Magic–Entanglement Complementarity

• Magic–entanglement complementarity is a framework that quantitatively relates nonstabilizerness (magic) and entanglement with measure-dependent trade-offs underpinning various quantum protocols.
• It illustrates how entanglement acts as a conduit for the nonlocal spreading of magic, with dynamics captured by stabilizer Rényi entropy and operator entanglement measures.
• The framework establishes distinct dynamical and phase transitions in quantum systems, highlighting that magic and entanglement are complementary yet irreducible resources in computational scenarios.

Searching arXiv for the core papers to ground the article in current literature. Magic–entanglement complementarity denotes a family of quantitative relations between nonstabilizerness and entanglement, rather than a single universal trade-off. In recent work, complementarity has appeared in at least four distinct forms: entanglement can facilitate the nonlocal spreading of locally injected magic, entanglement and magic can undergo distinct dynamical transitions, local dissipation can irreversibly destroy entanglement while later reviving magic, and operator entanglement can be upper-bounded by magic monotones in the Heisenberg picture (Hou et al., 26 Mar 2025, Fux et al., 2023, Cao, 21 May 2026, Dowling et al., 30 Jan 2025). Taken together, these results place magic and entanglement in a joint resource-theoretic landscape in which they are neither reducible to one another nor generically monotone with each other.

1. Resource-theoretic setting and measures

A pure stabilizer state is stabilized by an Abelian subgroup of the $N$-qubit Pauli group with $2^N$ elements; such states are prepared by Clifford circuits and admit efficient classical simulation. Magic quantifies deviation from stabilizer structure, and several recent works use stabilizer-entropic measures as the principal diagnostics. In the stabilizer-state setting, the linear stabilizer entropy is

$$Y(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^4,$$

with $Y(\rho)=1$ iff $\rho$ is a stabilizer state, $Y(\rho)<1$ otherwise, and $M_2(\rho)=-\log Y(\rho)$ the second stabilizer Rényi entropy. The more general family

$$Y_\alpha(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^{2\alpha}, \qquad M_\alpha(\rho) := \frac{1}{1-\alpha}\log Y_\alpha(\rho)$$

is used to track the leading dependence of magic under local injections (Hou et al., 26 Mar 2025).

Entanglement is quantified differently across settings. For bipartite stabilizer states on $AB$, the von Neumann entropy $E=-\operatorname{tr}(\rho_A\log_2\rho_A)$ is an integer and counts Bell pairs across the cut after local Clifford reduction. In monitored circuits, the same entropy across a half cut is used as the entanglement order parameter. In two-qubit studies, entanglement is quantified by the concurrence $2^N$0, while dissipative GHZ analyses use bipartite negativity and separability thresholds. In operator-space formulations, local operator entanglement (LOE) is the entanglement entropy of the vectorized Heisenberg operator across a doubled bipartition $2^N$1 (Fux et al., 2023, Roman et al., 26 Mar 2026, Cao, 21 May 2026, Dowling et al., 30 Jan 2025).

The same plurality holds on the magic side. Hybrid-circuit work uses the stabilizer $2^N$2-Rényi entropy defined from the Pauli-string expectation distribution $2^N$3. Dissipative studies use the robustness of magic $2^N$4, defined by the minimal signed-stabilizer decomposition cost. Two-qubit Pareto analyses use $2^N$5, and operator-space work uses operator stabilizer Rényi entropy, unitary nullity, and $2^N$6-count. This suggests that “complementarity” is measure-dependent and protocol-dependent, not a single invariant law.

2. Entanglement as a conduit for magic spreading

The clearest “highway” formulation appears for an initial pure stabilizer state $2^N$7 on $2^N$8 with bipartite entanglement $2^N$9, followed by a Haar random unitary $$Y(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^4,$$0 on a local subregion $$Y(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^4,$$1. To leading order for $$Y(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^4,$$2,

$$Y(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^4,$$3

Since lower $$Y(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^4,$$4 means higher magic, increasing $$Y(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^4,$$5 exponentially suppresses $$Y(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^4,$$6 and therefore enhances global magic generation. The same leading factor persists for the stabilizer Rényi family,

$$Y(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^4,$$7

so $$Y(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^4,$$8 grows linearly in $$Y(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^4,$$9 at leading order (Hou et al., 26 Mar 2025).

The mechanism is stabilizer counting. Stabilizers fully supported on $Y(\rho)=1$0 commute through $Y(\rho)=1$1 and contribute coherently to $Y(\rho)=1$2. For a bipartite stabilizer with entanglement $Y(\rho)=1$3, the number of stabilizers supported on $Y(\rho)=1$4 is $Y(\rho)=1$5. After normalization by $Y(\rho)=1$6, this yields the factor $Y(\rho)=1$7. In the operator-spreading picture, a Pauli on $Y(\rho)=1$8 conjugated by $Y(\rho)=1$9 proliferates into many Pauli strings, while the $\rho$0 logical pairs between $\rho$1 and $\rho$2 make that mixing extend into $\rho$3. The result is delocalization of locally injected magic across the full system.

A second complementarity appears when independent Haar random unitaries act on both sides. For $\rho$4 and $\rho$5,

$\rho$6

Because the Haar-random pure-state value is $\rho$7, the deviation from global Haar magic is controlled solely by $\rho$8. Numerically, $\rho$9 already saturates to the Haar value for all $Y(\rho)<1$0. In this regime, entanglement suppresses the residual gap left by locality, so a product local unitary can act “as if” it were global Haar on the metric $Y(\rho)<1$1 (Hou et al., 26 Mar 2025).

The same qualitative structure extends beyond the bipartite stabilizer case. For tripartite stabilizer states, the relevant data are the GHZ count $Y(\rho)<1$2, pairwise Bell counts $Y(\rho)<1$3, and local singles. Under $Y(\rho)<1$4, the leading exponent is controlled by the acted region $Y(\rho)<1$5 together with the entanglement across $Y(\rho)<1$6, while the product-vs-global gap is suppressed by $Y(\rho)<1$7. Non-stabilizer entanglement built from imperfect Bell pairs, and shallow brickwork circuits replacing Haar $Y(\rho)<1$8, preserve the same qualitative conclusion: increasing total entanglement enhances global magic spreading, although finite-depth architectures exhibit saturation effects at very small depths (Hou et al., 26 Mar 2025).

3. Distinct dynamical regimes and phase structure

In hybrid Clifford+$Y(\rho)<1$9 circuits with measurements, complementarity takes the form of separated critical behavior. The architecture is a 1D chain of $M_2(\rho)=-\log Y(\rho)$0 qubits with a brickwork pattern of random two-qubit Clifford gates on nearest neighbors, projective measurements in the computational basis with probability $M_2(\rho)=-\log Y(\rho)$1 per qubit per time step, and $M_2(\rho)=-\log Y(\rho)$2-gate injections with probability $M_2(\rho)=-\log Y(\rho)$3. For the principal case $M_2(\rho)=-\log Y(\rho)$4, entanglement undergoes the usual monitored-induced transition at

$M_2(\rho)=-\log Y(\rho)$5

while magic collapses only at a higher critical rate,

$M_2(\rho)=-\log Y(\rho)$6

for $M_2(\rho)=-\log Y(\rho)$7. The regime $M_2(\rho)=-\log Y(\rho)$8 is diagnostic: entanglement already shows area-law scaling, while magic remains (sub)-extensive. The resulting intermediate phase demonstrates that entanglement alone does not diagnose nonstabilizerness or computational hardness in monitored dynamics (Fux et al., 2023).

A related decoupling appears in gauge theory. In the $M_2(\rho)=-\log Y(\rho)$9-dimensional SU(2) lattice gauge theory formulated in a dressed-site basis, gauge-invariant entanglement entropy $$Y_\alpha(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^{2\alpha}, \qquad M_\alpha(\rho) := \frac{1}{1-\alpha}\log Y_\alpha(\rho)$$0 and stabilizer Rényi entropy $$Y_\alpha(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^{2\alpha}, \qquad M_\alpha(\rho) := \frac{1}{1-\alpha}\log Y_\alpha(\rho)$$1 were computed for systems up to $$Y_\alpha(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^{2\alpha}, \qquad M_\alpha(\rho) := \frac{1}{1-\alpha}\log Y_\alpha(\rho)$$2 ($$Y_\alpha(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^{2\alpha}, \qquad M_\alpha(\rho) := \frac{1}{1-\alpha}\log Y_\alpha(\rho)$$3 qubits). At the CFT point $$Y_\alpha(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^{2\alpha}, \qquad M_\alpha(\rho) := \frac{1}{1-\alpha}\log Y_\alpha(\rho)$$4, the entanglement scaling yields $$Y_\alpha(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^{2\alpha}, \qquad M_\alpha(\rho) := \frac{1}{1-\alpha}\log Y_\alpha(\rho)$$5, while magic obeys

$$Y_\alpha(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^{2\alpha}, \qquad M_\alpha(\rho) := \frac{1}{1-\alpha}\log Y_\alpha(\rho)$$6

with $$Y_\alpha(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^{2\alpha}, \qquad M_\alpha(\rho) := \frac{1}{1-\alpha}\log Y_\alpha(\rho)$$7, $$Y_\alpha(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^{2\alpha}, \qquad M_\alpha(\rho) := \frac{1}{1-\alpha}\log Y_\alpha(\rho)$$8, and $$Y_\alpha(\rho) := 2^{-N} \sum_{P\in P_N} \operatorname{tr}(P\rho)^{2\alpha}, \qquad M_\alpha(\rho) := \frac{1}{1-\alpha}\log Y_\alpha(\rho)$$9. At $AB$0, the entanglement entropy decreases monotonically with $AB$1, its sharpest change occurs near $AB$2, and $AB$3 shows a broad plateau before dropping rapidly only after the same crossover. The strong-coupling limit $AB$4 is a product stabilizer state with both $AB$5 and $AB$6. The intermediate region is therefore “magic-rich but low-entanglement” rather than maximally nonclassical in both senses (Jha et al., 8 Jun 2026).

The light-front formulation of the $AB$7-dimensional transverse-field Ising model provides a sharper basis-dependent contrast. In instant-form momentum space, the ground state is a product over $AB$8 of two-mode BCS-like states,

$AB$9

which carry pairwise entanglement between $E=-\operatorname{tr}(\rho_A\log_2\rho_A)$0 and $E=-\operatorname{tr}(\rho_A\log_2\rho_A)$1. The light-front Hamiltonian is diagonal in $E=-\operatorname{tr}(\rho_A\log_2\rho_A)$2, so the light-front ground state is the separable Fock vacuum $E=-\operatorname{tr}(\rho_A\log_2\rho_A)$3. Away from criticality, the instant-form ground state has nonzero momentum-space magic,

$E=-\operatorname{tr}(\rho_A\log_2\rho_A)$4

At the quantum critical point, however, both ground states are stabilizers: the instant-form state is a product of maximally entangled Bell-like pairs in momentum space, whereas the light-front ground state remains separable. High entanglement with zero magic is therefore possible, and the quantization scheme itself determines which resource is used (Alterman et al., 14 Jul 2025).

4. Dissipative complementarity and reborn magic

Under local amplitude damping, magic and entanglement respond in qualitatively different ways because separability is preserved by local CPTP maps, whereas stabilizer membership is not. For the $E=-\operatorname{tr}(\rho_A\log_2\rho_A)$5-qubit GHZ family

$E=-\operatorname{tr}(\rho_A\log_2\rho_A)$6

the evolved state under $E=-\operatorname{tr}(\rho_A\log_2\rho_A)$7 has only one surviving off-diagonal GHZ coherence $E=-\operatorname{tr}(\rho_A\log_2\rho_A)$8. On the real GHZ–X slice, stabilizer membership is exact: $E=-\operatorname{tr}(\rho_A\log_2\rho_A)$9 This produces two stabilizer thresholds: the lower entry $2^N$00, defined by $2^N$01, and the upper exit

$2^N$02

Magic is absent exactly on $2^N$03 and present on the two open branches $2^N$04 (Cao, 21 May 2026).

Entanglement, by contrast, exhibits sudden death at a single threshold independent of the bipartition: $2^N$05 At this same point the state is fully separable. The central complementarity identity is therefore

$2^N$06

valid for every $2^N$07 in the re-entrant regime $2^N$08. Its origin is the system–environment duality of amplitude damping,

$2^N$09

which mirrors the system-side magic-rebirth condition against the environment-side entanglement-death condition. In Regimes I and II of the ordering analysis, the entire reborn branch lies in fully separable states, and all proper marginals are stabilizer. Reborn magic is then nonlocal but not entanglement-bearing (Cao, 21 May 2026).

This nonlocal reborn magic can nevertheless be concentrated. Measuring the $2^N$10 commuting parity stabilizers $2^N$11 and postselecting the trivial syndrome projects onto the GHZ subspace with success probability

$2^N$12

A subsequent CNOT cascade and discarding of spectators yields a single-qubit state $2^N$13. The extraction is lossless in expectation for the robustness of magic: $2^N$14 For sufficiently large $2^N$15, the decoded branch enters the standard distillable regions for $2^N$16-type and $2^N$17-type protocols (Cao, 21 May 2026).

The same dissipative framework also splits pure stabilizer inputs into magic-generators and magic-insulators. Under homogeneous local amplitude damping, a pure stabilizer state is an insulator iff its computational-basis support has constant Hamming weight; otherwise it is a generator. At two qubits, the Bell states

$2^N$18

exemplify the split: $2^N$19 exits the stabilizer polytope immediately for any $2^N$20, whereas $2^N$21 remains stabilizer for all $2^N$22 (Cao, 21 May 2026).

5. Operator-space, complexity, and algorithmic consequences

In the Heisenberg picture, complementarity becomes a hierarchy of rigorous inequalities. For an $2^N$23-qubit unitary $2^N$24, a nontrivial Pauli $2^N$25, and a bipartition $2^N$26, the local operator entanglement of the Heisenberg-evolved operator $2^N$27 satisfies

$2^N$28

where $2^N$29 is the operator stabilizer Rényi entropy, $2^N$30 is unitary nullity, and $2^N$31 is the $2^N$32-count. Thus large LOE is impossible without large magic in operator space. For random ensembles, the same quantities nearly coincide on average. For the $2^N$33-doped Clifford ensemble $2^N$34,

$2^N$35

and for the $2^N$36-compressible ensemble $2^N$37,

$2^N$38

The implication is operational: any dynamics that are hard for tensor-network simulation because LOE is large must also be hard for stabilizer and Pauli-truncation methods (Dowling et al., 30 Jan 2025).

A closely related result states that a unitary map generates nonlocal magic if and only if it generates operator entanglement on Pauli strings. On that basis, an average measure of a unitary’s Pauli-entangling power is introduced as a proxy for nonlocal magic generation, with analytical formulas, a typical value, and upper bounds in terms of the nonstabilizerness properties of the evolution (Andreadakis et al., 12 Apr 2025).

The computational consequences in the Schrödinger picture are more asymmetric. Using stabilizer nullity $2^N$39 as the magic order parameter, Hilbert space can be partitioned into an entanglement-dominated (ED) phase, where the entanglement $2^N$40 significantly exceeds the state’s magic, and a magic-dominated (MD) phase, where magic dominates entanglement. For pure states with nullity $2^N$41, one has

$2^N$42

and more generally

$2^N$43

In the ED phase there are sample- and time-efficient, input-agnostic quantum algorithms for entanglement estimation, distillation, and dilution, with near-reversible resource conversion. In the MD phase, constant-relative-error estimation and constant-fraction distillation are provably intractable in general. This establishes a computational phase separation rather than a simple resource inequality (Gu et al., 2024).

6. Two-qubit frontier geometry and general lessons

For pure two-qubit states, the interplay between concurrence $2^N$44 and stabilizer Rényi-2 magic $2^N$45 can be solved exactly. The minimal-magic frontier is a single continuous curve,

$2^N$46

which vanishes at $2^N$47 and $2^N$48 and is strictly positive for $2^N$49. A Schmidt-family representative,

$2^N$50

saturates this lower boundary. Partially entangled two-qubit pure states therefore cannot have zero magic when $2^N$51 is the chosen nonstabilizerness measure (Roman et al., 26 Mar 2026).

The maximal-magic frontier is piecewise and has three segments: $2^N$52 where $2^N$53. The global maximum $2^N$54 occurs at two distinct entanglement values, $2^N$55 and $2^N$56. At $2^N$57, the upper boundary gives $2^N$58, while Bell stabilizers still realize $2^N$59. Entanglement is therefore neither a monotone lower bound nor a monotone upper bound for magic, even in the smallest nontrivial bipartite system (Roman et al., 26 Mar 2026).

Taken together, the contemporary literature supports a plural notion of magic–entanglement complementarity. In some settings, entanglement is a conduit that exponentially enhances the global spread of local magic. In others, entanglement and magic separate into distinct phases, distinct critical points, or even opposite responses to noise. In operator space, magic bounds entanglement from above; in two-qubit geometry, the feasible region is bounded by exact Pareto frontiers rather than a single curve. A plausible implication is that any general theory of quantum resources must track not only how much entanglement and magic are present, but also where they reside, how they are generated, and which operational task probes them.