Moduli Flow: Dynamics in Geometry & Physics

• Moduli Flow is defined as the evolution of parameter spaces of geometric or physical structures, driven by metrics, Hamiltonians, or stability conditions.
• It manifests in various settings such as gradient flows in string vacua, induced boundary RG flows in CFT, and radial attractor dynamics in black hole physics.
• Applications include determining stability in deformed systems and constructing compact moduli spaces in Morse, Floer, and gauge theories, offering actionable insights.

Searching arXiv for the cited works and closely related uses of “moduli flow” to ground the article. arXiv search query: "Moduli flow" Moduli flow denotes several related but technically distinct notions in contemporary mathematical physics and geometry. In the literature surveyed here, it can refer to induced renormalization-group motion under deformations of a conformal-field-theory modulus, gradient or attractor trajectories in scalar-field moduli spaces, Morse or Floer-type flow lines whose compactified trajectory spaces are themselves moduli spaces, and dynamical systems acting on moduli spaces or flat bundles over them. The unifying feature is not a single universal definition, but the systematic appearance of a parameter space of geometric or physical structures together with a preferred evolution law determined by a metric, a Hamiltonian, a stability condition, or a perturbation (Elitzur et al., 2012).

1. Taxonomy of usages

Across the cited works, “moduli flow” is used in several precise senses rather than as a single standardized term. The following table records the principal usages that appear.

Setting	Space	Meaning of “moduli flow”
Boundary CFT	$c=1$ circle/orbifold moduli space	bulk exactly marginal deformation induces boundary RG flow
String vacua	moduli space $\mathcal M$ with metric $g_{ij}$	gradient flow $\dot\phi^i=-\nabla^i\log T$
AdS black holes	scalar moduli in gauged supergravity	radial flow to an attractor horizon
Flat connections	loop space on $\mathcal M^g(P)$	heat flow compared with Yang–Mills $L^2$-flow
Higgs bundles	moduli space of twisted Higgs bundles	gradient flow lines of $\|\phi\|_{L^2}^2$
Teichmüller-type dynamics	bundles over moduli space	earthquake, Teichmüller, or Weil–Petersson-related flows

This multiplicity of meanings is explicit in the research record. “Induced Boundary Flow on the $c=1$ Orbifold Moduli Space” studies motion along the $c=1$ CFT moduli space and the resulting RG flow of boundary conditions (Elitzur et al., 2012). “Geodesic Gradient Flows in Moduli Space” defines moduli flow as the gradient flow of $-\log T$ for brane tensions or particle masses (Etheredge et al., 2023). “Moduli flow and non-supersymmetric AdS attractors” treats radial scalar evolution in gauged supergravity as both moduli flow and holographic RG flow (0711.0036). “Heat flow on the moduli space of flat connections and Yang-Mills theory” compares heat flow on a loop space of flat-connection moduli with Yang–Mills $\mathcal M$0-flow (Janner, 2010). “Flow lines on the moduli space of rank $\mathcal M$1 twisted Higgs bundles” studies the gradient flow lines of the Hitchin norm function on a Higgs-bundle moduli space (Wilkin, 2024).

A plausible implication is that “moduli flow” is best understood as a family resemblance term. What is stable across the different usages is the presence of a moduli space together with a distinguished evolution that selects preferred trajectories, strata, or asymptotic states.

2. Boundary conformal field theory and induced boundary flow

In boundary conformal field theory, moduli flow can mean the boundary RG flow induced by moving in bulk moduli space. For the $\mathcal M$2 free boson and its $\mathcal M$3 orbifold, the relevant mechanism is encoded by the bulk and boundary beta functions

$\mathcal M$4

$\mathcal M$5

Even if all boundary couplings initially vanish, a bulk exactly marginal perturbation can generate boundary couplings when $\mathcal M$6, forcing the brane to move in boundary moduli space (Elitzur et al., 2012).

On the circle branch, the canonical radius-changing operator is $\mathcal M$7. At the self-dual circle, generic conformal boundary states are labeled by

$\mathcal M$8

The bulk-boundary coefficient for the induced current obeys

$\mathcal M$9

Hence the coupling vanishes only for $g_{ij}$0 or $g_{ij}$1, namely the pure Neumann and pure Dirichlet loci. Generic states are therefore unstable under radius change. The induced infinitesimal motion changes $g_{ij}$2 and $g_{ij}$3 while keeping their phases fixed; increasing the radius drives the flow to $g_{ij}$4, while decreasing it drives the flow to $g_{ij}$5 (Elitzur et al., 2012).

The orbifold branch preserves the same basic mechanism after projecting onto $g_{ij}$6-invariant combinations. Generic untwisted orbifold branes again flow to Dirichlet-type configurations when the radius is increased and to Neumann-type configurations when the radius is decreased. Twisted branes at fixed loci are more constrained: the perturbation collapses the continuum present at special radii to discrete twisted Dirichlet or twisted Neumann endpoints. At rational radii, the same consistency statement persists for the allowed bulk marginal operators and known boundary conditions. At the exceptional multicritical point, the only states fixed under both the circle-branch and orbifold-branch flows are the eight discrete branes

$g_{ij}$7

The paper also checks the boundary $g_{ij}$8-theorem. For a single Dirichlet brane on a circle of radius $g_{ij}$9,

$\dot\phi^i=-\nabla^i\log T$0

while for a single Neumann brane,

$\dot\phi^i=-\nabla^i\log T$1

For rational radii, the endpoints are stacks of Dirichlet or Neumann branes, and the resulting $\dot\phi^i=-\nabla^i\log T$2 decreases whether the radius is increased or decreased. In this usage, moduli flow is therefore a coupled bulk-boundary phenomenon: bulk motion along an exactly marginal direction dynamically projects generic special-point branes onto the boundary conditions that survive after deformation (Elitzur et al., 2012).

3. Gradient, geodesic, attractor, and flow-tree pictures in string theory

A second major usage of moduli flow concerns scalar trajectories in string vacua. In “Geodesic Gradient Flows in Moduli Space,” the central object is the scalar charge-to-tension ratio vector

$\dot\phi^i=-\nabla^i\log T$3

For a scalar function $\dot\phi^i=-\nabla^i\log T$4, if

$\dot\phi^i=-\nabla^i\log T$5

and

$\dot\phi^i=-\nabla^i\log T$6

then the flow is a geodesic. Applied to $\dot\phi^i=-\nabla^i\log T$7, this yields the criterion that the gradient flow of $\dot\phi^i=-\nabla^i\log T$8 is geodesic whenever $\dot\phi^i=-\nabla^i\log T$9 has fixed norm on moduli space. The paper further notes that if $\mathcal M^g(P)$0, then $\mathcal M^g(P)$1 itself is a distance function, and along the resulting geodesic one obtains exponential decay laws

$\mathcal M^g(P)$2

Many examples with fixed-length $\mathcal M^g(P)$3-vectors are given, especially $\mathcal M^g(P)$4-BPS states in maximal or half-maximal settings (Etheredge et al., 2023).

In non-supersymmetric AdS attractor physics, moduli flow means radial evolution of scalars from asymptotic AdS data to charge-determined horizon values. For static, spherically symmetric extremal black holes in five-dimensional gauged supergravity, the scalar equations take the form

$\mathcal M^g(P)$5

with

$\mathcal M^g(P)$6

The attractor conditions are

$\mathcal M^g(P)$7

together with positivity of the Hessian. In the constant-potential case, regularity near the extremal horizon forces the scalar perturbations to decay there, so the horizon loses memory of asymptotic moduli. With higher-derivative Gauss–Bonnet terms, the same logic survives after replacing $\mathcal M^g(P)$8 by a corrected effective potential $\mathcal M^g(P)$9. The paper interprets this radial moduli flow holographically as RG flow from a UV AdS fixed point to an IR attractor horizon and exhibits a monotone $L^2$0-function (0711.0036).

A third string-theoretic realization is provided by attractor flow trees for D4-D2-D0 systems in the large-volume limit. Here the asymptotic Kähler modulus is $L^2$1, and a rooted binary tree encodes hierarchical charge splittings. The increment of the flow parameter along an edge is

$L^2$2

Positivity of these flow parameters is an existence condition for the tree. A central result is that their signs can be determined iteratively from the initial moduli $L^2$3, without explicit calculation of the full flow. In the same large-volume regime, the paper proves that the indefinite quadratic form controlling the D4-D2-D0 BPS mass becomes positive definite for flow trees with three or fewer endpoints, implying convergence of the corresponding partition-function contributions. It also argues that $L^2$4-duality requires using rational invariants

$L^2$5

rather than the integer invariants $L^2$6 (Manschot, 2010).

Taken together, these works show that string-theoretic moduli flow can be geodesic, radial, or combinatorial. This suggests that the shared structure is not the form of the evolution equation, but the extraction of physically distinguished trajectories from local charge, tension, or stability data.

4. Moduli spaces of flow lines in Morse, Floer, and gauge theory

A different but closely related usage of moduli flow appears in Morse and Floer theory, where the central objects are moduli spaces of trajectories. For a Morse–Smale pair $L^2$7, the parametrized trajectory space is

$L^2$8

and the unparametrized moduli space is

$L^2$9

The dimensions are

$\|\phi\|_{L^2}^2$0

Choosing orientations on unstable manifolds induces orientations on these moduli spaces; coherent orientations and gluing then identify the boundary of a $\|\phi\|_{L^2}^2$1-dimensional compactified moduli space with broken trajectories, giving the standard proof that $\|\phi\|_{L^2}^2$2 over $\|\phi\|_{L^2}^2$3 (Giroux, 2020).

“Compactness of Moduli Spaces of Gradient Flow Lines in the Uniform Topology” abstracts this picture. It considers classes of $\|\phi\|_{L^2}^2$4-gradient flow lines $\|\phi\|_{L^2}^2$5 satisfying

$\|\phi\|_{L^2}^2$6

with energy

$\|\phi\|_{L^2}^2$7

For trajectories on a half-line converging to a compact critical subset $\|\phi\|_{L^2}^2$8, the moduli space is

$\|\phi\|_{L^2}^2$9

Under two hypotheses—$c=1$0-compactness of bounded-energy trajectories and a shortening property near $c=1$1—and provided the energy satisfies $c=1$2, where $c=1$3 is a spectral gap, the paper proves compactness in the uniform topology on $c=1$4. The framework covers both Morse flow lines and Floer cylinders, and the main new analytic input in the Floer case is an exponential decay estimate with coefficient function continuously depending on the initial loop (Stalljohann, 1 Apr 2026).

A discrete analogue appears in “A combinatorial construction of the moduli space of flowlines in discrete Morse theory.” There, the moduli space of index-$c=1$5 flowlines between critical simplices is a simplicial $c=1$6-complex whose vertices are flowlines and whose edges are generated by elementary moves such as Flop, Insert, and Cancel. Its boundary points are the critical or broken flowlines, and opposite-signed endpoint pairings give a new proof that the discrete Morse differential squares to zero (Bleau, 2023).

The same logic is categorified in “Foundation of Floer homotopy theory I: Flow categories.” A flow category $c=1$7 is a non-unital category enriched in derived orbifolds and stratified by the combinatorics of broken trajectories. Once Floer solutions are assembled into such a flow category, the associated Floer homotopy type is realized as a mapping spectrum in a stable $c=1$8-category of flow categories (Abouzaid et al., 2024).

The moduli-space perspective also governs Janner’s comparison between heat flow on the loop space of the moduli space of flat connections and Yang–Mills $c=1$9-flow on $c=1$0. The loop-space energy is

$c=1$1

while the perturbed Yang–Mills functional is

$c=1$2

For sufficiently small $c=1$3, bounded Morse homology on the loop space of $c=1$4 is isomorphic to the bounded Morse homology of the Yang–Mills connection space (Janner, 2010).

5. Dynamical systems on moduli spaces and over them

In several works, moduli flow is a genuine dynamical system acting on a moduli space or on a bundle over it. A central example is the earthquake flow on the unit measured lamination bundle over moduli space. For a closed genus-$c=1$5 surface, the paper proves that if correlations satisfied a polynomial bound

$c=1$6

then necessarily

$c=1$7

Hence earthquake flow is at most polynomially mixing and in particular not exponentially mixing. The mechanism is long cusp excursions in the thin part of moduli space (Bonnafoux, 2022).

A related construction lifts Teichmüller dynamics to a flat bundle whose fibers are deformation spaces of representations $c=1$8. For compact connected $c=1$9, the resulting horizontal lift of Teichmüller flow to

$-\log T$0

is strongly mixing on each connected component of each stratum component bundle, while the analogous lift of the Weil–Petersson geodesic local flow is ergodic (Forni et al., 2017).

Other examples treat the moduli space as the target of a geometric PDE. In “Ergodic and Entropic Behavior of the Harmonic Map Heat Flow to the Moduli Space of Flat Tori,” the target is

$-\log T$1

with hyperbolic metric

$-\log T$2

The harmonic map heat flow

$-\log T$3

decreases the energy

$-\log T$4

and the time-averaged pushforward measures converge weak-$-\log T$5 to the normalized hyperbolic measure on $-\log T$6. Under additional assumptions, the relative entropy with respect to that measure decays to zero (Kolaei, 18 Apr 2025).

Algebro-geometric gradient dynamics on moduli spaces appear in the rank-$-\log T$7 twisted Higgs-bundle setting. There the Morse function is

$-\log T$8

the moment map for the $-\log T$9-action scaling the Higgs field. In rank $\mathcal M$00, flow lines between critical sets are classified by secant varieties of the underlying curve embedded in the projectivized negative eigenspace of the Hessian, and the Morse compactification by broken trajectories is identified with Bertram’s resolution of secant varieties (Wilkin, 2024).

Finally, the vortex literature supplies a spectral-flow variant. Over the centered BPS $\mathcal M$01-vortex moduli space, the natural local coordinate near coincidence is

$\mathcal M$02

not $\mathcal M$03, and the discrete positive fluctuation spectrum varies smoothly in $\mathcal M$04. The paper shows that at $\mathcal M$05 there are three positive discrete modes, one branch rises into the continuum as $\mathcal M$06 increases, and two branches approach the $\mathcal M$07-vortex shape-mode eigenvalue at large separation (Alonso-Izquierdo et al., 2023).

6. Terminological boundaries and recurrent mechanisms

A common misconception is to conflate moduli flow with modular flow. The latter belongs to Tomita–Takesaki theory and concerns a state/algebra pair $\mathcal M$08, with modular operator $\mathcal M$09, modular Hamiltonian $\mathcal M$10, and modular automorphism group

$\mathcal M$11

In Sorce’s presentation, modular flow is the unique thermal time-evolution compatible with the KMS condition for a cyclic and separating state (Sorce, 2023). In the free-fermion cylinder problem, modular flow is computed explicitly and can be non-local for generic periodic zero-energy states; in pure Ramond limits it can become local while still mixing chiralities (Cadamuro et al., 2024). This operator-algebraic notion is conceptually distinct from moduli flow.

Within the genuinely moduli-theoretic literature, several recurrent mechanisms stand out. First, many flows are gradient-like: $\mathcal M$12 in string vacua, $\mathcal M$13 for harmonic maps into a moduli orbifold, and $\mathcal M$14-gradient flow of $\mathcal M$15 on Higgs-bundle moduli space (Etheredge et al., 2023, Kolaei, 18 Apr 2025, Wilkin, 2024). Second, compactification by broken trajectories is central whenever the objects of interest are themselves moduli spaces of flow lines, as in Morse, Floer, and discrete Morse theory (Giroux, 2020, Bleau, 2023, Stalljohann, 1 Apr 2026). Third, stability or attractor conditions often replace direct integration of the flow: boundary CFT uses bulk-boundary OPE coefficients, AdS attractors use critical points of $\mathcal M$16, and D-brane flow trees use the signs of flow parameters and walls of marginal stability (Elitzur et al., 2012, 0711.0036, Manschot, 2010).

A plausible implication is that “moduli flow” is less a single theory than a transferable pattern. It organizes how geometric structures respond to deformation, how asymptotic data are forgotten or retained, and how dynamical selection rules emerge on spaces that already classify solutions, bundles, branes, or metrics.