Measurement-Induced Energy, Noise, and Backreaction

The energy cost and disturbance of a measurement belong to its physical instrument, not to its POVM alone. Even when two implementations have exactly the same outcome probabilities, they can require different control work, excite different probe degrees of freedom, and leave different field states. This page gives an exact validation check for one regulated smeared-field mode and then states what must be added before the result becomes a continuum-QFT claim. It does not derive an implementation-independent localization bound.

Required background. Local measurement instruments separates effects from update maps. Switching and smearing controls the ultraviolet and temporal profile of the coupling.

Chapter map. Use the route from a local coupling to a field instrument, the chapter claim-validity table, and the three independent validity questions to place this construction within the full measurement protocol.

Work supplied by a driven coupling

Let a field, a probe, and a classical control be described during the measurement by

$$H(t)=H_F+H_P+H_I(t), \qquad H_I(t)=\lambda\chi(t)\,Q_F\otimes P_P,$$

where $Q_F=\Phi(F)$ is a smeared field quadrature and $\chi$ vanishes before and after the interaction. The explicit time dependence represents the work source. For unitary evolution,

$$\frac{d}{dt}\langle H(t)\rangle =\left\langle\frac{\partial H_I(t)}{\partial t}\right\rangle, \qquad P_{\mathrm{ctrl}}(t) =\lambda\dot\chi(t)\langle Q_F\otimes P_P\rangle_t.$$

Because the interaction energy is zero at both endpoints,

$$W_{\mathrm{ctrl}} =\int dt\,P_{\mathrm{ctrl}}(t) =\Delta E_F+\Delta E_P.$$

This equality is exact for the closed field–probe system. It does not include later amplification, record storage, erasure, or probe reset. In a perturbative calculation every term in this balance must be expanded to the same order; a field-energy change computed to $O(\lambda^2)$ cannot be compared with a probe-energy change kept only to $O(\lambda)$.

Exact smeared-mode benchmark

Choose one normal mode of a regulated free scalar field, or equivalently declare an exact invariant one-mode truncation, with dimensionless quadratures $q,p$ satisfying $[q,p]=i$ and

$$H_F=\frac{\omega_F}{2}(q^2+p^2).$$

The probe has $[Q,P]=i$ and $H_P=\omega_P(Q^2+P^2)/2$. During a pulse short compared with both free periods, take the exactly soluble coupling

$$U=e^{-igqP}.$$

Prepare the field mode in its vacuum and the probe in a centered minimum-uncertainty Gaussian state,

$$\operatorname{Var}(q)=\operatorname{Var}(p)=\frac12, \qquad \operatorname{Var}(Q)=\sigma^2, \qquad \operatorname{Var}(P)=\frac{1}{4\sigma^2}.$$

The Heisenberg input–output relations are

$$q_{\mathrm{out}}=q, \quad p_{\mathrm{out}}=p-gP, \quad Q_{\mathrm{out}}=Q+gq, \quad P_{\mathrm{out}}=P.$$

Reading $m=Q_{\mathrm{out}}/g$ therefore measures $q$ with additive Gaussian noise. With the standard root-mean-square definitions of Ozawa 2003, Eqs. (7)–(10),

$$\epsilon^2(q) =\left\langle(m-q)^2\right\rangle =\frac{\sigma^2}{g^2}, \qquad \eta^2(p) =\left\langle(p_{\mathrm{out}}-p)^2\right\rangle =\frac{g^2}{4\sigma^2}.$$

This dilation saturates $\epsilon^2(q)\eta^2(p)=1/4$. That equality characterizes this independent-intervention model; it is not a universal formula for arbitrary instruments. Ozawa’s explicit quadrature amplifier gives the closely related input–output construction in Eqs. (17)–(19).

The field and probe energy changes follow from the same four relations:

$$\Delta E_F=\frac{\omega_Fg^2}{8\sigma^2}, \qquad \Delta E_P=\frac{\omega_Pg^2}{4}, \qquad W_{\mathrm{ctrl}}=\Delta E_F+\Delta E_P.$$

For the reproducible parameter choice

$$\omega_F=2, \qquad \omega_P=1, \qquad g=\frac12, \qquad \sigma=1,$$

the complete account is

Quantity	Exact value	Decimal value
Readout-noise variance $\epsilon^2(q)$	$4$	$4.0000$
Momentum-disturbance variance $\eta^2(p)$	$1/16$	$0.0625$
Field-energy change $\Delta E_F$	$1/16$	$0.0625$
Probe-energy change $\Delta E_P$	$1/16$	$0.0625$
Control work $W_{\mathrm{ctrl}}$	$1/8$	$0.1250$

Thus the work balance closes exactly, while the readout distribution is the sharp $q$ distribution convolved with a Gaussian of variance $4$. The benchmark is deliberately simple enough that a numerical implementation can be checked against every entry rather than only against its final probability distribution.

Same POVM, different backreaction

Let $M_m$ denote the Kraus-density realization of the preceding Gaussian readout. Append a field kick

$$V_\kappa=e^{-i\kappa q}, \qquad \widetilde M_m=V_\kappa M_m.$$

The effect density is unchanged,

$$\widetilde M_m^\dagger\widetilde M_m=M_m^\dagger M_m,$$

so every immediate outcome probability is identical. Nevertheless $V_\kappa^\dagger pV_\kappa=p-\kappa$. For the centered benchmark state, the kick adds

$$\Delta E_F^{(\mathrm{kick})}=\frac{\omega_F\kappa^2}{2}.$$

Taking $\kappa=1/2$ adds $1/4$ unit of field energy: the total field-energy change becomes $5/16$ and the total control work becomes $3/8$, while the POVM is still exactly the same. This is the explicit adversarial control. It proves that no unique energetic or disturbance cost follows from the POVM alone without specifying the update map and its physical dilation.

In continuum QFT, the corresponding nonselective channel is tested through

$$\Delta_{\mathcal E}B=\mathcal E^*(B)-B.$$

For $B$ in the causal complement of a genuinely supported coupling, this must vanish. The system–probe construction gives the nonselective update and its causal-complement action in Fewster and Verch 2020, § 3.3, Eq. (3.23). A feedback operation must pass that localization test independently; multiplying a Kraus operator by a unitary does not make the unitary local.

From the benchmark to continuum validation

The impulsive unitary isolates the algebra of noise and energy accounting, but it has unlimited frequency bandwidth and does not certify a compactly supported continuum coupling. A physical validation must replace it by a smooth, compactly supported switching family—or by another declared class with sufficient decay—and report the deviation from the short-pulse formulas. For compact switching, smoothness makes the detector response a well-defined distributional pairing, as shown in Satz 2007, § 2, Eqs. (2.2)–(2.5), and § 3; spatial smearing and its zero-size limit require a separate analysis, developed in Louko and Satz 2006, §§ 2–4.

A reproducible apparatus record should therefore state:

• the field and probe Hamiltonians, their zero points, and any renormalized or vacuum-subtracted field-energy observable;
• the initial joint state and all cross-correlations used in the energy balance;
• switching, smearing, support or tail tolerance, and the order of pointlike and sudden limits;
• the probe readout, estimator calibration, and separate noise and disturbance observables;
• the perturbative remainder and independent numerical integration error;
• the work source, including feedback, reset, amplification, and record costs when those are part of the claim.

Evidence boundary and limitations

Scientific evidence boundary, checked 2026-08-26. The cited results support apparatus-specific noise definitions, smooth/smeared response calculations, and local probe-induced operations. They do not establish a universal lower bound on the energy cost of every local QFT measurement. The exact numbers above validate one regulated mode, one pulse approximation, and one probe state. Cross-protocol lower bounds require an optimization over admissible instruments and belong to the later energy–information treatment.

Common pitfalls

Calling field-energy gain the work cost. The probe can lose or gain energy, and an explicit control supplies the remainder. Check $W_{\mathrm{ctrl}}=\Delta E_F+\Delta E_P$ before attaching a thermodynamic interpretation.

Inferring disturbance from resolution. The POVM fixes the readout distribution, not the update. The $V_\kappa$ control has identical resolution and a different energy injection.

Taking a short pulse literally. The exact benchmark is a regulator-level unit test. A causal continuum claim needs a smooth, supported coupling and an error bound connecting it to the pulse model.

Exercises

Derive all four input–output relations for $U=e^{-igqP}$ using the Baker–Campbell–Hausdorff series, and explain why the series terminates.

Solution

For example,

$$U^\dagger pU=p+[igqP,p]=p-gP,$$

because $[qP,p]=iP$ and the next commutator vanishes. Likewise $[igqP,Q]=gq$, while $q$ and $P$ commute with the generator. Hence $q\mapsto q$, $p\mapsto p-gP$, $Q\mapsto Q+gq$, and $P\mapsto P$. The generator is bilinear but each transformed quadrature acquires only a commuting quadrature, so no higher nested commutators survive.

Using the benchmark parameters, recompute $\epsilon^2(q)$, $\eta^2(p)$, $\Delta E_F$, $\Delta E_P$, and the work balance without using the table.

Solution

The probe variances are $\operatorname{Var}(Q)=1$ and $\operatorname{Var}(P)=1/4$. Therefore

$$\epsilon^2=\frac{1}{(1/2)^2}=4, \qquad \eta^2=\left(\frac12\right)^2\frac14=\frac1{16}.$$

The field energy rises by $(\omega_F/2)\eta^2=(2/2)(1/16)=1/16$. The probe pointer is shifted by $gq$, so its energy rises by $(\omega_P/2)g^2\operatorname{Var}(q)=(1/2)(1/4)(1/2)=1/16$. Their sum is $1/8$, equal to the control work.

Show that the kick $V_\kappa$ leaves the POVM fixed and find the total work for $\kappa=1/2$.

Solution

Unitarity gives $(V_\kappa M_m)^\dagger(V_\kappa M_m)=M_m^\dagger M_m$, so the effect density and all current outcome probabilities are unchanged. The kick shifts $p$ by $-\kappa$ and, in the centered state, adds $\omega_F\kappa^2/2=2(1/4)/2=1/4$ to the field energy. Adding this to the baseline work $1/8$ gives $3/8$.

References

• Fewster, C. J., and Verch, R. (2020). “Quantum Fields and Local Measurements.” Communications in Mathematical Physics 378, 851–889. DOI. Open PDF.
• Louko, J., and Satz, A. (2006). “How Often Does the Unruh–DeWitt Detector Click? Regularisation by a Spatial Profile.” Classical and Quantum Gravity 23, 6321–6344. DOI. Open PDF.
• Ozawa, M. (2003). “Universally Valid Reformulation of the Heisenberg Uncertainty Principle on Noise and Disturbance in Measurement.” Physical Review A 67, 042105. DOI. Open PDF.
• Satz, A. (2007). “Then Again, How Often Does the Unruh–DeWitt Detector Click If We Switch It Carefully?” Classical and Quantum Gravity 24, 1719–1732. DOI. Open PDF.