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

VerificationOpenContext 5.14

A solution to the problem which uses Aronszajn's inequality

by Rafael Wagner · accepted 20 Nov 2026 · partially AI-assisted

Main result

Below, I present an independent claim from the one by Yujie where it seems that Fable 5.2 found a solution to the problem. The interesting aspects is that it clearly has succeeded by building on the theory of Λ\Lambda-polytopes. Below, I mark all the points where I believe further exposition is needed with a triple star (⋆⋆⋆)(\star\star\star).

This is the main output:

Theorem (inradius of the stabilizer polytope). Let Λn\Lambda_n be the Λ\Lambda-polytope of nn qubits. Denote by d=2nd=2^n the Hilbert space dimension, and Mn:=max⁡A∈Λn Tr[A2].M_n:=\max_{A\in\Lambda_n}\,\mathrm{Tr}[A^2]. For every integer n≥1n\ge1,

Mn=2.M_n=2 .

Consequently, for every nn-qubit state ρ\rho,

Tr[ρ2]≤1d−12   ⟹   ρ∈STABn,\mathrm{Tr}[\rho^2]\le\frac{1}{d-\frac12}\ \implies\ \rho\in\mathrm{STAB}_n ,

and the constant 12\tfrac12 cannot be improved [1]. Moreover, A∈ΛnA\in\Lambda_n has Tr[A2]=2\mathrm{Tr}[A^2]=2 if and only if A=UA⋆U†A=UA^\star U^\dagger for a Clifford unitary UU, where A⋆=12(I+X+Y+Z)⊗∣0…0⟩⟨0…0∣A^\star=\tfrac12(\mathbb{I}+X+Y+Z)\otimes|0\ldots0\rangle\langle0\ldots0|.

This is the first contribution of two parts. First part proves Mn=2M_n=2 and shows that every maximiser has largest eigenvalue λ+=1+32\lambda_+=\tfrac{1+\sqrt3}{2}. The second part proves that such an operator lies in the Clifford orbit of A⋆A^\star and reports numerical experiments on sub-polytopes of STABn\mathrm{STAB}_n.

The proof is analytical; no step is computer-assisted. It relies on three earlier results, recalled below: the block lemma [4], the Liu–Liu inequality [3], and the theorem on the spectrum of a witness proved in a previous contribution. The full strategy is close to the one by Yujie Zhang.

Preliminary mathematical ideas

We keep the notation of the previous contributions. Λn\Lambda_n is the set of Hermitian operators AA with Tr[A]=1\mathrm{Tr}[A]=1 and ⟨s∣A∣s⟩≥0\langle s|A|s\rangle\ge0 for all stabilizer states, known as the Λ\Lambda-polytope [4, 2, 9, 10, 8, 11]. A state belongs to STABn\mathrm{STAB}_n if and only if Tr[Aρ]≥0\mathrm{Tr}[A\rho]\ge0 for all A∈ΛnA\in\Lambda_n [4]. By the Cauchy–Schwarz inequality every bound Mn≤mM_n\le m implies that Tr[ρ2]≤1/(d−1m)\mathrm{Tr}[\rho^2]\le1/(d-\tfrac1m) is sufficient for ρ∈STABn\rho\in\mathrm{STAB}_n (see OpenContext §5.12, the purity section). Define

v:=λ+=1+32,c:=−λ−=v−1,r:=32.v:=\lambda_+=\frac{1+\sqrt3}{2},\qquad c:=-\lambda_-=v-1,\qquad r:=\frac{\sqrt3}{2}.

It is simple to show that (⋆⋆⋆)(\star \star\star) vc=12,v2=v+12,c2=12−c,v2+c2=2,r=v−12=1−c2.vc=\tfrac12, \quad v^2=v+\tfrac12, \quad c^2=\tfrac12-c, \quad v^2+c^2=2, \quad r=v-\tfrac12=1-c^2. For A∈ΛnA\in\Lambda_n we write

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

These quantities are instrumental for the proof. For a Pauli operator P≠IP\ne\mathbb{I} let Π±=12(I±P)\Pi_\pm=\tfrac12(\mathbb{I}\pm P), B±=Π±AΠ±B_\pm=\Pi_\pm A\Pi_\pm and t±=Tr[B±]=12(1±aP)t_\pm=\mathrm{Tr}[B_\pm]=\tfrac12(1\pm a_P). All sums over PP run over the d2−1d^2-1 Pauli operators P≠IP\ne\mathbb{I}.

We use the following results discussed previously in OpenContext §5.12.

Lemma (block lemma [4]). The range of Π±\Pi_\pm is an (n−1)(n-1)-qubit code space whose stabilizer states are nn-qubit stabilizer states. Hence t±≥0t_\pm\ge0, and B±B_\pm is t±t_\pm times an element of Λn−1\Lambda_{n-1} (and B±=0B_\pm=0 if t±=0t_\pm=0). In particular

Tr[B±2]≤Mn−1 t±2.\mathrm{Tr}[B_\pm^2]\le M_{n-1}\,t_\pm^2 .

Theorem (spectrum of a witness, previous contribution). For all n≥1n\ge1 and A∈ΛnA\in\Lambda_n: −c≤λmin⁡(A)-c\le\lambda_{\min}(A) and λmax⁡(A)≤v\lambda_{\max}(A)\le v.

Lemma (Liu–Liu inequality [3]). For n≥2n\ge2 and A∈ΛnA\in\Lambda_n:  d q+1−2 Tr[A3]≤Mn−1 (d−q)\ d\,q+1-2\,\mathrm{Tr}[A^3]\le M_{n-1}\,(d-q).

We also use the rule for Pauli sums [7]: ∑PTr[PX]Tr[PY]=d Tr[XY]\sum_{P}\mathrm{Tr}[PX]\mathrm{Tr}[PY]=d\,\mathrm{Tr}[XY] and ∑PPXP=d Tr[X] I\sum_PPXP=d\,\mathrm{Tr}[X]\,\mathbb{I}, with the identity included in the sums. The last ingredient is a classical inequality for partitioned Hermitian matrices, attributed to Aronszajn [14]. We include its short proof (\star \star \star).

Lemma (Aronszajn's inequality). Let HH be a Hermitian operator on V1⊕V2V_1\oplus V_2 and let H1,H2H_1,H_2 be its compressions to V1V_1 and V2V_2. Then

λmax⁡(H)+λmin⁡(H)≤λmax⁡(H1)+λmax⁡(H2).\lambda_{\max}(H)+\lambda_{\min}(H)\le\lambda_{\max}(H_1)+\lambda_{\max}(H_2).

Proof: K:=H−λmin⁡(H)I≥0K:=H-\lambda_{\min}(H)\mathbb{I}\ge0. For a unit vector ψ=ψ1+ψ2\psi=\psi_1+\psi_2 with ψj∈Vj\psi_j\in V_j, the triangle inequality for K1/2K^{1/2} gives

⟨ψ∣K∣ψ⟩≤(∥ψ1∥k1+∥ψ2∥k2)2≤k1+k2,kj:=λmax⁡(Hj)−λmin⁡(H)≥0,\langle\psi|K|\psi\rangle\le\big(\|\psi_1\|\sqrt{k_1}+\|\psi_2\|\sqrt{k_2}\big)^2\le k_1+k_2,\qquad k_j:=\lambda_{\max}(H_j)-\lambda_{\min}(H)\ge0 ,

where the last step is the Cauchy–Schwarz inequality. Taking the supremum over ψ\psi gives λmax⁡(H)−λmin⁡(H)≤k1+k2\lambda_{\max}(H)-\lambda_{\min}(H)\le k_1+k_2. □\square

Block slacks

In this section and the next, n≥2n\ge2. Also, we assume Mn−1=2M_{n-1}=2, and A∈ΛnA\in\Lambda_n is fixed in order to structure an induction step later on. For P≠IP\ne\mathbb{I} we define the slacks as

s±P:=2 t±2−Tr[B±2].(slacks)s^P_\pm:=2\,t_\pm^2-\mathrm{Tr}[B_\pm^2] \tag{slacks}.

We also define the operator WW by

W:=2A−A2.W:=2A-A^2.

Additionally, we denote wP:=Tr[PW]w_P:=\mathrm{Tr}[PW] the expectation value of PP on WW. With these ingredients at hand, we can state and prove the following lemma (⋆⋆⋆)(\star \star \star).

Lemma (block slacks). For every P≠IP\ne\mathbb{I} we have s±P≥0s^P_\pm\ge0 and s+P−s−P=wPs^P_+-s^P_-=w_P. Moreover,

∑P(s+P+s−P)=32 d−d2−2d−22 θ,∑PwP2=d Tr[W2]−θ2.\sum_{P}\big(s^P_++s^P_-\big)=\frac32\,d-\frac{d^2-2d-2}{2}\,\theta,\qquad \sum_{P}w_P^2=d\,\mathrm{Tr}[W^2]-\theta^2 .

Proof: s±P≥0s^P_\pm\ge0 is the block lemma with Mn−1=2M_{n-1}=2. Let bP=Tr[PA2]b_P=\mathrm{Tr}[PA^2] and cP=Tr[PAPA]c_P=\mathrm{Tr}[PAPA]. Expanding Π±\Pi_\pm,

Tr[B±2]=Tr[Π±AΠ±A]=14 (q+cP±2bP).\mathrm{Tr}[B_\pm^2]=\mathrm{Tr}[\Pi_\pm A\Pi_\pm A]=\tfrac14\,(q+c_P\pm2b_P).

Since t+2−t−2=aPt_+^2-t_-^2=a_P and t+2+t−2=12(1+aP2)t_+^2+t_-^2=\tfrac12(1+a_P^2), we get

s+P−s−P=2aP−bP=wP,s+P+s−P=1+aP2−12 (q+cP).s^P_+-s^P_-=2a_P-b_P=w_P,\qquad s^P_++s^P_-=1+a_P^2-\tfrac12\,(q+c_P).

The rule for Pauli sums gives ∑PaP2=dq−1\sum_Pa_P^2=dq-1 and ∑PcP=d−q\sum_Pc_P=d-q. Hence ∑P(s+P+s−P)=d2−2+dq−12(d2q−2q+d)\sum_P(s^P_++s^P_-)=d^2-2+dq-\tfrac12(d^2q-2q+d), which is the claimed expression for q=2+θq=2+\theta. Finally, ∑PwP2=d Tr[W2]−Tr[W]2\sum_Pw_P^2=d\,\mathrm{Tr}[W^2]-\mathrm{Tr}[W]^2 and Tr[W]=2−q=−θ\mathrm{Tr}[W]=2-q=-\theta. □\square

Remark. The same computation gives 4d∑P(t−s+P+t+s−P)=2d−1−(d+2) q+2 Tr[A3]\tfrac4d\sum_P\big(t_-s^P_++t_+s^P_-\big)=2d-1-(d+2)\,q+2\,\mathrm{Tr}[A^3] (⋆⋆⋆)(\star \star \star). Since the left-hand side is non-negative, this is the Liu–Liu inequality for Mn−1=2M_{n-1}=2, so the proof below is self-contained once the block lemma and the theorem on the spectrum of a witness are granted.

The key inequality

The proposition above shows how one can bound the sum of the largest and smallest eigenvalues of AA in the Λ\Lambda-polytope (⋆⋆⋆)(\star \star \star).

Proposition. Let n≥2n\ge2, Mn−1=2M_{n-1}=2 and A∈ΛnA\in\Lambda_n. For every P≠IP\ne\mathbb{I},

λmax⁡(A)+λmin⁡(A) ≤ 2−∣wP∣ .\lambda_{\max}(A)+\lambda_{\min}(A)\ \le\ \sqrt{2-|w_P|}\,.

If moreover θ≥0\theta\ge0, there is a Pauli operator P≠IP\ne\mathbb{I} with

∣wP∣ ≥ 23(Tr[W2]−θ2d).|w_P|\ \ge\ \frac23\Big(\mathrm{Tr}[W^2]-\frac{\theta^2}{d}\Big).

Proof: Replacing PP by −P-P exchanges the two blocks and changes the sign of wPw_P, so we may assume wP≥0w_P\ge0. Then s+P=wP+s−P≥wPs^P_+=w_P+s^P_-\ge w_P, that is, Tr[B+2]≤2t+2−wP\mathrm{Tr}[B_+^2]\le2t_+^2-w_P. The largest eigenvalue of a Hermitian operator is at most its Hilbert–Schmidt norm, so the compressions of AA to the ranges of Π+\Pi_+ and Π−\Pi_- have largest eigenvalues at most 2t+2−wP\sqrt{2t_+^2-w_P} and Tr[B−2]≤2 t−=2 (1−t+)\sqrt{\mathrm{Tr}[B_-^2]}\le\sqrt2\,t_-=\sqrt2\,(1-t_+). By Aronszajn's inequality,

λmax⁡(A)+λmin⁡(A)≤g(t+),g(t):=2t2−wP+2 (1−t).\lambda_{\max}(A)+\lambda_{\min}(A)\le g(t_+),\qquad g(t):=\sqrt{2t^2-w_P}+\sqrt2\,(1-t).

Since g′(t)=2t/2t2−wP−2≥0g'(t)=2t/\sqrt{2t^2-w_P}-\sqrt2\ge0 and t+≤1t_+\le1, we have g(t+)≤g(1)=2−wPg(t_+)\le g(1)=\sqrt{2-w_P}.

For the second statement, ∣wP∣=∣s+P−s−P∣≤s+P+s−P|w_P|=|s^P_+-s^P_-|\le s^P_++s^P_-, so by the lemma ∑P∣wP∣≤32d\sum_P|w_P|\le\tfrac32d when θ≥0\theta\ge0. Therefore

d Tr[W2]−θ2=∑PwP2≤max⁡P∣wP∣ ∑P∣wP∣≤32 d max⁡P∣wP∣,d\,\mathrm{Tr}[W^2]-\theta^2=\sum_Pw_P^2\le\max_P|w_P|\,\sum_P|w_P|\le\tfrac32\,d\,\max_P|w_P| ,

which is the claim. □\square

For A⋆A^\star (see definition at the top of this document) all inequalities are equalities. To see this, note that W=r (∣φ⟩⟨φ∣−∣χ⟩⟨χ∣)W=r\,(|\varphi\rangle\langle\varphi|-|\chi\rangle\langle\chi|) for the two eigenvectors of A⋆A^\star, Tr[W2]=32\mathrm{Tr}[W^2]=\tfrac32, and ∣wP∣=1|w_P|=1 exactly for the 32d\tfrac32d Pauli operators that act as logical Pauli operators on the code space of A⋆A^\star; these carry the whole slack 32d\tfrac32d, and λmax⁡+λmin⁡=1\lambda_{\max}+\lambda_{\min}=1.

A lemma on spectra

Let f(λ):=(2λ−λ2)2f(\lambda):=(2\lambda-\lambda^2)^2, so that Tr[W2]=∑if(λi)\mathrm{Tr}[W^2]=\sum_if(\lambda_i) for the eigenvalues λi\lambda_i of AA (⋆⋆⋆)(\star \star \star).

Lemma (spectral inequality). Let λ1≥⋯≥λd\lambda_1\ge\dots\ge\lambda_d be real numbers, d≥4d\ge4, with 54≤λ1≤v\tfrac54\le\lambda_1\le v and −c≤λd≤0-c\le\lambda_d\le0. Put δ=v−λ1\delta=v-\lambda_1, k=c+λdk=c+\lambda_d, s=2−λ12−λd2s=2-\lambda_1^2-\lambda_d^2, and assume that σ:=∑1<i<dλi2=s+θ\sigma:=\sum_{1<i<d}\lambda_i^2=s+\theta with θ≥0\theta\ge0 and σ≤716\sigma\le\tfrac7{16}. Then

(λ1+λd)2+23∑if(λi)−2 ≥ Q(δ,k)+θ4,(\lambda_1+\lambda_d)^2+\frac23\sum_{i}f(\lambda_i)-2\ \ge\ Q(\delta,k)+\frac\theta4,

where

Q(δ,k)=(2+43)δ+(43−2)k−(9+143)δ2−10 δk+(173−7)k2,Q(\delta,k)=\Big(2+\tfrac{4}{\sqrt3}\Big)\delta+\Big(\tfrac{4}{\sqrt3}-2\Big)k-\Big(9+\tfrac{14}{\sqrt3}\Big)\delta^2-10\,\delta k+\Big(\tfrac{17}{\sqrt3}-7\Big)k^2 ,

and Q(δ,k)>0Q(\delta,k)>0 for 0≤δ≤v−540\le\delta\le v-\tfrac54, 0≤k≤c0\le k\le c, (δ,k)≠(0,0)(\delta,k)\ne(0,0).

Proof: Let a=δ(2c−δ)≥0a=\delta(2c-\delta)\ge0, b=k(2v−k)≥0b=k(2v-k)\ge0 and e=k−δe=k-\delta. Then λ1+λd=1+e\lambda_1+\lambda_d=1+e and, using v2+c2=2v^2+c^2=2 and v−c=1v-c=1, s=δ(2v−δ)+k(2c−k)=a+b−2es=\delta(2v-\delta)+k(2c-k)=a+b-2e.

The two extreme eigenvalues. 2λ1−λ12=1−(λ1−1)2=r+a2\lambda_1-\lambda_1^2=1-(\lambda_1-1)^2=r+a, because λ1−1=c−δ\lambda_1-1=c-\delta and 1−c2=r1-c^2=r. Similarly 2λd−λd2=1−(v−k)2=−(r−b)2\lambda_d-\lambda_d^2=1-(v-k)^2=-(r-b), because 1−λd=v−k1-\lambda_d=v-k and v2−1=rv^2-1=r. Hence f(λ1)=(r+a)2≥34+3 af(\lambda_1)=(r+a)^2\ge\tfrac34+\sqrt3\,a and f(λd)=(r−b)2=34−3 b+b2f(\lambda_d)=(r-b)^2=\tfrac34-\sqrt3\,b+b^2.

The other eigenvalues. For 1<i<d1<i<d we have λi≤max⁡(λ2,0)≤σ<2\lambda_i\le\max(\lambda_2,0)\le\sqrt\sigma<2, hence (2−λi)2≥(2−σ)2≥3(1−σ)(2-\lambda_i)^2\ge(2-\sqrt\sigma)^2\ge3(1-\sigma), the last step because (2−σ)2−3(1−σ)=(1−2σ)2(2-\sqrt\sigma)^2-3(1-\sigma)=(1-2\sqrt\sigma)^2. Therefore ∑1<i<df(λi)≥3σ(1−σ)\sum_{1<i<d}f(\lambda_i)\ge3\sigma(1-\sigma). The function u↦3u(1−u)u\mapsto3u(1-u) is concave with derivative at least 38\tfrac38 on [0,716][0,\tfrac7{16}], so 3σ(1−σ)≥3s(1−s)+38θ3\sigma(1-\sigma)\ge3s(1-s)+\tfrac38\theta.

Collecting. The left-hand side of the lemma is at least

(1+e)2−2+23[32+3 (a−b)+b2+3s−3s2]+θ4=−2e+e2+(2+23)a+(2−23)b+23b2−2s2+θ4,(1+e)^2-2+\tfrac23\Big[\tfrac32+\sqrt3\,(a-b)+b^2+3s-3s^2\Big]+\tfrac\theta4 =-2e+e^2+\Big(2+\tfrac2{\sqrt3}\Big)a+\Big(2-\tfrac2{\sqrt3}\Big)b+\tfrac23b^2-2s^2+\tfrac\theta4 ,

where we used s=a+b−2es=a+b-2e in the linear term. Now a=2cδ−δ2a=2c\delta-\delta^2, b=2vk−k2b=2vk-k^2 and (2+23) 2c=(2−23) 2v=43(2+\tfrac2{\sqrt3})\,2c=(2-\tfrac2{\sqrt3})\,2v=\tfrac4{\sqrt3}. Moreover b≥(v+1)kb\ge(v+1)k gives 23b2≥(2+3)k2\tfrac23b^2\ge(2+\sqrt3)k^2, and 0≤s≤2vδ+2ck0\le s\le2v\delta+2ck gives 2s2≤(8+43)δ2+8δk+(8−43)k22s^2\le(8+4\sqrt3)\delta^2+8\delta k+(8-4\sqrt3)k^2. Inserting these and e2=δ2−2δk+k2e^2=\delta^2-2\delta k+k^2 yields Q(δ,k)+θ4Q(\delta,k)+\tfrac\theta4.

Positivity of QQ. Rounding all coefficients against us, Q≥δ (4.309−17.083 δ−10k)+k (0.309+2.814 k)Q\ge\delta\,(4.309-17.083\,\delta-10k)+k\,(0.309+2.814\,k). Since δ≤v−54<0.1161\delta\le v-\tfrac54<0.1161, the first bracket exceeds 2.325−10k2.325-10k. If k≤0.2325k\le0.2325, both terms are non-negative, and they vanish only for k=δ=0k=\delta=0. If 0.2325<k≤c<0.36610.2325<k\le c<0.3661, the first term is at least −0.1161⋅1.336>−0.156-0.1161\cdot1.336>-0.156 and the second is at least 0.2325⋅0.963>0.2230.2325\cdot0.963>0.223. □\square

Proof of the theorem

We prove by induction on nn that Mn=2M_n=2, and that every A∈ΛnA\in\Lambda_n with Tr[A2]=2\mathrm{Tr}[A^2]=2 has λmax⁡(A)=λ+\lambda_{\max}(A)=\lambda_+. For n=1n=1, A=12(I+r⃗⋅σ⃗)A=\tfrac12(\mathbb{I}+\vec r\cdot\vec\sigma) with ∣ri∣≤1|r_i|\le1, so Tr[A2]=12(1+∣r⃗∣2)≤2\mathrm{Tr}[A^2]=\tfrac12(1+|\vec r|^2)\le2, with equality only if ∣r⃗∣=3|\vec r|=\sqrt3, i.e. λmax⁡(A)=λ+\lambda_{\max}(A)=\lambda_+. Since A⋆∈ΛnA^\star\in\Lambda_n has purity 22, we have Mn≥2M_n\ge2 for all nn.

Let n≥2n\ge2, assume Mn−1=2M_{n-1}=2, and let A∈ΛnA\in\Lambda_n have q=2+θq=2+\theta with θ≥0\theta\ge0. We show that θ=0\theta=0 and ℓ=v\ell=v.

Step 1 (location of the spectrum). With Mn−1=2M_{n-1}=2 the Liu–Liu inequality reads 2 Tr[A3]≥5+(d+2)θ2\,\mathrm{Tr}[A^3]\ge5+(d+2)\theta. Since every eigenvalue satisfies λ3≤ℓλ2\lambda^3\le\ell\lambda^2, also Tr[A3]≤ℓ (2+θ)\mathrm{Tr}[A^3]\le\ell\,(2+\theta). Hence

θ (d+2−2ℓ)≤4ℓ−5,\theta\,(d+2-2\ell)\le4\ell-5 ,

so ℓ≥54>1=Tr[A]\ell\ge\tfrac54>1=\mathrm{Tr}[A] and AA has a negative eigenvalue. By the theorem on the spectrum of a witness, 54≤ℓ≤v\tfrac54\le\ell\le v and 0<κ≤c0<\kappa\le c, so δ:=v−ℓ∈[0,v−54]\delta:=v-\ell\in[0,v-\tfrac54] and k:=c−κ∈[0,c)k:=c-\kappa\in[0,c). Moreover θ≤(4ℓ−5)/(6−2ℓ)<16\theta\le(4\ell-5)/(6-2\ell)<\tfrac16, and the sum σ\sigma of the squares of the other d−2d-2 eigenvalues satisfies

σ=2+θ−ℓ2−κ2≤2−ℓ2+4ℓ−56−2ℓ≤716,\sigma=2+\theta-\ell^2-\kappa^2\le2-\ell^2+\frac{4\ell-5}{6-2\ell}\le\frac7{16},

because the function of ℓ\ell on the right is decreasing on [54,v][\tfrac54,v] (its derivative is −2ℓ+14/(6−2ℓ)2<0-2\ell+14/(6-2\ell)^2<0) and equals 716\tfrac7{16} at ℓ=54\ell=\tfrac54.

Step 2 (the key inequality). By Step 1, ℓ−κ≥54−c>0\ell-\kappa\ge\tfrac54-c>0. We square the first inequality of the proposition for the Pauli operator of its second statement:

(ℓ−κ)2≤2−23(Tr[W2]−θ2d).(\ell-\kappa)^2\le2-\frac23\Big(\mathrm{Tr}[W^2]-\frac{\theta^2}{d}\Big).

Step 3 (conclusion). The eigenvalues of AA satisfy the assumptions of the lemma on spectra, with λ1=ℓ\lambda_1=\ell and λd=−κ\lambda_d=-\kappa. Together with Step 2,

Q(δ,k)+θ4 ≤ 23 θ2d ≤ θ36,Q(\delta,k)+\frac\theta4\ \le\ \frac23\,\frac{\theta^2}{d}\ \le\ \frac{\theta}{36},

where we used θ<16\theta<\tfrac16 and d≥4d\ge4. Since Q(δ,k)≥0Q(\delta,k)\ge0, this forces θ=0\theta=0 and Q(δ,k)=0Q(\delta,k)=0, hence δ=k=0\delta=k=0. So no element of Λn\Lambda_n has purity larger than 22, i.e. Mn=2M_n=2, and every A∈ΛnA\in\Lambda_n with Tr[A2]=2\mathrm{Tr}[A^2]=2 has λmax⁡(A)=λ+\lambda_{\max}(A)=\lambda_+ (and λmin⁡(A)=λ−\lambda_{\min}(A)=\lambda_-). This completes the induction.

The statement about states follows from Mn=2M_n=2 as recalled in the preliminaries. The second part shows that λmax⁡(A)=λ+\lambda_{\max}(A)=\lambda_+ implies A=UA⋆U†A=UA^\star U^\dagger. □\square

Permanent link:open-context.jan-bulling.com/contribution/5.14

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.

  • Rafael Wagner · 20 Nov 2026

    I would be nice to have Lean certificates of these proofs.

  • Rafael Wagner · 20 Nov 2026

    Below I'll work out the various parts in which I haven't fully checked in order to improve my understanding of it.

    The first bit is simple, and it is related to the first batch of claims. In particular, I wanted to understand better the following simple bit first:

    Define

    v:=λ+=1+32,c:=−λ−=v−1,r:=32.v:=\lambda_+=\frac{1+\sqrt3}{2},\qquad c:=-\lambda_-=v-1,\qquad r:=\frac{\sqrt3}{2}.

    It is simple to show that

    vc=12,v2=v+12,c2=12−c,v2+c2=2,r=v−12=1−c2.vc=\tfrac12,\qquad v^2=v+\tfrac12,\qquad c^2=\tfrac12-c,\qquad v^2+c^2=2,\qquad r=v-\tfrac12=1-c^2.

    Indeed, λ±\lambda_\pm are the two roots of λ2−λ−12=0\lambda^2-\lambda-\tfrac12=0. They are the eigenvalues of 12(I+X+Y+Z),\tfrac12(\mathbb{I}+X+Y+Z), and since we have that (X+Y+Z)2=3 I(X+Y+Z)^2=3\,\mathbb{I} we get these values for the eigenvalues. Here, these are simply the Pauli matrices.

    The sum and the product of the roots give v−c=1v-c=1 and vc=12vc=\tfrac12. The first relation is trivial, the product follows from

    vc=−λ+λ−=1+32(1+32−1)=1+32(3−12)=(3)1−124=12.vc = -\lambda_+\lambda_- = \frac{1+\sqrt{3}}{2} \left(\frac{1+\sqrt{3}}{2}-1\right) = \frac{1+\sqrt{3}}{2} \left(\frac{\sqrt{3}-1}{2}\right) = \frac{(\sqrt{3})^1-1^2}{4} = \frac{1}{2}.

    Now, for v2=v+12v^2=v+\tfrac12 and c2=12−cc^2=\tfrac12-c. Inserting each root in the equation gives

    v2=(1+32)2=4+234=1+32=12+(1+32⏟=v).v^2 = \left ( \frac{1+\sqrt{3}}{2} \right)^2 = \frac{4+2\sqrt{3}}{4} = 1 + \frac{\sqrt{3}}{2} = \frac{1}{2} + \left(\underbrace{\frac{1+\sqrt{3}}{2}}_{=v}\right).

    Then v2+c2=(v−c)2+2vc=12+2vc⏟=12=2v^2+c^2=(v-c)^2+2vc=1^2+2\underbrace{vc}_{=\frac{1}{2}} = 2 and 1−c2=12+c=v−12=r1-c^2=\tfrac12+c=v-\tfrac12=r. So, these are simply aspects.

  • Jan · 20 Nov 2026

    Wow, nice that this was solved! Amazing!