Pointer States

Pointer states are the states of a system that remain comparatively stable while the system interacts with an apparatus or environment. They are the states whose alternatives can be recorded, amplified, and later treated as effectively classical outcomes.

Pointer states are found from dynamics over a stated interval, not by diagonalizing a density matrix at one instant. A robust state may move along a predictable trajectory rather than remain stationary.

Helpful background. Coherence and Preferred Bases distinguishes a choice of coordinates from a physically stable set of alternatives.

Pointer states in a monitoring interaction

The simplest model of environmental monitoring has an interaction Hamiltonian

$$H_{\mathrm{int}} = \sum_a \lvert a\rangle\langle a\rvert \otimes B_a.$$

Assume a pure, initially uncorrelated environment and that the system Hamiltonian preserves the $a$ sectors or is negligible on the interaction timescale. If the system starts in $\lvert a\rangle$, this model does not mix the label:

$$\lvert a\rangle\lvert E_0\rangle \longmapsto \lvert a\rangle\lvert E_a(t)\rangle.$$

The environment changes, but the system state remains in the same alternative. By contrast, a superposition evolves as

$$\sum_a c_a\lvert a\rangle\lvert E_0\rangle \longmapsto \sum_a c_a\lvert a\rangle\lvert E_a(t)\rangle.$$

Tracing out the environment gives

$$\rho_{ab}(t) = \langle E_b(t)|E_a(t)\rangle \rho_{ab}(0).$$

When the environment states become distinguishable,

$$\langle E_b(t)|E_a(t)\rangle\approx0 \qquad (a\ne b),$$

coherence between different $\lvert a\rangle$ alternatives disappears from the reduced state. The same alternatives are robust under monitoring and are therefore pointer states in this idealized model.

What Robustness Means

Robustness has several related meanings. In a useful pointer-state model, the candidate states should satisfy most of the following:

• They are not rapidly converted into superpositions of other candidate states by the system-environment coupling.
• For an initially pure environment, they produce relatively little entanglement when prepared individually; reduced purity is a more general diagnostic for mixed environments.
• Superpositions of different candidates decohere quickly.
• Their records are amplified into apparatus or environmental degrees of freedom.
• They remain identifiable over the timescale on which the record is used.

These criteria are operational. A laboratory pointer is useful because its alternatives can be read later. A macroscopic detector current, a meter position, or a stable memory state is not merely a vector in Hilbert space; it is a dynamical record protected well enough against later interference.

Pointer Basis, Pointer Subspace, Pointer Set

The phrase “pointer basis” is common, but it is not always literally a basis.

If the robust states are orthogonal and discrete, they form a pointer basis:

$$\{\lvert a\rangle\}.$$

If the environment distinguishes only coarse-grained labels, the robust objects may be subspaces:

$$\mathcal H = \bigoplus_\alpha\mathcal H_\alpha.$$

Coherence between different $\mathcal H_\alpha$ sectors is suppressed, while coherence inside each sector may survive. This is the structure behind Decoherence-Free Subspaces and noiseless subsystem ideas.

In oscillator and field problems, the robust states may be an overcomplete nonorthogonal set. Coherent states under vacuum pure loss are a standard example: they remain coherent while their amplitude decays. A thermal bath generally makes them mixed, so the statement is model dependent.

For this reason, “pointer structure” is often more accurate than “pointer basis.”

For the distinction between basis-relative coherence and dynamically selected preferred structure, see Coherence and Preferred Bases. For the effective-superselection language built from pointer structure, see Einselection.

Predictability Sieve

The predictability sieve is a way to identify pointer states by ranking initial states according to how predictable they remain under open-system evolution.

A typical diagnostic is purity loss or entropy production. If

$$\rho_\psi(0) = \lvert\psi\rangle\langle\psi\rvert,$$

one evolves the reduced state and compares quantities such as

$$\operatorname{Tr}\rho_\psi(t)^2, \qquad S(\rho_\psi(t)) = -\operatorname{Tr}\rho_\psi(t)\log\rho_\psi(t),$$

or the fidelity with a simple expected trajectory.

States that retain high purity, low entropy, or high predictability over the relevant timescale are candidate pointer states. The method is called a “sieve” because it filters out states that rapidly entangle with the environment or become operationally unstable.

This is a criterion, not a universal closed-form algorithm. The answer depends on the Hamiltonian, coupling operators, bath state, timescale, and coarse graining.

Lindblad Diagnostic

For Markovian dynamics,

$$\frac{d\rho}{dt} = -\frac{i}{\hbar}[H,\rho] + \sum_\mu \left( L_\mu\rho L_\mu^\dagger - \frac12\{L_\mu^\dagger L_\mu,\rho\} \right),$$

The rates are absorbed into $L_\mu$, which therefore have units of inverse square-root time. In finite dimensions, or on states with the required finite moments for unbounded operators, one local test is the instantaneous purity loss of a pure state. Ignoring the Hamiltonian contribution to purity, for $\rho=\lvert\psi\rangle\langle\psi\rvert$ one finds

$$\left. \frac{d}{dt} \operatorname{Tr}\rho^2 \right|_{\psi} = -2 \sum_\mu \left( \langle L_\mu^\dagger L_\mu\rangle_\psi - |\langle L_\mu\rangle_\psi|^2 \right).$$

Each term is nonnegative by the Cauchy-Schwarz inequality. The purity loss vanishes when $\lvert\psi\rangle$ is an eigenstate of every relevant $L_\mu$:

$$L_\mu\lvert\psi\rangle = \ell_\mu\lvert\psi\rangle \qquad \text{for all }\mu.$$

This is an instantaneous criterion, not a guarantee of finite-time purity. The Hamiltonian can move the state out of this set. Even a nonnormal dissipator can do so: for $H=0$ and

$$L=\sqrt\kappa\begin{pmatrix}1&1\\0&0\end{pmatrix}, \qquad L|0\rangle=\sqrt\kappa|0\rangle,$$

one nevertheless obtains

$$\mathcal D[L](|0\rangle\langle0|) =-\frac\kappa2\begin{pmatrix}0&1\\1&0\end{pmatrix}.$$

The purity derivative is zero at that instant, but the state starts moving away from the eigenvector. A finite-time test must check the full generator and the evolving state, not just the initial variances.

Example: Pure Dephasing

For qubit pure dephasing in the $Z$ basis,

$$\frac{d\rho}{dt} = \frac{\Gamma_\phi}{2} \left( Z\rho Z-\rho \right),$$

the coupling monitors the $Z$ alternatives. The states

$$\lvert0\rangle, \qquad \lvert1\rangle$$

are pointer states. They remain pure under the dephasing channel, while superpositions such as

$$\frac{\lvert0\rangle+\lvert1\rangle}{\sqrt2}$$

lose coherence in the reduced description.

In this case the pointer basis coincides with the eigenbasis of the monitored observable. That coincidence is special to this simple model.

Example: Position Monitoring

For a massive particle interacting with a scattering environment, environmental records often carry information about position. A schematic coupling is

$$H_{\mathrm{int}} \sim x\otimes B.$$

Then spatially separated wavepacket components leave distinguishable environmental records, and the reduced density matrix in position representation has suppressed off-diagonal elements:

$$\rho(x,x';t)\approx e^{-F(x-x')t}\rho(x,x';0).$$

For well-resolved alternatives and enough collisions, this factor is small. At fixed time the scattering rate can saturate with separation, so it need not tend to zero merely by increasing separation. Approximate pointer states are localized wavepackets; exact position eigenstates are nonnormalizable, and narrow physical packets spread under the free Hamiltonian. The selected packet width reflects a competition between environmental localization and Hamiltonian spreading.

This is why macroscopic positions can behave classically even though exact position eigenstates are not physical states.

Example: Amplitude Damping

For zero-temperature qubit amplitude damping, the channel pulls every state toward the ground state $\lvert0\rangle$:

$$\mathcal A_\gamma(\lvert0\rangle\langle0\rvert) = \lvert0\rangle\langle0\rvert.$$

The ground state is stable. The excited state is not stable:

$$\lvert1\rangle\langle1\rvert \longmapsto \gamma\lvert0\rangle\langle0\rvert + (1-\gamma)\lvert1\rangle\langle1\rvert.$$

Thus amplitude damping does not select a symmetric two-state pointer basis in the same way pure dephasing does. It selects an attractor and a decay record. In a monitored quantum-jump picture, a detected decay event points to $\lvert0\rangle$, while the absence of a detected jump also updates the state.

This distinction is one reason Amplitude-Damping Channel should not be treated as a Pauli or dephasing channel.

Example: Damped Oscillator

For a harmonic oscillator undergoing pure loss into a vacuum environment, with $0\le\eta\le1$, coherent states are especially stable. A beam-splitter model of loss maps

$$\lvert\alpha\rangle_S \lvert0\rangle_E \longmapsto \lvert\sqrt{\eta}\,\alpha\rangle_S \lvert\sqrt{1-\eta}\,\alpha\rangle_E.$$

No entanglement is produced for a single coherent-state input in this idealized model. Superpositions of well-separated coherent states, however, entangle with distinguishable environment states and decohere quickly.

The environmental overlap multiplying a coherence dyad between coherent inputs $|\alpha\rangle$ and $|\beta\rangle$ has magnitude

$$|D|=\exp\!\left[-\frac{1-\eta}{2}|\alpha-\beta|^2\right].$$

For alternatives $\alpha=2$ and $\beta=-2$, transmitting $90\%$ of the intensity leaves each amplitude at $\sqrt{0.9}\approx0.949$ of its initial value, but gives $|D|=e^{-0.8}\approx0.449$. Individual coherent states remain pure even while a superposition of separated alternatives loses substantial interference. This illustrates why pointer-state stability and superposition fragility are different questions.

Relation to Measurement Pointers

In a measurement apparatus, the pointer states are the stable macroscopic record states: different dial positions, detector clicks, memory states, current levels, or field amplitudes.

The measurement interaction correlates the microscopic system with apparatus alternatives. Environmental decoherence then suppresses interference between those alternatives and helps make the apparatus record robust:

$$\sum_a c_a \lvert a\rangle_S \lvert A_{\mathrm{ready}}\rangle\lvert E_{\mathrm{ready}}\rangle \longmapsto \sum_a c_a \lvert a\rangle_S \lvert A_a\rangle \lvert E_a\rangle.$$

If

$$\langle E_b|E_a\rangle\approx0 \qquad (a\ne b),$$

local interference between distinct apparatus records is suppressed. The states $\lvert A_a\rangle$ are pointer states only if they also remain stable under later interactions. A record that immediately disperses or becomes unreadable is not a useful pointer.

Stable alternatives and measurement outcomes

Pointer-state analysis identifies which records resist interference and remain useful for later observations. Agreement among different observers additionally needs accessible, consistent records; stability alone does not guarantee this. Quantum Darwinism examines redundant records in environmental fragments.

Robustness also does not select one individual outcome from a global superposition or replace the Born rule. What Decoherence Does Not Solve explains that distinction. Decoherence as a Classical-Limit Bridge relates robust wavepackets to semiclassical motion.

Exercises

Monitoring Hamiltonian

For

$$H_{\mathrm{int}} = \sum_a \lvert a\rangle\langle a\rvert\otimes B_a,$$

assume

$$\lvert a\rangle\lvert E_0\rangle \mapsto \lvert a\rangle\lvert E_a\rangle.$$

Starting with $\sum_a c_a\lvert a\rangle\lvert E_0\rangle$, compute the reduced density matrix of the system and identify the pointer states.

Solution

The final joint state is

$$\lvert\Psi\rangle = \sum_a c_a \lvert a\rangle\lvert E_a\rangle.$$

The density operator is

$$\lvert\Psi\rangle\langle\Psi\rvert = \sum_{a,b} c_ac_b^* \lvert a\rangle\langle b\rvert \otimes \lvert E_a\rangle\langle E_b\rvert.$$

Tracing over the environment gives

$$\rho_S = \sum_{a,b} c_ac_b^* \langle E_b|E_a\rangle \lvert a\rangle\langle b\rvert.$$

When $\langle E_b|E_a\rangle\approx0$ for $a\ne b$, the state is approximately diagonal in the $\{\lvert a\rangle\}$ basis. The states $\lvert a\rangle$ are pointer states in this idealized monitoring model.

Lindblad purity loss

For a pure state $\rho=\lvert\psi\rangle\langle\psi\rvert$, show that a single Lindblad operator $L$ gives instantaneous purity loss

$$\left. \frac{d}{dt}\operatorname{Tr}\rho^2 \right|_{\psi} = -2 \left( \langle L^\dagger L\rangle_\psi - |\langle L\rangle_\psi|^2 \right).$$

What condition makes this loss vanish?

Solution

For the dissipator

$$\mathcal D[L]\rho = L\rho L^\dagger - \frac12\{L^\dagger L,\rho\},$$

the purity derivative is

$$\frac{d}{dt}\operatorname{Tr}\rho^2 = 2\operatorname{Tr}(\rho\,\mathcal D[L]\rho).$$

For $\rho=\lvert\psi\rangle\langle\psi\rvert$,

$$\operatorname{Tr}(\rho L\rho L^\dagger) = |\langle L\rangle_\psi|^2,$$

and

$$\operatorname{Tr}(\rho L^\dagger L\rho) = \operatorname{Tr}(\rho^2 L^\dagger L) = \langle L^\dagger L\rangle_\psi.$$

Thus

$$\frac{d}{dt}\operatorname{Tr}\rho^2 = -2 \left( \langle L^\dagger L\rangle_\psi - |\langle L\rangle_\psi|^2 \right).$$

The bracket vanishes exactly when equality holds in Cauchy-Schwarz, which means

$$L\lvert\psi\rangle = \ell\lvert\psi\rangle$$

for some complex number $\ell$. Thus a pure state that is an eigenstate of the monitored Lindblad operator has no instantaneous purity loss from that dissipator.

Dephasing versus damping

Compare the pointer-state structure of qubit pure dephasing with zero-temperature amplitude damping.

Solution

For pure dephasing in the $Z$ basis, the states $\lvert0\rangle$ and $\lvert1\rangle$ are both stable under the channel. Superpositions of them lose coherence. Thus the model selects the $Z$ basis as a pointer basis.

For zero-temperature amplitude damping, $\lvert0\rangle$ is stable but $\lvert1\rangle$ decays toward $\lvert0\rangle$. The process selects a ground-state attractor and a decay record rather than a symmetric two-state pointer basis.

Pointer states are not just eigenvectors of rho

Give a reason why diagonalizing $\rho_S(t)$ at one time is not enough to identify pointer states.

Solution

The eigenvectors of $\rho_S(t)$ can depend on accidental degeneracies, transient dynamics, or the initial state. Pointer states are selected by the dynamical coupling to the environment and by record stability. Two states with the same instantaneous density matrix spectrum can have different future robustness under different system-environment Hamiltonians.

References

• J. P. Paz, S. Habib, and W. H. Zurek, “Reduction of the wave packet: Preferred observable and decoherence time scale,” Physical Review D 47, 488–501 (1993). DOI.
• M. Schlosshauer, “Quantum decoherence,” Physics Reports 831, 1–57 (2019). Author’s text.
• W. H. Zurek, “Pointer basis of quantum apparatus: Into what mixture does the wave packet collapse?,” Physical Review D 24, 1516–1525 (1981). DOI.
• W. H. Zurek, S. Habib, and J. P. Paz, “Coherent states via decoherence,” Physical Review Letters 70, 1187–1190 (1993). DOI.