|
1 | 1 | --- |
2 | | -title: Global residual bound on all of K3 |
| 2 | +title: From a certified atlas to a certified metric on K3 |
3 | 3 | axis: geometry |
4 | | -status: open |
| 4 | +status: active |
5 | 5 | opened: 2026-06-11 |
6 | 6 | updated: 2026-09-23 |
7 | 7 | --- |
8 | 8 |
|
9 | | -**Statement.** Extend the certified bound on the Calabi-Yau volume-form |
10 | | -residual from a box-local statement (4000 open boxes, Krawczyk-verified, with |
11 | | -the variance aggregation re-computed inside Lean) to a global bound on all of |
12 | | -K3. |
| 9 | +**Statement.** Produce a certified evaluation of the Ricci-flat (Calabi-Yau) |
| 10 | +metric on one explicit K3 surface: an algorithm that evaluates the metric |
| 11 | +with a global, machine-checked error bound, rather than a closed formula. |
13 | 12 |
|
14 | | -**Why it matters.** The box-local certificate is the framework's rigor |
15 | | -anchor; a global bound would remove the locality caveat entirely. |
| 13 | +**Why it matters.** The geometry this program rests on is only known |
| 14 | +numerically. A certified evaluator would turn "a good approximation" into a |
| 15 | +statement with a proved error, everywhere on the surface. |
16 | 16 |
|
17 | | -**Correction (K7 erratum, 2026-08-22).** The box-local certificate does not |
18 | | -bound what this statement says: it bounds an auxiliary quantity, not the |
19 | | -Calabi-Yau volume-form residual. Measured on the same 4000 points in the |
20 | | -correct pullback convention, the residual variance is 5.3e-1 (ratio 3677 to |
21 | | -the certified value) and 522 points carry no positive-definite form. What |
22 | | -stands is the surface-level certification (each box contains an exact point |
23 | | -of K3) and the method. The premise of this problem is therefore withdrawn as |
24 | | -stated; see `K7/publications/ERRATUM_v3.5.md`. |
| 17 | +**Where it stands.** The first layer is public: |
| 18 | +[Arithmon/K3](https://github.com/Arithmon/K3) (concept DOI |
| 19 | +[10.5281/zenodo.22047469](https://doi.org/10.5281/zenodo.22047469)) certifies |
| 20 | +the analytic geometry of an explicit K3 surface, a finite holomorphic atlas |
| 21 | +whose chart domains, transitions and branch continuations are machine-checked. |
| 22 | +No Ricci-flat metric is claimed there. Work toward the certified evaluation of |
| 23 | +the metric is in progress; no result is claimed. |
25 | 24 |
|
26 | | -**Known constraints.** The certificate formerly quoted here (variance |
27 | | -envelope 1321/10^7, 7.57x margin) applies to the auxiliary quantity only. Scoping concluded that a genuinely global bound is research-level |
28 | | -and that the only identified route is a Positivstellensatz / sum-of-squares |
29 | | -certificate over the orbifold chart. |
| 25 | +**Correction (K7 erratum, 2026-08-22).** An earlier formulation of this problem |
| 26 | +started from a box-local "certified bound on the Calabi-Yau residual". That |
| 27 | +bound certifies an auxiliary quantity, not the Calabi-Yau residual: in the |
| 28 | +correct pullback convention the residual variance of that ansatz is 5.3e-1 and |
| 29 | +522 of its 4000 points carry no positive-definite form. See |
| 30 | +`K7/publications/ERRATUM_v3.5.md`. |
30 | 31 |
|
31 | | -**Next step.** SOS feasibility study on a reduced chart; no commitment until |
32 | | -the central lock (exact metric) settles priorities. |
| 32 | +**Next step.** Complete the certified evaluator and publish it the way the |
| 33 | +atlas was published: with the certificates and a one-command verification. |
0 commit comments