Gleason’s Theorem Foundations

Gleason’s theorem answers a conditional question: once sharp events are represented by Hilbert-space projections, which normalized probability assignments are consistent across all compatible decompositions? In complex dimension at least three, the answer is the density-operator trace rule. The full theorem gives the exact statement, proof strategy and infinite-dimensional hypotheses; here the issue is what that result explains about quantum probability.

Helpful background. Projectors represents sharp events; Born Rule supplies the probability rule being characterized.

Probability attached to an event

A projection $P$ may occur in several projective measurements. A single function $\mu(P)$ assigns it the same probability wherever it appears. For a complete family of mutually orthogonal projections,

$$\sum_iP_i=I,\qquad \sum_i\mu(P_i)=1.$$

In infinite dimension the theorem uses countable additivity, not only finite additivity. Positivity and normalization are probability assumptions. Reuse of the same $\mu(P)$ is noncontextuality of event probabilities, not a hidden assignment of definite values to every observable.

Given these premises and dimension at least three, a unique density operator satisfies

$$\mu(P)=\operatorname{Tr}(\rho P).$$

The theorem narrows an initially large class of conceivable event-weight functions to the quantum class. It has not inferred the Hilbert-space event structure from raw observations or explained why an individual outcome occurs.

Certainty fixes the familiar pure-state rule

Suppose a particular preparation assigns probability one to $P_\psi=|\psi\rangle\langle\psi|$. The trace conclusion gives $\langle\psi|\rho|\psi\rangle=1$. Since $\rho\ge0$ and $\operatorname{Tr}\rho=1$, all its weight lies in that one-dimensional subspace, so $\rho=P_\psi$. For another unit vector $\phi$,

$$\mu(P_\phi)=\operatorname{Tr}(P_\psi P_\phi) =|\langle\phi|\psi\rangle|^2.$$

This compact argument identifies exactly where certainty about the prepared ray enters. Without that condition, the general answer is mixed-state probability, not an arbitrarily selected pure-state rule.

As a qutrit example, prepare $|0\rangle$ and measure in the basis $(|0\rangle+|1\rangle)/\sqrt2$, $(|0\rangle-|1\rangle)/\sqrt2$, $|2\rangle$. The probabilities are $(1/2,1/2,0)$. The zero event retains probability zero when placed in any other compatible basis containing $|2\rangle$; its probability belongs to the event, not to the name of the apparatus setting.

Why dimension two and POVMs need care

For a qubit, every rank-one projective measurement consists of an antipodal Bloch-sphere pair. One can choose one ray from each pair to have value one and the other zero. This obeys normalization on every such basis but is not a density-operator trace assignment. Continuity alone does not repair the exception: the smooth assignment $\mu(P_{\boldsymbol n})=(1+n_z^3)/2$ also normalizes on every antipodal pair but is not linear in the Bloch vector as a trace rule must be. Additional structural assumptions, such as extending the probability function to all POVM effects, must be stated before claiming a qubit extension.

A theorem for all effects and POVMs changes the domain of the probability function. It can constrain qubits because effects participate in more decompositions than sharp projectors alone. It is a different premise set, not permission to drop the original theorem’s dimension hypothesis. See the full theorem for the relevant variants.

Relation to the measurement and Born-rule problems

Gleason-type reasoning is useful when the task is to characterize a probability rule from structural consistency. It does not tell us whether probability means objective chance, an agent’s expectation, a typicality measure or a branch weight. Nor does it derive a state-update instrument or an outcome ontology. Those are questions for Probability and the Born Rule and the Measurement Problem.

Exercise: a nonlinear probability guess

In dimension three try $\mu(P_\phi)=|\langle\phi|0\rangle|^4$. Compare normalization in the computational basis and in the rotated basis used above.

Solution

The computational basis gives $1+0+0=1$. The rotated basis gives $1/4+1/4+0=1/2$. A rule that works in one preferred measurement can fail when all contexts are required. Renormalizing separately in each basis would abandon assignment of one probability to each projection.

References

• Busch, Paul, “Quantum States and Generalized Observables: A Simple Proof of Gleason’s Theorem,” Physical Review Letters 91, 120403 (2003), doi:10.1103/PhysRevLett.91.120403 — an effects-based extension.
• Gleason, Andrew M., “Measures on the Closed Subspaces of a Hilbert Space,” Journal of Mathematics and Mechanics 6, 885–893 (1957), doi:10.1512/iumj.1957.6.56050 — original projection-measure theorem.