Definition 2.1
AKJCThe group \operatorname{Aut}(G)_g \cong \operatorname{Aut}(\mathfrak{g})_g is called the group of isometric automorphisms of G.
For each left-invariant Riemannian metric on simply connected nonunimodular Lie groups of dimension four, we determine the full group of isometries.
The group \operatorname{Aut}(G)_g \cong \operatorname{Aut}(\mathfrak{g})_g is called the group of isometric automorphisms of G.
A Lie group G is said to be of type (R) if, for every element x in the Lie algebra \mathfrak{g} of G, the endomorphism \operatorname{ad} x : \mathfrak{g} \to \mathfrak{g}, defined by y \mapsto [x, y], has only real eigenvalues. Some authors refer to this condition by saying that \mathfrak{g} has only real roots.
(Lemma 2.2 in [ha2012isometry]). The group L(G) of left translations on G is a normal subgroup of \operatorname{Isom}(G, g) if and only if \operatorname{Isom}(G, g)_e = \operatorname{Aut}(G)_g.
(Corollary (5.3) in [gordon1988isometry]). Let G be a simply connected solvable Lie group whose Lie algebra has only real roots. Let g and g' be two left-invariant metrics on G. Then g is isometric to g' if and only if g' = \psi^{\ast} g for some \psi \in \operatorname{Aut}(G).
(Lemma 11.2. page 62 in [helgason]). Let M be a Riemannian manifold, \phi and \psi two isometries of M onto itself. Suppose there exists a point p \in M for which \phi(p) = \psi(p) and d_p\phi = d_p\psi. Then \phi = \psi.
Let G be a simply connected solvable Lie group of type (R), then for each left invariant Riemannian metric g on G, \operatorname{Isom}(G, g) \cong G \rtimes \operatorname{Aut}(G)_g.
According to the lemma 2.3, it suffices to show that \operatorname{Isom}(G, g)_e = \operatorname{Aut}(G)_g. Let \theta \in \operatorname{Isom}(G, g)_e, then \theta(e) = e and \theta^{\ast}g = g, i.e. g(u, v) = g(\theta_{\ast}^{-1}u, \theta_{\ast}^{-1}v) \,\, \forall u, v\in \mathfrak{g}. On the other hand, since G is a simply connected solvable Lie group of type (R), we have by lemma 2.4, \psi^{\ast}g = g for some \psi \in \operatorname{Aut}(G), i.e. g(u, v) = g(\psi_{\ast}^{-1}u, \psi_{\ast}^{-1}v) \,\, \forall u, v\in \mathfrak{g}. Hence \theta_{\ast} = d_e\theta = \psi_{\ast} = d_e\psi and \theta(e) = \psi(e) = e. By lemma 2.5, we have:
\theta = \psi \in \operatorname{Aut}(G)_g. Therefore \operatorname{Isom}(G, g) \cong G \rtimes \operatorname{Aut}(G)_g. (See page 192 in [ha2012isometry]).
All the Lie algebras in Table 1 are solvable. We will focus on those of type (R) and use the preceding theorem to describe their isometry groups. In fact, as we will show in the following proposition, only four Lie algebras from Table 1 are not of type (R).
All the Lie algebras from Table 1 are solvable of type (R) except the following four Lie algebras: \operatorname{A}_{3, 7}^{\alpha} \oplus \operatorname{A}_1, \operatorname{A}_{4, 6}^{\alpha, \beta}, \operatorname{A}_{4, 11}^{\alpha} and \operatorname{A}_{4, 12}.
For the Lie algebra \operatorname{A}_{3, 7}^{\alpha} \oplus \operatorname{A}_1, the endomorphism \operatorname{ad} e_3 is represented by the matrix
We see that \operatorname{ad} e_3 has two complex eigenvalues which are -\alpha + i and -\alpha - i.
For the Lie algebra \operatorname{A}_{4, 6}^{\alpha, \beta}, the endomorphism \operatorname{ad} e_4 is represented by the matrix
We see that \operatorname{ad} e_4 has two complex eigenvalues which are -\beta + i and -\beta - i.
For the Lie algebra \operatorname{A}_{4, 11}^{\alpha}, the endomorphism \operatorname{ad} e_4 is represented by the matrix
Thus, \operatorname{ad} e_4 has two complex eigenvalues which are -\alpha + i and -\alpha - i.
For the Lie algebra \operatorname{A}_{4, 12}, the endomorphism \operatorname{ad} e_4 is represented by the matrix
Therefore, the element \operatorname{ad} e_4 has two complex eigenvalues which are \pm i.
For all other Lie algebras in Table 1, one can easily verify that \operatorname{ad} e_i, for all i = 1, \dots, 4, have only real eigenvalues.
The simply connected Lie groups associated with these Lie algebras will be described in details in the following subsections.
We can realize the Lie algebra \operatorname{A}_{3, 5}^{\alpha} as a semidirect sum \operatorname{A}_{3, 5}^{\alpha} \cong \mathbb{R}^2 \rtimes_{\sigma_{\alpha}}\mathbb{R} where \sigma_{\alpha}(t) = \begin{bmatrix} -t & 0\\ 0 & -t\alpha \end{bmatrix}.
Consider the following basis \mathscr{B} = \left\lbrace X, Y, Z\right\rbrace of \mathbb{R}^2 \rtimes_{\sigma_{\alpha}}\mathbb{R} where
We recall that if (x, t), (y, r) \in \mathbb{R}^2 \rtimes_{\sigma_{\alpha}}\mathbb{R}, where x, y \in \mathbb{R}^2 and t, r \in \mathbb{R}. Then their bracket is given by
where [, ]_{\mathbb{R}^2} and [, ]_{\mathbb{R}} are the brackets in \mathbb{R}^2 and \mathbb{R} respectively. According to this fact, we obtain that the brackets of elements of \mathscr{B} are
These are exactly the brackets of the Lie algebra \operatorname{A}_{3, 5}^{\alpha}. Hence \operatorname{A}_{3, 5}^{\alpha} \cong \mathbb{R}^2 \rtimes_{\sigma_{\alpha}}\mathbb{R}.
The Lie algebra \operatorname{A}_{3, 5}^{\alpha} is isomorphic to \mathfrak{g}_c for some c \in \mathbb{R}, in fact we have
It is clear that the Lie algebra \mathbb{R}^2 \rtimes_{\sigma_{\alpha}}\mathbb{R} is not isomorphic to \mathbb{R}^2 \rtimes_{\sigma_{I}}\mathbb{R}; Because if they are isomorphic, then there exists a matrix P \in \operatorname{GL}(2, \mathbb{R}) = \operatorname{Aut}(\mathbb{R}^2) myeditcolorand a nonzero real number \lambda such that \sigma_{I}(\lambda t) = P\sigma_{\alpha}(t)P^{-1}. This means that \sigma_{I}(\lambda t) and \sigma_{\alpha}(t) have the same eigenvalues, this implies that t = 0 or \lambda = -1 and \alpha = 1. Both cases are impossible because \operatorname{A}_{3, 5}^{\alpha} is not abelian and 0 < |\alpha| < 1. Therefore \operatorname{A}_{3, 5}^{\alpha} is isomorphic to \mathfrak{g}_c for some c \in \mathbb{R}. In fact, put \lambda = -\frac{1 + \alpha}{2} and c = \frac{4\alpha}{(1 + \alpha)^2}. Then the matrix \sigma_{c}(\lambda t) has eigenvalues -t and -\alpha t. Consequently \sigma_{c}(\lambda t) is similar to \sigma_{\alpha}(t) and hence \operatorname{A}_{3, 5}^{\alpha} and \mathfrak{g}_c are isomorphic.
The following map is bijective
It is clear that the map \varphi is well defined. For the injectivity, let B, C \in \operatorname{Tsup}_{n} such that \varphi(B) = \varphi(C). This is equivalent to the following equality B^{-1}C = (C^{-1}B)^{t}. Put M = C^{-1}B, then B^{-1}C = (C^{-1}B)^{t} \Leftrightarrow M^{-1} = M^{t}. Since \operatorname{Tsup}_{n} is a group, then M, M^{-1} \in \operatorname{Tsup}_{n}. Since M^{-1} = M^{t} \in \operatorname{Tinf}_n, then M \in \operatorname{Tsup}_n \cap \operatorname{Tinf}_n = \operatorname{D}_n^{+}. Hence we have M^{-1} = M^{t} = M, then M = I_n. This implies that B = C and \varphi is injective.
For the surjectivity, let A \in \operatorname{S}_{n} and consider the inner product on \mathbb{R}^{n} defined by
Let \mathcal{B} = \left\lbrace e_1, ..., e_n\right\rbrace be the canonical basis of \mathbb{R}^{n}. The Gram-Schmidt procedure produces an orthonormal basis \mathscr{B} = \left\lbrace v_1, ..., v_n\right\rbrace of the Euclidean space \left( \mathbb{R}^{n}, \langle ., .\rangle\right) such that
This means that Xe_i = v_i. Since \mathscr{B} = \{v_1, ..., v_n\} is an orthonormal basis of (\mathbb{R}^n, \langle ., . \rangle), then
Thus A = (X^{-1})^{t}(X^{-1}) = \varphi(X). Hence \varphi is surjective.
Therefore \varphi is a bijection. This result was already proven in [van2017metrics].
[van2017metrics] The set of inner products on \mathfrak{g} is (n^2 + n)/2-dimensional, and can be identified with the set of upper triangular matrices with positive diagonal entries.
In each nonunimodular 4-dimensional Lie group, we will investigate different metrics that yield different isometry groups. This means that for any other metric, its isometry group is isomorphic to that of one of the investigated metrics.
In this paper, we choose the first class of metrics to be the one with the maximal group of isometric automorphisms.
The isotropy subgroup \operatorname{Isom}(G, g)_e is compact. If \theta is an element of \operatorname{Isom}(G, g)_e, the differential \theta_{\ast} = d_e\theta : \mathfrak{g} \longrightarrow \mathfrak{g} of \theta at e is an orthogonal transformation on the vector space \mathfrak{g} = T_eG. That is, for any x, y \in T_eG: g(\theta_{\ast}^{-1}x, \theta_{\ast}^{-1}y) = g(x, y).
Therefore, by lemma 2.5, \operatorname{Isom}(G, g)_e can be injected in \operatorname{O}(\mathfrak{g}, g). Hence
Throughout this paper, \operatorname{O}(n) denotes the orthogonal group in n dimensions.
The group of isometric automorphisms of G_{2.1} \times \mathbb{R}^2 is given by
The automorphism group of \operatorname{A}_{2} \oplus 2\operatorname{A}_1 = \operatorname{Lie}\left(G_{2.1} \times \mathbb{R}^2\right) consists of elements of the form [christodoulakis2003automorphisms]
We first note that, by Remark 5.2, for any left-invariant Riemannian metric on the Lie group G, we have \operatorname{Aut}(G)_g \subset \operatorname{O}(4). This implies that a_5 = a_9 = a_{13} = 0 in A. Therefore, it suffices to consider automorphisms of the form A = \begin{bmatrix} 1 & 0 & 0 & 0\\ 0 & a_6 & 0 & 0\\ 0 & 0 & a_{11} & a_{12}\\ 0 & 0 & a_{15} & a_{16} \end{bmatrix}. Let A be an automorphism of this form, then
Hence \operatorname{Aut}\left(G_{2.1} \times \mathbb{R}^2\right)_{M_1} = \operatorname{diag}\left\lbrace 1, \pm1, \operatorname{O}(2)\right\rbrace.
For the metric M_2, we have
Hence \operatorname{Aut}\left(G_{2.1} \times \mathbb{R}^2\right)_{M_2} = \operatorname{diag}\left\lbrace 1, 1, \operatorname{O}(2)\right\rbrace \cong \operatorname{O}(2).
For the metric M_3, one can see that
Thus
For the metric M_4, one gets that
Thus
myeditcolorFinally, it is easy to verify that the automorphism \operatorname{diag}\left\lbrace 1, 1, 1, -1\right\rbrace does not preserve M_5. Hence \operatorname{Aut}\left(G_{2.1} \times \mathbb{R}^2\right)_{M_5} = \left\lbrace I_4\right\rbrace.
The isometry group of G_{2.1} \times \mathbb{R}^2 is given by
The possible groups of isometric automorphisms of G_2 are
The automorphism group of \operatorname{2A_2} = \operatorname{Lie}(G_2) consists of elements of the form [christodoulakis2003automorphisms]
Let A be an automorphism of G_2 of the above form , then we obtain that
Therefore \operatorname{Aut}\left(G_2\right)_{M_1} \cong \mathbf{D}(4). This the maximal group of isometric automorphisms on the Lie group G_2. For the metric M_2, it suffices to verify which of the elements of \operatorname{Aut}\left(G_2\right)_{M_1} preserve the metric M_2. We obtain that
Similarly we can see that
Finally, one can easily verify that: A \in \operatorname{Aut}\left(G_2\right)_{M_4} \Leftrightarrow A^tM_4A = M_4 \Leftrightarrow A = I_4.
If we choose the metric U = \begin{bmatrix}
\alpha & 0 & \gamma & 0\\
0 & 1 & 0 & 0\\
0 & 0 & \mu & 0\\
0 & 0 & 0 & 1
\end{bmatrix} \quad \alpha, \mu > 0, \quad \gamma \neq 0.
Its associated symmetric positive definite matrix is M_{\gamma} = \begin{bmatrix}
\frac{1}{\alpha^2} & 0 & \frac{-\gamma}{\alpha^2\mu} & 0\\
0 & 1 & 0 & 0\\
\frac{-\gamma}{\alpha^2\mu} & 0 & \frac{\gamma^2}{\alpha^2\mu^2} + \frac{1}{\mu^2} & 0\\
0 & 0 & 0 & 1
\end{bmatrix}.
Then the group of isometric automorphisms of M_{\gamma} is exactly the one of M_1 or M_2. In fact we have
If we choose the metric U = \begin{bmatrix}
\alpha & 0 & 0 & \nu\\
0 & 1 & 0 & 0\\
0 & 0 & \mu & 0\\
0 & 0 & 0 & 1
\end{bmatrix} \quad \alpha, \mu > 0, \quad \nu \neq 0.
Its associated symmetric positive definite matrix is M_{\nu} = \begin{bmatrix}
\frac{1}{\alpha^2} & 0 & 0 & \frac{-\nu}{\alpha^2}\\
0 & 1 & 0 & 0\\
0 & 0 & \frac{1}{\mu^2} & 0\\
\frac{-\nu}{\alpha^2} & 0 & 0 & \frac{\nu^2}{\alpha^2} + 1
\end{bmatrix}.
Then we have A \in \operatorname{Aut}\left(G_2\right)_{M_{\nu}} \Leftrightarrow A^tM_{\nu}A = M_{\nu} \Leftrightarrow A \in \left\lbrace I_4, \operatorname{diag}\left\lbrace 1, -1, 1, 1\right\rbrace \right\rbrace \cong \mathbb{Z}_2.
In fact, for any other left-invariant Riemannian metric on G_2, the group of isometric automorphisms is isomorphic to one of the groups listed in Theorem 5.5.
The isometry group of the Lie group G_2 is given by
The possible groups of isometric automorphisms of G_{3.2} \times \mathbb{R} are
The automorphism group of \operatorname{A}_{3, 2} \oplus \operatorname{A}_1 = \operatorname{Lie}(G_{3.2} \times \mathbb{R}) consists of elements of the form [christodoulakis2003automorphisms]
Let A be an automorphism of \operatorname{A}_{3, 2} \oplus \operatorname{A}_1 = \operatorname{Lie}(G_{3.2} \times \mathbb{R}) of the above form , then
This is the maximal group of isometric automorphisms of G_{3.2} \times \mathbb{R}.
For the metric M_2, it suffices to verify which of the elements of \operatorname{Aut}\left(G_{3.2} \times \mathbb{R}\right)_{M_1} preserve the metric M_2. We obtain that
Finally, one can easily verify that: A \in \operatorname{Aut}\left(G_{3.2} \times \mathbb{R}\right)_{M_3} \Leftrightarrow A^tM_3A = M_3 \Leftrightarrow A = I_4.
The isometry group of the Lie group G_{3.2} \times \mathbb{R} is given by
The possible groups of isometric automorphisms of G_{3.3} \times \mathbb{R} are
The automorphism group of \operatorname{A}_{3, 3} \oplus \operatorname{A}_1 = \operatorname{Lie}(G_{3.3} \times \mathbb{R}) consists of elements of the form [christodoulakis2003automorphisms]
Let A be an automorphism of the above form; then
Hence \operatorname{Aut}\left(G_{3.3} \times \mathbb{R}\right)_{M_1} = \left(\operatorname{diag}\left\lbrace \operatorname{O}(2), 1, \pm1\right\rbrace \right). This is the maximal group of isometric autmorphisms of G_{3.3} \times \mathbb{R}. For the metric M_2, one can see that
Thus \operatorname{Aut}\left(G_{3.3} \times \mathbb{R}\right)_{M_2} = \left(\operatorname{diag}\left\lbrace \operatorname{O}(2), 1, 1\right\rbrace \right) \cong \operatorname{O}(2).
For the metric M_3, it suffices to verify which of the elements of \operatorname{Aut}\left(G_{3.3} \times \mathbb{R}\right)_{M_1} preserve M_3. We consider an automorphism of the form A = \begin{bmatrix}
a_1 & a_2 & 0 & 0\\
a_5 & a_6 & 0 & 0\\
0 & 0 & 1 & 0\\
0 & 0 & 0 & a_{16}
\end{bmatrix}. Then
Hence
For the metric M_4, we obtain that
Thus, \operatorname{Aut}\left(G_{3.3} \times \mathbb{R}\right)_{M_4} = \left\lbrace I_4, \operatorname{diag}\left\lbrace 1, -1, 1, 1\right\rbrace \right\rbrace \cong \mathbb{Z}_2.
The isometry group of G_{3.3} \times \mathbb{R} is given by
The possible groups of isometric automorphisms of G_{c} \times \mathbb{R} are
The automorphism group of \operatorname{A}_{3, 5}^{\alpha} \oplus \operatorname{A}_1 = \operatorname{Lie}(G_{c} \times \mathbb{R}) consists of elements of the form [christodoulakis2003automorphisms]
Let A be an automorphism of the above form, then
Hence \operatorname{Aut}\left(G_{c} \times \mathbb{R}\right)_{M_1} = \operatorname{diag}\left\lbrace \pm1, \pm1, 1, \pm1\right\rbrace \cong \left(\mathbb{Z}_2\right)^3. This is the maximal group of isometric automorphisms on G_{c} \times \mathbb{R}. For the metric M_2, we have
Additionally, comparing the (4, 3)-th and (3, 2)-th components of the matrices A^tM_2A and M_2 shows that a_7 and a_{15} must satisfy the following system
This implies that a_7 = a_{15} = 0. Thus
There is a simple way to find this result. Since \operatorname{Aut}\left(G_{c} \times \mathbb{R}\right)_{M_1} \cong \left(\mathbb{Z}_2\right)^3 is the maximal group of isometric automorphisms, it suffices to verify which of its elements preserve the metric M_2. For the other two metrics, we use this simple way and we obtain that
The isometry group of G_{c} \times \mathbb{R} is given by
The possible groups of isometric automorphisms of G_{4.2}^{\alpha} are
The automorphism group of \operatorname{A}_{4, 2}^{\alpha} = \operatorname{Lie}(G_{4.2}^{\alpha}) consists of elements of the form [christodoulakis2003corrigendum]
Let A be an automorphism of the above form, then
Thus
For the metric M_2, we obtain that
If we choose the metric U where \alpha > 0 and \gamma \neq 0, its group of isometric automorphisms is still \mathbb{Z}_2 as above.
The isometry group of G_{4.2}^{\alpha} is given by
The possible groups of isometric automorphisms of G_{4.2}^{1} are
The automorphism group of \operatorname{A}_{4, 2}^{1} = \operatorname{Lie}(G_{4.2}^{1}) consists of elements of the form [christodoulakis2003corrigendum]
Let A be an automorphism of the above form, then
Thus
This is the maximal group of isometric automorphisms of G_{4.2}^{1}. For the metric M_2, we obtain that
The isometry group of G_{4.2}^{1} is given by
The group of isometric automorphisms of G_{4.4} is \operatorname{Aut}\left(G_{4.4}\right)_{M_1} \cong \mathbb{Z}_2.
The automorphism group of \operatorname{A}_{4, 4} = \operatorname{Lie}(G_{4.4}) consists of elements of the form [christodoulakis2003automorphisms]
Let A be an automorphism of the above form, then
Hence \operatorname{Aut}\left(G_{4.4}\right)_{M_1} = \left\lbrace I_4, \operatorname{diag}\left\lbrace -1, -1, -1, 1\right\rbrace\right\rbrace \cong \mathbb{Z}_2.
The group of isometries of G_{4.4} is given by: \operatorname{Isom}\left(G_{4.4}, M_1\right) \cong G_{4.4} \rtimes \mathbb{Z}_2.
The group of isometric automorphisms of G_{4.5}^{\alpha, \beta} is given by
The automorphism group of \operatorname{A}_{4, 5}^{\alpha, \beta} = \operatorname{Lie}(G_{4.5}^{\alpha, \beta}) consists of elements of the form [christodoulakis2003automorphisms]
Let A be an automorphism of the above form, then
Hence \operatorname{Aut}\left(G_{4.5}^{\alpha, \beta}\right)_{M_1} = \operatorname{diag}\left\lbrace \pm1, \pm1, \pm1, 1\right\rbrace \cong \left(\mathbb{Z}_2\right)^3.
For the metric M_2, it suffices to verify which of the elements of \operatorname{Aut}\left(G_{4.5}^{\alpha, \beta}\right)_{M_1} preserve M_2. We obtain that
Finally, one gets that
The isometry group of G_{4.5}^{\alpha, \beta} is given by
The group of isometric automorphisms of G_{4.7} is given by:
\operatorname{Aut}\left(G_{4.7}\right)_{M_1} \cong \mathbb{Z}_2, \qquad \operatorname{Aut}\left(G_{4.7}\right)_{M_2} = \left\lbrace I_4\right\rbrace.
The automorphism group of \operatorname{A}_{4, 7} = \operatorname{Lie}(G_{4.7}) consists of elements of the form [christodoulakis2003automorphisms]
Let A be an automorphism of the above form, then
Hence \operatorname{Aut}\left(G_{4.7}\right)_{M_1} = \left\lbrace I_4, \operatorname{diag}\left\lbrace 1, -1, -1, 1\right\rbrace\right\rbrace \cong \mathbb{Z}_2.
For the metric M_2, one gets that A \in \operatorname{Aut}\left(G_{4.7}\right)_{M_2} \Leftrightarrow A^tM_2A = M_2 \Leftrightarrow A = I_4.
The group of isometries of G_{4.7} is given by:
\operatorname{Isom}\left(G_{4.7}, M_1\right) \cong G_{4.7} \rtimes \mathbb{Z}_2, \quad \operatorname{Isom}\left(G_{4.7}, M_2\right) \cong G_{4.7}.
The group of isometric automorphisms of G_{4.8}^{\alpha} is given by
The automorphism group of \operatorname{A}_{4, 9}^{\beta} = \operatorname{Lie}(G_{4.8}^{\alpha}) consists of elements of the form [christodoulakis2003automorphisms]
Let A be an automorphism of the above form, then
Hence
For the metric M_2, we obtain that
Finally, one can see that A \in \operatorname{Aut}\left(G_{4.8}^{\alpha}\right)_{M_3} \Leftrightarrow A^tM_3A = M_3 \Leftrightarrow A \in \left\lbrace I_4 \right\rbrace.
The isometry group of G_{4.8}^{\alpha} is given by
The study of the isometry group of left-invariant Riemannian metrics on the following Lie groups is still an open question:
\operatorname{Sol}_{\mu}^4, \widetilde{\operatorname{Isom}_0(\mathbb{R}^2)} \times \mathbb{R} (unimodular solvable not of type (R), see [van2017metrics]).
\widetilde{\operatorname{SL}(2,\mathbb{R})} \times \mathbb{R}, \operatorname{SU}(2, \mathbb{R}) \times \mathbb{R} (unimodular non-solvable, see [van2017metrics]).
The simply connected Lie groups associated with the following Lie algebras:
\operatorname{A}_{3, 7}^{\alpha} \oplus \operatorname{A}_1, \operatorname{A}_{4, 6}^{\alpha, \beta}, \operatorname{A}_{4, 11}^{\alpha} and \operatorname{A}_{4, 12}. They are nonunimodular and not of type (R), as shown in this paper.
The classification of left-invariant Riemannian metrics on these Lie groups is given by [van2017metrics, aitbenhaddou2025classification]. The question is a bit complicated, but one can use the fact that the isotropy subgroup \operatorname{Isom}(G, g)_e can be injected into the intersection \operatorname{O}(\mathfrak{g}, g) \cap \mathcal{C}(\hat{r}), where \operatorname{O}(\mathfrak{g}, g) is the orthogonal group of \mathfrak{g} with respect to g, and \mathcal{C}(\hat{r}) is the subspace of matrices that commute with the Ricci transformation \hat{r} of (G, g). For more details, see the strategy used in the following paper [aitbenhaddou2024isometry]. The signature of the Ricci operator of left-invariant Riemannian metrics on four-dimensional Lie groups is studied in [kremlev2009signature, kremlev2010signature].
The text on this page is from arXiv:2501.17009v2 by Youssef Ayad, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.
The record of each item is in analysis/objects/2501/2501.17009.jsonl.