Recovering ordinary slice barcodes from signed hooks

Let \(M\) be an originally finitely presented persistence module on \(\mathbb R^n\), over an arbitrary field, with \(n\ge1\). Choose a positive affine line \(L(t)=b+ta\), where every \(a_i>0\). Its entrance time \(\tau_L(p)=\max_i((p_i-b_i)/a_i)\) determines the restrictions of hooks: a finite hook becomes \([\tau_L(p),\tau_L(q))\), and a principal hook becomes \([\tau_L(p),\infty)\). Collapsed finite intervals are omitted and every remaining occurrence is retained. The slice presentation construction derives an original finite presentation of \(M\circ L\), so it has the existing ordinary finite barcode \(B_L(M)\).

Write \(E(M)\) and \(O(M)\) for the even and odd hook multisets in the canonical signed hook barcode of \(M\). Applying the endpoint operation to each multiset gives the exact identity

\[B_L(M)+\operatorname{slice}_L(O(M)) =\operatorname{slice}_L(E(M)).\]

These are equalities of multisets of real interval endpoints, with infinite deaths retained. In particular, the odd sliced multiset is contained in the even one, and removing its occurrences recovers the ordinary barcode. finitelyPresented_sliceSignedHookBarcode_cancel proves the additive identity from the original finite presentation alone. The accompanying domination and subtraction theorems make the removal precise.

Rank counts determine the endpoints

For a finite multiset \(D\) of positive intervals and \(s\le t\), define

\[\rho_D(s,t)=\#\{[u,v)\in D:u\le s,\ t<v\}.\]

The count includes multiplicities; \(t<\infty\) holds for every real \(t\). The rank of the actual structure map of the interval realization is this count. RealBarcodeRankUniqueness reuses the existing interval and realization rank formulas and proves that equality of all these counts determines a positive finite endpoint multiset.

The proof places every birth and every finite death from the two multisets in a common finite ordered grid. Each occurrence becomes a finite index bar, with essential deaths represented by the terminal index. Finite-barcode uniqueness then gives equality of the index bars, and mapping back gives equality of the original real endpoints. This preserves repeated endpoints and also covers the empty grid.

Applying a rank-exact resolution

In a finite rank-exact resolution, the rank of each projective term is the sum of the ranks of its two adjacent kernels, for every comparable pair of parameters. Evaluate these equations at \(L(s)\le L(t)\) and telescope them. The terminal zero kernel yields

\[\operatorname{rank} M(L(s)\le L(t)) +\sum_{j\ \mathrm{odd}}\operatorname{rank} Q_j(L(s)\le L(t)) =\sum_{j\ \mathrm{even}}\operatorname{rank} Q_j(L(s)\le L(t)).\]

Each term's actual finite hook-sum presentation converts its rank into the corresponding sliced endpoint count. Omitted hooks contribute zero. The ordinary barcode realization supplies the count for the first term. Endpoint-count uniqueness proves the displayed multiset identity.

RankExactBarcodeCancellation provides the finite-resolution argument; AffineSliceRank supplies its hook-sum and parity comparisons. AffineSliceSignedBarcode obtains the ambient resolution and the ordinary slice realization internally from the original finite presentation. The proof uses rank exactness, without assuming that restriction preserves minimality or projectivity of the resolution.

The cancellation mechanism corresponds to Botnan, Oppermann and Oudot, §5.3, Corollaries 5.8–5.9. The source's Corollary 5.8 is stated for rectangles; the hook comparison here uses the direct support calculation above. The identity recovers the ordinary slice barcode. Distance comparisons use that ordinary barcode; common-summand cancellation is not asserted to preserve bottleneck distance.