Skip to content

Fubini–Study Geometry

The Fubini–Study geometry is the natural geometry of pure quantum states. It turns transition probabilities into distances on projective Hilbert space, so that two rays are close when they are hard to distinguish and far apart when they are nearly orthogonal.

This page uses the convention

dFS([ϕ],[ψ])=arccos⁡∣⟨ϕ∣ψ⟩∣d_{\rm FS}([\phi],[\psi]) = \arccos|\langle\phi|\psi\rangle|

for normalized representatives. Some authors use twice this distance. The factor matters in formulas for Bloch-sphere radii and quantum speed limits, so the convention will be stated explicitly whenever it matters.

Pulling this metric back to a restricted trial-state manifold produces the real tangent metric used by McLachlan variational dynamics. The projection equations and their conditioning belong there; this page owns the ambient pure-state distance geometry.

For normalized pure states, the transition probability is

Pψ→ϕ=∣⟨ϕ∣ψ⟩∣2.P_{\psi\to\phi} = |\langle\phi|\psi\rangle|^2.

It depends only on the rays, not on the chosen phases of ∣ψ⟩\lvert\psi\rangle and ∣ϕ⟩\lvert\phi\rangle. Thus it is a projective invariant.

The Fubini–Study distance is defined by

cos⁡2dFS=Pψ→ϕ,0≤dFS≤π2.\cos^2 d_{\rm FS} = P_{\psi\to\phi}, \qquad 0\le d_{\rm FS}\le\frac{\pi}{2}.

Equivalently,

dFS([ϕ],[ψ])=arccos⁡∣⟨ϕ∣ψ⟩∣.d_{\rm FS}([\phi],[\psi]) = \arccos|\langle\phi|\psi\rangle|.

The limiting cases are physically transparent:

Relation between raysTransition probabilityFubini–Study distance
identical1100
orthogonal00π/2\pi/2
nearly identicalclose to 11small

This is not a new postulate. It is the Born rule rewritten as geometry on the space of rays.

Let ∣ψ⟩\lvert\psi\rangle be a normalized representative of a ray. A small change ∣dψ⟩\lvert d\psi\rangle contains both physical motion in ray space and an unphysical phase change. The phase direction is removed by projecting orthogonally to the ray:

∣dψ⊥⟩=(I−∣ψ⟩⟨ψ∣)∣dψ⟩.\lvert d\psi_\perp\rangle = (I-\lvert\psi\rangle\langle\psi\rvert) \lvert d\psi\rangle.

The Fubini–Study line element is

dsFS2=⟨dψ⊥∣dψ⊥⟩.ds_{\rm FS}^2 = \langle d\psi_\perp|d\psi_\perp\rangle.

Equivalently,

dsFS2=⟨dψ∣dψ⟩−∣⟨ψ∣dψ⟩∣2.ds_{\rm FS}^2 = \langle d\psi|d\psi\rangle - |\langle\psi|d\psi\rangle|^2.

The subtraction removes the part of ∣dψ⟩\lvert d\psi\rangle that only changes the phase of the representative. If

∣dψ⟩=i dα ∣ψ⟩,\lvert d\psi\rangle = i\,d\alpha\,\lvert\psi\rangle,

then dsFS=0ds_{\rm FS}=0, as it must: the ray has not moved.

The formula for dsFS2ds_{\rm FS}^2 is invariant under a phase change of the representative,

∣ψ⟩↦eiχ∣ψ⟩.\lvert\psi\rangle \mapsto e^{i\chi}\lvert\psi\rangle.

This invariance is essential. A metric on projective Hilbert space must assign the same distance no matter which phase convention is used to lift a ray to a vector.

The same metric can be expressed through projectors. For two nearby pure projectors,

Πψ=∣ψ⟩⟨ψ∣,Πψ+dψ=Πψ+dΠ,\Pi_\psi = \lvert\psi\rangle\langle\psi\rvert, \qquad \Pi_{\psi+d\psi} = \Pi_\psi+d\Pi,

one has, with the convention of this page,

dsFS2=12Tr⁡(dΠ dΠ).ds_{\rm FS}^2 = \frac12\operatorname{Tr}(d\Pi\,d\Pi).

The projector expression makes phase independence manifest and is often useful when comparing pure-state geometry with density-operator methods.

The finite distance between two rays is the length of the shortest Fubini–Study path between them:

dFS([ϕ],[ψ])=arccos⁡∣⟨ϕ∣ψ⟩∣.d_{\rm FS}([\phi],[\psi]) = \arccos|\langle\phi|\psi\rangle|.

The absolute value appears because the phases of representatives are arbitrary. One may choose phases so that ⟨ϕ∣ψ⟩\langle\phi|\psi\rangle is real and nonnegative, and the shortest path lies in the two-dimensional complex subspace spanned by the two vectors.

If the rays are not orthogonal, a convenient geodesic representative is

∣ψ(s)⟩=sin⁡(dFS−s)∣ψ⟩+sin⁡s ∣ϕ∥⟩sin⁡dFS,0≤s≤dFS,\lvert\psi(s)\rangle = \frac{ \sin(d_{\rm FS}-s)\lvert\psi\rangle + \sin s\,\lvert\phi_{\parallel}\rangle } {\sin d_{\rm FS}}, \qquad 0\le s\le d_{\rm FS},

where ∣ϕ∥⟩\lvert\phi_{\parallel}\rangle is the phase-adjusted representative of the final ray with positive overlap with ∣ψ⟩\lvert\psi\rangle. This formula is meant as geometry, not as a claim that the system dynamically follows this path under a given Hamiltonian.

For a qubit,

P(C2)=CP1,\mathbb P(\mathbb C^2) = \mathbb{CP}^1,

which is the Bloch sphere. A normalized representative may be written

∣ψ⟩=cos⁡θ2∣0⟩+eiϕsin⁡θ2∣1⟩.\lvert\psi\rangle = \cos\frac{\theta}{2}\lvert0\rangle + e^{i\phi}\sin\frac{\theta}{2}\lvert1\rangle.

The Bloch vector is

n=(sin⁡θcos⁡ϕ,sin⁡θsin⁡ϕ,cos⁡θ).\mathbf n = (\sin\theta\cos\phi,\sin\theta\sin\phi,\cos\theta).

With the distance convention used here, the Fubini–Study line element becomes

dsFS2=14(dθ2+sin⁡2θ dϕ2).ds_{\rm FS}^2 = \frac14 \left( d\theta^2 + \sin^2\theta\,d\phi^2 \right).

Thus CP1\mathbb{CP}^1 is a sphere of radius 1/21/2 in this convention. The ordinary angular separation γ\gamma between two Bloch vectors satisfies

∣⟨ϕ∣ψ⟩∣2=1+nϕ⋅nψ2=cos⁡2γ2,|\langle\phi|\psi\rangle|^2 = \frac{1+\mathbf n_\phi\cdot\mathbf n_\psi}{2} = \cos^2\frac{\gamma}{2},

so

dFS=γ2.d_{\rm FS} = \frac{\gamma}{2}.

Antipodal Bloch vectors correspond to orthogonal qubit states and have dFS=π/2d_{\rm FS}=\pi/2.

On the qubit chart where the coefficient of ∣0⟩\lvert0\rangle is nonzero, write

z=c1c0,∣ψ(z)⟩=∣0⟩+z∣1⟩1+∣z∣2.z = \frac{c_1}{c_0}, \qquad \lvert\psi(z)\rangle = \frac{\lvert0\rangle+z\lvert1\rangle} {\sqrt{1+|z|^2}}.

Then the Fubini–Study line element is

dsFS2=∣dz∣2(1+∣z∣2)2.ds_{\rm FS}^2 = \frac{|dz|^2}{(1+|z|^2)^2}.

Using

z=eiϕtan⁡θ2z=e^{i\phi}\tan\frac{\theta}{2}

recovers the radius-1/21/2 sphere formula above. In higher-dimensional finite systems, similar local coordinates turn CPN−1\mathbb{CP}^{N-1} into a complex manifold with a natural Hermitian metric.

The metric speed is one side of the projective geometry. The complementary symplectic derivation of the same Schrödinger path is developed in Hamiltonian Flow on Projective Hilbert Space.

For Schrödinger evolution,

iℏddt∣ψ(t)⟩=H∣ψ(t)⟩,i\hbar\frac{d}{dt}\lvert\psi(t)\rangle = H\lvert\psi(t)\rangle,

the physical speed of the ray is

dsFSdt=ΔψHℏ,\frac{ds_{\rm FS}}{dt} = \frac{\Delta_\psi H}{\hbar},

where

(ΔψH)2=⟨H2⟩ψ−⟨H⟩ψ2.(\Delta_\psi H)^2 = \langle H^2\rangle_\psi - \langle H\rangle_\psi^2.

This formula has a clean interpretation. The component of H∣ψ⟩H\lvert\psi\rangle proportional to ∣ψ⟩\lvert\psi\rangle only changes the phase of the vector representative. The component orthogonal to ∣ψ⟩\lvert\psi\rangle changes the ray. Energy uncertainty measures exactly the size of that orthogonal component.

An energy eigenstate has ΔψH=0\Delta_\psi H=0, so its ray is stationary even though the vector accumulates a phase.

The length of any path in projective Hilbert space is at least the geodesic distance between its endpoints. Therefore Schrödinger evolution implies

arccos⁡∣⟨ψ(0)∣ψ(T)⟩∣≤1ℏ∫0Tdt Δψ(t)H.\arccos|\langle\psi(0)|\psi(T)\rangle| \le \frac{1}{\hbar} \int_0^T dt\,\Delta_{\psi(t)}H.

For time-independent ΔH\Delta H, this gives the Mandelstam–Tamm form

T≥ℏΔHarccos⁡∣⟨ψ(0)∣ψ(T)⟩∣.T \ge \frac{\hbar}{\Delta H} \arccos|\langle\psi(0)|\psi(T)\rangle|.

For evolution to an orthogonal state,

T≥πℏ2ΔH.T \ge \frac{\pi\hbar}{2\Delta H}.

This is a geometry statement about the minimum projective distance that must be traversed. Stronger or different speed limits can use additional assumptions, energy above the ground state, open-system metrics, or mixed-state geometry. This page only previews the pure-state Fubini–Study version.

The detailed bridge, including the quantum geometric tensor and the pullback of the Fubini–Study two-form to Berry curvature, is Relation to Berry Geometry.

The Fubini–Study metric is only part of the natural geometry of projective Hilbert space. The imaginary part of the Hilbert-space inner product gives a compatible symplectic form, and the phase bundle over ray space gives a natural Berry connection.

The metric answers:

How far apart are the rays?\text{How far apart are the rays?}

The Berry connection answers:

How does a phase convention twist along a path?\text{How does a phase convention twist along a path?}

Both structures come from the same Hilbert-space inner product, but they encode different physics. The metric controls distinguishability and speed; the connection controls holonomy and geometric phase.

  • Forgetting the convention factor: some references use 2arccos⁡∣⟨ϕ∣ψ⟩∣2\arccos|\langle\phi|\psi\rangle| instead of arccos⁡∣⟨ϕ∣ψ⟩∣\arccos|\langle\phi|\psi\rangle|.
  • Treating the Bloch sphere as radius 11 while using formulas normalized for radius 1/21/2.
  • Applying the Fubini–Study distance to mixed states without switching to an appropriate mixed-state metric.
  • Confusing geodesic distance with the actual dynamical path length under a specific Hamiltonian.
  • Forgetting that vector phase changes must have zero projective length.
  • Thinking energy expectation ⟨H⟩\langle H\rangle controls ray speed; it is energy uncertainty ΔH\Delta H that controls the Fubini–Study speed.
  • J. P. Provost and G. Vallee, “Riemannian structure on manifolds of quantum states,” Communications in Mathematical Physics 76, 289, 1980.
  • J. Anandan and Y. Aharonov, “Geometry of quantum evolution,” Physical Review Letters 65, 1697, 1990.
  • T. W. B. Kibble, “Geometrization of quantum mechanics,” Communications in Mathematical Physics 65, 189, 1979.
  • I. Bengtsson and K. Zyczkowski, Geometry of Quantum States, 2nd ed., Cambridge University Press, 2017.
  • D. C. Brody and L. P. Hughston, “Geometric quantum mechanics,” Journal of Geometry and Physics 38, 19, 2001.
  • L. Mandelstam and I. Tamm, “The uncertainty relation between energy and time in non-relativistic quantum mechanics,” Journal of Physics (USSR) 9, 249, 1945.
  • M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information, 10th anniversary ed., Cambridge University Press, 2010.
  1. Show that the infinitesimal Fubini–Study line element vanishes for pure phase motion.
Solution

Let

∣dψ⟩=i dα ∣ψ⟩\lvert d\psi\rangle = i\,d\alpha\,\lvert\psi\rangle

with ⟨ψ∣ψ⟩=1\langle\psi|\psi\rangle=1. Then

⟨dψ∣dψ⟩=dα2,⟨ψ∣dψ⟩=i dα.\langle d\psi|d\psi\rangle = d\alpha^2, \qquad \langle\psi|d\psi\rangle = i\,d\alpha.

Therefore

dsFS2=⟨dψ∣dψ⟩−∣⟨ψ∣dψ⟩∣2=dα2−dα2=0.ds_{\rm FS}^2 = \langle d\psi|d\psi\rangle - |\langle\psi|d\psi\rangle|^2 = d\alpha^2-d\alpha^2 = 0.

The vector moved along the phase fiber, but the ray did not move.

  1. Compute the Fubini–Study distance between orthogonal normalized states.
Solution

If ⟨ϕ∣ψ⟩=0\langle\phi|\psi\rangle=0, then

dFS([ϕ],[ψ])=arccos⁡0=π2.d_{\rm FS}([\phi],[\psi]) = \arccos 0 = \frac{\pi}{2}.

Orthogonal rays are maximally separated in this convention.

  1. Derive the qubit relation dFS=γ/2d_{\rm FS}=\gamma/2.
Solution

For qubit pure states with Bloch vectors nψ\mathbf n_\psi and nϕ\mathbf n_\phi,

∣⟨ϕ∣ψ⟩∣2=1+nϕ⋅nψ2.|\langle\phi|\psi\rangle|^2 = \frac{1+\mathbf n_\phi\cdot\mathbf n_\psi}{2}.

If γ\gamma is the ordinary angle between the Bloch vectors, then

nϕ⋅nψ=cos⁡γ.\mathbf n_\phi\cdot\mathbf n_\psi = \cos\gamma.

Hence

∣⟨ϕ∣ψ⟩∣2=1+cos⁡γ2=cos⁡2γ2.|\langle\phi|\psi\rangle|^2 = \frac{1+\cos\gamma}{2} = \cos^2\frac{\gamma}{2}.

Taking the positive square root and applying dFS=arccos⁡∣⟨ϕ∣ψ⟩∣d_{\rm FS}=\arccos|\langle\phi|\psi\rangle| gives

dFS=γ2,d_{\rm FS} = \frac{\gamma}{2},

with 0≤γ≤π0\le\gamma\le\pi.

  1. Show that Schrödinger evolution has Fubini–Study speed ΔH/ℏ\Delta H/\hbar.
Solution

Schrödinger evolution gives

∣ψ˙⟩=−iℏH∣ψ⟩.\lvert\dot\psi\rangle = -\frac{i}{\hbar}H\lvert\psi\rangle.

The Fubini–Study speed is

(dsFSdt)2=⟨ψ˙∣ψ˙⟩−∣⟨ψ∣ψ˙⟩∣2.\left(\frac{ds_{\rm FS}}{dt}\right)^2 = \langle\dot\psi|\dot\psi\rangle - |\langle\psi|\dot\psi\rangle|^2.

Compute the two terms:

⟨ψ˙∣ψ˙⟩=1ℏ2⟨H2⟩ψ,⟨ψ∣ψ˙⟩=−iℏ⟨H⟩ψ.\langle\dot\psi|\dot\psi\rangle = \frac{1}{\hbar^2}\langle H^2\rangle_\psi, \qquad \langle\psi|\dot\psi\rangle = -\frac{i}{\hbar}\langle H\rangle_\psi.

Therefore

(dsFSdt)2=1ℏ2(⟨H2⟩ψ−⟨H⟩ψ2)=(ΔψH)2ℏ2.\left(\frac{ds_{\rm FS}}{dt}\right)^2 = \frac{1}{\hbar^2} \left( \langle H^2\rangle_\psi - \langle H\rangle_\psi^2 \right) = \frac{(\Delta_\psi H)^2}{\hbar^2}.

Taking the positive square root gives dsFS/dt=ΔψH/ℏds_{\rm FS}/dt=\Delta_\psi H/\hbar.

  1. Use the speed formula to obtain the orthogonal-state Mandelstam–Tamm bound for constant ΔH\Delta H.
Solution

The projective path length obeys

L=∫0Tdt ΔHℏ=ΔHℏTL = \int_0^T dt\, \frac{\Delta H}{\hbar} = \frac{\Delta H}{\hbar}T

when ΔH\Delta H is constant. Any path connecting orthogonal states must have length at least the geodesic distance π/2\pi/2. Thus

ΔHℏT≥π2,\frac{\Delta H}{\hbar}T \ge \frac{\pi}{2},

or

T≥πℏ2ΔH.T \ge \frac{\pi\hbar}{2\Delta H}.