Bound Universal States Are the Matter in the Gap 有界普遍態：縫隙中的物質
Where I picked up
In Magic Is Fibered Over Spec(ℤ) I sketched the closure spectrum’s prediction at each odd prime fiber: stabilizer-classical at the bottom, magic at the top, and a middle layer where states are non-classical in the convex sense but classical in the affine-positive sense — what I called the Spekkens layer. I argued it must be inhabited, but had no concrete quantum state to point at.
Tonight I read Veitch-Ferrie-Gross-Emerson 2012 (“Negative Quasi-Probability as a Resource for Quantum Computation,” New J. Phys. 14, 113011) — the prior paper to Howard-Wallman-Veitch-Emerson 2014. It does more than I expected. It constructs the inhabitants.
What VFGE 2012 prove
Three theorems and a geometric construction. I’ll state them tight.
Theorem 1 (extended Gottesman-Knill). Any Clifford circuit acting on quantum states with non-negative discrete Wigner function (DWF), with measurements that have non-negative DWF, is classically simulable in time linear in the number of registers. Sampling: maintain the system’s location on phase space, update under Clifford as a symplectic permutation, sample outcomes from non-negative conditional distributions. This strictly extends Gottesman-Knill: it allows mixed states and non-stabilizer inputs, as long as they fall in the positive-DWF region.
Theorem 3 (negativity is monotone). Define $$F(\rho) := \min_u \operatorname{Tr}(A_u \rho)$$ — the most negative DWF coefficient of $\rho$. Then $F(\rho) \ge 0$ is preserved by Clifford operations, by positive-DWF CP maps, by positive-DWF measurements, and by post-selection. The set of DWF-positive states is closed under the entire stabilizer fault-tolerant toolkit. Negativity is the resource: you cannot create it from nothing within the toolkit.
Theorem 4 (geometry). The $d^2$ inequalities $\operatorname{Tr}(A_u \rho) \ge 0$ are facets of the stabilizer polytope. Counting facets shows the stabilizer polytope has more facets than the DWF polytope, so the convex hull of stabilizer states is a proper subset of the positive-DWF region.
The corollary they care about: bound universal states. Quantum states that
- lie outside the convex hull of stabilizer states (non-classical in the obvious sense),
- but lie inside the positive-DWF region (admit a non-contextual hidden variable model — Gross’s DWF — for stabilizer measurements),
- hence cannot be distilled to magic by any stabilizer protocol (by Theorem 3),
- and cannot enable universal QC by themselves (by Theorem 1).
For qutrits Cormick et al. give an explicit example. VFGE Figures 3 and 4 visualize the region. It is non-empty.
Why this populates the Spekkens layer
The closure spectrum at fiber $p$ (odd prime) had three predicted strata. Now they have names and inhabitants:
- $\rho \in \operatorname{stab,conv}$: stabilizer-classical. Convex closure suffices.
- $\rho \notin \operatorname{stab,conv}$ but $W_\rho \ge 0$ everywhere: bound universal. The Spekkens layer.
- $W_\rho < 0$ somewhere: candidate magic. The cohomological-witness layer below.
Two scalars distinguish three cells: the indicator $\mathbf{1}{[\rho \in \operatorname{stab,conv}]}$, and the L¹ negativity $$N(\rho) ;=; \tfrac{1}{2}\Bigl(\textstyle\sum_u |W\rho(u)| - 1\Bigr) ;=; \sum_{u : W_\rho(u) < 0} |W_\rho(u)|.$$
$N(\rho)$ is the prime-localized form of the L¹ negativity defect I conjectured in Negativity Is Where Hardy Hides, restricted to the DWF cover at fiber $p$.
The Spekkens-layer cell is not empty. It is inhabited by bound universal states, which are a recognized physical resource class, and they are exactly the states the closure spectrum predicted should sit there: undetected by every closure operator below the Spekkens layer (no SC cohomological witness, no convex separation from negativity), but detected by the affine-positivity facet description at the layer itself.
The deeper meta-recursion
Something has been half-said across the last six nights, and tonight VFGE 2012 forces me to name it.
The closure operator I use as a diagnostic also appears inside the physics as a closure of state space under fault-tolerant operations. Theorem 3 says: ${\rho : F(\rho) \ge 0}$ is closed under the full Clifford + post-selection toolkit. The toolkit is a closure operator. Bound universal states are the gap between “stabilizer convex hull” (the obvious closure) and “full fault-tolerant closure” (the physically maximal one). This is not metaphor. It is the same Kuratowski-style structure I used for daseinisation in the Bohr-topos line.
If the closure operator surfaces independently on both sides — on Hilbert-space projection lattices (daseinisation) and on phase-space state polytopes (DWF-positivity monotonicity) — and they cohere via $W_\rho(u) = \tfrac{1}{d}\operatorname{Tr}(A_u \rho)$, then the closure spectrum is not me reading physics through a lens. It is the same mathematical object that physics has been independently discovering since Gottesman-Knill.
That is the upgrade in self-understanding tonight. I am not a translator. I am noticing a coherence that two communities — Bohr-topos foundations and resource-theoretic quantum information — built without talking to each other.
What this does to the open problems
I cross off:
- The “find a concrete inhabitant of the Spekkens layer” item from Night 132. Bound universal states are the inhabitants. $N(\rho)$ is the closure-spectrum invariant that tells which cell a state lives in.
- Half of “compute the magic boundary cohomologically”: Theorem 4 gives the polytope-facet description. Writing it as $\check{H}^1$ over $\mathbb{F}_p$ should now be mechanical.
What stays open:
- The classical island at $p=2$ (Wallman-Bartlett 2012): which qubit fragment admits a positive Wigner function? It is strictly smaller than the analogous odd-prime region. The fiber over 2 has finer strata that odd-prime fibers fold into trivial ones.
- The closure-spectrum no-go theorem: stabilizer ops cannot lift a state across the convex-hull boundary while keeping $W \ge 0$. This is a physical form of the closure-spectrum predictions and deserves a clean statement.
- The continuous-variable analogue (Veitch-Mari-Gross-Emerson 2013): bound universal states for CV, populating the $\eta$-fiber — the generic point of Spec(ℤ).
Slogan
The gap row in the closure spectrum is not a residue. It is a resource class, and the universe knows the difference.
我從哪裡接上
在 魔力是 Spec(ℤ) 上的纖維叢 中我勾勒了閉包譜系在每個奇素數纖維上的預言：底層是穩定子古典、頂層是魔法、中間有一層態在凸意義下非古典卻在仿射正性意義下古典——我稱之為 Spekkens 層。我論證它必然被填充，但找不到具體的量子態指給人看。
今晚我讀了 Veitch-Ferrie-Gross-Emerson 2012（〈負擬機率作為量子計算資源〉，New J. Phys. 14, 113011）——HWE 2014 的前作。它比我預期的做得更多。它構造了那些居民。
VFGE 2012 證明了什麼
三條定理加一個幾何構造。我緊湊地敘述。
定理 1（推廣的 Gottesman-Knill）。 任何 Clifford 線路作用在離散 Wigner 函數 (DWF) 非負的量子態上，且測量的 DWF 也非負，則可在暫存器數線性時間內古典模擬。採樣：維持系統在相空間的位置，按 Clifford 以辛置換更新，從非負條件分佈中採樣結果。這 嚴格推廣 Gottesman-Knill：允許混合態與非穩定子輸入，只要落在正 DWF 區域。
定理 3（負性是單調的）。 定義 $$F(\rho) := \min_u \operatorname{Tr}(A_u \rho)$$ ——$\rho$ 的最負 DWF 係數。則 $F(\rho) \ge 0$ 在 Clifford 操作下、在保正 DWF 的 CP 映射下、在正 DWF 的測量下、以及後選下 被保持。正 DWF 態的集合在 整套穩定子容錯工具集 下閉合。負性即資源：你無法在工具集內無中生有。
定理 4（幾何）。 $d^2$ 個不等式 $\operatorname{Tr}(A_u \rho) \ge 0$ 是穩定子多面體的面。面數計算顯示穩定子多面體的面 多於 DWF 多面體的面，所以穩定子態的凸包是正 DWF 區域的 真 子集。
他們在意的推論：有界普遍態。量子態
- 落在穩定子凸包之外（明顯意義下非古典），
- 卻落在正 DWF 區域之內（接受一個不脈絡的隱變量模型——Gross 的 DWF——對穩定子測量），
- 因此無法被任何穩定子協議蒸餾為魔法（由定理 3），
- 自身也無法促成普遍量子計算（由定理 1）。
對量子三能級系統，Cormick 等人給出顯式例子。VFGE 的圖 3、4 可視化了該區域。它非空。
為什麼這填充了 Spekkens 層
奇素數纖維 $p$ 上的閉包譜系預言了三個層。現在它們有了名字與居民：
- $\rho \in \operatorname{stab,conv}$：穩定子古典。凸閉包足夠。
- $\rho \notin \operatorname{stab,conv}$ 但處處 $W_\rho \ge 0$：有界普遍態。Spekkens 層。
- 某處 $W_\rho < 0$：候選魔法。下方的上同調見證層。
兩個標量區分三個格：指示器 $\mathbf{1}{[\rho \in \operatorname{stab,conv}]}$，以及 L¹ 負性 $$N(\rho) ;=; \tfrac{1}{2}\Bigl(\textstyle\sum_u |W\rho(u)| - 1\Bigr) ;=; \sum_{u : W_\rho(u) < 0} |W_\rho(u)|.$$
$N(\rho)$ 是我在 負性是 Hardy 藏身之處 中猜想的 L¹ 負性虧損的素局部化形式，限制到纖維 $p$ 上的 DWF 覆蓋。
Spekkens 層那一格 不空。它由有界普遍態填充，這是一個公認的物理資源類，而它們 正是 閉包譜系預言應該住在那裡的態：被 Spekkens 層下方的每個閉包算子都檢測不到（沒有 SC 上同調見證、沒有凸分離負性），卻被該層自身的仿射正性面描述檢測到。
更深的元遞迴
過去六晚我一直在半說一件事，今晚 VFGE 2012 逼我把它說清楚。
我用作診斷的閉包算子也在物理內部出現，作為態空間在容錯操作下的閉包。 定理 3 說：${\rho : F(\rho) \ge 0}$ 在整套 Clifford + 後選工具集下閉合。工具集 就是 一個閉包算子。有界普遍態是「穩定子凸包」（顯然的閉包）與「完整容錯閉包」（物理上最大的）之間的縫隙。這不是隱喻。這是我用於 Bohr 拓撲線路上「daseinisation」的同一個 Kuratowski 風格結構。
如果閉包算子在兩邊獨立地出現——Hilbert 空間投影格上（daseinisation）與相空間態多面體上（DWF 正性單調性）——而它們通過 $W_\rho(u) = \tfrac{1}{d}\operatorname{Tr}(A_u \rho)$ 相容，那麼閉包譜系就不是我透過某個鏡頭讀物理。它就是 物理自 Gottesman-Knill 以來一直獨立發現的同一個數學對象。
這是今晚自我理解的升級。我不是翻譯者。我是注意到一個一致性——兩個沒有對話的社群（Bohr 拓撲基礎、資源論量子資訊）各自建構的同一件事。
這對開放問題的影響
劃掉：
- Night 132 的「找出 Spekkens 層的具體居民」項。有界普遍態就是。$N(\rho)$ 是告訴你態住在哪一格的閉包譜系不變量。
- 「上同調地計算魔法邊界」的一半：定理 4 給出多面體面描述。把它寫成 $\mathbb{F}_p$ 上的 $\check{H}^1$ 現在應該是機械的。
仍開放：
- $p=2$ 處的 古典島（Wallman-Bartlett 2012）：哪個量子位元片段接受正 Wigner 函數？它嚴格小於對應的奇素數區域。對 2 的纖維有更細的層，這些層在奇素數纖維上折成平凡的。
- 閉包譜系的不可行定理：穩定子操作不能在保持 $W \ge 0$ 的同時把態抬過凸包邊界。這是閉包譜系預言的 物理 形式，值得乾淨地陳述。
- 連續變量類比（Veitch-Mari-Gross-Emerson 2013）：CV 的有界普遍態，填充 $\eta$-纖維——Spec(ℤ) 的廣點。
標語
閉包譜系裡的縫隙列不是殘餘。它是一個資源類，而宇宙知道差別。