1 Erratum to: Ann Glob Anal Geom (2011) 40:389–409 DOI 10.1007/s10455-011-9261-5

We correct some of the entries of Table 1 in our original paper.

1. Those entries of Table 1 in the original paper listing the numbers of geometric structures on the space of geodesics \(L^{\epsilon }(S^{p,q})\) of the 3-dimensional space \(S^{p,q}\) of non-zero constant curvature with \(p+q =3\) are incorrect. In this note we give the correct values and arguments. In original paper we failed to note that the \((\mathfrak {so}(V) + \mathfrak {so}(V'))\)-module \(V \otimes V'\), where \(V,V'\) are 2-dimensional pseudo-Euclidian spaces, has not only the invariant (para)complex structures \(J = J_V \otimes 1,\, J' = 1 \otimes J_{V'}\), but that also their product \(J'' = J J' = J_V \otimes J_{V'}\) is an invariant (para)complex structure. We next analyze the impact of this observation in detail and correct Table 1 accordingly. The authors would like to thank Henri Anciaux for drawing their attention to this oversight.

2. Let \(E = E^{p+1,q}\) be a pseudo-Euclidean 4-dimensional vector space of signature \((p+1,q)= (4,0), (3,1)\), (2, 2) or (1, 3) with an orthonormal basis \(e_1,e_2,e_3, e_4\) with \(e_4^2 := g(e_4,e_4)=1 \). We denote by \(E' = e_4^{\perp }\) the orthogonal complement to \(e_4\). The pseudo-sphere of unit vectors \( S^{p,q} = \{x \in E,\, x^2=1\}\) is the orbit \(S^{p,q} = SO(E) e_4\). It has the reductive decomposition

$$\begin{aligned} \mathfrak {so}(E) = \mathfrak {so}(E') + e_4\wedge E' , \end{aligned}$$

where \(\mathfrak {so}(E') \simeq \mathfrak {so}_{3,0}, \mathfrak {so}_{2,1}, \mathfrak {so}_{1,2}, \mathfrak {so}_{0,3} \), respectively. The metric of E induces an \(SO(E')\)-invariant metric in the tangent space \(T_0S^{p,q} = e_4 \wedge E'\). The stability subalgebra of the geodesic \(\ell (e_4 \wedge e_3)\) through the origin \(o =e_4 \in S^{p,q}\) in the direction \(e_4 \wedge e_3 \in T_o S^{p,q} = e_4 \wedge E'\) is \(\mathfrak {so}(V) \times \mathfrak {so}(V^{\perp })\), where \(V = \mathrm {span}(e_1, e_2)\) and \(V^{\perp } = \mathrm {span}(e_3,e_4)\). The geodesic is spacelike if \( \epsilon = e_3^2 =1\) and timelike if \(\epsilon = e_3^2 =-1\). The space of such geodesics is identified with \(L^{\epsilon }(S^{p,q}) = SO(E)/ SO(V) \times SO(V^{\perp })\) (up to a covering).

We now denote by \(g_V\) the unique (up to a scale) metric in 2-dimensional SO(V)-module V and by \(\omega _V\) unique ( up to scale ) 2-form. Then for some positive \(\lambda \), the complex structure \(J_V = \lambda g_V^{-1} \circ \omega _V\) is the unique (up to sign) \(\mathfrak {so}(V)\)-invariant complex structure if the metric \(g_V\) has signature (2, 0) or (0, 2). It is the unique (up to sign) paracomplex structure if the signature of \(g_V\) is (1, 1). We use similar notations for \(V^{\perp }\). The reductive decomposition of the symmetric space \(L^{\epsilon }(S^{p,q})\) is given by

$$\begin{aligned} \mathfrak {so}(E) = (\mathfrak {so}(V)\oplus (\mathfrak {so}(V^{\perp })) + V \wedge V^{\perp } . \end{aligned}$$

The tangent space \(T_o L^{\epsilon }(S^{p,q})\) is identified with \( \mathfrak {m}:= V \wedge V^{\perp } \simeq V \otimes V^{\perp }\) with the natural action of the isotropy algebra \(\mathfrak {h}= \mathfrak {so}(V) + \mathfrak {so}(V^{\perp })\). We have therefore proved the following.

Table 1 Invariant geometric structures

Proposition

The \(\mathfrak {h}\)-invariant symmetric bilinear form (respectively, invariant 2-form) on \(\mathfrak {m}\) are of the form

$$\begin{aligned} h_{\lambda , \mu }= \lambda \omega _V \otimes \omega _{V^{\perp }} + \mu g_V \otimes g_{V^{\perp }} ,\;\; \omega _{\lambda , \mu } = \lambda g_V \otimes \omega _{V^{\perp }} + \mu \omega _V \otimes g_{V^{\perp }} \end{aligned}$$

for \(\lambda , \mu \in \mathbb {R}\).

The endomorphisms \(J:= J_V \otimes 1, J^{\perp } := 1 \otimes J_{V^{\perp }}, J' := J \circ J^{\perp } = J_V \otimes J_{V^{\perp }} \) are invariant (para)complex structure in \(\mathfrak {m}\). Moreover, \(J.J^{\perp }\) are skew-symmetric with respect to any metric \(h_{\lambda , \mu }\) and the endomorphism \(J'\) is symmetric.

Since \(S^{p,q}\) is an irreducible symmetric space, any invariant tensor field is parallel. We obtain

Corollary

Any non-degenerate form \(h_{\lambda , \mu }\) defines an invariant pseudo-Riemannian metric, any non-degenerate 2-form \(\omega _{\lambda , \mu }\) defines an invariant symplectic form and the endomorphisms \(J,J^{\perp }, J'\) define integrable (para)complex structures on \(L^{\epsilon }(S^{p,q})\). Furthermore, any pair of the form \((h_{\lambda , \mu }, J)\) or \( (h_{\lambda .\mu }, J^{\perp })\), where \(h_{\lambda , \mu }\) is as above , defines a (para)Kähler structure on \(\mathfrak {m}\).

3. In the following table we indicate the signature of the metrics \(g_V\) and \(g_{V^{\perp }}\) for different values of p and \(\epsilon \) and the type of the endomorphisms \(J= J_V \otimes 1, \, J^{\perp } = 1 \otimes J_{V^{\perp }}, J'= J_V \otimes J_V{\perp }\) ( cx \(=\) complex structure, para \(=\) paracomplex structure).

p

\(\epsilon \)

\(g_V\)

\(g_{V^{\perp }}\)

J

\(J^{\perp }\)

\(J'\)

3

\(+\)

(2,0)

(2,0)

cx

cx

para

2

\(+\)

(1,1)

(2,0)

para

cx

cx

2

(2,0)

(1,1)

cx

para

cx

1

\(+\)

(0,2)

(2,0)

cx

cx

para

1

(1,1)

(1,1)

para

para

cx

0

(0,2)

(1,1)

cx

para

cx

Note also that there is a natural isomorphism \(L^{\pm }(S^{p,q} = L^{\mp }(S^{q,p}_-)\), where \(S^{q,p}_- = \{ x \in E^{q,p+1}, x^2 =-1 \}\) . We finally reproduce Table 1 in its corrected form.