03.02.18 · differential-geometry / manifolds

Petrov classification of Lorentzian 4-curvature

shipped3 tiersLean: none

Anchor (Master): Penrose & Rindler, Spinors and Space-Time, Vol. 2 (Cambridge, 1986), Ch. 8; Sternberg, Curvature in Mathematics and Physics, Ch. 12; Stephani et al., Exact Solutions, Ch. 4

Intuition Beginner

Empty space far from any matter can still be curved, and that leftover curvature is the free gravitational field. In four-dimensional spacetime this field is recorded by one tensor, the Weyl tensor. The Petrov classification is a way of sorting spacetimes by the internal pattern of that tensor, the same way you might sort crystals by their symmetry.

The sorting is done by special directions called principal null directions, the directions light can travel along that the gravitational field singles out. A generic gravitational field has four of them, all different. Special fields have some of these directions merging together, like two roots of an equation coinciding. The pattern of mergers is the type.

When all four directions are separate, the field is the most generic kind, called type I. When directions merge, the field becomes more special and more symmetric. A black hole sitting still has two pairs of merged directions, the type called D. A gravitational wave passing through empty space has all four directions merged into one, the type called N. And when the Weyl tensor is zero, there is no free gravitational field at all, the type called O.

So the Petrov type is a label that tells you, at a glance, what kind of gravitational field you are looking at. It separates the stationary black holes from the passing waves from the generic mess, using nothing but the algebra of those special null directions.

Visual Beginner

Picture a single light cone at one event in spacetime, drawn as a cone opening into the future. On the surface of that cone, mark the special directions the gravitational field picks out. Each principal null direction is one ray on the cone. The Petrov type is just a count of how many of those rays are distinct and how they cluster.

In the most generic case, type I, four separate rays sit on the cone, none touching. As the field becomes more special the rays slide together: type II has one pair touching with two others apart; type D has two separate pairs, each a double ray; type III has three rays piled at one place and one apart; type N has all four rays piled into a single quadruple direction; and type O has the cone bare, no rays at all, because the Weyl tensor vanishes.

Reading the picture from left to right is reading the field from generic to special. Each arrow merges rays, and each merger raises the symmetry of the gravitational field by one notch.

Worked example Beginner

A still black hole, the Schwarzschild solution, is type D. We show what that means by counting the special directions without heavy computation.

Step 1. Schwarzschild spacetime is spherically symmetric and unchanging in time. Around any event outside the hole there are two obvious light directions: one going straight out, away from the hole, and one going straight in, toward it. These are the radial null directions.

Step 2. The gravitational field of a still mass is built entirely from this in-and-out radial structure. So the special directions the field picks out are exactly these two radial ones, and each is picked out twice over because the geometry is symmetric under swapping forward and backward along the radius.

Step 3. Count the merged directions. There are four principal null directions in total, as always in four dimensions. Here they collapse into the two radial directions, each one a double. Two doubles is the pattern that defines type D, the letter D standing for degenerate.

Step 4. Read off the meaning. The two double directions are why a black hole has such a clean, rigid field. The same type D holds for the rotating Kerr black hole, which is why both are grouped together as the stationary black holes of the universe.

What this tells us: the type of a spacetime can sometimes be read straight off its symmetry. A still, round mass forces the four special directions to collapse into two radial doubles, and that collapse is the definition of type D.

Check your understanding Beginner

Formal definition Intermediate+

Let $(M, g)$ be a four-dimensional Lorentzian manifold of signature $(-, +, +, +)$ with Weyl conformal tensor $C_{ab c d}$ , the totally trace-free part of the Riemann tensor introduced in 03.02.16. The classification rests on the action of the Hodge star on 2-forms and the resulting complex structure on bivectors.

On the six-dimensional space $Λ^{2} T_{p}^{*} M$ of 2-forms, the Hodge star $⋆$ acts by $(⋆ F)_{ab} = \frac{1}{2} ε_{ab}^{c d} F_{c d}$ , where $ε_{ab c d}$ is the volume form. In Lorentzian signature $⋆^{2} = - 1$ on 2-forms, so $⋆$ is a complex structure: complexifying, $Λ^{2} \otimes C$ splits into the $\pm i$ eigenspaces of $⋆$ , the self-dual bivectors $S^{+}$ and anti-self-dual bivectors $S^{-}$ , each of complex dimension three.

The Weyl tensor, regarded as a symmetric linear operator $C : Λ^{2} \to Λ^{2}$ by $F_{ab} \mapsto \frac{1}{2} C_{ab}^{c d} F_{c d}$ , commutes with $⋆$ and is trace-free. It therefore decomposes as $C = C^{+} + C^{-}$ where $C^{\pm}$ act on $S^{\pm}$ , and the two halves are complex conjugates of one another. The self-dual part $C^{+}$ is a symmetric trace-free complex-linear operator on the three-dimensional space $S^{+}$ : in a basis it is a symmetric trace-free $3 \times 3$ complex matrix $Ψ$ , the Weyl matrix.

Definition (principal null directions). A null direction $ℓ^{a}$ ( $g_{ab} ℓ^{a} ℓ^{b} = 0$ ) is a principal null direction (PND) of multiplicity $\geq 1$ when $ℓ_{[e} C_{a] b c [d} ℓ_{f]} ℓ^{b} ℓ^{c} = 0$ . Higher multiplicities are defined by successively stronger vanishing conditions, culminating in $C_{ab c [d} ℓ_{e]} ℓ^{b} = 0$ (multiplicity $\geq 3$ ) and $C_{ab c d} ℓ^{d} = 0$ with $ℓ$ a quadruple PND (multiplicity $4$ ). A Weyl tensor that admits a PND of multiplicity $\geq 2$ is called algebraically special.

Definition (Petrov type). The Petrov type of $C$ at $p$ is the multiplicity pattern of its four principal null directions:

Type	PND multiplicities	Eigenvalue/eigenvector data of $Ψ$
I	${1, 1, 1, 1}$	three distinct eigenvalues, diagonalisable
II	${2, 1, 1}$	two equal eigenvalues, one $2 \times 2$ Jordan block
D	${2, 2}$	two equal eigenvalues, $Ψ$ diagonalisable
III	${3, 1}$	all eigenvalues zero, one $3 \times 3$ nilpotent Jordan block of rank $2$
N	${4}$	all eigenvalues zero, single $3 \times 3$ nilpotent block of rank $1$
O	$C = 0$	$Ψ = 0$

The spinor formulation makes the multiplicity structure transparent. In the $S L (2, C)$ spinor calculus the self-dual Weyl operator is encoded by the totally symmetric Weyl spinor $Ψ_{A B C D} = Ψ_{(A B C D)}$ , a symmetric element of the fourfold-symmetric power of two-component spinor space. Over the algebraically closed field $C$ it factorises into a product of four one-index spinors,

$Ψ_{A B C D} = α_{(A} β_{B} γ_{C} δ_{D)},$

the principal spinors, unique up to ordering and scale. Each principal spinor $α_{A}$ determines a null direction $ℓ^{a} = σ_{A \dot{A}}^{a} α^{A} \overset{α}{ˉ}^{\dot{A}}$ , the associated PND, and the Petrov type is the coincidence pattern of the four principal spinors: ${1111} = I$ , ${211} = II$ , ${22} = D$ , ${31} = III$ , ${4} = N$ , and the degenerate $Ψ_{A B C D} = 0$ is type O.

Counterexamples to common slips

Type D is not "more degenerate" than type III in the partial order. The degeneration order is $I \to II \to D$ and $II \to III \to N \to O$ . Type D and type III are not comparable; D has two double roots while III has a triple, and neither specialises to the other.
The classification is pointwise and complex. A spacetime can have different Petrov types at different events, and the four PNDs are real null directions but the principal spinors live over $C$ ; reality of the metric forces the PNDs to be genuine real null vectors, not complex ones.
Algebraically special is multiplicity $\geq 2$ , not "type II". Types II, D, III, N, and O are all algebraically special; only type I (and a generic point) is algebraically general. The single repeated PND is the defining feature, shared across the special types.

Key theorem with proof Intermediate+

Theorem (Petrov; the algebraic classification of the Weyl spinor). Let $Ψ_{A B C D}$ be the Weyl spinor of a four-dimensional Lorentzian Weyl tensor at a point. Then $Ψ_{A B C D}$ factorises as a symmetrised product of four principal spinors $α_{(A} β_{B} γ_{C} δ_{D)}$ , and the coincidence pattern of these spinors partitions all four-dimensional Lorentzian Weyl tensors into exactly the six Petrov types I, II, D, III, N, O. Equivalently the self-dual Weyl operator $Ψ$ on the three-dimensional space of self-dual bivectors is a symmetric trace-free complex $3 \times 3$ matrix, and the Petrov type is its Segre (Jordan) type as a complex linear operator.

Proof. Fix a spin frame and write the symmetric spinor $Ψ_{A B C D}$ in components. Contracting with an arbitrary spinor $ξ^{A}$ four times produces the quartic form $Ψ_{A B C D} ξ^{A} ξ^{B} ξ^{C} ξ^{D}$ . In a basis $ξ^{A} = (1, z)$ this is a degree-four polynomial $P (z)$ in the single ratio $z$ , with coefficients the five independent components $Ψ_{0}, Ψ_{1}, Ψ_{2}, Ψ_{3}, Ψ_{4}$ of the totally symmetric spinor,

$P (z) = Ψ_{0} - 4 Ψ_{1} z + 6 Ψ_{2} z^{2} - 4 Ψ_{3} z^{3} + Ψ_{4} z^{4} .$

Over $C$ the quartic has four roots (counted with multiplicity), $z = ρ_{1}, ρ_{2}, ρ_{3}, ρ_{4}$ , possibly with a root at infinity when $Ψ_{4} = 0$ . Each root $ρ_{i}$ gives a spinor $α_{A}^{(i)} \propto (ρ_{i}, - 1)$ annihilating the form along its direction, and $P (z) = Ψ_{4} \prod_{i} (z - ρ_{i})$ is exactly the component expansion of $α_{(A}^{(1)} α_{B}^{(2)} α_{C}^{(3)} α_{D)}^{(4)} ξ^{A} ξ^{B} ξ^{C} ξ^{D}$ . This establishes the factorisation $Ψ_{A B C D} = α_{(A} β_{B} γ_{C} δ_{D)}$ .

The Petrov type is now the multiplicity partition of the four roots. A root $ρ$ of multiplicity $m$ corresponds to a principal spinor appearing $m$ times in the symmetrised product, hence to a PND of multiplicity $m$ . The five algebraically distinct partitions of four roots, together with the everywhere-zero polynomial, are: four simple roots ${1, 1, 1, 1}$ (type I); one double and two simple ${2, 1, 1}$ (type II); two double ${2, 2}$ (type D); one triple and one simple ${3, 1}$ (type III); one quadruple ${4}$ (type N); and $P \equiv 0$ , i.e. $Ψ_{A B C D} = 0$ (type O).

The matrix picture matches root by root. The self-dual operator $Ψ$ is symmetric and trace-free, with characteristic polynomial $det (Ψ - λ) = - λ^{3} + \frac{I}{2} λ - \frac{J}{3}$ using the invariants $I = Ψ_{A B C D} Ψ^{A B C D}$ and $J = Ψ_{A B}^{E F} Ψ_{E F}^{C D} Ψ_{C D}^{A B}$ . A simple eigenvalue is a single PND; a repeated eigenvalue with a full eigenspace is the diagonalisable degenerate case (type D), while a repeated eigenvalue with a Jordan block forces two principal spinors to coincide (type II). The nilpotent cases, all eigenvalues zero, distinguish a rank- $2$ nilpotent ( $3 \times 3$ Jordan block, type III) from a rank- $1$ nilpotent (type N). Each Jordan type corresponds to exactly one root-multiplicity partition, so the spinor and matrix classifications coincide. $□$

Bridge. This classification builds toward the Goldberg-Sachs theorem and the integration of the algebraically special vacuum solutions in the Master section, and it appears again in the peeling theorem governing how the gravitational field falls off along outgoing null rays, in the curvature-invariant detection of speciality through $I^{3} - 27 J^{2}$ , and in the spinor and twistor formulation of the Bianchi identities. The mechanism that makes the argument close is the passage to the algebraically closed field $C$ : a real quartic need not factor over $R$ , but the Weyl spinor lives in complexified bivector space where every quartic splits into linear factors, and the resulting root multiplicities are a complete and discrete invariant. The same complexification that turns the Hodge star into a genuine complex structure is what gives the principal null directions their algebraic meaning, and it is why the classification is sharp in exactly four dimensions and in exactly Lorentzian signature.

Exercises Intermediate+

Exercise 3 (medium, symbolic).

Show that for a type N Weyl tensor both invariants vanish: $I = 0$ and $J = 0$ . Use the spinor form $Ψ_{A B C D} = α_{A} α_{B} α_{C} α_{D}$ (a single quadruple principal spinor).

Hint

$I = Ψ_{A B C D} Ψ^{A B C D}$ and $J$ is a cubic contraction. Every term contains a contraction $α^{A} α_{A} = 0$ of a spinor with itself.

Answer

With $Ψ_{A B C D} = α_{A} α_{B} α_{C} α_{D}$ , the quadratic invariant is $I = Ψ_{A B C D} Ψ^{A B C D} = (α_{A} α^{A}) (α_{B} α^{B}) (α_{C} α^{C}) (α_{D} α^{D})$ . The antisymmetric spinor contraction $α_{A} α^{A} = ϵ^{A B} α_{A} α_{B} = 0$ since $ϵ^{A B}$ is antisymmetric and $α_{A} α_{B}$ is symmetric. Hence $I = 0$ . The cubic invariant $J$ is built from three factors of $Ψ$ contracted in a cycle; each such contraction again pairs two indices of the same $α$ spinor, producing a factor $α_{A} α^{A} = 0$ , so $J = 0$ . Vanishing of both invariants is necessary (not sufficient) for type N; types III and N share $I = J = 0$ .

Exercise 4 (medium, multiple choice).

A spacetime is algebraically special precisely when the speciality discriminant vanishes:

A. $I = 0$ B. $J = 0$ C. $I^{3} - 27 J^{2} = 0$ D. $det Ψ = 0$

Hint

The discriminant of the trace-free cubic characteristic polynomial vanishes exactly when it has a repeated root, i.e. when $Ψ$ has a repeated eigenvalue.

Answer

C. $I^{3} - 27 J^{2} = 0$ . The characteristic polynomial of the trace-free Weyl operator is $λ^{3} - \frac{I}{2} λ + \frac{J}{3}$ (up to sign convention), whose discriminant is proportional to $I^{3} - 27 J^{2}$ . Its vanishing signals a repeated eigenvalue and hence a repeated PND, the defining condition of algebraic speciality. Types II, D, III, N, and O all satisfy $I^{3} = 27 J^{2}$ ; type I does not. The conditions $I = 0$ or $J = 0$ alone are neither necessary nor sufficient, and $det Ψ = J /3$ vanishing is a different condition.

Exercise 5 (medium, symbolic).

Distinguish type D from type II using the invariants. Given that a Weyl tensor is algebraically special ( $I^{3} = 27 J^{2}$ ) with $I \neq = 0$ , show that the additional condition characterising type D is that $Ψ$ is diagonalisable, and express this through the matrix $Ψ - λ_{0}$ where $λ_{0}$ is the repeated eigenvalue.

Hint

For a $3 \times 3$ matrix with a repeated eigenvalue $λ_{0}$ (algebraic multiplicity $2$ ) and a simple eigenvalue, type D means the eigenspace of $λ_{0}$ is two-dimensional, type II means it is one-dimensional. The repeated eigenvalue of a trace-free cubic with $I^{3} = 27 J^{2}$ is $λ_{0} = - 3 J / I$ (in the convention where $- λ^{3} + \frac{I}{2} λ - \frac{J}{3}$ is the characteristic polynomial; recompute for your sign convention).

Answer

When $I^{3} = 27 J^{2}$ with $I \neq = 0$ the characteristic polynomial has a double root $λ_{0}$ and a simple root $- 2 λ_{0}$ (the trace is zero, so the three roots $λ_{0}, λ_{0}, - 2 λ_{0}$ sum to zero), with $λ_{0}$ determined by $I$ and $J$ . The matrix $N = Ψ - λ_{0} Id$ has $λ_{0}$ as a double eigenvalue shifted to zero. Type D holds when $N$ has rank $1$ on the generalised eigenspace, equivalently when $rank (Ψ - λ_{0}) = 1$ restricted appropriately so that $Ψ$ is diagonalisable; type II holds when $(Ψ - λ_{0})^{2} = 0$ but $Ψ - λ_{0} \neq = 0$ on the double-eigenvalue block, the genuine Jordan-block case. Concretely, type D requires $(Ψ - λ_{0})$ to annihilate a two-dimensional eigenspace, while type II leaves only a one-dimensional eigenspace; this is the diagonalisable-versus-defective dichotomy at the repeated eigenvalue.

Exercise 6 (hard, symbolic).

Classify the Weyl tensor whose self-dual operator in a suitable null basis has the single non-zero spinor component $Ψ_{4} \neq = 0$ and all others zero ( $Ψ_{0} = Ψ_{1} = Ψ_{2} = Ψ_{3} = 0$ ). Identify the Petrov type and the quadruple principal null direction.

Hint

Form the quartic $P (z) = Ψ_{0} - 4 Ψ_{1} z + 6 Ψ_{2} z^{2} - 4 Ψ_{3} z^{3} + Ψ_{4} z^{4}$ and read off its root structure. A quadruple root corresponds to a single PND of multiplicity $4$ .

Answer

With only $Ψ_{4} \neq = 0$ the quartic is $P (z) = Ψ_{4} z^{4}$ , which has a single root $z = 0$ of multiplicity four. By the factorisation theorem the Weyl spinor is $Ψ_{A B C D} = α_{A} α_{B} α_{C} α_{D}$ with $α_{A} \propto (0, - 1)$ corresponding to $z = 0$ , a single quadruple principal spinor. The Petrov type is N. The associated PND is the null direction $ℓ^{a} = σ_{A \dot{A}}^{a} α^{A} \overset{α}{ˉ}^{\dot{A}}$ built from that principal spinor; in the standard Newman-Penrose null tetrad $(ℓ, n, m, \overset{m}{ˉ})$ , having only $Ψ_{4}$ non-zero is exactly the statement that $n^{a}$ is the quadruple PND. This is the canonical algebraic form of a pure gravitational wave: a single transverse polarisation propagating along one null direction, with $Ψ_{4}$ the wave amplitude.

Exercise 7 (hard, symbolic).

Show that the Petrov types form a degeneration partial order, and prove the elementary fact that type D and type III are not comparable. Use the root-multiplicity partitions ${2, 2}$ and ${3, 1}$ .

Hint

One partition degenerates to another if it is obtained by merging roots. Merging can only increase a multiplicity by absorbing another root; check whether ${2, 2}$ can be reached from ${3, 1}$ by merging, or conversely.

Answer

A type $X$ degenerates to type $Y$ when the root-multiplicity partition of $Y$ is obtained from that of $X$ by merging two roots into one of higher multiplicity (equivalently, $Y$ is a coarser partition reachable by addition). From ${1, 1, 1, 1}$ merging two roots gives ${2, 1, 1}$ (I to II); merging again gives ${2, 2}$ or ${3, 1}$ (II to D, II to III); from ${3, 1}$ merging gives ${4}$ (III to N); and ${4}$ collapses to the zero tensor (N to O). Now compare ${2, 2}$ and ${3, 1}$ . To reach ${3, 1}$ from ${2, 2}$ one would need to move a unit of multiplicity from one double root to the other, producing a triple and a single, but merging only combines distinct roots and cannot split a double; the two double roots of ${2, 2}$ are distinct directions and merging them yields ${4}$ , not ${3, 1}$ . Conversely ${3, 1}$ cannot reach ${2, 2}$ because its triple root cannot be split. Hence neither degenerates to the other, so type D and type III are incomparable in the order.

Lean formalization Intermediate+

Mathlib has the Riemann curvature tensor and an undergraduate-core geometry layer, but none of the structure this classification requires. There is no Lorentzian-metric object carrying a time orientation, no induced complex structure (Hodge star with $⋆^{2} = - 1$ ) on the bundle of real 2-forms in signature $(-, +, +, +)$ , and so no self-dual / anti-self-dual splitting of complexified bivectors. The $S O (3, C)$ -equivariant repackaging of the Weyl tensor as a symmetric trace-free complex $3 \times 3$ matrix, the $S L (2, C)$ spinor calculus, the totally symmetric Weyl spinor $Ψ_{A B C D}$ , and its factorisation into four principal spinors over $C$ are all absent.

lean_status: none records the gap; no Lean module ships. the Mathlib gap analysis names the missing infrastructure in dependence order; the spinor bundle and the symmetric-spinor decomposition are load-bearing, and the Segre/root-multiplicity logic of the classification would build on them. Tyler attests intermediate-tier correctness; the spinor factorisation argument and the Goldberg-Sachs material at Master tier are flagged for external differential-geometry review.

Advanced results Master

Three results take the classification to the depth Penrose-Rindler Vol. 2 and Stephani et al. require and connect it to the integration of the Einstein equations.

The Goldberg-Sachs theorem. A vacuum spacetime ( $R_{ab} = 0$ ) is algebraically special if and only if it admits a shear-free null geodesic congruence, and the repeated principal null direction is then tangent to that congruence ^{[Stephani et al. Ch. 9]}. In Newman-Penrose form the statement is sharp: with $ℓ^{a}$ a repeated PND, the optical scalars satisfy $κ = σ = 0$ (the congruence is geodesic and shear-free) precisely when $Ψ_{0} = Ψ_{1} = 0$ . This theorem is the engine behind the construction of the algebraically special exact solutions: imposing a repeated PND reduces the vacuum field equations from a coupled nonlinear system to a sequence of integrable equations, which is how the Robinson-Trautman and Kerr-Schild families were found. The Kerr metric itself was discovered by Roy Kerr in 1963 by assuming type D and exploiting the two shear-free congruences the Goldberg-Sachs theorem guarantees.

The peeling theorem. Along an outgoing null geodesic running to future null infinity in an asymptotically flat spacetime, the Weyl tensor "peels" through the Petrov types as a power series in the affine parameter $r$ ^{[Sachs 1962]} ^{[Penrose 1960]}:

$Ψ_{n} = \frac{Ψ _{n}^{(N)}}{r} + \frac{Ψ _{n}^{(I I I)}}{r ^{2}} + \frac{Ψ _{n}^{(I I)}}{r ^{3}} + \frac{Ψ _{n}^{(I)}}{r ^{4}} + \dots,$

so the leading $1/ r$ term is type N (radiation), the $1/ r^{2}$ term type III, the $1/ r^{3}$ term type D / II (the Coulomb-like field of the source mass), and the $1/ r^{4}$ term type I. The physical content is that gravitational radiation, the news measured by a distant detector, is the type N piece of the field that dominates far from the source, while the stationary near-field mass term is the type D piece that dominates close in. The peeling pattern $N \to III \to II \to I$ along the outgoing ray is the asymptotic signature of an isolated radiating system, and it is the geometric basis of the Bondi-Sachs mass-loss formula that quantifies energy carried away by gravitational waves.

The Bel-Robinson tensor and the Bel criteria. Bel's fourth-order tensor $T_{ab c d} = C_{a ec f} C_{b}^{e}_{d}^{f} + ⋆ C_{a ec f} ⋆ C_{b}^{e}_{d}^{f}$ is the gravitational analogue of the electromagnetic stress-energy tensor, totally symmetric and trace-free in vacuum ^{[Bel 1959]}. Its contractions with a timelike observer split the Weyl tensor into a "gravitoelectric" part $E_{ab} = C_{a c b d} u^{c} u^{d}$ and a "gravitomagnetic" part $B_{ab} = ⋆ C_{a c b d} u^{c} u^{d}$ , each a symmetric trace-free spatial $3 \times 3$ tensor, exactly the real and imaginary parts of the complex matrix $Ψ$ relative to $u^{a}$ . The Bel criteria restate the Petrov types as super-energy conditions: type N is the case where the super-energy flux is null, the algebraic statement that pure gravitational radiation carries energy along a single null direction, paralleling a null electromagnetic field.

Synthesis. The Petrov classification is the four-dimensional Lorentzian shadow of a purely algebraic fact, that a symmetric trace-free complex $3 \times 3$ matrix has six Segre types, and the spinor factorisation $Ψ_{A B C D} = α_{(A} β_{B} γ_{C} δ_{D)}$ is what translates that algebra into the geometry of principal null directions. The classification organises the entire catalogue of exact solutions: the stationary black holes are type D, the gravitational waves are type N, the generic field is type I, and the degeneration order I to II to D, II to III to N to O is the order of increasing algebraic symmetry. Goldberg-Sachs converts an algebraic condition (a repeated PND) into a differential-geometric one (a shear-free null congruence) and thereby into a method for solving the Einstein equations, which is how Kerr's rotating black hole was found. The peeling theorem reads the same types radially, sorting the radiative far field from the Coulombic near field along outgoing null rays, and the Bel-Robinson tensor recasts the whole scheme as a theory of gravitational super-energy. Across these the pattern is one structure seen from four sides: a Segre type of a complex matrix, a multiplicity partition of four null directions, a method for integrating algebraically special spacetimes, and the invariant signature of gravitational radiation.

Full proof set Master

Proposition 1 (the self-dual Weyl operator is symmetric trace-free of dimension five). On a four-dimensional Lorentzian manifold the Hodge star satisfies $⋆^{2} = - 1$ on 2-forms, the complexified bivector space splits as $S^{+} \oplus S^{-}$ into the $\pm i$ eigenspaces, and the self-dual Weyl operator $C^{+} : S^{+} \to S^{+}$ is a symmetric trace-free complex-linear operator on a three-dimensional space, hence carries five complex parameters.

Proof. In signature $(-, +, +, +)$ the Hodge star on $p$ -forms in dimension $n$ satisfies $⋆^{2} = (- 1)^{p (n - p)} sgn (g)$ on the relevant degree; for $p = 2$ , $n = 4$ , and Lorentzian sign $sgn (g) = - 1$ , this gives $⋆^{2} = (- 1)^{2 \cdot 2} \cdot (- 1) = - 1$ . Thus $⋆$ is a complex structure on the six-real-dimensional space $Λ^{2}$ , and its complexification $Λ^{2} \otimes C$ decomposes into the eigenspaces $S^{\pm}$ for eigenvalues $\pm i$ , each of complex dimension three because $⋆$ has no real eigenvalues and the two eigenspaces are conjugate. The Weyl operator $C$ commutes with $⋆$ : this is the duality identity $⋆ C = C ⋆$ , equivalent to the trace-free first-Bianchi symmetry of $C_{ab c d}$ together with $C_{ab}^{c d} ε_{c d}^{e f} = ε_{ab}^{c d} C_{c d}^{e f}$ , which holds because the totally trace-free part of the curvature is the part on which left and right Hodge duals agree. Commuting with $⋆$ , $C$ preserves $S^{+}$ and $S^{-}$ , restricting to $C^{\pm}$ . On $S^{+}$ , $C^{+}$ is symmetric with respect to the complex bilinear form inherited from the bivector metric, and trace-free because $C$ is totally trace-free. A symmetric trace-free operator on a three-dimensional complex space has $6 - 1 = 5$ independent complex entries. $□$

Proposition 2 (the six Petrov types are exhaustive and mutually exclusive). Every four-dimensional Lorentzian Weyl tensor falls into exactly one of the types I, II, D, III, N, O, and the types are in bijection with the Segre characteristics of a symmetric trace-free complex $3 \times 3$ matrix together with the zero matrix.

Proof. By Proposition 1 the data is a symmetric trace-free complex $3 \times 3$ matrix $Ψ$ (or the zero matrix, type O). For a non-zero $Ψ$ , classify by its Jordan / Segre type over $C$ . The characteristic polynomial is cubic with roots summing to zero (tracelessness). The possibilities are: three distinct eigenvalues, forcing diagonalisability (Segre $[1, 1, 1]$ ); a double and a simple eigenvalue, either diagonalisable (Segre $[(1, 1), 1]$ ) or with a $2 \times 2$ block (Segre $[2, 1]$ ); or a triple eigenvalue, which by tracelessness must be zero, in the diagonalisable (only $Ψ = 0$ , excluded), rank-one nilpotent (Segre $[(2, 1)]$ , a single $2$ -block plus a $1$ ), or rank-two nilpotent (Segre $[3]$ ) form. Matching to PND multiplicities: $[1, 1, 1] =$ I, $[2, 1] =$ II, $[(1, 1), 1] =$ D, rank-two nilpotent $=$ III, rank-one nilpotent $=$ N. These five Segre types plus the zero matrix exhaust the symmetric trace-free $3 \times 3$ complex matrices up to similarity, and they are pairwise distinct as similarity classes, so each Weyl tensor has exactly one type. The reality structure of the Lorentzian metric does not split any class further: the involution relating $C^{+}$ to $C^{-}$ acts within each Segre class, so the complex classification descends unchanged to the real Lorentzian tensor. $□$

Proposition 3 (Schwarzschild and Kerr are type D). The Schwarzschild and Kerr vacuum metrics have Petrov type D at every event of their exterior, with the single non-zero Weyl scalar $Ψ_{2}$ in the principal Newman-Penrose tetrad aligned with the two repeated PNDs.

Proof. For Schwarzschild in Boyer-Lindquist-type coordinates choose the Newman-Penrose tetrad with $ℓ^{a}, n^{a}$ the outgoing and ingoing radial null vectors and $m^{a}, \overset{m}{ˉ}^{a}$ spanning the angular sphere. A direct computation of the Weyl scalars in this tetrad gives $Ψ_{0} = Ψ_{1} = Ψ_{3} = Ψ_{4} = 0$ and

$Ψ_{2} = - \frac{m}{r ^{3}},$

the single non-zero scalar (here $m$ is the mass, $r$ the area radius). Only $Ψ_{2} \neq = 0$ is the algebraic signature of type D: the quartic $P (z) = 6 Ψ_{2} z^{2}$ has a double root at $z = 0$ and a double root at $z = \infty$ , i.e. the partition ${2, 2}$ , with the two double PNDs being $ℓ^{a}$ (outgoing radial) and $n^{a}$ (ingoing radial). For Kerr the same tetrad choice (the Kinnersley tetrad along the principal null congruences) yields $Ψ_{0} = Ψ_{1} = Ψ_{3} = Ψ_{4} = 0$ and the single scalar

$Ψ_{2} = - \frac{m}{( r - ia cos θ ) ^{3}},$

with $a$ the spin parameter; again only $Ψ_{2} \neq = 0$ , so Kerr is type D with the two principal null congruences as repeated PNDs. The $a \to 0$ limit recovers the Schwarzschild scalar, and the existence of these two shear-free congruences is the Goldberg-Sachs content that allowed Kerr's discovery of the metric. $□$

Proposition 4 (a plane gravitational wave is type N). The pp-wave metric $g = - 2 d u d v + H (u, x, y) (d u)^{2} + d x^{2} + d y^{2}$ with $H$ harmonic in $(x, y)$ has Petrov type N (or O when $H$ is such that the Weyl tensor vanishes), with the covariantly constant null vector $\partial_{v}$ as the quadruple PND.

Proof. The vector $k^{a} = \partial_{v}$ is null ( $g (k, k) = 0$ ) and covariantly constant, $\nabla_{a} k^{b} = 0$ , since the metric components are independent of $v$ and $H$ enters only the $d u^{2}$ term. Choosing a Newman-Penrose tetrad with $ℓ^{a} = k^{a}$ and $n^{a}$ the conjugate null vector along $\partial_{u}$ (adjusted by $H$ ), the only non-vanishing Weyl scalar is

$Ψ_{4} = - \frac{1}{2} (\partial_{x}^{2} - \partial_{y}^{2} - 2 i \partial_{x} \partial_{y}) H,$

the complex combination of second transverse derivatives of $H$ . All other scalars vanish because the curvature is concentrated entirely in the $d u \otimes d u$ direction. With only $Ψ_{4} \neq = 0$ the quartic $P (z) = Ψ_{4} z^{4}$ has a single quadruple root, the partition ${4}$ , so the metric is type N with $\partial_{v}$ the quadruple PND (when $Ψ_{4} = 0$ , e.g. $H$ harmonic but with vanishing traceless Hessian, the Weyl tensor is zero and the type is O). The two real polarisations of the wave are the real and imaginary parts of $Ψ_{4}$ , the plus and cross modes; this is the algebraic form of a transverse-traceless gravitational wave propagating along $k^{a}$ . $□$

These propositions assemble the classification into its working physical form: Proposition 1 builds the five-complex-parameter operator, Proposition 2 enumerates its types, and Propositions 3 and 4 exhibit the two physically central cases, the stationary black hole (D) and the gravitational wave (N).

Connections Master

Weyl tensor and conformally flat metrics 03.02.16 supplies the object that the Petrov scheme classifies: the totally trace-free Weyl tensor, conformally invariant and identical to the full curvature in vacuum. The conformally flat case $C = 0$ of that unit is exactly Petrov type O here, so the classification refines the binary distinction "Weyl-flat or not" of the prior unit into the six-fold algebraic taxonomy, the additional structure coming entirely from Lorentzian signature in dimension four.
Sectional, Ricci, scalar curvature 03.02.05 provides the Riemann tensor and the curvature operator on bivectors whose self-dual restriction is the Weyl matrix $Ψ$ . The eigenvalue / eigenbivector analysis that drives the Petrov types is the Lorentzian-signature counterpart of the curvature-operator spectral analysis introduced there; the same bivector formalism, complexified by the Lorentzian Hodge star, produces the principal null directions.
Globally hyperbolic Lorentzian manifolds 13.09.01 draws on the Petrov type as the organising label of the exact-solution catalogue: type D for the Schwarzschild and Kerr backgrounds on which curved-spacetime field theory is built, and type N for the radiative spacetimes whose Cauchy data carry gravitational waves. The peeling theorem proved here controls the asymptotic falloff that makes the radiation field well-defined at null infinity, a prerequisite for the scattering theory of that unit.
Geodesics and parallel transport 13.02.02 receives the principal null directions as the distinguished null geodesic congruences along which the Goldberg-Sachs theorem forces shear-free propagation, and the peeling theorem as the radial degeneration of the Weyl type along outgoing null geodesics. The algebraic speciality this unit detects is the curvature condition that the optical scalars of that unit make geometric.

Historical & philosophical context Master

Aleksei Zinov'evich Petrov introduced the classification in a 1954 paper in the Scientific Notices of Kazan State University ^{[Petrov 1954]}, working from the algebraic structure of the curvature tensor regarded as a symmetric linear operator on the six-dimensional space of bivectors. Petrov found the eigenvalue degeneracy types by a real-bivector analysis; his original three "types I, II, III" were later refined into the six-fold scheme used today once the conformally-special subcases D, N, and O were separated out. Felix Pirani, in a 1957 Physical Review paper ^{[Pirani 1957]}, connected the classification to physics, identifying the algebraically special types with gravitational radiation and giving the first invariant criterion for the presence of gravitational waves through the Petrov type of the Weyl tensor.

The decisive reformulation came from Roger Penrose, whose 1960 Annals of Physics paper ^{[Penrose 1960]} recast the entire scheme in two-component spinor language, replacing Petrov's bivector eigenvalue problem with the factorisation of the totally symmetric Weyl spinor $Ψ_{A B C D}$ into four principal spinors. In Penrose's formulation the six types are simply the five partitions of four roots of a quartic plus the zero case, an algebraic transparency the bivector approach lacked. Rainer Sachs, in 1961 and 1962 Proceedings of the Royal Society papers ^{[Sachs 1962]}, proved the peeling theorem governing how the Weyl tensor degenerates through the Petrov types along outgoing null rays, establishing the algebraic structure of the gravitational radiation field at null infinity. Louis Bel introduced the fourth-order super-energy tensor in 1959 ^{[Bel 1959]}, giving the types an energetic interpretation. The full spinor-and-twistor synthesis is the subject of Penrose and Rindler's Spinors and Space-Time, whose second volume gives the definitive modern account of the classification and its analytic consequences.

Bibliography Master

@article{Petrov1954,
  author = {Petrov, Aleksei Z.},
  title = {Klassifikacya prostranstv opredelyayushchikh polya tyagoteniya},
  journal = {Uchenye Zapiski Kazan. Gos. Univ.},
  volume = {114},
  pages = {55--69},
  year = {1954},
  note = {English transl.: ``The classification of spaces defining gravitational fields'', Gen. Relativ. Gravit. 32 (2000) 1665--1685}
}

@article{Pirani1957,
  author = {Pirani, Felix A. E.},
  title = {Invariant formulation of gravitational radiation theory},
  journal = {Phys. Rev.},
  volume = {105},
  number = {3},
  pages = {1089--1099},
  year = {1957}
}

@article{Penrose1960,
  author = {Penrose, Roger},
  title = {A spinor approach to general relativity},
  journal = {Ann. Phys. (N.Y.)},
  volume = {10},
  number = {2},
  pages = {171--201},
  year = {1960}
}

@article{Sachs1961,
  author = {Sachs, Rainer K.},
  title = {Gravitational waves in general relativity. {VI}. The outgoing radiation condition},
  journal = {Proc. R. Soc. Lond. A},
  volume = {264},
  pages = {309--338},
  year = {1961}
}

@article{Sachs1962,
  author = {Sachs, Rainer K.},
  title = {Gravitational waves in general relativity. {VIII}. Waves in asymptotically flat space-time},
  journal = {Proc. R. Soc. Lond. A},
  volume = {270},
  pages = {103--126},
  year = {1962}
}

@article{Bel1959,
  author = {Bel, Louis},
  title = {Introduction d'un tenseur du quatri{\`e}me ordre},
  journal = {C. R. Acad. Sci. Paris},
  volume = {248},
  pages = {1297--1300},
  year = {1959}
}

@article{GoldbergSachs1962,
  author = {Goldberg, Joshua N. and Sachs, Rainer K.},
  title = {A theorem on {P}etrov types},
  journal = {Acta Phys. Polon. Suppl.},
  volume = {22},
  pages = {13--23},
  year = {1962}
}

@article{Kerr1963,
  author = {Kerr, Roy P.},
  title = {Gravitational field of a spinning mass as an example of algebraically special metrics},
  journal = {Phys. Rev. Lett.},
  volume = {11},
  number = {5},
  pages = {237--238},
  year = {1963}
}

@book{PenroseRindler1986,
  author = {Penrose, Roger and Rindler, Wolfgang},
  title = {Spinors and Space-Time, Vol. 2: Spinor and Twistor Methods in Space-Time Geometry},
  publisher = {Cambridge University Press},
  year = {1986}
}

@book{Stephani2003,
  author = {Stephani, Hans and Kramer, Dietrich and MacCallum, Malcolm and Hoenselaers, Cornelius and Herlt, Eduard},
  title = {Exact Solutions of Einstein's Field Equations},
  edition = {2nd},
  publisher = {Cambridge University Press},
  year = {2003}
}

@book{Sternberg2012,
  author = {Sternberg, Shlomo},
  title = {Curvature in Mathematics and Physics},
  publisher = {Dover},
  year = {2012}
}

Prerequisites

03.02.16
03.02.05

Tier anchors

beginner: Penrose, The Road to Reality, Ch. 19 (free gravitational field); d'Inverno, Introducing Einstein's Relativity, Ch. 21 (Petrov types in pictures)
intermediate: Sternberg, Curvature in Mathematics and Physics (Dover, 2012), Ch. 12 §4; Stephani et al., Exact Solutions of Einstein's Field Equations, 2nd ed. (Cambridge, 2003), Ch. 4 and 9
master: Penrose & Rindler, Spinors and Space-Time, Vol. 2 (Cambridge, 1986), Ch. 8; Sternberg, Curvature in Mathematics and Physics, Ch. 12; Stephani et al., Exact Solutions, Ch. 4

References

Sternberg, S. — Curvature in Mathematics and Physics (Dover, 2012) · Ch. 12 §4 (the Petrov classification of the Weyl tensor in Lorentzian signature)
Penrose, R. & Rindler, W. — Spinors and Space-Time, Vol. 2: Spinor and Twistor Methods (Cambridge, 1986) · Ch. 8 (the Weyl spinor, principal spinors, and the Petrov-Penrose classification)
Petrov, A. Z. — 'Klassifikacya prostranstv opredelyayushchikh polya tyagoteniya' (Classification of spaces defining gravitational fields) · Uchenye Zapiski Kazan. Gos. Univ. 114 (1954) 55-69 (the original algebraic classification)
Stephani, H., Kramer, D., MacCallum, M., Hoenselaers, C. & Herlt, E. — Exact Solutions of Einstein's Field Equations, 2nd ed. (Cambridge, 2003) · Ch. 4 (the Petrov classification, principal null directions, the invariants I and J) and Ch. 9 (algebraically special solutions)
Penrose, R. — 'A spinor approach to general relativity' · Ann. Phys. (N.Y.) 10 (1960) 171-201 (the spinor formulation and the peeling theorem)
Pirani, F. A. E. — 'Invariant formulation of gravitational radiation theory' · Phys. Rev. 105 (1957) 1089-1099 (Petrov types and the physical reading of gravitational radiation)
Sachs, R. K. — 'Gravitational waves in general relativity. VI. The outgoing radiation condition' · Proc. R. Soc. A 264 (1961) 309-338 and 'VIII. Waves in asymptotically flat space-time', Proc. R. Soc. A 270 (1962) 103-126 (the peeling theorem)
Bel, L. — 'Introduction d'un tenseur du quatrième ordre' · C. R. Acad. Sci. Paris 248 (1959) 1297-1300 (the Bel-Robinson tensor and the Bel criteria for Petrov types)

Estimated time

beginner: 18m
intermediate: 45m
master: 85m