Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
45 changes: 45 additions & 0 deletions .github/workflows/hosted-math.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,45 @@
name: Hosted mathematics

on:
pull_request:

permissions:
contents: read

jobs:
hosted-math:
name: Rendered mathematics
runs-on: ubuntu-24.04
timeout-minutes: 20
steps:
- name: Check out the exact proposed source
uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
with:
ref: ${{ github.event.pull_request.head.sha }}
persist-credentials: false
- name: Set up Node.js
uses: actions/setup-node@949feb2413d6458794dcd2491c4babbbce0c15c1 # v7
with:
node-version: '24'
- name: Install the pinned capture dependencies
run: |
npm install --prefix "$RUNNER_TEMP/hosted-deps" --no-audit --no-fund playwright@1.56.1 mathjax-full@3.2.1
"$RUNNER_TEMP/hosted-deps/node_modules/.bin/playwright" install --with-deps chromium
- name: Compare source mathematics with GitHub
env:
SOURCE_REVISION: ${{ github.event.pull_request.head.sha }}
run: |
NODE_PATH="$RUNNER_TEMP/hosted-deps/node_modules" node verification/hosted/capture_hosted_math.cjs \
--root "$GITHUB_WORKSPACE" \
--repository "$GITHUB_REPOSITORY" \
--revision "$SOURCE_REVISION" \
--contract verification/hosted/math-contract.json \
--output "$RUNNER_TEMP/hosted-evidence"
- name: Preserve the capture and reports
if: always()
uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7
with:
name: hosted-math-${{ github.event.pull_request.head.sha }}
path: ${{ runner.temp }}/hosted-evidence/
if-no-files-found: warn
retention-days: 7
8 changes: 6 additions & 2 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -9,11 +9,11 @@ The central result is a structural theorem. If the best input is identified and
| A best-input label, or the complete fidelity ordering | Every universally nondegrading SEP success map has the fidelity-preserving extraction form. |
| Unknown ordering | No nonzero SEP success map can always achieve the best input fidelity. |

Both statements hold for every fixed finite number of inputs $`n\ge2`$, every local dimension $`d\ge2`$, and every fidelity floor $`1/d\le f_\star<1`$. The inputs are independent isotropic pairs,
Both statements hold for every fixed finite number of inputs $`n\ge2`$, every local dimension $`d\ge2`$, and every fidelity floor $`1/d\le f_\star\lt 1`$. The inputs are independent isotropic pairs,

```math
\omega(F_i)=F_iP+(1-F_i)\frac{I-P}{d^2-1},
\qquad f_\star<F_i<1,
\qquad f_\star\lt F_i\lt 1,
```

with a known target $`P=|\Phi_d\rangle\langle\Phi_d|`$. Success can become arbitrarily rare near perfect inputs; the theorem assumes no success floor at that boundary.
Expand All @@ -34,6 +34,10 @@ Start with the [model, information assumptions and connected reading route](scie

The universality task and earlier no-go results are due to [Zang et al. (2025)](https://doi.org/10.1103/PhysRevLett.134.190803). The [comparison](references/RELATED_WORK.md#relation-to-the-2025-universal-purification-no-go) identifies the arbitrary-SEP high-fidelity strengthening and keeps the earlier catalytic Clifford results separate.

## Learn the argument

The single educational anchor is John Watrous, *The Theory of Quantum Information* (2018). The [Watrous bridge](learning/WATROUS_BRIDGE.md) connects its notation, fidelity convention, quantum operations and separability results to this project's proof. The [30-milestone map](learning/MILESTONES.md) develops 12 background milestones followed by 18 milestones tracing the argument and its physical boundaries. Each milestone links back to the canonical scientific account.

## Reproduce the checks

From the repository root:
Expand Down
464 changes: 464 additions & 0 deletions learning/MILESTONES.md

Large diffs are not rendered by default.

276 changes: 276 additions & 0 deletions learning/WATROUS_BRIDGE.md

Large diffs are not rendered by default.

6 changes: 6 additions & 0 deletions llms.txt
Original file line number Diff line number Diff line change
Expand Up @@ -19,15 +19,21 @@ Canonical entrypoints:
the distinct catalytic Clifford and unknown-target results.
- references/references.bib: structured primary-source bibliography.
- verification/README.md: finite-check commands, expected behavior and limits.
- learning/WATROUS_BRIDGE.md: the single Watrous (2018) educational anchor,
verified textbook pointers, convention translation and one worked example.
- learning/MILESTONES.md: exactly 30 project milestones, with 12 background
and 18 narrative milestones; canonical proofs remain in science/.

Reader routes:
- For the story, read model, interpretation and the worked examples.
- For the theorem, read model, branch lemmas and ordered purification in order.
- For attribution, use the exact theorem and source-version comparisons.
- For reproduction, run the two documented scripts with the pinned dependency.
- For foundations, use the Watrous bridge and follow the milestone dependencies.

Preserve these distinctions:
- Fidelity is overlap with the fixed target, without a square root.
It equals the square of Watrous's root fidelity against the pure Bell target.
- A nonzero success map has positive success on interior full-rank inputs,
but success may vanish at a perfect boundary. No boundary division is used.
- The map is fixed over the family. Adaptive measurements and classical
Expand Down
12 changes: 12 additions & 0 deletions references/references.bib
Original file line number Diff line number Diff line change
@@ -1,6 +1,18 @@
% Primary-source bibliography for Entanglement-Purification-Rigidity.
% Metadata and cited versions verified on 2026-11-03.

% Educational anchor; author's standard-layout manuscript: https://cs.uwaterloo.ca/~watrous/TQI/
@book{Watrous2018,
author = {Watrous, John},
title = {The Theory of Quantum Information},
publisher = {Cambridge University Press},
year = {2018},
doi = {10.1017/9781316848142},
isbn = {9781107180567},
url = {https://cs.uwaterloo.ca/~watrous/TQI/},
note = {Section and proposition pointers refer to the author's standard-layout manuscript}
}

% Publisher: https://doi.org/10.1103/PhysRevLett.134.190803
% Version checked: https://arxiv.org/abs/2407.21760v2
@article{Zang2025NoGo,
Expand Down
4 changes: 2 additions & 2 deletions science/model.md
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,7 @@ R=\frac{I-P}{d^2-1},
\omega(F)=FP+(1-F)R.
```

The Bell fidelity of a normalized output state $`\rho`$ is $`F(\rho)=\operatorname{Tr}(P\rho)`$. Fix a threshold $`f_\star\in[1/d,1)`$. The theorem will assume a guarantee only for isotropic inputs with independently variable fidelities in $`(f_\star,1)`$. Every such input is positive definite. In particular, the argument does not need low-fidelity or separable inputs.
The Bell fidelity of a normalized output state $`\rho`$ is $`F(\rho)=\mathrm{Tr}(P\rho)`$. Fix a threshold $`f_\star\in[1/d,1)`$. The theorem will assume a guarantee only for isotropic inputs with independently variable fidelities in $`(f_\star,1)`$. Every such input is positive definite. In particular, the argument does not need low-fidelity or separable inputs.

A fixed accepted branch is a nonzero completely positive, trace-nonincreasing map

Expand All @@ -42,7 +42,7 @@ For
```math
\Omega(\boldsymbol F)=\bigotimes_{i=1}^n\omega(F_i),
\qquad
p(\boldsymbol F)=\operatorname{Tr}\mathcal E(\Omega(\boldsymbol F)),
p(\boldsymbol F)=\mathrm{Tr}\mathcal E(\Omega(\boldsymbol F)),
```

the conditional output is $`\mathcal E(\Omega)/p`$. Because $`\mathcal E\ne0`$ and $`\Omega>0`$, one has $`p(\boldsymbol F)>0`$ at every interior tuple. This imposes no lower bound on that probability and no requirement that it remain positive at a boundary tuple.
Expand Down
44 changes: 23 additions & 21 deletions science/ordered_purification.md
Original file line number Diff line number Diff line change
Expand Up @@ -16,9 +16,9 @@ F_1\ge \max_{i>1}F_i,
the fixed separable success map satisfies

```math
\operatorname{Tr}\!\left[P\mathcal E(\Omega(\boldsymbol F))\right]
\ge F_1\operatorname{Tr}\mathcal E(\Omega(\boldsymbol F)).
\tag{1}
\mathrm{Tr}\!\left[P\mathcal E(\Omega(\boldsymbol F))\right]
\ge F_1\mathrm{Tr}\mathcal E(\Omega(\boldsymbol F)).
\qquad\text{(1)}
```

Then every nonzero product Kraus operator in any finite separable decomposition of $`\mathcal E`$ has the form
Expand All @@ -27,7 +27,7 @@ Then every nonzero product Kraus operator in any finite separable decomposition
A_r=s_rU_r\otimes\langle a_r|,
\qquad
B_r=t_r\overline{U_r}\otimes\langle b_r|,
\tag{2}
\qquad\text{(2)}
```

where $`U_r`$ is unitary on the first local input system, $`s_r,t_r\ne0`$, and the bras act on the local spectator systems $`2,\ldots,n`$.
Expand All @@ -39,15 +39,15 @@ Consequently, for every state $`\rho`$ of pair 1 and every spectator state $`\ta
=\sum_r q_r(\tau)V_r\rho V_r^\dagger,
\qquad
V_r=U_r\otimes\overline{U_r},
\tag{3}
\qquad\text{(3)}
```

with

```math
q_r(\tau)=|s_rt_r|^2
\langle a_r b_r|\tau|a_r b_r\rangle\ge0.
\tag{4}
\qquad\text{(4)}
```

Whenever $`\sum_rq_r(\tau)>0`$, the accepted output has Bell fidelity exactly $`F(\rho)`$. The acceptance probability depends only on the spectators. Conversely, every nonzero trace-nonincreasing map of the form (2) satisfies (1).
Expand All @@ -66,16 +66,16 @@ Fix one spectator tuple with every fidelity strictly inside $`(f_\star,1)`$, and
All first-pair fidelities sufficiently close to one satisfy the ordering promise. Taking $`F_1\uparrow1`$ in the homogeneous inequality (1) yields

```math
\operatorname{Tr}\!\left[(I-P)\mathcal E(P\otimes\tau)\right]=0.
\tag{15}
\mathrm{Tr}\!\left[(I-P)\mathcal E(P\otimes\tau)\right]=0.
\qquad\text{(15)}
```

Each summand in the product-Kraus decomposition is nonnegative. Equivalently,

```math
\left\|(I-P)K_r(P\otimes\tau^{1/2})\right\|_2^2=0
\quad\text{for every }r.
\tag{16}
\qquad\text{(16)}
```

The norm in (16) is the Hilbert–Schmidt norm. Since $`\tau`$ is positive definite, every $`K_r`$ maps $`\mathcal S_1`$ into the output Bell ray. [Lemma 2](product_branch_rigidity.md#product-kraus-extraction) divides the nonzero branches into two classes: extraction branches of the form (5), and branches annihilating $`\mathcal S_1`$.
Expand All @@ -87,7 +87,7 @@ w_r=\langle a_rb_r|\tau|a_rb_r\rangle>0,
\qquad
g_r=w_r|\lambda_r|^2,
\qquad
h_r=w_r\operatorname{Tr}(L_rRL_r^\dagger).
h_r=w_r\mathrm{Tr}(L_rRL_r^\dagger).
```

Its contribution to the accepted Bell numerator on $`\omega(F)\otimes\tau`$ is $`Fg_r`$, and its contribution to the success probability is $`Fg_r+(1-F)h_r`$. Lemma 3 gives $`h_r\ge g_r`$.
Expand All @@ -97,9 +97,9 @@ For an annihilating branch, put
```math
\sigma_r=K_r(R\otimes\tau)K_r^\dagger,
\qquad
c_r=\operatorname{Tr}(P\sigma_r),
c_r=\mathrm{Tr}(P\sigma_r),
\qquad
z_r=\operatorname{Tr}(\sigma_r).
z_r=\mathrm{Tr}(\sigma_r).
```

Its Bell numerator and success probability are $`(1-F)c_r`$ and $`(1-F)z_r`$, respectively. Lemma 4 gives $`z_r>c_r`$.
Expand All @@ -110,33 +110,35 @@ Sum the extraction coefficients to obtain $`g=\sum g_r`$, $`h=\sum h_r`$, and th
N(F)=Fg+(1-F)c,
\qquad
p(F)=Fg+(1-F)(h+z).
\tag{17}
\qquad\text{(17)}
```

For $`F<1`$, the requirement $`N(F)\ge Fp(F)`$ is equivalent to

```math
c+F(g-h-z)\ge0.
\tag{18}
\qquad\text{(18)}
```

Letting $`F\uparrow1`$ gives

```math
(g-h)+(c-z)\ge0.
\tag{19}
\qquad\text{(19)}
```

Both terms on the left are nonpositive. The second is strictly negative if any nonzero annihilating branch is present. Hence no such branch exists, and $`h=g`$. Every extraction weight $`w_r`$ is strictly positive, so each individual extraction branch must attain equality in Lemma 3. This gives exactly the unitary form (2).

For any independent input $`\rho\otimes\tau`$, that form gives (3)–(4). Every $`V_r=U_r\otimes\overline{U_r}`$ fixes $`|\Phi_d\rangle`$, so

```math
\operatorname{Tr}\!\left[P\mathcal E(\rho\otimes\tau)\right]
\mathrm{Tr}\!\left[P\mathcal E(\rho\otimes\tau)\right]
=F(\rho)\sum_rq_r(\tau),
\qquad
\operatorname{Tr}\mathcal E(\rho\otimes\tau)=\sum_rq_r(\tau).
\tag{20}
```

```math
\mathrm{Tr}\mathcal E(\rho\otimes\tau)=\sum_rq_r(\tau).
\qquad\text{(20)}
```

This proves exact fidelity preservation whenever the protocol succeeds, as well as the converse. For a promise of complete ordering, choose the fixed spectator fidelities strictly ordered. The same limit $`F_1\uparrow1`$ stays within that promise and proves the same Kraus classification. $`\square`$
Expand All @@ -148,12 +150,12 @@ This proves exact fidelity preservation whenever the protocol succeeds, as well
```math
F\!\left(\frac{\mathcal E(\Omega(\boldsymbol F))}{p(\boldsymbol F)}\right)
\ge\max_iF_i
\tag{21}
\qquad\text{(21)}
```

for every independently variable tuple $`F_i\in(f_\star,1)`$.

**Proof.** Such a map would satisfy Theorem 1 on the subfamily where pair 1 is best. Its Kraus structure would therefore preserve the first pair's fidelity for every independent isotropic tuple, including tuples outside that ordering region. Choose $`F_2>F_1`$. The input is positive definite, so the nonzero map succeeds with positive probability, but its conditional fidelity is $`F_1<F_2`$. This contradicts (21). $`\square`$
**Proof.** Such a map would satisfy Theorem 1 on the subfamily where pair 1 is best. Its Kraus structure would therefore preserve the first pair's fidelity for every independent isotropic tuple, including tuples outside that ordering region. Choose $`F_2>F_1`$. The input is positive definite, so the nonzero map succeeds with positive probability, but its conditional fidelity is $`F_1\lt F_2`$. This contradicts (21). $`\square`$

Since the promised isotropic family is a subset of the full set of high-fidelity states, the impossibility also applies to a guarantee over all independent states above the same fidelity threshold.

Expand Down
Loading