---
title: Fubini-Study Metric Explained
url: https://www.emergentmind.com/topics/fubini-study-metric-bcee11ee-be30-42f5-ba40-caa9bad788c9
type: topic
---

# Fubini-Study Metric Explained

The **Fubini–Study metric** is the canonical metric on complex projective space, the space of complex rays rather than phase-labeled vectors. For nonzero vectors \(z,w\) in a complex inner-product space, its global distance is
\[
d_{\mathrm{FS}}([z],[w])
=\arccos\!\left(\frac{|\langle z,w\rangle|}{\|z\|\,\|w\|}\right).
\]
It is invariant under nonzero complex rescaling and, in particular, under \(U(1)\) phase transformations. Infinitesimally, it is the real part of the quantum geometric tensor (QGT),
\[
g_{ij}^{\mathrm{FS}}
=\operatorname{Re}\!\left[
\langle\partial_i\psi|\partial_j\psi\rangle
-\langle\partial_i\psi|\psi\rangle
\langle\psi|\partial_j\psi\rangle
\right],
\]
so it measures physically distinguishable changes of a quantum ray after removal of the phase direction. The same geometry appears in complex differential geometry, Kähler and projective geometry, quantum information, Lie-group state manifolds, Bergman geometry, CR geometry, Grassmannians, non-archimedean geometry, and projective descriptions of representation drift.

## 1. Definition and differential-geometric structure

Let \(|\psi(\lambda)\rangle\) be a smooth family of normalized states. The ordinary derivative \(|\partial_i\psi\rangle\) contains a component parallel to \(|\psi\rangle\), corresponding to a change of phase or normalization convention. The orthogonal projector
\[
P_\perp=I-|\psi\rangle\langle\psi|
\]
removes this component. The gauge-invariant QGT is
\[
Q_{ij}
=\langle\partial_i\psi|P_\perp|\partial_j\psi\rangle
=\langle\partial_i\psi|\partial_j\psi\rangle
-\langle\partial_i\psi|\psi\rangle
\langle\psi|\partial_j\psi\rangle .
\]
Its decomposition is
\[
Q_{ij}=g_{ij}+i\sigma_{ij},
\]
where \(g_{ij}=\operatorname{Re}Q_{ij}\) is symmetric and \(\sigma_{ij}=\operatorname{Im}Q_{ij}\) is antisymmetric. The Fubini–Study line element is
\[
ds_{\mathrm{FS}}^{2}=g_{ij}\,d\lambda^i d\lambda^j.
\]

With the Berry connection
\[
A_i=i\langle\psi|\partial_i\psi\rangle,
\]
the Berry curvature is
\[
F_{ij}=\partial_iA_j-\partial_jA_i,
\qquad
F_{ij}=-2\,\operatorname{Im}Q_{ij}
\]
for this convention. Thus the QGT combines a Riemannian metric measuring ray distinguishability with a symplectic or curvature component governing geometric phase.

For neighboring normalized states, the fidelity satisfies
\[
|\langle\psi(\lambda)|\psi(\lambda+d\lambda)\rangle|^2
=1-g_{ij}\,d\lambda^i d\lambda^j+O(d\lambda^3).
\]
The corresponding global projective distance is
\[
d_{\mathrm{FS}}(\psi,\phi)
=\arccos|\langle\psi|\phi\rangle|.
\]
Conventions sometimes multiply this distance or metric by factors of \(2\) or \(4\); the pure-state quantum Fisher information is commonly \(4g_{\mathrm{FS}}\).

## 2. Complex projective and Kähler geometry

Complex projective space is
\[
\mathbb{CP}^{n}
=\{\text{one-dimensional complex subspaces of }\mathbb C^{n+1}\}.
\]
On the affine chart \(Z_0\neq0\), with \(z_j=Z_j/Z_0\), the standard Fubini–Study potential is
\[
\Phi_{\mathrm{FS}}(z)
=\log\!\left(1+\sum_{j=1}^{n}|z_j|^2\right),
\]
and the Kähler form is
\[
\omega_{\mathrm{FS}}
=i\,\partial\bar\partial
\log\!\left(1+\sum_{j=1}^{n}|z_j|^2\right).
\]
The metric coefficients are
\[
(g_{\mathrm{FS}})_{\alpha\bar\beta}
=
\frac{(1+|z|^2)\delta_{\alpha\beta}
-\bar z_\alpha z_\beta}
{(1+|z|^2)^2}.
\]

The metric is invariant under the projectivized unitary group \(PU(n+1)\), and \(U(n+1)\) acts transitively on \(\mathbb{CP}^n\). Up to multiplication by a nonzero constant, it is the unique \(U(n+1)\)-invariant Kähler metric. Its curvature is positive and has constant holomorphic sectional curvature; the numerical value depends on normalization. In one convention the standard metric has holomorphic sectional curvature \(2\), while another uses curvature \(4\). Under a positive rescaling \(g'=\lambda g_{\mathrm{FS}}\), holomorphic sectional curvature scales as
\[
\operatorname{HSC}(g')=\frac{\operatorname{HSC}(g_{\mathrm{FS}})}{\lambda}.
\]

The Fubini–Study metric is also the canonical metric induced by a projective embedding. If \(\mathcal L\) is a very ample line bundle over a compact Kähler manifold \(X\), then
\[
\iota:X\hookrightarrow
\mathbb P\bigl(H^0(X,\mathcal L)^*\bigr)
\]
is the Kodaira embedding. A positive-definite Hermitian form \(H\) on \(H^0(X,\mathcal L)\) determines a metric \(FS(H)\) by requiring that an \(H\)-orthonormal basis \(\{s_i\}\) satisfy
\[
\sum_i |s_i|_{FS(H)}^2=1.
\]
Locally, if \(s_i=f_i e\), then the induced weight is
\[
-\log|e|_{FS(H)}^2
=\log\!\left(\sum_i|f_i|^2\right),
\]
and its curvature is the pullback of the projective Fubini–Study form.

The Hilbert map sends a positively curved Hermitian metric \(h\) on \(\mathcal L\) to
\[
\operatorname{Hilb}(h)(s,t)
=\int_X h(s,t)\,\omega_h^n.
\]
For a very ample line bundle, the Hilbert map is surjective onto positive-definite Hermitian forms, while the Fubini–Study map is injective [1705.11025]. The Fubini–Study current associated with an orthonormal basis of sections is asymptotically governed by the curvature of the equilibrium metric, including for sequences of line bundles and continuous or non-positive metrics [2410.09265].

## 3. Rigidity, projective inducedness, and curvature

The Fubini–Study metric is characterized by several rigidity phenomena. On a closed connected Kähler manifold of real dimension at least four, degree of mobility \(D\geq3\) forces either affine equivalence of every \(h\)-projectively equivalent Kähler metric or, up to scaling,
\[
(M,cg,J)
=
(\mathbb{CP}^n,g_{\mathrm{FS}},J_{\mathrm{standard}}).
\]
Thus complex projective space is the unique closed Kähler model with genuinely non-affine \(h\)-projective freedom of this size [1009.5530].

In real dimension four, a complete Einstein metric with positive sectional curvature that is Hermitian with respect to an integrable complex structure is, up to scaling and isometry, the Fubini–Study metric on \(\mathbb{CP}^2\). The non-Kähler Einstein alternatives in LeBrun’s classification—the Page metric and the Chen–LeBrun–Weber metric—are excluded by the existence of disjoint totally geodesic surfaces, contradicting Frankel’s intersection theorem under positive sectional curvature [1112.4181].

Projective inducedness imposes additional algebraic restrictions. For a Kähler–Einstein manifold admitting a Kähler immersion into \(\mathbb{CP}^{n+k}\), rotation invariance and codimension \(k\leq3\) lead to only three possibilities:
\[
(\mathbb{CP}^n,g_{\mathrm{FS}}),\qquad
(\mathbb{CP}^2,2g_{\mathrm{FS}}),\qquad
(\mathbb{CP}^1\times\mathbb{CP}^1,
g_{\mathrm{FS}}\oplus g_{\mathrm{FS}}).
\]
The second is induced by the quadratic Veronese embedding
\[
\nu_2:\mathbb{CP}^2\to\mathbb{CP}^5,
\qquad
\nu_2^*g_{\mathrm{FS}}=2g_{\mathrm{FS}},
\]
and the third by the Segre embedding
\[
\sigma:\mathbb{CP}^1\times\mathbb{CP}^1\to\mathbb{CP}^3,
\qquad
\sigma^*g_{\mathrm{FS}}
=g_{\mathrm{FS}}\oplus g_{\mathrm{FS}}.
\]
The classification uses the logarithmic Fubini–Study potential, Calabi diastasis, rotation-invariant monomial expansions, the Kähler–Einstein Monge–Ampère equation, and codimension counting [1607.07177].

For Bergman spaces, an orthonormal basis \(\{\varphi_j\}\) defines the Bergman–Bochner map
\[
B=[\varphi_1,\varphi_2,\ldots]:M\to\mathbb P^\infty,
\]
and
\[
B^*\omega_{\mathrm{FS}}
=
i\partial\bar\partial\log K_M(z,z)
=
\omega_B.
\]
Thus the Bergman metric is literally the pullback of the Fubini–Study metric. Positive constant holomorphic sectional curvature forces the Bergman space to be finite-dimensional and the manifold to be biholomorphic to a domain in \(\mathbb P^n\). Negative constant curvature yields ball-type domains \(\mathbb B^n\setminus E\), where \(E\) is closed and pluripolar; global flat Bergman metrics are ruled out under the stated nondegeneracy assumptions [2302.13456].

## 4. Quantum geometry and dynamical applications

For a unitary evolution generated by a Hermitian operator \(H\),
\[
|\partial_t\psi\rangle=-iH|\psi\rangle,
\]
the Fubini–Study speed is
\[
\frac{ds_{\mathrm{FS}}}{dt}
=\frac{\Delta H}{\hbar},
\qquad
(\Delta H)^2
=\langle H^2\rangle-\langle H\rangle^2.
\]
This is the Anandan–Aharonov relation. A component of \(H\) proportional to the identity generates only a global phase and is removed by the projective projection.

For a parameterized circuit
\[
|\psi(\boldsymbol\theta)\rangle
=U(\boldsymbol\theta)|\psi_0\rangle,
\]
the metric is
\[
g_{ij}
=\operatorname{Re}\!\left[
\langle\partial_i\psi|\partial_j\psi\rangle
-\langle\partial_i\psi|\psi\rangle
\langle\psi|\partial_j\psi\rangle
\right].
\]
If a parameter locally generates \(G_i\), then
\[
g_{ij}
=
\operatorname{Re}\!\left[
\langle G_iG_j\rangle
-\langle G_i\rangle\langle G_j\rangle
\right].
\]
Consequently, the metric is a covariance metric on generator fluctuations. In Lie-group state manifolds, the structure constants determine the adjoint transport of generators, while the reference state determines their covariance matrix [1706.00250].

For a qubit in a magnetic field with spherical parameters \((\theta,\varphi)\),
\[
ds_{\mathrm{FS}}^2
=\frac14\left(d\theta^2+\sin^2\theta\,d\varphi^2\right),
\]
one quarter of the unit-sphere metric. Doubling the distance produces the ordinary unit Bloch-sphere metric. The Berry curvature is proportional to the sphere’s area form and corresponds, in magnetic-field coordinates, to a monopole of charge \(\pm\tfrac12\).

For mixed states, a purification and the square-root derivative
\[
C_i=\partial_i\sqrt{\rho}
\]
give a generalized QGT,
\[
g_{ij}
=
\operatorname{Tr}(C_i^\dagger C_j)
-
\operatorname{Tr}(\sqrt{\rho}\,C_i^\dagger)
\operatorname{Tr}(\sqrt{\rho}\,C_j).
\]
Under unitary evolution, the resulting line element is
\[
ds_{\mathrm{FS}}^2
=
-dt^2\operatorname{Tr}\bigl([\sqrt{\rho},H]^2\bigr),
\]
which captures the quantum, noncommutative part of Hamiltonian uncertainty rather than the full variance. Imposing monotonicity under completely positive trace-preserving maps yields a quantum Fisher-information metric. Generalized \(\alpha\)-metrics retain the projective construction while allowing dependence on powers of \(\rho\) and, in general, a nonzero dynamical phase [1503.04146].

The metric also enters variational quantum optimization. L2O-\(g^\dagger\) embeds a Fubini–Study preconditioner into a coordinate-wise LSTM learned optimizer for parameterized quantum circuits:
\[
B_t=(1-\boldsymbol\gamma_t)g_t^\dagger
+\boldsymbol\gamma_t I,
\]
followed by
\[
\boldsymbol\theta_{t+1}
=
\boldsymbol\theta_t
-\boldsymbol\eta_t\circ B_t\boldsymbol v_t.
\]
Here the learned interpolation balances metric-aware state-space motion and ordinary parameter-space updates [2407.14761].

## 5. Grassmannians, bands, and projective data

For a fixed-dimensional Grassmannian \(G_p(F^n)\), the Fubini–Study distance is the projective angle between Plücker images:
\[
d_{\mathrm{FS}}(V,W)
=
\arccos\!\left(\prod_{i=1}^{p}\cos\theta_i\right),
\]
where \(\theta_i\) are the principal angles. Equivalently, if \(A\) and \(B\) are unit decomposable \(p\)-vectors representing \(V\) and \(W\),
\[
d_{\mathrm{FS}}(V,W)
=
\arccos|\langle A,B\rangle|.
\]
The product of cosines is the volume contraction produced by orthogonal projection.

On the Total Grassmannian, where subspaces may have different dimensions, the directed extension is
\[
d_{\mathrm{FS}}(V,W)
=
\arccos\frac{\|P_WA\|}{\|A\|}.
\]
For \(\dim V\leq\dim W\), it is the product formula above; for \(\dim V>\dim W\), it equals \(\pi/2\). It is asymmetric:
\[
d_{\mathrm{FS}}(V,W)=0
\iff V\subset W,
\]
whereas both directed distances vanish only when \(V=W\). It satisfies the oriented triangle inequality
\[
d_{\mathrm{FS}}(U,W)
\leq d_{\mathrm{FS}}(U,V)+d_{\mathrm{FS}}(V,W),
\]
and is invariant under orthogonal or unitary transformations [2310.17865].

For Bloch bands, the metric measures the variation of the eigenstate ray across momentum. In the atomic \(s\)-\(p\) chain considered in [2303.02126], the Bloch eigenstates depend on an angle \(\omega_k\), and
\[
g_{kk}^{\pm}
=
\left(\frac{\partial_k\omega_k}{2}\right)^2.
\]
The two bands have the same metric. In the flat-band regime
\[
\bar\varepsilon=0,\qquad |\lambda|=1,
\]
one has
\[
\partial_k\omega_k=\pm l,
\qquad
g_{kk}^{\pm}=\frac{l^2}{4}.
\]
The winding number constrains the minimum total Fubini–Study length but does not determine its local distribution. Consequently, the metric can distinguish systems with identical winding number, including nontrivial systems with different flatness properties.

In non-archimedean geometry, a strict Cartesian ultrametric norm on the section space induces a Fubini–Study metric on the analytification of a projective variety. Its Monge–Ampère measure is supported on finitely many Shilov points, and the associated Monge–Ampère polytope is
\[
P_{\mathrm{MA}}
=
\sum_\alpha m_\alpha\Delta_{I_e(x_\alpha)}.
\]
For Chow-stable \((X,\mathcal O(1))\), criticality under \(\mathrm{SL}(E)\) is equivalent to
\[
0\in P_{\mathrm{MA}},
\]
to minimization of the Chow norm, and to residual semistability under the non-archimedean Kempf–Ness criterion [2212.02486].

## 6. Extensions, asymptotics, and limitations

The Fubini–Study construction extends beyond finite-dimensional projective manifolds. For filtered Toeplitz spaces on compact strictly pseudoconvex CR manifolds, spectral eigenfunctions define projective embeddings
\[
[F_k]:X\to\mathbb{CP}^{N_k-1}.
\]
The pullback Fubini–Study metric has an asymptotic expansion whose leading Reeb-direction term is of order \(k^2\),
\[
r_0
=
\operatorname{var}(|\chi|^2)\,
\alpha_P\otimes\alpha_P,
\]
while the first horizontal term is of order \(k\) and is governed by \(d\alpha_P\). Under Heisenberg-type rescaling, the embeddings recover the contact and Levi geometry. The same kernels determine Gaussian CR ensembles and the asymptotic distribution of zero divisors [2401.09143].

For sequences of metrized line bundles, Fubini–Study currents satisfy
\[
\gamma_p
-\frac1{A_p}c_1(L_p,h_p^{\mathrm{eq}})
\longrightarrow0
\]
weakly as currents under the hypotheses of [2410.09265]. In the tensor-power case this recovers
\[
\frac{\gamma_p}{p}
\longrightarrow c_1(L,h^{\mathrm{eq}}).
\]
The equilibrium metric, rather than the original metric when it lacks semipositive curvature, governs the limiting projective current.

The Fubini–Study metric also has dynamical and variational limitations. It is an unstable generalized stationary solution of Ricci flow on \(\mathbb{CP}^2\): numerical simulations indicate that certain conformal non-Kähler perturbations develop finite-time Type-I singularities whose rescalings approach the FIK03 blowdown shrinking Kähler–Ricci soliton with reversed complex orientation [2403.06427]. On complex Grassmannians, the Fubini–Study metric can possess nonzero infinitesimal Einstein deformations while remaining nonlinearly rigid because the deformations are obstructed at second order when the ambient dimension \(m+n\) is odd [2403.18757].

Finally, complete one-period endpoint data for a periodically driven quantum system do not generally determine its period-averaged parameter-space Fubini–Study metric. Such endpoint data determine the conjugation path of the monodromy but not its unitary lift. Centralizer-valued periodic micromotion can preserve every one-period endpoint while changing the intra-period state trajectory and hence
\[
\frac1T\int_0^T
\operatorname{Var}_{\rho_0}
\!\left(iU_\theta^\dagger(t)\partial_\theta U_\theta(t)\right)\,dt.
\]
Thus the Fubini–Study functional depends on the full parameter-dependent trajectory, not merely on endpoint propagators [2608.10051].

Across these settings, the common principle is that the Fubini–Study metric removes unphysical scalar freedom and measures geometry on projective objects: complex rays, embedded Kähler manifolds, quantum-state manifolds, Plücker lines, spectral CR embeddings, or non-archimedean projective models. Its numerical normalization varies, but its defining operation remains the orthogonal removal of the direction that does not change the underlying projective object.

Source: https://www.emergentmind.com/topics/fubini-study-metric-bcee11ee-be30-42f5-ba40-caa9bad788c9