---
title: Wigner–Araki–Yanase Theorem Overview
url: https://www.emergentmind.com/topics/wigner-araki-yanase-theorem
type: topic
---

# Wigner–Araki–Yanase Theorem Overview

The Wigner–Araki–Yanase theorem is a limitation theorem for quantum measurement under additive conservation laws. In its standard form, a system observable \(M\) is measured by coupling the system to an apparatus with a unitary \(U\), while a conserved quantity of the composite has the additive form \(L_{\rm total}=L_1\otimes 1+1\otimes L_2\) and satisfies \([U,L_{\rm total}]=0\). The theorem states that if the measurement is exact and either repeatable or subject to the Yanase condition, then the measured observable must commute with the system part of the conserved quantity, \([M,L_1]=0\). Quantitative, information-theoretic, and resource-theoretic reformulations replace this exact no-go statement by lower bounds on measurement error, disturbance, or implementation cost, and extend the theorem to continuous observables, unitary channels, and general quantum resource theories [1012.4362][2208.13494][2507.23760].

## 1. Classical statement and measurement model

The standard formulation uses a von Neumann measurement model with a system Hilbert space and an apparatus Hilbert space, a target observable on the system, a pointer observable on the apparatus, an initial apparatus “ready” state, and a unitary interaction. In one common notation, the measured observable has spectral decomposition
\[
S=\sum_i \lambda_i\,Q_i,
\]
the pointer observable is
\[
Z=\sum_i z_i\,P_i,
\]
and the interaction is chosen so that for a fixed initial apparatus state \(\xi\),
\[
U(\varphi_i\otimes \xi)=\varphi_i\otimes X_i,
\qquad
Q_i\varphi_j=\delta_{ij}\varphi_j,\;\;P_iX_j=\delta_{ij}X_j.
\]
The conservation law is imposed through an additive conserved quantity,
\[
H_{\rm tot}=H_S\otimes I_A+I_S\otimes H_A,
\qquad
[U,H_{\rm tot}]=0.
\]
Under these assumptions, the Araki–Yanase formulation yields the exact constraint
\[
[Q_i,H_S]=0\quad \forall i,
\]
equivalently \([S,H_S]=0\): only observables commuting with the system Hamiltonian admit an exact energy-conserving von Neumann measurement [1610.04607].

A closely related statement appears in the bounded-operator formulation. Let \(M\) be a discrete-spectrum self-adjoint operator on the object Hilbert space \(H\), let \(L_1,L_2\) be bounded self-adjoint operators on object and probe, and let the interaction satisfy
\[
[U,L_1\otimes 1+1\otimes L_2]=0.
\]
If the measurement is repeatable, or if the pointer satisfies the Yanase condition \([Z,L_2]=0\), then
\[
[M,L_1]=0.
\]
This is the classical WAY theorem in the Araki–Yanase sense [1012.4362].

Wigner’s original 1952 spin model already exhibited the essential obstruction. For a spin-\(\tfrac12\) system, Wigner considered measurement of \(S_x\) under conservation of total \(S_z\otimes 1+1\otimes J_z\). If the apparatus state is a superposition
\[
\phi=\sum_{\nu=-n+1}^{n-1}c_\nu\,|\nu\rangle_z,
\]
then the probability \(\eta\) of an inconclusive outcome satisfies
\[
\eta^2=\frac{1}{2n-1},
\]
so \(\eta\to 0\) only as \(n\to\infty\). In that model, accurate and repeatable measurement of an observable not commuting with the conserved quantity demands an apparatus with unbounded spread in the apparatus contribution to that conserved quantity [1012.4362][1012.4372].

## 2. Structural assumptions, repeatability, and the Yanase condition

The theorem depends on a specific conjunction of assumptions. The measurement scheme is required to satisfy probability reproducibility, meaning that pointer statistics after the interaction reproduce the statistics of the target observable, and in the original idealized form it is also taken to be repeatable: an immediate repetition yields the same outcome with certainty [1012.4362]. In this setting, additive conservation and exact measurement are not by themselves the whole story; the role of the pointer and of repeatability is central.

Yanase emphasized that the pointer readout should itself be compatible with the apparatus part of the conserved quantity,
\[
[Z,H_A]=0
\quad\Longleftrightarrow\quad
[P_i,H_A]=0\quad\forall i,
\]
so that reading out the pointer does not itself violate the same conservation law. This requirement is the Yanase condition. In the Araki–Yanase theorem, either repeatability or the Yanase condition suffices to derive the commutativity restriction on the system observable [1610.04607][1012.4362].

A notable refinement is due to Parrott. Under the simplifying assumption that the measured observable has discrete and non-degenerate eigenvalues, the hypotheses of the original WAY theorem imply that the apparatus “ready” state is an eigenstate of \(H_A\), that \(H_A\) is diagonal in the pointer basis, and finally that
\[
H_A=a\,I_A.
\]
In that analysis, the Yanase condition becomes trivial because \([P_i,H_A]=0\) holds automatically when \(H_A\propto I_A\) [1610.04607]. This sharpens the sense in which the exact theorem rests on highly idealized premises: strict additivity of the conserved quantity and exact von Neumann coupling. The same analysis argues that relaxing strict additivity, for example by allowing interaction terms in the total Hamiltonian, or relaxing exact ideal coupling, restores the possibility of measuring observables that do not commute with subsystem energy, while total energy remains conserved at the full closed-system level [1610.04607].

This point addresses a recurrent misconception. The theorem does not assert that quantum measurements in general must conserve the system’s energy or that measurements of noncommuting observables are physically impossible. Rather, it states that under a specific idealized model—exact, additive, and repeatable or Yanase-compatible measurement—such observables cannot be measured sharply [1012.4362][1610.04607].

## 3. Quantitative WAY bounds: error, disturbance, and apparatus spread

The qualitative no-go theorem evolved into quantitative trade-off relations. In the Heisenberg picture, one defines the noise operator
\[
N=Z(\tau)-M\otimes 1,
\]
where \(Z(\tau)=U^*(1\otimes Z)U\), and the state-dependent error
\[
\varepsilon(\varphi)^2=\langle \varphi\otimes \phi|N^2|\varphi\otimes \phi\rangle.
\]
Using Robertson-type bounds together with \([U,L_{\rm total}]=0\), one obtains
\[
\varepsilon^2\ge
\frac{\|[M\otimes 1,L_{\rm total}]\|^2}
{4\bigl((\Delta_\phi L_2)^2+(\Delta_\varphi L_1)^2\bigr)}.
\]
If the Yanase condition holds, this reduces to
\[
\varepsilon
\ge
\frac{|\langle \varphi|[M,L_1]|\varphi\rangle|}{2\,\Delta_\phi L_2}
\;\Rightarrow\;
\varepsilon\ge \frac{\|[M,L_1]\|}{2\,\Delta_\phi L_2}.
\]
An analogous inequality holds for approximate repeatability. Thus the error can be made small only if the apparatus variance in the conserved quantity becomes large [1012.4362].

In the energy language often associated with Ozawa-type bounds, the same principle appears as
\[
\varepsilon\;\ge\;\frac{\|[A,H_S]\|}{2\,\sqrt{\langle\eta|H_A^2|\eta\rangle}},
\]
or equivalently
\[
\Delta H_A\;\gtrsim\;\frac{\|[A,H_S]\|}{2\,\varepsilon},
\]
so measurement of an observable that does not commute with the conserved Hamiltonian requires large apparatus fluctuation in the corresponding conserved quantity [2507.23760].

A later unification due to Emori and Tajima formulates both error and disturbance as special cases of irreversibility. For a CPTP map \(L\) and a test ensemble \(\Omega=\{p_k,\rho_k\}\), irreversibility is defined as
\[
\delta(L,\Omega):=
\min_R
\sqrt{\sum_k p_k\,D_F\bigl(\rho_k,R\circ L(\rho_k)\bigr)^2},
\]
with \(D_F(\sigma,\tau)=\sqrt{1-F(\sigma,\tau)^2}\). They then define general error and disturbance by small-\(\theta\) limits of irreversibility for suitable “loss” channels. Under an additive conservation law, they derive universal inequalities
\[
\varepsilon(\rho,A,M)\ge
\frac{|\mathrm{Tr}([Y_S,A]\rho)|}{\sqrt{F^{\rm cost}_{P_M}+\Delta_F}},
\qquad
\eta(\rho,B,M)\ge
\frac{|\mathrm{Tr}([Y'_S,B]\rho)|}{\sqrt{F^{\rm cost}_{I_M}+\Delta'_F}},
\]
where \(F^{\rm cost}\) is a minimal quantum-Fisher-information cost under the conservation law. In this form, the WAY restriction is no longer tied to a single error-operator formalism; it applies to any error or disturbance notion representable as irreversibility of an associated loss channel followed by optimal recovery [2309.14172].

## 4. Information-theoretic and resource-theoretic reformulations

A major conceptual reformulation interprets the WAY theorem within the resource theory of asymmetry. Relative to a symmetry group \(G\), symmetric states, observables, and channels are those invariant under the corresponding group action, while asymmetric ones are resources. In this language, the problem becomes: can one simulate the measurement of an asymmetric observable using only \(G\)-covariant operations and an asymmetric resource state? The information-theoretic answer is that exact simulation is possible if and only if the resource state is perfectly asymmetric, meaning that its orbit
\[
\{U(g)\sigma U(g)^\dagger:g\in G\}
\]
consists of mutually orthogonal states [1212.3378].

This recasts the WAY theorem as a consequence of the no-programming theorem for projective measurements. If a family of distinct projective measurements is to be programmed into a fixed device, then the program states must be perfectly distinguishable. Applied to symmetry orbits, exact implementation of an asymmetric projective measurement by symmetric processing requires a resource state whose group orbit is orthogonal. A direct corollary is that if \(G\) is infinite and the apparatus Hilbert space is finite-dimensional, then no state can have an orthogonal orbit, so exact simulation of a genuinely asymmetric projective measurement is impossible [1212.3378].

The 2025 general-resource formulation pushes this logic beyond asymmetry to arbitrary resource theories. A resource monotone \(M(\rho)\) is assumed to satisfy monotonicity under free unitaries and partial trace, additivity on product states, and a mild continuity bound. For a target channel \(\Lambda\), the \(\epsilon\)-implementation cost is defined by
\[
M_c^\epsilon(\Lambda):=
\min\Bigl\{
M(\eta)\;\Bigm|\;
\exists\,U\in\mathfrak U_F,\;
\Lambda(\rho)\approx_\epsilon
\mathrm{Tr}_A[U(\rho\otimes\eta)U^\dagger]
\Bigr\},
\]
and irreversibility is quantified on a two-state ensemble \(\Omega_p=(\rho_1,\rho_2)_p\), \(F(\rho_1,\rho_2)=0\), by
\[
\delta(\Lambda,\Omega_p)^2
:=
\min_{\mathcal R}
\sum_{k=1}^2
p_k\,D_F\bigl(\rho_k,\mathcal R\circ\Lambda(\rho_k)\bigr)^2.
\]
The resulting universal resource–irreversibility trade-off is
\[
M_c(\Lambda)\;\ge\;
\frac{\bigl|(\Omega_p\mid \mathscr K)-(\Omega'_{p,\Lambda}\mid \mathscr K)\bigr|^2}
{16\,K_S\,\delta(\Lambda,\Omega_p)}
-c',
\]
with \(K_S\) the continuity constant and \(c'\) a small offset [2507.23760].

For indirect measurement, this yields the general-resource WAY bound
\[
M(\eta)\;\ge\;
\max_p\;
\frac{(\Omega_p\mid\mathscr K)^2}{16\,K_S\,\epsilon}
-O(1).
\]
Whenever \((\Omega_p\mid\mathscr K)>0\), perfect discrimination \((\epsilon\to 0)\) requires \(M(\eta)\to\infty\). The theorem thereby extends the WAY principle from energy to asymmetry, coherence, magic, athermality, and other resources. The cited examples include energy cost \(E(\rho)=\mathrm{Tr}[H\rho]\), asymmetry cost \(M(\rho)=\tfrac14\mathcal F_H(\rho)\), and coherence cost given by relative entropy of coherence, with the common asymptotic feature that the required resource grows as \(1/\epsilon\) for finite-error implementation [2507.23760].

## 5. Continuous observables, unbounded conserved quantities, and unitary channels

Classical proofs of the WAY theorem were restricted to bounded or discrete-spectrum conserved observables. Kuramochi and Tajima extend the theorem to possibly unbounded and continuous conserved observables by working with the exponentiated one-parameter unitary groups \(U_t=e^{itL}\) rather than directly with the generators. In a general measurement model with probe state \(\rho_P\), coupling \(U\), and output-probe POVM \(F_{P'}(X)\), suppose the additive conservation law
\[
U^\dagger\bigl(L_{S'}\otimes 1_{P'}+1_{S'}\otimes L_{P'}\bigr)U
=
L_S\otimes 1_P+1_S\otimes L_P
\]
holds and the Yanase condition
\[
[F_{P'}(X),L_{P'}]=0
\quad\forall X
\]
is imposed. Then any implemented system PVM \(E_S(X)\) must satisfy
\[
[E_S(X),L_S]=0
\quad\forall X.
\]
This yields a WAY theorem for continuous and unbounded observables under the Yanase condition [2208.13494].

Two concrete impossibility statements follow. First, exact projective measurement of position under momentum conservation is impossible when the probe readout commutes with probe momentum. Second, exact projective measurement of quadrature amplitude using linear optical instruments and photon counters is impossible under total photon-number conservation, because the quadrature does not commute with number while the readout satisfies the Yanase condition [2208.13494].

The same work analyzes unitary channels. If a target unitary channel \(\rho_S\mapsto U_S\rho_S U_S^\dagger\) is implemented under the same additive conservation law, then
\[
U_S^\dagger L'_S U_S=L_S+\gamma\,1_S
\]
for some real \(\gamma\). When \(H'_S=H_S\), \(L'_S=L_S\), and \(L_S\) is semi-bounded above or below, spectral comparison forces \(\gamma=0\), hence
\[
[U_S,L_S]=0.
\]
If instead the spectrum of \(L_S\) is upper and lower unbounded, a nonzero constant shift can occur. The momentum operator provides explicit examples, including \(U_S=e^{i\gamma \hat x_S}\), for which
\[
U_S^\dagger \hat p_S U_S=\hat p_S+\gamma\,1.
\]
This identifies the precise sense in which unitary implementations are restricted by the same WAY logic, and where two-sided unbounded spectra open a controlled exception [2208.13494].

## 6. Interpretations, explicit models, and modern significance

A relational interpretation emphasizes that, in the presence of symmetry, only invariant observables are fundamentally measurable. On this view, non-invariant “absolute” quantities are shorthand for invariant relative ones, and the apparatus serves a dual role: it is both the statistical probe and the physical reference system relative to which the measured quantity is defined. Formally, a relativisation map sends an operator \(A\) on the system to an invariant operator on system plus reference,
\[
A\mapsto
\int_G
\bigl[U_s(g)AU_s(g)^*\bigr]\otimes F_r(dg),
\]
and restricting that invariant observable with a sharply localized reference state yields an effective non-invariant observable on the system alone. In this perspective, the requirement of large apparatus spread in the conserved quantity is reinterpreted as the requirement that the apparatus function as a sufficiently good quantum reference frame [2006.07047].

An explicit modern spin-conserving model makes the same point in channel form. In the 2026 analysis of angular-momentum-conserving measurement, the induced system channel can be written with Kraus operators
\[
M_0=\sqrt{\beta_z}\,I_2,\qquad
M_1=\sqrt{1-\beta_z}\,|{\uparrow}\rangle_z\langle{\uparrow}|,\qquad
M_2=\sqrt{1-\beta_z}\,|{\downarrow}\rangle_z\langle{\downarrow}|,
\]
so that an arbitrary qubit state undergoes dephasing in the \(z\)-basis:
\[
\mathcal E(\rho)=
\begin{pmatrix}
\rho_{\uparrow\uparrow} & \beta_z\,\rho_{\uparrow\downarrow}\\
\beta_z\,\rho_{\downarrow\uparrow} & \rho_{\downarrow\downarrow}
\end{pmatrix}.
\]
If the pointer is read in the \(x\)-basis, the misidentification probability is
\[
P_{\rm error}=\tfrac12(1-\beta_z).
\]
Perfect accuracy requires \(\beta_z=1\), which the paper identifies with an apparatus having infinite \(z\)-spin uncertainty. The model therefore realizes the WAY limitation as an exactly solvable apparatus-size-dependent dephasing effect [2606.02861].

Several broader conclusions recur across the literature. First, the theorem constrains not only measurements but also quantum control and gate implementation under conservation laws [1012.4362]. Second, the modern resource-theoretic form unifies energy–error, angular-momentum–error, coherence–error, magic–error, and work–error trade-offs as instances of a single cost–irreversibility principle [2507.23760]. Third, it is inaccurate to read the theorem as a blanket prohibition on measuring observables that fail to commute with conserved quantities. A more precise statement is that exact, reversible, Yanase-compatible or repeatable implementations are forbidden unless the relevant commutator vanishes, while approximate implementations remain possible at finite error, with required apparatus resource diverging as the error tends to zero [1012.4362][2309.14172][2507.23760].

Source: https://www.emergentmind.com/topics/wigner-araki-yanase-theorem