signature=a5b30effd242f7dc2fdb380972cecc58,Three-dimensional conformally symmetric manifolds

We start with some previous lemmas, firstly considering the case when the manifold \((M,g)\) decomposes as a product.

Lemma 2

A three-dimensional pseudo-Riemannian product manifold \((M=\mathbb R \times N,g={\pm } \mathrm{d}t^2+g_N)\) is never essentially conformally symmetric.

Proof

Let \((a,x)\) and \((b,y)\) be vector fields on \((M,g)\). Hence, the Schouten tensor \(S\) satisfies

$$\begin{aligned} S((a,x),(b,y))&= \rho ((a,x),(b,y))-\frac{\tau }{4}g((a,x),(b,y))\\&= \frac{\tau }{2}g(x,y)-\frac{\tau }{4}({\pm }ab+g(x,y))\\&= \frac{\tau }{4}({\mp }ab+g(x,y))\\&= \frac{\tau }{4}({\mp }\mathrm{d}t^2+g_N)((a,x),(b,y)). \end{aligned}$$

Next, observe that \(g\) and \(h={\mp }\mathrm{d}t^2+g_N\) are metrics on \(M\) sharing the same Levi-Civita connection, and hence, a straightforward calculation shows that \(\nabla S =\frac{1}{4}\mathrm{d}\tau \otimes h\). As a consequence, the \((0,3)\)-Cotton tensor is given by

$$\begin{aligned} \mathrm{C }_{\alpha \beta \gamma }=(\nabla _\alpha S)_{\beta \gamma }-(\nabla _\beta S)_{\alpha \gamma }=\frac{1}{4}\left( \mathrm{d}\tau (\alpha )h_{\beta \gamma } -\mathrm{d}\tau (\beta )h_{\alpha \gamma }\right) , \end{aligned}$$

from where we get

$$\begin{aligned} \nabla _\mu \mathrm{C }_{\alpha \beta \gamma } =\dfrac{1}{4}\left( \left( \mathrm{Hes }_\tau \right) _{\mu \alpha }h_{\beta \gamma } -\left( \mathrm{Hes }_\tau \right) _{\mu \beta }h_{\alpha \gamma } \right) , \end{aligned}$$

where \(\mathrm{Hes }_\tau \) denotes the Hessian of the scalar curvature, \(\mathrm{Hes }_\tau (X,Y)=X(Y \tau )-(\nabla _XY)\tau \). Then, \(\nabla _i\mathrm{C }_{tjt}=\frac{1}{4}\left( \mathrm{Hes }_\tau \right) _{ij}\), and therefore, the product manifold is conformally symmetric if and only if \(\mathrm{Hes }_\tau \) vanishes, i.e., \(\nabla \tau \) is a parallel vector field on \(N\). Finally, if \(\nabla \tau \ne 0\), then \(N\) splits locally as \(N\equiv \mathbb R \times \mathbb R \), while if \(\nabla \tau = 0\) then \(N\) has constant curvature. In any case, one has that \(\nabla \mathrm{C }=0\) implies \(\mathrm{C }=0\) thus showing that \((M,g)\) is locally conformally flat, which finishes the proof.\(\square \)

Next, lemma shows that any essentially conformally symmetric three-dimensional manifold is locally indecomposable, but not irreducible, i.e., it admits a parallel degenerate line field. These manifolds have been extensively investigated in the literature and usually referred to as Brinkmann waves or Walker manifolds (see for example [4] and the references therein).

Lemma 3

Any three-dimensional essentially conformally symmetric manifold \((M,g)\) is a Walker manifold.

Proof

In what follows, we explicitly use that \(\mathrm{dim }\, M=3\) to associate a \((0,2)\)-tensor field to the usual \((0,3)\)-Cotton tensor. Let \(\star :\Lambda ^p(M)\rightarrow \Lambda ^{3-p}(M)\) denote the Hodge \(\star \)-operator and consider the Cotton \(2\)-form \(C_i= \frac{1}{2}\mathrm{C }_{nmi}\mathrm{d}x^n\wedge \mathrm{d}x^m\). Using the Hodge \(\star \)-operator, the \((0,2)\)-Cotton tensor is associated to the Cotton \(2\)-form by \(\star C_i=\frac{1}{2}\mathrm{C }_{nmi}\epsilon ^{nm\ell }\mathrm{d}x^\ell \), thus resulting

$$\begin{aligned} \tilde{\mathrm{C }}_{ij}=\frac{1}{2\sqrt{g}}\mathrm{C }_{nmi}\epsilon ^{nm\ell }g_{\ell j}, \end{aligned}$$

where \(\epsilon ^{123}=1\). Moreover, associated to the \((0,2)\)-Cotton tensor, the Cotton operator is defined by \(\tilde{\mathrm{C }}(x,y)=g(\hat{\mathrm{C }}(x),y)\). Since the metric tensor and the Hodge \(\star \)-operator are parallel, the conformal symmetry of any three-dimensional manifold can be equivalently stated in terms of the \((0,2)\)-Cotton tensor \(\tilde{\mathrm{C }}\), or in terms of the Cotton operator \(\hat{\mathrm{C }}\).

Next, we consider three-dimensional manifolds with parallel Cotton operator and analyze the different possibilities for the Jordan normal form of \(\hat{\mathrm{C }}\). Assume that the Cotton operator diagonalizes. If the Cotton operator is parallel, then the eigenvalues of \(\hat{\mathrm{C }}\) are constant, and the corresponding eigenspaces define parallel distributions on \(M\). Since the Cotton tensor is traceless, it has at least two different eigenvalues, unless the manifold is locally conformally flat. In any case, \(\hat{\mathrm{C }}\) always has a distinguished eigenvalue of multiplicity one and thus \((M,g)\) admits locally a de Rham decomposition as a product \((\mathbb R \times N, \pm \mathrm{d}t^2+g_N)\), where \(N\) is a surface. Then Lemma 2 shows that \((M,g)\) is locally conformally flat.

Assume that the Cotton operator has a complex eigenvalue. Then, with respect to an orthonormal local frame \(\{e_1,e_2,e_3\}\) of signature \((++-)\), one has

$$\begin{aligned} \hat{\mathrm{C }}=\left( \begin{array}{c@{\quad }c@{\quad }c} \lambda &{} 0 &{} 0\\ 0 &{} \alpha &{} \beta \\ 0 &{} -\beta &{} \alpha \end{array} \right) . \end{aligned}$$

Since the Cotton operator is parallel, the distribution defined by the eigenspace corresponding to the eigenvalue \(\lambda \) is parallel. Such distribution is spacelike, and hence, the manifold decomposes locally as a product \(\mathbb R \times N\), where \(N\) is a surface. Then, Lemma 2 implies that \((M,g)\) is locally conformally flat.

If the minimal polynomial of the Cotton operator \(\hat{\mathrm{C }}\) has a root of multiplicity two, then there exists a local frame \(\{e_1,e_2,e_3\}\), with \(g(e_1,e_1)=g(e_2,e_3)=1\) such that \(\hat{\mathrm{C }}\) expresses with respect to that frame as

$$\begin{aligned} \hat{\mathrm{C }}=\left( \begin{array}{c@{\quad }c@{\quad }c} \lambda &{} 0 &{} 0\\ 0 &{} \alpha &{} 1\\ 0 &{} 0 &{} \alpha \end{array} \right) . \end{aligned}$$

Suppose first that \(\lambda \ne 0\); in this case, since \(\hat{\mathrm{C }}\) is parallel, the spacelike distribution defined by the eigenspace corresponding to the eigenvalue \(\lambda \) is also parallel. Then, the manifold decomposes locally as a product, and Lemma 2 shows that \((M,g)\) is locally conformally flat. Next, assume \(\lambda =0\); since the Cotton operator is trace-free, it must be \(2\)-step nilpotent (i.e., \(\hat{\mathrm{C }}^2=0\), \(\hat{\mathrm{C }}\ne 0\)). Then, \(\mathrm{Im }(\hat{\mathrm{C }})=\langle \hat{\mathrm{C }}(e_3)\rangle =\langle e_2\rangle \). Moreover, taking into account that the Cotton operator is parallel, we have

$$\begin{aligned} 0=(\nabla _X \hat{\mathrm{C }})(e_3)=\nabla _X (\hat{\mathrm{C }}(e_3))-\hat{\mathrm{C }}(\nabla _X e_3), \end{aligned}$$

from where \(\hat{\mathrm{C }}(\nabla _X e_3)=\nabla _X (\hat{\mathrm{C }}(e_3))\), and hence, \(\mathrm{Im }(\hat{\mathrm{C }})\) is a null and parallel one-dimensional distribution on \((M,g)\), from where one concludes that \(g\) is a Walker metric.

Finally, consider the case when the minimal polynomial of the Cotton operator has a root of multiplicity \(3\). Since the Cotton operator is trace-free, it must be \(3\)-step nilpotent. Proceeding as before and using the fact that the Cotton operator is parallel one easily shows that \(\mathrm{Ker }(\hat{\mathrm{C }})\) is a null and parallel distribution, and thus, \((M,g)\) is a Walker manifold.\(\square \)

Remark 4

In the Riemannian case, the Cotton operator diagonalizes, and thus, the nonexistence of three-dimensional essentially conformally symmetric Riemannian manifolds follows from the proof of Lemma 3.

Proof of Theorem 1

Let \((M,g)\) be a three-dimensional essentially conformally symmetric Lorentzian manifold. It follows from Lemma 3 that \((M,g)\) is indecomposable but not irreducible, and hence a Walker manifold.

Three-dimensional Walker manifolds admit local coordinates \((t,x,y)\) where the metric expresses as (see [4] and the references therein)

$$\begin{aligned} g= \mathrm{d}t \mathrm{d}y + \mathrm{d}x^2 + f(t,x,y) \mathrm{d}y^2, \end{aligned}$$

(2)

for some smooth function \(f(t,x,y)\). In the special case when the parallel degenerate line field is spanned by a parallel null vector field, the coordinates above can be further specialize so that the metric takes the form (2) for some function \(f(x,y)\).

Then, the Levi-Civita connection is determined (up to the usual symmetries) by

$$\begin{aligned} \nabla _{\partial _t}\partial _y=\tfrac{1}{2}f_t\partial _t,\quad \nabla _{\partial _x}\partial _y=\tfrac{1}{2}f_x\partial _t,\quad \nabla _{\partial _y}\partial _y=\tfrac{1}{2} \left( f_y+f f_t\right) \partial _t-\tfrac{1}{2}f_x \partial _x -\tfrac{1}{2}f_t \partial _y, \end{aligned}$$

(3)

and the Ricci tensor is given by

$$\begin{aligned} \rho (\partial _t,\partial _y)=\tfrac{1}{2}f_{tt},\quad \rho (\partial _x,\partial _y)=\tfrac{1}{2}f_{tx}, \quad \rho (\partial _y,\partial _y)=\tfrac{1}{2}\left( f f_{tt}-f_{xx}\right) . \end{aligned}$$

(4)

Moreover, the \((0,2)\)-Cotton tensor of a Walker metric (2) is characterized by

$$\begin{aligned} \tilde{\mathrm{C }}(\partial _t,\partial _x)&= -\frac{1}{4} f_{ttt},\quad \tilde{\mathrm{C }}(\partial _t,\partial _y)=\frac{1}{4} f_{ttx}, \quad \tilde{\mathrm{C }}(\partial _x,\partial _x)=-\frac{1}{2} f_{ttx},\nonumber \\ \tilde{\mathrm{C }}(\partial _x,\partial _y)&= \frac{1}{4} \left( 2 f_{txx}+f_{tty}-f f_{ttt}\right) ,\\ \tilde{\mathrm{C }}(\partial _y,\partial _y)&= \frac{1}{4}\left( f_x f_{tt}-2 f_{xxx}-f_t f_{tx}-2 f_{txy}+2 f f_{ttx}\right) .\nonumber \end{aligned}$$

(5)

A long but straightforward calculation shows that a Walker metric (2) has parallel Cotton tensor if and only if

$$\begin{aligned} \left\{ \begin{array}{l} f_{tttt}=f_{tttx}=f_{ttxx}=f_{txxx}=0,\\ f_t f_{ttt}-2 f_{ttty}=0,\\ 2 f_{ttxy}-f_x f_{ttt}=0,\\ 4 f_{txxy}+ \left( 2 f_{txx}+ f_{tty}\right) f_t+2 f_{ttyy}-3 f_xf_{ttx}-f_y f_{ttt}-2 f f_{ttty}=0,\\ (f_{tx})^2+2 f_{xxxx}+f_t f_{txx}+2 f_{txxy}-f_{xx} f_{tt}-2 f_x f_{ttx} =0,\\ f_{tx} \left( f_t\right) ^2+\left( 2 f_{xxx}+3 f_{txy}\right) f_t+2 f_{xxxy}+f_{ty} f_{tx}\,\\ + 2 f_{txyy}-f_{xy} f_{tt}- \left( 2 f_{txx}+f_t f_{tt}+2 f_{tty}\right) f_x-\left( f_y+f f_t\right) f_{ttx}=0. \end{array} \right. \end{aligned}$$

(6)

From the first equation in (6), we get

$$\begin{aligned} f(t,x,y)=\alpha (x,y)+t \left( \beta (y)+x\, \xi (y)+x^2 \delta (y)\right) +t^2 (\mu (y)+x \,\phi (y))+t^3 \gamma (y). \end{aligned}$$

Now, differentiating the second equation in (6) twice with respect to \(t\), it follows that \(\gamma (y)=0\). Then, the third equation in (6) transforms into \(\phi ^{\prime }(y)=0\) and therefore \(\phi (y)=K\). We differentiate now the fourth equation in (6) twice with respect to \(t\) to obtain \(K=0\). Moreover, differentiating again the fourth equation in (6) twice with respect to \(x\), we get

$$\begin{aligned} \delta (y) \left( 2 \delta (y)+\mu ^{\prime }(y)\right) =0, \end{aligned}$$

(7)

and differentiating once again the fourth equation in (6), in this case with respect to \(t\), we obtain

$$\begin{aligned} \mu (y) \left( 2 \delta (y)+\mu ^{\prime }(y)\right) =0. \end{aligned}$$

(8)

Hence, Eqs. (7) and (8) imply that \(\delta (y)=-\frac{1}{2}\mu ^{\prime }(y)\). At this point, the only nonzero component of the Cotton tensor is given by

$$\begin{aligned} \tilde{\mathrm{C }}(\partial _y,\partial _y)&= \frac{1}{8} \{4 \mu (y) \alpha _x(x,y) - 4 \alpha _{xxx}(x,y) - x^3 \mu ^{\prime }(y)^2 + 3 x^2 \xi (y) \mu ^{\prime }(y)\\&- 2 \beta (y) \left( \xi (y) -x \mu ^{\prime }(y)\right) -2 x \xi (y)^2+4 x \mu ^{\prime \prime }(y)-4 \xi ^{\prime }(y) \}. \end{aligned}$$

Now, differentiating the last equation in (6) with respect to \(t\), a straightforward calculation shows that

$$\begin{aligned} \mu (y) \tilde{\mathrm{C }}(\partial _y,\partial _y)=0. \end{aligned}$$

If \(\tilde{\mathrm{C }}(\partial _y,\partial _y)=0\) at some point, then the Cotton tensor vanishes everywhere since it is parallel, and therefore, the manifold is locally conformally flat. Thus, we may assume that \(\mu (y)=0\). Now, the fifth equation in (6) transforms into

$$\begin{aligned} 2\alpha _{xxxx}(x,y)+\xi (y)=0, \end{aligned}$$

and therefore \(\alpha (x,y)=-\frac{1}{48}x^4\xi (y)^2+\mathcal D (y)x^3+\mathcal C (y)x^2 +\mathcal B (y)x+\mathcal A (y)\). Now, a long but straightforward calculation shows that the unique nonzero component of the \((0,2)\)-Cotton tensor is given by

$$\begin{aligned} \tilde{\mathrm{C }}(\partial _y,\partial _y)= -\tfrac{1}{4} \left( 12 \mathcal D (y)+\beta (y) \xi (y)+2 \xi ^{\prime }(y)\right) , \end{aligned}$$

and differentiating the last equation in (6) with respect to \(x\), we get

$$\begin{aligned} 8 \xi (y) \tilde{\mathrm{C }}(\partial _y,\partial _y)=0. \end{aligned}$$

As before, if \(\tilde{\mathrm{C }}(\partial _y,\partial _y)\) does not vanish identically, then one has \(\xi (y)=0\). Thus, the Cotton tensor is determined by \(\tilde{\mathrm{C }}(\partial _y,\partial _y)=-3\mathcal D (y)\), which implies that \(\mathcal D (y)\ne 0\) everywhere unless \((M,g)\) is locally conformally flat; moreover, the last equation in (6) reduces to \(\mathcal{D }(y)\beta (y)+\mathcal{D }^{\prime }(y)=0\), from where \(\beta (y)=-\frac{\mathcal{D }^{\prime }(y)}{\mathcal{D }(y)}\).

At this point, the metric (2) is determined by

$$\begin{aligned} f(t,x,y)=-\frac{\mathcal{D }^{\prime }(y)}{\mathcal{D }(y)} t+\mathcal{D }(y)x^3+\mathcal{C }(y)x^2+\mathcal{B }(y) x+\mathcal{A }(y). \end{aligned}$$

A straightforward calculation shows that the Ricci operator of this metric is \(2\)-step nilpotent. A three-dimensional Walker manifold with \(2\)-step nilpotent Ricci operator admits a null and parallel vector field [4], and therefore, the Walker coordinates \((t,x,y)\) can be specialized so that the metric expresses as

$$\begin{aligned} g_f=\mathrm{d}t\mathrm{d}y+\mathrm{d}x^2+f(x,y)\mathrm{d}y^2. \end{aligned}$$

(9)

Now, the only nonzero component of the Cotton tensor is

$$\begin{aligned} \tilde{\mathrm{C }}(\partial _y,\partial _y)=-\tfrac{1}{2}f_{xxx}, \end{aligned}$$

and the nonvanishing components of \(\nabla \tilde{\mathrm{C }}\) are given by

$$\begin{aligned} (\nabla _{\partial _x} \tilde{\mathrm{C }})(\partial _y,\partial _y)=-\tfrac{1}{2}f_{xxxx}, \qquad (\nabla _{\partial _y} \tilde{\mathrm{C }})(\partial _y,\partial _y)=-\tfrac{1}{2}f_{xxxy}. \end{aligned}$$

A direct calculation shows that a strict Walker metric is essentially conformally symmetric if and only if

$$\begin{aligned} f(x,y)=\kappa x^3+x^2 \mathcal A (y)+\mathcal B (y) x+\mathcal C (y), \end{aligned}$$

(10)

for arbitrary smooth functions \(\mathcal A \), \(\mathcal B \) and \(\mathcal C \) and a nonzero real constant \(\kappa \).

In what remains of the proof, we show that any metric (10) is locally isometric to some metric (1) for a suitable function \(\mathfrak a (y)\). We proceed as in [12]. Let \(g_f\) be a Walker metric defined by (9) and consider the application:

$$\begin{aligned} T(t,x,y)=(t-\phi _y x+ \psi ,x+\phi ,y), \end{aligned}$$

where \(\phi \) and \(\psi \) are smooth functions on \(y\). Then, \(T\) defines an isometry between \(g_f\) and another Walker metric \(g_{\tilde{f}}\) given by (9) for some function

$$\begin{aligned} \tilde{f}(x,y)=f(x+\phi ,y)-2x \phi _{yy}+\phi _y^2+2\psi _y. \end{aligned}$$

Now, consider a Walker metric \(g_\mathfrak{b ,\kappa }=\mathrm{d}t \mathrm{d}y +\mathrm{d}x^2+(\kappa x^3+\mathfrak b (y)x) \mathrm{d}y^2\) defined by some arbitrary smooth function \(\mathfrak b (y)\) and some nonzero constant \(\kappa \). Then, \(T\) defines an isometry between \(g_\mathfrak{b ,\kappa }\) and a Walker metric \(g_{\tilde{f}}\) where

$$\begin{aligned} \tilde{f}(x,y)&= \kappa (x+\phi )^3+\mathfrak b (y)(x+\phi )-2x \phi _{yy}+\phi _y^2+2\psi _y\\&= \kappa x^3+3\kappa \phi x^2 +(\mathfrak b (y)+3 k \phi ^2-2\phi _{yy})x+\mathfrak b (y)\phi +\kappa \phi ^2+\phi _y^2+2\psi _y. \end{aligned}$$

Setting \(\mathcal A =3\kappa \phi \), \(\mathcal B \) defined by \(\mathcal B =\mathfrak b +3 k \phi ^2-2\phi _{yy}\) and choosing \(\psi \) so that \(\mathcal C =\mathfrak b (y)\phi +\kappa \phi ^2+\phi _y^2+2\psi _y\), one has that \(T\) defines an isometry between \(g_\mathfrak{b ,\kappa }\) and a Walker metric \(g_f\) with \(f(x,y)\) given by Eq. (10).

Finally, observe that \(\kappa \) is not relevant in the previous discussion since

$$\begin{aligned} \tilde{T}(t,x,y)=\left( \sqrt{\varepsilon \kappa }t,\varepsilon x,\dfrac{1}{\sqrt{\varepsilon \kappa }}y\right) \end{aligned}$$

is an isometry between \(g_\mathfrak{b ,\kappa }\) and \(g_\mathfrak a =\mathrm{d}t \mathrm{d}y +\mathrm{d}x^2+(x^3+\mathfrak a (y)x)\mathrm{d}y^2\), where the function \(\mathfrak a (y)\) is given by \(\mathfrak a (y)=(|\kappa |)^{-1}\mathfrak b (|\kappa |^{-1/2}y)\) and \(\varepsilon =\mathrm{Sign }(\kappa )\). This concludes the proof.\(\square \)

Remark 5

Any essentially conformally symmetric metric \(g_\mathfrak{a }\) given by (1) is defined by a function \(\mathfrak a (y)\). Hence, it is natural to consider whether two different functions \(\mathfrak a (y)\) and \(\mathfrak b (y)\) determine the same isometry class. In answering this question, first of all observe that the kernel and the image of the Ricci operator \(\hat{\rho }\) (defined by \(\rho (X,Y)=g(\hat{\rho }X,Y)\)) of any metric given by (1) are generated by

$$\begin{aligned} \mathrm{Ker }\hat{\rho }=\langle \{ \partial _t,\partial _x\}\rangle ,\qquad \mathrm{Im }\hat{\rho }=\langle \{\partial _t\}\rangle . \end{aligned}$$

Therefore, any (local) isometry must preserve these subspaces.

Let \(\Phi =({}^1\Phi (t,x,y),{}^2\Phi (t,x,y),{}^3\Phi (t,x,y))\) be an isometry between \(g_\mathfrak a \) and \(g_\mathfrak b \), i.e., \(\Phi ^\star g_\mathfrak b =g_\mathfrak a \). Since \(\Phi \) has to preserve \(\mathrm{Ker }\hat{\rho }\) and \(\mathrm{Im }\hat{\rho }\), it follows that \({}^3\Phi \) depends only on the coordinate \(y\) and \({}^2\Phi \) is a function of the coordinates \(x\) and \(y\). Moreover,

$$\begin{aligned} 1=g_\mathfrak a (\partial _x,\partial _x)=g_\mathfrak b (\Phi _\star \partial _x, \Phi _\star \partial _x)= \left( {}^2\Phi _x\right) ^2\,, \end{aligned}$$

and thus, \({}^2\Phi (x,y)=\varepsilon _1 x+ \varphi (y)\) with \(\varepsilon _1^2=1\). Furthermore, any isometry has to preserve the Ricci tensor (i.e., \(\Phi ^\star \rho _\mathfrak b =\rho _\mathfrak a \), where \(\rho _\mathfrak{a }\) and \(\rho _\mathfrak{b }\) are the Ricci tensors of \(g_\mathfrak a \) and \(g_\mathfrak b \), respectively). Hence, at any point \(p=(t,x,y)\)

$$\begin{aligned} -3x=\rho _\mathfrak{a }(\partial _y,\partial _y)_{|_p} =\rho _\mathfrak{b }(\Phi _\star \partial _y,\Phi _\star \partial _y)_{|_{\Phi (p)}} =-3(\varepsilon _1 x+\varphi (y))\left( {}^3\Phi _y\right) ^2, \end{aligned}$$

from where it follows that \(\varphi (y)=0\), \({}^3\Phi (y)=\varepsilon _2 y+\alpha \) and \(\varepsilon _1=1\) with \(\varepsilon _2^2=1\) and \(\alpha \) an arbitrary constant. Now,

$$\begin{aligned} 0=g_\mathfrak a (\partial _x,\partial _y)=g_\mathfrak b (\Phi _\star \partial _x, \Phi _\star \partial _y)=\varepsilon _2\, {}^1\Phi _x, \end{aligned}$$

from where we obtain that \({}^1\Phi \) depends only on the coordinates \(t\) and \(y\), i.e., \({}^1\Phi (t,x,y) =\Psi (t,y)\). In addition,

$$\begin{aligned} 1=g_\mathfrak a (\partial _t,\partial _y)=g_\mathfrak b (\Phi _\star \partial _t, \Phi _\star \partial _y)=\varepsilon _2\, \Psi _t. \end{aligned}$$

Then, \(\Psi (t,y)=\varepsilon _2 t+\Upsilon (y)\). So, for any point \(p\) we have

$$\begin{aligned} x^3+\mathfrak a (y)x=g_\mathfrak a (\partial _y,\partial _y)_{|_p} =g_\mathfrak b (\Phi _\star \partial _y,\Phi _\star \partial _y)_{|_{\Phi (p)}} =x^3+\mathfrak b (\varepsilon _2y+\alpha )x+2\varepsilon _2\Upsilon _y. \end{aligned}$$

Hence, \(\Upsilon (y)=\beta \). Finally, \(\Phi \) is an isometry from \(g_\mathfrak a \) to \(g_\mathfrak b \) if and only if

$$\begin{aligned} \Phi =(\varepsilon _2\,t+\beta ,x,\varepsilon _2 y+\alpha ), \, \text{ and }\quad \mathfrak a (y)=\mathfrak b (\varepsilon _2 y +\alpha ), \quad \varepsilon _2^2=1. \end{aligned}$$

This shows that the moduli space of isometry classes of essentially conformally symmetric three-dimensional manifolds coincides with the space of smooth functions of one-variable

\(\mathfrak a (y)\), up to constant speed parametrization.

Remark 6

Any essentially conformally symmetric Lorentzian manifold of dimension \(n\ge 4\) has recurrent Ricci curvature [9]. This behavior also holds in dimension \(n=3\) since it can be easily shown that any Walker metric (1) (indeed, any Walker metric given by (9)) has recurrent Ricci curvature.

Clearly, any three-dimensional \(2\)-symmetric manifold (i.e., \(\nabla ^2 R=0\) but \(\nabla R\ne 0\)) has parallel Cotton tensor. From [1, 2], it is easy to show that any three-dimensional \(2\)-symmetric manifold is locally conformally flat. Therefore, a three-dimensional essentially conformally symmetric manifold cannot be \(2\)-symmetric.

It follows from the work in [12] that essentially conformally symmetric manifolds of dimension three are not locally homogeneous (even more, they cannot be 1-curvature homogeneous).