Let $(\mathbb{S}^n,g_0)$ be the standard sphere, $n\geq 3$, consider the Nirenberg problem$$ -k(n) \Delta_{g_0} u+R_0 u=R u^{\frac{n+2}{n-2}}, \quad u>0\,\text{ on }\, \mathbb{S}^n, $$ where $k(n)=\frac{4(n-1)}{n-2}$ and $R_0=n(n-1)$ is the scalar curvature of $g_0$. Let $t \in \mathbb{R}$, let $\varphi_{t}$ be the M"obius transformation on $\mathbb{S}^n$ which, under stereographic projection with respect to the north pole, sends $y$ to $ty$. For function $v$ defined on $\mathbb{S}^n$, we let $$ T_{t}v:=v\circ\varphi_t|\det d\varphi_t|^{\frac{n-2}{2n}} $$ where $d\varphi_t$ denotes the Jacobian of $\varphi_t$. In particular, the pull-back metric of $g_v = v^{\frac{4}{n-2}}g_0$ under $\varphi_t$ is given by $\varphi_t^* (g_v) = g_{T_t v}$. Is it true that $$ -k(n) \Delta_{g_0}\left(T_\phi u\right)+R_0\left(T_\phi u\right)=(R \circ \phi)\left(T_\phi u\right)^{\frac{n+2}{n-2}}\quad \text{ on }\, \mathbb{S}^n? $$ I don't konw how to verify the above equation since the $ |\det d\varphi_t|$ term in $\Delta_{g_0}$ is hard to calculate, Thanks a lot for any help.


1 Answer 1


Yes, it is true. This is a consequence of the conformal invariance of the conformal Laplacian.

Let $(M^n,g)$ be a Riemannian manifold. The conformal Laplacian is $$ L_2^g = -\Delta + \frac{n-2}{4(n-1)}R . $$ (I use the convention $-\Delta \geq 0$.) We require two properties of this operator.

First, it is natural: If $\Phi \colon M \to M$ is a diffeomorphism, then $$ \label{1}\tag{1} \Phi^\ast (L_2^gu) = L_2^{\Phi^\ast g}(\Phi^\ast u) $$ for all $u \in C^\infty(M)$, where $\Phi^\ast f = f \circ \Phi$ is the pullback of a function and $\Phi^\ast g$ is the pullback of the initial metric.

Second, it is conformally covariant: If $\Upsilon \in C^\infty(M)$, then $$ \label{2}\tag{2} L_2^{e^{2\Upsilon}g}(u) = e^{-\frac{n+2}{2}\Upsilon} L_2^g \left( e^{\frac{n-2}{2}\Upsilon}u \right) $$ for all $u \in C^\infty(M)$.

Now let’s specialize to your setting of the sphere $(S^n,g)$ with its metric of constant sectional curvature one. Then $R=n(n-1)$, so $L_2^g$ is (proportional to) the operator you wrote down. Let $\Phi \colon S^n \to S^n$ be a conformal diffeomorphism; i.e. $\Phi^\ast g = e^{2\Upsilon}g$ for some $\Upsilon \in C^\infty(M)$. (Your $\phi_t$ is one example of a conformal diffeomorphism.) Then the volume element $\mathrm{dv}$ transforms by $$ \label{3}\tag{3} \mathrm{dv}_{\Phi^\ast g} = e^{n\Upsilon}\mathrm{dv}_g , $$ and so $\lvert \det d\Phi \rvert = e^{n\Upsilon}$. Combining Equations \eqref{1}, \eqref{2}, and \eqref{3} yields \begin{align*} (L_2^gu) \circ \Phi & = L_2^{\Phi^\ast g}(\Phi^\ast u) \\ & = \lvert \det d\Phi \rvert^{-\frac{n+2}{2n}} L_2^g \left( \lvert \det d\Phi \rvert^{\frac{n-2}{2n}} (u\circ\Phi) \right) . \end{align*} In particular, if you set $T_\Phi u := \lvert \det d\Phi \rvert^{\frac{n-2}{2n}} (u \circ \Phi)$ and assume $$ L_2^gu = \frac{n-2}{4(n-1)}fu^{\frac{n+2}{n-2}} , $$ then you conclude that \begin{align*} \frac{n-2}{4(n-1)}(f \circ \Phi)(T_\Phi u)^{\frac{n+2}{n-2}} & = \frac{n-2}{4(n-1)}\lvert \det d\Phi \rvert^{\frac{n+2}{2n}} (fu^{\frac{n+2}{n-2}})\circ \Phi \\ & = \lvert \det d\Phi \rvert^{\frac{n+2}{2n}} (L_2^gu) \circ \Phi \\ & = L_2^g(T_\Phi u) . \end{align*} This is precisely your equation.

  • $\begingroup$ Good job! Many thans! $\endgroup$ Commented Apr 15, 2023 at 13:04
  • $\begingroup$ @ Davidi Cone If you find an answer to your question correct and useful, you should accept it by clicking on the "check " sign. I think your questions are interesting. Good Luck! $\endgroup$
    – Medo
    Commented Apr 25, 2023 at 15:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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