---
title: Bryant’s Dirichlet-Type Metric in G2 Structures
url: https://www.emergentmind.com/topics/bryant-s-dirichlet-type-metric
type: topic
---

# Bryant’s Dirichlet-Type Metric in G2 Structures

Searching arXiv for recent and foundational papers on Bryant's Dirichlet-type metric and related Dirichlet/gradient metrics.
Bryant’s Dirichlet-type metric is a formal Riemannian metric on the infinite-dimensional space of closed \(G_2\)-structures in a fixed cohomology class, defined by an \(L^2\)-pairing on canonical potentials of exact variations. In the current literature, the most precise use of the term occurs for the space \(\mathcal M\) of closed \(G_2\)-structures, where the metric organizes geodesics, geodesic concavity of Hitchin’s volume functional, and length contraction for the \(G_2\) Laplacian flow [2507.20456]. Closely related Dirichlet or gradient metrics were studied earlier on spaces of Kähler and Sasakian metrics, where they are likewise defined from first-derivative or gradient pairings and serve as weak Riemannian structures with nontrivial curvature and geodesic theory [1202.6610] [1405.1211].

## 1. \(G_2\)-geometric definition

Let \(M\) be a closed \(7\)-manifold, and let \([\varphi]\subset H^3(M,\mathbb R)\) be a cohomology class containing at least one \(G_2\)-structure. The relevant configuration space is
\[
\mathcal M,
\]
the space of all closed \(G_2\)-structures in \([\varphi]\). At each \(\varphi\in\mathcal M\), the tangent space is identified with exact \(3\)-forms,
\[
T_\varphi \mathcal M \cong d\Omega^2(M),
\]
so tangent vectors are written as
\[
X=d\alpha,\qquad Y=d\beta.
\]

The metric is defined by taking the \(L^2\) inner product of the canonical potentials of tangent vectors. If \(u,v\) are the \(d\)-exact \(3\)-forms representing \(X,Y\), then
\[
\mathcal G_\varphi(X,Y) =\int_M g_\varphi(\delta u,\delta v)\, \mathrm{vol}_\varphi,
\]
where \(g_\varphi\) is the Riemannian metric determined by the \(G_2\)-structure \(\varphi\), and \(\delta\) is the codifferential. The paper also presents the equivalent \(L^2\)-metric on \(2\)-forms,
\[
\mathcal G(\alpha,\beta)=\int_M g_\varphi(\alpha,\beta)\,\mathrm{vol}_\varphi, \qquad \alpha,\beta\in \Omega^2(M),
\]
and rewrites the metric in “gradient metric” form as
\[
\mathcal G_\varphi(X,Y)=\int_M g_\varphi(GX,Y)\,\mathrm{vol}_\varphi,
\]
where \(G\) is the Green operator chosen so that \(GX\) is the canonical potential of \(X\) [2507.20456].

This formulation makes the metric Hodge-theoretic: exact variations are measured through their canonical potentials rather than directly by the exact \(3\)-forms themselves. A plausible implication is that the geometry is tailored to the cohomological constraint defining \(\mathcal M\), rather than to an unconstrained \(L^2\)-geometry on all \(3\)-forms.

## 2. Geodesics and the normalization mechanism

The geodesic theory is expressed first on \(2\)-forms. A family \(\alpha(t)\) is an \(L^2\)-geodesic if it solves
\[
0=\alpha_t +g(\tau,\alpha)\alpha -\frac{1}{2}\delta\!\left(|\alpha|^2-{}_\alpha\alpha\right) -\frac{1}{2}{}_\alpha d\alpha.
\]
A family of closed \(G_2\)-structures \(\varphi(t)\) is then a geodesic if
\[
\varphi_t=d\alpha
\]
for some such \(L^2\)-geodesic \(\alpha(t)\) [2507.20456].

Several distinguished subclasses are singled out. A canonical geodesic satisfies \(\alpha=\delta u\). An \(\Omega^2_{14}\)-geodesic satisfies \(\alpha\in \Omega^2_{14}\). A gauge-fixing geodesic satisfies \(\alpha\in T\mathcal M_0\). An \(\mathcal N\)-geodesic satisfies \(\mathcal N(\alpha)=0\). An \(\mathcal N_{14}\)-geodesic satisfies both \(\alpha\in\Omega^2_{14}\) and \(\mathcal N(\alpha)=0\). The normalization condition is essential because the paper states explicitly that Hitchin’s volume is not concave along arbitrary geodesics [2507.20456].

The normalization functional \(\mathcal N(\alpha)\) is built from several pieces. The paper writes
\[
\mathcal N(\alpha) :=\|X^\bot\|^2_L+\mathcal L(\alpha)+\mathcal R() +\int_M\!\Big[ -g(\tau,\alpha)g(\,d\cdot) +g({}_\tau\tau,{}_d) \Big]\mathrm{vol},
\]
where \(X=d\alpha\), \(X^\bot\) is the gauge-fixed component, \(\mathcal L(\alpha)\) is built from the linearized operator
\[
L(\alpha)=\delta\!\left[\frac{4}{3}X_1+X_7-X_{27}\right],
\]
and \(\mathcal R()\) is a torsion/Ricci-type correction term involving scalar and Ricci curvature. For an \(\Omega^2_{14}\)-geodesic, the condition simplifies to an expression involving \(\|X^\bot\|_L^2\), \(\|X_1\|_L^2\), \(\|X_7\|_L^2\), \(\|X_{27}\|_L^2\), \(\mathcal R()\), and a torsion term [2507.20456].

The main Hessian statement is written as
\[
3\,\varphi_{tt} = -\|X^\bot\|_L^2+\mathcal N(\alpha)
\]
for an \(\Omega^2_{14}\)-geodesic \(\alpha(t)\). The paper concludes that along an \(\mathcal N_{14}\)-geodesic,
\[
\varphi_{tt}\le 0,
\]
so Hitchin’s volume functional is geodesically concave [2507.20456]. The normalization is therefore not auxiliary bookkeeping; it is the mechanism that converts the geodesic equation into a concavity statement.

## 3. Hitchin’s volume and the \(G_2\) Laplacian flow

Hitchin’s volume functional is
\[
\mathcal V(\varphi)=\int_M \mathrm{vol}_\varphi.
\]
In the hypersymplectic setting discussed later in the same work, it is also written as
\[
\mathcal V(\varphi)=\int_M u^{1/3}\,\mathrm{vol}_E.
\]
The paper’s central claim is that, under the normalization condition \(\mathcal N(\alpha)=0\), this functional becomes geodesically concave with respect to Bryant’s Dirichlet-type metric [2507.20456].

The same metric also governs the \(G_2\) Laplacian flow,
\[
\varphi_t = d\tau,
\]
where \(\tau\) is the torsion form. The paper identifies this as the gradient flow of Hitchin’s volume with respect to Bryant’s metric. For a path \(\varphi(t,s)\) with \(s\)-variation \(d\beta\), the length at fixed \(t\) is
\[
L(t)=\int_0^1 \| \partial_s\varphi \|_{\mathcal G}\,ds, \qquad \|\partial_s\varphi\|_{\mathcal G}^2=\mathcal G(d\beta,d\beta).
\]
The main contraction identity is
\[
\frac12\,\partial_t \|\partial_s\varphi\|_{\mathcal G}^2 = -\|X^\bot\|_L^2+\mathcal N(\beta).
\]
Hence, for \(\mathcal N\)-paths,
\[
\partial_t \|\partial_s\varphi\|_{\mathcal G}^2 \le 0.
\]
The paper summarizes this by stating that the gradient flow of the volume functional decreases the length of all normalised paths [2507.20456].

These statements place Bryant’s metric in a standard Riemannian variational role: it is simultaneously the metric defining geodesics and the metric relative to which a natural functional has a gradient flow. A plausible implication is that the normalization condition isolates those directions in which the flow exhibits genuine contractive behavior.

## 4. Hypersymplectic and hyper-symplectic realizations

A major source of examples is the product construction
\[
M=\mathbb T^3\times X^4,
\]
where a hyper-symplectic triple \(\underline\omega=(\omega_1,\omega_2,\omega_3)\) determines a closed \(G_2\)-structure
\[
\varphi = d\theta^{123} - d\theta^i\wedge \omega_i.
\]
Within this framework the paper studies spaces such as \(\mathcal M_1\), a cohomogeneity-one family on \(\mathbb T^7\), and \(\mathcal M_3\), a space of triples of such structures [2507.20456].

For the torus-fibration example \(\mathcal M_1\), the paper proves three geometric properties: the volume is geodesically concave along the canonical geodesic, the gradient flow decreases the distance, and the sectional curvature is nonnegative. The concavity statement is expressed by
\[
9\,\varphi_{tt}=-2\int_M \left[u^{-1}f-u^{-2}u_0\alpha_3\right]^2 \,\mathrm{vol}\le 0.
\]
In the more general triple setting \(\mathcal M_3\), the weighted volume functional
\[
\mathcal V_{\underline\chi}(\underline\omega) = \sum_i \int_M \chi_i(u_i)\,\mathrm{vol}_E
\]
is geodesically concave along the corresponding geodesic, and the associated weighted flow decreases the distance [2507.20456].

The hyper-symplectic motivation is structural. For closed \(M=\mathbb T^3\times X^4\), the torsion-free condition for the induced \(G_2\)-structure is equivalent to the hyper-symplectic structure being hyper-Kähler. The paper further states that the resulting geodesic concavity yields uniqueness results modulo diffeomorphisms [2507.20456]. In this way the metric is not merely formal: it produces effective rigidity statements in explicit geometric models.

## 5. Earlier Dirichlet and gradient metrics on Kähler and Sasakian spaces

Before the \(G_2\)-theoretic formulation, closely related Dirichlet or gradient metrics were developed on spaces of Kähler metrics. In the Kähler one-form model
\[
\mathcal A:=\{d\varphi \mid \omega_\varphi \text{ is a Kähler metric in } \mathcal H\},
\]
the Dirichlet metric is
\[
\langle d\psi,d\chi\rangle_{d\varphi}^{D} = \int_M (d\psi,d\chi)_{g_\varphi}\,\frac{\omega_\varphi^n}{n!},
\]
and, equivalently on potentials,
\[
\langle \psi,\chi\rangle_{\omega_\varphi}^{D} = \int_M (d\psi,d\chi)_{g_\varphi}\,\frac{\omega_\varphi^n}{n!}.
\]
Its Levi-Civita derivative is given explicitly by
\[
D_t d\psi = d\dot\psi + 2\,T_a\!\big(C[\dot\varphi]*d\psi\big),
\]
with
\[
C[f]=( \Delta f)\,\omega_\varphi - i\partial\bar\partial f.
\]
The paper derives explicit sectional-curvature formulas and proves that the curvature of any \(2\)-plane is bounded above and below by a constant depending on the base point and one tangent direction; it also states that when \(M\) is a Riemann surface, the Dirichlet metric is flat. The K-energy is shown to be convex at a constant scalar curvature Kähler point, and the pseudo-Calabi flow is described as the gradient flow of the Mabuchi K-energy for this metric [1202.6610].

A later paper adopts the name gradient metric for the same basic construction on the space of normalized Kähler potentials,
\[
g_G(v_1,v_2)_\varphi = \int_M (\nabla^{\omega_\varphi} v_1,\nabla^{\omega_\varphi} v_2)_{\omega_\varphi}\,\omega_\varphi^n,
\]
with geodesic equation
\[
2\,\Delta_\varphi \varphi'' - |\nabla \varphi'|^2_{\varphi} + (\Delta_\varphi\varphi')^2=0.
\]
That work proves local well-posedness of the geodesic Cauchy problem for the gradient metric and for the sum metric
\[
g = a\,g_M + b\,g_G + c\,g_C.
\]
It also establishes a Rauch-type comparison theorem with the Calabi metric and shows that, in Sasakian geometry, the Ebin metric restricted to type II deformations satisfies
\[
I^* g_E = 2\,(g_C + g_G).
\]
The paper explicitly presents the Dirichlet/gradient metric as a Sobolev- or energy-type weak Riemannian structure built from first derivatives of potentials [1405.1211].

Taken together, these Kähler and Sasakian developments provide the clearest antecedents of the \(G_2\) theory. The common feature is a weak Riemannian metric defined through first-order data—gradients, exact forms, or canonical potentials—rather than through zeroth-order \(L^2\)-pairings alone.

## 6. Related but distinct “Dirichlet-type” constructions

The phrase “Dirichlet-type” has several other established meanings that should not be conflated with Bryant’s \(G_2\)-metric.

In one complex variable, Dirichlet type spaces on the disk are analytic function spaces. One family is
\[
D^p_{p-1}=\left\{f\in H(\mathbb D): |f(0)|^p + \int_{\mathbb D} |f'(z)|^p (1-|z|^2)^{p-1}\, dA(z) <\infty \right\},
\]
and the embedding
\[
i: D^p_{p-1}\longrightarrow T^p_q(\mu)
\]
is bounded if and only if \(\mu\) is a \((q+1)\)-Carleson measure; compactness is characterized by vanishing \((q+1)\)-Carleson measures. The same paper relates these spaces to weighted Bergman spaces, the Bloch space, Volterra-type integral operators, and multiplier theory [1811.05172]. A different Dirichlet-type space \(D(\mu)\) is defined by
\[
\int_{\mathbb D} |f'(z)|^2 P_\mu(z)\, dA(z) < \infty,
\]
and is characterized by a double-integral criterion, by mean oscillation in the Bergman metric, by higher-order derivatives, and by an atomic decomposition adapted to Bergman balls and separated sequences [1304.5958]. These are function-space geometries, not Riemannian metrics on moduli spaces.

In geometric mapping theory, a Dirichlet-type energy on annuli is
\[
\mathcal E[h]=\int_{\mathbb A}\|Dh(x)\|^{\,n-1}\,|h(x)|^{\,n-1}\,dx,
\]
with auxiliary combined energy
\[
\mathbb E[a,b][h] = \int_{\mathbb A}\left( a^2\rho^{n-1}(x)|DS(x)|^{n-1} + b^2|\nabla \rho(x)|^{n-1}\rho^{n-1}(x) \right)\,dx.
\]
For \(n\ge 4\), the infimum of \(\mathcal E[h]\) is not attained in the full homeomorphism class, but there is a minimizing sequence inside the generalized radial class [2009.13617]. This is again a variational energy for maps, not Bryant’s metric on a space of geometric structures.

Finally, Bryant–Tupper diversity theory is a separate axiomatic framework. There a diversity is a set-function on finite subsets satisfying non-degeneracy and a triangle inequality, and recent work shows that metric complexity yields such a diversity via
\[
\kappa^t:=\exp\{C^t\}-1.
\]
The paper proves that both \(\kappa^t\) and \(C^t\) are diversities and studies Minkowski-superlinearity on compact subsets of \(\mathbb R\) [2507.09698]. Despite the shared name “Bryant,” this topic concerns Bryant–Tupper diversities rather than Bryant’s Dirichlet-type metric.

The terminological boundary is therefore sharp. In current geometric usage, Bryant’s Dirichlet-type metric refers most specifically to the formal \(L^2\)-metric on the space of closed \(G_2\)-structures built from canonical potentials and Hodge theory; earlier Kähler and Sasakian Dirichlet or gradient metrics are close analogues, while the function-space, annulus-energy, and diversity-theoretic notions are distinct constructions that share only the broader “Dirichlet-type” vocabulary.

Source: https://www.emergentmind.com/topics/bryant-s-dirichlet-type-metric