Skip to content
OpenContext
Contribution toProblem 5: Test problem: Hilbert–Schmidt inradius of the multiqubit stabilizer polytopeunder revision
0

Partial resultOpenContext 5.11

An improved computer-assisted bound at 0.4996 for all $n$, and a sharp 0.5 value for $n\leq 5$ (part I)

by Rafael Wagner · accepted 19 Nov 2026 · partially AI-assisted

Main result

After a couple of prompts, Claude Fable 5.2 found another advance:

Theorem (0.49960.4996 bound): Let ρ∈D(C2⊗⋯⊗C2⏟n times)\rho \in \mathcal{D}(\underbrace{\mathbb{C}^2 \otimes \cdots \otimes \mathbb{C}^2}_{n\text{ times}}) be an arbitrary nn-qubit state, and denote STABn\mathrm{STAB}_n the convex polytope of nn-qubit stabilizer states. For every integer n≥1n \geq 1, it holds that

Tr[ρ2]≤1dn−a∗  ⟹  ρ∈STABn,\mathrm{Tr}[\rho^2] \leq \frac{1}{d_n-a^*} \implies \rho \in \mathrm{STAB}_n,

where dn=2nd_n = 2^n is the dimension of the underlying Hilbert space and a∗=0.4996a^*=0.4996. Furthermore, for 1≤n≤51\leq n \leq 5 the conjecture is shown to hold, i.e., a∗=12a^*=\tfrac{1}{2}.

The results have used the strategy present in the notes by Y. Wang. "Partial Results on the Hilbert–Schmidt Inradius of the Multiqubit Stabilizer Polytope", in collaboration with ChatGPT 6 Astra.

How this contribution is organised. Because of its length, the argument is divided into two parts. This first part contains the preliminaries and two new results on the spectrum of the operators in the Λ\Lambda-polytope: the theorem on the spectrum of a witness and the star inequality. Both are proved analytically. The second part combines them with the Liu–Liu inequality of Ref. [3] to prove the theorem above.

Preliminary mathematical ideas

Recall that a crucial ingredient in this discussion is the notion of the Λ\Lambda-polytope from Ref. [4]. For more on Λ\Lambda-polytopes we refer to Refs. [8, 9, 10, 11, 4]. It can be defined as follows:

Λn={A=A†: Tr[A]=1, ⟨s∣A∣s⟩≥0 for all stabilizer states ∣s⟩}.\Lambda_n=\left \{A=A^\dagger:\ \mathrm{Tr}[A]=1,\ \langle s|A|s\rangle\ge 0\ \text{for all stabilizer states } \vert s\rangle \right\}.

Above the relevant operators are always defined relative to nn-qubit Hilbert spaces C2⊗⋯⊗C2\mathbb{C}^2 \otimes \cdots \otimes \mathbb{C}^2 of dimension dn=2nd_n = 2^n. Below, all sums taken over Paulis PP run over the d2−1d^2-1 Pauli operators which are different from the identity I\mathbb{I}. For such Pauli PP, we also denote

Π±=(I±P)/2,\Pi_\pm=(\mathbb{I}\pm P)/2,

aP=Tr[PA]a_P=\mathrm{Tr}[PA] as the trace of that Pauli with an operator AA, and B±=Π±AΠ±B_\pm=\Pi_\pm A\Pi_\pm. Let A1=(I+X+Y+Z)/2A_1=(\mathbb{I}+X+Y+Z)/2, with eigenvalues λ±=(1±3)/2\lambda_\pm=(1\pm\sqrt3)/2, and A⋆=A1⊗∣0…0⟩⟨0…0∣A^\star=A_1\otimes|0\dots0\rangle\langle0\dots0|. Throughout this document we define

v:=λ+,c:=−λ−=v−1.v:=\lambda_+,\qquad c:=-\lambda_-=v-1.

With these definitions, it is simple to show that the following equations all hold:

2v2=2v+1,vc=12,v2+c2=2,v3−c3=52.\qquad 2v^2=2v+1,\quad vc=\tfrac12,\quad v^2+c^2=2,\quad v^3-c^3=\frac52.

Building on my earlier contribution, we recall that we are using ideas related to the stabilizer theory, which is tied to quantum error correction. Because of that, the language of "codes" is used below, without loss of generality.

Lemma (Block lemma). For an arbitrary Pauli PP the range of Π±\Pi_\pm is an (n−1)(n-1)-qubit code space whose stabilizer states are nn-qubit stabilizer states. Hence B±B_\pm is non-negative on them,

Tr[B±]=(1±aP)/2≥0,\mathrm{Tr}[B_\pm]=(1\pm a_P)/2\ge0,

with aP=Tr[PA]a_P=\mathrm{Tr}[PA], and

B±/Tr[B±]∈Λn−1.B_\pm/\mathrm{Tr}[B_\pm]\in\Lambda_{n-1}.

As a specific case, if Tr[B±]=0\mathrm{Tr}[B_\pm]=0 then B±=0B_\pm=0.

This lemma is shown in Ref. [4] (their Lemma 1, part (2)) and discussed also in Ref. [3], Appendix A (see the "Pauli-compression recursion" subsection).

The following is another simple Lemma, which can be found, for example, in standard textbooks [7].

Lemma (a rule for Pauli sums). For arbitrary Hermitian operators X,YX,Y, with the identity included in the sums, we have that

∑PPXP=d Tr[X] I\sum_P PXP = d\, \mathrm{Tr}[X]\,\mathbb{I}

and

∑PTr[PX]Tr[PY]=d Tr[XY].\sum_P\mathrm{Tr}[PX]\mathrm{Tr}[PY]=d\,\mathrm{Tr}[XY].

The first rule in the above lemma says that conjugation by a uniformly random Pauli operator is the completely depolarizing channel, i.e. that the Pauli group is a unitary 1-design: Watrous, Eq. (4.109); Ref. [12]; for one qubit it is the fully depolarizing case of the quantum depolarizing channel written with the Kraus operators I,X,Y,Z\mathbb{I},X,Y,Z.

The second rule is Parseval's identity for the Pauli operators, which form an orthogonal basis of the space of operators for the Hilbert–Schmidt inner product: see Pauli matrices ("Algebraic properties") for one qubit, and J. Watrous [7], Section 4.1.2, Eq. (4.78) (equation numbers as in the author's online version), where it is stated for the discrete Weyl operators (for a qubit these are the Pauli matrices up to phases, Example 4.12).

The spectrum of a witness

In order to show the main result, we need to first show a few theorems. The first one is the following:

Theorem (spectrum of a witness). Let A1=(I+X+Y+Z)/2A_1=(\mathbb{I}+X+Y+Z)/2, with eigenvalues λ±\lambda_\pm. For all n≥1n\ge1 and A∈ΛnA\in\Lambda_n it holds that λ−≤λmin⁡(A)\lambda_-\le\lambda_{\min}(A) and λmax⁡(A)≤λ+\lambda_{\max}(A)\le\lambda_+. If λmax⁡(A)=λ+\lambda_{\max}(A)=\lambda_+ or λmin⁡(A)=λ−\lambda_{\min}(A)=\lambda_-, then AA has eigenvalues λ+,λ−\lambda_+,\lambda_- and d−2d-2 zeros.

Proof: The proof is an induction on nn. For n=1n=1, A=(I+r⋅σ)/2A=(\mathbb{I}+r\cdot\sigma)/2 with ∣ri∣≤1|r_i|\le1, so the eigenvalues (1±∣r∣)/2(1\pm|r|)/2 lie in [λ−,λ+][\lambda_-,\lambda_+], and they are λ±\lambda_\pm when a bound is attained. Assume the theorem for n−1n-1 and let A∈ΛnA\in\Lambda_n, n≥2n\ge2. The upper bound is proved in Section 2.1 and the lower bound in Section 2.2.

2.1 The largest eigenvalue

In this subsection n≥2n\ge2 and A∈ΛnA\in\Lambda_n are fixed, and the induction hypothesis is used in the form λmax⁡(B)≤v\lambda_{\max}(B)\le v for every B∈Λn−1B\in\Lambda_{n-1}. Let ℓ=λmax⁡(A)\ell=\lambda_{\max}(A) with unit eigenvector φ\varphi, let q=Tr[A2]q=\mathrm{Tr}[A^2] and let κ=max⁡(0,−λmin⁡(A))\kappa=\max(0,-\lambda_{\min}(A)). For every Pauli P≠IP\ne\mathbb{I} we define the real numbers

xP=⟨φ∣P∣φ⟩,αP=⟨φ∣PAP∣φ⟩,uP=v aP−(2ℓ−v) xP,σP=sign(uP),x_P=\langle\varphi|P|\varphi\rangle,\qquad \alpha_P=\langle\varphi|PAP|\varphi\rangle,\qquad u_P=v\,a_P-(2\ell-v)\,x_P,\qquad \sigma_P=\mathrm{sign}(u_P),

and we also define

R′=A+κ I−(ℓ+κ) ∣φ⟩⟨φ∣,eP=⟨φ∣P R′ P∣φ⟩,ω=v−ℓ+κ.R'=A+\kappa\,\mathbb{I}-(\ell+\kappa)\,|\varphi\rangle\langle\varphi|,\qquad e_P=\langle\varphi|P\,R'\,P|\varphi\rangle,\qquad \omega=v-\ell+\kappa .

The operator R′R' is positive semidefinite: A+κI≥0A+\kappa\mathbb{I}\ge0, and R′R' is this operator with its eigenvalue ℓ+κ\ell+\kappa at φ\varphi removed. Therefore eP=αP+κ−(ℓ+κ)xP2≥0e_P=\alpha_P+\kappa-(\ell+\kappa)x_P^2\ge0 for every PP. We also note that ∣aP∣≤1|a_P|\le1, because by the block lemma aP=Tr[B+]−Tr[B−]a_P=\mathrm{Tr}[B_+]-\mathrm{Tr}[B_-] is a difference of two non-negative numbers whose sum is Tr[A]=1\mathrm{Tr}[A]=1.

Lemma (per-Pauli inequality). For every Pauli P≠IP\ne\mathbb{I},

0 ≤ eP ≤ ω (1−xP2)−∣uP∣ (1−σPxP).0\ \le\ e_P\ \le\ \omega\,(1-x_P^2)-|u_P|\,(1-\sigma_Px_P).

Consequently, ω≥0\omega\ge0 and ∣uP∣≤ω (1+σPxP)|u_P|\le\omega\,(1+\sigma_Px_P) for every P≠IP\ne\mathbb{I}.

Proof: By the block lemma and the induction hypothesis, B±≤v Tr[B±] Π±B_\pm\le v\,\mathrm{Tr}[B_\pm]\,\Pi_\pm. We take the expectation value in φ\varphi and use Aφ=ℓφA\varphi=\ell\varphi, Tr[B±]=(1±aP)/2\mathrm{Tr}[B_\pm]=(1\pm a_P)/2 and ⟨φ∣Π±∣φ⟩=(1±xP)/2\langle\varphi|\Pi_\pm|\varphi\rangle=(1\pm x_P)/2. This gives

ℓ+αP±2ℓxP ≤ v (1±aP)(1±xP).\ell+\alpha_P\pm2\ell x_P\ \le\ v\,(1\pm a_P)(1\pm x_P).

Now insert αP=eP−κ+(ℓ+κ)xP2\alpha_P=e_P-\kappa+(\ell+\kappa)x_P^2. The left-hand side becomes ℓ(1±xP)2−κ(1−xP2)+eP\ell(1\pm x_P)^2-\kappa(1-x_P^2)+e_P, and collecting terms we find

eP ≤ (1±xP) (ω±ζP),ζP:=v aP−(ℓ+κ) xP.e_P\ \le\ (1\pm x_P)\,(\omega\pm\zeta_P),\qquad \zeta_P:=v\,a_P-(\ell+\kappa)\,x_P .

Since uP=ζP+ωxPu_P=\zeta_P+\omega x_P, the two bounds are ω(1−xP2)+(1+xP) uP\omega(1-x_P^2)+(1+x_P)\,u_P and ω(1−xP2)−(1−xP) uP\omega(1-x_P^2)-(1-x_P)\,u_P. The smaller of the two is ω(1−xP2)−∣uP∣(1−σPxP)\omega(1-x_P^2)-|u_P|(1-\sigma_Px_P), which is the claimed inequality.

For the consequences, note that there is a Pauli PP with ∣xP∣<1|x_P|<1 (because ∑PxP2=d−1<d2−1\sum_Px_P^2=d-1<d^2-1 by the rule for Pauli sums), and for this PP the inequality gives 0≤ω(1−xP2)0\le\omega(1-x_P^2), so ω≥0\omega\ge0. If σPxP≠1\sigma_Px_P\ne1, we divide ∣uP∣(1−σPxP)≤ω(1−xP2)|u_P|(1-\sigma_Px_P)\le\omega(1-x_P^2) by 1−σPxP1-\sigma_Px_P and obtain ∣uP∣≤ω(1+σPxP)|u_P|\le\omega(1+\sigma_Px_P). If σPxP=1\sigma_Px_P=1, the claim reads ∣uP∣≤2ω|u_P|\le2\omega, and it holds because ∣aP∣≤1|a_P|\le1 gives ∣uP∣=σPv aP−(2ℓ−v)≤2(v−ℓ)≤2ω|u_P|=\sigma_Pv\,a_P-(2\ell-v)\le2(v-\ell)\le2\omega. □\square

Theorem (star inequality). Let n≥2n\ge2 and A∈ΛnA\in\Lambda_n, with largest eigenvalue ℓ\ell, purity qq and κ=max⁡(0,−λmin⁡(A))\kappa=\max(0,-\lambda_{\min}(A)). Then

v2(dq−1)−2v(2ℓ−v)(dℓ−1)+(2ℓ−v)2(d−1) ≤ (v−ℓ+κ) G,v^2(dq-1)-2v(2\ell-v)(d\ell-1)+(2\ell-v)^2(d-1)\ \le\ (v-\ell+\kappa)\,G,

where

G=(v−ℓ) d(d−1)+d(ℓ−1)+2v(dℓ−1)−2(2ℓ−v)(d−1).G=(v-\ell)\,d(d-1)+d(\ell-1)+2v(d\ell-1)-2(2\ell-v)(d-1).

Equality holds for A=A⋆A=A^\star.

At this point of the argument the star inequality is proved under the induction hypothesis for n−1n-1. Once the theorem on the spectrum of a witness is established, it holds for every n≥2n\ge2 without any assumption, and this is how it is used in the second part.

Proof: The rule for Pauli sums gives the four identities (all sums run over P≠IP\ne\mathbb{I}, as before)

∑PxP2=d−1,∑PaPxP=dℓ−1,∑PaP2=dq−1,∑PαP=d−ℓ.\sum_P x_P^2=d-1,\qquad \sum_P a_Px_P=d\ell-1,\qquad \sum_P a_P^2=dq-1,\qquad \sum_P\alpha_P=d-\ell .

Therefore ∑PeP=d(1−ℓ)+κ d(d−1)\sum_Pe_P=d(1-\ell)+\kappa\,d(d-1), and summing the per-Pauli inequality over PP yields

∑P∣uP∣ (1−σPxP) ≤ ω d(d−1)−∑PeP=(v−ℓ) d(d−1)+d(ℓ−1).\sum_P|u_P|\,(1-\sigma_Px_P)\ \le\ \omega\,d(d-1)-\sum_Pe_P=(v-\ell)\,d(d-1)+d(\ell-1).

Next we multiply ∣uP∣≤ω(1+σPxP)|u_P|\le\omega(1+\sigma_Px_P) by ∣uP∣|u_P| and sum over PP. Since ∣uP∣ σP=uP|u_P|\,\sigma_P=u_P and ω≥0\omega\ge0, we get

∑PuP2 ≤ ω(∑P∣uP∣ (1−σPxP)+2∑PuPxP) ≤ ω G,\sum_Pu_P^2\ \le\ \omega\Big(\sum_P|u_P|\,(1-\sigma_Px_P)+2\sum_Pu_Px_P\Big)\ \le\ \omega\,G ,

where we used ∑PuPxP=v(dℓ−1)−(2ℓ−v)(d−1)\sum_Pu_Px_P=v(d\ell-1)-(2\ell-v)(d-1). Finally, the four identities give

∑PuP2=v2(dq−1)−2v(2ℓ−v)(dℓ−1)+(2ℓ−v)2(d−1),\sum_Pu_P^2=v^2(dq-1)-2v(2\ell-v)(d\ell-1)+(2\ell-v)^2(d-1),

which is the left-hand side of the star inequality. For A⋆A^\star one has ℓ=v\ell=v, κ=c\kappa=c and q=2q=2, and both sides are equal to d/2d/2. □\square

Proof of the upper bound λmax⁡(A)≤λ+\lambda_{\max}(A)\le\lambda_+: Suppose that ℓ=v+ε\ell=v+\varepsilon with ε≥0\varepsilon\ge0. Then ω=κ−ε≥0\omega=\kappa-\varepsilon\ge0. A direct expansion, using 2v2=2v+12v^2=2v+1, gives

G=dv−ε (d(d+2−2v)−4) ≤ dvG=dv-\varepsilon\,\big(d(d+2-2v)-4\big)\ \le\ dv

and

∑PuP2=d[v2(q−ℓ2)+(12−(2−v) ε)2]−4ε2.\sum_Pu_P^2=d\Big[v^2(q-\ell^2)+\big(\tfrac12-(2-v)\,\varepsilon\big)^2\Big]-4\varepsilon^2 .

Since ω≥0\omega\ge0 and G≤dvG\le dv, the star inequality implies ∑PuP2≤(κ−ε) dv\sum_Pu_P^2\le(\kappa-\varepsilon)\,dv. Dividing by dd and rearranging, this reads

v2 (q−ℓ2−κ2)+(vκ−12)2+(3−1) ε+((2−v)2−4d)ε2 ≤ 0.v^2\,(q-\ell^2-\kappa^2)+\big(v\kappa-\tfrac12\big)^2+(\sqrt3-1)\,\varepsilon+\Big((2-v)^2-\frac4d\Big)\varepsilon^2\ \le\ 0 .

The first term is non-negative because q≥ℓ2+κ2q\ge\ell^2+\kappa^2, and the second term is a square. If ε>0\varepsilon>0, the last two terms are positive. Indeed, the coefficient of ε2\varepsilon^2 is positive for d≥16d\ge16. For d=4d=4 and d=8d=8 it equals −0.60-0.60 and −0.10-0.10, but ε≤d−v\varepsilon\le\sqrt d-v (because ℓ2≤q≤d\ell^2\le q\le d, the last inequality following from ∣aP∣≤1|a_P|\le1), that is, ε<0.64\varepsilon<0.64 and ε<1.47\varepsilon<1.47 respectively, and then (3−1)+((2−v)2−4d)ε>0(\sqrt3-1)+\big((2-v)^2-\tfrac4d\big)\varepsilon>0. This contradiction shows that ε=0\varepsilon=0, i.e. ℓ≤λ+\ell\le\lambda_+. If ℓ=λ+\ell=\lambda_+, the displayed inequality forces vκ=12v\kappa=\tfrac12, that is κ=c=−λ−\kappa=c=-\lambda_-, and q=ℓ2+κ2q=\ell^2+\kappa^2, so that all the other eigenvalues of AA vanish. This proves the upper bound and its equality case.

2.2 The smallest eigenvalue

The argument is the mirror image of the previous one, and the induction hypothesis is now used in the form λmin⁡(B)≥−c\lambda_{\min}(B)\ge-c for every B∈Λn−1B\in\Lambda_{n-1}. If λmin⁡(A)≥0\lambda_{\min}(A)\ge0 there is nothing to prove. Otherwise let −κ=λmin⁡(A)<0-\kappa=\lambda_{\min}(A)<0 with unit eigenvector χ\chi, and let ℓ=λmax⁡(A)\ell=\lambda_{\max}(A). For every Pauli P≠IP\ne\mathbb{I} we define

yP=⟨χ∣P∣χ⟩,αP′=⟨χ∣PAP∣χ⟩,uP′=c aP−(2κ−c) yP,σP′=sign(uP′),y_P=\langle\chi|P|\chi\rangle,\qquad \alpha'_P=\langle\chi|PAP|\chi\rangle,\qquad u'_P=c\,a_P-(2\kappa-c)\,y_P,\qquad \sigma'_P=\mathrm{sign}(u'_P),

and

R′′=ℓ I−A−(ℓ+κ) ∣χ⟩⟨χ∣,fP=⟨χ∣P R′′ P∣χ⟩,ω′=c−κ+ℓ.R''=\ell\,\mathbb{I}-A-(\ell+\kappa)\,|\chi\rangle\langle\chi|,\qquad f_P=\langle\chi|P\,R''\,P|\chi\rangle,\qquad \omega'=c-\kappa+\ell .

The operator R′′R'' is positive semidefinite, since ℓ I−A≥0\ell\,\mathbb{I}-A\ge0 and R′′R'' is this operator with its eigenvalue ℓ+κ\ell+\kappa at χ\chi removed. Therefore fP=ℓ−αP′−(ℓ+κ)yP2≥0f_P=\ell-\alpha'_P-(\ell+\kappa)y_P^2\ge0.

Lemma (mirror inequalities). For every Pauli P≠IP\ne\mathbb{I},

0 ≤ fP ≤ ω′ (1−yP2)−∣uP′∣ (1−σP′yP),0\ \le\ f_P\ \le\ \omega'\,(1-y_P^2)-|u'_P|\,(1-\sigma'_Py_P),

and, as a consequence,

∑PuP′ 2 ≤ (c−κ+ℓ) G′,G′=(c−κ) d(d−1)+d(1+κ)−2c(dκ+1)−2(2κ−c)(d−1).\sum_Pu_P'^{\,2}\ \le\ (c-\kappa+\ell)\,G',\qquad G'=(c-\kappa)\,d(d-1)+d(1+\kappa)-2c(d\kappa+1)-2(2\kappa-c)(d-1).

Proof: By the block lemma and the induction hypothesis, B±≥−c Tr[B±] Π±B_\pm\ge-c\,\mathrm{Tr}[B_\pm]\,\Pi_\pm. Taking the expectation value in χ\chi and using Aχ=−κχA\chi=-\kappa\chi gives

−κ+αP′∓2κyP ≥ −c (1±aP)(1±yP).-\kappa+\alpha'_P\mp2\kappa y_P\ \ge\ -c\,(1\pm a_P)(1\pm y_P).

Inserting αP′=ℓ−(ℓ+κ)yP2−fP\alpha'_P=\ell-(\ell+\kappa)y_P^2-f_P and collecting terms, we find

fP ≤ (1±yP) (ω′±ζP′),ζP′:=c aP−(κ+ℓ) yP.f_P\ \le\ (1\pm y_P)\,(\omega'\pm\zeta'_P),\qquad \zeta'_P:=c\,a_P-(\kappa+\ell)\,y_P .

Since uP′=ζP′+ω′yPu'_P=\zeta'_P+\omega'y_P, the smaller of the two bounds is ω′(1−yP2)−∣uP′∣(1−σP′yP)\omega'(1-y_P^2)-|u'_P|(1-\sigma'_Py_P), exactly as in the per-Pauli inequality. In the same way we obtain ω′≥0\omega'\ge0 and ∣uP′∣≤ω′(1+σP′yP)|u'_P|\le\omega'(1+\sigma'_Py_P) for every PP (if σP′yP=1\sigma'_Py_P=1, then ∣uP′∣=σP′c aP−(2κ−c)≤2(c−κ)≤2ω′|u'_P|=\sigma'_Pc\,a_P-(2\kappa-c)\le2(c-\kappa)\le2\omega').

The rule for Pauli sums now gives ∑PyP2=d−1\sum_Py_P^2=d-1, ∑PaPyP=−dκ−1\sum_Pa_Py_P=-d\kappa-1, ∑PaP2=dq−1\sum_Pa_P^2=dq-1 and ∑PαP′=d+κ\sum_P\alpha'_P=d+\kappa. Hence ∑PfP=ℓ d(d−1)−d(1+κ)\sum_Pf_P=\ell\,d(d-1)-d(1+\kappa), and summing the first inequality over PP yields

∑P∣uP′∣ (1−σP′yP) ≤ (c−κ) d(d−1)+d(1+κ).\sum_P|u'_P|\,(1-\sigma'_Py_P)\ \le\ (c-\kappa)\,d(d-1)+d(1+\kappa).

Multiplying ∣uP′∣≤ω′(1+σP′yP)|u'_P|\le\omega'(1+\sigma'_Py_P) by ∣uP′∣|u'_P|, summing over PP, and using ∑PuP′yP=−c(dκ+1)−(2κ−c)(d−1)\sum_Pu'_Py_P=-c(d\kappa+1)-(2\kappa-c)(d-1), we arrive at ∑PuP′ 2≤ω′ G′\sum_Pu_P'^{\,2}\le\omega'\,G'. □\square

Proof of the lower bound λmin⁡(A)≥λ−\lambda_{\min}(A)\ge\lambda_-: Suppose that κ=c+ε\kappa=c+\varepsilon with ε≥0\varepsilon\ge0. A direct expansion, using c2+c=vc=12c^2+c=vc=\tfrac12, gives

G′=dc−ε (d(d+2+2c)−4) ≤ dcG'=dc-\varepsilon\,\big(d(d+2+2c)-4\big)\ \le\ dc

and

∑PuP′ 2=d[c2(q−κ2)+(12+(2+c) ε)2]−4ε2.\sum_Pu_P'^{\,2}=d\Big[c^2(q-\kappa^2)+\big(\tfrac12+(2+c)\,\varepsilon\big)^2\Big]-4\varepsilon^2 .

Since ω′=ℓ−ε≥0\omega'=\ell-\varepsilon\ge0 and G′≤dcG'\le dc, the mirror inequalities imply ∑PuP′ 2≤(ℓ−ε) dc\sum_Pu_P'^{\,2}\le(\ell-\varepsilon)\,dc. Dividing by dd and rearranging, this reads

c2 (q−ℓ2−κ2)+(c ℓ−12)2+(1+3) ε+((2+c)2−4d)ε2 ≤ 0.c^2\,(q-\ell^2-\kappa^2)+\big(c\,\ell-\tfrac12\big)^2+(1+\sqrt3)\,\varepsilon+\Big((2+c)^2-\frac4d\Big)\varepsilon^2\ \le\ 0 .

The first term is non-negative because q≥ℓ2+κ2q\ge\ell^2+\kappa^2, the second is a square, and the coefficients of ε\varepsilon and ε2\varepsilon^2 are positive. Hence ε=0\varepsilon=0, i.e. λmin⁡(A)≥λ−\lambda_{\min}(A)\ge\lambda_-. If λmin⁡(A)=λ−\lambda_{\min}(A)=\lambda_-, the displayed inequality forces c ℓ=12c\,\ell=\tfrac12, that is ℓ=v=λ+\ell=v=\lambda_+, and q=ℓ2+κ2q=\ell^2+\kappa^2, so that all the other eigenvalues of AA vanish. This completes the induction and the proof of the theorem on the spectrum of a witness. □\square

The theorem has a direct translation into a statement about states.

Corollary (dual form of the theorem on the spectrum). For every nn-qubit state ρ\rho,

λ+I−ρλ+d−1∈STABn\frac{\lambda_+\mathbb{I}-\rho}{\lambda_+d-1}\in\mathrm{STAB}_n

and

p ρ+(1−p) Id∈STABnwheneverp≤11+(3−1) 2 n−1.p\,\rho+(1-p)\,\frac{\mathbb{I}}{d}\in\mathrm{STAB}_n\qquad\text{whenever}\qquad p\le\frac{1}{1+(\sqrt3-1)\,2^{\,n-1}} .

Both constants are optimal for every nn.

Proof: A state σ\sigma belongs to STABn\mathrm{STAB}_n if and only if Tr[Aσ]≥0\mathrm{Tr}[A\sigma]\ge0 for all A∈ΛnA\in\Lambda_n [4]. By the theorem, λ−≤Tr[Aρ]≤λ+\lambda_-\le\mathrm{Tr}[A\rho]\le\lambda_+ for every state ρ\rho and every A∈ΛnA\in\Lambda_n. Hence Tr[A(λ+I−ρ)]=λ+−Tr[Aρ]≥0\mathrm{Tr}[A(\lambda_+\mathbb{I}-\rho)]=\lambda_+-\mathrm{Tr}[A\rho]\ge0, and Tr[A(pρ+(1−p)I/d)]≥−p c+(1−p)/d\mathrm{Tr}[A(p\rho+(1-p)\mathbb{I}/d)]\ge-p\,c+(1-p)/d, which is non-negative exactly when p≤1/(1+cd)p\le1/(1+cd), and c d=(3−1) 2n−1c\,d=(\sqrt3-1)\,2^{n-1}. For the optimality, take A=A⋆A=A^\star and let ρ\rho be its eigenvector with eigenvalue λ+\lambda_+ in the first statement and with eigenvalue λ−\lambda_- in the second. □\square

2.3 The star inequality near A⋆A^\star

For the application to purity in the second part, it is convenient to rewrite the star inequality in variables that measure the distance of a witness from A⋆A^\star. For A∈ΛnA\in\Lambda_n with a negative eigenvalue we write

ℓ=λmax⁡(A)=v−δ,κ=−λmin⁡(A)=c−k,q=Tr[A2]=2+θ.\ell=\lambda_{\max}(A)=v-\delta,\qquad \kappa=-\lambda_{\min}(A)=c-k,\qquad q=\mathrm{Tr}[A^2]=2+\theta .

By the theorem on the spectrum of a witness, δ≥0\delta\ge0 and k≥0k\ge0, and for A⋆A^\star we have δ=k=θ=0\delta=k=\theta=0.

Lemma (the star inequality near A⋆A^\star). Let n≥2n\ge2 and let A∈ΛnA\in\Lambda_n have a negative eigenvalue. Then

d v2 θ+d v k ≤ δ [(c−k) D−d (v+3)]+δ2 [D+4d(v−1)+4],d\,v^2\,\theta+d\,v\,k\ \le\ \delta\,\big[(c-k)\,D-d\,(v+3)\big]+\delta^2\,\big[D+4d(v-1)+4\big],

where D:=d(d+2−2v)−4D:=d(d+2-2v)-4.

Proof: We insert ℓ=v−δ\ell=v-\delta, κ=c−k\kappa=c-k and q=2+θq=2+\theta in the star inequality. Using 2v2=2v+12v^2=2v+1, vc=12vc=\tfrac12 and v2+c2=2v^2+c^2=2, one finds G=dv+δDG=dv+\delta D and

v2(dq−1)−2v(2ℓ−v)(dℓ−1)+(2ℓ−v)2(d−1)=d[12+v2θ+(2v+3) δ−4(v−1) δ2]−4δ2,v^2(dq-1)-2v(2\ell-v)(d\ell-1)+(2\ell-v)^2(d-1)=d\Big[\tfrac12+v^2\theta+(2v+3)\,\delta-4(v-1)\,\delta^2\Big]-4\delta^2 ,

while

(v−ℓ+κ) G=(c−k+δ)(dv+δD)=d2−d v k+d v δ+(c−k) D δ+D δ2.(v-\ell+\kappa)\,G=(c-k+\delta)(dv+\delta D)=\tfrac d2-d\,v\,k+d\,v\,\delta+(c-k)\,D\,\delta+D\,\delta^2 .

Subtracting the two expressions gives the claim. □\square

The second part combines this lemma with the Liu–Liu inequality of Ref. [3] to bound the purity Tr[A2]\mathrm{Tr}[A^2] of every witness, which is what the theorem stated at the beginning requires.

Remark. All the inequalities used in the proofs above were also tested numerically on elements of Λn\Lambda_n for n=2,3,4n=2,3,4. The argument is AI-generated and has not been peer reviewed.

Permanent link:open-context.jan-bulling.com/contribution/5.11

References

  1. Zhenhuan Liu, Tobias Haug, Qi Ye, Zi-Wen Liu, and Ingo Roth, Triangle Criterion: a mixed-state magic criterion with applications in distillation and detection, arXiv:2512.16777 (2025)
  2. Michael Zurel and Jack Davis, Basis-independent stabilizerness and maximally noisy magic states, arXiv:2602.22336 (2026)
  3. Zhenhuan Liu and Zi-Wen Liu, On the geometry, typicality, and detectability of quantum magic, arXiv:2609.03944 (2026)
  4. Michael Zurel, Cihan Okay, and Robert Raussendorf, Hidden Variable Model for Universal Quantum Computation with Magic States on Qubits, Phys. Rev. Lett. 125, 260404 (2020)
  5. Ben W. Reichardt, Quantum universality by state distillation, arXiv:quant-ph/0608085 (2006)
  6. Hilbert–Schmidt inradius of the multiqubit stabilizer polytope, Quantum Information and Quantum Computation Open Problem Zoo (QIQCOP Zoo), (2026)
  7. John Watrous, The Theory of Quantum Information, Cambridge University Press (2018)
  8. Cihan Okay, Michael Zurel, Robert Raussendorf, On the extremal points of the $\Lambda$-polytopes and classical simulation of quantum computation with magic states, Quantum Information and Computation Vol.21 No.13&14, 1533-7146 (2021) doi:10.26421/QIC21.13-14-2
  9. Michael Zurel, Cihan Okay, Robert Raussendorf, and Arne Heimendahl, Hidden variable model for quantum computation with magic states on qudits of any dimension, Quantum 8, 1323 (2024)
  10. Michael Zurel, Cihan Okay, and Robert Raussendorf, Simulating Quantum Computation: How Many “Bits” for “It”?, PRX Quantum 5, 030343 (2024)

Discussion

Sign in to join the discussion.

No comments yet. Questions about this result, checks of its steps and pointers are welcome here.