Misconceptions about Measurement and Open Systems

Most mistakes in measurement and open-system calculations come from asking a formula to answer more than it specifies. An outcome probability does not determine backaction; a channel at one time does not determine memory; a diagonal density matrix need not be thermal. The examples below show how to recognize these gaps before doing a longer calculation.

Required background. One Interaction, Different Measurement Records and Closed versus Open Systems distinguish the descriptions used here. Density Operators supplies the state language.

Measurement probabilities do not determine the output state

For effects $E_0=E_1=I/2$, both outcomes have probability $1/2$ for every input. One instrument can generate a random bit and leave the system untouched. Another can first measure $Z$, discard that result, and then report an independent random bit. The effects are identical, but the second instrument dephases the system.

Thus “an uninformative measurement” does not automatically mean “no disturbance.” Even a sharp PVM does not uniquely determine its instrument. The Lüders update is one important ideal implementation; a device may also rotate or reset its output. Use Quantum Instruments whenever later measurements depend on the output state.

Ignoring an outcome does not undo the measurement

An unread ideal $Z$ measurement sends $|+\rangle\langle+|$ to $I/2$. A subsequent $X$ measurement is then random, whereas it would have been certain without the first measurement. The interaction matters even if no human looks at its record.

Discarding a record later does not introduce a new physical kick. It changes the conditional information available for predictions. The nonselective state is the probability-weighted average of the conditional states, not their coherent sum. Selective and Nonselective Measurements develops this distinction.

Positivity on isolated states is not complete positivity

Matrix transposition sends every density matrix to another density matrix. Nevertheless, applying it to just one half of a maximally entangled qubit pair produces an operator with eigenvalue $-1/2$. It cannot describe a deterministic local operation on arbitrary states that may be entangled with a reference.

This is why Completely Positive Maps test the extension of a map to untouched reference systems. For a channel, trace preservation is a separate requirement. For a selected operation, trace decrease records its probability; normalization is performed only after selecting a nonzero-probability outcome.

Dephasing is neither thermalization nor a universal energy statement

Dephasing removes off-diagonal terms in a specified basis. It preserves energy when the Hamiltonian is diagonal in that basis. If it is not, dephasing can change the mean energy. Population preservation in an arbitrary basis is therefore insufficient to claim absence of energy exchange.

Even in the energy basis, dephasing leaves the initial populations unchanged. An excited energy eigenstate is already diagonal, yet is generally not a thermal state at the bath temperature. Dephasing versus Dissipation separates these processes; Thermal Master Equations explains what produces thermal populations.

Decoherence also does not select one unique outcome from the full entangled state. It explains suppression of local interference and, in appropriate models, robust records. The additional interpretive issue is discussed in What Decoherence Does Not Solve.

A time-local equation can describe dynamics with memory

The notation $\dot\rho(t)=\mathcal L_t\rho(t)$ does not settle whether a process is Markovian. Some non-Markovian reduced evolutions admit a time-local generator, possibly with negative canonical rates or singularities. Conversely, every map from the initial time may be a channel even when intermediate propagators are not completely positive.

Different definitions test different features. CP Divisibility concerns intermediate maps; Information Backflow concerns changes in distinguishability; stronger process descriptions include interventions at multiple times. State which definition is being used before comparing claims about memory.

A valid GKLS generator need not be the right physical model

In finite dimension a time-independent GKLS generator defines a completely positive, trace-preserving semigroup. That mathematical guarantee does not establish that its rates, steady state, or approximations fit a particular experiment. A fitted exponential over a short interval may miss later revivals or slow noise.

Nor are the displayed jump operators a unique microscopic account: a generator can have different operator representations. A detector model is needed before interpreting a particular jump as an observed event. Open Quantum Systems explains the approximations; Lindblad Operators explains the representation freedom.

A trajectory depends on how the system is monitored

Photon counting and homodyne detection can monitor the same output field but generate different conditional states. Averaging either set, with the appropriate probabilities and the same unconditioned setup, recovers the master equation. The trajectory is therefore a conditional description associated with a specified record, rather than a unique hidden path supplied by the master equation alone. See Unravelings.

An environment can also cool or stabilize a system. Noise does not always increase entropy: amplitude damping can take a maximally mixed qubit toward a pure ground state. Reservoir Engineering uses this possibility deliberately, subject to achievable couplings, rates, and imperfections.

Exercises

Distinguish two detectors. For the two instruments with effects $I/2$ described above, prepare $|+\rangle$ and then measure $X$. What statistics distinguish them?

Solution

The instrument that leaves the system untouched gives $X=+1$ with certainty. The one that performs an unread $Z$ measurement gives $X=\pm1$ with equal probability. The reported random bit supplies no distinction; the subsequent system measurement does.

Noise and purity. A qubit initially in $I/2$ undergoes amplitude damping with survival probability $s$ for the excited population, where $0\le s\le1$. Find its final purity.

Solution

The populations become $1-s/2$ and $s/2$, with no coherence. The purity is $(1-s/2)^2+(s/2)^2=1-s+s^2/2$. It rises from $1/2$ to $1$ as $s$ falls from one to zero. The bath carries away energy and correlations; subsystem entropy need not increase.

References

• Breuer, H.-P., and F. Petruccione. The Theory of Open Quantum Systems. Oxford University Press, 2002. Open dynamics and approximations.
• Nielsen, M. A., and I. L. Chuang. Quantum Computation and Quantum Information. 10th anniversary edition, Cambridge University Press, 2010. Instruments, complete positivity, and channels.
• Schlosshauer, M. Decoherence and the Quantum-to-Classical Transition. Springer, 2007. Decoherence mechanisms and interpretive limits.
• Wiseman, H. M., and G. J. Milburn. Quantum Measurement and Control. Cambridge University Press, 2010. Conditional dynamics and monitoring.