Let $(M,g)$ be an Einstein manifold with Levi-Civita connection $\nabla$ and whose Ricci tensor $\text{Rc}(g)=g$, in components $R_{ij}=g_{ij}$. The Lichnerowicz Laplacian of $g$ is the map \begin{align*} \Delta_L:\Gamma(S^2M)&\rightarrow\Gamma(S^2M)\\ h&\mapsto(\Delta_Lh)_{ij}=g^{st}\nabla_s\nabla_th_{ij}-g^{st}R_{is}h_{jt}-g^{st}T_{is}h_{it}-2g^{pq}g^{qt}R_{isjt}h_{pq}. \end{align*} The Bianchi operator is the map \begin{align*} \text{Bian}(g):\Gamma(S^2M)&\rightarrow\Gamma(T^*)\\ h&\mapsto\text{Bian}(g,h)=\delta_gg+\frac{1}{2}\text{d}(\text{tr}_gh) \end{align*} where the divergence is the map \begin{align*} \delta_g:\Gamma(S^2M)&\rightarrow\Gamma(T^*M)\\ h&\mapsto(\delta_gh)_k=-g^{st}\nabla_sh_{tk}. \end{align*} The formal $L^2$-adjoint of the divergence is the map \begin{align*} \delta^*_g:\Gamma(T^*M)&\rightarrow\Gamma(S^2M)\\ \omega&\mapsto(\delta^*_g\omega)_{ij}=\frac{1}{2}(\nabla_i\omega_j+\nabla_j\omega_i). \end{align*}

Question: Does it hold that \begin{align*} \Delta_L(\text{im}\,\delta^*_g)\subset\text{im}\,\delta^*_g\qquad\text{and}\qquad\Delta_L\big(\text{ker}\,\text{Bian}(g)\big)\subset\text{ker}\,\text{Bian}(g)? \end{align*} If these set inclusions do hold, how does one prove them?


Since you've done the rest of the work in your answer, I'll show that $DRc_g = $ id on im $\delta_g^*$. It ends up being just a consequence of the invariance of $Rc$ under the action of the diffeomorphism group on the metrics.

For $X$ a vector field and $\varphi_t$ the 1-parameter group of diffeomorphisms generated by $X$, this invariance gives $$ Rc(\varphi_t^*g) = \varphi_t^* Rc(g). $$ Differentiating this equation, we find, since $Rc(g) = g$, $$ DRc_g(\mathcal{L}_{X}g) = \mathcal{L}_X(Rc(g)) = \mathcal{L}_Xg. $$ But the Lie derivative of the metric is given by $$ (\mathcal{L}_Xg)_{ij} = \frac12(g_{kj}\partial_iX^k + g_{ki}\partial_jX^k). $$ So for $\alpha$ a 1-form and $\alpha^{\#}$ it's associated dual vector field, we get $$ (\mathcal{L}_{\alpha^{\#}}g)_{ij} = \frac12(\nabla_i\alpha_j + \nabla_j\alpha_i) = (\delta_g^*\alpha)_{ij}. $$ It follows that $$ DRc_g(\delta_g^*\alpha) = DRc_g(\mathcal{L}_{\alpha^{\#}}g) = \mathcal{L}_{\alpha^{\#}}g = \delta_g^*\alpha. $$


First inclusion: The first inclusion $\Delta_L(\text{im}\,\delta^*_g)\subset\text{im}\,\delta^*_g$ doesn't rely on $g$ being Einstein. It relies on the fact that the linearisation $D\text{Rc}(g):\Gamma(S^2M)\rightarrow\Gamma(S^2M)$ is the identity on the image $\text{im}\,\delta^*_g$, that is, for $\delta^*_g\omega\in\text{im}\,\delta_g^*$ we have \begin{align*} D\text{Rc}(g)(\delta^*_g\omega)=\frac{1}{2}\Delta_L(\delta^*_g\omega)+\delta^*_g\big(\text{Bian}(g,\delta^*_g\omega)\big)=\delta_g^*\omega \end{align*} so we get \begin{align*} \frac{1}{2}\Delta_L(\delta^*_g\omega)&=\delta^*_g\big(\omega-\text{Bian}(g,\delta^*_g\omega)\big)\\ &=\delta_g^*\alpha \end{align*} where $\alpha=\omega-\text{Bian}(g,\delta^*_g\omega)\in\Gamma(T^*M)$. So my new question is:

Question: How does one show that $D\text{Rc}(g)=\text{id}$ on the image $\text{im}\,\delta_g^*$?

Second inclusion: The second inclusion $\Delta_L\big(\text{ker}\,\text{Bian}(g)\big)\subset\text{ker}\,\text{Bian}(g)$ makes use of the Einstein condition. The linearisation of the map $g\mapsto B(g):=\text{Bian}\big(g,\text{Rc}(g)\big)=0$ in the direction of $h\in\text{Ker}\,\text{Bian}(g)$ yields \begin{align*} 0=DB(g)h&=\partial_1\text{Bian}\big(g,\text{Rc}(g)\big)h+\text{Bian}\big(g,D\text{Rc}(g)h\big)\\ &=\partial_1\text{Bian}\big(g,\text{Rc}(g)\big)h+\frac{1}{2}\text{Bian}(g,\Delta_Lh). \end{align*} The first term is the linearisation of the map $g\mapsto\text{Bian}(g,R)$ in the direction of $h$ where we set $R=\text{Rc}(g)$. For this we calculate that \begin{align*} \big(D\text{Bian}(g,R)h\big)_m&=g^{pq}R_{pm}\text{Bian}(g,h)_q-\frac{1}{2}g^{pq}g^{st}(\nabla_pR_{sm}+\nabla_sR_{pm}-\nabla_mR_{ps})h_{qt}\\ &=0. \end{align*} The first term is $0$ since $h\in\text{Ker}\,\text{Bian}(g)$. The second term is $0$ since $g$ is Einstein so $R_{ij}=g_{ij}$.

So out of this, my only remaining question is as above: Why is $D\text{Rc}(g)=\text{id}$ on $\text{im}\,\delta^*_g$?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.