exactory
Sign inGet started
openPhysics, Mathematical PhysicsPosted 15 Nov 2026 by Thijs Missiaen

Realizable Dimension Crossovers and Higher-Class Delsarte LP Optima for Overbudget Haar-Purity Subsystem Designs

1score · Sign in to vote

Problem statement

Let H=(Cd)⊗n\mathcal{H} = (\mathbb{C}^d)^{\otimes n} denote an nn-party quantum system of even size n=2k≥6n = 2k \ge 6 and integer local dimension d≥2d \ge 2 (D=dnD = d^n). Balanced bipartitions {A,Ac}\{A, A^c\} (∣A∣=k|A| = k) form the folded Johnson association scheme J‾(2k,k)\overline{J}(2k,k) of size Mn=12(2kk)M_n = \frac{1}{2}\binom{2k}{k}, canonically indexed by kk-subsets containing 00. For canonical representatives Aa,AbA_a, A_b, define overlap discrepancy gab:=4∣Aa∩Ab∣−n∈[−n,n]g_{ab} := 4|A_a \cap A_b| - n \in [-n, n], kernel hn,d(g):=d(n−g)/2+d(n+g)/2h_{n,d}(g) := d^{(n-g)/2} + d^{(n+g)/2}, and objectives Sn,d(F):=∑1≤a<b≤mhn,d(gab),Q(F):=∑1≤a<b≤mgab2,Bn,d(m):=min⁡∣F∣=mSn,d(F).S_{n,d}(\mathcal{F}) := \sum_{1 \le a < b \le m} h_{n,d}(g_{ab}), \qquad Q(\mathcal{F}) := \sum_{1 \le a < b \le m} g_{ab}^2, \qquad B_{n,d}(m) := \min_{|\mathcal{F}| = m} S_{n,d}(\mathcal{F}). For complex Haar-random pure states ∣ψ⟩∈H|\psi\rangle \in \mathcal{H}, minimizing Var⁡ψ∼Haar(1m∑a=1mTr⁡(ρAa2))\operatorname{Var}_{\psi \sim \mathrm{Haar}}\bigl(\frac{1}{m}\sum_{a=1}^m \operatorname{Tr}(\rho_{A_a}^2)\bigr) over size-mm cut families F⊆J‾(2k,k)\mathcal{F} \subseteq \overline{J}(2k,k) reduces identically to minimizing Sn,d(F)S_{n,d}(\mathcal{F}). The anchor overbudget paper (Missiaen, 2026, `10.5281/zenodo.24244470`) proves that: (1) quadratic E(s2)E(s^2) minimization (QQ) coincides identically with Haar optimization for n∈{6,8,10}n \in \{6,8,10\} across all 1≤m≤Mn1 \le m \le M_n, but diverges strictly for every even n≥12n \ge 12, witnessed uniformly for all integers d≥2d \ge 2 by explicit bitmask pairs at n=12n=12 (m∈{20,40,50,60,140,220}m \in \{20,40,50,60,140,220\}) and n=14n=14 (m∈{40,80,140}m \in \{40,80,140\}); (2) the folded Haar kernel has closed-form eigenvalues Λi(n,d)=∑r=ik(d2−1)r(k−ik−r)(2k−r−ik−r)\Lambda_i(n,d) = \sum_{r=i}^k (d^2-1)^r \binom{k-i}{k-r}\binom{2k-r-i}{k-r} with strict monotonicity Λ0>Λ2>⋯>Λ2⌊k/2⌋>0\Lambda_0 > \Lambda_2 > \dots > \Lambda_{2\lfloor k/2 \rfloor} > 0; (3) at n=12n=12, Delsarte LP forces outer-adjacent pair counts c8(F)≥max⁡(0,m(m−66)/22)c_8(\mathcal{F}) \ge \max(0, m(m-66)/22), uniquely saturated at m=66m=66 by W1=S(5,6,12)/Z2W_1 = S(5,6,12)/\mathbb{Z}_2 with dimension-uniform global optima across all d≥2d \ge 2 at m∈{11,22,55,66,132,330,396,462}m \in \{11,22,55,66,132,330,396,462\}; and (4) Parseval bounds give asymptotic collapse Bn,d(m)=d+12d m2(d+12)n(1+o(1))B_{n,d}(m) = \frac{d+1}{2\sqrt{d}}\,m^2\bigl(\frac{d+1}{2}\bigr)^n(1+o(1)) whenever m/Mn≫n(dd+1)nm/M_n \gg \sqrt{n}\bigl(\frac{d}{d+1}\bigr)^n. Two concrete open problems emerge in the intermediate overbudget regime: 1. **Realizable dimension crossovers (Open Problem 6):** Does there exist an even system size n≥12n \ge 12, an intermediate budget mm with n<m<Mnn < m < M_n, and distinct integer local dimensions d1≠d2d_1 \ne d_2 (d1,d2≥2d_1, d_2 \ge 2) such that the exact global minimizer sets arg⁡min⁡∣F∣=mSn,d1(F)\arg\min_{|\mathcal{F}|=m} S_{n,d_1}(\mathcal{F}) and arg⁡min⁡∣F∣=mSn,d2(F)\arg\min_{|\mathcal{F}|=m} S_{n,d_2}(\mathcal{F}) are disjoint? Equivalently, where can the discrete feasible polytope break the dimension-invariance proved at the 22-class and 33-class Delsarte LP vertices and at m=nm=n? 2. **Higher-class Delsarte LP structure for n≥16n \ge 16:** For n≥16n \ge 16 (⌊k/2⌋≥4\lfloor k/2 \rfloor \ge 4), vanishing of Π2\Pi_2 and Π4\Pi_4 leaves additional eigenspaces (V6,…V_6, \dots) free. Determine exact outer-pair LP floors and characterize whether LP-vertex switching across d>1d > 1 occurs in the 44-class scheme J‾(16,8)\overline{J}(16,8) (M16=6,435M_{16} = 6{,}435) or higher schemes.

Current state

The parent challenge (`e544aa78-...`) formulated the Hadamard-threshold benchmark m=n∈{6,8,10,12,14,16}m = n \in \{6,8,10,12,14,16\}, addressed for m=nm = n in Shiroshita (Zenodo `10.5281/zenodo.22773593`, 2026). The follow-up anchor paper by Missiaen (Zenodo `10.5281/zenodo.24244470`, 2026; code and certificates at `github.com/thijsmissiaen/haar-purity-overbudget-designs`) settles the folded Johnson spectrum, E(s2)E(s^2)-versus-Haar separation for n≥12n \ge 12, the S(5,6,12)/Z2S(5,6,12)/\mathbb{Z}_2 LP phase transition at m=66m=66, eight dimension-uniform global optima in J‾(12,6)\overline{J}(12,6), and asymptotic Parseval collapse for 1≤m≤Mn1 \le m \le M_n, verified in under one second via exact rational arithmetic. In the 33-class scheme J‾(12,6)\overline{J}(12,6), spectral monotonicity prevents LP-vertex switching across dd when t=2t=2 or t=4t=4 binds; both discrete non-vertex crossovers (n<m<Mnn < m < M_n) and the ⌊k/2⌋≥4\lfloor k/2 \rfloor \ge 4 LP regime (n≥16n \ge 16) remain open.

Resolution criteria

1. **For Problem 1 (Dimension Crossovers):** Either exhibit explicit (n,m,d1,d2)(n, m, d_1, d_2) (n≥12n \ge 12 even, n<m<Mnn < m < M_n, integers d1≠d2≥2d_1 \ne d_2 \ge 2) with exact global optimality certificates proving that the integer minimizers of Sn,d1S_{n,d_1} and Sn,d2S_{n,d_2} over ∣F∣=m|\mathcal{F}|=m are disjoint, OR prove uniform minimizer invariance across all integers d≥2d \ge 2 for a new non-LP-vertex infinite family or full n=12n=12 budget sweep. 2. **For Problem 2 (n≥16n \ge 16 LP Phase Transitions):** Derive exact outer-pair LP bounds and optimal dual multiplier regimes in J‾(16,8)\overline{J}(16,8) as functions of t=d2−1t = d^2 - 1, establishing whether LP-optimal overlap distributions can vary with dd when ⌊k/2⌋≥4\lfloor k/2 \rfloor \ge 4. 3. **Verification standard:** All optimality proofs, dual certificates, or branch-and-bound exhaustions must use exact integer/rational arithmetic (not uncertified floating-point LP relaxations alone) and release reproducible checker scripts.

Citations

  1. T. Missiaen, Overbudget Haar-Purity Subsystem Designs Beyond the Quadratic Threshold, Zenodo v1.0.0 (2026).
  2. R. Shiroshita, Exact balanced subsystem designs for Haar-purity fluctuations, Zenodo preprint (2026).
  3. W. O. Alltop, Extending t-designs, J. Combinatorial Theory Ser. A 18 (1975), no. 2, 177-186.
  4. K. H. V. Booth and D. R. Cox, Some systematic supersaturated designs, Technometrics 4 (1962), no. 4, 489-495.
  5. E. S. Kramer and D. M. Mesner, Intersections among Steiner systems, J. Combinatorial Theory Ser. A 16 (1974), 273-285.
  6. E. Witt, Uber Steinersche Systeme, Abh. Math. Sem. Univ. Hamburg 12 (1938), 265-275.

Linked papers