Common Misconceptions

This page collects recurring confusions that otherwise spread across measurement theory, quantum channels, decoherence, master equations, and trajectories. The common pattern is simple: a correct formula is used outside the modeling assumptions that make it correct.

The safest diagnostic is always:

$$\text{Which degrees of freedom are kept, which are ignored, and which records are conditioned on?}$$

That question prevents most mistakes in this volume.

Quick Corrections

Misconception	Better statement
Measurement requires a human observer.	Measurement theory describes physical correlations between a system, apparatus, environment, and record. A conscious observer is not part of the standard formalism.
Collapse must be a literal dynamical process inside the apparatus.	State update is an operational rule for assigning a post-measurement state, while interpretations differ about what, if anything, physically collapses.
A POVM is just a fuzzy projective measurement.	A POVM gives probabilities for outcomes; many inequivalent instruments can realize the same POVM with different post-measurement states.
A Kraus operator is the same thing as a POVM element.	A Kraus or measurement operator acts on states; a POVM effect is positive and determines probabilities.
Every nonunitary equation is allowed.	Physical state maps must preserve trace and positivity, and usually complete positivity is required for subsystem dynamics.
Decoherence is the same as dissipation.	Decoherence suppresses interference in a basis; dissipation changes energy or populations. They often coexist but are not identical.
Decoherence alone solves the measurement problem.	Decoherence explains suppression of local interference and pointer stability, but it does not by itself select a single outcome in every interpretation.
Markovian dynamics means the environment does not matter.	Markovian dynamics means environmental memory is negligible in the reduced description; the environment is still the source of rates and noise.
A Lindblad equation is always microscopically valid.	Lindblad form guarantees a legitimate Markovian semigroup, but deriving the generator from a Hamiltonian requires approximations.
Quantum trajectories are literal hidden paths.	A trajectory is a conditioned state assignment relative to a measurement record or unraveling. Different unravelings can have the same master equation.
Thermalization is just loss of phase coherence.	Thermalization requires relaxation toward a thermal state, not merely decay of off-diagonal terms.
Noise is always bad.	Noise can destroy coherence, but controlled noise, engineered dissipation, and measurement backaction can prepare states and stabilize dynamics.

Measurement Is Not Human Looking

In the operational formalism, a measurement is specified by possible outcomes and by the physical operation associated with those outcomes. The minimal probability rule for a generalized measurement is

$$p(i)=\operatorname{Tr}(\rho F_i), \qquad F_i\ge0, \qquad \sum_i F_i=I.$$

No term in this formula refers to consciousness. The detector may be a photodiode, a cloud chamber, a superconducting resonator, an ion fluorescence collection system, or an environment that stores which-alternative information. Human reading enters only later, when a record is used.

The stronger point is that “measurement” is not passive inspection of a pre-existing classical value. A measurement setup couples the system to other degrees of freedom. Depending on the setup, it may disturb the state, leave it nearly unchanged, weakly monitor it over time, or correlate it with an apparatus pointer.

Collapse Language Has Limited Scope

For an ideal projective measurement with projectors $\{P_a\}$, the selective Lüders update is

$$\rho \longmapsto \rho_a = \frac{P_a\rho P_a} {\operatorname{Tr}(P_a\rho)}.$$

This is an extremely useful rule. The mistake is to assume that this equation, by itself, is already a complete microscopic account of what happens inside every measuring apparatus.

The update rule answers an operational question: given outcome $a$, which state should be used for later predictions? A microscopic model may instead describe a system-apparatus-environment state evolving unitarily, followed by conditioning on or ignoring a record. Interpretations of quantum mechanics differ on whether the update is physical collapse, information update, branching, effective collapse, or something else. The formal calculations in this volume do not require choosing that interpretive layer unless the question is explicitly foundational.

For boundaries of the postulates, see What the Postulates Do Not Say.

POVMs Do Not Determine State Update

A POVM $\{F_i\}$ determines outcome probabilities. It does not, by itself, determine the conditional output state.

For a measurement described by operators $\{M_i\}$,

$$F_i=M_i^\dagger M_i, \qquad p(i)=\operatorname{Tr}(M_i\rho M_i^\dagger) = \operatorname{Tr}(\rho F_i),$$

and the conditional state is

$$\rho_i = \frac{M_i\rho M_i^\dagger} {\operatorname{Tr}(M_i\rho M_i^\dagger)}.$$

The same effect $F_i$ can be implemented by different $M_i$ up to additional unitary rotations, ancillary systems, or coarse-graining. Therefore the POVM answers “what are the probabilities?” but not “what state is left afterward?”

The more complete object is an instrument: a family of completely positive maps $\{\mathcal I_i\}$ such that

$$p(i)=\operatorname{Tr}\mathcal I_i(\rho), \qquad \rho_i= \frac{\mathcal I_i(\rho)} {\operatorname{Tr}\mathcal I_i(\rho)}.$$

The nonselective state is obtained by summing over outcomes:

$$\rho' = \sum_i\mathcal I_i(\rho).$$

This is why a detector model, not only a POVM, is needed whenever backaction matters.

Kraus Operators Are Not Effects

Kraus operators and POVM effects are related but have different roles.

Object	Symbol	Mathematical type	Role
measurement or Kraus operator	$M_i$ or $K_\alpha$	arbitrary operator subject to normalization constraints	transforms states
POVM effect	$F_i$	positive operator	gives outcome probabilities
channel	$\Phi$	completely positive trace-preserving map	gives unconditional state evolution
instrument element	$\mathcal I_i$	completely positive, trace-nonincreasing map	gives outcome probability and conditional output

For one-outcome labeling, the effect is often

$$F_i=M_i^\dagger M_i.$$

For a channel with Kraus operators $\{K_\alpha\}$,

$$\Phi(\rho) = \sum_\alpha K_\alpha\rho K_\alpha^\dagger, \qquad \sum_\alpha K_\alpha^\dagger K_\alpha=I.$$

The operators $K_\alpha$ are not unique. Replacing one Kraus representation by another can leave the same channel unchanged. Confusing representation-dependent Kraus operators with observable effects is a common source of false physical conclusions.

Nonunitary Does Not Mean Anything Goes

Closed-system dynamics is unitary. Open-system dynamics can be nonunitary for the reduced state, but physical reduced dynamics is still constrained.

A valid density operator must remain positive and normalized:

$$\rho\ge0, \qquad \operatorname{Tr}\rho=1.$$

A state map $\Phi$ that describes unconditional evolution should preserve trace and map density operators to density operators:

$$\operatorname{Tr}\Phi(\rho)=\operatorname{Tr}\rho, \qquad \Phi(\rho)\ge0 \quad \text{when} \quad \rho\ge0.$$

For a subsystem that may be entangled with a reference $R$, positivity of $\Phi$ alone is not enough. One requires complete positivity:

$$(\Phi\otimes I_R)(\rho_{SR})\ge0 \quad \text{for every reference } R \text{ and every } \rho_{SR}\ge0.$$

This condition is not decorative. Without it, a map can appear harmless on isolated system states but produce impossible negative probabilities when applied to part of an entangled state.

Effective non-Hermitian Hamiltonians are a frequent boundary case. They can describe conditional no-jump evolution, absorption out of a modeled subspace, or decay before renormalization. They are not ordinary closed-system Hamiltonians, and their missing probability must be accounted for by jumps, reservoirs, detectors, or omitted channels.

Decoherence Is Not Dissipation

Decoherence is loss of interference in a chosen reduced description. Dissipation is exchange of energy or relaxation of populations. A model can have either one without the other.

Pure dephasing of a qubit in the $\{\lvert0\rangle,\lvert1\rangle\}$ basis has the schematic form

$$\begin{pmatrix} \rho_{00} & \rho_{01} \\ \rho_{10} & \rho_{11} \end{pmatrix} \longmapsto \begin{pmatrix} \rho_{00} & \lambda\rho_{01} \\ \lambda^*\rho_{10} & \rho_{11} \end{pmatrix}, \qquad |\lambda|\le1.$$

The off-diagonal elements are suppressed while the populations remain fixed. That is decoherence without energy relaxation in this basis.

By contrast, amplitude damping changes populations:

$$\lvert1\rangle \longrightarrow \lvert0\rangle \quad \text{with some decay probability}.$$

It also typically reduces coherence, but the population transfer is the dissipative part. Keeping these mechanisms separate makes it easier to diagnose experiments and read master equations.

Decoherence Is Not a Complete Interpretation

Decoherence explains why interference between alternatives becomes locally inaccessible when the alternatives become entangled with many uncontrolled degrees of freedom. A simple branch structure is

$$\left(c_0\lvert0\rangle+c_1\lvert1\rangle\right) \lvert E_0\rangle \longrightarrow c_0\lvert0\rangle\lvert E_0'\rangle + c_1\lvert1\rangle\lvert E_1'\rangle.$$

After tracing over the environment, the reduced coherence is multiplied by an environmental overlap:

$$\rho_{01} \longmapsto \rho_{01}\langle E_1'|E_0'\rangle.$$

When $\langle E_1'|E_0'\rangle\approx0$, interference between $\lvert0\rangle$ and $\lvert1\rangle$ becomes negligible for local system measurements. This explains the practical stability of pointer-like alternatives.

What it does not do, by itself, is turn the total superposition into a single selected outcome in a way accepted by all interpretations. The measurement problem is therefore not the same question as the calculation of decoherence rates.

Markovian Does Not Mean Environment-Free

A Markovian reduced model is one in which the future reduced evolution depends on the present reduced state in a memoryless way, often through a time-local generator:

$$\frac{d\rho}{dt} = \mathcal L(\rho).$$

The environment has not disappeared. It has been compressed into rates, noise spectra, jump operators, temperatures, or correlation functions. In a weak-coupling derivation, Markovianity is typically justified by a short bath correlation time compared with the system evolution time.

The opposite mistake is also common: “non-Markovian” does not simply mean complicated, strong, or time dependent. It means that some memory or information backflow invalidates a memoryless reduced description, or that the chosen divisibility condition fails.

Lindblad Form Is Not a Microscopic Derivation

The Gorini–Kossakowski–Sudarshan–Lindblad form is the standard generator for a time-homogeneous Markovian quantum dynamical semigroup:

$$\frac{d\rho}{dt} = -\frac{i}{\hbar}[H,\rho] + \sum_\mu \left( L_\mu\rho L_\mu^\dagger - \frac{1}{2} \{L_\mu^\dagger L_\mu,\rho\} \right).$$

This equation is structurally powerful: it preserves trace, Hermiticity, and complete positivity under its assumptions. But writing an equation in Lindblad form is not the same as deriving the correct $H$, $L_\mu$, and rates from a real system.

Microscopic derivations may require assumptions such as weak coupling, initially factorized states, rapidly decaying bath correlations, secular approximation, rotating-wave approximation, or coarse graining. If those assumptions fail, a Lindblad equation may still be a useful effective model, but its parameters should not be overinterpreted.

Trajectories Are Conditioned Descriptions

Quantum trajectories represent conditioned state evolution associated with measurement records or with an unraveling of an unconditional master equation. A jump trajectory might describe an emitted photon record; a diffusive trajectory might describe a continuous homodyne current.

The key warning is that the same master equation can have different unravelings. If the record is ignored, the average over trajectories gives the unconditional state:

$$\rho(t) = \mathbb E[\rho_c(t)],$$

where $\rho_c(t)$ is the conditional state and $\mathbb E$ denotes averaging over records. The trajectory is not a hidden path followed by the system independent of the measurement arrangement. It is a state assignment conditioned on a specified monitoring scheme.

Thermalization Is More Than Dephasing

Thermalization means approach, under suitable conditions, to a thermal state such as

$$\rho_\beta = \frac{e^{-\beta H}} {\operatorname{Tr}(e^{-\beta H})}.$$

Dephasing can make a density matrix diagonal in some basis without producing the correct thermal populations. Conversely, a system can exchange energy with a bath and relax populations while retaining some coherences during intermediate times.

A useful test is:

$$\text{decoherence asks about off-diagonal terms,} \qquad \text{thermalization asks about the final state}.$$

For many applications both effects matter, but they answer different physical questions.

Noise Can Be a Resource

Uncontrolled noise usually limits coherence, precision, and gate fidelity. But open-system dynamics is not merely a list of failures. Engineered dissipation can cool systems, prepare entangled steady states, stabilize codes, reset qubits, implement reservoir engineering, and make weak signals observable through measurement backaction.

The correct question is not “is there noise?” but:

1. What channel or generator describes it?
2. Is it known, calibrated, monitored, or uncontrolled?
3. Does it commute with the information one wants to preserve?
4. Can feedback, error correction, or engineered reservoirs turn it into a useful operation?

This is one reason measurement, decoherence, and control belong in the same volume.

Cross-Links

• Overview
• Concept Map
• Closed vs Open Quantum Systems
• Projective Measurement
• POVMs: First Encounter
• Generalized Measurements Overview
• Density Operators
• Partial Trace
• Decoherence Preview
• Liouville–von Neumann Equation

References

• J. von Neumann, Mathematical Foundations of Quantum Mechanics, Princeton University Press, 1955.
• K. Kraus, States, Effects, and Operations: Fundamental Notions of Quantum Theory, Springer, 1983.
• P. Busch, P. J. Lahti, and P. Mittelstaedt, The Quantum Theory of Measurement, Springer, 1996.
• M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information, Cambridge University Press, 2010.
• H.-P. Breuer and F. Petruccione, The Theory of Open Quantum Systems, Oxford University Press, 2002.
• H. M. Wiseman and G. J. Milburn, Quantum Measurement and Control, Cambridge University Press, 2010.
• M. Schlosshauer, Decoherence and the Quantum-to-Classical Transition, Springer, 2007.
• E. Joos, H. D. Zeh, C. Kiefer, D. Giulini, J. Kupsch, and I.-O. Stamatescu, Decoherence and the Appearance of a Classical World in Quantum Theory, Springer, 2003.

Exercises

1. A two-outcome detector has effects $F_0=F_1=I/2$. Does this determine the post-measurement state?

Solution

No. The effects determine only $p(0)=p(1)=1/2$ for every input state. One implementation could ignore the system and output a random classical bit, leaving $\rho$ unchanged. Another could measure a projective observable and then randomize the reported label. These have the same outcome probabilities but different backaction. An instrument is needed to specify the post-measurement state.

2. A qubit undergoes the map

$$\rho \longmapsto \begin{pmatrix} \rho_{00} & e^{-\Gamma t}\rho_{01} \\ e^{-\Gamma t}\rho_{10} & \rho_{11} \end{pmatrix}.$$

Is this dissipation, decoherence, or both?

Solution

It is decoherence in the displayed basis. The populations are unchanged, so there is no energy relaxation in a Hamiltonian whose energy eigenbasis is $\{\lvert0\rangle,\lvert1\rangle\}$. If a different Hamiltonian is relevant, the interpretation must be checked against that Hamiltonian.

3. Explain why a positive map on a single system may still be rejected as an open-system evolution law.

Solution

The system may be entangled with an external reference. A physically valid local evolution must keep every joint state positive under $\Phi\otimes I_R$. Positivity of $\Phi$ alone only checks isolated system states; complete positivity checks consistency with entanglement to arbitrary references.

4. A master equation is fitted to data and has Lindblad form. What can be concluded, and what cannot be concluded without further modeling?

Solution

One can conclude that the fitted generator defines a trace-preserving completely positive Markovian semigroup, assuming the fitted rates are physically allowed. One cannot conclude that the listed jump operators are uniquely microscopic, that weak-coupling or secular approximations are valid, or that the environment has no memory outside the fitted regime. Those require a model of the system, bath, coupling, and time scales.