Magic Witness in Quantum Computation

• Magic witness is a criterion that verifies a quantum state is non-stabilizer (i.e., has 'magic') by lying outside the stabilizer polytope, making it essential for universal fault-tolerant quantum computation.
• The approach includes diverse paradigms such as Hermitian operator tests, nonlinear multicopy functionals, and operational thermodynamic measures, each balancing experimental efficiency with detection fidelity.
• These frameworks not only enable practical magic state distillation and resource quantification but also highlight computational hardness in exact verification while offering efficient estimability in realistic quantum platforms.

A magic witness is a criterion—most commonly a Hermitian operator, but in recent work also a nonlinear multicopy functional or an operational thermodynamic test—that certifies that a quantum state lies outside the stabilizer polytope and therefore possesses nonstabilizerness, or “magic,” as a resource for universal fault-tolerant quantum computation. In the stabilizer resource theory, free states are convex mixtures of pure stabilizer states, while magic states are precisely those that are not such mixtures. Recent arXiv work has developed several distinct witness paradigms: facet and hyperplane witnesses derived from the geometry of the stabilizer polytope, stabilizer-Rényi and polynomial multicopy witnesses, triangle inequalities among stabilizer fidelities, thermodynamic witnesses based on energy or heat, and operational witnesses tied to measurement-based quantum computation (Warmuz et al., 2024, Haug et al., 25 Apr 2025, Liu et al., 18 Dec 2025, Macêdo et al., 9 Apr 2026, Tang et al., 25 Jun 2026).

1. Resource-theoretic definition and sign conventions

Let $H=(\mathbb C^2)^{\otimes n}$ and let

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$

denote the stabilizer polytope. Mixed stabilizer states are exactly the elements of this convex hull, and a state has magic iff it is not a stabilizer state (Haug et al., 25 Apr 2025, Leone et al., 25 Feb 2026).

In its most standard linear form, a magic witness is a Hermitian operator $W\in\mathrm{Herm}(H)$ satisfying two conditions: $$\mathrm{Tr}[W\sigma]\le 0\quad \forall \sigma\in S_n,$$ and

$$\exists \rho\notin S_n \text{ such that } \mathrm{Tr}[W\rho]>0.$$

Equivalently, $W$ defines a separating hyperplane for the stabilizer polytope and detects at least one non-stabilizer state (Leone et al., 25 Feb 2026). This convention is not universal. Some papers adopt the opposite sign, requiring nonnegative expectation on free states and negative expectation on detected magic states; others use a standardized form

$\widetilde W=I-\frac{W}{\|W\|_\infty},$

with $\mathrm{Tr}[\widetilde W\,\sigma]\le 1$ for all stabilizer states and $\mathrm{Tr}[\widetilde W\,\rho]>1$ for some magic $\rho$ (Tang et al., 25 Jun 2026). The underlying geometric content is the same: a witness separates free from resourceful states.

The term “magic witness” has therefore broadened beyond single-copy linear observables. In current usage it includes nonlinear criteria based on polynomial functions of $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$0, multicopy expectation values, or restricted operational tasks such as energy or heat exchange, provided that violation is guaranteed to imply nonstabilizerness (Haug et al., 25 Apr 2025, Macêdo et al., 9 Apr 2026, Tang et al., 25 Jun 2026). A recurrent distinction is between faithful witnesses, which detect all and only magic states, and merely sufficient witnesses, which certify magic when violated but may leave some magic states undetected.

2. Facet witnesses from the stabilizer polytope

Warmuz et al. formulate the witness problem directly in the space of Pauli-string expectation values. For an $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$1-qubit state $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$2, label the $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$3 Pauli strings by $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$4 with $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$5, and define

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$6

Because $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$7, one has $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$8, and the remaining $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$9 coordinates specify $W\in\mathrm{Herm}(H)$0 completely (Warmuz et al., 2024).

In this representation, the stabilizer polytope is

$W\in\mathrm{Herm}(H)$1

where the pure stabilizer vertices satisfy $W\in\mathrm{Herm}(H)$2. Every facet can be written as an integer-coefficient hyperplane inequality

$W\in\mathrm{Herm}(H)$3

with $W\in\mathrm{Herm}(H)$4 and $W\in\mathrm{Herm}(H)$5. By construction, every stabilizer mixture satisfies all such inequalities, and conversely any point inside their intersection is a stabilizer mixture. Hence

$W\in\mathrm{Herm}(H)$6

This gives a necessary and sufficient condition for both pure and mixed states (Warmuz et al., 2024).

Each facet induces a witness operator

$W\in\mathrm{Herm}(H)$7

for which

$W\in\mathrm{Herm}(H)$8

on all stabilizer states, while $W\in\mathrm{Herm}(H)$9 certifies magic. Because $$\mathrm{Tr}[W\sigma]\le 0\quad \forall \sigma\in S_n,$$0 is a weighted sum of Pauli operators, its evaluation reduces to measuring the relevant Pauli strings, multiplying by the integer coefficients $$\mathrm{Tr}[W\sigma]\le 0\quad \forall \sigma\in S_n,$$1, summing, and comparing to $$\mathrm{Tr}[W\sigma]\le 0\quad \forall \sigma\in S_n,$$2. A single hyperplane with positive violation is sufficient for certification (Warmuz et al., 2024).

The same geometric construction yields a faithful magic monotone. In one form,

$$\mathrm{Tr}[W\sigma]\le 0\quad \forall \sigma\in S_n,$$3

with $$\mathrm{Tr}[W\sigma]\le 0\quad \forall \sigma\in S_n,$$4 chosen so that $$\mathrm{Tr}[W\sigma]\le 0\quad \forall \sigma\in S_n,$$5 on stabilizer states and $$\mathrm{Tr}[W\sigma]\le 0\quad \forall \sigma\in S_n,$$6 on magic states. Equivalent formulations are

$$\mathrm{Tr}[W\sigma]\le 0\quad \forall \sigma\in S_n,$$7

and the Minkowski-functional expression

$$\mathrm{Tr}[W\sigma]\le 0\quad \forall \sigma\in S_n,$$8

The monotone is faithful, Clifford-invariant, convex, and monotone under stabilizer channels (Warmuz et al., 2024).

For one qubit, the stabilizer polytope is the Bloch octahedron

$$\mathrm{Tr}[W\sigma]\le 0\quad \forall \sigma\in S_n,$$9

with eight facets

$$\exists \rho\notin S_n \text{ such that } \mathrm{Tr}[W\rho]>0.$$0

A corresponding witness is $$\exists \rho\notin S_n \text{ such that } \mathrm{Tr}[W\rho]>0.$$1, and $$\exists \rho\notin S_n \text{ such that } \mathrm{Tr}[W\rho]>0.$$2 is equivalent to $$\exists \rho\notin S_n \text{ such that } \mathrm{Tr}[W\rho]>0.$$3. For the pure $$\exists \rho\notin S_n \text{ such that } \mathrm{Tr}[W\rho]>0.$$4 state,

$$\exists \rho\notin S_n \text{ such that } \mathrm{Tr}[W\rho]>0.$$5

one has $$\exists \rho\notin S_n \text{ such that } \mathrm{Tr}[W\rho]>0.$$6 and therefore $$\exists \rho\notin S_n \text{ such that } \mathrm{Tr}[W\rho]>0.$$7 (Warmuz et al., 2024).

3. Entropic and polynomial multicopy witnesses

Haug and Tarabunga introduce a witness family based on the Pauli spectrum of a mixed state. For $$\exists \rho\notin S_n \text{ such that } \mathrm{Tr}[W\rho]>0.$$8,

$$\exists \rho\notin S_n \text{ such that } \mathrm{Tr}[W\rho]>0.$$9

and the $W$0-Rényi entropy is

$W$1

The mixed-state stabilizer-Rényi entropy is

$W$2

and the corresponding witness is

$W$3

Equivalently, $W$4. In the pure-state limit $W$5, $W$6 recovers the standard stabilizer-Rényi entropy (Haug et al., 25 Apr 2025).

These witnesses are Clifford-invariant and additive. Their range is

$W$7

For any $W$8, the key implication is: $W$9 They also satisfy the hierarchy

$\widetilde W=I-\frac{W}{\|W\|_\infty},$0

where

$\widetilde W=I-\frac{W}{\|W\|_\infty},$1

is the stabilizer norm, and

$\widetilde W=I-\frac{W}{\|W\|_\infty},$2

Thus $\widetilde W=I-\frac{W}{\|W\|_\infty},$3 gives a quantitative lower bound on log-free robustness and, via $\widetilde W=I-\frac{W}{\|W\|_\infty},$4, also constrains stabilizer-fidelity distance (Haug et al., 25 Apr 2025).

A central advantage is efficient estimability. For odd integer $\widetilde W=I-\frac{W}{\|W\|_\infty},$5, $\widetilde W=I-\frac{W}{\|W\|_\infty},$6 can be estimated by grouping $\widetilde W=I-\frac{W}{\|W\|_\infty},$7 copies into $\widetilde W=I-\frac{W}{\|W\|_\infty},$8 Bell pairs, applying local Bell unitaries, measuring all qubits in the $\widetilde W=I-\frac{W}{\|W\|_\infty},$9 basis, combining parities into a $\mathrm{Tr}[\widetilde W\,\sigma]\le 1$0 outcome, and averaging over $\mathrm{Tr}[\widetilde W\,\sigma]\le 1$1 repetitions. The complexity is

$\mathrm{Tr}[\widetilde W\,\sigma]\le 1$2

copies, $\mathrm{Tr}[\widetilde W\,\sigma]\le 1$3 depth, and

$\mathrm{Tr}[\widetilde W\,\sigma]\le 1$4

classical time. Under the promise $\mathrm{Tr}[\widetilde W\,\sigma]\le 1$5, estimating $\mathrm{Tr}[\widetilde W\,\sigma]\le 1$6 yields a poly$\mathrm{Tr}[\widetilde W\,\sigma]\le 1$7-copy property test that distinguishes low-magic from high-magic states in the sense of $\mathrm{Tr}[\widetilde W\,\sigma]\le 1$8 and $\mathrm{Tr}[\widetilde W\,\sigma]\le 1$9 (Haug et al., 25 Apr 2025).

The same paper applies $\mathrm{Tr}[\widetilde W\,\rho]>1$0 to noisy $\mathrm{Tr}[\widetilde W\,\rho]>1$1-gate certification and to experiment. For

$\mathrm{Tr}[\widetilde W\,\rho]>1$2

with $\mathrm{Tr}[\widetilde W\,\rho]>1$3 any mixed unital Clifford channel and $\mathrm{Tr}[\widetilde W\,\rho]>1$4, measuring $\mathrm{Tr}[\widetilde W\,\rho]>1$5 efficiently decides whether $\mathrm{Tr}[\widetilde W\,\rho]>1$6 or $\mathrm{Tr}[\widetilde W\,\rho]>1$7. On the IonQ quantum computer, the Bell-measurement protocol was used on $\mathrm{Tr}[\widetilde W\,\rho]>1$8-qubit random Clifford circuits interleaved with $\mathrm{Tr}[\widetilde W\,\rho]>1$9 $\rho$0-gates, and even with estimated depolarizing $\rho$1, one found $\rho$2 for all $\rho$3 (Haug et al., 25 Apr 2025).

A distinct multicopy route is the witness-expansion framework. Starting from a standardized seed witness $\rho$4, one twirls it over the Clifford group $\rho$5 and defines, for positive integer $\rho$6,

$\rho$7

If

$\rho$8

with

$\rho$9

then $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$00 is magic. For single qubits, choosing the $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$01-state projector as seed produces the degree-$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$02 polynomial

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$03

Stabilizer states satisfy $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$04, while the pure $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$05 state gives $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$06, so

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$07

This criterion detects every pure single-qubit non-stabilizer state (Tang et al., 25 Jun 2026).

4. Triangle witnesses, fidelity witnesses, and distillation

The Triangle Criterion introduces a different linear witness family based on triples of stabilizer fidelities. Choose three pure stabilizer states $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$08 satisfying

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$09

If

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$10

then $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$11 is magic. For single qubits this condition is necessary and sufficient, and for arbitrary-qubit pure states it detects all nonstabilizers (Liu et al., 18 Dec 2025).

For one qubit, the geometric interpretation is exact. The stabilizer states form the regular octahedron in the Bloch ball, and each facet is cut out by an equation of the form

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$12

with $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$13 three pairwise $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$14-overlapping stabilizer vertices. Violation of such a facet inequality places the state outside the octahedron and therefore outside the stabilizer set (Liu et al., 18 Dec 2025). This makes the Triangle Criterion conceptually close to the facet witnesses of the stabilizer polytope, but it is expressed directly through overlaps with three stabilizer projectors rather than through a full facet enumeration.

The same work contrasts triangle witnesses with fidelity-based witnesses of the form

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$15

where $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$16 is the maximal stabilizer fidelity with the chosen pure magic state $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$17. Such witnesses only certify a mixed state if its fidelity with $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$18 exceeds the stabilizer threshold $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$19. By contrast, each triangle witness

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$20

has unit trace and is rank-$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$21, is nonnegative on all stabilizers, and can be negative for many mixed states. Numerics reported in the paper show that triangle witnesses detect a much larger fraction of multi-qubit mixed states than any single-pure-state fidelity witness (Liu et al., 18 Dec 2025).

A particularly strong result is the distillation equivalence: $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$22 Thus an $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$23-qubit state can be converted by Clifford unitaries, Pauli measurements, and stabilizer ancillas into a single-qubit magic state iff it violates the Triangle Criterion (Liu et al., 18 Dec 2025). The paper further shows non-tensor-stability: a two-qubit $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$24 may fail all triangle tests while $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$25 violates one, yielding a genuinely two-copy distillation protocol. In that sense, multi-qubit distillation is strictly more powerful than all single-qubit schemes.

The same framework also identifies “unfaithful” mixed magic states. Using a minimal-purity result that approaches $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$26, the paper constructs states of the form

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$27

for traceless $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$28 with $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$29, such that for all pure $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$30,

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$31

so $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$32 evades every fidelity-based witness despite being magical, namely despite violating some Triangle inequality (Liu et al., 18 Dec 2025). This establishes a precise limitation of fidelity witnesses in the mixed-state regime.

5. Operational witnesses: thermodynamics and measurement-based computation

One operational approach replaces state reconstruction by physically accessible observables. Given an $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$33-qubit Hamiltonian $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$34, define the stabilizer ground-state energy

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$35

the true ground-state energy

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$36

and the stabilizer gap

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$37

If $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$38, no stabilizer can reach the true ground energy. Therefore any state $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$39 satisfying

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$40

cannot be stabilizer. Equivalently, one may define

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$41

which obeys $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$42 for all stabilizer states, while $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$43 certifies magic (Macêdo et al., 9 Apr 2026). This is a linear witness with the opposite sign convention from that used in some geometric formulations.

The same paper introduces a nonlinear heat witness. The unknown system $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$44 is coupled to a thermal ancilla $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$45 at inverse temperature $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$46 by an energy-preserving unitary $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$47, with an auxiliary memory $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$48 that must return to its initial state. Define the optimal cooling and heating heats,

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$49

under energy conservation and memory-catalysis. The extrema are obtained by mapping $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$50 to a Gibbs state $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$51 at an effective inverse temperature $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$52 solving

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$53

where

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$54

If one conditions on a measured average energy $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$55, then within the energy slice

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$56

one obtains a stabilizer heat window

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$57

Any observed pair $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$58 outside this interval certifies magic (Macêdo et al., 9 Apr 2026).

These witnesses need not be tight, but they can detect magic where direct energy measurements fail. The paper gives several examples. For the depolarized $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$59-state family under a Hamiltonian $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$60, the energy witness is inconclusive, but the heat witness detects magic up to the true stabilizer threshold $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$61. For a dephased $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$62-state under $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$63, the energy witness again fails, while the heat witness detects magic for $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$64. For the transverse-field Ising chain

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$65

one has

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$66

while the exact ground-state energy density is

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$67

so the stabilizer gap peaks at the critical point $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$68 (Macêdo et al., 9 Apr 2026).

A different operational notion appears in measurement-based quantum computation. In that setting, non-Pauli single-qubit measurements on an initial graph state inject magic. “Invested magic” is defined by a $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$69-decomposition of a target unitary $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$70: $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$71 while the “potential magic” of a graph state $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$72 is

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$73

The authors describe invested magic as serving both as a witness of magic resources and an upper bound for the realization of a desired unitary transformation (Li et al., 2024). For the $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$74-qubit quantum Fourier transform,

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$75

giving an explicit MQC-based upper bound on the $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$76-count. They also show

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$77

and experimentally demonstrate the framework in a four-photon setup (Li et al., 2024). This suggests that, within MQC, “witness” can also denote a resource-accounting quantity that certifies the presence and flow of injected non-Cliffordness.

6. Computational hardness, scope, and limitations

The existence of separating hyperplanes follows from convex geometry, but deciding whether a candidate operator is a valid magic witness is computationally difficult in the strongest sense currently known. The decision problem “Magic-Witness-Validity” asks, given $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$78, a Hermitian matrix $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$79 of dimension $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$80, and a precision parameter $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$81, whether

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$82

or instead

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$83

under the promise that one of these holds. Under the Exponential Time Hypothesis, any algorithm distinguishing these cases must run in time at least

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$84

even with inverse-polynomial gap $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$85 (Leone et al., 25 Feb 2026).

The proof sketch reduces $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$86-SAT on $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$87 Boolean variables to Magic-Witness-Validity by encoding assignments into stabilizer states, associating each clause with a local Hermitian penalty operator, and setting

$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$88

Then the sign of $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$89 distinguishes satisfiable from unsatisfiable formulas, implying super-exponential hardness in the number of qubits (Leone et al., 25 Feb 2026).

Several consequences follow. First, there is no fast universal witness constructor: no $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$90-time procedure can, for arbitrary $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$91, either output a valid witness with positive detection or certify $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$92. Second, every faithful mixed-state magic monotone is super-exponentially hard to compute to inverse-polynomial precision. Third, deciding whether a proposed operator is a valid witness is itself as hard as the underlying polytope-membership problem (Leone et al., 25 Feb 2026). The paper further states that experimentally implementable few-term Pauli witnesses can only detect restricted subsets of magic states, and that no single or small family of witnesses can cover all non-stabilizer states without incurring super-exponential complexity in $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$93.

This hardness does not negate the practical value of structured witnesses. Rather, it clarifies their scope. Facet searches with symmetry reduction can be efficient for small $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$94, with Warmuz et al. reporting runtimes $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$95 on a desktop up to $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$96, and optimization over $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$97 integer coefficients rather than over $S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$98 stabilizer-decomposition variables (Warmuz et al., 2024). Entropic, multicopy, and thermodynamic witnesses are similarly valuable because they replace universal exact membership testing by efficiently measurable sufficient criteria (Haug et al., 25 Apr 2025, Macêdo et al., 9 Apr 2026). A plausible implication is that the modern theory of magic witnesses is best understood not as a search for one universally efficient detector, but as a hierarchy of specialized criteria trading faithfulness, operational accessibility, and computational tractability in different ways.

7. Conceptual synthesis

Across these developments, a magic witness is not a single object but a family of certification paradigms tied together by the same resource-theoretic boundary: the stabilizer polytope. The hyperplane witnesses of Warmuz et al. provide an exact polyhedral picture and a faithful mixed-state monotone (Warmuz et al., 2024). The stabilizer-Rényi witnesses of Haug and Tarabunga give efficient sufficient criteria, quantitative lower bounds on robustness-like monotones, and poly$S_n=\mathrm{Conv}\{\,|\psi\rangle\langle\psi|:\ |\psi\rangle\ \text{a pure %%%%1%%%%-qubit stabilizer state}\,\}$99-copy testing under bounded $W\in\mathrm{Herm}(H)$00 (Haug et al., 25 Apr 2025). The Triangle Criterion provides a particularly compact family of linear inequalities with an exact single-qubit characterization and an operational equivalence to single-qubit magic distillation (Liu et al., 18 Dec 2025). Thermodynamic witnesses show that energy and heat alone can certify nonstabilizerness in physically motivated settings (Macêdo et al., 9 Apr 2026). Witness expansion unifies multicopy nonlinear detection by twirling seed witnesses over the Clifford group and recovers explicit polynomial criteria for qubit and qudit magic (Tang et al., 25 Jun 2026). Measurement-based constructions reinterpret witness notions in terms of invested and potential magic resources (Li et al., 2024).

Two general lessons recur. First, witness strength depends strongly on the target regime: some constructions are faithful for all mixed states, some only for pure states or single qubits, and some are intentionally coarse but experimentally lightweight. Second, recent work has made the limitations mathematically explicit: fidelity witnesses can miss “unfaithful” mixed magic, and exact universal witness validation is super-exponentially hard under ETH (Liu et al., 18 Dec 2025, Leone et al., 25 Feb 2026). In that sense, the study of magic witnesses has evolved from isolated detection tricks into a systematic interface between convex geometry, multicopy estimation, thermodynamics, distillation theory, and computational complexity.