Quantum-State Purity¶

The quadratic-trace measure of how concentrated a quantum density operator's eigenvalues are.

Version: v1 · 2026-10-03 · History
Domain-specific #: 13545
Domain group: Natural Sciences
Origin domain: Physics
Subdomains: Quantum Mechanics, Quantum Information → Physics
Aliases: Purity of a quantum state

Core Idea¶

Quantum-state purity is the number $P(\rho)=\operatorname{Tr}(\rho^2)$ assigned to a normalized density operator $\rho$. If $\rho$ has eigenvalues $p_1,\ldots,p_d$ in a $d$-dimensional Hilbert space, then $P(\rho)=\sum_i p_i^2$. Those eigenvalues are nonnegative and sum to one. Thus $1/d\leq P(\rho)\leq1$: the upper endpoint occurs exactly for a rank-one, pure state, while the lower endpoint occurs for the maximally mixed state $I/d$. IBM Quantum gives the quadratic-trace definition; Preskill's density-operator spectral account supplies the pure/mixed distinction and the premises for the bound.^[1]^[2]

This scalar tells us how concentrated the state's spectrum is, not why it is mixed. A low-purity state may be a reduced part of a pure entangled whole, an intentionally prepared statistical mixture, or the output of a particular channel. A change in purity is not, by itself, a diagnosis of decoherence or information loss to an environment. Under unitary conjugation $U\rho U^\dagger$, purity stays fixed by cyclicity of trace. Nonunitary channels can lower it or raise it: IBM's complete dephasing sends $|+\rangle\langle+|$ to $I/2$, while its reset channel sends $I/2$ to $|0\rangle\langle0|$.^[3]

For a known globally pure bipartite state, the reduced state's purity provides a precise entanglement test: $P(\rho_A)<1$ exactly when the joint pure state has Schmidt rank greater than one. Without that pure-global premise, reduced mixedness does not prove entanglement; a separable mixed preparation can give the same reduced density matrix.^[2]

Structural Signature¶

Sig role-phrases: normalized density operator → quadratic trace → spectral concentration → dimension baseline → interpretation context.

Normalized quantum-state carrier. The input is a positive, trace-one density operator on a stated finite-dimensional Hilbert space. Its positivity and normalization make the eigenvalues a probability vector. An arbitrary matrix is not licensed the same bounds or state interpretation.^[2]
Quadratic trace. The operation $\operatorname{Tr}(\rho^2)$ produces one scalar without selecting a measurement basis. This functional—not $\operatorname{Tr}(\rho)=1$, two-state fidelity or von Neumann entropy—is the named measure.^[1]
Spectral concentration. Squaring eigenvalues rewards a spectrum concentrated on fewer nonzero entries. Rank one gives $P=1$; more distributed eigenvalues give a smaller value. Equal probabilities in a chosen measurement basis alone do not establish a mixed density operator.^[2]
Dimension-dependent scale. $I/d$ has purity $1/d$, not zero. Comparing raw values across Hilbert spaces of different dimension requires preserving their different lower baselines.^[2]^[1]
Interpretive context. A global, reduced or outcome-conditioned density operator may have a different purity even when they describe related physical circumstances. An entanglement conclusion from reduced purity requires the joint state to be known pure; a causal conclusion about decoherence requires channel or preparation evidence.^[2]^[3]

The first four roles define and scale the number. The fifth is not needed to calculate it, but is indispensable for interpreting it without false causal or entanglement claims.

What It Is Not¶

It is not a probability that one chosen basis outcome occurs. A pure superposition can yield several probabilistic measurement outcomes while retaining $P=1$. Purity depends on the density operator's eigenvalues, not on a conveniently selected measurement basis.^[2]

It is not two-state fidelity. Fidelity asks how close $\rho$ is to another state $\sigma$; purity evaluates one state through $\operatorname{Tr}(\rho^2)$. Nor is it quantum-state purification, the construction of a pure state on a larger system whose reduction is $\rho$. The words share a root but name different operations.^[2]

It is not a monotone clock for every open process. The reset channel is a valid nonunitary quantum channel that increases the purity of a maximally mixed qubit. A dephasing channel may lower purity for an off-diagonal input. The direction belongs to the state and channel together, not to “openness” as such.^[3]

Scope of Application¶

In finite-dimensional quantum information, purity distinguishes rank-one states from mixed density operators and gives a scalar summary of intermediate mixedness. The lower bound is tied to the dimension, so qubit $I/2$ has $P=1/2$ and a four-dimensional $I/4$ has $P=1/4$.^[2]^[1]

In channel analysis, one can compare the input and output density matrices under a specified map. IBM describes unitary conjugation, complete dephasing, complete depolarization and reset as different channels. The purity calculation can register a change but needs the channel model to explain its direction and mechanism.^[3]

In bipartite-state analysis, purity can be applied separately to the joint state and a reduced subsystem. Preskill's Schmidt decomposition makes the inference exact for a pure joint state: a product pure state leaves pure reduced states, whereas an entangled pure state has mixed reduced states. That inference does not extend unchanged to an arbitrary globally mixed state.^[2]

Clarity¶

The functional resolves one question cleanly: is this normalized density operator rank one, and how concentrated is its spectrum? $P=1$ says rank one; $P<1$ says mixed, provided the finite-dimensional state assumptions hold. This is more precise than calling a state “coherent,” “clean,” or “well known,” terms that can refer to different properties.^[1]^[2]

It also keeps the level of description visible. The Bell pair $|\Phi^+\rangle$ is globally pure, yet each qubit's reduced state is $I/2$ and has purity $1/2$. “The state is pure” and “the subsystem is mixed” are both true because they refer to different density operators. Without writing the subsystem, the statements look contradictory.^[2]

Finally, it blocks causal overreading. Purity alone cannot reveal whether a mixed state arose from dephasing, ignorance of a preparation label, or tracing out a correlated partner. These histories may generate identical $\rho$ and therefore identical values of every functional of $\rho$.^[2]^[3]

Manages Complexity¶

Purity compresses a $d\times d$ density matrix to one basis-invariant number. Spectral diagonalization shows what the compression retains: the second moment $\sum_i p_i^2$ of the eigenvalue distribution. It ignores eigenvectors, the order of preparation events and much of the spectrum's detail. This makes it useful for a quick pure-versus-mixed test and for controlled comparisons within a fixed dimension.^[2]^[1]

The compression must not be mistaken for completeness. Distinct density operators can share the same purity, and at dimension greater than two different spectra can share the same sum of squares. When the decision needs a source of noise, an entanglement claim under mixed global states, or full state discrimination, retain more information than this scalar.^[2]

Abstract Reasoning¶

Start with a positive trace-one $\rho$ and identify its Hilbert-space dimension. Compute $\operatorname{Tr}(\rho^2)$ directly or diagonalize and sum squared eigenvalues. Compare it with $1$ and $1/d$. If a state changes under a unitary $U$, the expression $\operatorname{Tr}[(U\rho U^\dagger)^2]=\operatorname{Tr}(\rho^2)$ proves that any claimed purity change is incompatible with closed unitary conjugation alone.^[1]^[2]^[3]

If purity changes under a channel, do not jump directly from the sign to a mechanism. Check the channel's action and the initial state. IBM's complete dephasing maps $|+\rangle\langle+|$ to $I/2$, so $P$ goes $1\to1/2$; reset maps $I/2$ to $|0\rangle\langle0|$, so it goes $1/2\to1$. These explicit calculations refute a universal open-system decrease.^[3]

For entanglement, ask first whether the joint state is known pure. If yes, compute one reduction and test its purity: below one means Schmidt rank above one and thus entanglement. If not, a mixed reduced state is compatible with separable classical correlation, so the test cannot by itself decide the entanglement question.^[2]

Knowledge Transfer¶

The same formula transfers literally between a channel output, a reconstructed qubit state and a subsystem reduced from a larger quantum state. In each, the carrier remains a normalized density operator and the quadratic trace has the same mathematical meaning. The Interpretation changes with whether the operator is global, reduced or conditioned, and with what is known about preparation and dynamics.^[1]^[2]^[3]

Outside quantum theory, $\sum_i p_i^2$ resembles other concentration indices, but resemblance alone does not make those measures this named quantum-state purity. Its carrier, unitary invariance and conditional entanglement use require density-operator structure. A generic second-moment concentration prime might be a worthwhile unadmitted future-prime question; no such parent is asserted here. Live Density Matrix is the proposed strict prerequisite, not a synonym for the functional.

Examples¶

A qubit through dephasing and reset¶

Let $|+\rangle=(|0\rangle+|1\rangle)/\sqrt2$. Its projector $|+\rangle\langle+|$ has eigenvalues $(1,0)$ and purity $1$. IBM's complete dephasing removes the off-diagonal entries, leaving $I/2$ with eigenvalues $(1/2,1/2)$ and purity $1/2$. IBM's reset channel instead sends that $I/2$ to $|0\rangle\langle0|$, restoring purity $1$. Neither calculation claims the physical channel is reversible or that reset recovers the original $|+\rangle$ state.^[3]^[1]

Mapped back: normalized carrier = qubit input/output density matrices; quadratic trace = $\operatorname{Tr}(\rho^2)$ for each; spectral concentration = $(1,0)$ versus $(1/2,1/2)$; dimension baseline = $d=2$ and minimum $1/2$; interpretive context = specified unconditioned dephasing or reset map, explaining opposite changes.

Pure Bell pair versus a separable mixed pair¶

For $|\Phi^+\rangle=(|00\rangle+|11\rangle)/\sqrt2$, the global projector has purity $1$, but tracing out either qubit gives $I/2$ with purity $1/2$. Because the joint state is known pure, this reduced mixedness confirms entanglement by the Schmidt-rank criterion. Now take the separable mixed joint state $\tfrac12|00\rangle\langle00|+\tfrac12|11\rangle\langle11|$. Its reduced qubits are also $I/2$ with purity $1/2$, yet the joint state is an explicit mixture of product states. The same local number has different inferential force.^[2]

Mapped back: normalized carrier = the global two-qubit state and one-qubit reduction; quadratic trace = applied separately to joint and local operators; spectral concentration = global Bell $(1,0,0,0)$ versus local $(1/2,1/2)$; dimension baseline = $d=4$ globally and $d=2$ locally; interpretive context = pure-global premise present only for the Bell pair.

Structural Tensions¶

T1 — Compact scalar versus cause-specific diagnosis. A single basis-independent number makes a pure-versus-mixed check cheap and comparable, but it discards eigenvectors, much spectral detail and preparation history. Insisting on a cause from the scalar risks a false decoherence narrative; demanding full channel and state evidence for every simple classification forfeits the economy that makes purity useful. Diagnostic: Is the actual decision only about rank-one status or spectral concentration, or does it require a channel/preparation explanation that $P$ cannot supply alone?^[2]^[3]

T2 — Locally available reduction versus globally justified entanglement claim. Reduced purity may be accessible without full joint-state reconstruction. Under a verified pure global state it gives a sharp entanglement test. Relaxing that premise broadens use but admits separable mixed counterexamples; insisting on it narrows applicability while preserving logical force. Diagnostic: Is global purity independently established, or could a separable mixed state produce the same reduced spectrum?^[2]

Structural–Framed Character¶

Quantum-State Purity is strongly structural within quantum information yet framed by the state formalism. Evaluative weight: $\operatorname{Tr}(\rho^2)$ is an objective mathematical functional; a claim that higher purity is better for a task is a separate evaluation. Human-practice dependence: experimenters choose a preparation, subsystem and conditioning record, but once $\rho$ is fixed the number is not a matter of wording. Institutional origin: IBM's implementation names and textbook conventions express, rather than create, the spectral identity. Vocabulary travel: “purity” appears in many fields, while this entry's density-operator calculation travels literally only where that carrier exists. Import versus recognition: a reduced-state or channel example qualifies because it supplies $\rho$ and the same quadratic trace, not because its story sounds like cleanliness or isolation.^[1]^[2]

Its character: a structural quantum-state measure whose physical interpretation is conditional on dimension, subsystem and dynamics—not a general causal index of environmental loss.

Structural Core vs. Domain Accent¶

The portable skeleton is quadratic concentration of a normalized spectrum. That resemblance to second-moment concentration measures is real mathematical commonality, but no existing prime was identified that supplies the full skeleton as an admitted parent. Whether it merits one is an unadmitted future-prime question, not an implicit ontology mutation.

The domain accent is constitutive: a density operator is positive and trace one, eigenvalues encode quantum-state mixedness, unitary conjugation preserves the spectrum, and pure-global Schmidt structure makes reduced purity an entanglement test. A generic concentration index lacks that whole inferential setting. Live Density Matrix is proposed as the carrier prerequisite; Purity adds a new one-state functional and bounded interpretations.^[2]^[1]

This entry presupposes Density matrix. Quantum-state purity is defined on a density operator.

Relationships to Other Abstractions¶

Current abstraction Quantum-State Purity Domain-specific

Parents (1) — more general patterns this builds on

Quantum-State Purity presupposes Density matrix Domain-specific

Quantum-state purity is defined on a density operator.

Hierarchy path (1) — routes to 1 parentless root

Quantum-State Purity → Density matrix

Neighborhood in Abstraction Space¶

Quantum-State Purity sits in a moderately populated region (59^th percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.

Family — Quantum States & Information Measures (25 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With¶

Pure-state probability amplitudes may yield uncertain outcomes in a chosen basis while the state still has $P=1$. Linear entropy $1-P$ is a simple rescaling of this functional, not the logarithmic von Neumann entropy. Fidelity is a two-state comparison. Purification is an extension construction. Decoherence is a type of dynamics that can lower purity in particular cases, not the definition of low purity. Entanglement from a local mixed state is a valid inference only with the stated pure-global bipartite assumption.^[2]^[3]

References¶

[1] IBM Quantum, Quantum Information API reference, “Measures → purity,” defining $P(\rho)=\operatorname{Tr}(\rho^2)$. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k

[2] John Preskill, Lecture Notes for Ph219/CS219: Quantum Information and Computation, Chapter 2 (updated July 2015), §§2.3–2.4 for density-operator spectrum, pure/mixed states, partial trace, Schmidt decomposition and the pure-bipartite entanglement criterion. Author-hosted Caltech notes. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u ↩v ↩w ↩x

[3] IBM Quantum Learning, “Quantum channel basics”, sections “Unitary operations as channels” and “Examples of qubit channels,” including reset and completely dephasing maps. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k