---
title: Gleason’s Theorem in Quantum Mechanics
url: https://www.emergentmind.com/topics/gleason-s-theorem
type: topic
---

# Gleason’s Theorem in Quantum Mechanics

Gleason’s theorem is a result in the foundations of quantum mechanics and in the geometry of Hilbert spaces. In its standard Hilbert-space form, it states that for a real or complex Hilbert space with dimension at least \(3\), any non-contextual probability assignment to projections that is additive on orthogonal resolutions of the identity must be representable by a density operator through the Born rule, \(\mu(P)=\operatorname{Tr}(\rho P)\). The theorem therefore links probability measures on quantum propositions to the operator-state formalism, and it has become a central reference point for contextuality, quantum logic, state reconstruction, and operator-algebraic generalizations [1904.00533].

## 1. Classical Hilbert-space statement

In the standard formulation, one considers a Hilbert space \(H\) and the set \(\mathcal{P}(H)\) of orthogonal projections. A measurement is modeled as a projection-valued measure (PVM), that is, a finite family of projectors \(\{\Pi_i\}\) satisfying
\[
\Pi_i^2=\Pi_i,\qquad \Pi_i^\dagger=\Pi_i,\qquad \Pi_i\Pi_j=\delta_{ij}\Pi_i,\qquad \sum_i \Pi_i=I.
\]
A probability assignment is then a map on projectors such that probabilities sum to \(1\) on every PVM and depend only on the projector, not on which PVM contains it; in the frame-function language, this is a map \(f\) on rays or unit vectors satisfying \(\sum_i f(e_i)=1\) for every orthonormal basis [2511.15607].

The standard theorem can be stated as follows. Let \(H\) be a complex Hilbert space with \(\dim(H)\ge 3\). Suppose \(\mu:\mathcal{P}(H)\to[0,1]\) satisfies \(\mu(\Pi)\ge 0\) for all projectors and
\[
\sum_i \mu(\Pi_i)=1
\]
for every orthogonal resolution of the identity \(I=\sum_i \Pi_i\). Then there exists a density operator \(\rho\), with \(\rho\ge 0\) and \(\operatorname{Tr}\rho=1\), such that
\[
\mu(\Pi)=\operatorname{Tr}(\rho \Pi)
\]
for all projectors \(\Pi\) [2511.15607]. In equivalent rank-one language, for \(P_\psi=|\psi\rangle\langle\psi|\), one has
\[
f(\psi)=\langle \psi|\rho|\psi\rangle,
\]
so the theorem yields the Born rule on pure outcomes [2106.06182].

A constructive route begins from a frame function \(f\) and defines a quadratic form
\[
Q(x)=\|x\|^2 f\!\left(\frac{x}{\|x\|}\right),\qquad Q(0)=0.
\]
By polarization one obtains a bounded sesquilinear form \(B\), and by the Riesz representation theorem a unique bounded positive self-adjoint operator \(A\) such that \(B(x,y)=\langle x|Ay\rangle\). The frame-function weight gives \(\operatorname{Tr}(A)=1\), so \(A\) is the density operator \(\rho\) representing the measure [1904.00533]. In finite-dimensional complex Hilbert spaces, an \(L^2\) version of the theorem shows that if a complex frame function \(f:S(H)\to\mathbb{C}\) lies in \(L^2(S(H),\nu_H)\), then there exists a unique linear operator \(A\) with
\[
f(z)=(z|Az)
\]
for all \(z\in S(H)\); if \(f\) is real-valued, then \(A\) is Hermitean [1205.4504].

## 2. The two-dimensional obstruction

The classical dimension threshold is essential. In \(\mathcal{H}_2\), every rank-one projector \(\Pi\) appears only in a single two-outcome PVM \(\{\Pi,I-\Pi\}\). Consequently, the additivity constraint provides only
\[
\mu(\Pi)+\mu(I-\Pi)=1,
\]
without the higher-dimensional “intertwining” across different resolutions of the identity that drives the original theorem [2511.15607].

This leaves room for non-Born assignments. In Bloch form, a rank-one qubit projector can be written as
\[
\Pi_{\mathbf n}=\frac{I+\mathbf n\cdot\boldsymbol{\sigma}}{2},
\]
and a density operator as
\[
\rho=\frac{I+\mathbf r\cdot\boldsymbol{\sigma}}{2},\qquad \|\mathbf r\|\le 1.
\]
The Born rule for the PVM \(\{\Pi_{\mathbf n},I-\Pi_{\mathbf n}\}\) is
\[
p(\mathbf n)=\operatorname{Tr}(\rho \Pi_{\mathbf n})=\frac{1+\mathbf r\cdot\mathbf n}{2},
\]
but one can also define dispersion-free assignments \(\mu(\Pi)\in\{0,1\}\) consistent with \(\mu(\Pi)+\mu(I-\Pi)=1\) [2511.15607]. A geometric example is a hemisphere assignment on the Bloch sphere: \(\mu(\Pi_{\mathbf n})=1\) for \(\mathbf n\) in some hemisphere and \(0\) otherwise [2511.15607].

Explicit counterexamples recur throughout the modern literature. One example is
\[
g(P)=
\begin{cases}
0,& P=|0\rangle\langle 0|,\\
1,& P=|1\rangle\langle 1|,\\
\frac12,& \text{otherwise,}
\end{cases}
\qquad g(0)=0,\qquad g(I)=1,
\]
which respects every two-outcome PVM on \(\mathbb{C}^2\) but is not of Born form [1808.08091]. Another family is
\[
\mu_m(P_{\mathbf n})=\frac12[1+f(\mathbf m\cdot \mathbf n)],
\]
where \(\mathbf m\) is fixed and \(f:[-1,1]\to[-1,1]\) is continuous, odd, and satisfies \(f(1)=1\); unless \(f\) is linear, this is not representable as \(\operatorname{Tr}(\rho P_{\mathbf n})\) [1611.00613]. A Bloch-space analysis makes the contrast geometrically explicit: for \(d=2\), additivity reduces to a two-term constraint \(f(x)+f(-x)=0\), while for \(d\ge 3\) one obtains genuine multi-term constraints that force linearity [2603.07745].

## 3. Operator-algebraic and field-theoretic generalizations

The Hilbert-space theorem extends naturally to von Neumann algebras and, in part, to \(AW^\ast\)-algebras. For a von Neumann algebra \(M\) with no type \(\mathrm{I}_2\) direct summand, the Mackey–Gleason–Bunce–Wright theorem states that any bounded orthoadditive measure on \(\mathcal{P}(M)\) extends uniquely to a bounded linear operator on \(M\); scalar-valued positive measures extend to states [1408.4597]. In \(AW^\ast\)-algebras, a COrtho-morphism \(\Phi:\mathcal{P}(A)\to\mathcal{P}(B)\) preserving orthocomplements and arbitrary joins extends to a normal Jordan \(\ast\)-homomorphism when \(A\) has no type \(\mathrm{I}_2\) direct summand [1408.4597]. This formulation emphasizes a general principle already present in the Hilbert-space theorem: projection-level data determine global operator-algebraic structure [1408.4597].

The exclusion of type \(\mathrm{I}_2\) is the operator-algebraic counterpart of the qubit obstruction. In homogeneous type \(\mathrm{I}_n\) \(AW^\ast\)-algebras with \(n\ge 3\), bounded measures on projections extend to bounded linear maps, while the type \(\mathrm{I}_2\) case retains the classical obstruction [1408.4597].

Field dependence is also nontrivial. For real and complex Hilbert spaces, the classical trace formula remains basis-independent in the usual way. In quaternionic Hilbert spaces, however, the naive trace
\[
\operatorname{tr}_N(A)=\sum_{u\in N}\langle u,Au\rangle
\]
depends on the orthonormal basis \(N\) unless \(A\) is self-adjoint. The corrected common form of the theorem for \(\mathbb{R}\), \(\mathbb{C}\), and \(\mathbb{H}\) uses the real trace
\[
\operatorname{Re\,Tr}(A)=\operatorname{Re}\left(\sum_{u\in N}\langle u,Au\rangle\right),
\]
which is basis-independent on trace-class operators [1803.06882]. The resulting theorem states that for a separable Hilbert space over \(\mathbb{D}\in\{\mathbb{R},\mathbb{C},\mathbb{H}\}\) with \(\dim\mathcal{H}\ge 3\), every \(\sigma\)-additive probability measure on \(\mathcal{P}(\mathcal{H})\) is represented as
\[
\mu(P)=\operatorname{Re\,Tr}(\rho P),
\qquad
\rho=\rho^\ast\ge 0,\qquad
\operatorname{Re\,Tr}(\rho)=1
\]
[1803.06882]. This corrects the quaternionic formulation attributed to Varadarajan by replacing the naive trace with its real part [1803.06882].

## 4. Extensions beyond the original projection setting

A major line of work extends Gleason-type conclusions to \(d=2\) by enlarging the class of measurements from PVMs to POVMs. In Busch’s theorem, probabilities are assigned to effects \(E\ge 0\) with \(\sum_i E_i=I\), and under additivity and noncontextuality one obtains
\[
\mu(E)=\operatorname{Tr}(\rho E)
\]
for all effects, including the qubit case [2511.15607]. Closely related approaches due to Caves–Fuchs–Schack derive the same conclusion from consistency conditions for probabilities on effects [2511.15607]. A Parseval-frame reformulation shows that rank-one POVMs are essentially equivalent to Parseval frames, and that under appropriate assumptions Gleason functions for Parseval frames are quadratic forms \(x\mapsto \langle Ax,x\rangle\) [2001.06738].

A more restrictive route stays closer to the projective setting while still enlarging the operational assumptions. One theorem derives Born’s rule for all finite dimensions \(d\ge 2\) by requiring consistent probability assignments to projective measurements and their classical mixtures. In that setting, projective-simulable measurements \(\mathbf{PSM}_d\) are enough: if a frame function respects \(\mathbf{PSM}_d\), then
\[
f(e)=\operatorname{Tr}(\rho e)
\]
for all effects \(e\in\mathcal{E}_d\) [1808.08091]. In the qubit case, the proof uses only the two-outcome POVMs
\[
\mathbb{D}_e=\llbracket e,I-e\rrbracket
\]
and the restricted three-outcome simulable POVMs
\[
\mathbb{T}_e=\llbracket \tfrac e2,\tfrac e2,I-e\rrbracket,\qquad
\mathbb{T}_{e,e'}=\llbracket \tfrac e2,\tfrac {e'}2,I-\tfrac{e+e'}2\rrbracket,
\]
from which one derives
\[
f\!\left(\tfrac e2\right)=\tfrac12 f(e),\qquad
f(e+e')=f(e)+f(e')
\]
and hence the trace form [1808.08091].

Other generalizations weaken the global hypotheses rather than enlarging the measurement set. A generalized Gleason theorem shows that if the additivity and normalization constraints hold only locally on an open connected set of commuting projectors, then the probability assignment still has an affine-linear form
\[
\mu(E)=\operatorname{Tr}(\hat{\Delta}E)+K
\]
with \(\hat{\Delta}\) Hermitian and \(K\) constant on overlapping connected domains [2206.11830]. In hidden-variable models with a finite amount of information about the measurement context, this implies that conditional on the finite context index \(n\), probabilities for projective measurements with at least three outcomes are affine-linear in the projectors [2206.11830].

## 5. Qubits, composition, and composite systems

A distinct strategy for recovering the qubit case keeps the measurement class at PVMs but strengthens the structural assumptions about composition. One recent theorem assumes finite-dimensional Hilbert spaces, PVMs, the tensor-product composition axiom
\[
\mathcal{H}_{AB}=\mathcal{H}_A\otimes \mathcal{H}_B,
\]
the local measurement representation
\[
\Pi_x^A \mapsto \Pi_x^A\otimes I_B,
\]
and a strengthened state assumption that states of a \(d\)-dimensional system correspond to marginal frame functions [2511.15607]. The resulting statement is:
\[
f(\Pi)=\operatorname{Tr}(\rho \Pi)\qquad \text{for all } \Pi\in\mathcal{P}(\mathcal{H}_d),
\]
for every \(d\ge 2\), including \(d=2\), with \(\rho\) a density matrix [2511.15607].

The key additional condition is independence across composition:
\[
\mu_A(\Pi_A)=\mu_{AB}(\Pi_A\otimes I_B).
\]
Operationally, the probability assigned to a local outcome must be the same whether the system is treated on its own or as part of a larger one [2511.15607]. For a qubit \(A\), one chooses an ancilla \(B\) with \(\dim\mathcal{H}_B\ge 2\), so that \(\dim(\mathcal{H}_A\otimes\mathcal{H}_B)\ge 3\). Gleason’s theorem applies on \(AB\), yielding
\[
\mu_{AB}(\Pi_{AB})=\operatorname{Tr}(\rho_{AB}\Pi_{AB}).
\]
Then the partial trace identity gives
\[
\mu_A(\Pi_A)=\mu_{AB}(\Pi_A\otimes I_B)
=\operatorname{Tr}(\rho_{AB}(\Pi_A\otimes I_B))
=\operatorname{Tr}((\operatorname{Tr}_B\rho_{AB})\Pi_A),
\]
so
\[
\rho_A=\operatorname{Tr}_B(\rho_{AB}),\qquad
\mu_A(\Pi_A)=\operatorname{Tr}(\rho_A\Pi_A)
\]
[2511.15607].

This excludes dispersion-free assignments precisely because a non-Born qubit frame function cannot be written as \(\operatorname{Tr}(\rho_A\Pi_A)\) for any density matrix \(\rho_A\), hence cannot arise as a subsystem marginal of a composite-system frame function [2511.15607]. The same idea reappears in a broader operator-algebraic setting for bipartite systems: contextual probability data over product abelian subalgebras become a genuine state on the tensor-product von Neumann algebra once one adds compatibility with dilations and with dynamical correspondences [2205.00493]. In that setting, for \(N_1\) and \(N_2\) with no type \(\mathrm{I}_2\) summands, time-oriented global sections of the normal dilated probabilistic presheaf over \(V(N_1)\times V(N_2)\) correspond bijectively to normal states on \(N_1\bar{\otimes}N_2\) [2205.00493].

## 6. Relation to contextuality, symmetry, and modern interpretations

Gleason’s theorem and the Kochen–Specker theorem are closely related but distinct. The Kochen–Specker obstruction concerns \(0\)-\(1\) valuations on projectors, whereas Gleason’s theorem rules out all non-Born noncontextual probability measures in dimension at least \(3\). An elementary comparison based on linear equations for a function \(m\) on the unit sphere shows that Kochen–Specker contradictions can be exhibited on a finite set of rays, whereas the reduced Gleason argument requires constraints across the whole sphere to force the quadratic form \(m(x)=x^T A x\) in the real case [1411.4583]. In both cases, dimension \(2\) is exceptional [1411.4583].

The theorem also underlies modern symmetry results. In a complex separable Hilbert space with \(\dim H\ge 3\), if a map on rank-one projections sends complete orthogonal systems of rank-one projections to complete orthogonal systems, then for any density operator \(D\), the quantity
\[
O_D(P)=\operatorname{Tr}(D\,\phi(P))
\]
defines a frame function. Gleason’s theorem then gives a density operator \(E\) with
\[
\operatorname{Tr}(D\,\phi(P))=\operatorname{Tr}(EP),
\]
which can be used to show that \(\phi\) preserves transition probabilities and hence is implemented by a unitary or antiunitary operator [2106.06182]. This yields an optimal version of Wigner’s theorem under one-direction orthogonality preservation plus the image of a single complete orthogonal system [2106.06182].

Several contemporary derivations of the Born rule use Gleason’s theorem as the last structural step after establishing a frame-function condition by other means. One example combines envariance with a basis-independent sum rule over orthonormal sets and then invokes Gleason to obtain
\[
p(|\psi\rangle)=\langle \psi|A|\psi\rangle,
\]
followed by specialization to projective measurements and POVMs [1406.0138]. Another line, formulated in terms of contexts, systems, and modalities, uses Uhlhorn’s theorem to derive unitary changes of context and then uses Gleason to infer
\[
f(P)=\operatorname{Tr}(\rho P)
\]
from normalization over rank-one resolutions of the identity [2111.10758]. These approaches do not change the theorem’s mathematical content, but they reinterpret its hypotheses as consequences of other structural principles.

A recurring misconception is that Gleason’s theorem directly settles the qubit case. It does not in its original PVM form. The two-dimensional obstruction persists unless one adds further assumptions, such as POVMs, classical mixtures of projective measurements, or consistency under composition [1808.08091]. Another misconception is that the theorem is merely a statement about continuity; several modern proofs and reformulations derive the trace form from additivity, noncontextuality, and structural constraints without assuming continuity as an independent axiom [1808.08091].

In this sense, Gleason’s theorem remains both a precise representation theorem and a benchmark for extensions. Its classical content is the identification of noncontextual additive measures with density operators. Its modern extensions clarify exactly which extra assumptions—POVMs, classical mixing, dilations, dynamical correspondences, or composition consistency—are sufficient to recover Born-rule probabilities when the original hypotheses are too weak, most notably in dimension \(2\) [2511.15607].

Source: https://www.emergentmind.com/topics/gleason-s-theorem