03.09.18 · modern-geometry / spin-geometry

Berger holonomy classification and parallel spinors

shipped3 tiersLean: partial

Anchor (Master): Berger 1955 *Sur les groupes d'holonomie homogène* (Bull. Soc. Math. France 83); Wang 1989 *Parallel spinors and parallel forms* (Annals of Global Analysis and Geometry); Bryant 1987 *Metrics with exceptional holonomy* (Annals 126); Joyce 2007 *Riemannian Holonomy Groups and Calibrated Geometry*

Intuition [Beginner]

A Riemannian manifold has a notion of parallel transport: given a tangent vector at one point, one can transport it along a curve to a tangent vector at another point, in a way that preserves length and angle. The transport depends on the curve. If you go around a closed loop and come back to the starting point, the transported vector may differ from the original by a rotation. The collection of all such rotations, over all closed loops based at a fixed point, is a subgroup of the orthogonal group $O (n)$ called the holonomy group of the manifold.

For most metrics, the holonomy group is the full $SO (n)$ — the orthogonal group of all orientation-preserving rotations. Special metrics, however, have holonomy strictly smaller than $SO (n)$ . The classification of which subgroups of $SO (n)$ can occur as the holonomy of an irreducible non-symmetric Riemannian manifold was completed in 1955 by Marcel Berger. He proved that there are exactly nine possibilities: $SO (n)$ itself (the generic case), $U (n)$ , $SU (n)$ , $Sp (n)$ , $Sp (n) Sp (1)$ , plus the two exceptional cases $G_{2}$ in dimension 7 and $Spin (7)$ in dimension 8 — and the symmetric and reducible cases.

The deep connection to spin geometry is the Wang bijection of 1989: the holonomy groups on Berger's list that admit parallel spinors (sections of the spinor bundle that are everywhere preserved by parallel transport) are exactly the special holonomy groups $SU (n)$ , $Sp (n)$ , $G_{2}$ , and $Spin (7)$ . These are also exactly the holonomies whose ambient manifolds support Harvey-Lawson calibrated geometries. Berger's 1955 list, viewed through Wang's 1989 lens, is the structural foundation of the entire calibrated-geometry programme.

Visual [Beginner]

A Riemannian manifold with a closed loop, and a tangent vector being parallel-transported around the loop. Returning to the starting point, the vector has been rotated by an element of the holonomy group. To the side, a list of the nine possible holonomy groups: $SO (n)$ , $U (n)$ , $SU (n)$ , $Sp (n)$ , $Sp (n) Sp (1)$ , $G_{2}$ , $Spin (7)$ for irreducible non-symmetric metrics.

The four special-holonomy groups $SU (n), Sp (n), G_{2}, Spin (7)$ are exactly those that stabilise a non-zero spinor — the bridge to calibrated geometries.

Worked example [Beginner]

The complex projective space $C P^{n}$ with the Fubini-Study metric is the simplest example beyond $SO (n)$ . Its holonomy group is $U (n)$ , the unitary group, sitting inside $SO (2 n)$ as the stabiliser of the complex structure $J$ . Parallel transport on $C P^{n}$ preserves not just lengths and angles, but also the complex structure: a complex tangent vector remains complex after transport.

Why is the holonomy $U (n)$ rather than $SO (2 n)$ ? Because the Fubini-Study metric is Kähler: its symplectic form $ω$ is parallel. Preservation of $ω$ by parallel transport reduces the holonomy from $SO (2 n)$ to the symplectic-orthogonal subgroup, which on a Kähler manifold equals $U (n)$ .

If the manifold is Calabi-Yau — Kähler with vanishing first Chern class — the holonomy reduces further to $SU (n)$ , the special unitary group. The K3 surface in dimension 4, the Yau-Vrănceanu manifolds in higher dimension, and the Calabi-Yau threefolds of string theory are all examples. On these manifolds, a parallel spinor exists; this is the Wang bijection at work.

What this tells us: each entry on Berger's list corresponds to a structural feature of the manifold (Kähler structure, complex structure, hyperkähler structure, $G_{2}$ structure, etc.) preserved by parallel transport. Special holonomy is special geometry.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Holonomy group. Let $(M, g)$ be a connected Riemannian manifold with Levi-Civita connection $\nabla^{L C}$ . Fix a point $x \in M$ . The holonomy group at $x$ is $$ \mathrm{Hol}x(M, g) := \big{P\gamma : T_x M \to T_x M \mid \gamma \text{ a piecewise-smooth loop at } x\big} \subset \mathrm{O}(T_x M), $$ where $P_{γ}$ is parallel transport along $γ$ . It is a Lie subgroup of $O (T_{x} M)$ (Borel-Lichnerowicz reduction theorem). Choosing an orthonormal frame at $x$ identifies it with a subgroup of $O (n)$ , well-defined up to conjugation; the restricted holonomy $Hol_{x}^{0}$ — using only contractible loops — is connected and lies in $SO (n)$ .

A connected Riemannian manifold is irreducible if its tangent bundle decomposes only inert-ly under parallel transport, i.e., $Hol^{0}$ acts irreducibly on $T_{x} M$ . It is symmetric if $(M, g)$ is a Riemannian symmetric space — i.e., for every $x \in M$ there is an isometric involution fixing $x$ with isolated fixed point.

Theorem (Berger 1955 — holonomy classification). Let $(M, g)$ be a simply-connected, irreducible, non-symmetric Riemannian manifold of dimension $n$ . The restricted holonomy $Hol^{0}$ is one of:

$dim M$	Holonomy group	Geometry
$n$	$SO (n)$	generic
$2 n$ ( $n \geq 2$ )	$U (n) \subset SO (2 n)$	Kähler
$2 n$ ( $n \geq 2$ )	$SU (n) \subset SO (2 n)$	Calabi-Yau
$4 n$ ( $n \geq 2$ )	$Sp (n) \subset SO (4 n)$	hyperkähler
$4 n$ ( $n \geq 2$ )	$Sp (n) Sp (1) \subset SO (4 n)$	quaternionic Kähler
$7$	$G_{2} \subset SO (7)$	$G_{2}$ holonomy
$8$	$Spin (7) \subset SO (8)$	Spin(7) holonomy

The reducible case decomposes as a product of irreducibles; the symmetric case is the family of Riemannian symmetric spaces classified separately by Cartan. Quaternionic Kähler is sometimes excluded in modern statements (its scalar curvature is non-zero, distinguishing it from the others) ^{[Berger 1955]}.

Parallel spinor. Let $(M^{n}, g)$ be a Riemannian spin manifold. A spinor section $ψ \in Γ (S)$ is parallel if $\nabla^{L C, S} ψ = 0$ , where $\nabla^{L C, S}$ is the spin lift of the Levi-Civita connection. The space of parallel spinors is a vector space, of dimension $\geq 0$ , that is bounded by the dimension of the spinor representation $Δ_{n}$ .

Theorem (Wang 1989 — bijection holonomy ↔ parallel spinors). A simply-connected Riemannian spin manifold $(M^{n}, g)$ admits a non-zero parallel spinor if and only if its restricted holonomy is one of $SU (n /2), Sp (n /4), G_{2}$ (in dim 7), $Spin (7)$ (in dim 8). The dimension of the space of parallel spinors equals:

Holonomy	Dimension of $M$	Parallel spinors
$SU (n)$	$2 n$	2 (if $n$ even) or 2 (if $n$ odd)
$Sp (n)$	$4 n$	$n + 1$
$G_{2}$	7	1
$Spin (7)$	8	1

^{[Wang 1989]}. A non-zero parallel spinor is precisely a structure-group reduction to a Wang-special holonomy; this identifies parallel-spinor existence with a holonomy classification, putting these together gives the foundational Wang bijection.

Key theorem with proof [Intermediate+]

Theorem (Berger 1955). Let $(M^{n}, g)$ be a simply-connected, irreducible, non-symmetric Riemannian manifold. The restricted holonomy $Hol^{0} (M, g)$ is one of the seven groups listed above.

Proof sketch. Berger's argument is essentially a representation-theoretic case analysis. The Ambrose-Singer theorem (1953) identifies the Lie algebra $hol$ of $Hol^{0}$ as the span of all $R (X, Y)$ for $X, Y$ tangent vectors, where $R$ is the Riemann curvature tensor. So $hol$ is a Lie subalgebra of $so (n)$ generated by curvature tensors satisfying the Bianchi identities: $R (X, Y) = - R (Y, X)$ , $R (X, Y, Z, W) = R (Z, W, X, Y)$ , and the first Bianchi identity.

The classification reduces to: for which Lie subalgebras $g \subset so (n)$ is the space of curvature tensors satisfying the Bianchi identity non-zero and acting irreducibly on $R^{n}$ ? Berger went through every irreducible subalgebra of $so (n)$ (the irreducible Lie algebra subalgebras of $so (n)$ being a known classification due to Cartan and others) and computed the Bianchi-compatible curvature space dimension-by-dimension. Most subalgebras have zero curvature space — meaning the only manifold with that holonomy is locally symmetric. The seven cases above are exactly those with non-zero curvature space and non-symmetric examples.

Berger's original 1955 paper went through all the cases by hand. The argument was later simplified by Simons (1962), who introduced the holonomy reduction theorem, and by Olmos (2005), who gave a representation-theoretic proof. The key technical input is the first Bianchi identity: the cyclic sum of $R (X, Y) Z$ over the three cyclic permutations of $X, Y, Z$ vanishes. This identity provides strong constraints on the curvature tensor and rules out most subalgebras as candidate holonomies.

The case analysis: $SO (n)$ (generic, no constraint); $U (n)$ in $SO (2 n)$ (preserves a complex structure, equivalently a Kähler form $ω$ ); $SU (n)$ in $SO (2 n)$ (additionally preserves a complex volume form $Ω$ , hence Calabi-Yau); $Sp (n)$ in $SO (4 n)$ (preserves three Kähler forms $I, J, K$ satisfying quaternion relations, hence hyperkähler); $Sp (n) Sp (1)$ in $SO (4 n)$ (preserves a 4-dimensional family of compatible complex structures, but not individually parallel — quaternionic Kähler); $G_{2}$ in $SO (7)$ (preserves the Cayley 3-form on $Im O$ ); $Spin (7)$ in $SO (8)$ (preserves the Cayley 4-form on $O$ ). $□$

Theorem (Wang 1989). Let $(M^{n}, g)$ be a simply-connected Riemannian spin manifold. There exists a non-zero parallel spinor on $M$ if and only if the restricted holonomy is one of $SU (m)$ (in dim $2 m$ ), $Sp (m)$ (in dim $4 m$ ), $G_{2}$ (in dim 7), or $Spin (7)$ (in dim 8).

Proof. (One direction.) A parallel spinor $ψ$ on $(M, g)$ has $\nabla^{L C, S} ψ = 0$ . Compute the holonomy at $x$ : parallel transport around any loop preserves $ψ (x) \in S_{x} ≅ Δ_{n}$ , so $Hol^{0}$ lies in the stabiliser of $ψ (x)$ inside $Spin (n)$ . The stabiliser of a generic spinor in $Spin (n)$ is computed case-by-case and turns out to be one of $SU (n /2), Sp (n /4), G_{2} \subset Spin (7), Spin (7) \subset Spin (8)$ , depending on dimension and on whether $n$ is even or odd. The other cases of Berger's list ( $SO (n)$ , $U (n)$ , $Sp (n) Sp (1)$ ) do not stabilise any non-zero spinor in $Δ_{n}$ .

(The other direction.) When the holonomy is one of $SU (m), Sp (m), G_{2}, Spin (7)$ , one constructs a parallel spinor explicitly: in each case the spinor representation $Δ_{n}$ contains a one-dimensional (or higher) inert subrepresentation under the holonomy action, by direct branching-rule calculation. Choosing $ψ (x)$ in this inert subspace and parallel-transporting gives a globally parallel section.

The precise count of parallel spinors per holonomy is a representation-theoretic calculation: how many inert summands does the spin representation contain under the inclusion of the holonomy group into $Spin (n)$ . For $SU (n) ↪ Spin (2 n)$ : 2 inert summands. For $Sp (n) ↪ Spin (4 n)$ : $n + 1$ inert summands (the symmetric powers of the defining representation of $Sp (n)$ all carry parallel spinors). For $G_{2} ↪ Spin (7)$ : 1 inert summand. For $Spin (7) ↪ Spin (8)$ (the spin embedding into the diagonal): 1 inert summand. $□$

Bridge. The construction here builds toward later units of the strand, where the same pattern is taken up at higher structure. The defining pattern appears again in those units in a sharpened form, where the local data is glued or quotiented. Putting these together, the foundational insight is that the data of this unit gives the structural signature that the rest of the strand reads off.

Exercises [Intermediate+]

Exercise 3 (medium, proof).

Show that on a simply-connected Calabi-Yau $n$ -fold (holonomy $SU (n)$ ), the space of parallel spinors has dimension exactly 2.

Hint

Compute the dimension of the inert subrepresentation of $SU (n)$ inside $Δ_{2 n}$ , the spin representation in dimension $2 n$ .

Answer

The spin representation $Δ_{2 n}$ has dimension $2^{n}$ (complex). Under the inclusion $SU (n) ↪ Spin (2 n)$ , $Δ_{2 n}$ branches as follows. The complex structure $J$ on $R^{2 n}$ identifies the standard $Spin (2 n)$ -spinor space with $Λ^{∙} C^{n} = ⨁_{k} Λ^{k} C^{n}$ , on which $U (n)$ acts as $det^{- 1/2}$ tensored with the standard exterior power.

Restricting to $SU (n)$ kills the $det$ factor, and the inert subrepresentations of $SU (n)$ inside $Λ^{∙} C^{n}$ are exactly the top and bottom exterior powers $Λ^{0} C^{n}$ and $Λ^{n} C^{n}$ , both one-dimensional. So the space of parallel spinors has dimension 2 — generated by the two "extreme" exterior powers.

Geometrically, the two parallel spinors on a Calabi-Yau correspond to the constant function $1$ and the holomorphic volume form $Ω$ , respectively. Both are preserved by $SU (n)$ but not by the larger $U (n)$ . $□$

Exercise 4 (medium, proof).

Show that the holonomy of a hyperkähler manifold is contained in $Sp (n)$ and not in $Sp (n) Sp (1)$ strictly.

Hint

A hyperkähler manifold has three parallel complex structures $I, J, K$ with $I J = K$ etc. Their stabiliser inside $SO (4 n)$ is $Sp (n)$ .

Answer

A hyperkähler manifold $(M, g, I, J, K)$ has three parallel complex structures satisfying the quaternion relations $I^{2} = J^{2} = K^{2} = I J K = - 1$ . Each $I, J, K$ defines a parallel Kähler form $ω_{I}, ω_{J}, ω_{K}$ . The stabiliser of all three forms inside $SO (4 n)$ is the symplectic group $Sp (n) \subset SO (4 n)$ — the group of orthogonal transformations preserving all three quaternionic structures simultaneously.

Parallel transport on a hyperkähler manifold preserves $g, I, J, K$ , hence preserves $ω_{I}, ω_{J}, ω_{K}$ , hence is contained in $Sp (n)$ . The strict containment $Sp (n) ⊊ Sp (n) Sp (1)$ holds because $Sp (n) Sp (1)$ acts on the quaternionic structure as a whole (the bundle of imaginary quaternions $Im H_{x}$ inside $End (T_{x} M)$ ) but mixes $I, J, K$ among themselves — so individual $I, J, K$ are not parallel under $Sp (n) Sp (1)$ .

A quaternionic Kähler manifold has each $I, J, K$ defined only locally (parallel only modulo the $Sp (1)$ rotation among them) and has scalar curvature non-zero; a hyperkähler manifold has $I, J, K$ globally parallel and is Ricci-flat. $□$

Exercise 5 (medium, symbolic).

Identify the parallel spinor on a $G_{2}$ manifold $(M^{7}, ϕ)$ in terms of the $G_{2}$ -fixed vector in the spin representation $Δ_{7}$ .

Hint

The spin representation $Δ_{7}$ has complex dimension 8; under $G_{2} ↪ Spin (7)$ , it branches as $R \oplus V$ where $V$ is the standard 7-dimensional $G_{2}$ -representation.

Answer

The spin representation $Δ_{7}$ of $Spin (7)$ is real of dimension 8. Under the inclusion $G_{2} ↪ Spin (7)$ , $Δ_{7}$ branches as $$ \Delta_7 \cong \mathbb{R} \oplus V_7, $$ where $R$ is the inert $G_{2}$ -representation and $V_{7}$ is the standard 7-dimensional $G_{2}$ -representation on $Im O$ . The inert summand is one-dimensional, generated by a unit spinor $ψ_{0} \in Δ_{7}$ with $G_{2} \cdot ψ_{0} = ψ_{0}$ .

On a $G_{2}$ manifold, the parallel spinor is the section corresponding to $ψ_{0}$ at every point. Its existence is equivalent to the holonomy reducing from $Spin (7)$ to $G_{2}$ , by the Wang bijection.

The associative 3-form $ϕ$ on $Im O$ is recovered from the parallel spinor by the spinor-square formula $ϕ (u, v, w) = ⟨ u \cdot v \cdot w \cdot ψ_{0}, ψ_{0} ⟩$ where $\cdot$ is Clifford multiplication. This is the spinor-square construction of the calibrating form for associative geometry 03.09.19.

Exercise 6 (hard, proof).

Sketch Berger's case-analysis that rules out $SU (2) \subset SO (5)$ as a candidate holonomy in dimension 5.

Hint

Compute the dimension of the space of curvature tensors $R \in Sym^{2} (Λ^{2} R^{5})$ satisfying the first Bianchi identity and lying in $Sym^{2} (su (2))$ . Show this space is zero unless the manifold is locally symmetric.

Answer

In dimension $n = 5$ , candidate holonomies are subgroups of $SO (5)$ acting irreducibly on $R^{5}$ . The Lie algebra $so (5)$ has rank 2 and dimension 10. Irreducible subalgebras of $so (5)$ that act irreducibly on $R^{5}$ : the full $so (5)$ itself, and the inclusion of $su (2)$ via the irreducible 5-dimensional representation (the symmetric tensor power $Sym^{4} (C^{2}) ∣_{R}$ ).

The space of curvature tensors is $Sym^{2} (hol)$ subject to the first Bianchi identity. For $hol = su (2)$ acting on $R^{5}$ via $Sym^{4} (C^{2}) ∣_{R}$ , decompose $Sym^{2} (su (2)) = Sym^{2} (V_{3})$ where $V_{3}$ is the 3-dimensional adjoint $su (2)$ -representation. By Clebsch-Gordan, $Sym^{2} (V_{3}) = V_{5} \oplus V_{1}$ (5- and 1-dimensional irreducibles).

The first Bianchi identity is a $SU (2)$ -equivariant linear constraint; computing its kernel in $V_{5} \oplus V_{1}$ : the constraint forces the $V_{1}$ part to vanish (after a representation-theoretic manipulation involving the standard Casimir), and projects the $V_{5}$ part to a sub-representation that turns out to be empty.

Conclusion: the space of curvature tensors compatible with $su (2)$ acting irreducibly on $R^{5}$ is zero. So any 5-manifold with this holonomy must be flat or locally symmetric, ruling it out as an irreducible non-symmetric example. Berger ran this calculation for every irreducible subalgebra of $so (n)$ in every dimension and identified the seven non-empty cases. $□$

Exercise 7 (hard, proof).

Show that a Calabi-Yau manifold is automatically Ricci-flat.

Hint

The Ricci tensor of a Kähler manifold is the curvature of the canonical line bundle $K_{M} = Λ^{n} T^{*} M$ . On a Calabi-Yau manifold, $K_{M}$ is holomorphically inert as a line bundle (the structure-sheaf class), so its first Chern class vanishes; the Ricci form representing $c_{1}$ is therefore exact, and on the unique compatible Kähler-Einstein metric (Yau 1977) it vanishes identically.

Answer

A Kähler manifold $(M, g, J, ω)$ has Ricci form $ρ = Ric (J \cdot, \cdot)$ closed and representing $c_{1} (M) \in H^{2} (M; R)$ (up to a factor of $2 π$ ). On a Calabi-Yau $n$ -fold, $c_{1} (M) = 0$ in $H^{2} (M; R)$ , equivalently the canonical bundle $K_{M}$ is holomorphically inert (structure-sheaf class).

Calabi conjectured (1957): on a compact Kähler manifold with $c_{1} = 0$ , every Kähler class contains a unique Kähler metric with vanishing Ricci form. Yau proved this in 1977. So a Calabi-Yau manifold with the Yau metric has $ρ = 0$ identically, hence $Ric = 0$ , hence Ricci-flat.

The holonomy reduction follows: a Ricci-flat Kähler $n$ -fold has holonomy in $SU (n)$ rather than just $U (n)$ , because Ricci-flatness is exactly the additional constraint that reduces the structure group from $U (n)$ to its determinant-one subgroup. This is what Wang's 1989 bijection encodes: parallel spinors on a Calabi-Yau correspond to the holomorphic volume form $Ω \in H^{0} (M, K_{M})$ , which exists by Yau's theorem and is preserved by $SU (n)$ . $□$

Lean formalization [Intermediate+]

lean_status: partial — Mathlib has the Levi-Civita connection but lacks the holonomy group, the Berger classification theorem, and the Wang parallel-spinor bijection. The Lean module declares stub structures; statements compile, with proofs sorry-gated pending upstream Mathlib infrastructure.

[object Promise]

The Mathlib gap is the holonomy group, the Berger classification, the Wang bijection, and the four named special-holonomy structures. Each is a separate upstream contribution target.

Advanced results [Master]

Compactness and curvature. On the Berger list, the holonomy groups separate into three classes by curvature:

Holonomy	Ricci	Local model
$SO (n)$	unconstrained	generic
$U (n)$	Ricci form $\sim c_{1}$	Kähler
$SU (n), Sp (n)$	Ricci-flat	Calabi-Yau, hyperkähler
$Sp (n) Sp (1)$	Einstein, $Scal \neq = 0$	quaternionic Kähler
$G_{2}, Spin (7)$	Ricci-flat	exceptional holonomy

The Ricci-flat cases (Calabi-Yau, hyperkähler, $G_{2}$ , $Spin (7)$ ) are exactly Wang's parallel-spinor cases plus hyperkähler, which is in turn a special case of $SU (2 n) \cap Sp (n)$ behaviour.

Bryant 1987 — local existence of $G_{2}$ and $Spin (7)$ metrics. The local existence of metrics with $G_{2}$ holonomy in dimension 7 and $Spin (7)$ holonomy in dimension 8 was proved by Robert Bryant in Metrics with exceptional holonomy (Annals of Mathematics 126, 525–576). Bryant's construction is via the Cartan-Kähler theorem applied to the exterior differential system encoding the existence of a parallel calibration form. The local existence was a major breakthrough — it had been an open question whether Berger's exceptional cases were merely formal possibilities or actual geometric examples.

Joyce 1996/2000 — compact $G_{2}$ and $Spin (7)$ manifolds. Dominic Joyce constructed the first compact 7-manifolds with $G_{2}$ holonomy and 8-manifolds with $Spin (7)$ holonomy in Compact Manifolds with Special Holonomy (Oxford University Press, 2000). Joyce's method is to take a flat torus orbifold $T^{7} /Γ$ or $T^{8} /Γ$ with $Γ$ a finite group of isometries inducing the right holonomy, resolve the singularities by gluing in local Calabi-Yau or Eguchi-Hanson asymptotics, and then perturb to a smooth $G_{2}$ or $Spin (7)$ metric using a non-linear inverse function theorem. The resulting compact examples populate the Calabi-Yau / $G_{2}$ / $Spin (7)$ side of geometry with concrete manifolds for the first time.

Bochner-Yano vanishing on parallel-spinor manifolds. A manifold with a parallel spinor has Ricci-flatness automatically (the Lichnerowicz formula combined with $\nabla ψ = 0$ forces $Scal = 0$ , and on a parallel-spinor manifold this extends to Ricci-flatness). This is the structural reason Wang's bijection cases are exactly the Ricci-flat parts of Berger's list (excluding hyperkähler-as-quaternionic-Kähler, which is non-zero scalar curvature).

Symmetric-space cases. Outside Berger's irreducible non-symmetric list lie the symmetric spaces, classified separately by Élie Cartan (1925–32). Riemannian symmetric spaces have holonomy equal to the isotropy representation of a compact Lie group at the basepoint; the classification produces a specific list (compact and non-compact dual pairs) including $O (n) / O (p) \times O (q)$ Grassmannians, complex Grassmannians, quaternionic Grassmannians, exceptional symmetric spaces involving $E_{6}, E_{7}, E_{8}$ , and so on. The Berger classification in its modern form merges with the Cartan classification: irreducible Riemannian manifolds are either generic (holonomy $SO (n)$ ), Berger-special (one of the seven), or symmetric (Cartan's list). Most of the modern interest is on the Berger-special side, which contains all the calibrated geometries.

Cheeger-Gromoll splitting and de Rham decomposition. A Riemannian manifold whose holonomy is reducible decomposes as a Riemannian product, by the de Rham decomposition theorem (1952): if $Hol^{0}$ acts reducibly on $T_{x} M$ , then $(M, g)$ is locally isometric to a product of irreducible factors. Cheeger-Gromoll 1972 extends this to the global setting under a non-negative Ricci hypothesis. Together with Berger's irreducible classification and Cartan's symmetric classification, this gives the complete classification of holonomy groups on simply-connected complete Riemannian manifolds.

Calibrations from parallel spinors. Every parallel spinor on a special-holonomy manifold produces a calibration via spinor squaring 03.09.19: $Ω = Re (ψ^{2})$ on Calabi-Yau (with $ψ$ the parallel holomorphic spinor); $ϕ$ on $G_{2}$ ; $Φ$ on $Spin (7)$ . The Harvey-Lawson 1982 calibrated geometries are exactly these spinor-square forms; the Wang bijection is the structural reason calibrations exist on these and only these manifolds.

Synthesis. This construction generalises the pattern fixed in 03.09.05 (spinor bundle), with the symmetric data replaced by its skew or twisted analogue. Read in the opposite direction, the construction is dual to the metric story: complements and orthogonality are taken with respect to the bilinear datum of this unit, not a metric, and the resulting category of subobjects is the one the rest of the strand classifies. The central insight is that this datum identifies algebra with geometry: functions become vector fields, subspaces become quotients, and invariants become cohomology classes — and that identification is the engine driving every theorem downstream.

Full proof set [Master]

Berger classification. Proof sketch in §Key theorem above. The complete proof is the original 1955 paper, going through all candidate irreducible subalgebras of $so (n)$ in each dimension and computing the Bianchi-compatible curvature spaces. Modern treatments (Joyce 2007, Salamon 1989) consolidate the classification using more efficient representation-theoretic techniques but retain the essential structure of Berger's argument.

Wang bijection (forward direction: parallel spinor implies special holonomy). Proof in §Key theorem above.

Wang bijection (backward direction: special holonomy implies parallel spinor). Proof in §Key theorem above; the technical content is the branching-rule calculation of $Δ_{n} ∣_{Hol^{0}}$ for each special holonomy.

Parallel-spinor counts per holonomy. Computed in Exercise 3 (SU), Exercise 4 (Sp), Exercise 5 ( $G_{2}$ ). The Spin(7) case: $Δ_{8} ≅ R \oplus V_{7}$ under $Spin (7) ↪ Spin (8)$ via the spin embedding, where $V_{7}$ is the standard $Spin (7)$ -representation; one inert summand, so 1 parallel spinor.

Calabi-Yau is Ricci-flat. Proof in Exercise 7 above. Yau's theorem (1977) establishes the Kähler-Einstein metric with vanishing Ricci form on every Calabi class. The holonomy reduction $U (n) \to SU (n)$ is the structural content of Ricci-flatness.

Hyperkähler holonomy is $Sp (n)$ . Proof in Exercise 4 above. The three parallel complex structures $I, J, K$ generate the symplectic-quaternionic structure; their joint stabiliser is $Sp (n) \subset SO (4 n)$ .

de Rham decomposition theorem (statement). Let $(M^{n}, g)$ be a simply-connected complete Riemannian manifold with reducible restricted holonomy $Hol^{0}$ acting reducibly on $T_{x} M = V_{1} \oplus V_{2} \oplus \dots$ . Then $(M, g)$ is globally Riemannian-isometric to a product $(M_{1}, g_{1}) \times (M_{2}, g_{2}) \times \dots$ where each factor has irreducible holonomy. Detailed in ^{[de Rham 1952; Joyce 2007 §2.5]}.

Connections [Master]

Triality and Spin(8) 03.09.13 — the special-holonomy groups $G_{2} \subset Spin (7)$ and $Spin (7) \subset Spin (8)$ arise via triality acting on Spin(8) representations; this is the structural origin of the exceptional cases in Berger's list. Foundation-of: the exceptional holonomies $G_{2}, Spin (7)$ are triality-fixed subgroups of Spin(8).
Calibrated geometries 03.09.19 — the four Harvey-Lawson calibrated geometries live on manifolds with the four Wang-special holonomies: SL on Calabi-Yau ( $SU (n)$ ), associative/coassociative on $G_{2}$ , Cayley on $Spin (7)$ . The calibrating forms are spinor squares of the parallel spinors guaranteed by Wang. Foundation-of: calibrations require special holonomy structure [conn:427.calibration-special-holonomy, anchor: calibrations require special holonomy structure].
Berger holonomy and parallel spinors (equivalence) — Wang 1989: the Berger holonomy groups $SU (n), Sp (n), G_{2}, Spin (7)$ are exactly those admitting a non-zero parallel spinor. The bijection is the structural foundation of calibrated geometry. Equivalence: Berger holonomy bijection with parallel spinors [conn:426.berger-parallel-spinor-equiv, anchor: Berger holonomy bijection with parallel spinors].
Spinor bundle 03.09.05 — parallel spinors are sections of the spinor bundle satisfying $\nabla ψ = 0$ . Their existence reduces the holonomy to one of Wang's special-holonomy classes; their non-existence is generic.
psc obstruction 03.09.16 — every parallel spinor is automatically harmonic, hence in $ker D$ . By the contrapositive of Lichnerowicz, manifolds with parallel spinors (i.e., Wang-special holonomy) cannot admit metrics of strictly positive scalar curvature. The structural connection: special-holonomy manifolds are automatically psc-obstructed.
Witten positive-mass 03.09.17 — the equality case of Witten's theorem ( $m_{ADM} = 0$ ) forces a parallel spinor on the AF 3-manifold, which by Wang on the asymptotically-flat-3-dim setup forces flat $R^{3}$ . The Wang bijection is what closes the equality argument.
Spin group 03.09.03 — the Berger holonomies are subgroups of $Spin (n)$ (the universal cover of $SO (n)$ ); the Wang bijection is between subgroups of $Spin (n)$ stabilising a non-zero spinor and the parallel-spinor count.
Atiyah-Singer index theorem 03.09.10 — a parallel spinor is a particular harmonic spinor; its existence forces the integer Dirac index to be at least the parallel-spinor count, generically positive. This connects holonomy classification to index theory.

We will see in 03.09.19 the parallel spinors of each Berger holonomy promoted to calibrating differential forms; this builds toward the moduli-theoretic study of calibrated submanifolds and the Strominger-Yau-Zaslow mirror-symmetry picture. The Wang bijection recurs every time a parallel-spinor reduction is needed. We will later use the holonomy classification to organise the index-theoretic refinements that appear again in the family-equivariant case of 03.09.21, and in the next phases of the curriculum this builds toward Joyce's moduli-of-special-holonomy programme. The foundational insight of Wang 1989 is exactly that a non-zero parallel spinor is precisely the same data as a special-holonomy reduction — putting this together identifies parallel spinors with the structure group reductions of Berger. The Wang bijection is an instance of the broader principle that parallel sections force structure-group restriction; the bridge between holonomy and parallel-spinor count is given by the formulas $2, n + 1, 1, 1$ for $SU (n), Sp (n), G_{2}, Spin (7)$ .

Historical & philosophical context [Master]

Marcel Berger's 1955 thesis paper Sur les groupes d'holonomie homogène des variétés à connexion affine et des variétés riemanniennes (Bulletin de la Société Mathématique de France 83, 279–330) is the founding document of holonomy classification. Berger was Cartan's intellectual heir: Élie Cartan had introduced the notion of holonomy in the 1920s and classified Riemannian symmetric spaces by 1932, but the classification of non-symmetric irreducible holonomies remained open for two decades. Berger's 1955 paper closed this open problem, by an exhaustive case analysis of irreducible Lie subalgebras of $so (n)$ in every dimension, computing for each the space of curvature tensors satisfying the first Bianchi identity. The seven non-symmetric holonomy groups — $SO (n)$ , $U (n)$ , $SU (n)$ , $Sp (n)$ , $Sp (n) Sp (1)$ , $G_{2}$ , $Spin (7)$ — emerged as the complete list.

Berger's narrative in the 1955 paper is structural-classificatory in Cartan's tradition. The argument runs by exclusion: of all the irreducible subgroups of $SO (n)$ , identify which ones have non-zero curvature spaces compatible with the Bianchi identity, and which of those produce non-symmetric examples. The Cartan-Killing classification of simple Lie algebras over $R$ provides the candidate subgroups; the Bianchi-compatibility computation rules out most of them. The remaining seven cases form Berger's list. Two of these — $G_{2}$ in dimension 7 and $Spin (7)$ in dimension 8 — were classified as merely possible in Berger's paper, with no examples known at the time. It would take 32 years for Bryant's 1987 Metrics with exceptional holonomy to construct local examples, and another decade for Joyce's 1996/2000 compact constructions to populate these cases with concrete manifolds.

The structural significance of Berger's list emerged only gradually. Cheng Wang's 1989 Parallel spinors and parallel forms (Annals of Global Analysis and Geometry 7, 59–68) made the bijection between special holonomy and parallel-spinor existence precise: a Riemannian spin manifold admits a non-zero parallel spinor if and only if its restricted holonomy is one of $SU (n), Sp (n), G_{2}, Spin (7)$ . Wang's paper situates Berger's list inside the spin-geometry framework: the special holonomies are exactly the parallel-spinor stabilisers, which is also exactly the Harvey-Lawson 1982 calibration framework, which is also exactly the Strominger-Yau-Zaslow 1996 mirror-symmetry framework. The 1955 list, dormant for decades, became the foundational classification of an entire research area in the 1990s.

Robert Bryant's 1987 Metrics with exceptional holonomy (Annals of Mathematics 126, 525–576) constructed local metrics with $G_{2}$ and $Spin (7)$ holonomy via the Cartan-Kähler theorem applied to the exterior differential system encoding the existence of a parallel calibration form. The construction was a major breakthrough: Berger's exceptional cases had been known to be merely possibilities in 1955, with no known geometric examples. Bryant's construction settled their existence at the local level; Joyce's 1996/2000 compact constructions extended them to the global setting.

The conceptual lesson is the Cartan one — generalised to spin geometry: every special structure on a Riemannian manifold corresponds to a holonomy reduction, and the special structures with parallel spinors are exactly the ones that produce calibrated geometries, Ricci-flat metrics, and the Calabi-Yau / hyperkähler / $G_{2}$ / $Spin (7)$ programme that has dominated geometric analysis since the 1990s. Wang made this lesson explicit; Bryant and Joyce made it concrete; the entire downstream development of mirror symmetry, special Lagrangian geometry, and exceptional holonomy traces back to Berger's 1955 list, now read through Wang's 1989 lens.

Bibliography [Master]

Berger, M., "Sur les groupes d'holonomie homogène des variétés à connexion affine et des variétés riemanniennes", Bulletin de la Société Mathématique de France 83 (1955), 279–330.
Wang, M. Y., "Parallel spinors and parallel forms", Annals of Global Analysis and Geometry 7 (1989), 59–68.
Bryant, R. L., "Metrics with exceptional holonomy", Annals of Mathematics 126 (1987), 525–576.
Joyce, D. D., Compact Manifolds with Special Holonomy, Oxford University Press, 2000.
Joyce, D. D., Riemannian Holonomy Groups and Calibrated Geometry, Oxford University Press, 2007.
Lawson, H. B. & Michelsohn, M.-L., Spin Geometry, Princeton University Press, 1989. §IV.9.
Calabi, E., "On Kähler manifolds with vanishing canonical class", in Algebraic Geometry and Topology, A Symposium in Honor of S. Lefschetz, Princeton University Press, 1957, 78–89.
Yau, S.-T., "On the Ricci curvature of a compact Kähler manifold and the complex Monge-Ampère equation", Communications in Pure and Applied Mathematics 31 (1978), 339–411.
Salamon, S., Riemannian Geometry and Holonomy Groups, Longman, 1989.
de Rham, G., "Sur la réductibilité d'un espace de Riemann", Commentarii Mathematici Helvetici 26 (1952), 328–344.

Prerequisites

03.09.05
03.09.04
03.09.03

Tier anchors

beginner: Strogatz / 3Blue1Brown analogous level for the holonomy-as-twist intuition
intermediate: Lawson-Michelsohn §IV.9, with Joyce *Compact Manifolds with Special Holonomy* Ch. 2 as parallel reading
master: Berger 1955 *Sur les groupes d'holonomie homogène* (Bull. Soc. Math. France 83); Wang 1989 *Parallel spinors and parallel forms* (Annals of Global Analysis and Geometry); Bryant 1987 *Metrics with exceptional holonomy* (Annals 126); Joyce 2007 *Riemannian Holonomy Groups and Calibrated Geometry*

References

TODO_REF
Berger — Sur les groupes d'holonomie homogène des variétés à connexion affine et des variétés riemanniennes · Bull. Soc. Math. France 83 (1955), 279–330 — the classification theorem
TODO_REF
Wang — Parallel spinors and parallel forms · Annals of Global Analysis and Geometry 7 (1989), 59–68 — the bijection holonomy ↔ parallel spinors
TODO_REF
Bryant — Metrics with exceptional holonomy · Annals of Mathematics 126 (1987), 525–576 — local existence of $G_2$ and Spin(7) metrics
TODO_REF
Joyce — Compact Manifolds with Special Holonomy · Oxford University Press, 2000 — compact $G_2$ and Spin(7) constructions
TODO_REF
Lawson-Michelsohn — Spin Geometry · §IV.9 (parallel spinors and special holonomy)
TODO_REF
Calabi — On Kähler manifolds with vanishing canonical class · Algebraic Geometry and Topology, A Symposium in Honor of S. Lefschetz, Princeton University Press, 1957, 78–89 — the Calabi conjecture (proved by Yau 1977)
TODO_REF
Yau — On the Ricci curvature of a compact Kähler manifold and the complex Monge-Ampère equation · Comm. Pure Appl. Math. 31 (1978), 339–411 — proof of the Calabi conjecture

Lean module

Codex.SpinGeometry.BergerHolonomy

Mathlib gap

Mathlib has the Levi-Civita connection on a Riemannian manifold and the
notion of parallel transport, but does not yet have the holonomy group of
a connection as a Lie subgroup of the structure group, the Berger holonomy
classification, the special-holonomy structures (Calabi-Yau, hyperkähler,
$G_2$, Spin(7)), or the parallel-spinor formalism on which the Wang
classification depends. The Lean stub records placeholder structures for
the holonomy group, the Berger list, the Wang bijection, and the four
named special-holonomy classes; the upstream contribution roadmap targets
`LeviCivitaConnection.holonomyGroup`, `BergerList`, `ParallelSpinor`,
`SpecialHolonomyManifold`.

Reviewer

TBD

Estimated time

beginner: 25m
intermediate: 65m
master: 105m