2
$\begingroup$

The Fubini-Study metric $g:=g_{FS}$ is the unique $U(n+1)$-invariant Riemannian metric on the complex projective space $\mathbb{CP}^{n}$ the complex projective space which by $U(n+1)$-invariance can be wlog definined on tangent bundle over affine chart $ U_0 :=\{Z \in \mathbb{CP}^n \ \vert z_0 \neq 0 \} \subset \mathbb{CP}^n $ and then it extends uniquely by $U(n+1)$-action to all affine pieces $U_i$.

So we can restict our considerations to tangent space $T_Z \mathbb{CP}^n$ over a point

$$Z =[1:z_1:....z_n] \in U_0 :=\{Z \in \mathbb{CP}^n \ \vert z_0 \neq 0 \} \subset \mathbb{CP}^n$$

Let $\{\partial _{1},\ldots ,\partial _{n}\}$ the be frame of tangent space
$T_Z \mathbb{CP}^n$ at $Z$, ie the canonical $\mathbb{C}$-basis of $T_Z \mathbb{CP}^n$. Then the Fubini Study metric is defined by

$$\tag{FS} g_{i{\bar {j}}}=h(\partial _{i},{\bar {\partial }}_{j})= {\frac {(1+|\mathbf {z} |^{2})\delta _{i{\bar {j}}}- {\bar {z}}_{i}z_{j}}{(1+|\mathbf {z} |^{2})^{2}}} $$

At first glance one can think that $g$ somehow "has hallen from heaven" but it is well knows that it's not. One way to derive it is using Kähler potentials and canonical choice of a distinguished connection: Chern connection: See Here the excellent answer by Arctic Char.

The concern of this question is to find out if it also possible to derive $g$ on $\mathbb{CP}^{n}$ naturally by recognizing $(\mathbb{CP}^{n},g)$ as Riemannian submersion of $(\mathbb{C}^{n+1} \backslash \{0\},h)$ where $h$ is to restriction of standard Hermitian metric on $\mathbb{C}^{n+1}$ to $\mathbb{C}^{n+1} \backslash \{0\}$.

What do I mean by 'naturally"?

Well, $\mathbb{CP}^{n}$ can be considered as submersion of complex manifold $\mathbb{C}^{n+1} \backslash \{0\}$ via canonical projection

$$\pi: \mathbb{C}^{n+1} \backslash \{0\} \to \mathbb{CP}^n$$

The complex manifold is moreover Hermitian since $ \mathbb{C}^{n+1} \backslash \{0\}$ inherits naturally the standard Hermitian metric $h$ on $ \mathbb{C}^{n+1} $. Futhermore by $\pi$ the $\mathbb{CP}^{n}$ can also be considered as quotient $ \mathbb{C}^{n+1} \backslash \{0\} / \mathbb{C}^*$.

By definition a Riemannian submersion $F: (M,g_M) \to (N, g_N)$ is a submersion from one Riemannian manifold $M$ with Riemannian metric $g^M$ to another Riemannian manifold $N$ with Riemannian metric $g^N$ that respects the metrics, meaning that it is an orthogonal projection on tangent spaces.

Namely that means that for every $x \in M$ and $y=F(X)$ the restriction of the differential map $dF_x$ to $ker(dF_x)^{\perp} \subset T_x M $ gives an isometry $d_x F: ker(dF_x)^{\perp} \to T_y N$ such that the metric $g^N$ is induced by $g^M$ is following sense:

Let $A, B \in T_y N$ and denote by $\overline{A}, \overline{B} \in ker(dF_x)^{\perp}$ their unique preimages with respect the isometry $d_x F: ker(dF_x)^{\perp} \to T_y N$.

Then

$$\tag{RieSub} g^N _y(A,B)= g^M _x(\overline{A}, \overline{B})$$

QUESTION:

How can be proved that the Fubini-Study metric $g$ arises exactly on this way? ie that the canonical projection

$\pi: \mathbb{C}^{n+1} \backslash \{0\} \to \mathbb{CP}^n$

extends naturally to Riemannian submersion $\pi: (\mathbb{C}^{n+1} \backslash \{0\},h) \to (\mathbb{CP}^n, g)$

is the sense above, where $h$ is the restriction of the stadard Hermitian metric on $\mathbb{C}^{n+1}$ and $g$ is the Fubini- Study metric defined above in (FS).

In other words we need to verify that the metrics $h$ and $g$ are related to each other as in (RieSub).

Take as $y= Z =[1:z_1:....z_n]$ and $x =(1,z_1,..., z_n) \in \mathbb{C}^{n+1}$ and $\partial _{i}, \partial _{j} \in T_y \mathbb{CP}^n$ element of the canonical basis of the tangent space at $y$. Firstly we need an explicit identification between the tangent space $T \mathbb{CP}^n$ and $ker(d \pi)^{\perp}$. We need this isometry explicitly since in order to prove (FS) we have to work with unique preimages of elements in $T_y \mathbb{CP}^n$ living in $ker(d \pi_x)^{\perp}$.

But I don't see how can I show that indeed

$$g(\partial _{i},{\bar {\partial }}_{j})= {\frac {(1+|\mathbf {z} |^{2})\delta _{i{\bar {j}}}- {\bar {z}}_{i}z_{j}}{(1+|\mathbf {z} |^{2})^{2}}} $$

holds. On the left side we have already identified $\partial _{i}, \partial _{j} \in T_y \mathbb{CP}^n$ with their unique preimages in $ker(d_x \pi)^{\perp}$.

$\endgroup$
6
  • 3
    $\begingroup$ It can't be a Riemannian submersion. This is more easily seen in the case of the radial projection $\pi: \mathbb{R}^n\backslash\{0\} \rightarrow S^{n-1}$, which also is not a Riemannian submersion, because the restriction of $\pi$ to each sphere of radius $r$ is not an isometry. You have to put a scale invariant metric on $\mathbb{R}^n\backslash\{0\}$ or $\mathbb{C}^n\backslash\{0\}$ to make the projection map a Riemannian submersion. $\endgroup$
    – Deane Yang
    Commented Jul 27, 2020 at 23:57
  • 1
    $\begingroup$ Alternatively, you could restrict to the unit sphere, where $\pi$ becomes a Riemannian submersion. This may not look natural because the unit sphere is not a complex manifold and the standard Hermitian metric on $\mathbb C^n$ is somewhat arbitrary. But in a technical sense, the construction is still natural because the isometry group of $S^{n+1}$ maps surjectively to that of $\mathbb{CP}^n$. $\endgroup$ Commented Jul 28, 2020 at 15:08
  • $\begingroup$ @DeaneYang: Hi. What do you percisely mean by "scale invariant metric" on a manifold $M$? Altough I think I intuitively understand what you mean, I nowhere found literature defining it precisely. Below I will sketch what I guess you mean, but I'm not pretty sure if that's correct. The next important question is if it possible to obtain a "scale invariant metric" $g_{si}$ on $\mathbb{C}^n\backslash\{0\}$ by naive geometric modification of the stadard Hermitian metric $h$. $\endgroup$
    – user267839
    Commented Aug 3, 2020 at 16:28
  • $\begingroup$ My guess : Do you mean that a metric $g$ on complex manifold $M= \mathbb{C}^n\backslash\{0\}$ is a scale invariant metric if at every point $m \in M$, arbitrary tangent vectors $X, Y \in T_m M$ at $m$ and arbitrary scalars $a, b \in \mathbb{C} \backslash\{0\}$ following holds: $\endgroup$
    – user267839
    Commented Aug 3, 2020 at 16:29
  • $\begingroup$ Let decompose $X= X^{\perp}+ X^{\parallel}$ and $Y= Y^{\perp}+ Y^{\parallel}$ with unique $X^{\parallel}, Y^{\parallel} \in \mathbb{C} \cdot m$ and $X^{\perp}, Y^{\perp} \in (\mathbb{C} \cdot m)^{\perp}$. Then $g(X^{\perp}+ aX^{\parallel},Y^{\perp}+ bY^{\parallel})= g(X^{\perp}+ X^{\parallel},Y^{\perp}+ Y^{\parallel})$ for arbitrary scalars $a, b \in \mathbb{C} \backslash\{0\}$. By linearity this implies that $g_m$ is constant on subpace $\mathbb{C} \cdot m \subset T_m M$ for every $m \in M$. Is this the definition of scale invariant metric you have in mind? $\endgroup$
    – user267839
    Commented Aug 3, 2020 at 16:31

1 Answer 1

2
$\begingroup$

Let me start from scratch. Note that everything below uses only the definition of complex projective space and the natural Hermitian inner product on $\mathbb{C}^{n+1}$. Also, the construction is coordinate-independent in the sense that everything below can be done with an abstract complex vector space with a Hermitian inner product without using any basis.

Recall that canonical projection map is defined to be \begin{align*} \mathbb{C}^{n+1}\backslash\{0\} &\rightarrow \mathbb{C}P^n\\ z &\mapsto [z], \end{align*} where $[z] = \{ tz\ :\ t \in \mathbb{C} \}$. For each $z \ne 0$, the tangent space of $\mathbb{C}^{n+1}$ at $z$ splits naturally $$ T_z\mathbb{C}^{n+1} = [z] \oplus z^\perp, $$ where $z^\perp = \{ w \in \mathbb{C}^{n+1}\ :\ \langle w,z\rangle = 0 \}$. For each $z$, the pushforward map $\pi_*: T_z\mathbb{C}^{n+1} \rightarrow T_{[z]}\mathbb{C}P^n$ is an isomorphism if restricted to $z^\perp$.

We want to use the Hermitian inner product to define a Kähler metric $g$ on $\mathbb{C}P^n$. Observe that, given such a metric, $\pi^*g$ is an degenerate Hermitian $2$-tensor $h$ on $T_z\mathbb{C}$ that is scale invariant in the sense that if $R_t(z) = tz$, then $(R_t)^*h = h$. Conversely, a Hermitian metric on $\mathbb{C}^{n+1}$ defines one on $\mathbb{C}P^n$ only if its restriction to each $z^\perp$ is scale invariant.

If we want to define the metric using only the standard Hermitian inner product pointwise on $\mathbb{C}^{n+1}$, then the only possibility is one of the form $$ g = f(|z|^2)|dz|^2 $$ restricted to $z^\perp$, for each $z \in \mathbb{C}^{n+1}\backslash\{0\}$. On the other hand, if you work out all the definitions carefully, you find that $$ R_t^*g = f(|t|^2|z|^2)|t|^2|dz|^2, $$ which equals $g$ if and only if there exists a real constant $c$ such that $f(|z|^2) = c|z|^{-2}$. By the observations above, this metric can be pushed down to a metric on $\mathbb{C}P^n$, and it can be verified that it is, up to a scale factor, the Fubini-Study metric.

Note that the explanation above is just the definition of a Riemannian submersion but for the specific situation here.

$\endgroup$
7
  • $\begingroup$ Thank you for your excellent answer! One detain I still not understand: the scale invariance is defined by $(R_t)^*h = h$. In your case this means that $f(|z|^2)|dz|^2 = R_t^*g = f(|t|^2|z|^2)|t|^2|dz|^2$, thus $ f(|z|^2)= f(|t|^2|z|^2)|t|^2$. Why this already imply $f(|z|^2) = |z|^{-2}$? $\endgroup$
    – user267839
    Commented Aug 4, 2020 at 1:15
  • $\begingroup$ What I said in my answer was not exactly correct, I've revised it. As for your question, note that the identity for $f$ is equivalent to saying that $f(x) = f(ax)a^2$, for any positive $x, a$. Now set $y = ax$. $\endgroup$
    – Deane Yang
    Commented Aug 4, 2020 at 14:49
  • 1
    $\begingroup$ I'm not sure if this could be of any use here (maybe to simplify the notation?), but it's interesting to note that $T_{[z]}\Bbb CP^n$ is naturally identified with the vector space of linear maps ${\rm Hom}([z],z^{\perp})$. In general, $T_p{\rm Gr}(k,V)={\rm Hom}(p,V/p)$ even without the choice of a form in $V$, where ${\rm Gr}(k,V)$ denotes the Grassmannian of $k$-dimensional subspaces of a vector space $V$. $\endgroup$
    – Dry Bones
    Commented Aug 4, 2020 at 21:14
  • $\begingroup$ @DryBones, that's the right abstract definition of the tangent space, in the sense that it does not depend on coordinates or any inner product. So it's useful when you want to study properties of complex projective space that do not depend on the metric. $\endgroup$
    – Deane Yang
    Commented Aug 4, 2020 at 23:11
  • 1
    $\begingroup$ Choose $z$ so that $|z| = 1$. $\endgroup$
    – Deane Yang
    Commented Aug 5, 2020 at 19:02

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .