Skip to content
OpenContext
Contribution toProblem 5: Test problem: Hilbert–Schmidt inradius of the multiqubit stabilizer polytopeunder revision
+1

Claimed solutionOpenContext 5.13

A proposed analytic proof of the sharp bound cₙ = 1/2 for every number of qubits

by Yujie Zhang · accepted 20 Nov 2026 · partially AI-assisted

Main result

Proposed theorem. Write d=2nd=2^n. Let cnc_n be the largest constant such that every nn-qubit density operator of purity at most 1/(d−cn)1/(d-c_n) is a convex mixture of stabilizer states. Then

cn=12for every n≥1.\boxed{c_n=\tfrac12\qquad\text{for every }n\ge1.}

The centered Hilbert–Schmidt inradius is 1/d(2d−1)1/\sqrt{d(2d-1)}. The proof also determines every equality witness. If a trace-one Hermitian AA is nonnegative on all pure stabilizer states and Tr⁡(A2)=2\operatorname{Tr}(A^2)=2, then AA is Clifford-equivalent to

Vn=∣0n−1⟩⟨0n−1∣⊗I+X+Y+Z2.V_n=|0^{n-1}\rangle\langle0^{n-1}|\otimes\frac{I+X+Y+Z}{2}.

Which steps use a computer? The argument below is analytic throughout. This is a fully AI-generated proposed proof for independent review. Its external spectral input is Rafael Wagner’s upper eigenvalue bound in OpenContext 5.11, §2.1. All other estimates below follow from the displayed Pauli identities and inequalities. The earlier computer-assisted estimates in 5.10 and the grid checks in 5.12 are unnecessary for this proof. OpenContext 5.10 · OpenContext 5.11 · OpenContext 5.12

Proof strategy. Suppose an nn-qubit witness violates the bound. We show that some Pauli measurement outcome leaves a violating witness on n−1n-1 qubits. To prove this, we classify the witnesses exactly on the boundary for which every outcome obeys the lower-dimensional bound. Wagner’s theorem forces large Pauli expectations to carry more than 73.8%73.8\% of a particular probability measure. Their commutation relations leave at most two logical qubits. The two remaining cases can then be settled explicitly.

1. The boundary statement and induction

Write Λn\Lambda_n for the normalized witnesses above. For a nonidentity Hermitian Pauli PP, let Π=(I±P)/2\Pi=(I\pm P)/2, B=ΠAΠB=\Pi A\Pi and p=Tr⁡Bp=\operatorname{Tr}B. The outcome has a stabilizer basis, so p≥0p\ge0. If p>0p>0, identifying the code with n−1n-1 qubits gives B/p∈Λn−1B/p\in\Lambda_{n-1}. If p=0p=0, extend each logical stabilizer vector to a stabilizer basis. Every expectation in that basis vanishes, since they are nonnegative and sum to zero. Logical stabilizer projectors span the Hermitian operators, hence B=0B=0.

Boundary lemma. If A∈ΛnA\in\Lambda_n, Tr⁡(A2)=2\operatorname{Tr}(A^2)=2, and every signed Pauli outcome satisfies Tr⁡(B2)≤2p2\operatorname{Tr}(B^2)\le2p^2, then AA is Clifford-equivalent to VnV_n.

Here is why the lemma suffices. Interpolate a hypothetical witness with Tr⁡(A2)>2\operatorname{Tr}(A^2)>2 toward I/dI/d until its squared norm equals 22. Every stabilizer expectation of the interpolated witness is strictly positive. Each Clifford image of VnV_n has a zero stabilizer expectation. The lemma therefore forces some outcome to violate ∥B∥HS≤2 p\|B\|_{\mathrm{HS}}\le\sqrt2\,p. The same outcome violates the inequality on the original witness, since the inequality defines a convex cone containing the compression of I/dI/d. This gives the descent. On one qubit each Bloch coordinate of a witness lies in [−1,1][-1,1], giving squared norm at most 22, with equality precisely at the cube vertices. Induction excludes every violation. The lemma then gives the asserted equality classification at every nn.

2. The weight carried by large Pauli expectations

Assume the hypotheses of the boundary lemma. Put t=Tr⁡(A3)t=\operatorname{Tr}(A^3) and ℓ=λmax⁡(A)\ell=\lambda_{\max}(A). Wagner proves ℓ≤u=(1+3)/2\ell\le u=(1+\sqrt3)/2. For ℓ≤1\ell\le1, summing λ3≤ℓλ2\lambda^3\le\ell\lambda^2 gives t≤2t\le2. For ℓ>1\ell>1, the remaining eigenvalues have squared sum 2−ℓ22-\ell^2, hence

t≤ℓ3+(2−ℓ2)3/2≤u3+(u−1)3=332.t\le\ell^3+(2-\ell^2)^{3/2} \le u^3+(u-1)^3=\frac{3\sqrt3}{2}.

The middle expression increases for 1≤ℓ≤21\le\ell\le\sqrt2, and u2+(u−1)2=2u^2+(u-1)^2=2. OpenContext 5.11, §2.1

Label phase-free Paulis by v∈F22nv\in\mathbb F_2^{2n}, with [v,w]=1[v,w]=1 when their Hermitian representatives anticommute. Choose P0=IP_0=I and put

av=Tr⁡(APv),fv=av2,kv=Tr⁡(APvAPv),zv=Tr⁡((2A−A2)Pv).a_v=\operatorname{Tr}(AP_v),\quad f_v=a_v^2,\quad k_v=\operatorname{Tr}(AP_vAP_v),\quad z_v=\operatorname{Tr}((2A-A^2)P_v).

The two nonnegative Pauli outcome probabilities give ∣av∣≤1|a_v|\le1. Direct expansion with Π±=(I±Pv)/2\Pi_\pm=(I\pm P_v)/2, B±=Π±AΠ±B_\pm=\Pi_\pm A\Pi_\pm, gives

4(2p±2−Tr⁡(B±2))=2fv−kv±2zv.4(2p_\pm^2-\operatorname{Tr}(B_\pm^2))=2f_v-k_v\pm2z_v.

Thus 2fv−kv≥2∣zv∣2f_v-k_v\ge2|z_v|, and at v=0v=0 both sides vanish. Define μv=(2fv−kv)/(3d)\mu_v=(2f_v-k_v)/(3d). Pauli orthogonality gives ∑vfv=2d\sum_v f_v=2d and kv=d−1∑w(−1)[v,w]fwk_v=d^{-1}\sum_w(-1)^{[v,w]}f_w. Hence ∑vkv=d\sum_v k_v=d, so μ\mu is a probability distribution. Taking the Fourier transform again gives

Pr⁡w∼μ([v,w]=1)=dμv+1−fv2.(1)\Pr_{w\sim\mu}([v,w]=1)=d\mu_v+\frac{1-f_v}{2}.\tag{1}

Let H={v≠0:∣av∣>3/4}\mathcal H=\{v\ne0:|a_v|>3/4\} and β=μ(H)\beta=\mu(\mathcal H). By Pauli orthogonality and the bound on ∣zv∣|z_v|,

4−t=1d∑vavzv≤32∑vμv∣av∣≤32(34+β4).4-t=\frac1d\sum_v a_vz_v \le\frac32\sum_v\mu_v|a_v| \le\frac32\left(\frac34+\frac{\beta}{4}\right).

Consequently β≥(23−8t)/3≥(23−123)/3>0.738\beta\ge(23-8t)/3\ge(23-12\sqrt3)/3>0.738.

3. The anticommutation graph leaves at most two logical qubits

Join two labels in H\mathcal H when they anticommute. Three induced graphs on four vertices are impossible. Up to Clifford conjugation their four observables are

two disjoint edgesXI,ZI,IX,IZfour-vertex pathXI,ZI,XX,IZtriangle with a pendant vertexXI,ZI,YX,IZ\begin{array}{c|c} \text{two disjoint edges}&XI,ZI,IX,IZ\\ \text{four-vertex path}&XI,ZI,XX,IZ\\ \text{triangle with a pendant vertex}&XI,ZI,YX,IZ \end{array}

Here XI=X⊗IXI=X\otimes I, and identities on additional qubits are implicit. Each graph has an odd number of perfect matchings, so its alternating adjacency matrix is nonsingular over F2\mathbb F_2. Its four Pauli labels therefore form a symplectically nondegenerate basis of their span, which proves the asserted equivalence to the displayed realizations. Independence also lets Pauli conjugation prescribe all four signs.

For each row the sum of the four absolute expectations is at most 33. To check this, choose signs giving positive expectations. For commuting signed Paulis U,VU,V the operator I+U+V+UVI+U+V+UV is four times a stabilizer code projector. For the path take the six triples (U,V,UV)(U,V,UV) below, two per line.

(IX,−XI,−XX)(−IX,−ZI,ZX)(−IZ,−XI,XZ)(−IZ,−ZI,ZZ)(−XX,−YY,−ZZ)(−XZ,YY,−ZX)\begin{array}{cc} (IX,-XI,-XX)&(-IX,-ZI,ZX)\\ (-IZ,-XI,XZ)&(-IZ,-ZI,ZZ)\\ (-XX,-YY,-ZZ)&(-XZ,YY,-ZX) \end{array}

For the two disjoint edges, replace the first triple by (−IX,−XI,XX)(-IX,-XI,XX). For the triangle with a pendant vertex use

(IY,−XI,−XY)(−IY,−ZI,ZY)(−IZ,−XI,XZ)(−IZ,−ZI,ZZ)(XY,−YX,−ZZ)(−XZ,−YX,−ZY).\begin{array}{cc} (IY,-XI,-XY)&(-IY,-ZI,ZY)\\ (-IZ,-XI,XZ)&(-IZ,-ZI,ZZ)\\ (XY,-YX,-ZZ)&(-XZ,-YX,-ZY). \end{array}

In each case the six operators sum to 6I6I minus twice the sum of the four observables. Stabilizer positivity proves the claimed upper bound, whereas four elements of H\mathcal H have total absolute expectation greater than 33.

All edges now lie in one connected component, since two edges in different components would be disjoint and induced. The absence of an induced four-vertex path bounds the component’s diameter by two. If a,ba,b are adjacent and cc is adjacent to neither, take a common neighbor ee of a,ca,c in this component. When ee meets bb there is a triangle with pendant cc. Otherwise c,e,a,bc,e,a,b is an induced path. Thus nonadjacency, including equality, partitions the component into independent parts, with all edges between distinct parts.

Put W=span⁡(H)W=\operatorname{span}(\mathcal H) and L=W∩W⊥L=W\cap W^\perp. Every maximal commuting label space FF has μ(F)≤1/3\mu(F)\le1/3. Indeed, F=F⊥F=F^\perp and ∣F∣=d|F|=d, so ∑v∈Fkv=∑v∈Ffv\sum_{v\in F}k_v=\sum_{v\in F}f_v by the Fourier formula. The latter sum divided by dd is the collision probability of the stabilizer basis distribution for FF, hence at most one. Since β>1/3\beta>1/3, the graph has an edge.

Choose one representative viv_i from each of the m≥2m\ge2 parts. They pairwise anticommute. Labels in a part have identical neighbors in H\mathcal H, so their differences lie in LL, as do all isolated labels. Thus H⊂Ω=L∪⋃i(vi+L)\mathcal H\subset\Omega=L\cup\bigcup_i(v_i+L). The representatives span W/LW/L. Their Gram matrix has ones off the diagonal and zeros on it, so its rank is 2r2r, where m=2rm=2r or 2r+12r+1. In the odd case their sum lies in LL. Since a nondegenerate complement of LL in WW has dimension 2r2r, ∣L∣≤2n−r|L|\le2^{n-r}.

A label orthogonal to LL anticommutes with at most 2r2r entire cosets of Ω\Omega, by the sum relation when mm is odd. A label outside L⊥L^\perp anticommutes with half of each coset. Summing (1) over Ω\Omega gives dβ≤2r∣L∣d\beta\le2r|L|, hence β≤2r/2r\beta\le2r/2^r. This rules out r≥4r\ge4. Also ∣L∣≤2n−r−1|L|\le2^{n-r-1} would give β≤r/2r≤1/2\beta\le r/2^r\le1/2. Therefore dim⁡L=n−r\dim L=n-r and W=L⊥W=L^\perp. A joint eigenspace of LL, called a syndrome space below, encodes rr logical qubits.

For r=3r=3, complete the six or seven representatives to seven by adjoining their sum if needed. Their only relation modulo LL is that all seven sum to zero. Each logical label is the sum of at most three of them. Let Ω′\Omega' be LL and the seven representative cosets, and set M=μ(Ω′)M=\mu(\Omega'). There are 8∣L∣=d8|L|=d labels in Ω′\Omega'. A sum of one, two or three generators anticommutes with respectively six, two or four generators. Thus a physical label in Ω′\Omega' anticommutes with at most 3d/43d/4 labels in Ω′\Omega'. A label outside anticommutes with at most d/2d/2, including exactly half of each coset when it lies outside L⊥L^\perp. Sum dμv≤Pr⁡([v,w]=1)d\mu_v\le\Pr([v,w]=1) from (1) over Ω′\Omega'. Then M≤3M/4+(1−M)/2M\le3M/4+(1-M)/2, so β≤M≤2/3\beta\le M\le2/3, a contradiction. Only r=1,2r=1,2 remain.

4. Bounding the blocks outside one syndrome

Fix signed generators of LL. Write EsE_s for their joint syndrome projectors, Bs=EsAEsB_s=E_sAE_s and ps=Tr⁡Bsp_s=\operatorname{Tr}B_s. The psp_s are nonnegative and sum to one. Fourier orthogonality on LL and averaging (1) over v∈Lv\in L give

S0:=∑sps2=1∣L∣∑v∈Lfv=μ(L⊥)+2d∣L∣μ(L)≥β.S_0:=\sum_s p_s^2=\frac1{|L|}\sum_{v\in L} f_v =\mu(L^\perp)+\frac{2d}{|L|}\mu(L)\ge\beta.

Let EE have the largest weight p=1−ηp=1-\eta. Since p≥S0≥βp\ge S_0\ge\beta and S0≤p2+η2S_0\le p^2+\eta^2, we get η<0.156\eta<0.156 (at η=0.156\eta=0.156 the last upper bound is 0.7366720.736672). Label EE by syndrome zero. In blocks on E⊕E⊥E\oplus E^\perp write A=(BHH†C)A=\left(\begin{smallmatrix}B&H\\H^\dagger&C\end{smallmatrix}\right), h=Tr⁡(HH†)h=\operatorname{Tr}(HH^\dagger) and qB=Tr⁡(B2)q_B=\operatorname{Tr}(B^2).

For each nonzero syndrome parity aa, apply the outcome hypothesis to a⋅i=1a\cdot i=1. Among all parities a fixed nonzero ii is selected half the time, and two distinct nonzero indices i,ji,j are both selected one quarter of the time. Summing the squared block norms on the left and the squared traces on the right gives

∥C∥HS2+∑i≠0∥Cii∥HS2≤2η2+2∑i≠0pi2≤4η2.\|C\|_{\mathrm{HS}}^2+\sum_{i\ne0}\|C_{ii}\|_{\mathrm{HS}}^2 \le 2\eta^2+2\sum_{i\ne0}p_i^2\le4\eta^2.

Let Π\Pi project onto either outcome of a nonidentity logical Pauli on EE, and put b=Tr⁡(ΠB)b=\operatorname{Tr}(\Pi B). Multiply that signed Pauli by a uniformly random syndrome parity. The resulting physical outcome selects Π\Pi on EE and selects a uniformly random one of the two complementary logical eigenspaces on each nonzero syndrome. Their total selected trace TT satisfies 0≤T≤η0\le T\le\eta and ET=η/2\mathbb ET=\eta/2. Keep the 0000 and 0i,i00i,i0 blocks in the squared norm of the physical compression and average. Each nonzero-syndrome projector has mean I/2I/2, giving

Tr⁡[(ΠBΠ)2]+Tr⁡(ΠHH†)≤2E(b+T)2≤2b2+2bη+η2.\operatorname{Tr}[(\Pi B\Pi)^2]+\operatorname{Tr}(\Pi HH^\dagger) \le2\mathbb E(b+T)^2\le2b^2+2b\eta+\eta^2 .

The first term is at least b2/rank⁡Πb^2/\operatorname{rank}\Pi. Hence

Tr⁡(ΠHH†)≤{(b+η)2,r=1,32b2+2bη+η2,r=2.(2)\operatorname{Tr}(\Pi HH^\dagger)\le \begin{cases} (b+\eta)^2,&r=1,\\ \tfrac32b^2+2b\eta+\eta^2,&r=2 . \end{cases}\tag{2}

For a logical axis with a physical lift in H\mathcal H, sign that lift to have expectation >3/4>3/4. Its opposite outcome has total trace <1/8<1/8. On EE its trace bb is at most 1/81/8, since each other syndrome contributes a nonnegative trace. The estimates also hold for a single syndrome, when the complementary blocks are empty.

5. The one-logical-qubit case

All three logical Pauli axes must occur in H\mathcal H. Otherwise H\mathcal H would lie in the union of two maximal commuting spaces, of total mass at most 2/32/3. For the opposite projectors on these signed axes write bj≤1/8b_j\le1/8, S=∑jbjS=\sum_j b_j and S2=∑jbj2≤S/8S_2=\sum_j b_j^2\le S/8. The one-qubit Pauli expansion gives qB=2p2−2pS+2S2q_B=2p^2-2pS+2S_2. The three projectors sum to an operator of least eigenvalue (3−3)/2(3-\sqrt3)/2. Summing (2) thus gives 2h≤K(S2+2ηS+3η2)2h\le K(S_2+2\eta S+3\eta^2), with K=4/(3−3)<3.16K=4/(3-\sqrt3)<3.16. Substitution into 2=qB+2h+∥C∥HS22=q_B+2h+\|C\|_{\mathrm{HS}}^2 yields

0≤η[−4+(6+3K)η]+S[−2+(2+2K)η+(2+K)/8].0\le\eta[-4+(6+3K)\eta]+S[-2+(2+2K)\eta+(2+K)/8].

Both brackets are negative at η=0.156\eta=0.156 and increase with η\eta. Hence η=S=0\eta=S=0. The complement and coherence bounds give C=H=0C=H=0, and all three signed logical coordinates are +1+1. These are precisely the Clifford images of VnV_n.

6. The two-logical-qubit case

Complete the four or five anticommuting representatives to five axes R1,…,R5R_1,\ldots,R_5, adjoining their sum when needed. The other ten nonzero logical labels are the pair sums Ri+RjR_i+R_j. For each logical label jj choose a lift TjT_j and let Sj=∑s(Tr⁡(BsTj))2S_j=\sum_s(\operatorname{Tr}(B_sT_j))^2. Fourier orthogonality on the syndromes gives Sj=∣L∣−1∑v∈j+LfvS_j=|L|^{-1}\sum_{v\in j+L}f_v. For a set JJ of logical labels write J+LJ+L for its full preimage in L⊥L^\perp. Summing the Fourier formula for kvk_v over its cosets gives

12μ(J+L)=2∑j∈JSj−14∑v∈L⊥/LSv∑j∈J(−1)[j,v].12\mu(J+L)=2\sum_{j\in J}S_j- \frac14\sum_{v\in L^\perp/L} S_v\sum_{j\in J}(-1)^{[j,v]} .

Take JJ to consist of zero and the five axes. Let UU sum the SjS_j on the axes and VV sum them on the pairs. The character sums over JJ are 6,−2,26,-2,2 on zero, axes and pairs respectively. For JcJ^c they are 10,2,−210,2,-2. Thus

24μ(J+L)=S0+5U−V,24μ(Jc+L)=−5S0−U+5V.24\mu(J+L)=S_0+5U-V,\qquad 24\mu(J^c+L)=-5S_0-U+5V .

The second mass is nonnegative and the first contains H\mathcal H, so β≤μ(J+L)≤U/5\beta\le\mu(J+L)\le U/5.

Order αi=∣Tr⁡(BRi)∣\alpha_i=|\operatorname{Tr}(BR_i)| decreasingly on EE, and set si=p2−αi2s_i=p^2-\alpha_i^2, Δ=∑isi\Delta=\sum_i s_i. At least four axes have a lift in H\mathcal H, so the opposite-outcome estimate in §4 gives αi≥p−1/4\alpha_i\ge p-1/4 on the top four. Every logical coefficient on any other syndrome has magnitude at most psp_s. Consequently

0≤Δ≤5S0−U≤5(p2+η2−β).(3)0\le\Delta\le5S_0-U\le5(p^2+\eta^2-\beta).\tag{3}

For each pair Q=Ri+RjQ=R_i+R_j choose the index with smaller ss inside the pair and the two with smallest ss outside it, breaking ties to avoid index five. Let mQm_Q be the sum of the three chosen deficits. The chosen axes form a triangle and QQ is adjacent only to the chosen index inside the pair. The four-observable inequality in §3 applied to B/pB/p gives

∣Tr⁡(BQ)∣≤∑chosen k(p−αk)≤mQ2p−1/4.|\operatorname{Tr}(BQ)|\le\sum_{\text{chosen }k}(p-\alpha_k) \le \frac{m_Q}{2p-1/4}.

The chosen deficit inside the pair is bounded by half its sum, and the two outside by two thirds of their sum. Thus mQ≤2Δ/3m_Q\le2\Delta/3. Summing over pairs gives ∑QmQ≤6Δ\sum_Qm_Q\le6\Delta since every index is inside four pairs and outside six. Hence ∑QmQ2≤4Δ2\sum_Qm_Q^2\le4\Delta^2. Expanding BB in the fifteen nonidentity Paulis gives

qB≤32p2+Δ(Δ(2p−1/4)2−14).(4)q_B\le\tfrac32p^2+\Delta\left(\frac{\Delta}{(2p-1/4)^2}-\tfrac14\right).\tag{4}

For a lower bound take four signed axes with lifts in H\mathcal H. Their signed sum squares to 4IE4I_E, so their opposite projectors sum to at least IEI_E. Sum (2) with bj≤1/8b_j\le1/8 and use ∥C∥HS2≤4η2\|C\|_{\mathrm{HS}}^2\le4\eta^2. This gives

qB≥2916−2η−12η2=32p2+g(η),g(η)=516+η−272η2.(5)q_B\ge\frac{29}{16}-2\eta-12\eta^2 =\tfrac32p^2+g(\eta),\qquad g(\eta)=\tfrac5{16}+\eta-\tfrac{27}{2}\eta^2.\tag{5}

To see the contradiction using exact numbers, weaken β\beta to 13/1813/18, so η≤1/6\eta\le1/6. Let D=5[(1−η)2+η2−13/18]D=5[(1-\eta)^2+\eta^2-13/18] and a=7/4−2ηa=7/4-2\eta. By (3), Δ≤D\Delta\le D. The derivative of D/a2D/a^2 has the sign of −43/18+3η-43/18+3\eta, hence this ratio decreases. On [0,1/12][0,1/12] we have D≤25/18D\le25/18 and D/a2≤200/441D/a^2\le200/441. The correction in (4) above 3p2/23p^2/2 is at most

2518(200441−14)<2996≤g(η).\frac{25}{18}\left(\frac{200}{441}-\frac14\right)<\frac{29}{96}\le g(\eta).

On [1/12,1/6][1/12,1/6], D/a2≤90/361<1/4D/a^2\le90/361<1/4, making the correction nonpositive, whereas g(η)≥5/48g(\eta)\ge5/48. The lower bounds on gg follow by concavity and substitution at the two endpoints of each interval. Both cases contradict (5). This excludes two logical qubits and completes the boundary lemma.

7. Sharpness and the purity threshold

Postselecting a pure stabilizer state onto the defining code of VnV_n gives zero or a multiple of a logical stabilizer state. Thus Vn∈ΛnV_n\in\Lambda_n, and direct multiplication gives Tr⁡(Vn2)=2\operatorname{Tr}(V_n^2)=2. By polarity the stabilizer polytope is the intersection of the half-spaces Tr⁡(Aρ)≥0\operatorname{Tr}(A\rho)\ge0 for A∈ΛnA\in\Lambda_n. The squared distance from I/dI/d to such a hyperplane is (1/d)2/(Tr⁡(A2)−1/d)(1/d)^2/(\operatorname{Tr}(A^2)-1/d). The minimum is 1/[d(2d−1)]1/[d(2d-1)], so adding 1/d1/d gives the threshold 1/(d−1/2)1/(d-1/2).

The state ρ∗=(2I−Vn)/(2d−1)\rho_*=(2I-V_n)/(2d-1) reaches the hyperplane for VnV_n and is positive definite. Moving slightly farther on the ray from I/dI/d through ρ∗\rho_* therefore gives physical magic states approaching the threshold from above. This establishes sharpness for every nn.

A trace-distance consequence. With Dtr(ρ,σ)=∥ρ−σ∥1/2D_{\mathrm{tr}}(\rho,\sigma)=\|\rho-\sigma\|_1/2, the centered inradius is 1/(d3)1/(d\sqrt3). Both spectral bounds in 5.11 give λmax⁡(A)−λmin⁡(A)≤3\lambda_{\max}(A)-\lambda_{\min}(A)\le\sqrt3. A traceless perturbation at trace distance rr from zero has positive and negative parts of trace rr, so it can lower Tr⁡(AI/d)=1/d\operatorname{Tr}(AI/d)=1/d by at most r3r\sqrt3. Equality is attained along the direction P−−P+P_--P_+, where P±P_\pm are extremal eigenprojectors of VnV_n. The touching state I/d+(P−−P+)/(d3)I/d+(P_--P_+)/(d\sqrt3) is positive definite. OpenContext 5.11

Attribution. Liu, Haug, Ye, Liu and Roth posed the purity conjecture and proved the upper bound on its threshold. Zurel and Davis related it to maximal witness norm. Pauli compression as an inradius recursion was developed by Liu and Liu and explored here in Yixuan Wang’s notes. Wagner’s §2.1 supplies the spectral input to §2 above. The additional step in this proof is to retain the commutation relations between large Pauli coefficients and bound each remaining logical block. Triangle Criterion, Conjecture 1 · Zurel and Davis · Liu and Liu, Appendix A · Wang, OpenContext 5.6 · Wagner, OpenContext 5.11

Permanent link:open-context.jan-bulling.com/contribution/5.13

Discussion

Sign in to join the discussion.

  • Rafael Wagner · 20 Nov 2026

    I think that my Claude's solution is along similar lines