Kochen–Specker Theorem

The Kochen–Specker theorem rules out a deterministic, measurement-noncontextual assignment of values to all sharp quantum observables in Hilbert-space dimension at least three. In projector form:

Let $\mathcal H$ be a finite-dimensional real or complex Hilbert space with $\dim\mathcal H=d\geq3$. There is no map

$$v:\mathcal P_1(\mathcal H)\longrightarrow\{0,1\}$$

on all rank-one projectors such that, for every orthonormal basis $\{|\psi_i\rangle\}$,

$$\sum_{i=1}^{d} v\!\left(|\psi_i\rangle\langle\psi_i|\right) =1.$$

The map would assign exactly one “yes” outcome to every complete rank-one projective measurement, while giving a shared projector the same value in every context. The theorem says those requirements cannot be satisfied globally. It does not rule out contextual hidden-variable models, stochastic models under weaker assumptions, or the two-dimensional projective case.

Required background. Contextuality defines the traditional value-assignment problem and distinguishes it from generalized operational contextuality; Projectors supplies sharp events; Projection-Valued Measures supplies complete sharp contexts.

Helpful background. Gleason’s Theorem turns any normalized additive measure on projectors into a density-operator trace rule in dimension at least three.

The valuation problem

Consider a deterministic ontological model at a fixed ontic state $\lambda$. Each sharp yes-no event $P$ receives a value

$$v_\lambda(P)\in\{0,1\}.$$

If $\mathcal C=\{P_1,\ldots,P_d\}$ is a complete rank-one PVM, exclusivity and exhaustivity require

$$\sum_{i=1}^{d}v_\lambda(P_i)=1.$$

Measurement noncontextuality removes the context label from the value. If $P_1$ also belongs to another PVM $\mathcal C'$, the model uses the same $v_\lambda(P_1)$ in both. Without this identification, one can trivially choose one outcome in every basis independently and no contradiction follows.

For an observable

$$A=\sum_a aP_a,$$

the equivalent formulation assigns one spectral value $v_\lambda(A)$ and preserves functional relations,

$$v_\lambda(f(A)) = f(v_\lambda(A)).$$

For jointly measurable observables, the valuation likewise respects their joint polynomial relations. The projector rule follows by applying this condition to spectral indicator functions.

A coloring problem on overlapping bases

Represent every rank-one projector by a vertex and every orthonormal basis by a hyperedge containing $d$ vertices. A Kochen–Specker valuation is then a zero-one coloring with exactly one vertex colored one on every hyperedge.

An isolated basis is always colorable. The obstruction comes from overlaps: one ray can belong to several bases, so its color propagates between hyperedges. Kochen and Specker exhibited a finite set of directions in three-dimensional real space for which all such propagation choices end in a contradiction. Later constructions reduced or reorganized the finite sets; they do not change the theorem’s assumptions.

This hypergraph form separates two issues:

• local consistency: every one of the listed PVMs has exactly one selected outcome;
• global consistency: a projector shared by several PVMs has one value across all of them.

Quantum mechanics supplies the compatibility hypergraph. The theorem proves that it has no global deterministic noncontextual section.

Why dimension three is the threshold

For a qubit,

$$P_{\boldsymbol n} = \frac12 \left(I+\boldsymbol n\cdot\boldsymbol\sigma\right),$$

and every rank-one context is the antipodal pair

$$\{P_{\boldsymbol n},P_{-\boldsymbol n}\}.$$

Choose a subset $H$ of the Bloch sphere containing exactly one vector from each antipodal pair and define

$$v(P_{\boldsymbol n}) = \begin{cases} 1,&\boldsymbol n\in H,\\ 0,&-\boldsymbol n\in H. \end{cases}$$

Then

$$v(P_{\boldsymbol n})+v(P_{-\boldsymbol n})=1$$

for every projective qubit measurement. The selection can be discontinuous; continuity is not among the Kochen–Specker assumptions. This explicit coloring shows why the theorem cannot simply omit its dimensional hypothesis.

In dimension three, orthogonality connects triples of rays in a sufficiently rich network to make a global coloring impossible. The same obstruction holds in every dimension greater than three.

The qubit exception concerns rank-one projective measurements only. It does not guarantee a generalized-noncontextual model for arbitrary qubit preparations, POVMs, or transformations. Nor does it apply to two qubits, whose joint Hilbert space has dimension four.

Gleason’s theorem as a corollary route

Suppose the rank-one zero-one valuation in the theorem existed on a complex Hilbert space of dimension at least three. Define a function on unit vectors by

$$f(\psi) = v\!\left(|\psi\rangle\langle\psi|\right).$$

It is nonnegative and has weight one on every orthonormal basis:

$$\sum_{i=1}^{d}f(\psi_i)=1.$$

The frame-function form of Gleason’s theorem therefore gives a density operator $\rho$ satisfying

$$v\!\left(|\psi\rangle\langle\psi|\right) = \langle\psi|\rho|\psi\rangle = \operatorname{Tr} \left( \rho|\psi\rangle\langle\psi| \right)$$

for every unit vector $|\psi\rangle$. In the equivalent formulation on all projectors, this is the unique normalized additive measure $\mu(P)=\operatorname{Tr}(\rho P)$.

No density operator has only zero-one probabilities on every rank-one projector. If $\rho$ has an eigenvalue strictly between zero and one, its eigenprojector already gives an intermediate value. If $\rho$ is pure, $\rho=|\psi\rangle\langle\psi|$, choose a normalized $|\phi\rangle$ that is neither parallel nor orthogonal to $|\psi\rangle$. Then

$$\operatorname{Tr} \left( \rho|\phi\rangle\langle\phi| \right) = |\langle\phi|\psi\rangle|^2 \in(0,1).$$

Either way, the assumed zero-one valuation is impossible.

This is an efficient corollary, not a finite Kochen–Specker construction. Gleason’s theorem assumes an assignment on the full projection lattice and proves its trace form. A finite uncolorable set instead provides a bounded logical witness involving only finitely many contexts.

The Peres–Mermin square

A compact state-independent parity proof lives on two qubits. Let $X$, $Y$, and $Z$ be the Pauli matrices. Arrange nine observables as

$$\begin{array}{ccc} X\otimes I & I\otimes X & X\otimes X\\ I\otimes Y & Y\otimes I & Y\otimes Y\\ X\otimes Y & Y\otimes X & Z\otimes Z \end{array}.$$

The three observables in each row commute, and the three in each column commute. Direct Pauli multiplication gives

$$R_1=R_2=R_3=+I, \qquad C_1=C_2=+I, \qquad C_3=-I,$$

where $R_i$ and $C_j$ denote the operator products in a row or column. For example,

$$\begin{aligned} (X\otimes Y)(Y\otimes X)(Z\otimes Z) &= (XYZ)\otimes(YXZ)\\ &=I, \end{aligned}$$

whereas

$$(X\otimes X)(Y\otimes Y)(Z\otimes Z) = (XYZ)\otimes(XYZ) = -I.$$

Assume a deterministic noncontextual value

$$a_{ij}\in\{+1,-1\}$$

for each cell, with functional consistency inside every commuting context. The row products require

$$\prod_{j=1}^3a_{ij}=+1$$

for $i=1,2,3$. Multiplying those equations gives

$$\prod_{i,j}a_{ij}=+1.$$

The first two column products are $+1$ and the last is $-1$, so multiplying the column equations gives

$$\prod_{i,j}a_{ij}=-1.$$

Both left-hand sides contain every assigned value exactly once, so they cannot differ. This parity contradiction excludes the proposed valuation without using a quantum state.

Each $\pm1$ observable $A$ has spectral projectors

$$P_A^\pm=\frac{I\pm A}{2},$$

so the square can be translated into projector-valued contexts. It is a Kochen–Specker-type witness in dimension four using degenerate observables, not the original three-dimensional ray set. Its virtue is that the entire contradiction fits into six commuting contexts.

State independence and probability

The theorem is state independent because no state or Born probability appears in the coloring contradiction. It says that the same ideal sharp-measurement structure cannot be endowed with deterministic noncontextual values for any underlying state.

State independence does not mean that every experimental statistic violates the same inequality by the same amount. An experiment observes frequencies, not a literal coloring. Turning the logical proof into a noise-robust test requires a noncontextuality inequality or another statistical witness, a specified preparation, and an account of imperfect measurements.

The Gleason route uses probabilities, but only as a proof strategy: a zero-one valuation would be a special additive probability measure, and Gleason’s theorem shows that no such measure exists. The Kochen–Specker theorem does not derive the Born rule or measurement dynamics.

Experimental scope

An ideal proof identifies the same observable across different compatible contexts and assumes the stated operator products exactly. Real devices introduce three separate questions:

1. How is the shared measurement event operationally identified across implementations?
2. How closely are nominally compatible measurements jointly measurable or nondisturbing?
3. How do finite precision, loss, and statistical uncertainty modify the noncontextual bound?

A laboratory claim must answer these questions in its model and error analysis. Substituting nearby directions into an exact coloring proof is not by itself a robust experiment. Generalized operational contextuality provides a framework for replacing ideal operator identity by tested operational equivalences; that framework carries additional assumptions and should not be silently conflated with the original theorem.

What the theorem proves

• There is no global zero-one valuation on all rank-one projectors when $\dim\mathcal H\geq3$ that assigns exactly one value one to every orthonormal basis.
• Equivalently, sharp observables cannot all possess predetermined, measurement-noncontextual values that preserve their functional relations.
• Finite sets of contexts can witness the inconsistency; the Peres–Mermin square gives a compact dimension-four parity witness.
• The contradiction is state independent and requires no spacelike separated parties.

What the theorem does not prove

• It does not rule out contextual hidden-variable theories.
• It does not establish Bell local causality or its failure; locality is not a premise of this theorem.
• It does not produce a projective contradiction for a single qubit.
• It does not by itself cover stochastic generalized-noncontextual models, POVMs, preparations, or transformations.
• It does not say that experimental imperfections are irrelevant.
• It does not select a unique interpretation or explain the physical mechanism of an individual outcome.

Common pitfalls

Dropping the shared-value premise. Independent one-hot assignments to each basis are easy. Noncontextuality is what identifies a repeated projector across bases.

Calling noncommutativity contextuality. Every context in a Kochen–Specker proof consists of compatible observables. The contradiction is global across overlapping compatible contexts.

Applying the theorem to a single qubit. The rank-one projective theorem starts at dimension three. Qubit generalized-contextuality results use a broader premise set.

Presenting the Peres–Mermin square as the original qutrit proof. The square acts on two qubits, hence dimension four, and uses degenerate $\pm1$ observables.

Treating an exact proof as a finished experiment. A robust test needs a statistical inequality and explicit bounds on compatibility and operational equivalence errors.

Exercises

1. Projector values from functional consistency

Let $P$ be a projector and assume a valuation preserves polynomial functions. Show that $v(P)\in\{0,1\}$. If $\sum_iP_i=I$ is a complete orthogonal family, explain why exactly one $P_i$ has value one.

Solution

Because $P^2=P$, functional consistency gives

$$v(P)^2=v(P).$$

The only real solutions are zero and one. For a complete projective measurement, additivity on the commuting orthogonal family gives

$$\sum_i v(P_i)=v(I)=1.$$

Since every term is zero or one, exactly one term equals one.

2. Complete the Gleason contradiction

Suppose Gleason’s theorem gives $v(P)=\operatorname{Tr}(\rho P)$ and all rank-one values are zero or one. Show directly that neither a mixed nor a pure $\rho$ can satisfy this condition.

Solution

If $\rho$ is mixed, at least one eigenvalue lies strictly between zero and one. For its rank-one eigenprojector $P_j$,

$$v(P_j)=\operatorname{Tr}(\rho P_j)$$

equals that intermediate eigenvalue, a contradiction.

If $\rho=|\psi\rangle\langle\psi|$ is pure, select a normalized $|\phi\rangle$ with

$$0<|\langle\phi|\psi\rangle|^2<1.$$

Then the rank-one projector $P_\phi$ has

$$v(P_\phi)=\operatorname{Tr}(\rho P_\phi) =|\langle\phi|\psi\rangle|^2,$$

again neither zero nor one.

3. Audit the Peres–Mermin signs

Verify the products of the third row and third column of the square. Use $XY=iZ$, $YX=-iZ$, and $XYZ=iI$.

Solution

For the third row,

$$\begin{aligned} (X\otimes Y)(Y\otimes X)(Z\otimes Z) &= (XY Z)\otimes(YX Z)\\ &= (iI)\otimes(-iI)\\ &=+I. \end{aligned}$$

For the third column,

$$\begin{aligned} (X\otimes X)(Y\otimes Y)(Z\otimes Z) &= (XYZ)\otimes(XYZ)\\ &= (iI)\otimes(iI)\\ &=-I. \end{aligned}$$

The single negative context product makes the parity contradiction possible.

4. Construct an explicit qubit coloring

Define $H$ by the lexicographic rule: $\boldsymbol n\in H$ when the first nonzero component in the ordered list $(n_z,n_x,n_y)$ is positive. Prove that $H$ contains exactly one vector from each antipodal pair.

Solution

Every unit vector has a first nonzero component in the stated ordered list. Changing $\boldsymbol n$ to $-\boldsymbol n$ reverses the sign of that component. Therefore exactly one of $\boldsymbol n$ and $-\boldsymbol n$ satisfies the positivity rule. Assigning value one to $P_{\boldsymbol n}$ for $\boldsymbol n\in H$ and zero to its antipode gives

$$v(P_{\boldsymbol n})+v(P_{-\boldsymbol n})=1$$

for every rank-one qubit context.

References

• J. S. Bell, “On the Problem of Hidden Variables in Quantum Mechanics,” Reviews of Modern Physics 38, 447–452, 1966, doi:10.1103/RevModPhys.38.447.
• A. Cabello, J. M. Estebaranz, and G. García-Alcaine, “Bell–Kochen–Specker Theorem: A Proof with 18 Vectors,” Physics Letters A 212, 183–187, 1996, doi:10.1016/0375-9601(96)00134-X.
• A. M. Gleason, “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.
• S. Kochen and E. P. Specker, “The Problem of Hidden Variables in Quantum Mechanics,” Indiana University Mathematics Journal 17, 59–87, 1968, doi:10.1512/iumj.1968.17.17004.
• N. D. Mermin, “Simple Unified Form for the Major No-Hidden-Variables Theorems,” Physical Review Letters 65, 3373–3376, 1990, doi:10.1103/PhysRevLett.65.3373.
• A. Peres, “Two Simple Proofs of the Kochen–Specker Theorem,” Journal of Physics A: Mathematical and General 24, L175–L178, 1991, doi:10.1088/0305-4470/24/4/003.