Symmetry, Integrability and Geometry: Methods and Applications SIGMA 3 (2007), 095, 13 pages
Stanilov–Tsankov–Videv Theory?
Miguel BROZOS-VA´ZQUEZ †
1
, Bernd FIEDLER †
2
, Eduardo GARCI´A-RI´O †
1
,
Peter GILKEY †
3
, Stana NIKCˇEVIC´ †
4
, Grozio STANILOV †
5
,
Yulian TSANKOV †
5
, Ramo´n VA´ZQUEZ-LORENZO †
1
and Veselin VIDEV †
6
†1 Department of Geometry and Topology, Faculty of Mathematics,
University of Santiago de Compostela, Santiago de Compostela 15782, Spain
E-mail: mbrozos@usc.es, xtedugr@usc.es, ravazlor@usc.es
†2 Eichelbaumstr. 13, D-04249 Leipzig, Germany
E-mail: bfiedler@fiemath.de
†3 Mathematics Department, University of Oregon, Eugene Oregon 97403-1222, USA
E-mail: gilkey@uoregon.edu
†4 Mathematical Institute, SANU, Knez Mihailova 35, p.p. 367, 11001 Belgrade, Serbia
E-mail: stanan@mi.sanu.ac.yu
†5 Sofia University “St. Kl. Ohridski”, Sofia, Bulgaria
E-mail: stanilov@fmi.uni-sofia.bg, ucankov@fmi.uni-sofia.bg
†6 Mathematics Department, Thracian University, University Campus,
6000 Stara Zagora, Bulgaria
E-mail: videv@uni-sz.bg
Received August 07, 2007, in final form September 22, 2007; Published online September 28, 2007
Original article is available at http://www.emis.de/journals/SIGMA/2007/095/
Abstract. We survey some recent results concerning Stanilov–Tsankov–Videv theory, con-
formal Osserman geometry, and Walker geometry which relate algebraic properties of the
curvature operator to the underlying geometry of the manifold.
Key words: algebraic curvature tensor; anti-self-dual; conformal Jacobi operator; confor-
mal Osserman manifold; Jacobi operator; Jacobi–Tsankov; Jacobi–Videv; mixed-Tsankov;
Osserman manifold; Ricci operator; self-dual; skew-symmetric curvature operator; skew-
Tsankov; skew-Videv; Walker manifold; Weyl conformal curvature operator
2000 Mathematics Subject Classification: 53B20
This article is dedicated to the memory of N. Blazˇic´ (who passed away 10 October
2005) and to the memory of T. Branson (who passed away 11 March 2006). They
were coauthors, friends, and talented mathematicians.
1 Introduction
In this article we shall survey just a few of the many recent developments in Differential Geome-
try which relate algebraic properties of various operators naturally associated with the curvature
of a pseudo-Riemannian manifold to the underlying geometric properties of the manifolds in-
volved.
We introduce the following notational conventions. Let M = (M, g) be a pseudo-Riemannian
manifold of signature (p, q) and dimension m = p+ q. We say that M is Riemannian if p = 0,
?This paper is a contribution to the Proceedings of the 2007 Midwest Geometry Conference in honor of
Thomas P. Branson. The full collection is available at http://www.emis.de/journals/SIGMA/MGC2007.html
2 M. Brozos-Va´zquez et al.
i.e. if g is positive definite. We say that M is Lorentzian if p = 1. Let
S±P (M) = {ξ ∈ TPM : g(ξ, ξ) = ±1}
be the pseudo-spheres of unit spacelike (+) and unit timelike (−) vectors. Let ∇ be the Levi-
Civita connection and let
R(x, y) := ∇x∇y −∇y∇x −∇[x,y]
be the associated skew-symmetric curvature operator. If {ei} is a local frame for the tangent
bundle, we let gij := g(ei, ej) and let gij be the inverse matrix. The Jacobi operator and the
Ricci operator are the self-adjoint endomorphisms defined, respectively, by:
J (x) : y → R(y, x)x and ρ : x→
∑
ij
gijR(x, ei)ej . (1.1)
One also defines the curvature tensor R ∈ ⊗4T ∗M , the scalar curvature τ , the Weyl conformal
curvature operator W, and the conformal Jacobi operator JW , respectively, by:
R(x, y, z, w) = g(R(x, y)z, w),
τ := Tr(ρ) =
∑
ijkl g
ilgjkR(ei, ej , ek, el), (1.2)
W(x, y) : z → R(x, y)z − {(m− 1)(m− 2)}−1τ{g(y, z)x− g(x, z)y}
+ (m− 2)−1 {g(ρy, z)x− g(ρx, z)y + g(y, z)ρx− g(x, z)ρy} ,
JW (x) : y →W(y, x)x.
Motivated by the seminal paper of Osserman [23], one studies the spectral properties of the
Jacobi operator J and of the conformal Jacobi operator JW and makes the following definitions:
Definition 1. Let M be a pseudo-Riemannian manifold.
1. M is pointwise Osserman if J has constant eigenvalues on S+P (M) and on S
−
P (M) for
every P ∈M .
2. M is pointwise conformally Osserman if JW has constant eigenvalues on S
+
P (M) and
on S−P (M) for every P ∈M .
We refer to [16] for a more complete discussion of Osserman geometry as that lies beyond
the scope of our present endeavors.
Similarly, motivated by the seminal papers of Stanilov and Videv [26], of Tsankov [27], and
of Videv [28] one studies the commutativity properties of these operators:
Definition 2. Let M be a pseudo-Riemannian manifold.
1. M is Jacobi–Tsankov if J (ξ1)J (ξ2) = J (ξ2)J (ξ1) for all ξi.
2. M is mixed-Tsankov if R(ξ1, ξ2)J (ξ3) = J (ξ3)R(ξ1, ξ2) for all ξi.
3. M is skew-Tsankov if R(ξ1, ξ2)R(ξ3, ξ4) = R(ξ3, ξ4)R(ξ1, ξ2) for all ξi.
4. M is Jacobi–Videv if J (ξ)ρ = ρJ (ξ) for all ξ.
5. M is skew-Videv if R(ξ1, ξ2)ρ = ρR(ξ1, ξ2) for all ξi. This has also been called Ricci
semi-symmetric by some authors.
In this brief note, we survey some recent results concerning these concepts; we refer to [16,
17, 18] for a discussion of some previous results in this area.
Our first task is to pass to the algebraic setting.
Stanilov–Tsankov–Videv Theory 3
Definition 3. Let 〈·, ·〉 be a non-degenerate bilinear form of signature (p, q) on a finite dimen-
sional real vector space V . Let R ∈ ⊗4V ∗ be a 4-tensor. We say thatM = (V, 〈·, ·〉, R) is a model
and that R is an algebraic curvature tensor if R satisfies the usual curvature identities for all
x, y, z, and w:
R(x, y, z, w) = −R(y, x, z, w) = R(z, w, x, y),
R(x, y, z, w) +R(y, z, x, w) +R(z, x, y, w) = 0.
The associated algebraic curvature operator R is then defined by using the inner product to
raise indices; this skew-symmetric operator is characterized by the identity:
〈R(x, y)z, w〉 = R(x, y, z, w).
The Jacobi operator, the Ricci operator, the Weyl conformal curvature operator, and the con-
formal Jacobi operator are then defined as in equations (1.1) and (1.2). The concepts of
Definitions 1 and 2 extend naturally to this setting.
If P is a point of a pseudo-Riemannian manifold M, then the associated model is defined by
M(M, P ) := (TPM, gP , RP ).
We note that every model M is geometrically realizable; this means that given M, there is
(M, P ) such that M(M, P ) is isomorphic to M – see, for example, the discussion in [17].
One has the following examples of algebraic curvature tensors.
Example 1.
1. If ψ is self-adjoint with respect to 〈·, ·〉, one defines an algebraic curvature tensor
Rψ(x, y, z, w) = 〈ψx,w〉〈ψy, z〉 − 〈ψx, z〉〈ψy,w〉.
Taking ψ = id and rescaling yields the algebraic curvature tensor of constant sectional
curvature c:
Rc(x, y, z, w) = c{〈x,w〉〈y, z〉 − 〈x, z〉〈y, w〉}.
One says that a model M or a pseudo-Riemannian manifold M has constant sectional
curvature c if R = Rc for some constant c.
2. If φ is skew-adjoint with respect to 〈·, ·〉, one defines an algebraic curvature tensor
Rφ(x, y, z, w) = 〈φy, z〉〈φx,w〉 − 〈φx, z〉〈φy,w〉 − 2〈φx, y〉〈φz,w〉.
Remark 1. The space of algebraic curvature tensors is spanned as a linear space by the tensors
given in Example 1 (1) or in Example 1 (2) [13]; we also refer to [12].
Our first result is the equivalence of conditions (1) and (2) and of (4) and (5) in Definition 2;
if M is a model or if M is a pseudo-Riemannian manifold, then Jacobi–Tsankov and mixed-
Tsankov are equivalent conditions. Similarly Jacobi–Videv and skew-Videv are equivalent con-
ditions. This follows from the following result [20]:
Theorem 1. Let M be a model and let T be a self-adjoint linear transformation of V . Then
the following assertions are equivalent:
1. R(x, y)T = TR(x, y) for all x, y ∈ V .
4 M. Brozos-Va´zquez et al.
2. J (x)T = TJ (x) for all x ∈ V .
3. R(Tx, y, z, w) = R(x, Ty, z, w) = R(x, y, Tz, w) = R(x, y, z, Tw) for all x, y, z, w in V .
Here is a brief outline of the remainder of this article. In Section 2, we study Jacobi–
Tsankov models and manifolds. In Section 3, we study skew-Tsankov models and manifolds. In
Section 4, we study Jacobi–Videv models and manifolds. In Section 5, we recall some general
results concerning conformal Osserman geometry. In Section 6, we study these concepts in the
context of Walker manifolds of signature (2,2).
2 Jacobi–Tsankov models and manifolds
We first turn to the Riemannian setting in the following result [9]:
Theorem 2. If M is a Jacobi–Tsankov Riemannian model, then R = 0.
Proof. We can sketch the proof as follows. Since {J (x)}x∈V form a family of commuting self-
adjoint operators, we can simultaneously diagonalize these operators to decompose V = ⊕λVλ
so J (x) = λ(x) id on Vλ. If x ∈ V , decompose x = ⊕xλ for xλ ∈ Vλ. Let
O = {x ∈ V : xλ 6= 0 for all λ};
this is an open dense subset of V . If x ∈ O, since J (x)x = 0, λ(x) = 0 for all λ. Since O
is dense and λ(·) is continuous, λ(x) = 0 for all x so J (x) = 0 for all x; the usual curvature
symmetries now imply the full curvature tensor R vanishes. 
Definition 4. One says that a model M or a pseudo-Riemannian manifold M is orthogonally
Jacobi–Tsankov if J (x)J (y) = J (y)J (x) for all vectors x and y with x ⊥ y.
One has the following classification result [9]; we also refer to a related result [27] if M is
a hypersurface in Rm+1.
Theorem 3.
1. Let M = (V, 〈·, ·〉, R) be a Riemannian model. Then M is orthogonally Jacobi–Tsankov if
and only if one of the following conditions holds:
(a) R = cRid has constant sectional curvature c for some c ∈ R.
(b) dim(V ) is even and R = cRΘ is defined by Example 1 (2) where Θ is a Hermitian
almost complex structure on (V, 〈·, ·〉) and where c ∈ R.
2. Let M be a Riemannian manifold of dimension m.
(a) If m > 2, then M is orthogonally Jacobi–Tsankov if and only if M has constant
sectional curvature c.
(b) If m = 2, then M is always orthogonally Jacobi–Tsankov.
Definition 5. We say that a model M or a pseudo-Riemannian manifold M is conformally
Jacobi–Tsankov if JW (x)JW (y) = JW (y)JW (x) for all x and y. We say that M or M is
orthogonally conformally Jacobi–Tsankov if JW (x)JW (y) = JW (y)JW (x) for all vectors x and y
with x ⊥ y.
Remark 2. These are conformal notions – if M is conformally equivalent to M1, then M is
conformally Jacobi–Tsankov (resp. orthogonally conformally Jacobi–Tsankov) if and only if M1
is conformally Jacobi Tsankov (resp. orthogonally conformally Jacobi–Tsankov). We refer to [3]
for further details.
Stanilov–Tsankov–Videv Theory 5
We have the following useful result:
Theorem 4. A Riemannian model M is orthogonally conformally Jacobi–Tsankov if and only
if W = 0.
Proof. Let W be the associated Weyl conformal curvature operator. Then W is an algebraic
curvature tensor which is orthogonally-Jacobi Tsankov. Thus Theorem 3 yields either that
W = cRid or that W = cRΘ. Since the scalar curvature defined by the tensors Rid and RΘ is
non-zero, we may conclude c = 0. 
There are non-trivial examples of Jacobi–Tsankov manifolds and models in the higher signa-
ture setting.
Definition 6. We say that a modelM = (V, 〈·, ·〉, R) is indecomposable if there is no non-trivial
orthogonal decomposition V = V1 ⊕ V2 which induces a decomposition R = R1 ⊕R2.
We refer to [7] for the proof of the following result:
Theorem 5. Let M be a model.
1. If M is Jacobi–Tsankov, then J (x)2 = 0 for all x in V .
2. If M is Jacobi–Tsankov and Lorentzian, then R = 0.
3. Let M be indecomposable with dim(M) < 14. The following conditions are equivalent:
(a) V = U ⊕ U¯ and R = RU ⊕ 0 where U and U¯ are totally isotropic subspaces.
(b) M is Jacobi–Tsankov.
Either (3a) or (3b) implies that R(x, y)z ∈ U¯ and that R(x, y)R(u, v)z = 0 for all x, y,
z, u, v ∈ V , that J (x)J (y) = 0 for all x, y ∈ V , and that M is skew-Tsankov.
The condition J (x)2 = 0 does not imply that M is Jacobi–Tsankov [7]:
Example 2. Let 〈·, ·〉 be an inner product of signature (4, 4) on R8. Choose skew-symmetric
endomorphisms {e1, e2, e3, e4} so that
e21 = e
2
2 = id, e
2
3 = e
2
4 = − id, and eiej + ejei = 0 for i 6= j.
Note that this gives a suitable Clifford module structure to R8. Set
φ1 = e1 + e3 and φ2 = e2 + e4.
Adopt the notation of Example 1 (2) to define Rφi . Then
M := (R8, 〈·, ·〉, Rφ1 +Rφ2)
is not Jacobi–Tsankov but satisfies J (x)2 = 0 for all x.
We have the following example [7] that shows that the structure of Theorem 5 (3a) is geo-
metrically realizable:
Example 3. Let (x1, . . . , xp, y1, . . . , yp) be the usual coordinates on R2p. Let M = (R2p, g)
where g(∂xi , ∂yj ) = δij and let g(∂xi , ∂xj ) = gij(x). Then there exists a decomposition T (R
2p) =
U ⊕ U¯ where U and U¯ are totally isotropic so that R(x, y)z ∈ U¯ and that R(x, y)R(u, v)z = 0
for all x, y, z, u, v ∈ V . Furthermore, for generic g, the model M(M, P ) is indecomposable for
all P ∈ R2p.
6 M. Brozos-Va´zquez et al.
The restriction in Theorem 5 that dim(V ) < 14 is essential. We have the following [7]:
Example 4. Let {αi, α∗i , βi,1, βi,2, β4,1, β4,2}1≤i≤3 be a basis for R
14. Define M6,8 by:
〈αi, α
∗
i 〉 = 〈βi,1, βi,2〉, 1 ≤ i ≤ 3; 〈β4,1, β4,1〉 = 〈β4,2, β4,2〉 = −
1
2 ; 〈β4,1, β4,2〉 =
1
4 ;
Rα2,α1,α1,β2,1 = Rα3,α1,α1,β3,1 = Rα3,α2,α2,β3,2 = 1,
Rα1,α2,α2,β1,2 = Rα1,α3,α3,β1,1 = Rα2,α3,α3,β2,2 = 1,
Rα1,α2,α3,β4,1 = Rα1,α3,α2,β4,1 = Rα2,α3,α1,β4,2 = Rα2,α1,α3,β4,2 = −
1
2 .
Then M6,8 has signature (6, 8), M6,8 is Jacobi–Tsankov, M6,8 is not skew-Tsankov, and there
exist x and y so that J (x)J (y) 6= 0.
Furthermore, this example is geometrically realizable [10]:
Example 5. Take coordinates {xi, x∗i , yi,1, yi,2, y4,1, y4,2}i=1,2,3 for R
14. Let ai,j ∈ R and let
M6,8 := (R14, g) where:
g(∂xi , ∂x∗i ) = g(∂yi,1 , ∂yi,2) = 1, g(∂y4,1 , ∂y4,1) = g(∂y4,2 , ∂y4,2) = −
1
2 ,
g(∂y4,1 , ∂y4,2) =
1
4 , g(∂x1 , ∂x1) = −2a2,1x2y2,1 − 2a3,1x3y3,1,
g(∂x2 , ∂x2) = −2a3,2x3y3,2 − 2a1,2x1y1,2, g(∂x3 , ∂x3) = −2a1,1x1y1,1 − 2a2,2x2y2,2,
g(∂x1 , ∂x2) = 2(1− a2,1)x1y2,1 + 2(1− a1,2)x2y1,2,
g(∂x2 , ∂x3) = x1y4,1 + 2(1− a3,2)x2y3,2 + 2(1− a2,2)x3y2,2,
g(∂x1 , ∂x3) = x2y4,2 + 2(1− a3,1)x1y3,1 + 2(1− a1,1)x3y1,1.
Then M has the model M6,8 and M is locally symmetric if and only if
a1,1 + a2,2 + a3,1a3,2 = 2, 3a2,1 + 3a3,1 + 3a1,2a1,1 = 4,
3a1,2 + 3a3,2 + 3a2,1a2,2 = 4.
We note that the relations of Example 5 have non-trivial solutions. One may take, for
example, a1,1 = a2,2 = 1, a1,2 = a2,1 = 23 , and a3,1 = a3,2 = 0.
3 Skew-Tsankov models and manifolds
Riemannian skew-Tsankov models are completely classified [8]:
Theorem 6. Let M be a Riemannian skew-Tsankov model. Then there is an orthogonal direct
sum decomposition V = V1⊕ · · · ⊕ Vk ⊕U where dim(Vk) = 2 and where R = R1⊕ · · · ⊕Rk ⊕ 0.
Proof. One has that {R(ξ, η)}ξ,η∈V is a collection of commuting skew-adjoint endomorphisms.
As the inner product is definite, there exists an orthogonal decomposition of V so that each
endomorphism R(ξ, η) decomposes as a direct sum of 2× 2 blocks
(
0 a(ξ, η)
−a(ξ, η) 0
)
.
The desired result then follows from the curvature symmetries. 
The situation in the geometric context is less clear. We refer to [8] for the following 3-di-
mensional and 4-dimensional examples which generalize previous examples found in [27]. We
say that M is an irreducible Riemannian manifold if there is no local product decomposition.
Stanilov–Tsankov–Videv Theory 7
Example 6.
1. Let M = (0,∞) × N where N is a Riemann surface with scalar curvature τN 6≡ 1. Give
M the warped product metric ds2 = dt2 + t2ds2N . Then M := (M, gM ) is an irreducible
skew-Tsankov manifold with τM = t−2(τN − 1).
2. Let (x1, x2, x3, x4) be the usual coordinates on R4. Let Mβ = (R4, g) where ds2 = x23dx
2
1+
(x3 + βx4)2dx22 + dx
2
3 + dx
2
4. Then Mβ is an irreducible skew-Tsankov manifold with
τβ = −2x
−1
3 (x3 + βx4)
−1. Mβ is not isometric to Mβ¯ for 0 < β < β¯.
In the higher signature setting, we note that Example 3 provides examples of neutral signature
pseudo-Riemannian manifolds with R(x, y)R(z, w) = 0 for all x, y, z, w. There are, however,
less trivial examples.
Definition 7. We say M is 3-skew nilpotent if
1. There exist ξi with R(ξ1, ξ2)R(ξ3, ξ4) 6= 0 and
2. For all ξi, one has R(ξ1, ξ2)R(ξ3, ξ4)R(ξ5, ξ6) = 0.
We refer to [14] for the proof of:
Example 7. Let (x, u1, . . . , um−2, y) be coordinates on Rm. Let f = f(~u) be smooth. Let
Ξ be a non-degenerate bilinear form on Rm−2. Consider M := (Rm, g) where the non-zero
components of g are given by:
g(∂x, ∂x) = −2f(~u), g(∂x, ∂y) = 1, g(∂ua , ∂ub) = Ξab.
Then M is skew-Tsankov and 3-skew nilpotent; it need not be Jacobi–Tsankov.
4 Jacobi–Videv models and manifolds
One says M is Einstein if ρ is a scalar multiple of the identity. More generally:
Definition 8. One says M is pseudo-Einstein if ρ either has a single real eigenvalue λ or has
exactly two eigenvalues which are complex conjugates µ and µ¯.
It is immediate that pseudo-Einstein implies Einstein in the Riemannian setting as ρ is
diagonalizable if the metric is positive definite.
We refer to [19] for the proof of the following result; see also [22] for related work in the
4-dimensional context.
Theorem 7. Let M be an indecomposable model which is Jacobi–Videv. Then M is pseudo-
Einstein.
Proof. Let m := dim(V ). Let λ ∈ C have non-negative real part. Set
Vλ := {v ∈ V : (T − λ)
m(T − λ¯)mv = 0}.
We then have the Jordan decomposition V = ⊕λVλ as an orthogonal direct sum of generalized
eigenspaces of ρ. Since J (x) preserves this decomposition, it follows that J = ⊕λJλ. The
curvature symmetries then imply that R = ⊕λRλ. Since M is assumed indecomposable, there
is only one Vλ 6= {0} and thus M is pseudo-Einstein. 
This shows, in the Riemannian setting, that an indecomposable model is Jacobi–Videv if and
only if it is Einstein. The condition that M is pseudo-Einstein does not, however, imply that
M is Jacobi–Videv in the higher signature setting as the following [20] shows:
8 M. Brozos-Va´zquez et al.
Example 8. Let {x1, x2, x3, x4} be coordinates on R4. Let M = (R4, g) where
g(∂x1 , ∂x4) = g(∂x2 , ∂x2) = g(∂x3 , ∂x3) = 1 and g(∂x1 , ∂x3) = e
x2 .
Then M is a homogeneous Lorentz manifold and M is pseudo-Einstein with Rank(ρ) = 2,
Rank(ρ2) = 1, and Rank(ρ3) = 0. Thus M is pseudo-Einstein. However M is not Jacobi–
Videv.
We also have [20]
Example 9. Let {x1, x2, x3, x4} be coordinates on R4. Let M = (R4, g) where
g(∂x1 , ∂x3) = g(∂x2 , ∂x4) = 1, g(∂x3 , ∂x3) = −g(∂x4 , ∂x4) = sx1x2,
g(∂x3 , ∂x4) =
s
2(x
2
2 − x
2
1).
Then M is locally symmetric of signature (2, 2), M is Jacobi–Videv, M is skew-Tsankov, and
M is conformal Osserman. M is neither Jacobi–Tsankov nor Osserman. M is pseudo-Einstein
with ρ2 = −s2 id.
Example 10. Setting
g(∂x1 , ∂x3) = g(∂x2 , ∂x4) = 1, g(∂x3 , ∂x3) = −g(∂x4 , ∂x4) =
s
2(x
2
2 − x
2
1),
g(∂x3 , ∂x4) = −sx1x2
yields a local symmetric space of signature (2, 2) which is Einstein. This manifold is Jacobi–
Videv and skew-Tsankov. It is neither Jacobi–Tsankov, Osserman, nor conformal Osserman.
We can give a general ansatz which constructs such examples in the algebraic setting; we do
not know if these examples are geometrically realizable in general:
Example 11. Let M = (V, (·, ·), R) be a model. We complexify and let
U := V ⊗R C.
We extend (·, ·) and R to be complex multi-linear. Let {ei} be an orthonormal basis for V .
Let {e+i := ei, e
−
i :=
√
−1ei} be a basis for the underlying real vector space U := V ⊕
√
−1V .
Let < and = denote the real and imaginary parts of a complex number, respectively. It is then
immediate that
〈·, ·〉 := <{(·, ·)} and S(·, ·, ·, ·) = ={R(·, ·, ·, ·)}
define a model N := (U, 〈·, ·〉, S). One has that the non-zero components of 〈·, ·〉 are 〈e+i , e
+
i 〉 = 1
and 〈e−i , e
−
i 〉 = −1. Thus the metric has neutral signature. Furthermore, the non-zero compo-
nents of S are given by:
S(e−i , e
+
j , e
+
k , e
+
l ) = S(e
+
i , e
−
j , e
+
k , e
+
l ) = S(e
+
i , e
+
j , e
−
k , e
+
l )
= S(e+i , e
+
j , e
+
k , e
−
l ) = R(ei, ej , ek, el),
S(e+i , e
−
j , e
−
k , e
−
l ) = S(e
−
i , e
+
j , e
−
k , e
−
l ) = S(e
−
i , e
−
j , e
+
k , e
−
l )
= S(e−i , e
−
j , e
−
k , e
+
l ) = −R(ei, ej , ek, el).
We refer to [20] for the proof of the following result:
Theorem 8. Adopt the notation of Example 11. If M is a Riemannian Einstein model with
ρM = s id, then N is a Jacobi–Videv pseudo-Einstein neutral signature model with ρ2N = −4s
2 id.
Stanilov–Tsankov–Videv Theory 9
Definition 9. Let M = (V, 〈·, ·〉, R) be a model. Let {v1, . . . , vk} be an orthonormal basis for
a non-degenerate k-plane pi ⊂ V . Let εi := 〈vi, vi〉 be ±1. One defines the higher order Jacobi
operator by setting:
J (pi) :=
k∑
i=1
εiJ (vi).
The operator J (pi) is independent of the particular orthonormal basis chosen; we refer to
[21, 24, 25] for a further discussion of this operator. If pi = V , then J (pi) = ρ. If pi =
Span(x) where x is a unit spacelike vector, then J (pi) = J (x). Thus J (pi) can be thought of as
interpolating between the Jacobi operator and the Ricci operator.
Definition 10. Let M be a model of signature (p, q). We say that (r, s) is admissible if and
only if
0 ≤ r ≤ p, 0 ≤ s ≤ q, and 1 ≤ r + s ≤ m− 1.
Equivalently, (r, s) is admissible if and only if the Grassmannian of linear subspaces of signature
(r, s) has positive dimension.
One has the following useful characterization [19]:
Theorem 9. The following properties are equivalent for M = (V, 〈·, ·〉, R):
1. M is Jacobi–Videv, i.e. J (x)ρ = ρJ (x) for all x ∈ V .
2. There exists (r, s) admissible so J (pi)J (pi⊥) = J (pi⊥)J (pi) for every non-degenerate sub-
space pi of signature (r, s).
3. There exists (r, s) admissible so J (pi)ρ = ρJ (pi) for every non degenerate subspace pi of
signature (r, s).
4. J (pi)J (pi⊥) = J (pi⊥)J (pi) for every non-degenerate linear subspace pi.
5. J (pi)ρ = ρJ (pi) for every non-degenerate linear subspace pi ⊂ V .
5 Conformal Osserman geometry
We refer to [1, 3] for the proof of the following result:
Theorem 10. Let M be a conformally Osserman pseudo-Riemannian manifold of dimension m.
1. If M is Riemannian and if m is odd, then M is locally conformally flat.
2. If M is Riemannian, if m ≡ 2 mod 4, if m ≥ 10, and if W(P ) 6= 0, then there is an
open neighborhood of P in M which is conformally equivalent to an open subset of either
complex projective space with the Fubini–Study metric or the negative curvature dual.
3. If M is Lorentzian, then M is locally conformally flat.
We also recall the following result [2, 5]:
Theorem 11. Let M be a 4-dimensional model of arbitrary signature.
1. M is conformally Osserman if and only if M is either self-dual or anti-self-dual.
2. If M is Riemannian, then M is conformally Osserman if and only if there exists a quater-
nion structure {I, J,K} on V and constants λI , λJ , λK with λI + λJ + λK = 0 so that
R = λIRI + λJRJ + λKRK where RI , RJ , and RK are given by Example 1 (2).
10 M. Brozos-Va´zquez et al.
6 Walker geometry
One says M is a Walker manifold of signature (2, 2) if it admits a parallel totally isotropic
2-plane field; this implies [29, 30] that locally M is isometric to a metric on R4 with non-zero
components
g(∂x1 , ∂x3) = g(∂x2 , ∂x4) = 1, g(∂x3 , ∂x4) = g34,
g(∂x3 , ∂x3) = g33, g(∂x4 , ∂x4) = g44.
The geometry of Walker manifolds with g34 = 0 has been studied in [11]. We impose
a different condition by setting g33 = g44 = 0 so the non-zero components of the metric are
given by:
g(∂x1 , ∂x3) = g(∂x2 , ∂x4) = 1 and g(∂x3 , ∂x4) = g34. (6.1)
By Theorem 11, M is conformally Osserman if and only if M is either self-dual or anti-self-
dual. One has [5] that:
Theorem 12. Let M = (R4, g) where g is given by equation (6.1).
1. M is self-dual if and only if g34 = x1p(x3, x4) + x2q(x3, x4) + s(x3, x4).
2. M is anti-self-dual if and only if g34 = x1p(x3, x4) + x2q(x3, x4) + s(x3, x4) + ξ(x1, x4) +
η(x2, x3) with p/3 = q/4 and g34p/3 − x1p/34 − x2p/33 − s/34 = 0.
We refer to [4] for the following results:
Theorem 13. Let M = (R4, g) where g is given by equation (6.1).
1. The following conditions are equivalent:
(a) M is Osserman.
(b) M is Einstein.
(c) ρ = 0.
(d) g34 = x1p(x3, x4) + x2q(x3, x4) + s(x3, x4) where p = −2a4(a0 + a3x3 + a4x4)−1, and
q = −2a3(a0 + a3x3 + a4x4)−1 for (a0, a3, a4) 6= (0, 0, 0).
(e) J (x)2 = 0 for all x.
(f) M is Jacobi–Tsankov.
2. The following conditions are equivalent:
(a) M is Jacobi–Videv.
(b) M is skew-Tsankov.
(c) g34 = x1p(x3, x4) + x2q(x3, x4) + s(x3, x4) where p/3 = q/4.
A feature of these examples is that the warping functions are affine functions of x1 and x2.
We return to the general setting of Walker signature (2, 2) geometry. Let ∇ be a torsion free
connection on a 2-dimensional manifold N . Let (x3, x4) be local coordinates on N . We expand
∇∂xi∂xj =
∑
k
Γij
k∂xk for i, j, k = 3, 4
to define the Christoffel symbols of ∇. Let ω = x1dx3+x2dx4 ∈ T ∗N ; the pair (x1, x2) gives the
dual fiber coordinates. Let ξ = ξij(x3, x4) ∈ C∞(S2(T ∗N)) be an auxiliary symmetric bilinear
form.
Stanilov–Tsankov–Videv Theory 11
Definition 11. The deformed Riemannian extension is the Walker metric on T ∗N defined by
setting [15]
g(∂x1 , ∂x3) = g(∂x2 , ∂x4) = 1,
g(∂x3 , ∂x3) = −2x1Γ33
3(x3, x4)− 2x2Γ33
4(x3, x4) + ξ33(x3, x4),
g(∂x3 , ∂x4) = −2x1Γ34
3(x3, x4)− 2x2Γ34
4(x3, x4) + ξ34(x3, x4),
g(∂x4 , ∂x4) = −2x1Γ44
3(x3, x4)− 2x2Γ44
4(x3, x4) + ξ44(x4, x4).
Definition 12. Let ρN (x, y) := Tr(z → R∇(z, x)y) be the affine Ricci tensor. We may decom-
pose this 2-tensor into symmetric and anti-symmetric parts by defining:
ρsN (x, y) :=
1
2(ρN (x, y) + ρN (y, x)) and,
ρaN (x, y) :=
1
2(ρN (x, y)− ρN (y, x)).
The Jacobi operator is defined by setting J∇(x) : y → R∇(y, x)x. We say that N := (N,∇) is
affine Osserman if J∇(x) is nilpotent or, equivalently, if Spec{J∇(x)} = {0} for all x.
We refer to [4] for the proof of the following result:
Theorem 14.
1. M is skew-Tsankov if and only if ρaN = 0.
2. M is Osserman if and only if N is affine Osserman if and only if ρsN = 0.
3. ρaN = 0 or ρ
s
N = 0 if and only if M is Jacobi–Videv.
4. ρN = 0 if and only if M is Jacobi–Tsankov.
Remark 3. This shows the notions Jacobi–Videv, and Jacobi–Tsankov, and skew-Tsankov are
inequivalent notions.
If M is conformally Osserman, let mλ be the minimal polynomial of JW and let SpecW be
the spectrum of JW . One has [5]:
Theorem 15. Let M = (R4, g) be the Walker manifold with non-zero metric components:
g(∂x1 , ∂x3) = g(∂x2 , ∂x4) = 1, and g(∂x3 , ∂x4) = g34.
The following choices of g34 make M conformal Osserman with:
1. The Jordan normal form does not change from point to point:
(a) If g34 = x21 − x
2
2, then mλ = λ(λ
2 − 14) and SpecW = {0, 0,±
1
2}.
(b) If g34 = x21 + x
2
2, then mλ = λ(λ
2 + 14) and SpecW =
{
0, 0,±
√
−1
2
}
.
(c) If g34 = x1x4 + x3x4, then mλ = λ2 and SpecW = {0}.
(d) If g34 = x21, then mλ = λ
3 and SpecW = {0}.
2. SpecW = {0} but the Jordan normal form changes from point to point.
(a) If g34 = x2x24 + x
2
3x4, then mλ = λ
3 if x4 6= 0, mλ = λ2 if x4 = 0 and x3 6= 0, and
mλ = λ if x3 = x4 = 0.
(b) If g34 = x2x24 + x3x4, then mλ = λ
3 if x4 6= 0, and mλ = λ2 if x4 = 0.
(c) If g34 = x1x23, then mλ = λ
3 if x3 6= 0, and mλ = λ if x3 = 0.
(d) If g34 = x1x3+x2x4, then mλ = λ2 if x1x3+x2x4 6= 0, and mλ = λ if x1x3+x2x4 = 0.
12 M. Brozos-Va´zquez et al.
3. The eigenvalues can change from point to point:
(a) If g34 = x41 + x
2
1 − x
4
2 − x
2
2, then SpecW =
{
0, 0,±12
√
(6x21 + 1)(6x
2
2 + 1)
}
.
(b) If g34 = x41 + x
2
1 + x
4
2 + x
2
2, then SpecW =
{
0, 0,±12
√
−(6x21 + 1)(6x
2
2 + 1)
}
.
(c) If g34 = x31 − x
3
2, then SpecW = {0, 0,±
3
2
√
x1x2}.
We conclude our discussion with the following result [6]:
Theorem 16. Of the manifolds given above in Theorem 15, only the manifold with g34 = x21 is
curvature homogeneous and only the manifold with g34 = x1x4 + x3x4 is geodesically complete.
Acknowledgements
The research of M. Brozos-Va´zquez and of P. Gilkey was partially supported by the Max
Planck Institute for Mathematics in the Sciences (Germany) and by Project MTM2006-01432
(Spain). The research of E. Garc´ıa–R´ıo and of R. Va´zquez-Lorenzo was partially supported
by PGIDIT06PXIB207054PR (Spain). The research of S. Nikcˇevic´ was partially supported by
Project 144032 (Serbia). It is a pleasure to acknowledge helpful conversations with C. Dunn,
E. Puffini, and Z. Zhelev concerning these and related matters.
References
[1] Blazˇic´ N., Gilkey P., Conformally Osserman manifolds and conformally complex space forms, Int. J. Geom.
Methods Mod. Phys. 1 (2004), 97–106, math.DG/0311263.
[2] Blazˇic´ N., Gilkey P., Conformally Osserman manifolds and self-duality in Riemannian geometry, in Pro-
ceedings of the Conference “Differential Geometry and Its Applications” (August 30 – September 3, 2004,
Charles University, Prague, Czech Republic), Editors J. Bures, O. Kowalski, D. Krupka and J. Slovak,
MATFYZPRESS, 2005, 15–18, math.DG/0504498.
[3] Blazˇic´ N., Gilkey P., Nikcˇevic´ S., Simon U., The spectral geometry of the Weyl conformal tensor, Banach
Center Publ. 69 (2005), 195–203, math.DG/0310226.
[4] Brozos-Va´zquez M., Garc´ıa–R´ıo E., Gilkey P., Va´zquez-Lorenzo R., Examples of signature (2, 2) manifolds
with commuting curvature operators, J. Phys. A: Math. Theor., to appear, arXiv:0708.2770.
[5] Brozos-Va´zquez M., Garc´ıa–R´ıo E., Va´zquez-Lorenzo R., Conformally Osserman four-dimensional manifolds
whose conformal Jacobi operators have complex eigenvalues, Proc. Royal Soc. A 462 (2006), 1425–1441.
[6] Brozos-Va´zquez M., Garc´ıa–R´ıo E., Gilkey P., Va´zquez–Lorenzo R., Completeness, Ricci blowup, the Os-
serman and the conformal Osserman condition for Walker signature (2,2) manifolds, in Proceedings of XV
International Workshop on Geometry and Physics, to appear, math.DG/0611279.
[7] Brozos-Va´zquez M., Gilkey P., Pseudo-Riemannian manifolds with commuting Jacobi operators, Rend. Circ.
Mat. Palermo 55 (2006), 163–174, math.DG/0608707.
[8] Brozos-Va´zquez M., Gilkey P., The global geometry of Riemannian manifolds with commuting curvature
operators, J. Fixed Point Theory Appl. 1 (2007), 87–96, math.DG/0609500.
[9] Brozos-Va´zquez M., Gilkey P., Manifolds with commuting Jacobi operators, J. Geom. 86 (2007), 21–30,
math.DG/0507554.
[10] Brozos-Va´zquez M., Gilkey P., Nikcˇevic´ S., Jacobi–Tsankov manifolds which are not 2-step nilpotent, in
Proceedings of the Conference “Contemporary Geometry and Related Topics” (June 26 – July 2, 2005,
Belgrade, Serbia and Montenegro), Editors N. Bokan, M. Djoric´, A.T. Fomenko, Z. Rakic, B. Wegner and
J. Wess, University of Belgrade, Serbia, 2006, 63–79, math.DG/0609565.
[11] Chaichi M., Garc´ıa–R´ıo E., Matsushita Y., Curvature properties of four-dimensional Walker metrics, Clas-
sical Quantum Gravity 22 (2005), 559–577.
[12] Dı´az-Ramos J.C., Garc´ıa-R´ıo E., A note on the structure of algebraic curvature tensors, Linear Algebra
Appl. 382 (2004), 271–277.
[13] Fiedler B., Determination of the structure of algebraic curvature tensors by means of Young symmetrizers,
Seminaire Lotharingien de Combinatoire B48d (2003), 20 pages, math.CO/0212278.
Stanilov–Tsankov–Videv Theory 13
[14] Fiedler B., Gilkey P., Nilpotent Szabo´, Osserman and Ivanov–Petrova pseudo Riemannian manifolds, Con-
temp. Math. 337 (2003), 53–64, math.DG/0211080.
[15] Garc´ıa–R´ıo E., Kupeli D.N., Va´zquez-Abal M.E., Va´zquez-Lorenzo R., Affine Osserman connections and
their Riemann extensions, Differential Geom. Appl. 11 (1999), 145–153.
[16] Garc´ıa-R´ıo E., Kupeli D., Va´zquez-Lorenzo R., Osserman manifolds in semi-Riemannian geometry, Lecture
Notes in Mathematics, Vol. 1777, Springer-Verlag, Berlin, 2002.
[17] Gilkey P., Geometric properties of natural operators defined by the Riemann curvature tensor, World Scien-
tific, 2001.
[18] Gilkey P., The geometry of curvature homogeneous pseudo Riemannian manifolds, Imperial College Press,
2007.
[19] Gilkey P., Puffini E., Videv V., Puffini–Videv models and manifolds, J. Geom., to appear,
math.DG/0605464.
[20] Gilkey P., Nikcˇevic´ S., Pseudo-Riemannian Jacobi–Videv manifolds, Int. J. Geom. Methods Mod. Phys. 4
(2007), 727–738, arXiv:0708.1096.
[21] Gilkey P., Stanilov G., Videv V., Pseudo Riemannian manifolds whose generalized Jacobi operator has
constant characteristic polynomial, J. Geom. 62 (1998) 144–153.
[22] Ivanova M., Videv V., Zhelev Z., Four-dimensional Riemannian manifolds with commuting higher order
Jacobi operators, math.DG/0701090.
[23] Osserman R., Curvature in the eighties, Amer. Math. Monthly 97 (1990), 731–756.
[24] Stanilov G., Videv V., On a generalization of the Jacobi operator in the Riemannian geometry, God. Sofij.
Univ., Fak. Mat. Inform. 86 (1994) 27–34.
[25] Stanilov G., Videv V., Four dimensional pointwise Osserman manifolds, Abh. Math. Sem. Univ. Hamburg
68 (1998), 1–6.
[26] Stanilov G., Videv V., On the commuting of curvature operators, in Proceedings of the 33rd Spring Con-
ference of the Union of Bulgarian Mathematicians Borovtes “Mathematics and Education in Mathematics”
(April 1–4, 2004, Sofia), Sofia, 2004, 176–179.
[27] Tsankov Y., A characterization of n-dimensional hypersurface in Euclidean space with commuting curvature
operators, Banach Center Publ. 69 (2005), 205–209.
[28] Videv V., A characterization of the 4-dimensional Einstein Riemannian manifolds using curvature operators,
Preprint.
[29] Walker A.G., Canonical form for a Riemannian space with a parallel field of null planes, Quart. J. Math.,
Oxford Ser. (2) 1 (1950), 69–79.
[30] Walker A.G., Canonical forms. II. Parallel partially null planes, Quart. J. Math., Oxford Ser. (2) 1 (1950),
147–152.