Regge–Ricci Flow: Discrete Geometric Evolution

• Regge–Ricci Flow is a discrete analogue of Hamilton’s Ricci flow that replaces smooth metrics with simplicial edge lengths and averages hinge deficit angles to mimic curvature evolution.
• It constructs discrete Ricci curvature by leveraging a dual-edge formulation where curvature is computed from weighted averages of hinge deficit angles over circumcentric dual areas.
• Validated by models like the 600-cell and icosahedral cylinder, the method demonstrates convergence to continuum behavior and exact reproduction of symmetric Ricci flow solutions.

Regge–Ricci Flow (RRF) is a discrete analogue of Hamilton’s Ricci flow on a $d$-dimensional piecewise-flat simplicial manifold, constructed within Regge calculus by replacing the smooth metric tensor with simplicial edge lengths and replacing smooth Ricci curvature with edge-based averages of hinge curvature derived from deficit angles and circumcentric dual areas. In the formulation introduced under the title “Simplicial Ricci Flow,” the primary discrete equation is first written on dual edges as $$\frac{1}{\lambda}\frac{\partial \lambda}{\partial t}=-Rc_\lambda$$, and then projected to simplicial edges as $$\left\langle \frac{1}{\lambda}\frac{\partial \lambda}{\partial t}\right\rangle_\ell=-Rc_\ell$$, where the actual metric degrees of freedom are the primal edge lengths $\ell$ (Miller et al., 2013). The construction depends on the first geometric discretization of the Ricci tensor in arbitrary dimension, developed as an integrated Ricci one-form on both the simplicial lattice and its circumcentric dual (Alsing et al., 2011). Conceptually, RRF is meant to imitate the smooth Ricci flow $\frac{\partial}{\partial t}g=-2\,\Ric(g)$, understood as a geometric heat flow that smooths curvature while retaining the possibility of singularity formation (Khan, 2022).

1. Smooth Ricci flow as the model for discretization

The smooth prototype for RRF is Hamilton’s Ricci flow,

$\frac{\partial}{\partial t}g=-2\,\Ric(g), \qquad \frac{\partial g_{ij}}{\partial t}=-2\,\mathrm{Ric}_{ij},$

with $g=g(t)$ a time-dependent Riemannian metric and $\Ric$ the Ricci tensor (Khan, 2022). In the smooth theory, the metric coefficients encode local lengths, angles, areas, volumes, geodesics, and curvature, so Ricci flow is literally an evolution equation for the geometry itself.

The foundational intuition emphasized in the Ricci-flow background literature is that Ricci flow is a “geometric heat flow.” Ordinary heat flow spreads temperature until it becomes more uniform, while Ricci flow spreads curvature until the geometry becomes more uniform. The analogy is sharpened by two related observations. First, the Laplacian can be interpreted as a comparison between a function and its average on small spheres. Second, Ricci curvature can be interpreted as an average of sectional curvatures through planes containing a fixed direction,

$\Ric(X,X)=\frac12\frac{(n-1)}{\omega(\mathbb{S}^{n-2})}\oint_{\|Y\|=1,\;X\perp Y} K(X,Y)\,d\mathbb{S}^{n-2}(Y).$

This average-of-directions viewpoint is especially suggestive for discretization, because it recasts Ricci curvature as a directional average rather than a purely differential expression (Khan, 2022).

A second important smooth motivation is the harmonic-coordinate identity

$\Ric_{ij}=-\frac12\Delta g_{ij}+\text{lower order terms}.$

This indicates that Ricci flow behaves roughly like diffusion of metric coefficients, with nonlinear corrections. In a simplicial setting, where the natural metric variables are edge lengths rather than tensor components $$\frac{1}{\lambda}\frac{\partial \lambda}{\partial t}=-Rc_\lambda$$0, this suggests that a discrete flow should evolve edge-length data according to neighboring curvature information rather than by a purely pointwise rule (Khan, 2022).

2. Regge-calculus framework and piecewise-flat geometry

RRF is formulated on a compact $$\frac{1}{\lambda}\frac{\partial \lambda}{\partial t}=-Rc_\lambda$$1-dimensional piecewise-flat simplicial lattice $$\frac{1}{\lambda}\frac{\partial \lambda}{\partial t}=-Rc_\lambda$$2, assumed well-centered Delaunay for simplicity, together with its circumcentric Voronoi dual $$\frac{1}{\lambda}\frac{\partial \lambda}{\partial t}=-Rc_\lambda$$3 (Miller et al., 2013). A $$\frac{1}{\lambda}\frac{\partial \lambda}{\partial t}=-Rc_\lambda$$4-simplex $$\frac{1}{\lambda}\frac{\partial \lambda}{\partial t}=-Rc_\lambda$$5 has a dual $$\frac{1}{\lambda}\frac{\partial \lambda}{\partial t}=-Rc_\lambda$$6-cell $$\frac{1}{\lambda}\frac{\partial \lambda}{\partial t}=-Rc_\lambda$$7, and the basic orthogonality relation

$$\frac{1}{\lambda}\frac{\partial \lambda}{\partial t}=-Rc_\lambda$$8

is central to the entire construction.

For each primal-dual pair $$\frac{1}{\lambda}\frac{\partial \lambda}{\partial t}=-Rc_\lambda$$9, one defines a $$\left\langle \frac{1}{\lambda}\frac{\partial \lambda}{\partial t}\right\rangle_\ell=-Rc_\ell$$0-dimensional hybrid cell $$\left\langle \frac{1}{\lambda}\frac{\partial \lambda}{\partial t}\right\rangle_\ell=-Rc_\ell$$1 with volume

$$\left\langle \frac{1}{\lambda}\frac{\partial \lambda}{\partial t}\right\rangle_\ell=-Rc_\ell$$2

These hybrid cells tile the geometry and provide the natural support domains for integrated curvature quantities (Miller et al., 2013).

In Regge calculus, curvature is concentrated on codimension-2 simplices, called hinges. If $$\left\langle \frac{1}{\lambda}\frac{\partial \lambda}{\partial t}\right\rangle_\ell=-Rc_\ell$$3 is a hinge and $$\left\langle \frac{1}{\lambda}\frac{\partial \lambda}{\partial t}\right\rangle_\ell=-Rc_\ell$$4 $$\left\langle \frac{1}{\lambda}\frac{\partial \lambda}{\partial t}\right\rangle_\ell=-Rc_\ell$$5-simplices meet at $$\left\langle \frac{1}{\lambda}\frac{\partial \lambda}{\partial t}\right\rangle_\ell=-Rc_\ell$$6 with hyperdihedral angles $$\left\langle \frac{1}{\lambda}\frac{\partial \lambda}{\partial t}\right\rangle_\ell=-Rc_\ell$$7, the deficit angle is

$$\left\langle \frac{1}{\lambda}\frac{\partial \lambda}{\partial t}\right\rangle_\ell=-Rc_\ell$$8

The dual of $$\left\langle \frac{1}{\lambda}\frac{\partial \lambda}{\partial t}\right\rangle_\ell=-Rc_\ell$$9 is a 2-dimensional polygon $\ell$0, and the discrete sectional curvature is

$\ell$1

This is the basic curvature density from which the discrete Ricci tensor is assembled (Miller et al., 2013).

The hinge-based curvature hierarchy used in the RRF literature treats curvature as uniformly distributed over the hinge hybrid block $\ell$2, with

$\ell$3

This local proportionality is one of the structural assumptions of the piecewise-flat framework (Miller et al., 2013).

Geometric support	Curvature quantity	Discrete representative
Hinge $\ell$4	Sectional / Riemann-type curvature	$\ell$5
Dual edge $\ell$6	Ricci one-form	$\ell$7
Simplicial edge $\ell$8	Projected Ricci one-form	$\ell$9
Vertex $\frac{\partial}{\partial t}g=-2\,\Ric(g)$0	Scalar curvature	$\frac{\partial}{\partial t}g=-2\,\Ric(g)$1

This support structure is not merely organizational. It determines how curvature contractions are performed: sectional curvature is hinge-based, Ricci curvature is naturally dual-edge-based, and scalar curvature is vertex-based (Miller et al., 2013).

3. The discrete Ricci tensor on dual and simplicial lattices

The first explicit geometric discretization of the Ricci tensor in Regge calculus was constructed as an integrated Ricci one-form on dual edges $\frac{\partial}{\partial t}g=-2\,\Ric(g)$2 and then transferred to simplicial edges $\frac{\partial}{\partial t}g=-2\,\Ric(g)$3, with the two formulations proved equivalent (Alsing et al., 2011). This result is the key ingredient that makes higher-dimensional Regge–Ricci Flow possible.

On the dual lattice, the Ricci one-form is defined by summing hinge curvatures over all dual polygons $\frac{\partial}{\partial t}g=-2\,\Ric(g)$4 containing $\frac{\partial}{\partial t}g=-2\,\Ric(g)$5: $\frac{\partial}{\partial t}g=-2\,\Ric(g)$6 Using the Regge curvature formula, this becomes

$\frac{\partial}{\partial t}g=-2\,\Ric(g)$7

so the dual Ricci one-form is a volume-weighted average of hinge Riemann curvatures around the dual edge (Alsing et al., 2011).

For Ricci flow, however, the relevant metric variables are not dual-edge lengths but simplicial edge lengths $\frac{\partial}{\partial t}g=-2\,\Ric(g)$8. The dual Ricci tensor is therefore lowered to the simplicial lattice: $\frac{\partial}{\partial t}g=-2\,\Ric(g)$9 After rearranging sums, one obtains the integrated simplicial Ricci one-form

$\frac{\partial}{\partial t}g=-2\,\Ric(g), \qquad \frac{\partial g_{ij}}{\partial t}=-2\,\mathrm{Ric}_{ij},$0

For $\frac{\partial}{\partial t}g=-2\,\Ric(g), \qquad \frac{\partial g_{ij}}{\partial t}=-2\,\mathrm{Ric}_{ij},$1, this yields the compact edge-based ratio

$\frac{\partial}{\partial t}g=-2\,\Ric(g), \qquad \frac{\partial g_{ij}}{\partial t}=-2\,\mathrm{Ric}_{ij},$2

where the $\frac{\partial}{\partial t}g=-2\,\Ric(g), \qquad \frac{\partial g_{ij}}{\partial t}=-2\,\mathrm{Ric}_{ij},$3-weighted average is

$\frac{\partial}{\partial t}g=-2\,\Ric(g), \qquad \frac{\partial g_{ij}}{\partial t}=-2\,\mathrm{Ric}_{ij},$4

The Ricci tensor is therefore an edge-based weighted average of deficit angles divided by an edge-based weighted average of dual areas (Alsing et al., 2011).

This formulation makes explicit why the simplicial Ricci tensor is adapted to RRF. In Regge calculus the metric is encoded by edge lengths, so a curvature flow driven by Ricci curvature requires an edge-based curvature object. The dual-edge formulation reflects the natural contraction of the hinge-based Riemann tensor, while the simplicial-edge formulation translates that contraction into the actual Regge degrees of freedom (Alsing et al., 2011).

The dimensional dependence is also explicit. In $\frac{\partial}{\partial t}g=-2\,\Ric(g), \qquad \frac{\partial g_{ij}}{\partial t}=-2\,\mathrm{Ric}_{ij},$5, hinges are edges and the Ricci tensor carries the full curvature information of the manifold. In $\frac{\partial}{\partial t}g=-2\,\Ric(g), \qquad \frac{\partial g_{ij}}{\partial t}=-2\,\mathrm{Ric}_{ij},$6, where Ricci curvature is determined by scalar curvature, the simplicial Ricci one-form becomes the arithmetic average of the endpoint vertex curvatures,

$\frac{\partial}{\partial t}g=-2\,\Ric(g), \qquad \frac{\partial g_{ij}}{\partial t}=-2\,\mathrm{Ric}_{ij},$7

This 2-dimensional collapse is one reason why higher-dimensional RRF required a separate edge-based Ricci construction (Alsing et al., 2011).

4. Definition of Regge–Ricci Flow

The defining insight of RRF is that the smooth mixed-index Ricci flow equation can be written in a local orthogonal frame as

$\frac{\partial}{\partial t}g=-2\,\Ric(g), \qquad \frac{\partial g_{ij}}{\partial t}=-2\,\mathrm{Ric}_{ij},$8

This turns the flow into a relation between a fractional rate of change of a diagonal metric coefficient and the corresponding mixed Ricci component (Miller et al., 2013).

On a circumcentric primal-dual lattice pair, the mixed Ricci tensor is naturally associated with dual edges. If the metric component along a dual edge $\frac{\partial}{\partial t}g=-2\,\Ric(g), \qquad \frac{\partial g_{ij}}{\partial t}=-2\,\mathrm{Ric}_{ij},$9 is $g=g(t)$0, then

$g=g(t)$1

Matching this with the smooth mixed-index equation leads to the dual-edge RRF equation

$g=g(t)$2

This is the first, and conceptually primary, form of Regge–Ricci Flow (Miller et al., 2013).

The dual-edge system is not yet the practical evolution law, because the entire geometry of both $g=g(t)$3 and $g=g(t)$4 is determined by the simplicial edge lengths $g=g(t)$5, while the set of dual edges is typically larger. The full collection of dual-edge equations is therefore generally overdetermined. The remedy is to project the dual-edge flow onto simplicial edges: $g=g(t)$6 with

$g=g(t)$7

This simplicial-edge equation is the main operational form of RRF (Miller et al., 2013).

A notable feature is that the left-hand side is not $g=g(t)$8. Regge–Ricci Flow does not directly identify the simplicial-edge fractional change with $g=g(t)$9. Instead, the natural tensorial variable is the fractional change of dual-edge lengths, and the simplicial equation is a volume-weighted projection of that dual-edge evolution onto the primal lattice. This is one of the distinctive structural differences between RRF and more ad hoc edge-update rules (Miller et al., 2013).

The Ricci term on simplicial edges can also be written as

$\Ric$0

so the edge Ricci tensor is a double weighted average of hinge sectional curvatures. This formula makes the contraction pattern explicit: hinge curvature is first averaged to dual edges and then to simplicial edges (Miller et al., 2013).

5. Dimensional structure and solved model geometries

The RRF construction is dimension-independent. In $\Ric$1 dimensions, hinges are $\Ric$2-simplices, duals of hinges are 2-cells, Ricci curvature naturally lives on dual edges $\Ric$3, and scalar curvature naturally lives on vertices. The hybrid volume formula

$\Ric$4

is used uniformly across dimensions (Miller et al., 2013).

In $\Ric$5, where hinges are edges, the formulas simplify substantially. For an edge $\Ric$6, the RRF equation becomes an explicit coupled nonlinear first-order ODE system for the simplicial edge lengths. Because each dual edge $\Ric$7 depends on nearby primal edges,

$\Ric$8

the projected equations close on the primal edge variables (Miller et al., 2013).

Two classes of solved examples are central in the literature.

The first is the homogeneous 3-sphere, approximated by the boundaries of the regular 4-polytopes 5-cell, 16-cell, and 600-cell. In the smooth continuum, the round $\Ric$9 of radius $\Ric(X,X)=\frac12\frac{(n-1)}{\omega(\mathbb{S}^{n-2})}\oint_{\|Y\|=1,\;X\perp Y} K(X,Y)\,d\mathbb{S}^{n-2}(Y).$0 satisfies

$\Ric(X,X)=\frac12\frac{(n-1)}{\omega(\mathbb{S}^{n-2})}\oint_{\|Y\|=1,\;X\perp Y} K(X,Y)\,d\mathbb{S}^{n-2}(Y).$1

For the regular simplicial models, all primal edges are equal, and the RRF equation reduces to

$\Ric(X,X)=\frac12\frac{(n-1)}{\omega(\mathbb{S}^{n-2})}\oint_{\|Y\|=1,\;X\perp Y} K(X,Y)\,d\mathbb{S}^{n-2}(Y).$2

After matching the polytope boundary volume to $\Ric(X,X)=\frac12\frac{(n-1)}{\omega(\mathbb{S}^{n-2})}\oint_{\|Y\|=1,\;X\perp Y} K(X,Y)\,d\mathbb{S}^{n-2}(Y).$3, the effective collapse coefficient approaches the continuum value $\Ric(X,X)=\frac12\frac{(n-1)}{\omega(\mathbb{S}^{n-2})}\oint_{\|Y\|=1,\;X\perp Y} K(X,Y)\,d\mathbb{S}^{n-2}(Y).$4. The 600-cell model reproduces the continuum Ricci coefficient within about $\Ric(X,X)=\frac12\frac{(n-1)}{\omega(\mathbb{S}^{n-2})}\oint_{\|Y\|=1,\;X\perp Y} K(X,Y)\,d\mathbb{S}^{n-2}(Y).$5, and the convergence from 16-cell to 600-cell is approximately second order in the deficit angle (Miller et al., 2013).

The second is the icosahedral 3-cylinder, a discrete model of the continuum metric

$\Ric(X,X)=\frac12\frac{(n-1)}{\omega(\mathbb{S}^{n-2})}\oint_{\|Y\|=1,\;X\perp Y} K(X,Y)\,d\mathbb{S}^{n-2}(Y).$6

In the simplicial model, spherical cross-sections are approximated by icosahedra of edge length $\Ric(X,X)=\frac12\frac{(n-1)}{\omega(\mathbb{S}^{n-2})}\oint_{\|Y\|=1,\;X\perp Y} K(X,Y)\,d\mathbb{S}^{n-2}(Y).$7, adjacent sections are connected by axial edges $\Ric(X,X)=\frac12\frac{(n-1)}{\omega(\mathbb{S}^{n-2})}\oint_{\|Y\|=1,\;X\perp Y} K(X,Y)\,d\mathbb{S}^{n-2}(Y).$8, and symmetry reduces the system to two edge types. The resulting RRF equations are

$\Ric(X,X)=\frac12\frac{(n-1)}{\omega(\mathbb{S}^{n-2})}\oint_{\|Y\|=1,\;X\perp Y} K(X,Y)\,d\mathbb{S}^{n-2}(Y).$9

Using the area match

$\Ric_{ij}=-\frac12\Delta g_{ij}+\text{lower order terms}.$0

one obtains exactly

$\Ric_{ij}=-\frac12\Delta g_{ij}+\text{lower order terms}.$1

with no axial evolution, matching the continuum cylinder solution exactly (Miller et al., 2013).

These examples show two different validation modes for RRF: convergence to the smooth flow under refined regular triangulation in the spherical case, and exact reproduction of a nontrivial continuum Ricci-flow solution in a highly symmetric cylindrical case (Miller et al., 2013).

6. Scope, misconceptions, and adjacent directions

Regge–Ricci Flow is specifically a piecewise-flat, simplicial, Regge-calculus discretization of Ricci flow. It is therefore important to separate it from several neighboring theories that use similar terminology but address different problems.

The most common terminology trap is “RG-2 flow.” In that literature, “RG” means renormalization group, not Regge calculus. The flow

$\Ric_{ij}=-\frac12\Delta g_{ij}+\text{lower order terms}.$2

is the 2-loop renormalization-group deformation of Ricci flow arising from the nonlinear sigma model, not a simplicial or piecewise-flat Regge construction (Gimre et al., 2013). The distinction is substantive: RG-2 flow is a continuum curvature-squared deformation of Ricci flow, whereas RRF is a discrete curvature flow built from deficit angles and dual volumes.

Other adjacent but non-Regge directions are also relevant only by analogy. Generalized Ricci flows coupled to a $\Ric_{ij}=-\frac12\Delta g_{ij}+\text{lower order terms}.$3 gauge field and a $\Ric_{ij}=-\frac12\Delta g_{ij}+\text{lower order terms}.$4-field, with short-time existence proved after gauge fixing, are continuum PDE systems rather than simplicial curvature flows (He et al., 2011). Conceptual attempts to connect Ricci flow with the Hamilton–Jacobi formulation of general relativity use Ricci flow mainly as a source of an external evolution parameter and do not provide any Regge-calculus formalism (Alzain, 2022). Topological quantum-gravity constructions in which Ricci-type equations arise as localization equations isolate DeTurck-type gauge fixing and Perelman-type functionals, but they likewise remain continuum theories with no triangulated or deficit-angle formulation (Frenkel et al., 2020).

Within the RRF literature itself, several structural assumptions and limitations are explicit. The construction assumes a compact, well-centered Delaunay simplicial lattice with circumcentric Voronoi dual; the circumcentric orthogonality is essential to the simple factorized hybrid-volume formulas (Miller et al., 2013). Curvature in each hinge hybrid block is treated as uniformly distributed, and the block is treated locally as an Einstein space so that the hinge-based $\Ric_{ij}=-\frac12\Delta g_{ij}+\text{lower order terms}.$5, $\Ric_{ij}=-\frac12\Delta g_{ij}+\text{lower order terms}.$6, and $\Ric_{ij}=-\frac12\Delta g_{ij}+\text{lower order terms}.$7 are proportional there (Miller et al., 2013). The raw dual-edge flow system is typically overdetermined and must be projected to the primal edge variables (Miller et al., 2013).

The literature also records that, like continuum Ricci flow, the discrete equations can display instability, and that modified or gauge-fixed formulations may be needed. Proposed future directions include neck-pinching singularities, surgery on simplicial manifolds, normalized or gauge-fixed versions of Ricci flow, simplicial diffeomorphisms, Perelman-type modifications, and stability analysis. Preliminary studies of perturbations of the 600-cell and of dumbbell or neck-pinch geometries are specifically noted (Miller et al., 2013).

Taken together, these results place Regge–Ricci Flow at the intersection of geometric analysis and piecewise-flat gravity. Its defining content is not merely that it discretizes a curvature flow, but that it does so by preserving the Regge-calculus hierarchy

$\Ric_{ij}=-\frac12\Delta g_{ij}+\text{lower order terms}.$8

with the simplicial edge lengths serving as the discrete metric variables and circumcentric duality supplying the contraction mechanism (Alsing et al., 2011). This makes RRF the natural simplicial candidate for a higher-dimensional discrete Ricci flow in the Regge sense.