Almost Symmetric Submanifolds

• Almost symmetric submanifolds are submanifolds that display near-full symmetry by fixing the affine normal space and having a one-dimensional invariant tangent direction.
• In the Euclidean framework, their homogeneous and inhomogeneous classifications reveal that almost symmetry tightly controls submanifold geometry, often resulting in cohomogeneity-one structures.
• Curvature-adapted approaches in symmetric spaces offer analogous definitions and rigidity results, opening pathways for further research on symmetry deformations and ambient curvature alignment.

Searching arXiv for recent and foundational papers on almost symmetric submanifolds and closely related notions. Almost symmetric submanifolds form a family of submanifolds situated strictly between arbitrary submanifolds and fully symmetric or extrinsically symmetric models. In the Euclidean setting introduced by Gorodski and Olmos, an almost symmetry at a point $p\in M\subset V$ is an involutive ambient isometry $\sigma_p\in \operatorname{Iso}(V)$ such that $\sigma_p(M)=M$, $\sigma_p$ fixes the affine normal space $p+\nu_pM$ pointwise, and the fixed point set of $(d\sigma_p)_p$ in $T_pM$ is $1$-dimensional (Gorodski et al., 27 Sep 2025). This weakens extrinsic symmetry by allowing one tangent direction to remain fixed rather than requiring the whole tangent space to be reflected. Closely related frameworks appear in curvature-adapted submanifold theory in symmetric spaces, where the extrinsic geometry is forced to align with the ambient curvature decomposition, yielding a strong form of “almost symmetry” without requiring homogeneity (Murphy, 2011), and in the recent global Euclidean classification of inhomogeneous examples (Gorodski et al., 12 Jan 2026).

1. Euclidean definition and basic geometric meaning

In Euclidean space $V$, a submanifold $S\subset V$ is extrinsically symmetric if for every $\sigma_p\in \operatorname{Iso}(V)$0, the orthogonal reflection in the affine normal space $\sigma_p\in \operatorname{Iso}(V)$1 preserves $\sigma_p\in \operatorname{Iso}(V)$2. By Ferus, these are exactly the Euclidean submanifolds with parallel second fundamental form, a class used as the comparison model in the almost symmetric theory (Gorodski et al., 12 Jan 2026). The Euclidean notion of almost symmetry modifies only the tangent action: the ambient involution still fixes $\sigma_p\in \operatorname{Iso}(V)$3 pointwise, but on $\sigma_p\in \operatorname{Iso}(V)$4 it acts as $\sigma_p\in \operatorname{Iso}(V)$5 on a line and $\sigma_p\in \operatorname{Iso}(V)$6 on its orthogonal complement (Gorodski et al., 27 Sep 2025, Gorodski et al., 12 Jan 2026).

This definition is explicitly extrinsic. It depends on ambient Euclidean isometries, affine normal spaces, and the induced decomposition

$\sigma_p\in \operatorname{Iso}(V)$7

where $\sigma_p\in \operatorname{Iso}(V)$8 is the $\sigma_p\in \operatorname{Iso}(V)$9-dimensional tangent fixed-point set of $\sigma_p(M)=M$0 (Gorodski et al., 27 Sep 2025). In this sense, almost symmetric submanifolds are “one tangent direction away” from extrinsic symmetry.

The homogeneous theory also admits a broader “co-index of extrinsic symmetry” viewpoint. In the terminology of Gorodski and Olmos, symmetric submanifolds have co-index $\sigma_p(M)=M$1, while almost symmetric submanifolds have co-index $\sigma_p(M)=M$2 (Gorodski et al., 27 Sep 2025). This suggests a hierarchy of increasingly weaker reflectional symmetries, though only the co-index-one case is developed there.

A separate but influential geometric interpretation comes from symmetric-space submanifold theory. Murphy’s curvature-adapted condition requires, for each normal direction $\sigma_p(M)=M$3, that the normal Jacobi operator

$\sigma_p(M)=M$4

preserve $\sigma_p(M)=M$5 and commute with the shape operator

$\sigma_p(M)=M$6

For hypersurfaces this reduces to simultaneous diagonalizability of $\sigma_p(M)=M$7 and $\sigma_p(M)=M$8, so principal curvature directions are aligned with ambient curvature eigendirections (Murphy, 2011). This is not the same definition as Euclidean almost symmetry, but it provides a widely used symmetric-space analogue of “almost symmetric” behavior.

2. Homogeneous Euclidean theory

The first systematic compact homogeneous theory is due to Gorodski and Olmos (Gorodski et al., 27 Sep 2025). Let $\sigma_p(M)=M$9 be a full irreducible compact homogeneous submanifold of Euclidean space of codimension at least two. Their structural theorem states that if $\sigma_p$0 is almost symmetric, then in codimension greater than three it must be an orbit of an irreducible $\sigma_p$1-representation (Gorodski et al., 27 Sep 2025). Thus high-codimension compact homogeneous almost symmetric submanifolds remain tightly controlled by the classical orbit geometry of symmetric spaces.

More precisely, if such a homogeneous almost symmetric submanifold is an orbit of an $\sigma_p$2-representation, then it is either a most singular orbit or an almost most singular orbit (Gorodski et al., 27 Sep 2025). The almost most singular case admits a geometric description as a partial holonomy tube $\sigma_p$3 over a symmetric submanifold $\sigma_p$4, where $\sigma_p$5 lies in a $\sigma_p$6-dimensional irreducible factor of the normal holonomy of $\sigma_p$7. Conversely, any holonomy tube over a symmetric submanifold through a sufficiently small vector in a two dimensional factor of the normal holonomy is an almost symmetric submanifold (Gorodski et al., 27 Sep 2025).

A key mechanism in the proof is uniqueness of the almost symmetry outside the $\sigma_p$8-orbit case. If there are two distinct almost symmetries at a point, then the submanifold is forced to be an orbit of an $\sigma_p$9-representation (Gorodski et al., 27 Sep 2025). This uses the homogeneous normal connection and the distribution

$p+\nu_pM$0

where $p+\nu_pM$1 compares the normal connection with the canonical $p+\nu_pM$2-invariant one. When $p+\nu_pM$3, equivalently $p+\nu_pM$4, one recovers the Olmos–Sánchez criterion that the homogeneous submanifold is an orbit of an irreducible $p+\nu_pM$5-representation (Gorodski et al., 27 Sep 2025).

The compact homogeneous classification itself splits into $p+\nu_pM$6-orbit examples and three additional families of principal orbits of reducible cohomogeneity-three representations (Gorodski et al., 27 Sep 2025). The $p+\nu_pM$7-orbit list contains 15 families, including $p+\nu_pM$8, $p+\nu_pM$9, $(d\sigma_p)_p$0, $(d\sigma_p)_p$1, $(d\sigma_p)_p$2, various real and complex Grassmannian products, and exceptional examples involving $(d\sigma_p)_p$3, $(d\sigma_p)_p$4, and $(d\sigma_p)_p$5 (Gorodski et al., 27 Sep 2025). The non-$(d\sigma_p)_p$6-orbit families are $(d\sigma_p)_p$7, $(d\sigma_p)_p$8, and $(d\sigma_p)_p$9, each realized as an arbitrary principal orbit of a reducible cohomogeneity-three representation (Gorodski et al., 27 Sep 2025).

An important intrinsic byproduct is that, except for one $T_pM$0-dimensional odd-contact example, the classified compact homogeneous almost symmetric submanifolds are embeddings of contact sub-Riemannian symmetric spaces, and their $T_pM$1-quotients are compact Hermitian symmetric spaces (Gorodski et al., 27 Sep 2025). This places Euclidean almost symmetry in direct contact with intrinsic symmetric geometries weaker than full Riemannian symmetry.

3. Inhomogeneous Euclidean classification

The global inhomogeneous theory is completed in “Inhomogeneous almost symmetric submanifolds” (Gorodski et al., 12 Jan 2026). There the ambient space is Euclidean, and the main objects are full irreducible properly embedded almost symmetric submanifolds. The principal conclusion is a complete description of the inhomogeneous case: every full irreducible inhomogeneous properly embedded almost symmetric submanifold is a union of parallel symmetric submanifolds, parametrized by a connected properly embedded $T_pM$2-dimensional submanifold $T_pM$3 in a flat section (Gorodski et al., 12 Jan 2026).

The geometric model is built as follows. One decomposes

$T_pM$4

where for each $T_pM$5, $T_pM$6 acts irreducibly on $T_pM$7 as an $T_pM$8-representation, and $T_pM$9 is a fixed nontrivial extrinsic symmetric orbit (Gorodski et al., 12 Jan 2026). Let

$1$0

and let $1$1 be a connected properly embedded $1$2-dimensional submanifold through $1$3, satisfying nonconstancy and nonvanishing conditions on its projections to each $1$4, and fullness of the projection to $1$5 (Gorodski et al., 12 Jan 2026). Then $1$6 is a full irreducible properly embedded almost symmetric submanifold.

The general classification theorem states that every full irreducible inhomogeneous properly embedded almost symmetric submanifold of Euclidean space is of one of the forms

$1$7

with at most two extra rotational factors corresponding to singular orbits of the cohomogeneity-one action (Gorodski et al., 12 Jan 2026). These examples are inhomogeneous unless they are round spheres (Gorodski et al., 12 Jan 2026).

The essential geometry is cohomogeneity one. Let $1$8 be the closure of the group generated by the almost symmetries. Then $1$9 acts with cohomogeneity one, a fact already established in the earlier paper and used as input here (Gorodski et al., 12 Jan 2026). For a regular point $V$0, the orbit

$V$1

is a compact extrinsic symmetric submanifold. Choosing a normal vector $V$2 to $V$3 inside $V$4, the normal geodesic $V$5 defines a transverse curve

$V$6

with

$V$7

Since $V$8 meets all $V$9-orbits, one has

$S\subset V$0

and each $S\subset V$1 is a parallel symmetric submanifold of $S\subset V$2 in $S\subset V$3 (Gorodski et al., 12 Jan 2026). This is the precise meaning of the “union of parallel symmetric submanifolds” description.

The singular-orbit analysis is encoded in the scalar functions $S\subset V$4. If some $S\subset V$5, then $S\subset V$6 lies on a singular orbit, the corresponding factor is a standard $S\subset V$7-representation, and the curve $S\subset V$8 is invariant under reflection in the hyperplane $S\subset V$9 (Gorodski et al., 12 Jan 2026). Since a cohomogeneity one action has at most two singular orbits, there can be at most two such extra rotational factors (Gorodski et al., 12 Jan 2026).

4. Intrinsic almost symmetric spaces and cohomogeneity one

The intrinsic theory in the Euclidean program abstracts away the embedding. A complete Riemannian manifold is almost symmetric if for each point there is an involutive isometry fixing the point and having $\sigma_p\in \operatorname{Iso}(V)$00-dimensional fixed-point set in the tangent space (Gorodski et al., 27 Sep 2025). Let $\sigma_p\in \operatorname{Iso}(V)$01 be the closure of the group generated by the almost symmetries. If $\sigma_p\in \operatorname{Iso}(V)$02 is not transitive, then $\sigma_p\in \operatorname{Iso}(V)$03 acts with cohomogeneity $\sigma_p\in \operatorname{Iso}(V)$04 (Gorodski et al., 27 Sep 2025).

This cohomogeneity-one theorem is one of the most rigid consequences of the notion. On the open dense regular set $\sigma_p\in \operatorname{Iso}(V)$05, the almost symmetry at a point is unique (Gorodski et al., 27 Sep 2025). If $\sigma_p\in \operatorname{Iso}(V)$06 is a $\sigma_p\in \operatorname{Iso}(V)$07-orbit of codimension $\sigma_p\in \operatorname{Iso}(V)$08, then there is a connected normal subgroup $\sigma_p\in \operatorname{Iso}(V)$09 acting trivially on $\sigma_p\in \operatorname{Iso}(V)$10, whose isotropy representation on

$\sigma_p\in \operatorname{Iso}(V)$11

is $\sigma_p\in \operatorname{Iso}(V)$12. In particular, the isotropy algebra contains an ideal isomorphic to $\sigma_p\in \operatorname{Iso}(V)$13 (Gorodski et al., 27 Sep 2025). The connected components of singular $\sigma_p\in \operatorname{Iso}(V)$14-orbits are totally geodesic, and the connected components of all $\sigma_p\in \operatorname{Iso}(V)$15-orbits are intrinsic symmetric spaces (Gorodski et al., 27 Sep 2025).

The simply connected complete inhomogeneous intrinsic classification then states that such a manifold is one of three types (Gorodski et al., 27 Sep 2025):

Type	Model	Symmetric ingredient
Multiply warped	$\sigma_p\in \operatorname{Iso}(V)$16	$\sigma_p\in \operatorname{Iso}(V)$17 simply-connected symmetric
Product with Euclidean factor	$\sigma_p\in \operatorname{Iso}(V)$18	$\sigma_p\in \operatorname{Iso}(V)$19 simply-connected symmetric
Product with spherical factor	$\sigma_p\in \operatorname{Iso}(V)$20	$\sigma_p\in \operatorname{Iso}(V)$21 simply-connected symmetric

In the first case, each component of the warping map scales independently each irreducible factor of $\sigma_p\in \operatorname{Iso}(V)$22 (Gorodski et al., 27 Sep 2025). This shows that even intrinsically, inhomogeneous almost symmetric spaces are assembled from symmetric pieces arranged in a cohomogeneity-one manner.

5. Curvature-adapted submanifolds as symmetric-space analogues

The phrase “almost symmetric submanifold” is also used informally for submanifolds of symmetric spaces whose principal curvature directions are dictated by the ambient curvature. Murphy’s study of curvature-adapted submanifolds makes this precise (Murphy, 2011). A submanifold $\sigma_p\in \operatorname{Iso}(V)$23 is curvature-adapted if, for every point $\sigma_p\in \operatorname{Iso}(V)$24 and every unit normal vector $\sigma_p\in \operatorname{Iso}(V)$25,

$\sigma_p\in \operatorname{Iso}(V)$26

and

$\sigma_p\in \operatorname{Iso}(V)$27

For hypersurfaces the first condition is automatic, so curvature-adaptedness reduces to commutation of the normal Jacobi operator and shape operator (Murphy, 2011).

Because $\sigma_p\in \operatorname{Iso}(V)$28 and $\sigma_p\in \operatorname{Iso}(V)$29 are self-adjoint, they are simultaneously diagonalizable. In compact type one can choose a common orthonormal eigenbasis $\sigma_p\in \operatorname{Iso}(V)$30 with

$\sigma_p\in \operatorname{Iso}(V)$31

This alignment of the extrinsic and ambient spectral decompositions is the fundamental “almost symmetric” feature in the symmetric-space setting (Murphy, 2011).

The Riccati equation for parallel hypersurfaces then simplifies drastically. Along a normal geodesic $\sigma_p\in \operatorname{Iso}(V)$32, if $\sigma_p\in \operatorname{Iso}(V)$33 is the shape operator of the parallel hypersurface $\sigma_p\in \operatorname{Iso}(V)$34, then

$\sigma_p\in \operatorname{Iso}(V)$35

and curvature-adaptedness scalarizes this into

$\sigma_p\in \operatorname{Iso}(V)$36

in compact type, with explicit solution

$\sigma_p\in \operatorname{Iso}(V)$37

where $\sigma_p\in \operatorname{Iso}(V)$38 (Murphy, 2011). Thus focal geometry is encoded directly by the common eigendirections and their scalar Riccati evolution.

Murphy’s Theorem 1.3 states that for a curvature-adapted hypersurface of a compact symmetric space,

$\sigma_p\in \operatorname{Iso}(V)$39

In rank one, the Jacobi eigenvalues are automatically constant, recovering the familiar equivalence between isoparametricity and constant principal curvatures for complete curvature-adapted hypersurfaces (Murphy, 2011).

The paper then develops classification results in concrete symmetric spaces. In the Cayley projective plane,

$\sigma_p\in \operatorname{Iso}(V)$40

satisfies

$\sigma_p\in \operatorname{Iso}(V)$41

(Theorem 1.4) (Murphy, 2011). Thus constant-principal-curvature curvature-adapted hypersurfaces in $\sigma_p\in \operatorname{Iso}(V)$42 are exactly homogeneous tubes around totally geodesic singular orbits.

The complex two-plane Grassmannian

$\sigma_p\in \operatorname{Iso}(V)$43

provides a contrasting higher-rank case. With its simultaneous Kähler structure $\sigma_p\in \operatorname{Iso}(V)$44 and quaternionic-Kähler structure $\sigma_p\in \operatorname{Iso}(V)$45, Murphy proves strong partial nonexistence results for generic-angle curvature-adapted hypersurfaces. In particular, Theorem 1.5 excludes curvature-adapted hypersurfaces satisfying

$\sigma_p\in \operatorname{Iso}(V)$46

under a natural constancy assumption along normal geodesics (Murphy, 2011). Similar exclusions hold in the noncompact dual (Murphy, 2011). This shows that higher-rank almost-symmetry phenomena are highly constrained and often collapse back to known homogeneous models.

Koike pushes this perspective further for higher codimension (Koike, 2013). In a symmetric space $\sigma_p\in \operatorname{Iso}(V)$47 of compact or noncompact type, he considers complete curvature-adapted submanifolds with maximal flat section and trivial normal holonomy. Writing the curvature eigendistributions as $\sigma_p\in \operatorname{Iso}(V)$48 and the shape eigendistributions as $\sigma_p\in \operatorname{Iso}(V)$49, he proves that for every parallel normal field $\sigma_p\in \operatorname{Iso}(V)$50, the eigenvalues $\sigma_p\in \operatorname{Iso}(V)$51 of the normal Jacobi operator are constant on $\sigma_p\in \operatorname{Iso}(V)$52 (Theorem A) (Koike, 2013). Under the additional inclusion assumption

$\sigma_p\in \operatorname{Iso}(V)$53

Theorem B yields global constancy of all principal curvatures $\sigma_p\in \operatorname{Iso}(V)$54, and hence isoparametricity in compact type, or in noncompact type under real analyticity (Koike, 2013).

The strongest conclusions identify such submanifolds with genuinely symmetric homogeneous models. In compact type and rank $\sigma_p\in \operatorname{Iso}(V)$55, Theorem C states that $\sigma_p\in \operatorname{Iso}(V)$56 is congruent to a principal orbit of the isotropy action of $\sigma_p\in \operatorname{Iso}(V)$57 (Koike, 2013). In noncompact type and rank $\sigma_p\in \operatorname{Iso}(V)$58, Theorem D identifies $\sigma_p\in \operatorname{Iso}(V)$59 as a principal orbit of a Hermann action under an additional boundary condition (Koike, 2013). Thus sufficiently rigid ambient-curvature alignment forces full symmetry.

6. Related rigidity benchmarks and neighboring theories

Several nearby theories clarify the boundary between exact symmetry, almost symmetry, and weaker curvature constraints.

In Euclidean geometry, extrinsically symmetric submanifolds provide the exact reference class. The paper “Extrinsically Symmetric Spaces, Submanifolds of Clifford Type and a Theorem of Harish-Chandra” proves that a connected submanifold $\sigma_p\in \operatorname{Iso}(V)$60 is a compact extrinsically symmetric space iff it is intrinsically a compact symmetric space and every maximal torus of $\sigma_p\in \operatorname{Iso}(V)$61 is a Clifford torus in $\sigma_p\in \operatorname{Iso}(V)$62 (Eschenburg et al., 26 Feb 2025). This criterion detects exact ambient symmetry via the geometry of maximal flats, and thereby furnishes a sharp boundary point for almost symmetric Euclidean phenomena.

The geometry of totally geodesic submanifolds of symmetric spaces serves as another benchmark. Berndt and Olmos define the index

$\sigma_p\in \operatorname{Iso}(V)$63

and prove

$\sigma_p\in \operatorname{Iso}(V)$64

for irreducible symmetric spaces (Berndt et al., 2014). Their classification of spaces with equality $\sigma_p\in \operatorname{Iso}(V)$65 and of maximal totally geodesic submanifolds identifies the most rigid large symmetric pieces available in an ambient symmetric space (Berndt et al., 2014). For almost symmetric theories this gives a natural codimension benchmark: genuinely symmetric submanifolds are already rare in low codimension.

In the same vein, Chen–Nagano theory and later stability analysis show that not every exactly symmetric submanifold is variationally rigid. The paper “Totally geodesic submanifolds of symmetric spaces, III” develops a Casimir-operator method for deciding stability of compact totally geodesic submanifolds, especially the basic $\sigma_p\in \operatorname{Iso}(V)$66 and $\sigma_p\in \operatorname{Iso}(V)$67 submanifolds (Chen et al., 2013). This is relevant because almost symmetric submanifolds should be understood not only through symmetry conditions but also through deformation modes around exact symmetric models.

Another neighboring rigidity notion is the CPC condition: the principal curvature spectrum is independent of the chosen unit normal direction. In irreducible noncompact symmetric spaces of rank at least two, Berndt and Sanmartín-López classify many homogeneous CPC submanifolds, proving they are always austere and minimal and include totally geodesic submanifolds, homogeneous austere hypersurfaces, and singular orbits of cohomogeneity one actions (Berndt et al., 2018). Although CPC does not coincide with almost symmetry, it captures a strong normal-isotropy condition reminiscent of symmetric behavior.

Finally, the 2026 paper “Attached Submanifolds Beyond Symmetric Spaces” extends Tamaru’s attached-submanifold construction from noncompact symmetric spaces to solvmanifolds of strong Iwasawa type (Kerr et al., 19 Jan 2026). These submanifolds preserve several symmetry-like features—minimality, Ricci restriction, and a root-orthogonality total-geodesy criterion—without requiring a symmetric ambient space. Their central algebraic criterion, the Jacobi Star Condition, characterizes when the intrinsic Ricci tensor of the submanifold equals the restriction of the ambient Ricci tensor (Kerr et al., 19 Jan 2026). This offers another example of how a residue of ambient symmetry can be isolated and used to define an “almost symmetric” class beyond the classical setting.

7. Scope, interpretations, and open directions

The term “almost symmetric submanifold” is therefore used in two distinct but compatible ways in current research. The first is the precise Euclidean notion of Gorodski and Olmos: ambient involutions fix affine normal spaces pointwise and have $\sigma_p\in \operatorname{Iso}(V)$68-dimensional fixed tangent spaces (Gorodski et al., 27 Sep 2025, Gorodski et al., 12 Jan 2026). The second is an interpretive symmetric-space notion: submanifolds need not be homogeneous or totally geodesic, but their extrinsic geometry is so tightly synchronized with ambient curvature or root data that they behave like deformations or relatives of symmetric orbits (Murphy, 2011, Koike, 2013, Kerr et al., 19 Jan 2026).

A common misconception is that almost symmetry should mean “slightly nonhomogeneous” or “approximately invariant under many isometries.” The Euclidean classification shows that the condition is much more rigid: inhomogeneous examples are still cohomogeneity one and globally determined by a properly embedded curve sweeping out parallel symmetric orbits (Gorodski et al., 12 Jan 2026). Similarly, in symmetric-space settings, curvature-adaptedness combined with flat normal geometry often forces full isoparametricity or actual identification with principal orbits of isotropy or Hermann actions (Koike, 2013).

Another misconception is that almost symmetry is merely intrinsic. In the Euclidean definition it is decisively extrinsic, depending on affine normal spaces and ambient involutions (Gorodski et al., 27 Sep 2025). By contrast, the intrinsic cohomogeneity-one theorem concerns almost symmetric spaces rather than submanifolds, and its classification must be kept separate (Gorodski et al., 27 Sep 2025).

Several open directions are explicit in the literature. Gorodski and Olmos suggest studying higher co-index of extrinsic symmetry beyond the co-index-one case (Gorodski et al., 27 Sep 2025). Murphy’s symmetric-space analysis points toward completing the classification in $\sigma_p\in \operatorname{Iso}(V)$69, removing technical hypotheses in Grassmannian nonexistence results, and extending curvature-adapted/equifocal theory to broader higher-codimension settings (Murphy, 2011). Koike’s work indicates that further understanding of normal holonomy and flat-section assumptions could clarify how far one can weaken homogeneity before principal-orbit behavior becomes unavoidable (Koike, 2013). The attached-submanifold program suggests that “almost symmetric” geometry may persist in solvable ambient spaces whenever the appropriate root-theoretic algebra survives (Kerr et al., 19 Jan 2026).

Taken together, these developments show that almost symmetric submanifolds are not a diffuse metaphor but a family of rigorously structured geometries. Whether defined via ambient involutions in Euclidean space or via curvature alignment in symmetric spaces, they occupy a narrow corridor between arbitrary submanifold geometry and full symmetry, where cohomogeneity, holonomy, focal structure, and root data become the dominant organizing principles.