---
title: Busch's Extension of Gleason's Theorem
url: https://www.emergentmind.com/topics/busch-s-extension-of-gleason-s-theorem
type: topic
---

# Busch's Extension of Gleason's Theorem

Searching arXiv for Busch's extension of Gleason's theorem and related discussions on additivity, qubits, and later generalizations.
Busch’s extension of Gleason’s theorem is a characterization of quantum probability measures on the set of all effects, rather than only on the lattice of projection operators. In the form discussed by Hall, if $\mathcal H$ is a two-dimensional or higher-dimensional Hilbert space and $\mathcal E=\{E\mid 0\le E\le I\}$ is the set of all effects, then any generalized probability measure $\mu:\mathcal E\to[0,1]$ satisfying normalization and additivity on sums of effects that remain bounded by the identity has the trace form $\mu(E)=\mathrm{Tr}[\rho E]$ for a unique density operator $\rho$. This yields a derivation of the Born rule that applies even in dimension $2$, where Gleason’s original theorem does not apply [1611.00613].

## 1. Formulation of the extension

Let $\mathcal H$ be a two-dimensional (or higher-dimensional) Hilbert space, and let
\[
\mathcal E=\{E\mid 0\le E\le I\}
\]
denote the set of all effects on $\mathcal H$. A function
\[
\mu:\mathcal E\to[0,1]
\]
is called a generalized probability measure if it satisfies:

1. Normalisation: $\mu(I)=1$.
2. Additivity on compatible outcomes: whenever $\{E_1,E_2,\dots,E_n\}\subset\mathcal E$ obeys
   \[
   E_1+E_2+\cdots+E_n\le I,
   \]
   then
   \[
   \mu(E_1+E_2+\cdots+E_n)=\mu(E_1)+\mu(E_2)+\cdots+\mu(E_n).
   \]

Busch’s theorem asserts that for any such $\mu$—whether $\mathcal H$ has dimension $2$ or more—there exists a unique density operator $\rho$ on $\mathcal H$ such that
\[
\mu(E)=\mathrm{Tr}[\rho E]\qquad\text{for all }E\in\mathcal E.
\]

A later formulation states the theorem on any real or complex separable Hilbert space $\mathcal H$ of dimension $\ge 2$, with $\mu$ defined on all effects, nonnegative, normalized by $\mu(I)=1$, and countably additive on any collection of pairwise disjoint effects $\{A_i\}$ satisfying $\sum_i A_i\le I$; the conclusion is again that $\mu(A)=\mathrm{Tr}[\rho A]$ for a unique density operator $\rho$ [2603.06211].

## 2. Relation to Gleason’s original theorem

Gleason’s 1957 theorem assumed only that a probability measure is defined on the lattice of all orthogonal projectors in $\mathcal H$, is additive on sets of mutually orthogonal projectors, and satisfies $\mu(I)=1$. Under those assumptions, the theorem shows $\mu(P)=\mathrm{Tr}(\rho P)$ whenever $\dim\mathcal H\ge 3$. Busch’s extension replaces projectors by the larger set of all effects and requires additivity for sums of effects that remain $\le I$; with that stronger additivity hypothesis, the trace-form conclusion holds already in dimension $2$ [1611.00613].

| Feature | Gleason | Busch |
|---|---|---|
| Domain | Orthogonal projectors | All effects $0\le E\le I$ |
| Additivity | Orthogonal projector sums | Sums of effects with total $\le I$ |
| Dimension range | $\dim\mathcal H\ge 3$ | $\dim\mathcal H\ge 2$ |
| Representation | $\mu(P)=\mathrm{Tr}(\rho P)$ | $\mu(E)=\mathrm{Tr}(\rho E)$ |

The qubit case lies outside the original theorem because in dimension $2$ one can construct non-Born-type frame functions on the projectors alone. Hall gives the family
\[
\mu_f(P_n)=\frac12[1+f(m\cdot n)],
\]
with any odd, continuous $f$ mapping $[-1,1]$ onto itself and $f(1)=1$. These assignments satisfy the projector additivity conditions and normalization, but violate linearity in $P$. Hence the original axioms are too weak to force the trace rule in the qubit case [1611.00613].

## 3. Assumptions and proof structure

Compared with Gleason’s original assumptions, Busch’s single new ingredient is to replace the domain of definition by the larger set of all effects and to require additivity for any finite sum of effects that remains $\le I$. Hall emphasizes that no additional continuity hypothesis or “existence of eigenstates” is needed once one admits additivity over all effects [1611.00613].

The proof architecture described in later analyses breaks into a sequence of linearity results. First, for rational scalars $r\in\mathbb Q$ with $0\le r\le 1$, additivity yields homogeneity,
\[
\mu(rA)=r\,\mu(A).
\]
Second, nonnegativity and normalization give continuity at zero, $\lim_{\epsilon\to 0^+}\mu(\epsilon A)=0$. Third, rational homogeneity plus continuity extend homogeneity to all real $0\le r\le 1$. Fourth, additivity extends to arbitrary real combinations: if $A,B\in\mathcal E$ with $A+B\le I$, then for any reals $\alpha,\beta\ge 0$ with $\alpha A+\beta B\le I$,
\[
\mu(\alpha A+\beta B)=\alpha\,\mu(A)+\beta\,\mu(B).
\]
This turns $\mu$ into a positive linear functional on the operator space. One then invokes the finite-dimensional or separable Riesz representation theorem to obtain a unique density operator $\rho$ such that
\[
\mu(A)=\mathrm{Tr}[\rho A].
\]
In Hall’s summary, the same logic is presented as the extension of $\mu$ to a positive linear functional on the full algebra $\mathcal B(\mathcal H)$, followed by trace representation [2603.06211].

In this framework, nonnegativity and normalization are used to exclude pathological additive functions and to guarantee continuity of the functional. A plausible implication is that the success of the theorem depends not only on additivity but on the compatibility of additivity with positivity and normalization, since those ingredients are what permit the linear-functional machinery to go through [2603.06211].

## 4. Explicit qubit representation

For qubits, every effect $E$ can be written in Pauli-basis form
\[
E=\frac12(a_0 I+a\cdot \sigma),
\qquad 0\le E\le I,
\]
so that $(a_0,a)$ lies in the Bloch ball
\[
|a|\le a_0\le 2-|a|.
\]
If the density operator has Bloch vector $r$, so that
\[
\rho=\frac12(I+r\cdot \sigma),
\]
then Busch’s theorem yields
\[
\mu(E)=\mathrm{Tr}[\rho E]=\frac12[a_0+r\cdot a].
\]

For rank-1 projectors
\[
P_n=\frac12(I+n\cdot \sigma),\qquad |n|=1,
\]
one recovers
\[
\mu(P_n)=\mathrm{Tr}[\rho P_n]=\frac12(1+r\cdot n).
\]
If $\rho$ is pure, $r$ is a unit vector $m$ and
\[
\mu(P_n)=\frac12(1+m\cdot n)=|\langle m|n\rangle|^2,
\]
which is the usual Born rule. In this sense, Busch’s extension gives an explicit Born-rule representation for qubit effects and projectors alike [1611.00613].

## 5. Additivity, counterexamples, and the De Zela dispute

A central issue in the literature is whether the Born rule in dimension $2$ can be recovered from weaker assumptions than Busch’s. De Zela claimed that Gleason’s theorem for probability measures on the lattice of projection operators can be extended to qubits by adding assumptions related to continuity and the existence of “eigenstates.” Hall’s response is that these extra constraints still admit nonlinear assignments of the form
\[
\mu_f(P_n)=\frac12[1+f(m\cdot n)],
\]
and therefore do not suffice to recover $\mu(P)=\mathrm{Tr}(\rho P)$ [1611.00613].

By contrast, Busch’s single, strong assumption—additivity over all effects—cannot be satisfied by those nonlinear projector-only constructions. They fail to extend consistently to sums of non-projective effects $E$ with $E_1+E_2\le I$. Once one demands
\[
\mu(E_1+E_2)=\mu(E_1)+\mu(E_2)
\]
for arbitrary positive operators summing below $I$, the only surviving measures are exactly those of trace form. Hall therefore characterizes Busch’s theorem as trading the “logical” motivation of Gleason—additivity on commuting projectors—for the more powerful assumption of additivity on all positive operators $\le I$, that is, all POVM elements [1611.00613].

A later analysis strengthens this point by arguing that additivity cannot be derived from non-contextuality and normalization alone. It examines Gleason’s theorem, Busch’s extension, the Deutsch-Wallace theorem, Zurek’s envariance proof, and the Finkelstein-Hartle theorem, and concludes that these derivations either depend heavily on the additivity assumption or lead to obvious loopholes due to the lack of additivity. The same work gives a simple nonadditive example on a $d$-dimensional space,
\[
\mu(A)=\frac{(\mathrm{Tr}[A])^2}{d^2},
\]
which satisfies $\mu(I)=1$ and $\mu(A)\ge 0$ but fails additivity on orthogonal rank-1 projections, so no density operator reproduces it [2603.06211].

## 6. Later generalizations and weakened hypotheses

Busch’s theorem has itself been generalized in two distinct directions. First, Barnett, Cresser, Jeffers, and Pegg formulate a more general probability rule without assuming that positive measurement outcome operators are effects or that they form a probability operator measure. For a finite set of bounded positive operators $\{M_i\}$ associated with the possible recorded outcomes of a measurement procedure $x$, and a density operator $\rho$ associated with preparation $s$, they define
\[
X\equiv \sum_i M_i
\]
and obtain
\[
p(m_i\mid s,x)=\frac{\mathrm{Tr}(M_i\rho)}{\mathrm{Tr}(X\rho)}.
\]
When there is no measurement outcome information available, non-contextuality with respect to $x$ forces $X\propto I$, so the standard POVM Born rule is recovered as a special case. In that framework, Busch’s normalization condition $v(I)=1$ emerges as a corollary rather than an initial postulate [1308.0946].

Second, the Parseval-frame approach recasts POVMs in frame-theoretic language. A finite sequence $X=\{x_j\}_{j=1}^N\subset \mathcal H$ is a Parseval frame if and only if the rank-1 operators
\[
E_j:=x_j\otimes x_j^*
\]
satisfy
\[
\sum_{j=1}^N E_j=I_{\mathcal H},
\]
so that $\{E_j\}$ is a POVM. In this setting one proves that bounded Gleason functions for all finite Parseval frames are quadratic forms,
\[
g(x)=\langle x,Tx\rangle,
\]
and uses that result to weaken the finite-dimensional Busch theorem. Specifically, if $v:\mathcal E(\mathcal H)\to[0,1]$ satisfies $0\le v(E)\le 1$, $v(I)=1$, and there exists an integer $N\ge d+2$ such that whenever $\{E_1,\dots,E_N\}\subset\mathcal E(\mathcal H)$ satisfies $\sum_{i=1}^N E_i=I$, one has
\[
\sum_{i=1}^N v(E_i)=1,
\]
then $v$ is automatically countably additive on all countable sums in $\mathcal E(\mathcal H)$ and hence
\[
v(E)=\mathrm{tr}(\rho E)
\]
for a density operator $\rho$. This suggests that, at least in finite dimension, the full additivity demanded in Busch’s original statement can be recovered from a much weaker finite-cardinality hypothesis tied to a single sufficiently overcomplete frame [2001.06738].

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