Quantum Steering Ellipsoid Volume

• Quantum steering ellipsoid volume is defined as the volume of the Bloch-ball ellipsoid representing all steered single-qubit states, serving as a geometric descriptor of two-qubit correlations.
• Its determination uses analytic determinant formulas involving the correlation matrix and reduced Bloch radii to provide criteria for separability, entanglement, and discord.
• The volume acts as a practical correlation indicator that obeys volume-monogamy constraints in multipartite systems and remains monotone under local noise operations.

Quantum steering ellipsoid volume is the volume of the Bloch-ball ellipsoid formed by all single-qubit states to which one subsystem of a two-qubit state can be steered by local measurements on the other subsystem. In the standard two-qubit formalism, the steering set is always an ellipsoid, possibly degenerate to a disk, line segment, or point, and its volume provides a compact geometric scalar linking Bloch-representation data, partial-transpose structure, separability, steerability, and multipartite monogamy. Because the ellipsoid is invariant under appropriate local filtering and can be normalized by the Bloch-sphere volume $4\pi/3$, the quantity functions both as a faithful geometric descriptor and as a correlation indicator (Jevtic et al., 2013).

1. Analytic definition and determinant formulas

For a two-qubit state

$$\rho=\frac14 \sum_{\mu,\nu=0}^3 \Theta_{\mu\nu}\,\sigma_\mu\otimes\sigma_\nu, \qquad \Theta= \begin{pmatrix} 1 & \boldsymbol b^T\ \boldsymbol a & T \end{pmatrix},$$

the steering ellipsoid of Alice, generated by Bob’s measurements, is characterized by a center

$$\boldsymbol c_A=\frac{\boldsymbol a-T\boldsymbol b}{1-b^2},$$

and an ellipsoid matrix

$$Q_A=\frac{1}{1-b^2}\left(T-\boldsymbol a\boldsymbol b^T\right) \left(1+\frac{\boldsymbol b\boldsymbol b^T}{1-b^2}\right) \left(T^T-\boldsymbol b\boldsymbol a^T\right).$$

The semiaxis lengths are the square roots of the eigenvalues of $Q_A$, so the volume is

$V_A=\frac{4\pi}{3}\sqrt{\det Q_A}.$

Using the block determinant identity $\det\Theta=\det(T-\boldsymbol a\boldsymbol b^T)$, one obtains

$$V_A=\frac{4\pi}{3}\frac{\bigl|\det(T-\boldsymbol a\boldsymbol b^T)\bigr|}{(1-b^2)^2} =\frac{4\pi}{3}\frac{|\det\Theta|}{(1-b^2)^2}.$$

With

$\det\Theta=16\bigl(\det\rho^{T_B}-\det\rho\bigr),$

this becomes the determinant formula emphasized in the foundational literature,

$$V_A=\frac{64\pi}{3}\,\frac{\bigl|\det\rho-\det\rho^{T_B}\bigr|}{(1-b^2)^2}.$$

By exchanging Alice and Bob one gets the directional counterpart

$$\rho=\frac14 \sum_{\mu,\nu=0}^3 \Theta_{\mu\nu}\,\sigma_\mu\otimes\sigma_\nu, \qquad \Theta= \begin{pmatrix} 1 & \boldsymbol b^T\ \boldsymbol a & T \end{pmatrix},$$0

and the two directional volumes satisfy

$$\rho=\frac14 \sum_{\mu,\nu=0}^3 \Theta_{\mu\nu}\,\sigma_\mu\otimes\sigma_\nu, \qquad \Theta= \begin{pmatrix} 1 & \boldsymbol b^T\ \boldsymbol a & T \end{pmatrix},$$1

These formulas show that QSE volume is controlled jointly by the correlation block $$\rho=\frac14 \sum_{\mu,\nu=0}^3 \Theta_{\mu\nu}\,\sigma_\mu\otimes\sigma_\nu, \qquad \Theta= \begin{pmatrix} 1 & \boldsymbol b^T\ \boldsymbol a & T \end{pmatrix},$$2, the reduced Bloch radii, and the determinant difference between $$\rho=\frac14 \sum_{\mu,\nu=0}^3 \Theta_{\mu\nu}\,\sigma_\mu\otimes\sigma_\nu, \qquad \Theta= \begin{pmatrix} 1 & \boldsymbol b^T\ \boldsymbol a & T \end{pmatrix},$$3 and its partial transpose (Jevtic et al., 2013, Slater, 2020).

2. Canonical geometry, physicality, and boundary structure

The geometric content becomes especially transparent in canonical form. After Bob-side filtering that makes Bob’s reduced state maximally mixed, $$\rho=\frac14 \sum_{\mu,\nu=0}^3 \Theta_{\mu\nu}\,\sigma_\mu\otimes\sigma_\nu, \qquad \Theta= \begin{pmatrix} 1 & \boldsymbol b^T\ \boldsymbol a & T \end{pmatrix},$$4, the steering map reduces to

$$\rho=\frac14 \sum_{\mu,\nu=0}^3 \Theta_{\mu\nu}\,\sigma_\mu\otimes\sigma_\nu, \qquad \Theta= \begin{pmatrix} 1 & \boldsymbol b^T\ \boldsymbol a & T \end{pmatrix},$$5

so the ellipsoid is literally the image of the unit sphere under the affine map $$\rho=\frac14 \sum_{\mu,\nu=0}^3 \Theta_{\mu\nu}\,\sigma_\mu\otimes\sigma_\nu, \qquad \Theta= \begin{pmatrix} 1 & \boldsymbol b^T\ \boldsymbol a & T \end{pmatrix},$$6. In this representation, the center is $$\rho=\frac14 \sum_{\mu,\nu=0}^3 \Theta_{\mu\nu}\,\sigma_\mu\otimes\sigma_\nu, \qquad \Theta= \begin{pmatrix} 1 & \boldsymbol b^T\ \boldsymbol a & T \end{pmatrix},$$7 and the shape is determined by $$\rho=\frac14 \sum_{\mu,\nu=0}^3 \Theta_{\mu\nu}\,\sigma_\mu\otimes\sigma_\nu, \qquad \Theta= \begin{pmatrix} 1 & \boldsymbol b^T\ \boldsymbol a & T \end{pmatrix},$$8. The same ellipsoid can therefore be viewed either as a set of remotely preparable states or as a canonical affine image of the Bloch sphere (Jevtic et al., 2013).

This geometric representation is faithful, but not every ellipsoid inside the Bloch sphere corresponds to a physical two-qubit state. In canonical aligned form, physicality depends on the center distance $$\rho=\frac14 \sum_{\mu,\nu=0}^3 \Theta_{\mu\nu}\,\sigma_\mu\otimes\sigma_\nu, \qquad \Theta= \begin{pmatrix} 1 & \boldsymbol b^T\ \boldsymbol a & T \end{pmatrix},$$9, the shape matrix $$\boldsymbol c_A=\frac{\boldsymbol a-T\boldsymbol b}{1-b^2},$$0, the term $$\boldsymbol c_A=\frac{\boldsymbol a-T\boldsymbol b}{1-b^2},$$1, and the chirality

$$\boldsymbol c_A=\frac{\boldsymbol a-T\boldsymbol b}{1-b^2},$$2

The physical–unphysical boundary is therefore stricter than mere inclusion inside the Bloch ball. The same framework yields the “no-pancake theorem”: a physical steering ellipsoid cannot touch the Bloch sphere at more than two points unless it is the whole sphere (Milne et al., 2014).

Tangency to the Bloch sphere has a direct steering meaning. For entangled two-qubit states, pure steered states correspond exactly to tangency points between the steering ellipsoid and the Bloch sphere. The possible numbers of tangency points are

$$\boldsymbol c_A=\frac{\boldsymbol a-T\boldsymbol b}{1-b^2},$$3

and Bob’s ellipsoid has exactly $$\boldsymbol c_A=\frac{\boldsymbol a-T\boldsymbol b}{1-b^2},$$4 tangency points if and only if Alice’s ellipsoid does. In particular, if one party can steer the other to at least one pure state with nonzero probability, then the state is two-way EPR steerable (Song et al., 18 May 2026).

3. Entanglement, separability, discord, and steering

The volume immediately yields a sharp sufficient entanglement criterion. The largest ellipsoid volume compatible with separability is the volume of the insphere of the largest tetrahedron inscribed in the Bloch sphere,

$$\boldsymbol c_A=\frac{\boldsymbol a-T\boldsymbol b}{1-b^2},$$5

Hence any two-qubit state with

$$\boldsymbol c_A=\frac{\boldsymbol a-T\boldsymbol b}{1-b^2},$$6

must be entangled. The converse does not hold: entangled states can have $$\boldsymbol c_A=\frac{\boldsymbol a-T\boldsymbol b}{1-b^2},$$7. Exact separability is given instead by the nested tetrahedron theorem: a two-qubit state is separable if and only if its steering ellipsoid fits inside a tetrahedron that itself fits inside the Bloch sphere. Volume is therefore a sufficient separability test, not a complete one (Jevtic et al., 2013).

The same geometry distinguishes other correlation classes. For zero-discord states, the steering ellipsoid degenerates to a line segment on a diameter, so the volume vanishes. Nonzero volume—termed “obesity” in the steering-ellipsoid literature—is strictly stronger than discord but strictly weaker than entanglement: zero-discord states necessarily have zero volume, but separable states with nonzero volume do exist (Jevtic et al., 2013, Milne et al., 2014).

A common misconception is that QSE volume is itself a steering monotone. It is not. For Bell-diagonal states, the normalized volume is

$$\boldsymbol c_A=\frac{\boldsymbol a-T\boldsymbol b}{1-b^2},$$8

yet steerability under two projective measurements is determined by

$$\boldsymbol c_A=\frac{\boldsymbol a-T\boldsymbol b}{1-b^2},$$9

where $$Q_A=\frac{1}{1-b^2}\left(T-\boldsymbol a\boldsymbol b^T\right) \left(1+\frac{\boldsymbol b\boldsymbol b^T}{1-b^2}\right) \left(T^T-\boldsymbol b\boldsymbol a^T\right).$$0 are the two largest eigenvalues of $$Q_A=\frac{1}{1-b^2}\left(T-\boldsymbol a\boldsymbol b^T\right) \left(1+\frac{\boldsymbol b\boldsymbol b^T}{1-b^2}\right) \left(T^T-\boldsymbol b\boldsymbol a^T\right).$$1. In that restricted scenario, a Bell-diagonal state is steerable by two projective measurements if and only if it violates CHSH, and the largest normalized volume among Bell-diagonal states that are not steerable by two projective measurements is

$$Q_A=\frac{1}{1-b^2}\left(T-\boldsymbol a\boldsymbol b^T\right) \left(1+\frac{\boldsymbol b\boldsymbol b^T}{1-b^2}\right) \left(T^T-\boldsymbol b\boldsymbol a^T\right).$$2

Thus a larger ellipsoid does not necessarily imply stronger restricted-measurement steering (Quan et al., 2016).

This distinction is reflected in geometric steerability witnesses. One construction compares the steering ellipsoid with the LHS surface of the closest unsteerable assemblage and defines

$$Q_A=\frac{1}{1-b^2}\left(T-\boldsymbol a\boldsymbol b^T\right) \left(1+\frac{\boldsymbol b\boldsymbol b^T}{1-b^2}\right) \left(T^T-\boldsymbol b\boldsymbol a^T\right).$$3

If $$Q_A=\frac{1}{1-b^2}\left(T-\boldsymbol a\boldsymbol b^T\right) \left(1+\frac{\boldsymbol b\boldsymbol b^T}{1-b^2}\right) \left(T^T-\boldsymbol b\boldsymbol a^T\right).$$4, the state is steerable. The point is not that volume alone is decisive, but that steering is revealed by the excess of the quantum ellipsoid over the optimal LHS benchmark (Ku et al., 2017).

4. Extremal families and special state classes

Several state families furnish canonical benchmarks for QSE volume.

State class	Ellipsoid geometry	Volume statement
Maximally entangled pure state	Full Bloch sphere	$$Q_A=\frac{1}{1-b^2}\left(T-\boldsymbol a\boldsymbol b^T\right) \left(1+\frac{\boldsymbol b\boldsymbol b^T}{1-b^2}\right) \left(T^T-\boldsymbol b\boldsymbol a^T\right).$$5
Bell-diagonal state	Centered at origin	Normalized volume $$Q_A=\frac{1}{1-b^2}\left(T-\boldsymbol a\boldsymbol b^T\right) \left(1+\frac{\boldsymbol b\boldsymbol b^T}{1-b^2}\right) \left(T^T-\boldsymbol b\boldsymbol a^T\right).$$6
Werner state	Sphere centered at origin	Separable up to radius $$Q_A=\frac{1}{1-b^2}\left(T-\boldsymbol a\boldsymbol b^T\right) \left(1+\frac{\boldsymbol b\boldsymbol b^T}{1-b^2}\right) \left(T^T-\boldsymbol b\boldsymbol a^T\right).$$7
Zero-discord state	Line segment on a diameter	Volume $$Q_A=\frac{1}{1-b^2}\left(T-\boldsymbol a\boldsymbol b^T\right) \left(1+\frac{\boldsymbol b\boldsymbol b^T}{1-b^2}\right) \left(T^T-\boldsymbol b\boldsymbol a^T\right).$$8
Maximal fixed-center physical ellipsoid	Oblate spheroid	$$Q_A=\frac{1}{1-b^2}\left(T-\boldsymbol a\boldsymbol b^T\right) \left(1+\frac{\boldsymbol b\boldsymbol b^T}{1-b^2}\right) \left(T^T-\boldsymbol b\boldsymbol a^T\right).$$9

For a fixed center $Q_A$0, the largest-volume physical ellipsoid is an oblate spheroid with semiaxes

$Q_A$1

and volume

$Q_A$2

The associated canonical state is a rank-2 $Q_A$3-state,

$Q_A$4

with

$Q_A$5

and this family is Choi-isomorphic to the amplitude-damping channel (Milne et al., 2014).

At fixed linear entropy

$Q_A$6

the states maximizing QSE volume fall into three families: Werner states $Q_A$7 in the high-mixedness regime, maximally discordant states $Q_A$8 for

$Q_A$9

and a mixture family $V_A=\frac{4\pi}{3}\sqrt{\det Q_A}.$0 connecting $V_A=\frac{4\pi}{3}\sqrt{\det Q_A}.$1 to concurrence-maximizing states. In the steerable regime, these same families maximize the steerable weight, although QSE volume itself remains distinct from genuine EPR-steering quantifiers (McCloskey et al., 2016).

5. Multipartite volume monogamy, noise, and open-system control

Normalized QSE volume obeys nontrivial monogamy laws in multipartite systems. For a pure three-qubit state,

$V_A=\frac{4\pi}{3}\sqrt{\det Q_A}.$2

This relation is stronger than the Coffman–Kundu–Wootters inequality, and nontrivial saturation occurs exactly for $V_A=\frac{4\pi}{3}\sqrt{\det Q_A}.$3-class states, including bipartite limiting cases. GHZ-class states lie strictly below the saturation curve (Cheng et al., 2016).

The linear pure-state bound does not extend to arbitrary mixed three-qubit states. A universal weaker inequality is

$V_A=\frac{4\pi}{3}\sqrt{\det Q_A}.$4

and analogous relations extend to larger systems. For pure four-qubit states,

$V_A=\frac{4\pi}{3}\sqrt{\det Q_A}.$5

while for arbitrary $V_A=\frac{4\pi}{3}\sqrt{\det Q_A}.$6-qubit states,

$V_A=\frac{4\pi}{3}\sqrt{\det Q_A}.$7

These relations make QSE volume a geometric multipartite resource constrained by shareability, rather than a merely bipartite descriptor (Cheng et al., 2016).

The volume is also monotone under local noise: for arbitrary local CPTP maps, the steering ellipsoid volume cannot increase. This makes any valid volume-monogamy relation robust under local noise. For isotropic noise, explicit formulas show how the normalized volume shrinks with the depolarizing parameter $V_A=\frac{4\pi}{3}\sqrt{\det Q_A}.$8 (Cheng et al., 2016).

These predictions have been tested experimentally in photonic systems. Reconstructed steering ellipsoids matched the theoretical ellipsoidal geometry, $V_A=\frac{4\pi}{3}\sqrt{\det Q_A}.$9-class states lay close to the line

$\det\Theta=\det(T-\boldsymbol a\boldsymbol b^T)$0

and a mixed entangled state was observed to violate the pure-state monogamy bound while still satisfying the weaker mixed-state inequality (Zhang et al., 2018).

More recent open-system work has shown that monogamy can be driven to an extreme asymmetrical form. In independent bosonic reservoirs, selective bound-state engineering can make an untrusted party’s QSE volume decay to zero while the trusted parties’ QSE volumes remain finite. In the regime

$\det\Theta=\det(T-\boldsymbol a\boldsymbol b^T)$1

one obtains

$\det\Theta=\det(T-\boldsymbol a\boldsymbol b^T)$2

while the trusted-party volumes remain finite, realizing “extreme volume monogamy” (Zhang et al., 1 Jul 2026).

6. Volume as a measure on state space and as a transferable geometric tool

Beyond individual states, QSE volume has been used as a weight on the full $\det\Theta=\det(T-\boldsymbol a\boldsymbol b^T)$3-dimensional convex set of two-qubit states. The estimated ratio of the integral of this measure over separable states to its integral over all states is

$\det\Theta=\det(T-\boldsymbol a\boldsymbol b^T)$4

substantially smaller than the Hilbert–Schmidt and Bures benchmark values

$\det\Theta=\det(T-\boldsymbol a\boldsymbol b^T)$5

Under the same QSE-weighted sampling, the estimated absolutely separable probability is

$\det\Theta=\det(T-\boldsymbol a\boldsymbol b^T)$6

Whether the QSE separability ratio can be obtained exactly, and whether there exists a metric whose volume form reproduces the QSE measure, remain open questions. The Bloch-norm dependence also appears atypical: in the QSE setting the separability probability appears to increase as the pure-state boundary is approached (Slater, 2020).

Subsequent work tested alternative weightings by dividing the QSE volume by the eigenvalue parts of the Hilbert–Schmidt and Bures volume elements. These procedures yielded separability estimates

$\det\Theta=\det(T-\boldsymbol a\boldsymbol b^T)$7

while inclusion of the Hilbert–Schmidt unitary factor led to highly variable estimates, with median

$\det\Theta=\det(T-\boldsymbol a\boldsymbol b^T)$8

mean

$\det\Theta=\det(T-\boldsymbol a\boldsymbol b^T)$9

and variance

$$V_A=\frac{4\pi}{3}\frac{\bigl|\det(T-\boldsymbol a\boldsymbol b^T)\bigr|}{(1-b^2)^2} =\frac{4\pi}{3}\frac{|\det\Theta|}{(1-b^2)^2}.$$0

The numerical instability of these constructions indicates that the measure-theoretic status of QSE volume is still unresolved (Slater, 2021).

The formalism has also been transferred from bipartite states to single-qubit channels by means of the Choi–Jamiołkowski state. For a qubit channel $$V_A=\frac{4\pi}{3}\frac{\bigl|\det(T-\boldsymbol a\boldsymbol b^T)\bigr|}{(1-b^2)^2} =\frac{4\pi}{3}\frac{|\det\Theta|}{(1-b^2)^2}.$$1, the channel ellipsoid has center

$$V_A=\frac{4\pi}{3}\frac{\bigl|\det(T-\boldsymbol a\boldsymbol b^T)\bigr|}{(1-b^2)^2} =\frac{4\pi}{3}\frac{|\det\Theta|}{(1-b^2)^2}.$$2

ellipsoid matrix

$$V_A=\frac{4\pi}{3}\frac{\bigl|\det(T-\boldsymbol a\boldsymbol b^T)\bigr|}{(1-b^2)^2} =\frac{4\pi}{3}\frac{|\det\Theta|}{(1-b^2)^2}.$$3

semiaxes $$V_A=\frac{4\pi}{3}\frac{\bigl|\det(T-\boldsymbol a\boldsymbol b^T)\bigr|}{(1-b^2)^2} =\frac{4\pi}{3}\frac{|\det\Theta|}{(1-b^2)^2}.$$4, and volume

$$V_A=\frac{4\pi}{3}\frac{\bigl|\det(T-\boldsymbol a\boldsymbol b^T)\bigr|}{(1-b^2)^2} =\frac{4\pi}{3}\frac{|\det\Theta|}{(1-b^2)^2}.$$5

In that setting the volume serves as a visual quantifier of single-qubit quantum memory through the bound

$$V_A=\frac{4\pi}{3}\frac{\bigl|\det(T-\boldsymbol a\boldsymbol b^T)\bigr|}{(1-b^2)^2} =\frac{4\pi}{3}\frac{|\det\Theta|}{(1-b^2)^2}.$$6

and non-entanglement-breaking channels are characterized geometrically by the failure of the tetrahedral containment condition inherited from steering-ellipsoid theory (Chang et al., 2023).

Taken together, these developments place quantum steering ellipsoid volume at the intersection of affine Bloch geometry, determinant invariants, separability theory, steering diagnostics, and multipartite correlation constraints. Its main strength is algebraic and geometric compactness; its main limitation is that size alone does not exhaust the operational content of steerability.