Progress · independent proof diagram
A period criterion for individual cycle classes
A class belongs to the pairing domain if its specified local obstruction periods vanish at every relevant bad place. This criterion can be applied to individual classes and their rational linear combinations.
Conditional criterion · Source edition v0.5.1
Domain of the result
Classes and satisfying the period criterion in their respective degrees.
Read the diagrams from top to bottom. Select a numbered connection for its exact hypotheses and deduction, or follow a node to its source. Analytic, model and comparison assumptions retain their stated scope.
How to read the logic
Read from top to bottom. Parallel premises meet at a numbered connection, then lead to a conclusion. Select a connection to read its hypotheses and deduction; select a node to read the linked statement.
The full scope of each occurrence is recorded below its diagram. A proved conditional implication and an open geometric premise keep their separate statuses.
LOGIC 01
Finite obstruction periods put specified classes in the pairing domain
Fix the witness and comparison before the cycle, check the relevant bad-place coordinates, and treat all other places by good reduction.
Scope. Let X/F be smooth projective geometrically integral. The FD17 fixed arithmetic witness, VT17 comparison, original-field reconstruction and height hypotheses are retained. Take z∈Vʳ and w∈Vⁿ⁺¹⁻ʳ.
Scroll inside the frame to follow a large branch. “Fit width” shows the whole layout; “Readable size” restores the text to 100%.
Read the 1 connection
Each numbered card gives the exact inputs, the deduction and its sources.
01AND · BOTH degree-wise certificates and the height construction
Premises and stated hypotheses
- HypothesisThe selected periods vanish in BOTH complementary degrees
Use one fixed arithmetic witness chosen before the classes and a finite set S containing its bad places and the places above an auxiliary prime. For each w∈S use the full local degree [Ew:Fv].
Scope & sources - ConditionalEach zero period vector gives the original local kernel condition
The comparison is literal equality with rational numerical coordinates. It preserves the original actual sheaf and every test in the Green conclusion.
Scope & sources - HypothesisGood reduction kills the remaining local coordinates
Smooth proper specialization sends the zero geometric generic class to zero on the special fibre, so all algebraic degree pairings vanish. The model detector kills the upstairs obstruction; dominant and finite-constant injection return to Fv.
Scope & sources - ConditionalH16 pairs the two admitted factors
All reciprocity, moving, good-place and archimedean normalization premises remain in force.
Scope & sources
How the deduction works
Apply the same argument separately in the two complementary degrees. Each period vector has the correct local kernel; vanishing therefore puts each class in its H factor.
Conclusion / target
The criterion applies to individual classes and their rational linear combinations. It does not prove that all homologically trivial classes have zero periods.
Scope & sourcesEvidence and further reading
Back to connection 01 in the diagram ↑All 6 nodes: scope, status and sources
ConditionalThe two certified classes admit the global height pairing
The criterion applies to individual classes and their rational linear combinations. It does not prove that all homologically trivial classes have zero periods.
Back to this node in the diagram ↑HypothesisThe selected periods vanish in BOTH complementary degrees
Use one fixed arithmetic witness chosen before the classes and a finite set S containing its bad places and the places above an auxiliary prime. For each w∈S use the full local degree [Ew:Fv].
Back to this node in the diagram ↑ConditionalEach zero period vector gives the original local kernel condition
The comparison is literal equality with rational numerical coordinates. It preserves the original actual sheaf and every test in the Green conclusion.
Back to this node in the diagram ↑HypothesisGood reduction kills the remaining local coordinates
Smooth proper specialization sends the zero geometric generic class to zero on the special fibre, so all algebraic degree pairings vanish. The model detector kills the upstairs obstruction; dominant and finite-constant injection return to Fv.
Back to this node in the diagram ↑ConditionalH16 pairs the two admitted factors
All reciprocity, moving, good-place and archimedean normalization premises remain in force.
Back to this node in the diagram ↑OpenEvery homologically trivial class has all these periods zero
The finite-dimensional description neither proves this vanishing nor supplies an effective algorithm for all numerical cycle groups. Coordinate choices change with the witness; the kernel does not.
Back to this node in the diagram ↑LOGIC 02
The fixed model quotient and the actual obstruction have the same kernel
Chow localization makes the coordinates well-defined; numerical duality and the stated comparison prove the two kernel inclusions.
Scope. Fix K, r and a finite L/K, a generically finite projective dominant g:Y→XL, and a regular projective reduced strictly semistable model of Y, before choosing Z. These are the FD17 witness data.
Scroll inside the frame to follow a large branch. “Fit width” shows the whole layout; “Readable size” restores the text to 100%.
Read the 3 connections
Each numbered card gives the exact inputs, the deduction and its sources.
01AND · BOTH covariance and the reverse model comparison
Premises and stated hypotheses
- HypothesisActual obstruction pullback and native finite-constant comparison
Keep the original absolute value and the full finite-constant degree.
Scope & sources - HypothesisVT17: zero obstruction forces the model image condition
This explicit comparison remains a companion hypothesis; a numerical dimension count does not replace it.
Scope & sources
How the deduction works
The necessary direction depends on VT17. It is not inferred from the existence of an arbitrary selected Green-pair construction alone.
Conclusion / target
Pull zero obstruction through finite constants and g; VT17 places the model restriction in im Aᵣ₋₁, so every coordinate against its transpose kernel vanishes.
Scope & sourcesEvidence and further reading
Back to connection 01 in the diagram ↑02AND · ALL correction, detection and descent steps
Premises and stated hypotheses
- Conditionalν = 0 gives rᵣΘW ∈ im Aᵣ₋₁
The perfect rational component pairing identifies Cʳ/im Aᵣ₋₁ with (ker Aₙ₋ᵣ)∨. Subtract a rational vertical preimage to obtain componentwise numerical zero.
Scope & sources - HypothesisThe full vertical numerical criterion kills upstairs Ob
All vertical testing cycles and native degrees are retained.
Scope & sources - HypothesisDominant injection followed by full finite-constant injection
First descend along g, then XL→X using m=[L:K]. No geometric sheet count is substituted.
Scope & sources
How the deduction works
Choose rational Chow representatives of a numerical preimage. This preserves the generic Chow class and does not require a specified cycle to pull back in proper position.
Conclusion / target
Perfect numerical duality gives a vertical correction. The model detector kills the corrected upstairs curvature; two injective pullbacks then return the zero class to X/K.
Scope & sourcesEvidence and further reading
Back to connection 02 in the diagram ↑03AND · ALL well-definedness and two kernel inclusions
Premises and stated hypotheses
- ConditionalChanging the Chow extension does not change ν
Two extensions differ by actual rational vertical cycles Σjᵢ*uᵢ. Their restrictions differ by Aᵣ₋₁u, which annihilates ker Aₙ₋ᵣ by transpose duality. Adding extensions gives rational linearity.
Scope & sources - ConditionalOb = 0 implies ν = 0
Pull zero obstruction through finite constants and g; VT17 places the model restriction in im Aᵣ₋₁, so every coordinate against its transpose kernel vanishes.
Scope & sources - Conditionalν = 0 implies Ob = 0
Perfect numerical duality gives a vertical correction. The model detector kills the corrected upstairs curvature; two injective pullbacks then return the zero class to X/K.
Scope & sources
How the deduction works
The source proves an equality of kernels, not merely a finite list of possible numerical tests.
Conclusion / target
This equality is conditional on VT17, the numerical detector, actual covariance, and dominant and full finite-constant injection.
Scope & sourcesEvidence and further reading
Back to connection 03 in the diagram ↑All 10 nodes: scope, status and sources
definitionVmodel = ker Aₙ₋ᵣ and ν(z)(v)=〈rᵣΘW,v〉
Cᵃ is the direct sum of rational numerical cycle spaces on the native special components. W represents g*p*z and ΘW is a model extension. A is restriction after vertical Gysin.
Back to this node in the diagram ↑Conditionalν is rational linear, representative-independent, and ker ν = ker Ob
This equality is conditional on VT17, the numerical detector, actual covariance, and dominant and full finite-constant injection.
Back to this node in the diagram ↑ConditionalChanging the Chow extension does not change ν
Two extensions differ by actual rational vertical cycles Σjᵢ*uᵢ. Their restrictions differ by Aᵣ₋₁u, which annihilates ker Aₙ₋ᵣ by transpose duality. Adding extensions gives rational linearity.
Back to this node in the diagram ↑ConditionalOb = 0 implies ν = 0
Pull zero obstruction through finite constants and g; VT17 places the model restriction in im Aᵣ₋₁, so every coordinate against its transpose kernel vanishes.
Back to this node in the diagram ↑HypothesisActual obstruction pullback and native finite-constant comparison
Keep the original absolute value and the full finite-constant degree.
Back to this node in the diagram ↑HypothesisVT17: zero obstruction forces the model image condition
This explicit comparison remains a companion hypothesis; a numerical dimension count does not replace it.
Back to this node in the diagram ↑Conditionalν = 0 implies Ob = 0
Perfect numerical duality gives a vertical correction. The model detector kills the corrected upstairs curvature; two injective pullbacks then return the zero class to X/K.
Back to this node in the diagram ↑Conditionalν = 0 gives rᵣΘW ∈ im Aᵣ₋₁
The perfect rational component pairing identifies Cʳ/im Aᵣ₋₁ with (ker Aₙ₋ᵣ)∨. Subtract a rational vertical preimage to obtain componentwise numerical zero.
Back to this node in the diagram ↑HypothesisThe full vertical numerical criterion kills upstairs Ob
All vertical testing cycles and native degrees are retained.
Back to this node in the diagram ↑HypothesisDominant injection followed by full finite-constant injection
First descend along g, then XL→X using m=[L:K]. No geometric sheet count is substituted.
Back to this node in the diagram ↑LOGIC 03
Realizing every rational coordinate by an original-field actual test
Framed vertical tests, a graph adjoint and the full finite trace realize the coordinates exactly, not just up to an unspecified scalar.
Scope. The witness and basis are fixed before the rational cycle Z. The statement concerns all rational Chow classes locally, with no homological assumption; its comparison hypotheses remain explicit.
Scroll inside the frame to follow a large branch. “Fit width” shows the whole layout; “Readable size” restores the text to 100%.
Read the 2 connections
Each numbered card gives the exact inputs, the deduction and its sources.
01AND · ALL steps yielding exact rational periods
Premises and stated hypotheses
- HypothesisVT17 constructs fixed global D-closed βj on Y
Choose rational basis cycles once. The framed comparison gives ∫Wβj=cL〈rᵣ[W̄],vj〉 for every W, with the same cL and the complete resolution differential.
Scope & sources - HypothesisAD17 gives global D-closed bj on XL
Against every separately closed ω, ∫XLbjω=∫Yβjg*ω. Curved Green identities and Stokes remove the global D-boundary between g*ΩZL and ΩW. No proper-position support premise is inserted.
Scope & sources - HypothesisFull actual finite trace and native measure comparison
Trace commutes with D and pairs with every actual cycle test, including native flat lengths. This is a full actual trace input, not merely the restricted scalar–curvature trace.
Scope & sources
How the deduction works
The graph-adjoint equality gives ∫ZLbj=cLν. Full finite trace gives an original-K actual form; native K integration upstairs is [L:K] times native L integration, canceling the factor exactly.
Conclusion / target
Here cL=−log|πL|=cK/e(L/K). Use the extending absolute value, full field degree and literal D-closedness.
Scope & sourcesEvidence and further reading
Back to connection 01 in the diagram ↑02AND · ALL construction and both directions of the criterion
Premises and stated hypotheses
- Conditionalηj = TrKXL/X(bj) / ([L:K] cL)
Here cL=−log|πL|=cK/e(L/K). Use the extending absolute value, full field degree and literal D-closedness.
Scope & sources - ConditionalStrong G forces every fixed period to vanish
Projectivity makes each global ηj compact as a test. Insert Dηj=0 into the strong Green identity; its residue term is therefore zero.
Scope & sources - ConditionalZero periods give zero obstruction and the prescribed Green form
A basis makes zero periods equivalent to ν=0. The kernel comparison gives Ob=0; the supported pair and global actual primitive correct the very specified Z.
Scope & sources
How the deduction works
These finitely many forms detect the obstruction. The resulting Green identity still holds against EVERY compact actual test.
Conclusion / target
For t=dim Vmodel, fixed η₁,…,ηt satisfy ∫Zηj=ν([Z])(vj)∈ℚ. Under this full comparison, strong G for Z is equivalent to all these periods vanishing.
Scope & sourcesEvidence and further reading
Back to connection 02 in the diagram ↑All 7 nodes: scope, status and sources
ConditionalFixed D-closed actual periods detect strong G on the exact support
For t=dim Vmodel, fixed η₁,…,ηt satisfy ∫Zηj=ν([Z])(vj)∈ℚ. Under this full comparison, strong G for Z is equivalent to all these periods vanishing.
Back to this node in the diagram ↑Conditionalηj = TrKXL/X(bj) / ([L:K] cL)
Here cL=−log|πL|=cK/e(L/K). Use the extending absolute value, full field degree and literal D-closedness.
Back to this node in the diagram ↑HypothesisVT17 constructs fixed global D-closed βj on Y
Choose rational basis cycles once. The framed comparison gives ∫Wβj=cL〈rᵣ[W̄],vj〉 for every W, with the same cL and the complete resolution differential.
Back to this node in the diagram ↑HypothesisAD17 gives global D-closed bj on XL
Against every separately closed ω, ∫XLbjω=∫Yβjg*ω. Curved Green identities and Stokes remove the global D-boundary between g*ΩZL and ΩW. No proper-position support premise is inserted.
Back to this node in the diagram ↑HypothesisFull actual finite trace and native measure comparison
Trace commutes with D and pairs with every actual cycle test, including native flat lengths. This is a full actual trace input, not merely the restricted scalar–curvature trace.
Back to this node in the diagram ↑ConditionalStrong G forces every fixed period to vanish
Projectivity makes each global ηj compact as a test. Insert Dηj=0 into the strong Green identity; its residue term is therefore zero.
Back to this node in the diagram ↑ConditionalZero periods give zero obstruction and the prescribed Green form
A basis makes zero periods equivalent to ν=0. The kernel comparison gives Ob=0; the supported pair and global actual primitive correct the very specified Z.
Back to this node in the diagram ↑LOGIC 04
Extending a complete-intersection Green difference across canceled components
A supplied comparison certificate produces a relative Aeppli decomposition; a support-controlled cutoff turns it into an exact-support Green form.
Scope. Fix n≥r≥3, one allowed metric on the same line bundle L, and two regular rank-r section tuples s₀,s₁. Let Yi=V(si) with full scheme multiplicities, Z=[Y₁]−[Y₀], S=|Y₀|∪|Y₁|, U=Xᵃⁿ∖|Z|, W=Xᵃⁿ∖S and CU=S∩U.
Scroll inside the frame to follow a large branch. “Fit width” shows the whole layout; “Readable size” restores the text to 100%.
Read the 3 connections
Each numbered card gives the exact inputs, the deduction and its sources.
01AND · ALL kernel, certificate and cutoff ingredients
Premises and stated hypotheses
- HypothesisThe two complete-intersection kernels have the SAME curvature Rʳ
φs=minᵢ(−log∥si∥), As=R−Dφs, and Gs=φsΣAsʲRʳ⁻¹⁻ʲ. The all-test identity is D[Gs]+δYs=[Rʳ], with nonreduced multiplicities. Their difference is initially defined only on W.
Scope & sources - ConditionalA valid certificate gives a relative Aeppli decomposition near CU
The certificate must be supplied in the precise analytic-matrix or conormal-word form. Rational equivalence by itself is not this datum.
Scope & sources - HypothesisRelative-support cutoff, extension by zero and all-test Stokes
Retain support closed in U, ζ=1 near the canceled locus, and the differentiated partition terms. The source relative-support Stokes theorem proves the identity on every actual test.
Scope & sources
How the deduction works
On N∩W the certificate gives G₁−G₀=d′α+d″β+E. With ζ supported closed in U inside N and equal to 1 near CU, subtract d′ext₀(ζα)+d″ext₀(ζβ). It equals E near CU and is actual there.
Conclusion / target
Start with G₁−G₀ on W. Near CU, subtract derivatives of cutoff primitives and glue to the actual error E.
Scope & sourcesEvidence and further reading
Back to connection 01 in the diagram ↑02AND · BOTH the common Chow class and the existing height
Premises and stated hypotheses
- HypothesisBoth tuples are regular sections of the SAME L⊕r
The metric is also fixed for the analytic curvature cancellation.
Scope & sources - HypothesisThe complementary class is admitted for the global height
All global reciprocity hypotheses and the complementary local-kernel condition remain necessary.
Scope & sources
How the deduction works
Use equality in rational Chow and the kernel-domain height on classes, rather than claiming that every local representative integral vanishes.
Conclusion / target
Each regular zero scheme represents cr(L⊕r)∩[X], with its original multiplicities. Global Chow descent and bilinearity then give zero height against every admitted complementary class.
Scope & sourcesEvidence and further reading
Back to connection 02 in the diagram ↑03AND · BOTH the analytic support correction and the Chow-class calculation
Premises and stated hypotheses
- ConditionalA strong actual Green form on U, with no extra deleted support
Start with G₁−G₀ on W. Near CU, subtract derivatives of cutoff primitives and glue to the actual error E.
Scope & sources - Conditional[Y₁]=[Y₀]=c₁(L)ʳ, so [Z]=0 in Chow
Each regular zero scheme represents cr(L⊕r)∩[X], with its original multiplicities. Global Chow descent and bilinearity then give zero height against every admitted complementary class.
Scope & sources
How the deduction works
The analytic branch constructs a form for the specified difference cycle. The algebraic branch independently shows its global Chow class is zero.
Conclusion / target
Under one of the specified comparison certificates, the form extends across canceled components to U. For global data, hX([Z],w)=0 for every admitted complementary w; individual local integrals may be nonzero.
Scope & sourcesEvidence and further reading
Back to connection 03 in the diagram ↑All 8 nodes: scope, status and sources
ConditionalThe exact-support Green difference has global height zero
Under one of the specified comparison certificates, the form extends across canceled components to U. For global data, hX([Z],w)=0 for every admitted complementary w; individual local integrals may be nonzero.
Back to this node in the diagram ↑ConditionalA strong actual Green form on U, with no extra deleted support
Start with G₁−G₀ on W. Near CU, subtract derivatives of cutoff primitives and glue to the actual error E.
Back to this node in the diagram ↑HypothesisThe two complete-intersection kernels have the SAME curvature Rʳ
φs=minᵢ(−log∥si∥), As=R−Dφs, and Gs=φsΣAsʲRʳ⁻¹⁻ʲ. The all-test identity is D[Gs]+δYs=[Rʳ], with nonreduced multiplicities. Their difference is initially defined only on W.
Back to this node in the diagram ↑ConditionalA valid certificate gives a relative Aeppli decomposition near CU
The certificate must be supplied in the precise analytic-matrix or conormal-word form. Rational equivalence by itself is not this datum.
Back to this node in the diagram ↑HypothesisRelative-support cutoff, extension by zero and all-test Stokes
Retain support closed in U, ζ=1 near the canceled locus, and the differentiated partition terms. The source relative-support Stokes theorem proves the identity on every actual test.
Back to this node in the diagram ↑Conditional[Y₁]=[Y₀]=c₁(L)ʳ, so [Z]=0 in Chow
Each regular zero scheme represents cr(L⊕r)∩[X], with its original multiplicities. Global Chow descent and bilinearity then give zero height against every admitted complementary class.
Back to this node in the diagram ↑HypothesisBoth tuples are regular sections of the SAME L⊕r
The metric is also fixed for the analytic curvature cancellation.
Back to this node in the diagram ↑HypothesisThe complementary class is admitted for the global height
All global reciprocity hypotheses and the complementary local-kernel condition remain necessary.
Back to this node in the diagram ↑LOGIC 05
Which comparison data justify crossing the canceled locus
The ordinary analytic three-row matrix and the finite conormal word are alternative sufficient certificates; neither is inferred from rational equivalence.
Scope. Keep the complete-intersection setup and CU from the preceding tree, n≥r≥3. All coefficient lifts, row labels, inverses, cap offsets and partitions specified by the source are part of the certificate.
Scroll inside the frame to follow a large branch. “Fit width” shows the whole layout; “Readable size” restores the text to 100%.
Read the 3 connections
Each numbered card gives the exact inputs, the deduction and its sources.
01AND · ALL exact matrix and relative-face inputs
Premises and stated hypotheses
- HypothesisThe exact matrix identity and the two-pivot cover are supplied
Invertibility, the active three rows, identity on all other rows and the specified pivot cover are part of the hypothesis.
Scope & sources - HypothesisThe companion relative three-row face identities
They provide the complete relative primitives and include the differentiated partition and mixed-face contributions.
Scope & sources
How the deduction works
Apply the relative three-row face identities and glue with a two-pivot partition controlled by v(d)−v(a). Its derivative and mixed terms are retained in E.
Conclusion / target
On an ordinary analytic neighbourhood N of CU, s₁=M s₀ with an invertible three-row block and identity elsewhere. The specified first-column entries a,d have nonvanishing opens covering CU.
Scope & sourcesEvidence and further reading
Back to connection 01 in the diagram ↑02AND · ALL word, lift and germ-compatibility inputs
Premises and stated hypotheses
- HypothesisThe full word, coefficients, inverses and common partition data
At each fixed word position all coefficient lifts use the SAME cap offsets and inner/outer partitions; every three-row block retains its two-pivot cover.
Scope & sources - ConditionalThe final discrepancy is 1+E₀ with E₀ vanishing on the reduced zero scheme
Shrink until ∥E₀∥<1. The tuple sup norm and the entire complete-intersection expression are then unchanged; this does not replace the original nonreduced cycles by their reductions.
Scope & sources - HypothesisCoefficient-uniform primitive germs and locally finite gluing
The complete primitive germs, not only currents or pointwise values, agree on overlaps. The differentiated discrepancy extends by zero.
Scope & sources
How the deduction works
Lift the same ordered word locally and execute it on the actual intermediate tuples. The coefficient-uniform germ theorem makes the complete primitive germs agree across lifts; a locally finite partition glues them.
Conclusion / target
On a Zariski neighbourhood of CU the two ideals agree. The reduced conormal comparison is a finite ordered word in shears, one-entry units, invertible two-row blocks, and allowed three-row blocks.
Scope & sourcesEvidence and further reading
Back to connection 02 in the diagram ↑03OR · ANY one of the TWO supplied certificates
Alternative complete routes
- ConditionalAn exact analytic three-row comparison with a two-pivot cover
On an ordinary analytic neighbourhood N of CU, s₁=M s₀ with an invertible three-row block and identity elsewhere. The specified first-column entries a,d have nonvanishing opens covering CU.
Scope & sources - ConditionalA finite supplied word for the reduced conormal comparison
On a Zariski neighbourhood of CU the two ideals agree. The reduced conormal comparison is a finite ordered word in shears, one-entry units, invertible two-row blocks, and allowed three-row blocks.
Scope & sources
How the deduction works
Each branch retains all of its comparison data. The theorem does not assert that arbitrary pairs of regular tuples admit one of these certificates.
Conclusion / target
This is the local decomposition needed by the support-controlled cutoff construction. The α and β primitives are actual on the overlap with W; E is actual across CU.
Scope & sourcesEvidence and further reading
Back to connection 03 in the diagram ↑All 8 nodes: scope, status and sources
ConditionalG₁−G₀=d′α+d″β+E near the canceled locus, with E actual across it
This is the local decomposition needed by the support-controlled cutoff construction. The α and β primitives are actual on the overlap with W; E is actual across CU.
Back to this node in the diagram ↑ConditionalAn exact analytic three-row comparison with a two-pivot cover
On an ordinary analytic neighbourhood N of CU, s₁=M s₀ with an invertible three-row block and identity elsewhere. The specified first-column entries a,d have nonvanishing opens covering CU.
Back to this node in the diagram ↑HypothesisThe exact matrix identity and the two-pivot cover are supplied
Invertibility, the active three rows, identity on all other rows and the specified pivot cover are part of the hypothesis.
Back to this node in the diagram ↑HypothesisThe companion relative three-row face identities
They provide the complete relative primitives and include the differentiated partition and mixed-face contributions.
Back to this node in the diagram ↑ConditionalA finite supplied word for the reduced conormal comparison
On a Zariski neighbourhood of CU the two ideals agree. The reduced conormal comparison is a finite ordered word in shears, one-entry units, invertible two-row blocks, and allowed three-row blocks.
Back to this node in the diagram ↑HypothesisThe full word, coefficients, inverses and common partition data
At each fixed word position all coefficient lifts use the SAME cap offsets and inner/outer partitions; every three-row block retains its two-pivot cover.
Back to this node in the diagram ↑ConditionalThe final discrepancy is 1+E₀ with E₀ vanishing on the reduced zero scheme
Shrink until ∥E₀∥<1. The tuple sup norm and the entire complete-intersection expression are then unchanged; this does not replace the original nonreduced cycles by their reductions.
Back to this node in the diagram ↑HypothesisCoefficient-uniform primitive germs and locally finite gluing
The complete primitive germs, not only currents or pointwise values, agree on overlaps. The differentiated discrepancy extends by zero.
Back to this node in the diagram ↑Remaining problem
The open problem is to prove these original periods vanish for every homologically trivial class. A finite list of periods alone does not establish that vanishing.