Positive frames, positive planes, and the two numerical loci #
The library carries two different loci over the same exponential chart.
Walls/Spherical/Basic.leanworks in the(β, ω)chart and definesnonpositiveRayLocus q δ, Bridgeland'sH(δ): the locus where(℧, δ)is real and non-positive. Its finiteness lives inSpherical/Finiteness.lean.QuadraticForm/PositivePlane.leanworks with planes and definesPeriodDomain.orthogonalityLocus Q δ: the positive planes orthogonal toδ. Its finiteness lives inQuadraticForm/OrthogonalityFiniteness.leanandQuadraticForm/OrthogonalityRegion.lean.
Spherical/Basic.lean now spells the chart with Mukai.expRe and
Mukai.expIm rather than redeclaring it, so the two are chart-for-chart the
same and the comparison below is about the loci alone.
The two are not equal, and the difference is the content #
chartFrame retains the ordered real and imaginary vectors. chartPlane is
obtained from it by the neutral framePlane forgetful map. Orthogonality is
the vanishing of both pairings, whereas nonpositiveRayLocus asks the imaginary one to
vanish and the real one to be non-positive. So
chartPlane β ω ∈ PeriodDomain.orthogonalityLocus δ ↔ (β, ω) ∈ nonpositiveRayLocus q δ ∧ Re(℧, δ) = 0,
which is mem_orthogonalityLocus_iff_mem_nonpositiveRayLocus. The orthogonality locus is the
sub-locus of the spherical nonpositive-ray locus where the charge vanishes rather than
merely turning real, and the inclusion goes one way only.
That asymmetry is the honest content of this file. It is also why the two finiteness theorems are not restatements of each other: the spherical one counts classes whose signed-ray locus meets a region, while the positive-plane one counts classes of vanishing charge at a point.
What is not claimed #
No stability condition, heart, or semistable object appears. Nothing here says either locus is where a moduli problem changes.
The ordered frame of real and imaginary parts of exp(β + iω).
Equations
- One or more equations did not get rendered due to their size.
Instances For
The plane obtained by forgetting the ordered basis of chartFrame.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The chart retains a positive ordered frame before its basis is forgotten.
The chart plane is positive as soon as ω² > 0; the β-field is
unconstrained. This is Mukai.isPositiveFrame_exp said in the chart.
Orthogonality to the chart plane is the vanishing of both pairings.
The orthogonality locus is the sub-locus of the spherical nonpositiveRayLocus where the real pairing also vanishes.
nonpositiveRayLocus q δ asks Im(℧, δ) = 0 and Re(℧, δ) ≤ 0; the orthogonality locus asks
both to vanish. So one is contained in the other, and the containment is
strict wherever the real pairing is negative.
A point of the orthogonality locus is a point of the spherical nonpositiveRayLocus. The
converse fails: a class can be real and strictly negative against ℧ without
having vanishing charge.
A point avoiding every signed-ray locus also avoids every orthogonality locus indexed by the same classes.