= Solution
Use ambient <quadratic form> $Q=-(X^0)^2-(X^4)^2+(X^1)^2+(X^2)^2$. The printed sine coordinate has a duplicated superscript: it must be $X^2$, not another $X^1$. With that correction the embedding lies on $Q=-1$. Put $s=\sqrt{1+r^2}$. Differentiating the two timelike coordinates gives
$$
(dX^0)^2+(dX^4)^2=\frac{r^2}{1+r^2}\,dr^2+(1+r^2)\,dt^2,
$$
whereas the two spatial coordinates give $(dX^1)^2+(dX^2)^2=dr^2+r^2d\theta^2$. Their difference is the induced <Lorentzian metric>
$$
g=-(1+r^2)\,dt^2+\frac{dr^2}{1+r^2}+r^2d\theta^2.
$$
The polar coordinate $\theta$ collapses at $r=0$; this is a coordinate degeneracy, not a degeneracy of the metric.
For the <matrix model of anti-de Sitter three-space>, define
$$
M(X)=\begin{pmatrix}X^4+X^1&X^2+X^0\\X^2-X^0&X^4-X^1\end{pmatrix}.
$$
Then $\det M=-Q$. This linear identification of ambient space with real two-by-two matrices identifies $Q=-1$ with $SL(2,\mathbb R)$. Its inverse recovers each $X$ by sums and differences of matrix entries, so the identification is global. The identity corresponds to $X^4=1$ and the other coordinates zero. For a traceless tangent matrix $U$, the induced squared norm is
$$
g_I(U,U)=-\det U=\tfrac12\operatorname{tr}(U^2).
$$
Polarization gives $g_I(U,V)=\tfrac12\operatorname{tr}(UV)$.
The <Killing form> of $\mathfrak{sl}_2(\mathbb R)$ is $B(U,V)=4\operatorname{tr}(UV)$. This normalization can be checked without assuming it: for $H=\operatorname{diag}(1,-1)$, $E=\left(\begin{smallmatrix}0&1\\0&0\end{smallmatrix}\right)$ and $F=\left(\begin{smallmatrix}0&0\\1&0\end{smallmatrix}\right)$, the brackets are $[H,E]=2E$, $[H,F]=-2F$, $[E,F]=H$. Taking traces of their three-dimensional adjoint matrices gives $B(H,H)=8$, $B(E,F)=B(F,E)=4$, and all other basis pairings zero. These agree with $4\operatorname{tr}(UV)$.
Both left and right multiplication by matrices in $SL(2,\mathbb R)$ preserve $-\det M$, hence its polarized ambient <bilinear form> and the induced metric. Equality at the identity therefore propagates throughout the group:
$$
\boxed{g=\tfrac18 B.}
$$
In particular the <Killing-form Einstein metric> calculation gives $\operatorname{Ric}=-2g$, consistent with unit-radius three-dimensional <Anti-de Sitter spacetime>.
Every element of $SO(2,2)$ preserves the ambient <quadratic form>, hence the hyperboloid and its induced metric, and so acts by <isometries>. Define the <left-right double cover of SO0(2,2)> by
$$
\Phi(A,B)(M)=AMB^{-1},\qquad A,B\in SL(2,\mathbb R).
$$
This is a <group homomorphism>, since successive actions multiply the two factors in the same order. It preserves $\det M$; as a transformation of the four-dimensional matrix space it has determinant $(\det A)^2(\det B^{-1})^2=1$. Thus it takes values in $SO(2,2)$.
If its action is the identity, applying it to $M=I$ gives $A=B$. Commuting with every matrix then forces $A$ to be scalar. The condition $\det A=1$ leaves only $A=I$ or $A=-I$. Hence
$$
\ker\Phi=\{(I,I),(-I,-I)\}.
$$
The differential is $(a,b):M\mapsto aM-Mb$. A zero differential implies $a=b$ and that $a$ is scalar; tracelessness makes $a=b=0$. It is therefore an isomorphism onto the target <Lie algebra>, since both dimensions are six. The image is an open subgroup. Since $SL(2,\mathbb R)$ is connected, the image is connected and is exactly the <identity component> $SO_0(2,2)$: an open subgroup of a connected group is also closed and must be the whole group. Thus the precise covering statement is
$$
\boxed{SO_0(2,2)\cong\big(SL(2,\mathbb R)\times SL(2,\mathbb R)\big)/\{\pm(I,I)\}.}
$$
It is not onto the disconnected full $SO(2,2)$. For example, conjugation $M\mapsto DMD^{-1}$, $D=\operatorname{diag}(1,-1)$, sends $(X^0,X^4,X^1,X^2)$ to $(-X^0,X^4,X^1,-X^2)$ and belongs to the other component of $SO(2,2)$. It reverses the orientation of the negative two-plane and is not in the image of the connected source.
The <Adjoint representation of a Lie group> of $SL(2,\mathbb R)$ preserves its <Killing form> of <metric signature> $(2,1)$ and gives a second double cover,
$$
\boxed{SL(2,\mathbb R)/\{\pm I\}\cong SO_0(2,1).}
$$
Its kernel is the scalar centre; its differential is injective and both dimensions are three, so the same connected-image argument proves surjectivity onto the identity component. The full $SO(2,1)$ again has an additional component. Thus $SO_0(2,1)$ is locally the same <Lie group> geometry as $SL(2,\mathbb R)$, with its <Killing metric>, and globally is its central quotient.
With the printed periodic time, the group model has topology $S^1\times\mathbb R^2$ and closed timelike curves, since varying $t$ at fixed spatial position gives a timelike circle. The causally unwrapped version of <Anti-de Sitter spacetime> instead has $t\in\mathbb R$ and is the <universal covering Lie group>. It must not be substituted silently for the periodic spacetime specified here.
Back to article page