
    mask: Float[Array, "Ny Nx"]
    shape: tuple[int, int] = eqx.field(static=True)
    dx: float
    dy: float
    lambda_: float = eqx.field(static=True)
    base_bc: str = eqx.field(static=True)

    def __call__(
        self,
        rhs: Float[Array, "Ny Nx"],
    ) -> Float[Array, "Ny Nx"]:
        """Solve (∇² − λ)ψ = rhs on the masked domain.

        The generic solver enforces ψ = 0 at the inner-boundary points and
        returns the corrected field; this method reshapes between fields and
        flat vectors and zeroes the exterior via the mask.

        Parameters
        ----------
        rhs : Float[Array, "Ny Nx"]
            Right-hand side on the full rectangular grid.
            Values outside the physical domain (mask = False) are ignored.

        Returns
        -------
        Float[Array, "Ny Nx"]
            Solution ψ on the full rectangular grid.  Exactly zero outside
            the mask, and ψ ≈ 0 at inner-boundary points.
        """
        ny, nx = self.shape
        # Honor the documented contract: exterior (mask = False) values are
        # ignored by zeroing them before the base solve.
        rhs_masked = rhs * self.mask
        psi_flat = self.solver(rhs_masked.reshape(ny * nx))
        return psi_flat.reshape(ny, nx) * self.mask

__call__(rhs)

Solve (∇² − λ)ψ = rhs on the masked domain.

The generic solver enforces ψ = 0 at the inner-boundary points and returns the corrected field; this method reshapes between fields and flat vectors and zeroes the exterior via the mask.

Parameters:

Name Type Description Default
rhs Float[Array, 'Ny Nx']

Right-hand side on the full rectangular grid. Values outside the physical domain (mask = False) are ignored.

 required

Returns:

Type Description
Float[Array, 'Ny Nx']

Solution ψ on the full rectangular grid. Exactly zero outside the mask, and ψ ≈ 0 at inner-boundary points.
Source code in .venv/lib/python3.12/site-packages/spectraldiffx/_src/fourier/capacitance.py 
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
def __call__(
    self,
    rhs: Float[Array, "Ny Nx"],
) -> Float[Array, "Ny Nx"]:
    """Solve (∇² − λ)ψ = rhs on the masked domain.

    The generic solver enforces ψ = 0 at the inner-boundary points and
    returns the corrected field; this method reshapes between fields and
    flat vectors and zeroes the exterior via the mask.

    Parameters
    ----------
    rhs : Float[Array, "Ny Nx"]
        Right-hand side on the full rectangular grid.
        Values outside the physical domain (mask = False) are ignored.

    Returns
    -------
    Float[Array, "Ny Nx"]
        Solution ψ on the full rectangular grid.  Exactly zero outside
        the mask, and ψ ≈ 0 at inner-boundary points.
    """
    ny, nx = self.shape
    # Honor the documented contract: exterior (mask = False) values are
    # ignored by zeroing them before the base solve.
    rhs_masked = rhs * self.mask
    psi_flat = self.solver(rhs_masked.reshape(ny * nx))
    return psi_flat.reshape(ny, nx) * self.mask

Factory

finitevolx.build_capacitance_solver(mask, dx, dy, lambda_=0.0, base_bc='fft')

Pre-compute the capacitance matrix and return a ready-to-use solver.

This is an offline function that performs N_b rectangular spectral solves (N_b = number of inner-boundary points). The result is a :class:CapacitanceSolver whose __call__ method is JIT-compilable.

Parameters:

Name Type Description Default
mask np.ndarray of bool shape (Ny, Nx), or Mask2D

Physical domain mask. True = interior (ocean/fluid), False = exterior (land/walls).

When an :class:Mask2D is passed, the psi staggering mask is extracted automatically.

 required
dx float

Grid spacing in x.

 required
dy float

Grid spacing in y.

 required
lambda_ float

Helmholtz parameter λ. Use 0.0 for pure Poisson.

 0.0
base_bc ('fft', 'dst', 'dct')

Rectangular spectral solver used as the base.

 "fft"

Returns:

Type Description
CapacitanceSolver

A callable equinox Module with all precomputed arrays baked in.
Source code in finitevolx/_src/solvers/elliptic.py 
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
def build_capacitance_solver(
    mask: np.ndarray | Mask2D,
    dx: float,
    dy: float,
    lambda_: float = 0.0,
    base_bc: str = "fft",
) -> CapacitanceSolver:
    """Pre-compute the capacitance matrix and return a ready-to-use solver.

    This is an **offline** function that performs *N_b* rectangular spectral
    solves (``N_b`` = number of inner-boundary points).  The result is a
    :class:`CapacitanceSolver` whose ``__call__`` method is JIT-compilable.

    Parameters
    ----------
    mask : np.ndarray of bool shape (Ny, Nx), or Mask2D
        Physical domain mask.  ``True`` = interior (ocean/fluid),
        ``False`` = exterior (land/walls).

        When an :class:`Mask2D` is passed, the ``psi``
        staggering mask is extracted automatically.
    dx : float
        Grid spacing in x.
    dy : float
        Grid spacing in y.
    lambda_ : float
        Helmholtz parameter λ.  Use ``0.0`` for pure Poisson.
    base_bc : {"fft", "dst", "dct"}
        Rectangular spectral solver used as the base.

    Returns
    -------
    CapacitanceSolver
        A callable equinox Module with all precomputed arrays baked in.
    """
    if isinstance(mask, Mask2D):
        mask = np.asarray(mask.xy_corner_strict, dtype=bool)
    return _build_capacitance_solver_base(mask, dx, dy, lambda_, base_bc)

Masked Laplacian

finitevolx.masked_laplacian(psi, mask, dx, dy, lambda_=0.0)

Apply the masked discrete Helmholtz operator (∇² − λ)·ψ.

Enforces homogeneous Dirichlet conditions at the mask boundary by zeroing psi outside the mask before applying the 5-point stencil. The output is also zeroed outside the mask.

Neighbors at the rectangle edges wrap around (periodic roll), which is consistent with using the FFT as a preconditioner.

Parameters:

Name Type Description Default
psi Float[Array, 'Ny Nx']

Field to which the operator is applied.

 required
mask Float[Array, 'Ny Nx'] or Mask2D

Binary mask: 1 inside the physical domain, 0 outside (land/exterior). When an :class:Mask2D is passed, the psi staggering mask is used (all four surrounding h-cells must be wet).

 required
dx float

Grid spacing in x.

 required
dy float

Grid spacing in y.

 required
lambda_ float

Helmholtz parameter. Default: 0.0 (pure Laplacian).

 0.0

Returns:

Type Description
Float[Array, 'Ny Nx']

Result of (∇² − λ)·(ψ·mask), zeroed outside the mask.
Source code in finitevolx/_src/solvers/iterative.py 
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
def masked_laplacian(
    psi: Float[Array, "Ny Nx"],
    mask: Float[Array, "Ny Nx"] | Mask2D,
    dx: float,
    dy: float,
    lambda_: float = 0.0,
) -> Float[Array, "Ny Nx"]:
    """Apply the masked discrete Helmholtz operator (∇² − λ)·ψ.

    Enforces homogeneous Dirichlet conditions at the mask boundary by zeroing
    *psi* outside the mask before applying the 5-point stencil.  The output
    is also zeroed outside the mask.

    Neighbors at the rectangle edges wrap around (periodic roll), which is
    consistent with using the FFT as a preconditioner.

    Parameters
    ----------
    psi : Float[Array, "Ny Nx"]
        Field to which the operator is applied.
    mask : Float[Array, "Ny Nx"] or Mask2D
        Binary mask: 1 inside the physical domain, 0 outside (land/exterior).
        When an :class:`Mask2D` is passed, the ``psi`` staggering
        mask is used (all four surrounding h-cells must be wet).
    dx : float
        Grid spacing in x.
    dy : float
        Grid spacing in y.
    lambda_ : float
        Helmholtz parameter.  Default: 0.0 (pure Laplacian).

    Returns
    -------
    Float[Array, "Ny Nx"]
        Result of (∇² − λ)·(ψ·mask), zeroed outside the mask.
    """
    if isinstance(mask, Mask2D):
        mask_arr = mask.xy_corner_strict.astype(psi.dtype)
    else:
        mask_arr = mask
    psi_m = psi * mask_arr  # enforce zero outside domain
    # 5-point finite-difference stencil with periodic roll at edges
    lap = (
        jnp.roll(psi_m, 1, axis=1) + jnp.roll(psi_m, -1, axis=1) - 2.0 * psi_m
    ) / dx**2 + (
        jnp.roll(psi_m, 1, axis=0) + jnp.roll(psi_m, -1, axis=0) - 2.0 * psi_m
    ) / dy**2
    return (lap - lambda_ * psi_m) * mask_arr