I · Descent of the polarized mixed tower

52 NAG Tags

Use the valuation-ring tower (T) and the action conventions of Chapter H.

Theorem. NAG-003010 (Descent of a compatible mixed equivariant tower). Assume (T), choose 0≠π∈mO0\ne\pi\in\mathfrak m_O, and put Bn=B/πnBB_n=B/\pi^nB. Suppose given a cartesian tower of proper, flat, finitely presented BnB_n-schemes XnX_n with compatible semilinear GG-actions. Assume that C=X1×B1lC=X_1\times_{B_1}l is at-worst-nodal of pure dimension one and that II acts identically on the entire scheme CC.

Choose any ample AC\mathcal A_C, compatible ample lifts An\mathcal A_n, and put Nn=⨂g∈Gρg,n∗An\mathcal N_n=\bigotimes_{g\in G}\rho_{g,n}^*\mathcal A_n. There exist a projective, flat, finitely presented nodal OO-scheme YY of relative dimension one and a relatively ample invertible sheaf P\mathcal P, with compatible equivariant polarized identifications (Y,P)×OBn≃(Xn,Nn)(n≥1).(15)(Y,\mathcal P)\times_O B_n \simeq (X_n,\mathcal N_n)\qquad(n\geq1). \tag{15} The chosen polarized tower determines this pair uniquely up to its unique compatible polarized isomorphism. Its residue polarization is characterized after extension to ll by P∣Yk⊗kl≃⨂h∈H(ρˉh∗AC)⊗e,(16)\mathcal P|_{Y_k}\otimes_k l\simeq \bigotimes_{h\in H}(\bar\rho_h^*\mathcal A_C)^{\otimes e}, \tag{16} with the HH-descent datum induced by the orbit linearization. An equivariant realization of the underlying tower over OO necessarily has the stated whole-special-scheme II-identity.

Proof. The exact result supplies AC\mathcal A_C and its compatible ample lifts, or lifts any prescribed ample AC\mathcal A_C. Apply Lemma G.0.8 at every level. Its naturality makes the tensor transition maps GG-equivariant, and the full II-action on Nn∣C\mathcal N_n|_C is the identity. Formula (12) gives the right side of (16); it is not replaced by AC⊗∣G∣\mathcal A_C^{\otimes |G|}.

Apply to the ordinary polarized tower. This constructs a projective, flat, finitely presented nodal pair (YB,NB)(Y_B,\mathcal N_B) over the complete valuation ring BB. Rank-one completeness gives B=lim←⁡BnB=\varprojlim B_n. For each g∈Gg\in G, the given tower automorphism and its linearization are compatible polarized isomorphisms into the base twist. The passing exact lemma algebraizes them uniquely. For every g,hg,h, their appropriate composite and the map for ghgh have the same reductions; faithfulness makes them equal. The identity relation is checked in the same way. Thus this is a full GG-action and linearization, including relations in a possibly nonsplit extension (8). Reduction at level one shows the required entire residual II-identities on the scheme and the bundle.

Theorem H.0.2 now applies to the actual algebraic pair. It descends the scheme and the existing NB\mathcal N_B to (Y,P)(Y,\mathcal P), without a second orbit tensor. Reducing the algebraic comparisons gives (15), all transition compatibilities, and (16). The corresponding O/πnOO/\pi^nO tower consists of the reductions of this actual proper flat finitely presented polarized scheme. Neither ramified residual descent nor a valuation-ring module theorem has been applied to the quotient rings BnB_n.

For uniqueness, two such pairs become algebraizations of the same polarized tower over BB. Polarized full faithfulness gives the unique comparison there; it is GG-equivariant because its equivariance equations hold at every level. Inertia functoriality and fully faithful unramified descent descend that comparison uniquely. Finally, if the underlying tower is already the equivariant base change of an OO-model, its residue scheme is obtained by extending kk to ll, on which II acts trivially. This proves the last assertion. ◻

Primitive cuts on a projective DVR model

Let KK be complete and nontrivially discretely valued, with valuation ring RR, residue field kk, uniformizer π\pi, and fixed absolute value. Set v=ord⁡R:K×⟶Z,λ=−log⁡∣π∣>0,−log⁡∣a∣=λv(a).v=\operatorname{ord}_R:K^\times\longrightarrow\mathbf Z, \qquad \lambda=-\log|\pi|>0, \qquad -\log|a|=\lambda v(a). We put v(0)=+∞v(0)=+\infty when taking contents. No algebraic closure or rescaling of the absolute value is used. Fix a free RR-module EE of rank d+1d+1, d≥2d\geq2. Our common model is P(E)\mathbf P(E), with projective points interpreted as lines, so its linear sections are elements of E∗E^*. This is a proper flat normal model with one reduced special component V=P(Ek)V=\mathbf P(E_k). At every induction stage we reuse EE for a free module of rank r+1r+1, 0≤r≤d0\leq r\leq d, and reuse the corresponding projective-model notation. The initial stage is r=dr=d.

For 0≤i≤r0\leq i\leq r, choose mi∈Z>0,ci∈Q,Li=O(mi),si=ℓimi,ℓi∈EK∗.m_i\in\mathbf Z_{>0},\qquad c_i\in\mathbf Q,\qquad L_i=\mathcal O(m_i),\qquad s_i=\ell_i^{m_i}, \qquad \ell_i\in E_K^*. The rows ℓi\ell_i form an invertible matrix AA in an RR-basis of EE. Thus every subfamily is independent over KK; the generic divisors meet properly and all r+1r+1 supports have empty total intersection. Use the metric ∥si([z])∥i=∣π∣ci∣ℓi(z)∣mi(max⁡j∣zj∣)mi.\begin{equation} \|s_i([z])\|_i =|\pi|^{c_i}\frac{|\ell_i(z)|^{m_i}}{ (\max_j|z_j|)^{m_i}}. \end{equation} The homogeneous expression is interpreted on analytic coordinate charts, or with coordinates in the completed residue field of an analytic point. It is independent of rescaling zz.

These are algebraic root model metrics. Choose ei≥1e_i\geq1 with eici∈Ze_ic_i\in\mathbf Z and use O(eimi)\mathcal O(e_im_i) on P(E)\mathbf P(E), whose standard model frame is identified generically with π−eici\pi^{-e_ic_i} times the standard frame. Taking the eie_ith root gives (I.0.1). The special-fibre line is O(eimi)\mathcal O(e_im_i) and is generated by global sections, hence has nonnegative degree on every vertical curve. Thus the metrics are semipositive. The rational shifts change the metric identification, not that special-fibre positivity. Changing the clearing denominator eie_i does not change the metric.

For a nonzero row ℓ∈EK∗\ell\in E_K^* write a=min⁡jv(ℓj),u=π−aℓ∈E∗.a=\min_j v(\ell_j),\qquad u=\pi^{-a}\ell\in E^*. The row uu is primitive: at least one coefficient is a unit. Its integer content aa is independent of the chosen RR-basis.

Lemma. NAG-002928 (Primitive cuts and determinant recursion). Assume r≥1r\geq1. There is an RR-basis e0,…,ere_0,\ldots,e_r of EE with u(e0)=1u(e_0)=1 and E′=ker⁡u=Re1⊕⋯⊕RerE'=\ker u=Re_1\oplus\cdots\oplus Re_r. The horizontal model cut is P(E′)≃PRr−1\mathbf P(E')\simeq\mathbf P_R^{r-1}. For r≥1r\geq1, a section s=ℓms=\ell^m with metric shift cc and clearing exponent ee as above has normalized model divisor 1ecyc⁡(div⁡O(em)(s⊗e))=m[P(E′)]+(ma+c)[V].\begin{equation} \frac1e\operatorname{cyc}\bigl(\operatorname{div}_{\mathcal O(em)}(s^{\otimes e})\bigr) =m[\mathbf P(E')]+(ma+c)[V]. \end{equation} The generic divisor is m[P(EK′)]m[\mathbf P(E'_K)]. A negative ma+cma+c is allowed as a rational vertical coefficient. All coefficients are computed by ordinary algebraic DVR orders, with actual quotient lengths.

If ℓ=ℓi\ell=\ell_i is a row of an invertible (r+1)(r+1)-row matrix AA, the other rows restrict to an invertible rr-row matrix AiA_i on EK′E'_K, and v(det⁡A)=ai+v(det⁡Ai).\begin{equation} v(\det A)=a_i+v(\det A_i). \end{equation} The restricted standard metric is the standard metric for the lattice E′E', with the same shift cjc_j and exponent mjm_j for every remaining line. No primitivity of these restricted rows is assumed.

9MIProofInputsElementary integral row reduction; ordinary DVR length convention of NAG-002552. The proof does not invoke the pending model log/divisor packet or a general intersection product.

Proof. Permute the coordinates to make a unit coefficient first, divide that coordinate vector by the unit, and subtract multiples of it from the other coordinate vectors. These are invertible RR-operations and give the displayed basis. The content is the fractional ideal generated by the row coefficients, so this ideal and aa are invariant under an invertible RR-basis change. The primitive equation is now z0=0z_0=0. Its model quotient is a polynomial projective hyperplane over RR, which proves the claim about P(E′)\mathbf P(E').

In a model frame, the coefficient of s⊗es^{\otimes e} has a factor πe(ma+c)uem\pi^{e(ma+c)}u^{em}. At the generic point of the horizontal hyperplane its order is emem, and at the generic point of VV its order is e(ma+c)e(ma+c). These are the only components of the divisor. At a DVR with parameter tt, the successive powers in A/(tn)A/(t^n) give length nn; for a rational coefficient one subtracts the lengths of numerator and denominator. Thus no nilpotent layer in uem=0u^{em}=0 is removed. Divide the coefficients by ee to obtain (I.0.2).

Move row ii to the first position and use the constructed basis. The matrix becomes (πai0∗Ai).\begin{pmatrix}\pi^{a_i}&0\\ *&A_i\end{pmatrix}. Row permutation and the basis matrix have determinant a unit up to sign, proving (I.0.3); its nonzero determinant also proves that AiA_i is invertible over KK. Finally, a matrix in GLr+1(R)\mathrm{GL}_{r+1}(R) and its inverse both have operator norm at most one for the maximum norm over every valued extension of KK. Consequently the maximum norm is unchanged. Restriction to the direct summand E′E' gives exactly the standard metric in its basis. ◻

DVR divisor coefficients and conditional model recursion

Let KK be complete and nontrivially discretely valued, with its specified absolute value, valuation ring RR, residue field kk, uniformizer π\pi, and λ=−log⁡∣π∣>0.\lambda=-\log|\pi|>0. We do not renormalize λ\lambda to one or pass to an algebraic closure. Let XX be an integral proper KK-variety of dimension d≥2d\geq2. Take a proper flat integral normal finite-type RR-model X\mathcal X with identified generic fibre XX. Retain explicitly the following local data: every height-one local ring used below is a DVR, and the special fibre has finitely many irreducible components VV, all of dimension dd. These are hypotheses, not assertions about arbitrary analytic local rings. Smoothness or strict semistability of the model is not required.

Write AV=OX,ηVA_V=\mathcal O_{\mathcal X,\eta_V} and normalize ord⁡V\operatorname{ord}_V by a uniformizer of this DVR. Then bV=ord⁡V(π)=length⁡AV(AV/πAV)>0,[Xk]=∑VbV[V].\begin{equation} b_V=\operatorname{ord}_V(\pi)=\operatorname{length}_{A_V}(A_V/\pi A_V)>0, \qquad [\mathcal X_k]=\sum_V b_V[V]. \end{equation} Thus a nonreduced special fibre is allowed. Its bVb_V is not discarded.

Choose line bundles L1,…,Ld,ML_1,\ldots,L_d,M on XX and model lines Li∣X≃Li⊗ei,M∣X≃M⊗e0,ei,e0∈Z>0.\mathcal L_i|_X\simeq L_i^{\otimes e_i},\qquad \mathcal M|_X\simeq M^{\otimes e_0},\qquad e_i,e_0\in\mathbf Z_{>0}.

Let ss be a nonzero rational section of MM. A model frame at ηV\eta_V writes s⊗e0=aVeVs^{\otimes e_0}=a_V e_V, and put cV=ord⁡V(aV)c_V=\operatorname{ord}_V(a_V). A change of model frame multiplies aVa_V by a DVR unit, so cVc_V is well-defined.

Proposition. NAG-002834 (Horizontal and vertical divisor coefficients). Assume the height-one DVR hypotheses given above also for the horizontal prime divisors. Then, as codimension-one cycles on X\mathcal X with rational coefficients, 1e0cyc⁡(div⁡X(s⊗e0))=cyc⁡(div⁡X(s))‾+∑VcVe0[V].\begin{equation} \frac1{e_0}\operatorname{cyc}\bigl(\operatorname{div}_{\mathcal X}(s^{\otimes e_0})\bigr) =\overline{\operatorname{cyc}(\operatorname{div}_X(s))} +\sum_V\frac{c_V}{e_0}[V]. \end{equation} Here the bar means coefficientwise closure of the horizontal prime cycles. At any height-one point with DVR AA, a local coefficient f=a/b∈Frac⁡(A)×f=a/b\in\operatorname{Frac}(A)^\times, a,b∈A∖{0}a,b\in A\setminus\{0\}, has order ord⁡A(f)=length⁡A(A/aA)−length⁡A(A/bA).\begin{equation} \operatorname{ord}_A(f)=\operatorname{length}_A(A/aA)-\operatorname{length}_A(A/bA). \end{equation} For an effective local equation ff, its coefficient is length⁡A(A/fA)\operatorname{length}_A(A/fA), using the actual quotient, including its nilpotents. This is algebraic Cartier-to-cycle compatibility; it does not identify an analytic current or an F8 tropical end cycle.

Proof. If tt is a uniformizer and a=tmua=t^m u with uu a unit, the chain of powers of tt in A/(tm)A/(t^m) has mm simple quotients, each its residue field. Thus length⁡A(A/aA)=m\operatorname{length}_A(A/aA)=m; a unit has quotient zero and length zero. Subtracting the two orders proves (I.0.6), independently of the chosen fraction. In particular replacing A/(tm)A/(t^m) by its reduction would incorrectly replace mm by one when m>0m>0.

A height-one prime of X\mathcal X is either a vertical component VV or horizontal. For a horizontal prime HH, the element π\pi is already a unit in OX,ηH\mathcal O_{\mathcal X,\eta_H}. Localizing the model to its generic fibre therefore leaves this DVR unchanged. If HKH_K is the corresponding prime divisor of XX, the local coefficient of s⊗e0s^{\otimes e_0} has order e0ord⁡HK(s)e_0\operatorname{ord}_{H_K}(s). At VV its order is cVc_V by definition. Comparing these coefficients proves (I.0.5).

There is no special-fibre factor bVb_V in a horizontal coefficient. For a finite residue extension, a simple residue-field factor has length its full field dimension after restriction of scalars; this is exactly the convention in NAG-002543 and NAG-002552. It counts inseparable degrees as well as separable degrees, without counting embeddings. ◻

Let ZZ have dimension r≥2r\geq2, and suppose a proposed class of normal proper models and sections supplies the following operations for the data in question:

  • HiZ\mathcal H_iZ is the normalized horizontal divisor cycle of sis_i, with actual multiplicities, and the required model data on its components;

  • VZ(i;J)\mathcal V_Z(i;J) is the normalized vertical term for cutting with sis_i and retaining the metrics indexed by JJ;

  • HC(J)H_C(J) is already defined, linear on the required cycles in lower dimensions, symmetric in its remaining entries, and satisfies the same horizontal–vertical recursion there.

All restrictions are required to be defined; no component is silently reduced or deleted. The operations and the vertical term are extended linearly to cycles. On a common normal DVR model the proposed vertical term, in integer model data, is VZ(i;J)=λei∏j∈Jej∑Vord⁡V(si⊗ei)deg⁡(Lj)j∈J,k(V).\begin{equation} \mathcal V_Z(i;J)=\frac{\lambda}{e_i\prod_{j\in J}e_j} \sum_V \operatorname{ord}_V(s_i^{\otimes e_i}) \deg_{(\mathcal L_j)_{j\in J},k}(V). \end{equation} This display specifies the algebraic term to be compared; it does not assume the pending packet’s general logarithmic identification.

Lemma. NAG-002940 (The interchange defect). Let i≠ji\ne j, let JJ index the other r−1r-1 entries, and put Ri=HHiZ(j,J)+VZ(i;j,J),Rj=HHjZ(i,J)+VZ(j;i,J).R_i=H_{\mathcal H_iZ}(j,J)+\mathcal V_Z(i;j,J),\qquad R_j=H_{\mathcal H_jZ}(i,J)+\mathcal V_Z(j;i,J). Under the preceding explicit operation and lower-dimensional recursion hypotheses, their difference is exactly Ri−Rj=HHjHiZ−HiHjZ(J)+VHiZ(j;J)−VHjZ(i;J)+VZ(i;j,J)−VZ(j;i,J).\begin{align} R_i-R_j={}&H_{\mathcal H_j\mathcal H_iZ-\mathcal H_i\mathcal H_jZ}(J)\\ &+\mathcal V_{\mathcal H_iZ}(j;J)-\mathcal V_{\mathcal H_jZ}(i;J)\\ &+\mathcal V_Z(i;j,J)-\mathcal V_Z(j;i,J). \end{align} Consequently equality of the horizontal codimension-two cycles and the vertical interchange identity VHiZ(j;J)+VZ(i;j,J)=VHjZ(i;J)+VZ(j;i,J)\begin{equation} \mathcal V_{\mathcal H_iZ}(j;J)+\mathcal V_Z(i;j,J) =\mathcal V_{\mathcal H_jZ}(i;J)+\mathcal V_Z(j;i,J) \end{equation} are sufficient for equality of the two first choices. Without that horizontal-cycle equality, vanishing of the full right side of (I.0.8) is the exact necessary and sufficient condition. If the appropriate identities hold at every node, and the curve/point base cases are compatible, the recursion is independent of its order.

Proof. Apply the lower-dimensional recursion to HHiZ(j,J)H_{\mathcal H_iZ}(j,J) and to HHjZ(i,J)H_{\mathcal H_jZ}(i,J). Their horizontal terms are respectively HHjHiZ(J)H_{\mathcal H_j\mathcal H_iZ}(J) and HHiHjZ(J)H_{\mathcal H_i\mathcal H_jZ}(J), and their vertical terms are the first two vertical expressions in (I.0.8). Subtract, use cycle linearity, and retain the two original vertical terms. This proves the exact formula and both stated criteria. Induct on dimension for every possible first choice, using the criteria to compare any two first choices and the induction hypothesis for later choices. The separately assumed curve and point cases start this induction. No equality of Chow classes is substituted for a required cycle or functional identity. ◻

Determinant norms and Cartier sections

Here q:Y→Xq:Y\to X may be any norm-admissible finite morphism between the fixed KK-spaces, with q∗OYq_*\mathcal O_Y locally free of constant rank n≥1n\geq1 on ordinary opens. In particular this applies to the finite constant-extension map above. We work with algebraic determinants of coherent modules, before introducing any metric on a norm bundle.

Define the determinant norm line Nm⁡q(L)=det⁡(q∗L)⊗det⁡(q∗OY)∨.\begin{equation} \operatorname{Nm}_q(\mathcal L)=\det(q_*\mathcal L)\otimes\det(q_*\mathcal O_Y)^\vee. \end{equation} Here det⁡\det is the nnth exterior power. Exterior powers and the displayed line are constructed on ordinary local frames and glued by their determinant transition maps. This also defines the norm of an isomorphism.

Proposition. NAG-002539 (Norm projection). For line bundles L\mathcal L on YY and M\mathcal M on XX, there is a canonical isomorphism Nm⁡q(L⊗q∗M)≃Nm⁡q(L)⊗M⊗n.\begin{equation} \operatorname{Nm}_q(\mathcal L\otimes q^*\mathcal M) \simeq \operatorname{Nm}_q(\mathcal L)\otimes\mathcal M^{\otimes n}. \end{equation} In particular Nm⁡q(q∗M)≃M⊗n\operatorname{Nm}_q(q^*\mathcal M)\simeq\mathcal M^{\otimes n}. The formulas commute with analytic base change. Consequently the kernel of Pic⁡(X)→Pic⁡(Y)\operatorname{Pic}(X)\to\operatorname{Pic}(Y) is killed by nn.

Proof. The tensor projection isomorphism gives q∗(L⊗q∗M)≃q∗L⊗Mq_*(\mathcal L\otimes q^*\mathcal M)\simeq q_*\mathcal L\otimes\mathcal M. For any rank-nn locally free module EE, local exterior multiplication gives det⁡(E⊗M)≃det⁡(E)⊗M⊗n.\det(E\otimes\mathcal M)\simeq \det(E)\otimes\mathcal M^{\otimes n}. Changing a frame of M\mathcal M by aa multiplies the top wedge by ana^n, and changing a frame of EE by a matrix multiplies both sides by the same determinant. These local maps therefore glue canonically. Cancel det⁡(q∗OY)\det(q_*\mathcal O_Y) to obtain (I.0.11). The norm line of OY\mathcal O_Y has its canonical trivialization, which gives the specialization.

Finite coherent direct image commutes with analytic base change, including nonflat base change. Exterior powers of locally free modules and the displayed frame maps do too. This proves the stated compatibility. If q∗Mq^*\mathcal M is trivial, apply its determinant norm and the specialization to get M⊗n≃OX\mathcal M^{\otimes n}\simeq\mathcal O_X. No injectivity of the unadorned Picard pullback is asserted. ◻

Atomic calibration and projective model induction

Use the same projective DVR and root-metric data.

Write ξE\xi_E for the divisorial point of VV. On a coordinate chart z0=1z_0=1 it is the unit-radius Gauss point. For r=0r=0 it is the unique point of PK0\mathbf P^0_K.

For these model data the finite model measure is specified by its atom at ξE\xi_E. Its weight is the mixed special-fibre degree of the root model lines divided by the product of their clearing exponents, with the multiplicity of the special component included. Here that multiplicity is one.

The comparison with the primary induction refers to the supplied recursion below when it has the same cycle, finite model measure and dimension-zero value. HZ(si,remaining data)=Hcyc⁡(div⁡Zsi)(remaining data)+∫Zan−log⁡∥si∥ dμremaining data.\begin{equation} H_Z(s_i,\text{remaining data}) =H_{\operatorname{cyc}(\operatorname{div}_Zs_i)}(\text{remaining data}) +\int_{Z^\mathrm{an}}-\log\|s_i\|\,d\mu_{\text{remaining data}}. \end{equation}

Proposition. NAG-002931 (Atomic calibration on the projective model). For any set JJ of rr remaining lines on P(E)\mathbf P(E), the specified mixed model measure and its mass are μJ=DJδξE,DJ=∏j∈Jmj=deg⁡(Lj)j∈J,K(P(EK)).\begin{equation} \mu_J=D_J\delta_{\xi_E},\qquad D_J=\prod_{j\in J}m_j =\deg_{(L_j)_{j\in J},K}(\mathbf P(E_K)). \end{equation} The empty product in dimension zero is one. For a further nonzero row ℓi\ell_i of content aia_i, ∫P(EK)an−log⁡∥si∥i dμJ=λ(miai+ci)DJ.\begin{equation} \int_{\mathbf P(E_K)^\mathrm{an}}-\log\|s_i\|_i\,d\mu_J =\lambda(m_i a_i+c_i)D_J. \end{equation} The logarithm is finite at the atom, and every proper algebraic closed subset has measure zero. For r≥1r\geq1 the normalized horizontal coefficient is mim_i. For r=0r=0 the generic divisor is empty and the horizontal part is zero. The normalized vertical coefficient is miai+cim_i a_i+c_i, and the special-fibre multiplicity is one. There is no additional factorial.

9MIFollowerInputsNAG-000182 and NAG-001707 for Gauss seminorms and analytification; Lemma I.0.2; the specified finite model-measure definition. The coordinate quotient calculation below calibrates the algebraic degrees and unreduced lengths directly.

Proof. For projective space, the dimension of homogeneous polynomials of degree nqnq is (nq+rr)\binom{nq+r}{r}, with leading term nrqr/r!n^rq^r/r!. Thus deg⁡O(n)Pr=nr\deg_{\mathcal O(n)}\mathbf P^r=n^r. Polarizing this polynomial shows that the mixed degree of O(n1),…,O(nr)\mathcal O(n_1),\ldots,\mathcal O(n_r) is ∏jnj\prod_jn_j: the coefficient of t1⋯trt_1\cdots t_r in (∑jnjtj)r(\sum_jn_jt_j)^r is r!∏jnjr!\prod_jn_j. The same normalization has a direct complete-intersection check. Choose the rr coordinate equations zjnj=0z_j^{n_j}=0, 1≤j≤r1\leq j\leq r. They meet at [1:0:⋯:0][1:0:\cdots:0], and their local algebra is k[t1,…,tr]/(t1n1,…,trnr).k[t_1,\ldots,t_r]/(t_1^{n_1},\ldots,t_r^{n_r}). Its monomials with exponents smaller than njn_j form a basis of size ∏jnj\prod_jn_j. These are regular successive coordinate powers, and the quotient retains all nilpotents. Its residue field is kk, so no residue-degree factor is suppressed. The same calculation holds over KK.

For the root model lines O(ejmj)\mathcal O(e_jm_j), this mixed special-fibre degree is ∏jejmj\prod_j e_jm_j. The BPS defining atomic weight divides by ∏jej\prod_je_j and multiplies by the special-fibre multiplicity, which is one. This proves (I.0.13) independently of the chosen clearing exponents. The constant shifts do not change these special-fibre lines.

The Gauss seminorm is multiplicative by NAG-000182. Its kernel on the affine coordinate ring is zero, so its scheme point is generic; this proves the zero-mass assertion. The Gauss value of the row ℓi\ell_i is the maximum absolute value of its coefficients, namely ∣π∣ai|\pi|^{a_i}. The denominator in (I.0.1) is one in this chart. Hence −log⁡∥si(ξE)∥=λ(miai+ci)-\log\|s_i(\xi_E)\|=\lambda(m_i a_i+c_i). Integrating a finite value against the one-atom measure proves (I.0.14). For r=0r=0 the row is a nonzero scalar and the same calculation applies. ◻

Define a model functional recursively on these tuples. In dimension zero, define it to be the negative logarithm of the nonzero section’s norm. In dimension r≥1r\geq1, choose any index ii, take the primitive horizontal model P(Ei)\mathbf P(E_i) of Lemma I.0.2, and set HE(i)((Lj,sj)j=0r)=miHEi((Lj∣Ei,sj∣Ei)j≠i)+λ(miai+ci)∏j≠imj.\begin{equation} H_E^{(i)}((L_j,s_j)_{j=0}^r) =m_i H_{E_i}((L_j|_{E_i},s_j|_{E_i})_{j\ne i}) +\lambda(m_i a_i+c_i)\prod_{j\ne i}m_j. \end{equation} Every subsequent stage again chooses an index. The first term uses the actual generic divisor cycle mi[P(Ei,K)]m_i[\mathbf P(E_{i,K})], and the second is the full model logarithmic integral, not a guessed correction.

Theorem. NAG-002935 (Symmetric projective model induction). The recursion is defined at every stage for every permutation of the indices. Every order gives HE=λ(D v(det⁡A)+∑i=0rciDmi),D=∏i=0rmi.\begin{equation} H_E=\lambda\left(D\,v(\det A) +\sum_{i=0}^r c_i\frac{D}{m_i}\right), \qquad D=\prod_{i=0}^r m_i. \end{equation} It is therefore symmetric in the labelled pairs (Li,si)(L_i,s_i), independent of the chosen integral bases and primitive complements, and independent of the clearing exponents eie_i. The value changes by λqD/mi\lambda q D/m_i if cic_i is increased by q∈Qq\in\mathbf Q. Replacing ℓi\ell_i by biℓib_i\ell_i, bi∈K×b_i\in K^\times, changes the value by −D∑ilog⁡∣bi∣-D\sum_i\log|b_i|.

This is a self-contained model recursion on the stated projective class. It identifies the restriction of the primary BPS induction to this class because its base value, divisor cycle and measure are the same. No native smooth-integral comparison, analytic adjointness, metric limit, model-change invariance or extension to arbitrary pseudo-divisors is part of the assertion.

9MIFollowerInputsLemmas and formulas I.0.2–I.0.6; induction and determinant arithmetic. The reference to the primary induction is a comparison of specified definitions, not an imported symmetry theorem.

Proof. The dimension-zero value is λ(m0v(ℓ0)+c0)\lambda(m_0v(\ell_0)+c_0), which is (I.0.16). Suppose the formula has been proved in dimension r−1r-1, for all choices of data and orders. The matrix of remaining rows on EiE_i is invertible, so the induction hypothesis applies after any first choice ii. Writing Di=D/miD_i=D/m_i, the first term of (I.0.15) is miλ(Div(det⁡Ai)+∑j≠icjDimj)=λ(Dv(det⁡Ai)+∑j≠icjDmj).m_i\lambda\left(D_i v(\det A_i) +\sum_{j\ne i}c_j\frac{D_i}{m_j}\right) =\lambda\left(Dv(\det A_i)+\sum_{j\ne i}c_j\frac D{m_j}\right). Its second term is λ(Dai+ciD/mi)\lambda(Da_i+c_iD/m_i). Equation (I.0.3) gives precisely (I.0.16). This proves the claim for every first index and every later order, without assuming order-independence in dimension rr.

Permuting rows changes the determinant by a sign of absolute value one; a change of RR-basis multiplies it by a unit. The norms and the child lattices have the basis compatibilities proved in Lemma I.0.2. The formula itself contains no eie_i. The two change rules follow by replacing cic_i and by multiplying the matrix rows by bib_i, respectively. Finally, the primary induction (I.0.12), if evaluated on this class, has exactly the same finite base value and exactly the two terms in (I.0.15). Finite induction identifies its value with the displayed formula; the new proof of symmetry did not use that primary theory. ◻

Proposition. NAG-002937 (Generic avoidance does not make model cuts proper). Fix d≥2d\geq2 and n≥1n\geq1 on PRd\mathbf P_R^d, with unshifted standard metrics and mi=1m_i=1. Take ℓ0=z0,ℓi=z0+πnzi(1≤i≤d).\begin{equation} \ell_0=z_0,\qquad \ell_i=z_0+\pi^n z_i \quad(1\leq i\leq d). \end{equation} Every model section avoids the generic point of the special component. Every generic-fibre subfamily meets properly, and the full generic intersection is empty. Nevertheless the simultaneous model zero scheme has special-fibre support {z0=0}≃Pkd−1\{z_0=0\}\simeq\mathbf P_k^{d-1}, so it is not a finite vertical intersection to which a final quotient-length formula could be applied. The symmetric model recursion is finite and has value dnλdn\lambda.

After the first horizontal cut z0=0z_0=0, each remaining section has vertical order nn. Its restriction to that vertical component is the zero section. Thus using the remaining restricted sections themselves to cut that vertical component fails; its contribution must use the mixed line-bundle degree, with the correct multiplicities.

9MIFollowerInputsTheorem I.0.7; explicit model equations. This is an obstruction to the proposed naive cut procedure, not to known refined intersection theory or to the primary local height.

Proof. Each row is primitive, so its order at the generic point of the special fibre is zero. Subtract row zero from all other rows. The determinant is πnd\pi^{nd}, hence all rows are independent over KK and their hyperplanes meet properly there. The simultaneous ideal is (z0,πnz1,…,πnzd).(z_0,\pi^n z_1,\ldots,\pi^n z_d). Modulo π\pi, its support is the stated projective hyperplane; its dimension is d−1>0d-1>0. For instance on z1=1z_1=1 it contains the quotient R[z2,…,zd]/(πn)R[z_2,\ldots,z_d]/(\pi^n) after setting z0=0z_0=0, which has positive-dimensional special support. Thus a finite local intersection length is not supplied by the whole tuple.

On the horizontal model z0=0z_0=0, the remaining rows are πnzi\pi^n z_i. Each therefore contributes the vertical component with order nn and vanishes identically when restricted to that component. Its line-bundle restriction is still O(1)\mathcal O(1), whose required mixed degree is one; the zero section is not a Cartier equation for that degree. Equation (I.0.16) gives H=λv(πnd)=dnλH=\lambda v(\pi^{nd})=dn\lambda. Equivalently, the first logarithmic term is zero, and the later dd stages, including the final point, each contribute nλn\lambda. ◻

Cartier sections under determinant norm

Use the same finite locally free rank and determinant norm conventions.

Corollary. NAG-002541 (Pulled-back Cartier sections). Let ss be a section of a line bundle M\mathcal M which is a non-zero-divisor in every local trivialization. Then q∗sq^*s is again a Cartier section. The determinant of multiplication by q∗sq^*s gives a section of Nm⁡q(q∗M)\operatorname{Nm}_q(q^*\mathcal M) which, under (I.0.11), is s⊗ns^{\otimes n}. Its effective Cartier divisor is n div⁡(s)n\,\operatorname{div}(s).

9DSFollowerInputsProposition I.0.5, NAG-001624; finite local freeness of qq

Proof. Choose an affinoid neighbourhood V=M(A)V=\mathcal M(A) on which M\mathcal M and q∗OYq_*\mathcal O_Y are trivial. The frame coefficient aa of ss is a non-zero-divisor in AA: its multiplication kernel sheaf has zero ordinary stalks, so the affinoid coherent equivalence makes its section kernel zero. The free AA-algebra module B=Γ(q−1V,OY)B=\Gamma(q^{-1}V,\mathcal O_Y) preserves this injection. The action of aa on the rank-nn module q∗OYq_*\mathcal O_Y is aIna I_n, with determinant ana^n. These equations agree on changes of frame, proving the assertion on sections. The local Cartier equation ana^n is precisely that of n div⁡(s)n\,\operatorname{div}(s). ◻

Use the data X,C,U,f,D,ηX,C,U,f,D,\eta and the actual chart cover ViV_i of the preceding chart lemma. There is no claim that η∣Vi\eta|_{V_i} has support in Int⁡(Vi/K)\operatorname{Int}(V_i/K). It can reach the original boundary of that chart. The finite union Q=⋃iViQ=\bigcup_iV_i is a compact ordinary neighbourhood of CC; it need not itself be good. The frozen finite-chart integrator covers exactly this situation. No locally finite partition, global paracompactness, or extension of a chart unit outside its domain is needed. On an overlap select one restricted representative; joint-tuple comparison is made only on that actual overlap.

Supplied fixed-test interface (IL). For a fixed ordinary compact form with one bidegree nn, regularity of ff implies vanishing on an ordinary neighbourhood of its zero space. The scalar logarithmic pairing on top forms is then the frozen ordinary compact integral away from that space. In particular this applies to d′′ηd''\eta and d′d′′ηd'd''\eta, and later to the cut test. This interface needs neither S1 nor S3.

Proposition. NAG-002955 (Fixed-test stabilization; F8SA-E). Assume IL. For every fixed compact ordinary top form β\beta supported in CC, there is RβR_\beta such that the continuous truncation ℓR=max⁡(log⁡∣f∣,−R)\ell_R=\max(\log|f|,-R), extended by −R-R at DD, satisfies ℓRβ=log⁡∣f∣ β(R≥Rβ),\ell_R\beta=\log|f|\,\beta\quad(R\geq R_\beta), where the right side is extended by zero across a neighbourhood on which β=0\beta=0. These sufficiently deep products are ordinary smooth compact forms, and their scalar integrals are exactly equal.

More specifically, fix a smooth χ\chi equal to one sufficiently far left and zero for x≥b0x\geq b_0. Set η1=χ(log⁡∣f∣)η\eta_1=\chi(\log|f|)\eta, extending χ(log⁡∣f∣)\chi(\log|f|) by one at DD. Given b>b0b>b_0, there exists finite a∗<b0a_*<b_0 such that for every a≤a∗a\leq a_*, Lf(d′d′′η1)=∫Q∩{ea≤∣f∣≤eb}log⁡∣f∣ d′d′′η1.\begin{equation} L_f(d'd''\eta_1)=\int_{Q\cap\{e^a\leq|f|\leq e^b\}} \log|f|\,d'd''\eta_1. \end{equation} Both d′′η1d''\eta_1 and d′d′′η1d'd''\eta_1 vanish on an ordinary neighbourhood of the removed lower cap.

Proof. IL [NAG-002806] makes the closed compact support of β\beta disjoint from DD. Hence ∣f∣|f| has a positive minimum on that support. Choose the cut strictly below this minimum. Outside the support the form is locally zero, so the asserted equality is equality of ordinary forms after zero extension, not just of integrals.

The cut scalar is smooth near DD, because it is locally constant there; away from DD, ff is a unit and its logarithm is an ordinary smooth moment function. Define its product with η\eta on UU, and extend by zero on X∖CX\setminus C. These definitions agree on their ordinary open overlap. All resulting supports remain in CC. Apply IL [NAG-002806] to the two indicated derivatives. Their compact supports avoid DD, so one positive lower bound for ∣f∣|f| works for both. Choose a∗a_* strictly below its logarithm and in the constant-one range of χ\chi. They vanish on a neighbourhood of Q∩{∣f∣≤ea∗}Q\cap\{|f|\leq e^{a_*}\}. They also vanish wherever log⁡∣f∣>b0\log|f|>b_0. Compact-neighbourhood independence of the frozen integral [NAG-002425] now gives (I.0.18), including its closed endpoints. ◻

Fix a complete non-Archimedean field KK, possibly trivially valued, with its specified absolute value. Affinoids are generalized affinoids with their canonical quotient and finite-module norms. An analytic space is good and Hausdorff, and purity always means analytic pure dimension. A regular function means a non-zero-divisor in the relevant affinoid rings, or equivalently the stated injectivity of its coherent sheaf multiplication when working locally. The zero ring and empty spaces contribute zero. We take n≥1n\geq1.

For an affinoid X=M(A)X=\mathcal M(A) pure of dimension nn, write Min⁡A={q1,…,qs},Xi=M(A/qi),ii:Xi↪X,mi=length⁡AqiAqi.\begin{equation} \operatorname{Min}A=\{\mathfrak q_1,\ldots,\mathfrak q_s\},\qquad X_i=\mathcal M(A/\mathfrak q_i),\quad i_i:X_i\hookrightarrow X,\quad m_i=\operatorname{length}_{A_{\mathfrak q_i}}A_{\mathfrak q_i}. \end{equation} All mim_i are positive integers. For regular f∈Af\in A, put fi=f∣Xif_i=f|_{X_i}, D=M(A/(f))D=\mathcal M(A/(f)), and Di=M(A/(qi,f))D_i=\mathcal M(A/(\mathfrak q_i,f)), retaining their actual quotient structures. The following interfaces are deliberately separate.

Proposition. NAG-002967 (Weighted component formula for native integrals; F8CR-N). For every ordinary compact (n,n)(n,n)-form ω\omega on the affinoid XX of (I.0.19), ∫Xω=∑i=1smi∫Xiii∗ω.\begin{equation} \int_X\omega=\sum_{i=1}^s m_i\int_{X_i}i_i^*\omega. \end{equation} The same formula holds with ω\omega replaced by ϕω\phi\omega for every ordinary smooth scalar ϕ\phi on XX. No boundary support assumption, zero-space purity, or scalar PL is used.

Proof. Here is the restriction check needed to keep the original mim_i when different moment representatives are used. Let V=M(B)V=\mathcal M(B) be an actual affinoid domain of XX. The map A→BA\to B is flat by ordinary single-domain flatness. If r∈Min⁡B\mathfrak r\in\operatorname{Min}B, then q=r∩A\mathfrak q=\mathfrak r\cap A is minimal in AA: flat going down (the already registered algebraic input Stacks 00HS) would otherwise give a prime properly below r\mathfrak r. The local flat map Aq→BrA_{\mathfrak q}\to B_{\mathfrak r} has Artinian source and target. Tensor a composition series of AqA_{\mathfrak q} with BrB_{\mathfrak r}. Exactness gives length⁡BrBr=mq length⁡Br(Br/qBr).\begin{equation} \operatorname{length}_{B_{\mathfrak r}}B_{\mathfrak r} =m_{\mathfrak q}\, \operatorname{length}_{B_{\mathfrak r}}(B_{\mathfrak r}/\mathfrak qB_{\mathfrak r}). \end{equation} For every other qi∈Min⁡A\mathfrak q_i\in\operatorname{Min}A, choose an element of qi∖q\mathfrak q_i\setminus\mathfrak q. It is invertible in BrB_{\mathfrak r}, so (B/qiB)r=0(B/\mathfrak q_iB)_{\mathfrak r}=0. Consequently (I.0.21) is exactly length⁡BrBr=∑imilength⁡Br(B/qiB)r.\operatorname{length}_{B_{\mathfrak r}}B_{\mathfrak r} =\sum_i m_i\operatorname{length}_{B_{\mathfrak r}} (B/\mathfrak q_iB)_{\mathfrak r}. In particular no assertion that a flat fibre has length one has entered this calculation.

Apply the generic finite-module weight formula NAG-002362 to these modules on VV, and then its global chain construction closed finite-module chain. For any chosen unit tuple gg on VV, it gives Tg(B)=∑imiTg(B/qiB).\mathcal T_g(B)=\sum_i m_i\mathcal T_g(B/\mathfrak q_iB). Actual domain base change NAG-001846 and finite tensor comparison NAG-001565 identify M(B/qiB)\mathcal M(B/\mathfrak q_iB) with V×XXiV\times_X X_i, retaining its full structure. This is an actual domain of XiX_i, hence pure of dimension nn when nonempty by NAG-002293. The compact finite-pushforward formula, on this single actual affinoid domain and for this closed immersion and the identity map on the common moment coordinates, identifies its integral with the target finite-module integral. There is no lattice or residue factor in addition to the recorded chain weights.

Represent ω\omega on a finite actual affinoid cover of a compact neighbourhood of its support. Use the same cover pulled back to each XiX_i, discarding empty pieces. All actual intersections have the same calculation. The finite inclusion–exclusion integral NAG-002416, followed by NAG-002425, gives (I.0.20). No unit tuple is extended beyond its chart; joint tuples and their projection pullbacks are used only on actual overlaps. The argument applies verbatim to the compact form ϕω\phi\omega. ◻

Fix a complete non-Archimedean field KK, possibly trivially valued, with its specified absolute value. Affinoids are generalized affinoids with their canonical quotient and finite-module norms. An analytic space is good and Hausdorff, and purity always means analytic pure dimension. A regular function means a non-zero-divisor in the relevant affinoid rings, or equivalently the stated injectivity of its coherent sheaf multiplication when working locally. The zero ring and empty spaces contribute zero. We take n≥1n\geq1.

For an affinoid X=M(A)X=\mathcal M(A) pure of dimension nn, write Min⁡A={q1,…,qs},Xi=M(A/qi),ii:Xi↪X,mi=length⁡AqiAqi.\begin{equation} \operatorname{Min}A=\{\mathfrak q_1,\ldots,\mathfrak q_s\},\qquad X_i=\mathcal M(A/\mathfrak q_i),\quad i_i:X_i\hookrightarrow X,\quad m_i=\operatorname{length}_{A_{\mathfrak q_i}}A_{\mathfrak q_i}. \end{equation} All mim_i are positive integers. For regular f∈Af\in A, put fi=f∣Xif_i=f|_{X_i}, D=M(A/(f))D=\mathcal M(A/(f)), and Di=M(A/(qi,f))D_i=\mathcal M(A/(\mathfrak q_i,f)), retaining their actual quotient structures. The following interfaces are deliberately separate.

Supplied hypothesis ZPloc_{\rm loc}. In each affinoid localization used here, DD and every DiD_i are empty or pure of analytic dimension n−1n-1.

Proposition. NAG-002979 (Cartier cycles with full generic weights; F8CR-C). For X,fX,f as above, assume ZPloc_{\rm loc}. Then, with the actual analytic Cartier coefficients and native pure-cycle integrals, div⁡X(f)=∑imi(ii)∗div⁡Xi(fi),∫Dη∣D=∑imi∫Di(ii∗η)∣Di\begin{equation} \operatorname{div}_X(f)=\sum_i m_i(i_i)_*\operatorname{div}_{X_i}(f_i),\qquad \int_D\eta|_D=\sum_i m_i\int_{D_i}(i_i^*\eta)|_{D_i} \end{equation} for every compact ordinary (n−1,n−1)(n-1,n-1)-form η\eta. The effective local length formula itself does not require ZP; its use here as a common pure-cycle identity does.

Proof. First check that no extra component is hidden in the right side. If r\mathfrak r is minimal over (qi,f)(\mathfrak q_i,f), choose p⊆r\mathfrak p\subseteq\mathfrak r minimal over (f)(f) [Stacks 00E0(4)]. ZPloc_{\rm loc} makes both integral affinoids M(A/p)\mathcal M(A/\mathfrak p) and M(A/r)\mathcal M(A/\mathfrak r) have analytic dimension n−1n-1. A strict inclusion would contain a nonzero element of the domain A/pA/\mathfrak p; the dimension-drop bound NAG-002374 would give dimension at most n−2n-2 for the latter zero locus (empty when n−1=0n-1=0). This is impossible. Thus r=p\mathfrak r=\mathfrak p. Conversely, if p∈Min⁡A(f)\mathfrak p\in\operatorname{Min}_A(f) contains qi\mathfrak q_i, no prime strictly between them can contain ff, so p/qi\mathfrak p/\mathfrak q_i is minimal over (fi)(f_i). Every p\mathfrak p contains at least one minimal prime of AA. This establishes the exact component correspondence, allowing the same Cartier component to occur for several ii.

At each such p\mathfrak p, Proposition C.1.14 gives the coefficient identity with every local Artin length. The frozen effective Cartier comparison NAG-002772 identifies each of these algebraic lengths with the length of the actual analytic quotient stalk at a maximal-Abhyankar point of that zero-space component. More explicitly, NAG-002769 applies to the pure quotient affinoids A/(f)A/(f) and (A/qi)/(fi)(A/\mathfrak q_i)/(f_i) and their finite structure modules; its ordinary and pointed-domain Artinian length comparisons retain the full closed fibre. Its regular local comparison, not flatness alone, is what justifies that bridge. No S1S_1 property of AA is required by either of these effective coefficient assertions. On the common reduced component the closed immersion has identical quotient A/pA/\mathfrak p and generic residue field, hence generic degree one.

Finally NAG-002778 identifies the full quotient integration with these analytic Cartier coefficients and their intrinsic component chains. Its pure-zero-space hypothesis is exactly the one retained here. Finite chart localization and its ordinary compact-form clause give the second equality in (I.0.23). No additional tropical multiplicity, lattice index, or reduced-fibre replacement is allowed. ◻

The tame inertia criterion for the supplied model

PDF-derived transcription. Historical English reader, Section 1.55, Theorem 1.55.5, printed pages 221–222. The historical original TeX remains unlocated.

Theorem. NAG-003147 (The full tame inertia criterion for a supplied model). Let L/KL/K be finite Galois satisfying (Tg), put O=K∘O=K^\circ, B=L∘B=L^\circ, and let G,I,lG,I,l be as in Theorem 1.55.3. Let YB/BY_B/B be a supplied projective, flat, finitely presented nodal scheme of relative dimension one with a semilinear GG-action. There is a projective, flat, finitely presented nodal OO-scheme YY of relative dimension one and a GG-equivariant comparison Y×OB≃YB\begin{equation} \tag{22} Y\times_O B\simeq Y_B \end{equation} if and only if II acts identically on the entire scheme (YB)l(Y_B)_l. Such a descent is unique up to the unique isomorphism compatible with (22).

A supplied GG-linearized relatively ample invertible sheaf LB\mathcal L_B descends, with its specified linearization and comparison, if and only if its entire special-fibre II-action is the identity. Its descent is relatively ample and unique with that comparison. There is no identity requirement on the residue Galois quotient G/IG/I and no further tensor power of LB\mathcal L_B is needed.

Proof.

Assume the whole-special-scheme identity. Take (19) and descend from Fj−1∘F_{j-1}^\circ to Fj∘F_j^\circ successively using Lemma 1.55.4, with P=G/Jj−1P=G/J_{j-1} and Q=Jj/Jj−1Q=J_j/J_{j-1}. Normality in GG makes these groups and their actions on the corresponding fields well defined. At each step projectivity, flatness, finite presentation and nodal relative dimension one are retained. Every coefficient residue field is ll. Reducing the actual equivariant comparison therefore identifies the new special scheme, with its remaining inertia action, with the previous one. The remaining inertia is still the identity. This argument uses the constructed comparison; it does not assume that invariants commute with reduction.

If r≥1r\ge1, it suffices to perform the first r−1r-1 steps. The last field configuration K⊂M⊂Fr−1K\subset M\subset F_{r-1} then satisfies exactly (T) of NAG-003009: M/KM/K is its actual unramified Galois lift, Fr−1/KF_{r-1}/K is Galois, and Fr−1/MF_{r-1}/M is one exact Kummer step. That frozen theorem performs the final Kummer and unramified descents. Composing the comparisons proves (22).

For completeness, when I=1I=1 the field L=ML=M itself is the actual unramified lift. The torsor identity in NAG-002262 turns the full GG-action into an fpqc descent datum. It is effective for this projective scheme by Stacks 0CCJ. Properness and finite presentation descend by Stacks 02L1 and Stacks 02L0, and flatness and nodal relative dimension one by Stacks 0C5E. The finite locally free surjective pullback has an ample invertible sheaf; Stacks 0BD4 gives one downstairs. Properness over the affine base then gives projectivity by Stacks 0B45.

If LB\mathcal L_B has full residual II-identity, apply the bundle part of Lemma 1.55.4 at these same steps. The new residual module is identified with the old one by the actual bundle comparison, so the required identity persists. For r≥1r\ge1 the bundle clause of NAG-003009 finishes the descent of this very bundle. In the unramified case I=1I=1, fpqc sheaf descent Stacks 023T is effective; finite local freeness descends by Stacks 05B2, and its rank is one by surjective base change. Relative ampleness descends by Stacks 0D2P.

Conversely, (22) identifies the special scheme with Yk×klY_k\times_k l. Inertia fixes ll and acts only on coefficients, so acts identically on that whole scheme. The same calculation on a pulled-back line bundle gives identity on its full residue module. Uniqueness follows at each ramified step from the canonical invariant descent and its functoriality. At the unramified step it is full faithfulness in Stacks 02W0 and Stacks 023T. This also proves independence of the chosen series and Kummer generators. The quotient G/IG/I retains its semilinear action on the complete special scheme, which is the descent datum used over l/kl/k. ◻

Descent of the supplied equivariant tower

PDF-derived transcription. Historical English reader, Section 1.55, Corollary 1.55.6, printed page 222. The historical original TeX remains unlocated.

Corollary. NAG-003148 (The supplied equivariant tower interface). Keep the field hypotheses and notation of Theorem 1.55.5, and choose 0≠π∈mK0\ne\pi\in\mathfrak m_K. Put Bn=B/πnBB_n=B/\pi^nB. Let Xn/BnX_n/B_n be a supplied cartesian tower of proper, flat, finitely presented schemes, with compatible semilinear GG-actions. Suppose C=X1×B1lC=X_1\times_{B_1}l is nodal of pure dimension one and II acts identically on the entire scheme CC. Choose an ample AC\mathcal A_C and compatible ample lifts An\mathcal A_n, and put Nn=⨂g∈Gρg,n∗An\mathcal N_n=\bigotimes_{g\in G}\rho^*_{g,n}\mathcal A_n. Then there is a projective, flat, finitely presented nodal pair (Y,P)(Y,\mathcal P) over OO, with P\mathcal P relatively ample, and compatible equivariant polarized identifications (Y,P)×OBn≃(Xn,Nn).\begin{equation} \tag{23} (Y,\mathcal P)\times_O B_n\simeq(X_n,\mathcal N_n). \end{equation} It is unique with these comparisons. The residue polarization, after extension to ll, is P∣Yk⊗kl≃⨂h∈G/I(ρˉh∗AC)⊗∣I∣,\begin{equation} \tag{24} \left.\mathcal P\right|_{Y_k}\otimes_k l \simeq\bigotimes_{h\in G/I}(\bar\rho_h^*\mathcal A_C)^{\otimes|I|}, \end{equation} with the quotient descent datum. An equivariant realization over OO necessarily has the stated entire-special-scheme inertia identity.

Proof. The exact input NAG-002910 supplies the indicated ample sheaf and any requested compatible lifts. The general finite-group lemma NAG-003007 makes Nn\mathcal N_n compatible and GG-linearized, with identity II-action on its entire restriction to CC; its coset formula is (24). This lemma does not assume cyclic inertia or invertibility of ∣G/I∣|G/I|.

Apply NAG-002911 to this ordinary polarized tower over the complete valuation ring BB. It gives a projective flat finitely presented nodal pair over BB. For each gg, the maps of the polarized tower into its base twist algebraize uniquely by NAG-002985. The base twists preserve πnB\pi^nB because gg fixes KK. Full faithfulness makes all group and linearization equations hold: both sides have the same reduction at every level. Thus the actual algebraic pair has its full GG-action and full residual II-identities. Theorem 1.55.5 descends that pair and that existing polarization without another orbit tensor, giving (23) and (24).

Two such realizations become algebraizations of the same polarized tower over BB. Polarized full faithfulness gives the unique comparison there and ensures its GG-equivariance. The uniqueness and functoriality in Theorem 1.55.5 descend it. Necessity follows by reducing an equivariant OO-realization to ll, where inertia has trivial coefficient action. The ramified affine theorem has been used only over valuation rings after algebraization, never over the nonreduced rings BnB_n. ◻

Meromorphic pairings and line-bundle frames

Let KK be complete and non-Archimedean with its specified absolute value; trivial valuation and generalized affinoid radii are allowed. Let XX be good, Hausdorff, and pure of analytic dimension n≥1n\geq1. Fix an ordinary open U⊆Int⁡(X/K)U\subseteq\operatorname{Int}(X/K) on which all ordinary local rings are S1S_1. All identities below are on UU. This also handles a single compact test on XX: choose such a neighbourhood of its support and restrict the data there. No S1S_1 claim outside that neighbourhood is needed. Ordinary open subspaces retain purity by NAG-002413; NAG-001630 and NAG-001704 show that every point of UU, and of its ordinary open subspaces, is interior over KK. Compact supports in these subspaces are therefore genuine analytic-interior supports.

Use the ordinary moment sheaf NAG-002424, the native integral NAG-002425, and the interleaved positive generator d′t1∧d′′t1∧⋯∧d′tn∧d′′tnd't_1\wedge d''t_1\wedge\cdots\wedge d't_n\wedge d''t_n of NAG-002388. There is no additional factor of 22, 2π2\pi, or a change from +log⁡∣f∣+\log|f|. We use the following exact inputs.

R-SPL (supplied hypothesis)

For a regular analytic aa on an ordinary neighbourhood of the compact support of an ordinary (n−1,n−1)(n-1,n-1)-test η\eta, La(d′d′′η)=∫V(a)η,La(α)=∫U∖V(a)log⁡∣a∣ α.\begin{equation} L_a(d'd''\eta)=\int_{V(a)}\eta, \qquad L_a(\alpha)=\int_{U\setminus V(a)}\log|a|\,\alpha. \end{equation} Regular means injective multiplication on the analytic GG-structure sheaf. The full actual quotient O/(a)\mathcal O/(a) occurs on the right. The review retains the good Hausdorff pure S1S_1 category and ordinary compact interior tests.

R-ZP (supplied hypothesis)

Actual zero spaces of such regular functions on pure affinoid refinements are empty or pure of dimension n−1n-1, with the full quotient structure. This also applies to products and to every actual common refinement. Its proof does not require S1S_1, but no non-S1S_1 scalar PL is inferred here.

Frozen inputs.

NAG-002772/G.0.17 compare and glue Cartier coefficients and native cycle integrals in their pure-zero-space scope. NAG-002806 gives actual neighbourhood vanishing of an ordinary form with one bidegree nn near the zeros of a regular analytic function. NAG-002427/C.1.60 give ordinary smooth cutoffs and finite compact partitions.

A regular meromorphic function is a section locally represented on ordinary opens by a/ba/b, with both a,ba,b regular in the above GG-sheaf sense. This is the unit part of the meromorphic sheaf in the good-space scope of frozen NAG-002772: its proof supplies ordinary quotient representatives and shrinking of coherent kernels. We do not replace a regular denominator by an arbitrary nonzero element. A line bundle is a coherent OUG\mathcal O_{U_G}-module free of rank one on ordinary open neighbourhoods, matching the established project category. A regular meromorphic section is a framewise regular meromorphic unit; its existence is data, not asserted for every line bundle.

Proposition. NAG-002989 (Quotient independence and full Cartier gluing; F8MG-Q). For a regular meromorphic ff on UU, the local native pairings La/b=La−LbL_{a/b}=L_a-L_b define a unique linear pairing Lf:AU,cn,n→RL_f:\mathcal A^{n,n}_{U,c}\to\mathbb R. It is additive under multiplication of regular meromorphic functions. The local full Cartier cycles div⁡(a)−div⁡(b)\operatorname{div}(a)-\operatorname{div}(b) glue to div⁡(f)\operatorname{div}(f), and their native pairings on compact ordinary (n−1,n−1)(n-1,n-1)-tests glue independently of presentations. The cycle assertion uses R-ZP and the exact frozen restricted Cartier comparison; the logarithmic pairing does not need R-SPL or a new measure representation.

For local regular quotient representatives, use the ordinary quotient charts. The logarithmic pairing argument uses actual compact top tests and supplied regular quotient factors. In that argument the compact integral and finite ordinary localization apply without an S1S_1 hypothesis. The Cartier comparison below retains the stated ambient and R-ZP hypotheses.

Proof. For a compact top test in one quotient neighbourhood, frozen NAG-002806 makes its closed support disjoint from the zeros of both aa and bb. On that support both logarithms are bounded, ordinary smooth functions, and their products with the test have well-defined native compact integrals, extended by zero where the test vanishes. If a/b=c/da/b=c/d, refine to actual common affinoid neighbourhoods. Localization at regular elements is injective, so ad=bcad=bc there. The test vanishes near the zeros of all four factors, and on the remaining support log⁡∣a∣−log⁡∣b∣=log⁡∣c∣−log⁡∣d∣\log|a|-\log|b|=\log|c|-\log|d|. Thus the pairings agree. This also proves logarithmic additivity for a product, after adjoining its finitely many regular factors.

For a general compact top test α\alpha, take a finite ordinary partition ρj\rho_j subordinate to quotient neighbourhoods, summing to one near Supp⁡α\operatorname{Supp}\alpha, and set Lf(α)=∑j(Laj−Lbj)(ρjα)L_f(\alpha)=\sum_j(L_{a_j}-L_{b_j})(\rho_j\alpha). Two choices are compared by the finite product partition ρjρk′\rho_j\rho'_k. Its supports lie on actual common neighbourhoods, where the preceding equality applies. Their sums equal the original test as ordinary sections. This proves independence and linearity; no countable partition or limiting interchange is used.

For cycles, R-ZP makes every numerator, denominator, and regular cross-product zero space empty or pure of one common dimension n−1n-1 on each common affinoid. The antichain hypothesis of NAG-002772 is then satisfied: a strict containment between two component primes of that analytic dimension would contradict NAG-002374. Its scoped product additivity applied to ad=bcad=bc gives equality of the divisor differences, including all generic lengths. Domain compatibility and ordinary gluing in that same theorem give the cycle; NAG-002778 supplies its native pairings with the full coherent coefficients. It does not replace these by reduced supports or count geometric points. The compact partition argument proves the asserted pairing independence. ◻

Theorem. NAG-002990 (Meromorphic scalar PL from the reviewed analytic input; F8MG-P). Under R-SPL and R-ZP in the category above, every regular meromorphic ff satisfies, for every ordinary η∈AU,cn−1,n−1\eta\in\mathcal A^{n-1,n-1}_{U,c}, Lf(d′d′′η)=δdiv⁡(f)(η),δD(η):=∫Dη.\begin{equation} L_f(d'd''\eta)=\delta_{\operatorname{div}(f)}(\eta), \qquad \delta_D(\eta):=\int_D\eta. \end{equation} This is first an equality of fixed-test pairings, before a continuity claim is made.

Apply the supplied scalar identity (I.0.24) to each regular factor on its actual compact quotient chart. The resulting differences refer to the two comparisons in Proposition I.0.15.

Proof. Use quotient neighbourhoods and a finite compact smooth partition ρj\rho_j near Supp⁡η\operatorname{Supp}\eta. Put ηj=ρjη\eta_j=\rho_j\eta. Each ηj\eta_j has compact support in its ordinary quotient neighbourhood and in the true analytic interior. R-SPL applies separately to aja_j and bjb_j on that neighbourhood. Subtract: Lf(d′d′′ηj)=∫V(aj)ηj−∫V(bj)ηj=δdiv⁡(f)(ηj).L_f(d'd''\eta_j) =\int_{V(a_j)}\eta_j-\int_{V(b_j)}\eta_j =\delta_{\operatorname{div}(f)}(\eta_j). Here Supp⁡d′d′′ηj⊆Supp⁡ηj\operatorname{Supp}d'd''\eta_j\subseteq\operatorname{Supp}\eta_j, so the local pairing agrees with the global pairing of Proposition I.0.15. Since η=∑jηj\eta=\sum_j\eta_j as ordinary sections, differentiation gives d′d′′η=∑jd′d′′ηjd'd''\eta=\sum_jd'd''\eta_j. In particular the terms d′d′′(ρjη)=ρjd′d′′η+d′ρj∧d′′η−d′′ρj∧d′η+d′d′′ρj∧ηd'd''(\rho_j\eta)=\rho_jd'd''\eta+d'\rho_j\wedge d''\eta -d''\rho_j\wedge d'\eta+d'd''\rho_j\wedge\eta are all retained. Their sums, rather than individual terms, use ∑ρj=1\sum\rho_j=1 near the support. Adding the finite identities proves (I.0.25). No component reduction, new annular Green assertion, or limit in a current topology enters this proof. ◻

Proposition. NAG-002992 (Frame gluing and its signs; F8MG-B). Let L\mathcal L have ordinary local frames eie_i on UiU_i, with ej=uijeie_j=u_{ij}e_i on overlaps, uiju_{ij} analytic units. A smooth metric means a norm on each one-dimensional H(x)\mathcal H(x)-fibre whose weights hi=−log⁡∥ei∥h_i=-\log\|e_i\| are ordinary smooth scalar functions. The compatibility equation is hj=hi−log⁡∣uij∣.\begin{equation} h_j=h_i-\log|u_{ij}|. \end{equation} For a regular meromorphic section s=fieis=f_i e_i, one has fj=uij−1fif_j=u_{ij}^{-1}f_i. Its full divisor DsD_s glues from div⁡(fi)\operatorname{div}(f_i), its ordinary smooth curvature form glues as c1(L‾)∣Ui=d′d′′hi,c_1(\overline{\mathcal L})|_{U_i}=d'd''h_i, and the local top-test pairings Gs,i(α)=∫Uihiα−Lfi(α)\begin{equation} G_{s,i}(\alpha)=\int_{U_i}h_i\alpha-L_{f_i}(\alpha) \end{equation} glue independently of frames and compact partitions. Off the local zero and pole loci this is the potential gs=−log⁡∥s∥=hi−log⁡∣fi∣g_s=-\log\|s\|=h_i-\log|f_i|.

Unit frame changes use the ordinary joint unit coordinates. Apply the ordinary zero-neighborhood vanishing to the actual top-test support on each regular quotient chart.

Proof. The metric rule ∥aei(x)∥=∣a∣exp⁡(−hi(x))\|a e_i(x)\|=|a|\exp(-h_i(x)) is independent of the frame precisely when (I.0.26) holds. On any smaller unit moment neighbourhood, log⁡∣uij∣\log|u_{ij}| is the negative of one moment coordinate. Its mixed second differential is zero as an ordinary form. Thus d′d′′hj=d′d′′hid'd''h_j=d'd''h_i, and the sheaf property NAG-002424 glues the curvature; it is killed by both d′d' and d′′d''. No PL theorem is needed to prove this unit identity.

The section transformation follows by comparing fiei=fjejf_i e_i=f_j e_j. The divisor of a unit is zero, so Proposition I.0.15 and frozen Cartier additivity give the same full divisor on overlaps. For local top tests, logarithmic additivity gives Lfj=Lfi−LuijL_{f_j}=L_{f_i}-L_{u_{ij}}, and Luij(α)=∫log⁡∣uij∣αL_{u_{ij}}(\alpha)=\int\log|u_{ij}|\alpha, since the unit has no zero. Equation (I.0.26) therefore proves equality of (I.0.27) on every admissible local top test. In a common regular-quotient chart the actual support avoids the regular numerator and denominator zero loci, by the ordinary top-form vanishing used in the quotient pairing. The larger compact set used to choose the charts may meet those loci. Finite compact partitions and their product refinements glue it just as in Proposition I.0.15. This proves independence, rather than merely a pointwise equality off a divisor. ◻

We now specify the topology rather than infer continuity from the preceding fixed-test equalities. Write Er,s(U)=AU,cr,s\mathcal E^{r,s}(U)=\mathcal A^{r,s}_{U,c}. A frozen cutoff presentation i=(V,g,Ω,χ)i=(V,g,\Omega,\chi) has an actual affinoid V⊂UV\subset U, analytic units gg on VV, an open Ω⊃tg(V)\Omega\supset t_g(V), and ordinary smooth compact χ\chi with support in Int⁡top(V,U)\operatorname{Int}_{\rm top}(V,U). Its map is Qiα=ext⁡0(χtg∗α),tg=−log⁡∣g∣.Q_i\alpha=\operatorname{ext}_0(\chi t_g^*\alpha),\qquad t_g=-\log|g|. The coefficient space Er,s(Ω)E^{r,s}(\Omega) has the usual seminorms pL,m(α)=max⁡I,J,∣a∣≤msup⁡L∣∂aαI,J∣p_{L,m}(\alpha)=\max_{I,J,|a|\leq m}\sup_L |\partial^a\alpha_{I,J}|. Definition NAG-002472 defines τcut\tau_{\rm cut} by those seminorms PP on E\mathcal E for which every PQiP Q_i is continuous. Frozen NAG-002473 proves that this equals the final ordinary compatible-presentation topology τord\tau_{\rm ord}. We use that comparison without reproving it. The topology on interior tests is this construction on UU; no unproved identification with a subspace topology on boundary-reaching tests of XX is made.

Proposition. NAG-002999 (Order-zero coefficient representations; F8MG-O). For every cutoff presentation, each applicable pairing Lf(r=s=n),Gs(r=s=n),δD(r=s=n−1),Tω(α)=∫ω∧αL_f\quad(r=s=n),\qquad G_s\quad(r=s=n),\qquad \delta_{D}\quad(r=s=n-1),\qquad T_\omega(\alpha)=\int\omega\wedge\alpha (where ω\omega is a fixed ordinary smooth complementary form) has a bound ∣T(Qiα)∣≤Ciptg(V),0(α),Ci<∞.\begin{equation} |T(Q_i\alpha)|\leq C_i p_{t_g(V),0}(\alpha), \qquad C_i<\infty. \end{equation} Here DD is any of the locally finite Cartier differences above. More precisely, the functional on each coefficient presentation has a finite representation T(Qiα)=∑ℓ,I,J∫σℓbℓ,IJ(w)αIJ(Pℓw) dλℓ(w),Pℓ(σℓ)⊆tg(V),\begin{equation} T(Q_i\alpha)=\sum_{\ell,I,J} \int_{\sigma_\ell}b_{\ell,IJ}(w) \alpha_{IJ}(P_\ell w)\,d\lambda_\ell(w), \qquad P_\ell(\sigma_\ell)\subseteq t_g(V), \end{equation} with fixed compact polytopes, lattice-normalized real cell integrals, and fixed smooth bounded densities. They retain all actual integer cycle weights and finite-cover signs. This is a local order-zero representation on coefficient spaces; no native signed Radon measure on the analytic space is assumed or constructed here.

For a globally represented quotient f=a/bf=a/b, smooth constant truncations of the two logarithms give pairings exactly equal to LfL_f on the entire image of each fixed top-degree QiQ_i, once the truncation is sufficiently deep. The threshold depends on ii. The logarithmic parts of this proposition require regular factors and the frozen logarithmic lemma, not ambient S1S_1 or scalar PL.

The logarithmic step for supplied regular factors uses the logarithmic pairing argument of NAG-002989. The Green pairing uses Proposition I.0.17 where applicable. The compact cells use the finite-cell functional and unit augmentation on the same affinoid, with exact finite-cover inclusion–exclusion.

Proof. First suppose f=a/bf=a/b on a neighbourhood of Supp⁡χ\operatorname{Supp}\chi. For every constant ambient basis form eIJ=d′tI∧d′′tJe_{IJ}=d't_I\wedge d''t_J of bidegree (n,n)(n,n), put βIJ=QieIJ\beta_{IJ}=Q_i e_{IJ}. There are finitely many of these fixed compact ordinary forms. By frozen NAG-002806, each vanishes on an ordinary neighbourhood of V(a)∪V(b)V(a)\cup V(b). Hence Ki=⋃I,JSupp⁡βIJK_i=\bigcup_{I,J}\operatorname{Supp}\beta_{IJ} is compact and disjoint from these zero loci. Every Qiα=∑I,JαIJ(tg)βIJQ_i\alpha=\sum_{I,J}\alpha_{IJ}(t_g)\beta_{IJ} has support in this same KiK_i, independently of the variable coefficients. This finite-basis argument is the uniform step; fixed-test integrability alone would not suffice. If there are no coefficient slots, the entire presentation is zero.

The logarithms are bounded on KiK_i. Choose a smooth compact cutoff θ\theta, supported where both factors are units and equal to one near KiK_i, using NAG-002427/C.1.60. The zero extension b0=ext⁡0(θlog⁡∣a/b∣)b_0=\operatorname{ext}_0(\theta\log|a/b|) is an ordinary smooth function and Lf(Qiα)=∫b0QiαL_f(Q_i\alpha)=\int b_0Q_i\alpha. For a meromorphic function given only locally, choose finitely many quotient neighbourhoods and a fixed smooth partition ρj\rho_j near Supp⁡χ\operatorname{Supp}\chi. Apply the same argument to the finitely many fixed basis forms ρjβIJ\rho_j\beta_{IJ} on those neighbourhoods. It produces finitely many fixed smooth multipliers bjb_j and reduces LfQiL_f Q_i to integrals of bjρjQiαb_j\rho_j Q_i\alpha. These functions and their support bounds are chosen before α\alpha. In particular this argument uses no result from the unreviewed component packet.

For any fixed smooth multiplier or complementary form, cover the fixed compact support by finitely many actual ordinary affinoid neighbourhoods inside VV, and take joint unit tuples representing that fixed data and gg. The variable coefficient appears only as αIJ(Pw)\alpha_{IJ}(P w), where PP is projection to the original gg-coordinates. The native finite-chart integral NAG-002405/5.50.6/5.52.4 supplies finitely many compact cells. Restriction of the two covector factors to a lattice tangent basis contributes fixed minors. Multiplying those minors, the fixed coefficient functions, actual weights, and inclusion– exclusion signs gives the densities in (I.0.29). This proves that formula for LfL_f, TωT_\omega, and therefore GsG_s after a finite frame partition. For δD\delta_D, the same construction is made on the actual pure (n−1)(n-1)-dimensional zero spaces, using R-ZP and NAG-002778. Closed-immersion pullback preserves compact support because its underlying image is closed. Its moment coordinates are the restricted gg, and the full quotient weights are retained.

Every summand is bounded in absolute value by λℓ(σℓ)∥bℓ,IJ∥∞ptg(V),0(α)\lambda_\ell(\sigma_\ell)\|b_{\ell,IJ}\|_\infty p_{t_g(V),0}(\alpha). The finite sum gives (I.0.28). These integrals also make sense for continuous coefficient functions on the compact moment image: the continuous real polytope integral is constructed in frozen NAG-002389, with a boundary-box estimate and uniform continuity. Smooth restrictions are uniformly dense in continuous functions on a compact subset of Ω\Omega: take finitely many Euclidean smooth bumps of small support, normalize their sum near the compact set, and use the value of the continuous function at one point in each small piece. Uniform continuity bounds the error. Thus (I.0.29) extends the coefficient functional uniquely in its zeroth-order norm. This is distinct from allowing arbitrary measurable multipliers in the frozen grid integral or from representing all these refinements by one analytic-space measure.

For the final assertion choose a smooth λM:R→R\lambda_M:\mathbb R\to\mathbb R which equals tt for t≥−Mt\geq-M and is the constant −M-M for t≤−M−1t\leq-M-1; the real smooth profile in NAG-002427 supplies one. Define ℓM(a)=λM(log⁡∣a∣)\ell_M(a)=\lambda_M(\log|a|) on the nonzero locus and −M-M at its zeros. It is ordinary smooth, being constant on a neighbourhood of the zeros and a smooth function of a unit moment coordinate elsewhere. Set uM=ℓM(a)−ℓM(b)u_M=\ell_M(a)-\ell_M(b). On the common KiK_i the two logarithms have fixed finite lower bounds. For all sufficiently large MM, uM=log⁡∣a/b∣u_M=\log|a/b| there, so their products with every QiαQ_i\alpha are identical. The cutoff presentations span compact tests by the finite partition theorem; hence this also gives eventual equality on every individual top test. No threshold uniform over varying presentations is asserted. ◻

Whole fibres on the original source

Fix a complete non-Archimedean field KK with its specified real-valued absolute value, possibly trivial. Affinoids may have arbitrary positive real presentation radii. Finite maps are norm-admissible finite, with their canonical finite-module topologies. All domains, analytic fibres, nilpotents and residue fields are retained.

Let Ω\Omega be a good Hausdorff KK-analytic space, and let f=(f1,…,fn)f=(f_1,\ldots,f_n), n≥1n\geq1, be global analytic units. Write ϕ:Ω⟶T=(Gmn)an,t=trop⁡∘ϕ=−log⁡∣f∣.\phi:\Omega\longrightarrow T=(\mathbb G_m^n)^{\rm an}, \qquad t=\operatorname{trop}\circ\phi=-\log|f|. For a closed rectangular box P⊂RnP\subset\mathbb R^n, put TP=trop⁡−1(P)T_P=\operatorname{trop}^{-1}(P) and ΩP=t−1(P)\Omega_P=t^{-1}(P), with the latter inverse image taken in all of the intended source Ω\Omega. The exact polytorus is TwT_w, its Gauss point is ηw\eta_w, and FwΩ=t−1(w)F_w^\Omega=t^{-1}(w).

Interior on the actual charts.

For an affinoid domain XX of the good source Ω\Omega, the formulas used below are Int⁡(Ω/T)=Int⁡(Ω/K),Int⁡(X/K)=X∩Int⁡(Ω/K)∩Int⁡top(X,Ω).(1)\operatorname{Int}(\Omega/T)=\operatorname{Int}(\Omega/K), \qquad \operatorname{Int}(X/K) =X\cap\operatorname{Int}(\Omega/K)\cap\operatorname{Int}_{\rm top}(X,\Omega). \tag{1} For the first formula take a Laurent affinoid neighbourhood of the target point with all bounds strict at that point, and an ordinary affinoid neighbourhood of the source point mapping into it. The target point is absolutely interior by this presentation. The source-factor inclusion and forward composition give the equivalence on these two affinoids. The ordinary local test then gives the asserted equivalence on the good spaces.

For the second formula, let y∈Xy\in X, choose an ordinary affinoid neighbourhood WW of yy in Ω\Omega, and choose an affinoid neighbourhood VV of yy in XX contained in the ordinary interior of WW. Both domain presentations use the induced analytic structure, so VV is an affinoid domain of WW. Absolute interiority in XX is equivalent to absolute interiority in VV by the ordinary local test. The source-factor inclusion for V→W→M(K)V\to W\to\mathcal M(K) then gives relative interiority in V/WV/W. The affinoid domain criterion makes VV an ordinary neighbourhood at yy in WW. Thus yy is ordinary interior in XX, and the local test on WW and Ω\Omega gives absolute interiority in Ω\Omega. Conversely, if yy is ordinary interior in XX and absolutely interior in Ω\Omega, take one ordinary affinoid neighbourhood of yy in Ω\Omega contained in XX and use the local test on both spaces.

Theorem. NAG-002857 (A criterion for an affinoid containing the whole interior fibre). Assume w∈t(Ω)w\in t(\Omega), and suppose that for one closed full-dimensional box P0P_0, with w∈int⁡(P0)w\in\operatorname{int}(P_0), the whole domain ΩP0\Omega_{P_0} is compact Hausdorff. The following conditions are equivalent.

  1. There is an affinoid domain X⊂ΩX\subset\Omega such that FwΩ⊂Int⁡(X/K).(2)F_w^\Omega\subset\operatorname{Int}(X/K). \tag{2}

  2. Every point of FwΩF_w^\Omega is in Int⁡(Ω/T)\operatorname{Int}(\Omega/T), and for every analytic point a∈Twa\in T_w the underlying set of the entire analytic fibre Ωa\Omega_a is finite. Empty fibres are allowed.

  3. There is a closed full-dimensional box P⊂int⁡(P0)P\subset \operatorname{int}(P_0), with w∈int⁡(P)w\in\operatorname{int}(P), such that ϕ:ΩP⟶TPis norm-admissible finite.(3)\phi:\Omega_P\longrightarrow T_P \quad\hbox{is norm-admissible finite}. \tag{3} In particular ΩP\Omega_P is affinoid and FwΩ⊂Int⁡(ΩP/K)F_w^\Omega\subset\operatorname{Int}(\Omega_P/K).

The boxes in condition 3 may be made arbitrarily small about ww. Restriction to any analytic point of TPT_P, or any exact polytorus Tv⊂TPT_v\subset T_P, retains the entire original-source analytic fibre with its actual algebra and canonical norm.

If Ω\Omega is boundaryless over KK, the relative-interior part of condition 2 is automatic. Alternatively, if ΩP0→TP0\Omega_{P_0}\to T_{P_0} is analytically proper, it supplies both compactness and the required relative interiority at the central level. In either case the finite-fibre condition still quantifies over all of TwT_w, not only ηw\eta_w. No flatness assumption is made.

Proof. Assume condition 1. The whole moment fibre in XX at ww is exactly FwΩF_w^\Omega, since that entire set lies in XX. The frozen theorem 5.38.4 therefore gives a finite collar in XX for the original tuple ff. For every a∈Twa\in T_w, its fibre is the entire fibre in Ω\Omega; hence its underlying set is finite, by 2.35.1. Equation (1) propagates the absolute interiority from XX to Ω\Omega, and identifies it with relative interiority for ϕ\phi. This proves condition 2. The whole-fibre finite collar theorem itself is inherited rather than resubmitted.

Assume condition 2 and set A0=ΩP0A_0=\Omega_{P_0}. This analytic domain is good: intersect an ordinary affinoid neighbourhood in Ω\Omega with the finitely many bounds on fjf_j and fj−1f_j^{-1}. The intersection is an actual affinoid domain and is an ordinary neighbourhood in A0A_0. Thus A0A_0 is a compact good Hausdorff analytic space. The map p0:A0→TP0p_0:A_0\to T_{P_0} is compact: every compact target domain is closed in the Hausdorff target, and its inverse image is closed in the compact source.

Fix a∈Twa\in T_w. Its fibre under p0p_0 is the entire Ωa\Omega_a. At every point of that fibre, both A0⊂ΩA_0\subset\Omega and TP0⊂TT_{P_0}\subset T are ordinary neighbourhoods, since all box inequalities are strict at weight ww. Ordinary local testing transfers the relative interiority in condition 2 to p0p_0. The frozen lemma 5.56.1 therefore supplies an ordinary open neighbourhood Na⊂TP0N_a\subset T_{P_0} over which the full restriction of p0p_0 is finite. This includes the empty-fibre case. The union N=⋃a∈TwNaN=\bigcup_{a\in T_w}N_a contains the entire exact polytorus, and p0−1(N)→Np_0^{-1}(N)\to N is finite by target locality. No branch or Gauss lift has been selected.

The compact set TP0∖NT_{P_0}\setminus N has compact tropical image, and that image misses ww. Choose a closed box PP about ww, inside int⁡(P0)\operatorname{int}(P_0), avoiding that image and lying in any prescribed real neighbourhood of ww. Then TP⊂NT_P\subset N. Finite base change yields the finite map p0−1(TP)=ΩP⟶TP.p_0^{-1}(T_P)=\Omega_P\longrightarrow T_P. Its source is affinoid by the definition of a finite analytic morphism, and its Banach algebra has the canonical finite norm. For y∈FwΩy\in F_w^\Omega, the target point is absolutely interior in TPT_P, and the finite map is relatively interior. The affinoid composition formula gives y∈Int⁡(ΩP/K)y\in\operatorname{Int}(\Omega_P/K). This proves condition 3. Condition 3 immediately implies 1.

All source identifications are inverse-image domain identities, using 2.40.11. Actual finite base change and 2.33.11 give the stated analytic fibre identifications; finite tensor comparison preserves the complete algebras. In particular, neither reduction nor a selection of a local Artinian factor occurs.

If Ω\Omega is boundaryless, (1) proves the final interior assertion. For an analytically proper p0p_0, compactness is part of properness and every source point is relatively interior. At the central level ordinary local testing again transfers that interiority to Ω/T\Omega/T. This is exactly the project convention 2.53.1; topological compactness alone does not include boundarylessness. ◻

Theorem. NAG-002861 (Whole-fibre capture in a boundaryless affinoid exhaustion). Suppose that Ω\Omega is boundaryless over KK and has increasing affinoid domains U1⊂U2⊂⋯U_1\subset U_2\subset\cdots whose ordinary interiors in Ω\Omega cover Ω\Omega. Let ff be any tuple of global analytic units and let w∈t(Ω)w\in t(\Omega). If ΩP0\Omega_{P_0} is compact for some closed full-dimensional box P0P_0 with w∈int⁡(P0)w\in\operatorname{int}(P_0), then there are an index mm and arbitrarily small closed full-dimensional boxes P⊂int⁡(P0)P\subset\operatorname{int}(P_0), with w∈int⁡(P)w\in\operatorname{int}(P), such that ΩP⊂Int⁡top(Um,Ω),ΩP⟶TPis norm-admissible finite,FwΩ⊂Int⁡(ΩP/K).(4)\Omega_P\subset\operatorname{Int}_{\rm top}(U_m,\Omega), \qquad \Omega_P\longrightarrow T_P\quad\hbox{is norm-admissible finite}, \qquad F_w^\Omega\subset\operatorname{Int}(\Omega_P/K). \tag{4} The inverse images are full inverse images in Ω\Omega. The fibre-finiteness assertion in condition 2 of Theorem I.1.1 is a consequence here, not an additional hypothesis.

This applies, in particular, to the analytification of an affine finite-type KK-scheme, provided the whole weight-box inverse image is compact. The unit functions may be global analytic functions; no algebraicity of the tuple is required.

Proof. The fibre FwΩF_w^\Omega is a closed subset of the compact ΩP0\Omega_{P_0}. Finitely many of the interiors of the UjU_j cover it. Taking the largest index gives FwΩ⊂Int⁡top(Um,Ω).F_w^\Omega\subset\operatorname{Int}_{\rm top}(U_m,\Omega). Boundarylessness and (1) imply FwΩ⊂Int⁡(Um/K)F_w^\Omega\subset\operatorname{Int}(U_m/K). Theorem I.1.1 applies and gives a full finite collar in Ω\Omega. To put the collar inside the specified UmU_m, use the compact set C=ΩP0∖Int⁡top(Um,Ω).C=\Omega_{P_0}\setminus\operatorname{Int}_{\rm top}(U_m,\Omega). Its tropical image is compact and misses ww. Shrink the finite collar box to avoid t(C)t(C), still keeping ww in its interior. This proves every assertion of (4). In particular, the same compact argument controls escape outside UmU_m; no completeness over Ω\Omega is inferred merely from a collar inside UmU_m.

For Ω=(Spec⁡B)an\Omega=(\operatorname{Spec}B)^{\rm an}, choose a finite algebra presentation B=K[X1,…,XN]/IB=K[X_1,\ldots,X_N]/I. The exact radius-piece construction 2.33.23 gives affinoid domains with ∣Xj∣≤Rm|X_j|\le R_m, for any increasing sequence Rm→∞R_m\to\infty. Every point has all coordinates strictly below some RmR_m, so their ordinary interiors cover. The construction is good and Hausdorff, and 2.33.27 supplies absolute boundarylessness. The preceding argument uses only these properties and the stated compactness of the full inverse image. It places no extra restriction on the global analytic unit tuple. ◻

Proposition. NAG-002864 (A full-rank Gauss lift and a vertical projective fibre). Over any complete field KK as above, put Talg=Spec⁡K[T1±1,T2±1]T_{\rm alg}=\operatorname{Spec}K[T_1^{\pm1},T_2^{\pm1}] and define the closed projective incidence surface Y={(T1,T2,[U:V])∈Talg×PK1:(T1−1)V=(T2−1)U}.(5)Y=\{(T_1,T_2,[U:V])\in T_{\rm alg}\times\mathbf P^1_K: (T_1-1)V=(T_2-1)U\}. \tag{5} Let Ω=Yan\Omega=Y^{\rm an}, and let ϕ:Ω→(Gm2)an\phi:\Omega\to(\mathbb G_m^2)^{\rm an} be the projection, with moment tuple f=(T1,T2)f=(T_1,T_2) and weight w=(0,0)w=(0,0). Then Ω\Omega is good, Hausdorff and boundaryless, ϕ\phi is proper, and every whole weight-box inverse image is compact. The fibre over the torus Gauss point ηw\eta_w consists of one point xx, with unchanged completed residue field and graded moment rank two. Nevertheless no affinoid domain in Ω\Omega contains the entire FwΩF_w^\Omega.

Moreover, there is a pure two-dimensional affinoid domain X⊂ΩX\subset\Omega containing xx in its absolute interior for which FwX⊄Int⁡(X/K).(6)F_w^X\not\subset\operatorname{Int}(X/K). \tag{6} It has the actual presentation, on the chart U≠0U\ne0 with z=V/Uz=V/U, X={12≤∣T1∣,∣T2∣≤2,∣z∣≤2,T2=1+(T1−1)z}.(7)X=\left\{ \tfrac12\le |T_1|,|T_2|\le2,\quad |z|\le2,\quad T_2=1+(T_1-1)z \right\}. \tag{7} Thus even a good moment chart about an interior full-rank Gauss lift in a proper boundaryless ambient source need not have an interior whole moment fibre. The restriction of the proper ambient map to this cut is compact but has relative boundary at weight ww; it is not analytically proper.

Proof. On the chart U≠0U\ne0, the equation eliminates T2T_2 and gives the affine ring K[T1,z,T1−1,(1+(T1−1)z)−1].(8)K[T_1,z,T_1^{-1},(1+(T_1-1)z)^{-1}]. \tag{8} The chart V≠0V\ne0 has the symmetric description. These charts give a finite-type separated scheme, and its projection to the torus is projective by (5). The exact analytification results 2.33.23, 2.33.29, and 2.33.27 give goodness, Hausdorffness, and boundarylessness of both the structural map and the projection. The compactness theorem 2.33.30 makes ϕ\phi compact, hence proper. In particular the inverse image of each compact polyannulus TPT_P is compact. No proper-pushforward or integration result is used.

Away from p=(1,1)p=(1,1), the inverse to the projection is explicitly (T1,T2)⟼(T1,T2,[T1−1:T2−1]).(T_1,T_2)\longmapsto (T_1,T_2,[T_1-1:T_2-1]). At ηw\eta_w, both ∣Tj−1∣|T_j-1| are one, so this point is outside pp, and ϕ\phi is an analytic isomorphism on an ordinary neighbourhood of it. It follows that its entire Gauss fibre is {x}\{x\} and H(x)=H(ηw)\mathcal H(x)=\mathcal H(\eta_w). The initial forms of T1,T2T_1,T_2 at the weighted Gauss point are algebraically independent over K~gr\widetilde K^{\rm gr}: a nonzero homogeneous Laurent polynomial has the prescribed Gauss norm and cannot vanish in the graded field. Thus the graded moment rank is two. In chart (8), ∣z(x)∣=∣T2−1∣/∣T1−1∣=1|z(x)|=|T_2-1|/|T_1-1|=1.

The fibre over pp, by contrast, is the whole E=(PK1)anE=(\mathbf P^1_K)^{\rm an}, with its actual analytic structure: the incidence equation becomes zero. It is a closed analytic subspace of Ω\Omega and is contained in FwΩF_w^\Omega. If an affinoid domain A⊂ΩA\subset\Omega contained all of FwΩF_w^\Omega, the pullback of the closed immersion E↪ΩE\hookrightarrow\Omega to AA would identify EE with a closed analytic subspace of an affinoid. Hence EE would itself be affinoid. It is boundaryless over KK, so 2.43.8 would make it finite over KK. Its spectrum would then be a finite set. This is impossible: the weighted Gauss points in its affine line chart, indexed by all radii ρ>0\rho>0, are distinct because the value of the affine coordinate is ρ\rho. This proves the failure of any affinoid container for the whole original moment fibre.

We now verify the cut, rather than presuming its boundary behaviour. The domain (7) has affinoid algebra D=K{2−1T1,2−1S1,2−1T2,2−1S2,2−1z}/(T1S1−1,T2S2−1,T2−1−(T1−1)z).(9)D=K\{2^{-1}T_1,2^{-1}S_1,2^{-1}T_2,2^{-1}S_2,2^{-1}z\} /(T_1S_1-1,T_2S_2-1,T_2-1-(T_1-1)z). \tag{9} The domain universal property, applied in chart (8), identifies this with the actual indicated analytic domain, including its Banach structure. Equivalently, it is the generalized rational domain of the bidisc ∣T1∣,∣z∣≤2|T_1|,|z|\le2 cut out by the lower bound on T1T_1 and the two bounds on 1+(T1−1)z1+(T_1-1)z. The Gauss norm makes the bidisc integral. Its completed residue fields are completions of fields generated by at most two coordinate classes, so 3.22.3 bounds their residual dimensions by two. The weighted Gauss field attains two by 3.22.4; 3.22.7 therefore gives analytic dimension two. The frozen domain-purity theorem 5.35.3 therefore makes the nonempty XX pure of dimension two.

All inequalities in (7) are strict at xx. The chart is ordinary open in Ω\Omega, so XX contains an ordinary neighbourhood of xx. Since Ω\Omega is boundaryless, (1) gives x∈Int⁡(X/K)x\in\operatorname{Int}(X/K). For ρ>0\rho>0, let yρy_\rho be the Gauss point of radius ρ\rho on the exceptional affine chart T1=T2=1T_1=T_2=1, with coordinate zz. Then y2∈Xy_2\in X, whereas yρ∉Xy_\rho\notin X for ρ>2\rho>2. As ρ↓2\rho\downarrow2, these points converge to y2y_2. Indeed polynomial evaluations are the continuous functions max⁡j∣aj∣ρj\max_j|a_j|\rho^j; on any fixed larger closed disc, uniform polynomial approximation gives the same continuity for all analytic functions, which determine the spectrum topology. These are actual points over the original field, also when its absolute value is trivial.

Thus y2∉Int⁡top(X,Ω)y_2\notin\operatorname{Int}_{\rm top}(X,\Omega). Equation (1) gives y2∉Int⁡(X/K)y_2\notin\operatorname{Int}(X/K), but t(y2)=(0,0)t(y_2)=(0,0). This proves (6). The map from XX to the target polyannulus of (7) is compact. At y2y_2 the target bounds are strict, so the affinoid composition formula identifies relative interiority there with absolute interiority in XX. Consequently that map has relative boundary and is not proper. Likewise no full collar (t∣X)−1(P)→TP(t|_X)^{-1}(P)\to T_P, with ww strictly inside PP, can be finite: finite-map innerness and target innerness would make y2y_2 absolutely interior in that domain and, by composition, in XX, a contradiction. ◻

Meromorphic Cartier coefficients

Quotient representatives must be regular on actual ordinary charts. Use the verified effective coefficient comparison and actual-domain transport. The pairing below uses the full pure zero-space integral.

Theorem. NAG-002791 (Meromorphic coherent-cycle comparison; F8CA-I). Let XX be good, Hausdorff, and pure of analytic dimension n≥1n\geq1. Let ff be regular meromorphic and admit ordinary local presentations a/ba/b, with a,ba,b regular and their actual zero spaces empty or pure of dimension n−1n-1. Restrict comparisons to such presentations. Then the differences of the effective Cartier cycles are independent of presentation, glue on actual common domains, and have the coherent-cycle integration pairing asserted in the pure-dimensional clause of G.0.17, on ordinary (n−1,n−1)(n-1,n-1)-tests with compact support in Int⁡(X/K)\operatorname{Int}(X/K).

Proof. Two representatives can be restricted to an actual common ordinary affinoid neighbourhood, say M(A)\mathcal M(A), where their four equations a,b,c,ea,b,c,e are defined and regular and a/b=c/ea/b=c/e. The zero spaces of these restricted equations are actual affinoid domains in the original zero spaces, by quotient/domain base change NAG-001702, 2.36.9. Frozen domain purity NAG-002413 makes each one empty or pure of dimension n−1n-1. Thus Lemma C.1.57 applies to (a,e)(a,e) and (b,c)(b,c), and also to any further products used for a common denominator.

Localization at regular elements is injective, so equality of the fractions gives ae=bcae=bc in this actual common ring. Theorem F.1.13 now gives div⁡(a)+div⁡(e)=div⁡(ae)=div⁡(bc)=div⁡(b)+div⁡(c).\operatorname{div}(a)+\operatorname{div}(e)=\operatorname{div}(ae)=\operatorname{div}(bc) =\operatorname{div}(b)+\operatorname{div}(c). Subtracting proves independence. The already verified effective actual-domain comparison makes these equalities agree on refinements and gives ordinary gluing. For triple overlaps use a common further affinoid refinement; the same equality is transitive there.

The effective chain identity of G.0.17 retains the actual whole Gauss algebra, all generic lengths, full residue degrees, and one lattice-index division. Subtract its two effective formulas. The independence just proved now justifies the finite localization and partition sum for a compact test, using NAG-002425 and C.1.60. Its restriction to each zero space is compact and inside that space’s analytic interior: a closed immersion is finite and relatively interior, and NAG-001704 composes interiors. No boundary term, logarithmic limit, or test-space convergence is introduced. ◻

Full facet values and finite covers

The complete non-Archimedean field KK keeps its specified absolute value. Trivial valuation and generalized positive real radii are allowed. Let XX be good, Hausdorff, and pure of analytic dimension n≥1n\geq1, and let F\mathcal F be coherent on XGX_G. All forms are ordinary smooth moment forms. All coefficients integrated in this packet are smooth or continuous coefficients on finitely many compact polyhedra. No measurable extension of the frozen grid integral is used.

The frozen convention NAG-002388 has all generators anticommuting. The involution JJ exchanges primed and double-primed generators and fixes functions; d′′=Jd′Jd''=Jd'J. On an integral rr-plane with coordinates y1,…,yry_1,\ldots,y_r, the positive generator is ωr=d′y1∧d′′y1∧⋯∧d′yr∧d′′yr=sr d′y1∧⋯∧d′yr∧d′′y1∧⋯∧d′′yr,sr=(−1)r(r−1)/2.\omega_r=d'y_1\wedge d''y_1\wedge\cdots\wedge d'y_r\wedge d''y_r =s_r\,d'y_1\wedge\cdots\wedge d'y_r\wedge d''y_1\wedge\cdots\wedge d''y_r, \qquad s_r=(-1)^{r(r-1)/2}. Thus Jωr=(−1)rωrJ\omega_r=(-1)^r\omega_r. A symmetric (p,p)(p,p)-form means Jα=(−1)pαJ\alpha=(-1)^p\alpha in this convention. The different convention

Let V=M(A)V=\mathcal M(A) be pure of dimension nn, MM a finite AA-module, and gg a tuple of units. Put C=Tg(M)C=\mathcal T_g(M). Choose a finite common subdivision of its actual moment image and boundary image. For every (n−1)(n-1)-face τ\tau, choose a basis v1,…,vn−1v_1,\ldots,v_{n-1} of its tangent lattice. The primitive normal uσ/τu_{\sigma/\tau} points into the incident nn-cell σ\sigma. Set bC(τ)=∑σ⊃τmσuσ/τ(mod(Nτ)R),ϵn=(−1)n(n+1)/2.b_C(\tau)=\sum_{\sigma\supset\tau}m_\sigma u_{\sigma/\tau} \pmod{(N_\tau)_{\mathbb R}},\qquad \epsilon_n=(-1)^{n(n+1)/2}. For a smooth ambient (n−1,n)(n-1,n)-form β~\widetilde\beta, define λC,τ,w(β~)=ϵnβ~w(v1′,…,vn−1′;v1′′,…,vn−1′′,bC(τ)′′),ΦC(β~)=∑τ∫τλC,τ,w(β~) dμτ(w).\begin{align} \lambda_{C,\tau,w}(\widetilde\beta) &=\epsilon_n\widetilde\beta_w (v_1',\ldots,v_{n-1}';v_1'',\ldots,v_{n-1}'',b_C(\tau)''), \\ \Phi_C(\widetilde\beta) &=\sum_\tau\int_\tau\lambda_{C,\tau,w}(\widetilde\beta) \,d\mu_\tau(w). \end{align} Here μτ\mu_\tau gives the chosen tangent basis covolume one. Alternation removes dependence on the normal lift, and the two determinant factors remove orientation dependence. NAG-002389 gives the proved identity ΦTg(M)(β~)=IM,g(d′β~).\begin{equation} \Phi_{\mathcal T_g(M)}(\widetilde\beta) =\mathcal I_{M,g}(d'\widetilde\beta). \end{equation} Every facet is included, including unbalanced interior facets and original analytic-boundary images. Nothing has yet been set equal to zero.

The weights in this formula are mσ=dim⁡HQ(M⊗AEQ)/[Zn:Q(Nσ)]m_\sigma=\dim_{H_Q}(M\otimes_A E_Q)/[\mathbb Z^n:Q(N_\sigma)] on a complete transverse Gauss fibre, in the notation of NAG-002362. Thus full module lengths, nilpotents, and full residue degrees enter once. No selected geometric lift, reduced fibre, or separable degree is substituted. Coherent modules, and finite integer linear combinations of their chains, are allowed. No real-weight analytic cycle extension is asserted.

R-F7. On a good Hausdorff pure nn-dimensional analytic space, for coherent F\mathcal F, an ordinary (n−1,n)(n-1,n)-form β\beta with compact support in Int⁡(X/K)\operatorname{Int}(X/K) satisfies ∫X,Fd′β=0\int_{X,\mathcal F}d'\beta=0; the corresponding d′′d'' statement holds in bidegree (n,n−1)(n,n-1).

Theorem. NAG-002875 (The full ordinary boundary functional; F8AG-T). For β∈AX,cn−1,n(X)\beta\in\mathcal A^{n-1,n}_{X,c}(X), the finite facet construction below gives a canonical linear functional BX,F(β)\mathcal B_{X,\mathcal F}(\beta) with BX,F(β)=∫X,Fd′β.\begin{equation} \mathcal B_{X,\mathcal F}(\beta)=\int_{X,\mathcal F}d'\beta. \end{equation} With R-F7, it factors through the actual ordinary trace map r∂X:AX,cn−1,n(X)⟶∏z∈∂KXAX,zn−1,n,β⟼(βz)z.\begin{equation} r_{\partial X}:\mathcal A^{n-1,n}_{X,c}(X)\longrightarrow \prod_{z\in\partial_KX}\mathcal A^{n-1,n}_{X,z},\qquad \beta\longmapsto(\beta_z)_z. \end{equation} That is, there is a unique functional on im⁡r∂X\operatorname{im}r_{\partial X} whose composite is BX,F\mathcal B_{X,\mathcal F}. This is a full boundary-germ functional, not an assertion that a boundary trace vanishes.

Proof. Let S=Supp⁡βS=\operatorname{Supp}\beta. Choose finitely many ordinary affinoid neighbourhoods ViV_i, whose ordinary interiors cover SS, such that β∣Vi=tgi∗β~i\beta|_{V_i}=t_{g_i}^*\widetilde\beta_i. Their union QQ is a compact analytic neighbourhood of SS. For each nonempty index set II, put VI=⋂i∈IViV_I=\bigcap_{i\in I}V_i, and use one i∈Ii\in I to restrict its tuple and representative to VIV_I. The coefficient module is the actual MI=Γ(VI,F)M_I=\Gamma(V_I,\mathcal F). Define BX,F(β)=∑I≠∅(−1)∣I∣+1ΦTgi∣VI(MI)(β~i).\begin{equation} \mathcal B_{X,\mathcal F}(\beta) =\sum_{I\ne\varnothing}(-1)^{|I|+1} \Phi_{\mathcal T_{g_i|V_I}(M_I)}(\widetilde\beta_i). \end{equation} Empty intersections contribute zero. Domain purity supplies the correct analytic dimension for every nonempty intersection, even if its moment image has lower dimension.

Apply (I.1.3) term by term. Exterior differentiation is well-defined on ordinary forms by NAG-002424, and Supp⁡(d′β)⊂S\operatorname{Supp}(d'\beta)\subset S. The finite compact integration theorem NAG-002416 identifies the sum with IQ,F(d′β)\mathcal I_{Q,\mathcal F}(d'\beta), and NAG-002425 identifies this with the right side of (I.1.4). Those theorems prove independence of the choice of ii, all representatives, subdivisions, the presenting cover, and the compact neighbourhood. Thus the explicit facet construction has the claimed intrinsic value. A single global moment representative on an affinoid recovers (I.1.2) itself.

The kernel of (I.1.5) consists exactly of the compact sections zero on an ordinary neighbourhood of the actual analytic boundary. Such a section has compact support in Int⁡(X/K)\operatorname{Int}(X/K). R-F7 makes its right side in (I.1.4) zero. Elementary linear factorization proves the boundary-germ assertion. This factor is defined only on the image of the trace map; it does not assign values to arbitrary incompatible germs along the boundary. ◻

Theorem I.1.6. (Full trace inclusion–exclusion; F8AG-C). Let QQ be a finite union of affinoid domains in a good Hausdorff pure nn-dimensional analytic space. Let F\mathcal F be coherent on QGQ_G and let β\beta be compatible finite moment-chart data of degree (n−1,n)(n-1,n) on QQ. Its facet construction defines BQ,F(β)=IQ,F(d′β)\mathcal B_{Q,\mathcal F}(\beta)=\mathcal I_{Q,\mathcal F}(d'\beta). For any finite affinoid-domain cover V1,…,VsV_1,\ldots,V_s of QQ, BQ,F(β)=∑∅≠I⊆{1,…,s}(−1)∣I∣+1BVI,F∣VI(β∣VI),VI=⋂i∈IVi.\begin{equation} \mathcal B_{Q,\mathcal F}(\beta) =\sum_{\varnothing\ne I\subseteq\{1,\ldots,s\}} (-1)^{|I|+1}\mathcal B_{V_I,\mathcal F|V_I}(\beta|V_I), \qquad V_I=\bigcap_{i\in I}V_i. \end{equation} This holds with different unit tuples on different members. Every term is a full trace on its own actual domain. It does not require or imply that the individual terms vanish. It is unchanged under finite refinements and applies to the flux in Green.

Proof. All intersections are actual affinoid domains, by NAG-002411; all nonempty ones are pure of dimension nn, by NAG-002413. The coefficient on an intersection is the single actual restriction of F\mathcal F, equivalently Mi⊗O(Vi)O(VI)M_i\otimes_{\mathcal O(V_i)}\mathcal O(V_I) in a finite-module presentation. It is not an independently chosen multiplicity.

For a cover on which β\beta has one representative per member, apply frozen finite inclusion–exclusion NAG-002416 to d′βd'\beta, then (I.1.3) on each intersection. This is (I.1.7). On VIV_I, the two unit tuples from any two chosen members are both defined and can be adjoined. Their joint pullbacks represent the same ordinary form; their derivatives have equal tangential top restrictions by NAG-002422/5.52.3. Global chain coordinate augmentation NAG-002405 gives the same integral. No unit is extended outside its actual domain.

For two arbitrary finite covers (Vi)(V_i) and (Wj)(W_j), refine on VI∩WJV_I\cap W_J. The same frozen formula gives the finite double sum with coefficient (−1)∣I∣+∣J∣(-1)^{|I|+|J|}. Its forms, derivatives, and coherent modules are restrictions of the same data on every actual intersection. Interchanging these finite sums recovers either cover. This also permits a nonpresenting member to be replaced by its finite presenting cover, proving the formula as stated. Triple overlaps are handled by the same actual restrictions; no unmatched jets are identified across unrelated closed pieces.

If one tuple and representative are defined on all of an affinoid QQ, NAG-002401 gives equality of the weighted top chains on one common real subdivision. Applying the linear normal expression (I.1.1) gives facetwise cancellation there. With different tuples, the invariant assertion is (I.1.7); joint-coordinate comparisons are made on actual overlaps. Intersecting unrelated moment images is not a substitute for intersecting the source domains. ◻

Fixed-stage logarithmic truncation

Write UU for a good Hausdorff analytic space over a complete non-Archimedean field, with arbitrary valuation. For an application with interior tests, take UU to be an ordinary open subset of the true analytic interior of the ambient space. Put E=Acp,q(U)E=\mathcal A_c^{p,q}(U), with the frozen ordinary moment sheaf and actual ordinary compact supports. The two final topologies of NAG-002472 are equal by NAG-002473; denote their common value by τa\tau_a. A cutoff generator has the form Qi:Ep,q(Ωi)⟶E,Qiη=ext⁡0(χitfi∗η),tfi=−log⁡∣fi∣.Q_i:E^{p,q}(\Omega_i)\longrightarrow E,\qquad Q_i\eta=\operatorname{ext}_0(\chi_i t_{f_i}^*\eta), \qquad t_{f_i}=-\log|f_i|. Here ViV_i is affinoid, fif_i is a unit tuple, Ωi\Omega_i contains tfi(Vi)t_{f_i}(V_i), and supp⁡χi\operatorname{supp}\chi_i is compact in Int⁡top(Vi,U)\operatorname{Int}_{\rm top}(V_i,U). The coefficient space has all compact C∞C^\infty seminorms. A seminorm on EE is τa\tau_a-continuous exactly when all its composites with QiQ_i are continuous. The images of the QiQ_i span EE.

Assume now that UU is pure of dimension n≥1n\geq1 and F\mathcal F is a coherent analytic module. Let g,h∈OU(U)g,h\in\mathcal O_U(U) be non-zero-divisors in every local ring, set Z=V(g)∪V(h)Z=V(g)\cup V(h), and assume the exact condition (Rn−1)(\mathrm R_{n-1}) of NAG-002479: for each z∈Zz\in Z there is an ordinary neighbourhood WzW_z, a good Hausdorff analytic space YzY_z of local dimension at most n−1n-1, a closed immersion into WzW_z whose image contains zz, and an analytic retraction onto YzY_z. The image need not contain the rest of Z∩WzZ\cap W_z. This is a stated hypothesis, not an existence assertion for arbitrary singular divisors.

Put u=log⁡∣g/h∣u=\log|g/h| off ZZ. Use the smooth truncation of the frozen NAG-002480: choose 0≤ρ≤10\leq\rho\leq1 smooth, zero on (−∞,0](-\infty,0] and one on [1,∞)[1,\infty), and define λM(t)=−M+(t+M)ρ(t+M+1).\lambda_M(t)=-M+(t+M)\rho(t+M+1). Extend λM(log⁡∣a∣)\lambda_M(\log|a|) by the constant −M-M at zeros of aa; write the result as ℓM(a)\ell_M(a). Set uM=ℓM(g)−ℓM(h)u_M=\ell_M(g)-\ell_M(h) and T(α)=∫U∖Z,Fuα,TM(α)=∫U,FuMα,DM=TM−T.T(\alpha)=\int_{U\setminus Z,\mathcal F}u\alpha,\quad T_M(\alpha)=\int_{U,\mathcal F}u_M\alpha,\quad D_M=T_M-T. Here E=Acn,n(U)E=\mathcal A_c^{n,n}(U), and the integral is the frozen native continuous coefficient integral. No arbitrary measurable multiplier is assigned a grid integral.

Theorem. NAG-003094 (Strong truncation in the frozen retraction scope). Under the preceding exact hypotheses, TMT_M and TT are continuous for τa\tau_a, and TM=TT_M=T on each τa\tau_a-bounded test set for all sufficiently large real MM.

More quantitatively, every cutoff presentation admits constants Ci,Hi,g,Hi,h≥0C_i,H_{i,g},H_{i,h}\geq0, with Hi=max⁡(Hi,g,Hi,h)H_i=\max(H_{i,g},H_{i,h}), such that ∣DMQiη∣≤Ai(M)pi(η),Ai(M)≤Ci((Hi,g−M)++(Hi,h−M)+),\begin{equation} |D_M Q_i\eta|\leq A_i(M)p_i(\eta),\qquad A_i(M)\leq C_i\bigl((H_{i,g}-M)_++(H_{i,h}-M)_+\bigr), \end{equation} where pi(η)=max⁡I,Jsup⁡tfi(Vi)∣ηI,J∣p_i(\eta)=\max_{I,J}\sup_{t_{f_i}(V_i)}|\eta_{I,J}|. For each a>0a>0, the weights wi=1+Ci(Hi,g+Hi,h)eaHiw_i=1+C_i(H_{i,g}+H_{i,h})e^{aH_i} define a continuous gauge PaP_a by (8.15.6) and give sup⁡α∈B∣(TM−T)(α)∣≤e−aMsup⁡α∈BPa(α)\begin{equation} \sup_{\alpha\in B}|(T_M-T)(\alpha)| \leq e^{-aM}\sup_{\alpha\in B}P_a(\alpha) \end{equation} on every τa\tau_a-bounded set BB. All coherent-cycle weights, lattice volumes, cutoff coefficients, pullback minors and finite inclusion–exclusion terms enter CiC_i.

Proof. The frozen NAG-002480 supplies TT, its continuity, and a compact Ki⊂U∖ZK_i\subset U\setminus Z containing the supports of all QiηQ_i\eta. It also supplies eventual equality on the whole presentation, uniformly in its unrestricted coefficients. Multiplication by the fixed smooth uMu_M and the finite native cell formula give continuity of TMQiT_MQ_i; hence TMT_M is continuous by finality. Theorem 8.15.12 already proves uniform eventual equality on bounded sets.

For the explicit constants, choose a fixed smooth cutoff equal to one near KiK_i and supported where g,hg,h are units. Use finitely many actual joint moment charts there, adjoining fi,g,hf_i,g,h and representations of this cutoff and χi\chi_i. The native finite cell formula has the form DMQiη=∑σ,I,Jϵσmσ∫σbσ,I,J(w)ηI,J(Pσw)(δM(lg,σ(w))−δM(lh,σ(w))) dλσ,\begin{equation} D_MQ_i\eta=\sum_{\sigma,I,J}\epsilon_\sigma m_\sigma \int_\sigma b_{\sigma,I,J}(w)\eta_{I,J}(P_\sigma w) \bigl(\delta_M(l_{g,\sigma}(w)) -\delta_M(l_{h,\sigma}(w))\bigr)\,d\lambda_\sigma, \end{equation} where δM(t)=λM(t)−t\delta_M(t)=\lambda_M(t)-t, the cells are fixed and compact, ϵσ∈{1,−1}\epsilon_\sigma\in\{1,-1\} are the finite cover signs, mσm_\sigma are the actual integer coherent-cycle multiplicities, λσ\lambda_\sigma is lattice-normalized volume, and the smooth fixed coefficients bb include the covector pullback minors. The maps PσP_\sigma are the coefficient-coordinate projections, with images in tfi(Vi)t_{f_i}(V_i), and lg,σ,lh,σl_{g,\sigma},l_{h,\sigma} are the finite affine logarithms of the adjoined unit coordinates. The formula uses only the frozen native integration leaves NAG-002389, NAG-002405, NAG-002416 and NAG-002425, on the fixed zero-avoiding support.

Writing s=t+Ms=t+M shows directly that 0≤δM(t)≤(−t−M)+.0\leq\delta_M(t)\leq(-t-M)_+. Indeed it vanishes for s≥0s\geq0, and for s<0s<0 it is (−s)(1−ρ(s+1))(-s)(1-\rho(s+1)). Define Ai(M)=∑σ,I,J∣mσ∣∫σ∣bσ,I,J∣((−lg,σ−M)++(−lh,σ−M)+)dλσ,Ci=∑σ,I,J∣mσ∣∫σ∣bσ,I,J∣dλσ,Hi,a=max⁡σ,w∈σ(−la,σ(w))+(a=g,h).\begin{align*} A_i(M)&=\sum_{\sigma,I,J}|m_\sigma| \int_\sigma |b_{\sigma,I,J}| \bigl((-l_{g,\sigma}-M)_++(-l_{h,\sigma}-M)_+\bigr)d\lambda_\sigma,\\ C_i&=\sum_{\sigma,I,J}|m_\sigma| \int_\sigma |b_{\sigma,I,J}|d\lambda_\sigma,\qquad H_{i,a}=\max_{\sigma,w\in\sigma}(-l_{a,\sigma}(w))_+ \quad(a=g,h). \end{align*} Empty sums have Ci=Hi,g=Hi,h=0C_i=H_{i,g}=H_{i,h}=0. Every integrand in these definitions is continuous on a compact cell. Thus these are legitimate frozen real integrals; no measure extension or cancellation of total variations is needed. Taking absolute values term by term in (I.1.10) proves (I.1.8). If M>HiM>H_i the bound is zero. If M≤HiM\leq H_i, then Ai(M)eaM≤Ci(Hi,g+Hi,h)eaHi<wi.A_i(M)e^{aM}\leq C_i(H_{i,g}+H_{i,h})e^{aH_i}<w_i. Proposition 8.15.13 with ε(M)=e−aM\varepsilon(M)=e^{-aM} gives (I.1.9). The constants use actual multiplicities rather than an unweighted support or a claimed uniform comparison of charts. ◻

Projected fibres, sectors and Gauss lifts.

Let KK be complete for its specified real-valued non-Archimedean absolute value, possibly trivial. All affinoid algebras and domains are actual generalized affinoid objects with their given Banach structures. A finite map is required to have the canonical finite-module topology. Fibre algebras retain their nilpotents. Purity refers to analytic dimension.

Proposition. NAG-002595 (One whole degree is not one cone weight). For every d≥1d\geq1 there is an interior analytic point with a pure dd-dimensional pointed fan and an entire moment-fibre algebra finite with its canonical norm for two projections injective on every top-cone lattice, for which the whole-fibre Gauss degree divided by the index on a specified cone has different values. The intrinsic cone weights do exist in this example; the two whole degrees measure projected sums, not individual cone weights.

Proof. Take a trivially valued field KK, for example Q\mathbb Q with the trivial norm; it is complete. On Z=Spec⁡K[t,t−1,(1−t)−1,z2±1,…,zd±1]Z=\operatorname{Spec}K[t,t^{-1},(1-t)^{-1}, z_2^{\pm1},\ldots,z_d^{\pm1}] use f=(t,1−t,z2,…,zd)f=(t,1-t,z_2,\ldots,z_d). Let xx have the trivial norm on K(t,z2,…,zd)K(t,z_2,\ldots,z_d), and let XX be the affinoid domain where the coordinates of tft_f lie in [−ϵ,ϵ][-\epsilon,\epsilon], with ϵ>0\epsilon>0. These bounds are strict at xx. It is a rational domain in a sufficiently large closed dd-polydisc, with the inverses of tt, 1−t1-t, and the zjz_j imposed by Laurent inequalities. Thus 5.35.3 gives purity of dimension dd. The polydisc coordinates and these inequalities are strict at xx, giving absolute interior. Its moment initials generate K(t,z2,…,zd)K(t,z_2,\ldots,z_d), of transcendence degree dd.

The pointed fan is C=(R≥0(1,0)∪R≥0(0,1)∪R≥0(−1,−1))×Rd−1.C=(\mathbb R_{\geq0}(1,0)\cup\mathbb R_{\geq0}(0,1) \cup\mathbb R_{\geq0}(-1,-1))\times\mathbb R^{d-1}. The ultrametric inequality for t+(1−t)=1t+(1-t)=1 allows exactly these possibilities for the first two norms. Weighted Gauss norms in tt, 1−t1-t, or t−1t^{-1}, together with the zjz_j, realize every one. This is also the graded-relation fan at xx, so 5.55.8 gives this pointed germ for every sufficiently small ordinary neighbourhood. Write σ1,σ2,σ3\sigma_1,\sigma_2,\sigma_3 for its maximal cones.

The entire moment fibre at zero has algebra D=K⟨T,T−1,U,U−1,Z2±1,…,Zd±1⟩/(T+U−1).D=K\langle T,T^{-1},U,U^{-1},Z_2^{\pm1},\ldots,Z_d^{\pm1}\rangle /(T+U-1). For a trivial norm restricted series are finite sums; this is the coordinate ring of ZZ with its trivial Banach norm. No analytic point or component of the fibre is omitted.

For positive integers a,ba,b set Qa,b(u,v,λ)=(au+bv,λ),S=Ta(1−T)b.Q_{a,b}(u,v,\lambda)=(au+bv,\lambda),\qquad S=T^a(1-T)^b. The three saturated cone lattices have indices a,b,a+ba,b,a+b; each restriction of Qa,bQ_{a,b} is injective. The exact target is R=K⟨S±1,Z2±1,…,Zd±1⟩R=K\langle S^{\pm1},Z_2^{\pm1},\ldots,Z_d^{\pm1}\rangle. There is an actual algebra identification D=R[T]/(Ta(1−T)b−S).D=R[T]/(T^a(1-T)^b-S). Indeed the relation makes both TT and 1−T1-T invertible, with T−1=S−1Ta−1(1−T)b,(1−T)−1=S−1Ta(1−T)b−1.T^{-1}=S^{-1}T^{a-1}(1-T)^b,\qquad (1-T)^{-1}=S^{-1}T^a(1-T)^{b-1}. The polynomial is monic up to a unit and has degree a+ba+b. Polynomial division gives the basis 1,T,…,Ta+b−11,T,\ldots,T^{a+b-1}. The given norm and the canonical finite-module norm are both trivial, so this is also a bounded analytic identification. The entire Gauss fibre consequently has degree δa,b=a+b.\delta_{a,b}=a+b.

There are genuine full finite collars as well. Use the same monic presentation over a target polyannulus in (S,Z)(S,Z); it is norm-admissible finite by 2.33.18. If ∣t∣<1|t|<1, then ∣S∣=∣t∣a|S|=|t|^a; if ∣1−t∣<1|1-t|<1, then ∣S∣=∣1−t∣b|S|=|1-t|^b; and if ∣t∣>1|t|>1, then ∣S∣=∣t∣a+b|S|=|t|^{a+b}. The remaining case has both norms one. Hence for a sufficiently small target logarithmic box its entire inverse image lies in XX, with strict defining bounds. The finite analytification theorem 2.33.26, followed by actual domain base change, identifies this full inverse image with the constructed finite analytic algebra. The obstruction is not a failure of norm-admissible finiteness.

On σ1\sigma_1, both Q1,1Q_{1,1} and Q1,2Q_{1,2} have index one; their whole-fibre degrees are 22 and 33. Thus the proposed quotients differ. The actual generic weight of each cone is one: on σ1\sigma_1 its entire fibre is the exact polytorus in (t,z)(t,z), because ∣t∣<1|t|<1 forces ∣1−t∣=1|1-t|=1. Projection to (t,z)(t,z) has degree and index one, and 5.37.1 gives weight one. Use (1−t,z)(1-t,z) on σ2\sigma_2 and (t,z)(t,z) on σ3\sigma_3.

For a small positive target first weight ss, the full projected level has two original moment weights (s/a,0,λ),(0,s/b,λ),(s/a,0,\lambda),\quad (0,s/b,\lambda), with contributions aa and bb. At a negative first weight only σ3\sigma_3 contributes, with index a+ba+b. These explain the whole degree a+ba+b. Thickening by ϵ′h=0\epsilon'^h=0 multiplies all degrees and all cone weights by hh with unchanged support; its generic nilpotent length has not disappeared. ◻

Throughout, KK is complete for its specified real-valued non-Archimedean absolute value, possibly trivial. All affinoids, domains and completed tensors are the actual generalized analytic objects with their given Banach structures. Finite maps must carry the canonical finite-module topology. All fibre algebras keep their nilpotents.

Definition I.1.9. (Two saturated finite collars with common Gauss lifts). Let X=M(A)X=\mathcal M(A) be nonempty and pure of analytic dimension d≥1d\geq1, let f=(f1,…,fn)∈(A×)nf=(f_1,\ldots,f_n)\in(A^\times)^n, and put t=tf=−log⁡∣f∣t=t_f=-\log|f|. Let MM be a finite AA-module. Fix w0∈Rnw_0\in\mathbb R^n and distinct integral maps Qi:Zn→ZdQ_i:\mathbb Z^n\to\mathbb Z^d, i=1,2i=1,2. For each ii, put ui0=Qiw0u_i^0=Q_iw_0 and choose a full-dimensional closed rectangular box PiP_i with ui0u_i^0 in its interior. Assume that Ci=(Qit)−1(Pi)⊂X,ψi:Ci⟶TPiC_i=(Q_i t)^{-1}(P_i)\subset X,\qquad \psi_i:C_i\longrightarrow\mathbb T_{P_i} is norm-admissible finite, where ψi\psi_i is given by the actual units fqi1,…,fqidf^{q_{i1}},\ldots,f^{q_{id}}. The equality defining CiC_i is for the entire inverse image in XX; it is the saturation hypothesis.

Let ηi\eta_i be the target Gauss point at ui0u_i^0. Assume the equality of actual point sets ψ1−1(η1)=ψ2−1(η2)=S={x1,…,xr}⊂t−1(w0),1≤r<∞.\psi_1^{-1}(\eta_1)=\psi_2^{-1}(\eta_2) =S=\{x_1,\ldots,x_r\}\subset t^{-1}(w_0), \qquad 1\leq r<\infty. These are analytic Gauss lifts, not selected lifts after scalar extension. The fibre algebras may have arbitrary finite lengths and residue-field degrees. No isomorphism of the two central fibre algebras is assumed. Require xj∈Int⁡(X/K)x_j\in\operatorname{Int}(X/K) and graded moment rank dd at every xjx_j.

Let CxjC_{x_j} be the pure dd-dimensional pointed fan supplied by 5.55.8, and choose a common rational fan subdivision Σ\Sigma of their union so that each CxjC_{x_j} is a subfan. Assume that both QiQ_i are injective on the real span of every dd-cone of Σ\Sigma. For a top cone σ\sigma, write Nσ=Zn∩span⁡R(σ),ei,σ=[Zd:QiNσ].N_\sigma=\mathbb Z^n\cap\operatorname{span}_{\mathbb R}(\sigma), \qquad e_{i,\sigma}=[\mathbb Z^d:Q_iN_\sigma]. These saturated-lattice indices are finite.

These are hypotheses for a local class, not an existence theorem for such pairs of collars at arbitrary moment points. In particular, separate finite presentations and equal numbers of Gauss lifts do not supply the displayed equality of point sets inside the same XX. The functions and the coefficient module must also be the same original-field data.

Choose pairwise disjoint ordinary open sets OjO_j about xjx_j. After shrinking, choose ordinary affinoid neighbourhoods Vj⊂Oj∩C1∩C2V_j\subset O_j\cap C_1\cap C_2 such that t(Vj)⊂w0+Cxj,\begin{equation} t(V_j)\subset w_0+C_{x_j}, \end{equation} and the pointed-image equality holds for every smaller ordinary neighbourhood of xjx_j. Here the collar inequalities are strict at xjx_j, so C1∩C2C_1\cap C_2 is an ordinary neighbourhood of SS. The stated choice follows from the full-stabilization theorem, continuity of tt, and the ordinary affinoid-neighbourhood basis recorded in 5.52.3: first restrict to a real ball where the upper fan containment holds, then take a smaller ordinary affinoid neighbourhood inside its inverse image. No new pointed-support theorem is being proved.

Lemma. NAG-002621 (Whole projected levels in a common union of neighbourhoods). For each jj, let U1j,U2j⊂VjU_{1j},U_{2j}\subset V_j be ordinary affinoid neighbourhoods of xjx_j. There are ordinary affinoid neighbourhoods Wj⊂U1j∩U2jW_j\subset U_{1j}\cap U_{2j} of xjx_j, open real neighbourhoods Bi⊂int⁡(Pi)B_i\subset\operatorname{int}(P_i) of ui0u_i^0, and finite unions HiH_i of proper integral-normal affine hyperplanes in Rd\mathbb R^d such that {y∈X:Qit(y)=u}⊂⋃j=1rInt⁡top(Wj,X)(u∈Bi∖Hi).\begin{equation} \{y\in X:Q_it(y)=u\} \subset\bigcup_{j=1}^r\operatorname{Int}_{\rm top}(W_j,X) \qquad(u\in B_i\setminus H_i). \end{equation} This includes every analytic point of each projected tropical level. Its intersections with the WjW_j form a disjoint finite affinoid cover and hence give a bounded product decomposition of its actual algebra and its finite coefficient modules.

Proof. Choose the WjW_j using the ordinary affinoid-neighbourhood basis, and set G=⋃jInt⁡top(Wj,X)G=\bigcup_j\operatorname{Int}_{\rm top}(W_j,X). For each ii, the set Ci∖GC_i\setminus G is compact. Its image under ψi\psi_i is closed in the Hausdorff target and misses ηi\eta_i, because the entire analytic fibre there is S⊂GS\subset G. Compactness and the evaluation topology are 2.24.1. Thus Oi′=TPi∖ψi(Ci∖G)O_i'=\mathbb T_{P_i}\setminus\psi_i(C_i\setminus G) is an ordinary open neighbourhood of ηi\eta_i with full inverse image inside GG. Apply the exact reviewed lemma C.1.43 to Oi′O_i'. It puts the entire target polytorus Tu\mathbb T_u inside Oi′O_i' for u∈Bi∖Hiu\in B_i\setminus H_i. Saturation of CiC_i then gives (I.1.12).

The level is an actual affinoid, with algebra Bi(u)=O(Ci)⊗^O(TPi)Ri(u),B_i(u)=\mathcal O(C_i) \widehat\otimes_{\mathcal O(\mathbb T_{P_i})}R_i(u), where Ri(u)=O(Tu)R_i(u)=\mathcal O(\mathbb T_u). It is a canonical finite Ri(u)R_i(u)-module by 2.33.19. Intersections with the WjW_j are actual affinoid domains by 2.36.9 and 2.36.10. They cover by (I.1.12) and are disjoint because Wj⊂OjW_j\subset O_j. Apply 2.38.1 to this finite cover. All pairwise intersections have zero algebra, so its norm-strict degree-zero equalizer is a bounded isomorphism from Bi(u)B_i(u) to the product of the restriction algebras. For a finite module the same theorem gives the corresponding module product. The characteristic idempotents are actual elements of Bi(u)B_i(u), and each factor has the quotient finite-module topology. No reduction or selection of a maximal ideal has occurred. ◻

Theorem. NAG-002624 (Sectorwise full-fibre overlap and projection comparison). Use Definition I.1.9, the VjV_j, and the neighbourhoods Uij,WjU_{ij},W_j of Lemma I.1.10. After shrinking the real neighbourhoods, the following hold.

For each top cone σ\sigma there is a finite union EσE_\sigma of proper affine hyperplanes in the affine plane w0+span⁡Rσw_0+\operatorname{span}_{\mathbb R}\sigma such that, whenever v∈w0+relint⁡σv\in w_0+\operatorname{relint}\sigma is sufficiently close to w0w_0 and v∉Eσv\notin E_\sigma, the whole original moment fibre Fv=t−1(v)F_v=t^{-1}(v) satisfies Fv=∐jFj,v,Fj,v=Fv∩Wj=Fv∩U1j=Fv∩U2j.F_v=\coprod_j F_{j,v},\qquad F_{j,v}=F_v\cap W_j=F_v\cap U_{1j}=F_v\cap U_{2j}. Every nonempty Fj,vF_{j,v} lies in the absolute interior of all three domains in the displayed equality. The equalities are equalities of actual analytic domains, so their full algebras are canonically the same algebra Dj,vD_{j,v}.

For u=Qivu=Q_iv, let Λi(u)\Lambda_i(u) be the finite set of original moment weights on the entire projected level {y∈X:Qit(y)=u}\{y\in X:Q_it(y)=u\}. For every such regular uu, restriction gives a bounded isomorphism with bounded inverse Bi(u) ≃ ∏v′∈Λi(u) ∏j=1rDj,v′.\begin{equation} B_i(u)\ \simeq\ \prod_{v'\in\Lambda_i(u)}\ \prod_{j=1}^r D_{j,v'}. \end{equation} Zero factors are omitted. The corresponding idempotent ei,j,v′e_{i,j,v'} is uniquely determined by the actual labelled domain Fj,v′F_{j,v'}, not by a numerical match of lengths. All original moment weights and all Gauss-lift neighbourhoods are included.

Let Hi(v)H_i(v) be the Gauss field of Ri(Qiv)R_i(Q_iv) and put Ei,j(v)=Dj,v⊗^Ri(Qiv)Hi(v),Li,j(v)=M⊗AEi,j(v).E_{i,j}(v)=D_{j,v} \widehat\otimes_{R_i(Q_iv)}H_i(v),\qquad L_{i,j}(v)=M\otimes_AE_{i,j}(v). Then Dj,vD_{j,v} is norm-admissible finite over both indicated target algebras, and dim⁡H1(v)L1,j(v)e1,σ=dim⁡H2(v)L2,j(v)e2,σ=mM∣Wj,f(v)∈Z≥0.\begin{equation} \frac{\dim_{H_1(v)}L_{1,j}(v)}{e_{1,\sigma}} =\frac{\dim_{H_2(v)}L_{2,j}(v)}{e_{2,\sigma}} =m_{M|W_j,f}(v)\in\mathbb Z_{\geq0}. \end{equation} The same value is obtained using U1jU_{1j} or U2jU_{2j}. An empty factor has value zero.

For either projection the numerator retains the full arithmetic: dim⁡Hi(v)Li,j(v)=∑m∈Max⁡Ei,j(v)length⁡Ei,j(v)mLi,j(v)m [κ(m):Hi(v)].\begin{equation} \dim_{H_i(v)}L_{i,j}(v) =\sum_{\mathfrak m\in\operatorname{Max}E_{i,j}(v)} \operatorname{length}_{E_{i,j}(v)_{\mathfrak m}} L_{i,j}(v)_{\mathfrak m} \,[\kappa(\mathfrak m):H_i(v)]. \end{equation} This includes all local factors, full inseparable degrees and nilpotent lengths. Generic lengths of the restricted module enter through the inherited module-weight formula.

Proof. We give the choices uniformly; there is no choice of an analytic sheet for an individual projection.

Control of all inverse weights. Since QiQ_i is injective on the span of each cone of the finite fan Σ\Sigma, compactness of its intersection with a unit sphere gives ci>0c_i>0 such that ∥Qia∥≥ci∥a∥(a∈∣Σ∣).\|Q_i a\|\geq c_i\|a\| \qquad(a\in|\Sigma|). Use (I.1.11) for every WjW_j and UijU_{ij}. The full-stabilization theorem gives one ρ>0\rho>0 on which all their moment images agree with their respective w0+Cxjw_0+C_{x_j}. Shrink BiB_i to have ∥u−ui0∥<ciρ\|u-u_i^0\|<c_i\rho. By (I.1.12), every point of a projected level is in some WjW_j, and its original moment weight is consequently in w0+∣Σ∣w_0+|\Sigma| and within distance ρ\rho of w0w_0. Thus no distant inverse weight remains in that projected level.

Let JiJ_i contain, for every cone τ\tau of Σ\Sigma of dimension less than dd, an affine hyperplane containing ui0+Qiτu_i^0+Q_i\tau. These hyperplanes are proper. Replace HiH_i by Hi∪JiH_i\cup J_i. For u∈Bi∖Hiu\in B_i\setminus H_i, every inverse weight belongs to the relative interior of a top cone. There is at most one inverse weight on each top cone, by injectivity there. Hence Λi(u)\Lambda_i(u) is finite. For a fixed σ\sigma, pull back both exceptional sets along Qi:w0+span⁡σ→RdQ_i:w_0+\operatorname{span}\sigma\to\mathbb R^d. This map is an affine isomorphism, so the pullbacks are proper hyperplanes. They give EσE_\sigma.

Common whole factors. Take vv as in the statement. Saturation implies that the entire FvF_v belongs to both CiC_i. The retention lemma puts it inside the union of the ordinary interiors of the WjW_j. Because Uij⊂OjU_{ij}\subset O_j and the OjO_j are disjoint, intersecting with UijU_{ij} gives exactly Fv∩WjF_v\cap W_j. These are actual domain identities, not identifications of reduced spectra. The domain universal property and 2.36.9 identify their actual algebras and their restrictions of MM.

For the entire projected level at uu, first split its finite set of moment values Λi(u)\Lambda_i(u), then intersect with the WjW_j. Each exact moment fibre is an affinoid domain obtained by imposing all the equal-radius conditions for the fkf_k. The intersections are affinoid by the actual domain theorems. This is a finite disjoint affinoid cover of the entire projected level. For any v′∈Λi(u)v'\in\Lambda_i(u), retention from this one projection already proves equality of its factors in WjW_j, U1jU_{1j} and U2jU_{2j}. The flat-germ argument below therefore compares both projections on that factor, even if Qkv′Q_kv' lies in the exceptional set for the other projection kk. Apply 2.38.1, exactly as in Lemma I.1.10. It gives (I.1.13) including the module product and the bounded inverse. The characteristic idempotents are unique because the restriction isomorphism fixes every factor. Since Bi(u)B_i(u) is a canonical finite Ri(u)R_i(u)-module, its idempotent factors have the canonical finite Ri(u)R_i(u)-module topology. The finite-module comparisons are 1.18.3 and 1.18.7.

Interior and the two normalized degrees. All points of the projected level have target weights strictly inside PiP_i. They are absolute interior in CiC_i: the target polyannulus bounds are strict, the finite map is relatively interior by 2.33.18, and interior composes by 2.33.21. By the source-factor inclusion and ordinary local testing they are absolute interior in XX. Each point labelled by jj lies in the ordinary interiors of WjW_j, U1jU_{1j} and U2jU_{2j}. The domain interior theorem 2.44.3 and composition then make them absolute interior in these domains as well.

Every nonempty WjW_j and UijU_{ij} is pure of dimension dd by 5.35.3. At vv in the relative interior of σ\sigma, their local moment images are either empty or contained in the affine plane v+(Nσ)Rv+(N_\sigma)_{\mathbb R}, by the chosen fan subdivision and the uniform stabilization radius. Apply 5.40.2 to each nonempty whole fibre in WjW_j and to its restrictions from the two UijU_{ij}. All of these computations use the same Dj,vD_{j,v} and the same module, so that theorem gives (I.1.14). We have not compared two sums with their original-weight labels erased.

Lengths and field normalization. The algebra Ei,j(v)E_{i,j}(v) is finite dimensional over Hi(v)H_i(v) and hence Artinian. Decompose it into its local factors and use a composition series of the finite module in each factor. Every simple quotient is its residue field; taking dimensions proves (I.1.15). For Aj=O(Wj)A_j=\mathcal O(W_j) and Mj=M⊗AAjM_j=M\otimes_AA_j, the generic version is exactly the inherited formula mMj,f(v)=∑p∈Min⁡Ajlength⁡(Aj)p(Mj)p mAj/p,f(v).m_{M_j,f}(v) =\sum_{\mathfrak p\in\operatorname{Min}A_j} \operatorname{length}_{(A_j)_{\mathfrak p}} (M_j)_{\mathfrak p}\, m_{A_j/\mathfrak p,f}(v). Finite tensors are already complete by 1.18.7.

If radii are normalized after a common complete isometric extension and ∣ak∣=e−vk|a_k|=e^{-v_k}, use ∏kakqiℓk\prod_k a_k^{q_{i\ell k}} for the projected unit fqiℓf^{q_{i\ell}}. The idempotents in (I.1.13) base change as actual idempotents; retain every factor into which they split. The finite Gauss-field comparison 5.36.2 and model normalization 5.36.3 preserve the total dimensions and their full length-degree formulas. This is a finite-fibre calculation, with no selected geometric lift and no assertion about scalar extension of an integral. ◻

Proposition. NAG-002629 (Point labels and sector labels are both necessary). The hypotheses of Definition I.1.9 occur in every dimension with r=2r=2, a nonflat three-cone fan, and different nilpotent lengths on the two Gauss lifts. For this example Q1,1Q_{1,1} and Q1,2Q_{1,2} have whole degrees 66 and 99, and even after separation by Gauss lift their degrees are respectively (2,4)(2,4) and (3,6)(3,6). After separation by original moment weight, the normalized coefficients on each cone are (1,2)(1,2) for both projections.

Moreover, on either individual lift there is no analytic idempotent on a full projected neighbourhood of the central Gauss fibre that selects the first positive sector and excludes the second on every sufficiently close positive Gauss fibre. Thus the sector idempotents need not extend across the vertex.

Proof. Use the exact reviewed example I.1.8 over trivially valued K=QK=\mathbb Q: Z=Spec⁡K[t,t−1,(1−t)−1,z2±1,…,zd±1],f=(t,1−t,z2,…,zd).Z=\operatorname{Spec} K[t,t^{-1},(1-t)^{-1},z_2^{\pm1},\ldots,z_d^{\pm1}], \qquad f=(t,1-t,z_2,\ldots,z_d). Take its bounded moment-box affinoid YY about zero and form the disjoint union of the two actual finite thickenings X=Y ∐ M(O(Y)[ϵ]/(ϵ2)).X=Y\ \coprod\ \mathcal M\bigl(\mathcal O(Y)[\epsilon]/(\epsilon^2)\bigr). Use the same moment functions on both components and M=AM=A. The two points x1,x2x_1,x_2 are the trivial generic norms on the two reductions; their thicknesses are h1=1,h2=2h_1=1,h_2=2. The reviewed example proves the pure dimension, interior, full pointed fan and norm-admissible saturated finite collars. Finite products and these canonical thickenings preserve them. The fan on each component is (R≥0(1,0)∪R≥0(0,1)∪R≥0(−1,−1))×Rd−1.\bigl(\mathbb R_{\geq0}(1,0)\cup \mathbb R_{\geq0}(0,1)\cup \mathbb R_{\geq0}(-1,-1)\bigr)\times\mathbb R^{d-1}. For Qa,b(u,v,λ)=(au+bv,λ)Q_{a,b}(u,v,\lambda)=(au+bv,\lambda), a,b>0a,b>0, the whole finite algebra in the reviewed example has degree a+ba+b. Its central analytic Gauss algebra on the jjth component is K(t,z2,…,zd)[ϵ]/(ϵhj)overK(ta(1−t)b,z2,…,zd).K(t,z_2,\ldots,z_d)[\epsilon]/(\epsilon^{h_j}) \quad\hbox{over}\quad K(t^a(1-t)^b,z_2,\ldots,z_d). For hj=1h_j=1 this means the field itself. These fields have their trivial complete norms; the full degree is a+ba+b. Each algebra is local, so the central analytic Gauss fibre consists of precisely xjx_j. Thus both projections have the same full two-point set SS, although their central algebras have different dimensions.

The cone indices are a,b,a+ba,b,a+b. At a small positive first projected weight ss, the two original moment weights are (s/a,0,λ),(0,s/b,λ).(s/a,0,\lambda),\qquad(0,s/b,\lambda). Their Gauss dimensions on component jj are hjah_ja and hjbh_jb, by the reviewed cone-weight calculation and finite thickening. At a negative projected weight only the third cone contributes, with dimension hj(a+b)h_j(a+b). Consequently the full degrees are (h1+h2)(a+b)(h_1+h_2)(a+b), giving 66 and 99, and the point-labelled degrees are as stated. Dividing each sector factor by its own index gives hjh_j, hence (1,2)(1,2) for both projections. Dividing a mixed point-labelled degree by the first cone index still gives 22 versus 33 on the first component. Matching the two point labels by an arbitrary permutation would likewise compare 11 with 22. The same points and the same original weights, rather than counts or degree multisets, fix the correct matching.

Finally fix a component jj and one Qa,bQ_{a,b}. Its central analytic Gauss fibre is the singleton xjx_j. Suppose an analytic idempotent on the full inverse image of an ordinary target neighbourhood of the central Gauss point made the proposed separation for all small s>0s>0. Its value at xjx_j is 00 or 11, and continuity makes it constant on an ordinary neighbourhood of xjx_j. Shrink to an affinoid target neighbourhood to use compactness. The image of the compact complement of that source neighbourhood misses the central Gauss point. The target Gauss points of weights (s,0,…,0)(s,0,\ldots,0) converge to the central Gauss point, so their entire analytic fibres eventually lie in that one constant-idempotent locus. Both positive-sector factors are nonzero: their Gauss dimensions are hja>0h_ja>0 and hjb>0h_jb>0. The idempotent therefore cannot be 11 on one and 00 on the other. This contradiction proves the nonextension assertion. It does not prevent the actual sector idempotents at each nonzero regular projected level. ◻

The ground field KK is complete for its specified real-valued non-Archimedean absolute value, which may be trivial. Every finite analytic map below is required to carry the canonical finite-module norm, up to uniform equivalence. All fibres are actual analytic fibres over KK, with their full algebras and nilpotents. No scalar extension is used in the construction.

Theorem. NAG-002693 (Simultaneous collars after one fixed finite cover). Use Definition C.1.46. Suppose that π:X=M(B)⟶Yε\pi:X=\mathcal M(B)\longrightarrow Y_\varepsilon is norm-admissible finite, that XX is nonempty and pure of dimension dd, and that S=π−1(η)S=\pi^{-1}(\eta) is nonempty. Let f=π∗gf=\pi^*g and let MM be any finite BB-module.

For any two distinct positive arrays a,a′a,a', choose boxes satisfying (2), separately for the two arrays. Then Ca=(Qatf)−1(Pa),Ca′=(Qa′tf)−1(Pa′)C_a=(Q_at_f)^{-1}(P_a),\qquad C_{a'}=(Q_{a'}t_f)^{-1}(P_{a'}) are saturated norm-admissible finite collars in the same XX. Their entire central analytic Gauss fibres are the same finite point set S⊂tf−1(0)S\subset t_f^{-1}(0). Every x∈Sx\in S is absolutely interior and has graded moment rank dd.

The full pointed fan at every x∈Sx\in S is F=F1×⋯×Fp×Rk,(7)F=F_1\times\cdots\times F_p\times\mathbb R^k, \tag{7} where FℓF_\ell is (5). The same rational subdivision of Rk\mathbb R^k may be used for both projections. On each top cone σ\sigma the projections are injective on its span, and their saturated-lattice indices are ea,σ=[Zd:QaNσ]=∏ℓ=1p{aℓj,σ uses the puncture ray j,Aℓ,σ uses the last ray.Nσ=Zn∩span⁡Rσ.(8)e_{a,\sigma}=[\mathbb Z^d:Q_aN_\sigma] =\prod_{\ell=1}^p \begin{cases} a_{\ell j},&\sigma\hbox{ uses the puncture ray }j,\\ A_\ell,&\sigma\hbox{ uses the last ray}. \end{cases} \qquad N_\sigma=\mathbb Z^n\cap\operatorname{span}_{\mathbb R}\sigma . \tag{8} Thus these data satisfy all hypotheses of the independently reviewed definition I.1.9 and theorem I.1.11, including its common entire Gauss set and saturation requirements. If some mℓ≥2m_\ell\geq2, FF is not a dd-plane.

Proof. Write Ya=qa−1(TPa)Y_a=q_a^{-1}(\mathbb T_{P_a}). Lemma C.1.47 gives Ya⊂YεY_a\subset Y_\varepsilon as the full inverse image, with its actual finite algebra. Domain base change and finite base change, 2.36.9 and 2.33.19, identify Ca=π−1(Ya),O(Ca)=B⊗O(Yε)O(Ya).C_a=\pi^{-1}(Y_a),\qquad \mathcal O(C_a)= B\otimes_{\mathcal O(Y_\varepsilon)}\mathcal O(Y_a). The tensor is already complete with its canonical finite norm. The map Ca→YaC_a\to Y_a is finite, as is Ya→TPaY_a\to\mathbb T_{P_a}. Composing finite free quotient presentations proves finiteness with the canonical norm for the composite. This construction also proves the displayed saturation equality for all of XX. The same argument applies to a′a'.

By (3), a point of CaC_a maps to the central target Gauss point if and only if its image under π\pi is η\eta. The entire fibre is therefore exactly SS, independently of aa. Its finiteness follows from the actual finite-dimensional algebra over H(η)\mathcal H(\eta); all local factors are kept. The map π\pi is everywhere relatively interior by 2.33.18. Absolute interiority of η\eta and the composition inclusion 2.33.21 give absolute interiority of every x∈Sx\in S.

We identify the fan before transferring it to XX. The norm calculation in the preceding lemma gives global containment tg(Zan)⊂Ft_g(Z^{\rm an})\subset F. Every vector of FF sufficiently near 00 is realized as follows. For a puncture ray of size s≥0s\geq0, use the Gauss norm in tℓ−cℓjt_\ell-c_{\ell j} of radius e−se^{-s}. For the last ray use the Gauss norm in tℓt_\ell of radius ese^s. Give each zbz_b its prescribed Gauss radius. Take the simultaneous Gauss norm in these dd independent variables. These points have precisely the stated moments. As all parameters tend to zero, every such family tends to η\eta. Indeed at radius 11 translation by an element of K∘K^\circ preserves the unit Gauss norm, and polynomial Gauss evaluations are continuous in the radii. The same is true for the inverses of the puncture coordinates, hence, by uniform approximation, for the affinoid functions on a fixed surrounding domain. There are only finitely many choices of rays. Every ordinary neighbourhood of η\eta therefore realizes every sufficiently small vector in FF. The frozen full-stabilization theorem identifies its pointed fan with FF.

For x∈Sx\in S, the isometric field embedding H(η)↪H(x)\mathcal H(\eta)\hookrightarrow\mathcal H(x) induces an injective map of graded residue fields. The graded relations of the initial forms of f(x)f(x) over K~gr\widetilde K^{\rm gr} are consequently exactly the relations of g(η)g(\eta), not a larger ideal. The initial forms of tℓ−cℓ1t_\ell-c_{\ell1} and zbz_b are algebraically independent, and all other initial forms are obtained from them by the displayed translations. Thus the graded moment rank is dd. In 5.55.5 the fan is defined from precisely this homogeneous relation ideal. Together with 5.55.8, this proves that every CxC_x equals FF, even when XX is nonreduced.

For a top cone, its saturated tangent lattice is the direct sum of the primitive ray lattice of each block and the full Zk\mathbb Z^k. Each primitive block vector maps to aℓja_{\ell j} or −Aℓ-A_\ell times the corresponding target basis vector. The matrix is diagonal, with the identity on the last kk directions. This proves injectivity and (8). If mℓ≥2m_\ell\geq2, the span of FℓF_\ell has dimension mℓ>1m_\ell>1, while that fan has dimension one. Thus the span of FF has dimension greater than dd, and FF is not a dd-plane.

We have verified, individually, purity, actual units and module, both saturated finite collars, the same full finite Gauss set, absolute interiority, graded rank, and simultaneous genericity on every top cone. These are exactly the hypotheses of the reviewed sector comparison. ◻

Fix a complete non-Archimedean field KK with its specified real-valued absolute value, possibly trivial. All affinoids are generalized affinoids, with arbitrary positive real presentation radii. Finite maps are norm-admissible finite: their given Banach structures are equivalent to the canonical finite-module structures. All domains, fibres and modules below have their actual analytic structures; nilpotents are retained. Write tg=−log⁡∣g∣t_g=-\log|g| for a tuple of analytic units. For a closed rectangular box P⊂RmP\subset\mathbb R^m, TP\mathbb T_P is the entire inverse image of PP in the analytic mm-torus. The point ηs\eta_s is its weighted Gauss point at s∈Ps\in P.

For a point xx and an actual unit tuple ff, write rf(x)=trdeg⁡K~grK~gr(f1(x)~,…,fn(x)~).r_f(x)=\operatorname{trdeg}_{\widetilde K^{\rm gr}} \widetilde K^{\rm gr} \bigl(\widetilde{f_1(x)},\ldots,\widetilde{f_n(x)}\bigr). The full pointed fan CxC_x is the one supplied by 5.55.8. It is pure of dimension rf(x)r_f(x).

Lemma. NAG-002711 (Interior Gauss recognition under a finite fan projection). Let ZZ be a generalized affinoid, let x∈Int⁡(Z/K)x\in\operatorname{Int}(Z/K), and let ff be an actual unit tuple with tf(x)=wt_f(x)=w. Suppose that CxC_x is contained in the support of a finite pure dd-dimensional rational fan Σ⊂Rn\Sigma\subset\mathbb R^n, where d≥1d\geq1. Let Q:Zn→ZdQ:\mathbb Z^n\to\mathbb Z^d be integral and injective on the real span of every top cone of Σ\Sigma. For the actual monomial tuple g=fQg=f^Q, g(x)=ηQw⟺rf(x)=d.(7)g(x)=\eta_{Qw} \quad\Longleftrightarrow\quad r_f(x)=d. \tag{7} This is an equality criterion for analytic points of the target torus. It does not identify the different projected fibre algebras with one another.

Proof. First, for any dd-tuple gg, the target point g(x)g(x) is the weighted Gauss point at tg(x)t_g(x) if and only if its dd initial forms are algebraically independent over K~gr\widetilde K^{\rm gr}. Here is the original-field argument. For a Laurent polynomial ∑νaνTν\sum_\nu a_\nu T^\nu, put R=max⁡ν∣aν∣∏j∣gj(x)∣νj.R=\max_\nu |a_\nu|\prod_j|g_j(x)|^{\nu_j}. The terms attaining RR define a nonzero homogeneous polynomial of degree RR in the graded variables. Independent initial forms make its value nonzero; consequently the evaluated polynomial has norm exactly RR. These equalities identify the seminorm with the weighted Gauss seminorm. Conversely, lift the finitely many homogeneous coefficients of a proposed graded relation to KK. A nonzero homogeneous component of that relation has one common weighted degree. Its lifted Laurent polynomial has that Gauss norm, whereas the relation would force a strictly smaller evaluated norm. This is impossible at a Gauss point. This argument uses neither realizability of the radii in ∣K×∣|K^\times| nor a change of the given absolute value.

The monomial initial forms of gg belong to the graded field generated by those of ff. Hence rg(x)≤rf(x)r_g(x)\leq r_f(x). The containment Cx⊂∣Σ∣C_x\subset|\Sigma| and the full pointed-fan theorem give rf(x)=dim⁡Cx≤dr_f(x)=\dim C_x\leq d. If g(x)g(x) is Gauss, the preceding criterion gives rg(x)=dr_g(x)=d, and therefore rf(x)=dr_f(x)=d.

Now suppose rf(x)=dr_f(x)=d. Choose a dd-cone τ\tau of the pointed fan. Its intersections with the finitely many cones of Σ\Sigma cover it. At least one such intersection has dimension dd: a finite union of proper linear subspaces cannot contain the relative interior of a dd-cone. Thus some dd-dimensional subcone τ′⊂τ\tau'\subset\tau lies in a top cone of Σ\Sigma. The map QQ is injective on its span, so Qτ′Q\tau' has dimension dd.

For every sufficiently small ordinary neighbourhood WW of xx, pointed stabilization supplies all sufficiently small vectors of w+τ′w+\tau' in tf(W)t_f(W). Their images occur in tg(W)=Qtf(W)t_g(W)=Qt_f(W). Applying full pointed stabilization also to gg, and choosing WW small enough for both tuples, shows that its pointed image fan contains a dd-dimensional conical germ. Its dimension is rg(x)r_g(x); hence rg(x)≥dr_g(x)\geq d. There are only dd generators, so rg(x)=dr_g(x)=d. The Gauss criterion proved in the first paragraph completes the proof. This uses a dimension lower bound from the actual neighbourhood images, not a claim that arbitrary projection commutes with an uncontrolled image germ. ◻

Proposition. NAG-002699 (A nonflat pointed germ with no finite saturated box). Let K=QK=\mathbb Q with its trivial absolute value, d≥1d\geq1, and use one puncture block 0,10,1, with k=d−1k=d-1. Fix ε>0\varepsilon>0 and 0<r<10<r<1, and define the affinoid domain Xr=Yε∩{∣t−2∣≥r},f=(t,t−1,z1,…,zd−1).X_r=Y_\varepsilon\cap\{|t-2|\geq r\},\qquad f=(t,t-1,z_1,\ldots,z_{d-1}). Let x=ηx=\eta. Then XrX_r is pure of dimension dd, x∈Int⁡(Xr/K)x\in\operatorname{Int}(X_r/K), and the pointed moment fan at xx is the nonflat three-ray fan times Rd−1\mathbb R^{d-1}, with graded rank dd.

For every a,b>0a,b>0, the map ψa,b=(ta(t−1)b,z1,…,zd−1)\psi_{a,b}=(t^a(t-1)^b,z_1,\ldots,z_{d-1}) is finite at xx in the ordinary-neighbourhood sense. Its full fibre over the target unit Gauss point is {x}\{x\}. Nevertheless, for no full-dimensional closed rectangular box PP with 0∈int⁡(P)0\in\operatorname{int}(P) is the saturated map (Qa,btf)−1(P)⟶TP(11)(Q_{a,b}t_f)^{-1}(P)\longrightarrow\mathbb T_P \tag{11} norm-admissible finite. This failure already occurs separately for Q1,1Q_{1,1} and Q1,2Q_{1,2}; hence the same fan, common analytic Gauss set, interior moment point, and ordinary local finiteness do not imply the two finite-collar hypotheses.

Proof. The lower bound on t−2t-2 is a Laurent-domain condition. Domain purity gives the asserted dimension. At xx its inequality is strict, since ∣t−2∣(x)=1>r|t-2|(x)=1>r; all original moment bounds are also strict. Adjoining the inverse of t−2t-2 with radius r−1>1r^{-1}>1 to the presentation in Lemma C.1.47 witnesses absolute interiority of xx. As XrX_r is an ordinary neighbourhood of xx in YεY_\varepsilon, full pointed stabilization and Theorem I.1.13 with the identity cover give the same rank and fan. The fibre assertion follows from (3), since xx remains in the domain.

For completeness, ordinary local finiteness can be exhibited using only the frozen local ingredients. Choose a small box P0P_0 satisfying (2), and let Y0=qa,b−1(TP0)⊂YεY_0=q_{a,b}^{-1}(\mathbb T_{P_0})\subset Y_\varepsilon. It is compact and finite over the full polyannulus. Inside Y0Y_0, the open set G={∣t−2∣>r}G=\{|t-2|>r\} contains its entire central analytic Gauss fibre {x}\{x\}. The image of Y0∖GY_0\setminus G is compact, hence closed in the Hausdorff target, and misses the target Gauss point. Choose an ordinary affinoid target neighbourhood VV of that point, inside the complementary open set and the ordinary interior of TP0\mathbb T_{P_0}. The ordinary affinoid-neighbourhood basis is the one used in 5.52.3. Finite base change gives qa,b−1(V)→Vq_{a,b}^{-1}(V)\to V finite, and its full inverse image lies in G⊂XrG\subset X_r. It is an ordinary neighbourhood of xx. The domain-interior and composition formulae (with ordinary openness for this finite free map) identify the ordinary interior of qa,b−1(V)q_{a,b}^{-1}(V) in Y0Y_0 with the inverse image of the ordinary interior of VV in TP0\mathbb T_{P_0}: compute relative interior over TP0\mathbb T_{P_0} using the two finite factorizations. Since VV lies in the ordinary interior of that polyannulus, these are also ordinary open neighbourhoods in the ambient source and target. Restricting gives a finite map between them, as required by finiteness at xx. Only this ordinary target neighbourhood, not a tropical box, has been constructed.

Suppose (11) were norm-admissible finite. Restrict its target to a sufficiently small closed box about zero. Finite base change preserves the supposed property. We may therefore assume that PP satisfies (2). Set YP=qa,b−1(TP),C=Xr∩YP.Y_P=q_{a,b}^{-1}(\mathbb T_P),\qquad C=X_r\cap Y_P. Then CC is an affinoid domain of YPY_P, and is the entire inverse image in (11).

For 0<ρ<10<\rho<1, let yρy_\rho be the simultaneous Gauss point in the coordinates (t−2,z1,…,zd−1)(t-2,z_1,\ldots,z_{d-1}), of radii (ρ,1,…,1)(\rho,1,\ldots,1). The trivial ground norm gives ∣t(yρ)∣=∣t(yρ)−1∣=∣zb(yρ)∣=1,∣t(yρ)−2∣=ρ.|t(y_\rho)|=|t(y_\rho)-1|=|z_b(y_\rho)|=1,\qquad |t(y_\rho)-2|=\rho. Every yρy_\rho lies in YPY_P and has original and projected weight zero. The map ρ↦yρ\rho\mapsto y_\rho is continuous: polynomial Gauss norms are continuous maxima, the puncture inverses have norm 11, and uniform approximation in an ambient affinoid presentation handles every analytic function. But yρ∈Cy_\rho\in C exactly when ρ≥r\rho\geq r. Thus yr∈Cy_r\in C is not in the ordinary interior of CC inside YPY_P.

The supposed norm-admissible finiteness of C→TPC\to\mathbb T_P makes every point relatively interior, by 2.33.18. Apply the frozen composition identity (using the source-factor implication) to C↪YP→TPC\hookrightarrow Y_P\to\mathbb T_P: Int⁡(C/TP)=Int⁡(C/YP)∩(C↪YP)−1Int⁡(YP/TP).\operatorname{Int}(C/\mathbb T_P) =\operatorname{Int}(C/Y_P)\cap (C\hookrightarrow Y_P)^{-1}\operatorname{Int}(Y_P/\mathbb T_P). It forces yr∈Int⁡(C/YP)y_r\in\operatorname{Int}(C/Y_P). The domain-interior theorem 2.44.3 says this is exactly ordinary interior, contradicting the preceding path. This proves the obstruction. ◻

The end cycle of a finite balanced graph

Use the finite graph datum of definition C.1.22, the cylinder description of lemma C.1.23, and the interleaved convention of lattice normalization. The smooth real profile is the one used in the compact-cutoff proof. Write N=ZrN=\mathbb Z^r, E=NR×RE=N_{\mathbb R}\times\mathbb R, with coordinates (y,t)(y,t), projection π:E→NR\pi:E\to N_{\mathbb R}, and ambient lattice N⊕ZN\oplus\mathbb Z. A polyhedron below is closed and given by finitely many affine linear inequalities, with rational directions and arbitrary real constants. For its tangent space LσL_\sigma, its lattice is the saturated lattice Nσ=Lσ∩(N⊕Z)N_\sigma=L_\sigma\cap(N\oplus\mathbb Z).

After a common subdivision of the polytopes QσQ_\sigma, define the weighted end chain D∞(C)=∑σ unbounded, dim⁡σ=nmσ[Qσ].(2.2)D_{\infty}(C)=\sum_{\sigma\ {\rm unbounded},\,\dim\sigma=n} m_\sigma[Q_\sigma]. \tag{2.2} Coincident projected pieces are added with all their weights. Equation (2.1) fixes the vertical index at one; degrees already present in mσm_\sigma must still be retained.

Let W⊂NRW\subset N_{\mathbb R} be open. Assume that the restriction of CC to π−1W\pi^{-1}W is a balanced local nn-cycle, with the indicated integer weights, exactly as in NAG-002384. This means that every relevant codimension-one normal sum is tangent to its facet; it includes all finite lower and side faces meeting this open set.

Theorem. NAG-002563 (Weak end-cycle identity; F8LA-P3R). For every η∈Acn−1,n−1(W)\eta\in\mathcal A_c^{n-1,n-1}(W), extended by zero outside WW, IC((−t)d′d′′π∗η)=ID∞(C)(η).(3.1)I_C\bigl((-t)d'd''\pi^*\eta\bigr) =I_{D_{\infty}(C)}(\eta). \tag{3.1} The left side is absolutely convergent in the cell sense of definition C.1.22.

Choose ρ∈Cc∞((0,1))\rho\in C_c^\infty((0,1)), ρ≥0\rho\geq0, ∫ρ=1\int\rho=1, and define hR(t)=−t+∫R(t−R−s)+ρ(s) ds.(3.2)h_R(t)=-t+\int_{\mathbb R}(t-R-s)_+\rho(s)\,ds. \tag{3.2} Then hR=−th_R=-t for t≤Rt\leq R, hR=−R−μh_R=-R-\mu for t≥R+1t\geq R+1, where μ=∫sρ(s) ds\mu=\int s\rho(s)\,ds, and hR′′=ρ(t−R)h_R''=\rho(t-R). For R≥HR\geq H, IC(d′d′′hR∧π∗η)=ID∞(C)(η).(3.3)I_C\bigl(d'd''h_R\wedge\pi^*\eta\bigr) =I_{D_{\infty}(C)}(\eta). \tag{3.3} Thus these pairings stabilize exactly, on this prescribed class of tests.

Proof. For completeness, the smooth-profile construction in the proof of NAG-002427 gives E(t)=e−1/tE(t)=e^{-1/t} for t>0t>0 and E(t)=0E(t)=0 otherwise. One may take θ(t)=1−E(t)/(E(t)+E(1−t))\theta(t)=1-E(t)/(E(t)+E(1-t)). To obtain a density with support strictly inside (0,1)(0,1), normalize E(t−14)E(34−t)E(t-\tfrac14)E(\tfrac34-t) by its positive continuous integral. These formulas supply the θ\theta and ρ\rho used here. The integral in (3.2) has a continuous, compactly supported coefficient; the established one-dimensional integral and fundamental theorem give the stated derivatives.

Put β=π∗η\beta=\pi^*\eta. Since d′d′′β=π∗(d′d′′η)d'd''\beta=\pi^*(d'd''\eta) has bidegree (n,n)(n,n), theorem C.1.24 proves absolute convergence in (3.1).

Choose θ∈C∞(R)\theta\in C^\infty(\mathbb R) equal to one on (−∞,0](-\infty,0] and zero on [1,∞)[1,\infty), and put wR(t)=−tθ(t−R)w_R(t)=-t\theta(t-R). The forms wRd′′βw_Rd''\beta and d′wR∧βd'w_R\wedge\beta have compact support on the local cycle over WW: horizontally their supports lie in the fixed compact support of η\eta, and vertically they lie in [a,R+1][a,R+1]. Two applications of Theorem 5.46.3 give 0=IC(d′(wRd′′β))=IC(d′wR∧d′′β)+IC(wRd′d′′β),0=IC(d′′(d′wR∧β))=−IC(d′d′′wR∧β)−IC(d′wR∧d′′β).\begin{align*} 0&=I_C\bigl(d'(w_Rd''\beta)\bigr) =I_C(d'w_R\wedge d''\beta)+I_C(w_Rd'd''\beta),\\ 0&=I_C\bigl(d''(d'w_R\wedge\beta)\bigr) =-I_C(d'd''w_R\wedge\beta)-I_C(d'w_R\wedge d''\beta). \end{align*} Consequently IC(wRd′d′′β)=IC(d′d′′wR∧β).(3.4)I_C(w_Rd'd''\beta)=I_C(d'd''w_R\wedge\beta). \tag{3.4} For R≥HR\geq H, the left side is the left side of (3.1). The right side is supported in the cylindrical region R≤t≤R+1R\leq t\leq R+1. On a cylinder choose an integral basis of NQσN_{Q_\sigma} followed by ete_t. The product lattice has covolume one. Moving the pair d′t∧d′′td't\wedge d''t past the even-degree horizontal top form gives sign +1+1. Iterated real integration therefore yields IC(d′d′′wR∧β)=(∫RR+1wR′′(t) dt)ID∞(C)(η)=ID∞(C)(η),I_C(d'd''w_R\wedge\beta) =\left(\int_R^{R+1}w_R''(t)\,dt\right) I_{D_{\infty}(C)}(\eta) = I_{D_{\infty}(C)}(\eta), because wR′(R)=−1w_R'(R)=-1 and wR′(R+1)=0w_R'(R+1)=0. This proves (3.1). Replacing wR′′w_R'' by hR′′=ρ(t−R)h_R''=\rho(t-R), whose integral is also one, proves (3.3). ◻

The ordinary tube and its absolute-boundary image

The ordinary image-tube scope also applies over each actual Gauss residue field used below. The dimension calculation uses the valued-field tower formula and the equality of completed residue fields in the actual boundary cover.

Lemma. NAG-002795 (Uniform tube and boundary defect; F8RC-T). Let V=M(A)V=\mathcal M(A) be affinoid, f∈Af\in A, and let the nonempty actual zero scheme D=M(A/(f))D=\mathcal M(A/(f)) have dimension at most dd. For any unit tuple gg, put P0=tg(D)P_0=t_g(D). There are ρ>0\rho>0 and a finite integral affine polyhedral set Δ⊂P0\Delta\subset P_0, with dim⁡Δ≤d−1\dim\Delta\leq d-1, such that tg({∣f∣<ρ})⊂P0,tg(∂KV∩{∣f∣<ρ})⊂Δ.\begin{align} t_g(\{|f|<\rho\})&\subset P_0,\\ t_g(\partial_KV\cap\{|f|<\rho\})&\subset\Delta. \end{align} Here Δ=∅\Delta=\varnothing if d=0d=0. If DD is empty, a sufficiently small tube is empty. The same assertions hold over any specified complete base field and with arbitrary radii.

Proof. By NAG-002358, P0P_0 is a compact finite integral affine polyhedral set of dimension at most dd. The inverse image of each compact cell is an actual affinoid Laurent domain: integral affine inequalities become bounds on monomials in the units gig_i, with positive real radii. Their finite union W0=tg−1(P0)W_0=t_g^{-1}(P_0) is an analytic domain, obtained by gluing its affinoid pieces along their actual intersections. It contains the closed subspace DD. Lemma H.0.3 gives an ordinary neighbourhood OO of DD contained in W0W_0. On the compact complement V∖OV\setminus O, ∣f∣|f| has a positive minimum, if the complement is nonempty. This proves (I.1.16) for a sufficiently small positive ρ\rho.

Use NAG-002356 to write the entire absolute boundary as a finite union of actual compact analytic fibres VjV_j over fields HjH_j, where d(Hj/K)=1d(H_j/K)=1 and the maps Vj→VV_j\to V preserve completed residue fields. Let DjD_j be the actual zero scheme of f∣Vjf|_{V_j}. Its points are exactly D∩VjD\cap V_j. For such a point zz, the tower formula and the maximum dimension bound give dHj(z)=dK(z)−1≤d−1.d_{H_j}(z)=d_K(z)-1\leq d-1. Consequently dim⁡HjDj≤d−1\dim_{H_j}D_j\leq d-1, by NAG-001591. If d=0d=0, no such point exists. Each nonempty DjD_j has a finite polyhedral image Δj=tg(Dj)\Delta_j=t_g(D_j) of dimension at most d−1d-1, by NAG-002358 over HjH_j.

Apply the first paragraph over HjH_j, with Vj,DjV_j,D_j, to get tg(Vj∩{∣f∣<ρj})⊂Δjt_g(V_j\cap\{|f|<\rho_j\})\subset\Delta_j. If DjD_j is empty, compactness gives a positive minimum of ∣f∣|f| on VjV_j, so its small tube is empty. Shrink ρ\rho to work for all finitely many jj, and take Δ=⋃jΔj\Delta=\bigcup_j\Delta_j. The maps preserve the absolute values of the original units, so this is a subset of P0P_0. It gives (I.1.17) for the actual absolute boundary, not merely for a polyhedral boundary of a chosen chart. ◻

Example. NAG-002484 (A higher-dimensional local application). Let AA be a generalized KK-affinoid algebra and let X=M(A{r1−1T1,…,rm−1Tm}),rj>0,X=\mathcal M\bigl(A\{r_1^{-1}T_1,\ldots,r_m^{-1}T_m\}\bigr), \qquad r_j>0, be good and pure of analytic dimension nn, and assume that each coordinate hyperplane Yj={Tj=0}Y_j=\{T_j=0\} used below has analytic dimension at most n−1n-1. Evaluation at Tj=0T_j=0 and the inclusion of series independent of TjT_j are bounded algebra maps whose composition is the identity. The actual quotient by (Tj)(T_j) is the corresponding algebra in the other variables, by its power-series coefficients and Corollary 1.18.4. The induced spectral maps give the closed immersion Yj↪XY_j\hookrightarrow X and retraction X→YjX\to Y_j. For an analytic unit uu and integers aja_j, choose g=u∏aj>0Tjaj,h=∏aj<0Tj−aj.g=u\prod_{a_j>0}T_j^{a_j},\qquad h=\prod_{a_j<0}T_j^{-a_j}. Assume these chosen g,hg,h satisfy the local non-zero-divisor hypotheses of Theorem G.0.15. Their zero and pole set is contained in the union of the coordinate hyperplanes. At every point of that set, choose one coordinate hyperplane containing it. Its coordinate retraction supplies (Rn−1)(\mathrm R_{n-1}). Thus the theorem gives the logarithmic current and the uniform stationarity on each top-degree presentation. The original absolute value and arbitrary positive radii are retained.

Fix a complete field KK with its specified real-valued non-Archimedean absolute value. Trivial valuation and generalized affinoid radii are allowed. Spaces in this supplement are good and Hausdorff. Coherent modules live on the analytic GG-site. A line bundle means a coherent module which is free of rank one on ordinary open neighbourhoods. The symbol q∗q^* on coherent modules denotes ordinary tensor pullback; it is not a derived functor.

Proposition. NAG-002525 (Coherent tensor projection). Let q:Y→Xq:Y\to X be norm-admissible and finite. For F∈Coh⁡(YG)\mathcal F\in\operatorname{Coh}(Y_G) and E∈Coh⁡(XG)\mathcal E\in\operatorname{Coh}(X_G) there is a canonical isomorphism q∗(F⊗OYq∗E)≃q∗F⊗OXE.\begin{equation} q_*(\mathcal F\otimes_{\mathcal O_Y}q^*\mathcal E) \simeq q_*\mathcal F\otimes_{\mathcal O_X}\mathcal E. \end{equation} It commutes with affinoid-domain restriction and analytic base change. Neither F\mathcal F nor E\mathcal E is required to be locally free. If X,YX,Y are pure of the same analytic dimension dd and α∈AX,cd,d(X)\alpha\in\mathcal A^{d,d}_{X,c}(X) is an ordinary smooth moment form, then ∫Y/K, F⊗q∗Eq∗α=∫X/K, q∗F⊗Eα.\begin{equation} \int_{Y/K,\,\mathcal F\otimes q^*\mathcal E}q^*\alpha =\int_{X/K,\,q_*\mathcal F\otimes\mathcal E}\alpha. \end{equation} Both integrals use the fixed ground field KK.

Proof. On V=M(A)V=\mathcal M(A) and its full inverse image M(B)\mathcal M(B), let MM and NN represent E\mathcal E and F\mathcal F. The map N⊗B(B⊗AM)⟶N⊗AM,n⊗(b⊗m)⟼bn⊗mN\otimes_B(B\otimes_A M)\longrightarrow N\otimes_A M,\qquad n\otimes(b\otimes m)\longmapsto bn\otimes m has inverse n⊗m↦n⊗(1⊗m)n\otimes m\mapsto n\otimes(1\otimes m). These are maps of finite AA-modules, so they are bounded, with bounded inverses, for the canonical norms. The finite-tensor theorem identifies the modules with the analytic completed constructions.

For a bounded affinoid base change A→CA\to C, put D=B⊗ACD=B\otimes_A C. The same formulas identify both base changes with N⊗AM⊗ACN\otimes_A M\otimes_A C. In particular they respect domain restrictions. The finite coherent direct-image theorem and affinoid coherent equivalence therefore glue them to (I.1.18). This argument concerns an isomorphism of tensors, and asserts no exactness for tensoring with an arbitrary E\mathcal E. Finally apply the fixed-field finite projection theorem to the coherent module F⊗q∗E\mathcal F\otimes q^*\mathcal E. It supplies compact support of q∗αq^*\alpha and the integral equality with coefficient q∗(F⊗q∗E)q_*(\mathcal F\otimes q^*\mathcal E). Substitute (I.1.18). ◻

Fix a complete field KK with its specified real-valued non-Archimedean absolute value. Trivial valuation and arbitrary positive affinoid radii are allowed. All analytic spaces below are good and Hausdorff. A finite morphism is norm-admissible finite: full affinoid inverse images have finite algebras with their canonical finite-module topology. No algebra is replaced by its reduction. Line bundles are coherent modules with frames on ordinary open neighbourhoods.

Let q:Y→Xq:Y\to X be finite, and assume that E=q∗OYE=q_*\mathcal O_Y is locally free of constant rank n≥1n\ge1 on ordinary opens. Fix an analytic line bundle M\mathcal M on XX with a continuous metric. The working class upstairs consists of the specified pullback L=q∗M\mathcal L=q^*\mathcal M with its pulled-back metric. For a frame ee of M\mathcal M, put ϕe=−log⁡∥e∥\phi_e=-\log\|e\|. Call the metric ordinary moment-smooth when these weights belong to AX0,0\mathcal A_X^{0,0} in the frozen ordinary smooth moment sheaf. This condition is independent of the frame: ϕue=ϕe−log⁡∣u∣\phi_{ue}=\phi_e-\log|u|, and the logarithm of an analytic unit is a smooth moment function.

The reviewed determinant projection isomorphism is ιq:Nm⁡q(q∗M):=det⁡(q∗q∗M)⊗det⁡(E)∨→ ∼ M⊗n.\begin{equation} \iota_q:\operatorname{Nm}_q(q^*\mathcal M) :=\det(q_*q^*\mathcal M)\otimes\det(E)^\vee \xrightarrow{\ \sim\ }\mathcal M^{\otimes n}. \end{equation} Transport the tensor-power metric through ιq\iota_q. This defines what this note calls the projection-compatible norm metric on this class. It is specified by the downstairs metrized line and its pullback presentation. No construction for an arbitrary metrized line upstairs is presumed.

For a section ss of q∗Mq^*\mathcal M on a full inverse image q−1Uq^{-1}U, the map a↦asa\mapsto as from E∣UE|_U to (q∗q∗M)∣U(q_*q^*\mathcal M)|_U gives a determinant section Nm⁡q(s)\operatorname{Nm}_q(s). If ee is a frame on UU and s=b q∗es=b\,q^*e, its coefficient under (I.1.20) is ιq(Nm⁡q(s))=N(b)e⊗n,N(b)=det⁡(mb:E∣U→E∣U).\begin{equation} \iota_q(\operatorname{Nm}_q(s))=N(b)e^{\otimes n},\qquad N(b)=\det(m_b:E|_U\to E|_U). \end{equation} The useful link is between the vanishing order of this coefficient and the ordinary local intersection lengths of ss.

Theorem. NAG-002554 (Compatibility on the pulled-back metric class). Use the finite locally free analytic map and the metrized line M\mathcal M of Section 1.

  1. The projection-compatible metric on Nm⁡q(q∗M)\operatorname{Nm}_q(q^*\mathcal M) is continuous. If the metric on M\mathcal M is ordinary moment-smooth, so is this metric. In the norm frame corresponding to e⊗ne^{\otimes n}, its weight is exactly nϕen\phi_e.

  2. For s=b q∗es=b\,q^*e, the determinant section satisfies (I.1.21). Wherever N(b)N(b) is an analytic unit, −log⁡∥Nm⁡q(s)∥=−log⁡∣N(b)∣+nϕe.\begin{equation} -\log\|\operatorname{Nm}_q(s)\|=-\log|N(b)|+n\phi_e. \end{equation} In the smooth case this is an ordinary smooth moment function on that locus. The line, section, metric and this formula commute with analytic base change between good Hausdorff spaces, with the full finite inverse image and its actual norm retained.

  3. Fix x∈Ux\in U such that the ordinary local ring A=OX,xA=\mathcal O_{X,x} is a DVR. Write k=κ(x)k=\kappa(x) for its ordinary residue field. Suppose the coefficient germ by∈OY,yb_y\in\mathcal O_{Y,y} is a non-zero-divisor for every y∈q−1(x)y\in q^{-1}(x). Then the local Cartier multiplicity of the determinant section is ord⁡xNm⁡q(s)=∑y∈q−1(x)[κ(y):κ(x)] length⁡OY,y(OY,y/byOY,y).\begin{equation} \operatorname{ord}_x\operatorname{Nm}_q(s) =\sum_{y\in q^{-1}(x)} [\kappa(y):\kappa(x)]\, \operatorname{length}_{\mathcal O_{Y,y}}(\mathcal O_{Y,y}/b_y\mathcal O_{Y,y}). \end{equation} The field degrees here are those of the ordinary local residue fields, not a count of analytic points or embeddings. The lengths include every nilpotent layer. This local multiplicity is independent of the chosen metric and frame.

The section ss need not be the pullback of a downstairs section. The DVR and non-zero-divisor hypotheses in part (3) are additional; they are not assertions about all analytic points or arbitrary base changes.

Proof. The isomorphism (I.0.11) sends the norm of the pulled-back frame to e⊗ne^{\otimes n}. The tensor-power metric therefore gives weight nϕen\phi_e. If e′=uee'=u e for an analytic unit uu, then ϕe′=ϕe−log⁡∣u∣,(e′)⊗n=une⊗n.\phi_{e'}=\phi_e-\log|u|,\qquad (e')^{\otimes n}=u^n e^{\otimes n}. These are exactly the metric transition rules. They prove the construction glues and is continuous. Scalar multiplication preserves the ordinary smooth moment sheaf, so it also preserves the asserted regularity. A new frame of the norm line multiplies this frame by an analytic unit vv; subtracting log⁡∣v∣\log|v| keeps the weight smooth. This verifies regularity for every frame, not just the displayed one.

Locally trivialize EE. The map a↦asa\mapsto as is then the matrix mbm_b tensored with ee, so exterior multiplication gives (I.1.21). Under e′=uee'=ue, the coefficient becomes b′=(q∗u)−1bb'=(q^*u)^{-1}b. Multiplication by q∗uq^*u on EE is uInuI_n; hence N(b′)=u−nN(b)N(b')=u^{-n}N(b). The displayed determinant section and (I.1.22) are therefore independent of ee. For example, the two changes in the right side are +nlog⁡∣u∣+n\log|u| and −nlog⁡∣u∣-n\log|u|, which cancel. Where N(b)N(b) is a unit, its negative logarithm is the one-coordinate linear moment representative t↦tt\mapsto t for that unit. The frozen sheaf theorem consequently proves the last regularity claim.

Finite coherent direct image commutes with analytic base change by NAG-001752. On an affinoid base change, the new finite algebra is the full ordinary tensor with its canonical complete finite-module norm, not its reduction. The matrix of multiplication and its determinant are the base changes of the old matrix and determinant. The reviewed projection isomorphism has the same compatibility. Pullback of metric weights is composition on points, and ordinary smooth pullback is provided by NAG-002424. Thus both constructions have weight nn times the pulled-back ϕe\phi_e, proving the claimed isometry and section compatibility. No claim that DVR orders are unchanged by base change is made.

For the local identity, frozen NAG-001735 gives the actual ordinary-stalk decomposition B=(q∗OY)x≃∏y∈q−1(x)OY,y.B=(q_*\mathcal O_Y)_x\simeq \prod_{y\in q^{-1}(x)}\mathcal O_{Y,y}. The module BB is free of rank nn over AA by hypothesis. Each factor is finite over AA, and the decomposition includes the whole fibre. The assumptions on byb_y say precisely that multiplication by bb is injective on this product. Lemma C.1.95 therefore applies. The germ of N(b)N(b) is its determinant, and length⁡AA/(N(b))=ord⁡AN(b)\operatorname{length}_A A/(N(b))=\operatorname{ord}_A N(b) is the local Cartier multiplicity at the DVR. The Artinian factors in the lemma are exactly the nonzero quotients of the displayed local-ring factors. This proves (I.1.23) with the stated residue fields. Changing the frame multiplies N(b)N(b) by a unit, whose order is zero. The metric enters only the regular term nϕen\phi_e in (I.1.22), not the determinant or any quotient ring. ◻

Fix a complete field KK with its specified real-valued non-Archimedean absolute value. In the new propositions the value may be trivial. All finite algebras and maps retain their canonical finite-module topology, their full fibres, and their nilpotents. A finite analytic map means a norm-admissible finite map. Analytic spaces are good and Hausdorff. All analytic integrals are taken over this same KK.

For an ordinary locally framed analytic line bundle and a metric hh, write ϕe,h=−log⁡∥e∥h,h′=e−uh⟺∥v∥h′=e−u∥v∥h.\phi_{e,h}=-\log\|e\|_h,\qquad h'=e^{-u}h \quad\Longleftrightarrow\quad \|v\|_{h'}=e^{-u}\|v\|_h. Thus ϕe,h′−ϕe,h=u\phi_{e,h'}-\phi_{e,h}=u. We use v(a)=−log⁡∣a∣v(a)=-\log|a|, without a factor 22, 2π2\pi, [L:K]−1[L:K]^{-1}, or a rescaling of ∣ ∣|\,|. For an ordinary moment-smooth metric, put Θh=d′d′′ϕe,h\Theta_h=d'd''\phi_{e,h}. This form is independent of the frame: if e′=aee'=ae, then ϕe′,h=ϕe,h−log⁡∣a∣\phi_{e',h}=\phi_{e,h}-\log|a|; the unit aa can be adjoined as a moment coordinate, on which −log⁡∣a∣-\log|a| is affine and has d′d′′d'd'' equal to zero. The frozen germ sheaf makes restriction, wedge product, and pullback compatible with this calculation.

For bounded variation in positive dimension we use a compact space XX pure of dimension dd, ordinary moment-smooth metrics h1,…,hdh_1,\ldots,h_d, and θ=Θh1∧⋯∧Θhd.\theta=\Theta_{h_1}\wedge\cdots\wedge\Theta_{h_d}. When an estimate or a uniform extension is asserted, we explicitly assume that θ\theta is positive in the frozen top-form sense. We do not identify this hypothesis with a model-theoretic nef condition. The initial variation metric h0h_0 may be any continuous metric; the ratio uu must be smooth or belong to the uniform closure specified below. No assertion that every model metric satisfies these conditions is made.

For a finite projection q:Y→Xq:Y\to X we additionally require E=q∗OYE=q_*\mathcal O_Y to be locally free of constant rank n≥1n\geq1 on ordinary opens when a determinant norm or a scalar degree formula is used. This is the precise finite locally free hypothesis here. It is not inferred from a proper-finite collar. Both spaces have the same pure dimension in the analytic integral statements. The metrics upstairs are the specified pullbacks of the downstairs metrics.

Proposition. NAG-002638 (Coherent bounded projection). Let q:Y→Xq:Y\to X be finite between good Hausdorff spaces pure of the same dimension dd, with XX compact. Let G\mathcal G be coherent on YGY_G, let u∈AX0,0(X)u\in\mathcal A_X^{0,0}(X), and let θ∈AXd,d(X)\theta\in\mathcal A_X^{d,d}(X). Put VX,F(u;θ)=∫X,Fuθ.V_{X,\mathcal F}(u;\theta)=\int_{X,\mathcal F}u\theta . Then YY is compact and VY,G(q∗u;q∗θ)=VX,q∗G(u;θ).\begin{equation} V_{Y,\mathcal G}(q^*u;q^*\theta) =V_{X,q_*\mathcal G}(u;\theta). \end{equation} If q∗Gq_*\mathcal G is locally free of constant rank ρ\rho on the analytic G-site, this is ρVX,OX(u;θ)\rho V_{X,\mathcal O_X}(u;\theta).

If θ\theta is positive, then, for every coherent F\mathcal F, ∣VX,F(u;θ)∣≤∥u∥∞ mX,F(θ),mX,F(θ)=∫X,Fθ≥0.\begin{equation} |V_{X,\mathcal F}(u;\theta)| \leq \|u\|_\infty\,m_{X,\mathcal F}(\theta), \qquad m_{X,\mathcal F}(\theta)=\int_{X,\mathcal F}\theta\geq0. \end{equation} This functional extends uniquely and continuously to u∈B(X)u\in\mathcal B(X), retaining (I.1.25). For positive θ\theta, pullback sends B(X)\mathcal B(X) into B(Y)\mathcal B(Y) and (I.1.24), including the rank case, holds for these extensions.

Proof. A finite affinoid cover of compact XX has full inverse images that are affinoid; their finite union is YY, so YY is compact. Multiplication and pullback in NAG-002424 give q∗(uθ)=q∗u q∗θq^*(u\theta)=q^*u\,q^*\theta. Apply NAG-002430 to the compactly supported form uθu\theta. Its rank clause gives the second assertion.

Let C=∥u∥∞C=\|u\|_\infty. The two forms (C+u)θ(C+u)\theta and (C−u)θ(C-u)\theta are positive. Linearity and positivity in NAG-002425 therefore give −CmX,F(θ)≤VX,F(u;θ)≤CmX,F(θ)-C m_{X,\mathcal F}(\theta)\leq V_{X,\mathcal F}(u;\theta) \leq C m_{X,\mathcal F}(\theta). Applying this inequality to differences shows that the integral values of a uniformly Cauchy smooth sequence are Cauchy. Their limits depend only on the uniform limit, by the same bound. This constructs the unique continuous extension to the specified closure, including the zero-mass case.

Pullback is continuous for the uniform norm since ∥q∗u∥∞≤∥u∥∞\|q^*u\|_\infty\leq\|u\|_\infty. The pullback of a positive top form is positive: on a common moment presentation the coefficient on a source tangent dd-plane is the corresponding target coefficient multiplied by the square of the real determinant of the induced tangent map; a rank drop gives zero. This is a pointwise superform calculation. Thus the same bounds apply to q∗θq^*\theta on YY. Taking limits in the proved smooth identity gives the asserted identity for B(X)\mathcal B(X) and its rank specialization. ◻

Proposition. NAG-002641 (Norm and bounded metric variation). Assume XX is compact, dim⁡X=dim⁡Y=d\dim X=\dim Y=d purely, and q∗OYq_*\mathcal O_Y is ordinary-locally free of constant rank nn. Let h0,h0′h_0,h'_0 be continuous metrics on the same line L0L_0, with h0′=e−uh0h'_0=e^{-u}h_0 and u∈B(X)u\in\mathcal B(X). Let θ\theta be a positive ordinary smooth top form on XX, in particular the positive wedge of smooth curvature forms specified in Section 1. Then (h0′)N=e−nuh0N(h'_0)^N=e^{-nu}h_0^N and VY,OY(q∗u;q∗θ)=VX,OX(nu;θ)=nVX,OX(u;θ).\begin{equation} V_{Y,\mathcal O_Y}(q^*u;q^*\theta) =V_{X,\mathcal O_X}(nu;\theta) =nV_{X,\mathcal O_X}(u;\theta). \end{equation} The absolute value of each expression is at most n∥u∥∞mX,OX(θ)n\|u\|_\infty m_{X,\mathcal O_X}(\theta). For smooth uu, the equalities hold without the positivity hypothesis; the stated bound and uniform extension use positivity.

Proof. In a frame ee of L0L_0, the two norm-frame weights are nϕe,h0n\phi_{e,h_0} and nϕe,h0′n\phi_{e,h'_0}. Their difference is nunu. It is independent of ee, so it gives the displayed global metric ratio. Proposition I.1.21 with G=OY\mathcal G=\mathcal O_Y gives the first and third expressions in (I.1.26); linearity gives the middle one. Its bound is (I.1.25) multiplied by nn. ◻

A common class and the two comparison objects

Let KK be an algebraically closed, complete, nontrivially valued non-Archimedean field. Set v(a)=−log⁡∣a∣v(a)=-\log|a|, without rescaling it. Fix d≥1d\geq1 and integers mi≥1m_i\geq1, and put X=(PK1)d,T=(Gm,K)d,Li=pr⁡i∗OP1(mi∞).X=(\mathbf P^1_K)^d,\qquad T=(\mathbf G_{m,K})^d,\qquad L_i=\operatorname{pr}_i^*\mathcal O_{\mathbf P^1}(m_i\infty). Use the usual analytification of this explicit smooth projective variety, with its good Hausdorff structure and pure dimension dd. Write ti=−log⁡∣Ti∣t_i=-\log|T_i| on TanT^\mathrm{an}. The continuous extension tˉ:Xan⟶R‾,d,R‾=[−∞,+∞],\bar t:X^\mathrm{an}\longrightarrow\overline{\mathbf R}^{,d},\qquad \overline{\mathbf R}=[-\infty,+\infty], sends Ti=0T_i=0 to +∞+\infty and Ti=∞T_i=\infty to −∞-\infty. The lattice is N=ZdN=\mathbf Z^d, dual to the character lattice M=ZdM=\mathbf Z^d. Lebesgue measure dtdt gives NN covolume one.

Choose convex functions ϕi∈C∞(R)\phi_i\in C^\infty(\mathbf R) and numbers Ai<BiA_i<B_i with ϕi(t)=−mit+ci−(t≤Ai),ϕi(t)=ci+(t≥Bi).\begin{equation} \phi_i(t)=-m_i t+c_i^-\quad(t\leq A_i),\qquad \phi_i(t)=c_i^+\quad(t\geq B_i). \end{equation} In the frame ei,0=1e_{i,0}=1 on the finite coordinate chart, give LiL_i the metric hih_i with weight −log⁡∥ei,0∥=ϕi(ti)-\log\|e_{i,0}\|=\phi_i(t_i). In the frame ei,∞=Timiei,0e_{i,\infty}=T_i^{m_i}e_{i,0} its weight is ϕi(ti)+miti\phi_i(t_i)+m_i t_i. The latter is constant near ∞\infty; the former is constant near zero. Thus these formulas define an actual global metric. They give ordinary smooth moment functions and smooth functions on algebraic tropical charts in the primary theory. At a boundary point, the weight is constant in that coordinate, so a very affine coordinate chart avoiding two other points of P1\mathbf P^1 suffices. There is no logarithm of a vanishing coordinate in that local representative.

This class is nonempty. If ρi∈Cc∞((Ai,Bi))\rho_i\in C_c^\infty((A_i,B_i)) is nonnegative and has integral one, then ϕi(t)=mi∫t∞(s−t)ρi(s) ds\begin{equation} \phi_i(t)=m_i\int_t^\infty(s-t)\rho_i(s)\,ds \end{equation} satisfies (I.2.1) and ϕi′′=miρi\phi_i''=m_i\rho_i. Put θi=d′d′′ϕi(ti),θ=θ1∧⋯∧θd,W(t)=∏i=1dϕi′′(ti),D=∏i=1dmi.\begin{equation} \theta_i=d'd''\phi_i(t_i),\qquad \theta=\theta_1\wedge\cdots\wedge\theta_d,\qquad W(t)=\prod_{i=1}^d\phi_i''(t_i),\qquad D=\prod_{i=1}^d m_i. \end{equation} The curvature expressions glue because frame changes add affine functions of logarithmic absolute values. In the primary smooth δ\delta-metric convention, the same smooth representative has zero residual corner term. This is a comparison of the explicit local representatives, not a claim identifying the two general sheaves of forms.

We use the positive generator ωd=d′t1∧d′′t1∧⋯∧d′td∧d′′td,∫a ωd=∫a(t) dt.\omega_d=d't_1\wedge d''t_1\wedge\cdots\wedge d't_d\wedge d''t_d, \qquad\int a\,\omega_d=\int a(t)\,dt. With grouped primed and double-primed factors the coefficient changes by (−1)d(d−1)/2(-1)^{d(d-1)/2}. This is the convention of the frozen polyhedral integral. In particular θ=t∗(Wωd)\theta=t^*(W\omega_d) on TanT^\mathrm{an}, with no factorial in this mixed wedge.

Let S\mathcal S denote the smooth functions FF on Rd\mathbf R^d which are eventually independent of each coordinate at both ends, uniformly in all the other coordinates. Explicitly, for each ii there exist ai<bia_i<b_i such that FF is independent of tit_i on each region ti<ait_i<a_i and ti>bit_i>b_i. They extend continuously to R‾,d\overline{\mathbf R}^{,d}. This class contains 11 and all compactly supported smooth functions. The function uF=F∘tˉu_F=F\circ\bar t is globally smooth in both stated chart conventions: near a boundary point ignore the coordinates in which FF is constant, and use the remaining units. Define also the larger, explicitly limited test space Crad={uF=F∘tˉ:F∈C(R‾,d)}.\mathcal C_{\rm rad} =\{u_F=F\circ\bar t:F\in C(\overline{\mathbf R}^{,d})\}. No assertion that these are all continuous functions on XanX^\mathrm{an} is made.

Let Si=supp⁡(ϕi′′)S_i=\operatorname{supp}(\phi_i'') and S=∏iSiS=\prod_iS_i. Choose a finite box B=∏i[ai′,bi′]B=\prod_i[a_i',b_i'] whose interior contains SS. Then Q=t−1(B)={e−bi′≤∣Ti∣≤e−ai′ for all i}\begin{equation} Q=t^{-1}(B) =\{e^{-b_i'}\leq|T_i|\leq e^{-a_i'}\text{ for all }i\} \end{equation} is a compact affinoid polyannulus. Its ordinary interior in XanX^\mathrm{an} contains supp⁡(θ)\operatorname{supp}(\theta), and θ\theta is zero off QQ. Every fibre used below is the whole fibre in this polyannulus. In particular QQ is not replaced by a set of Gauss points.

Write IK(uθ)=∫X/K,OXuθI_K(u\theta)=\int_{X/K,\mathcal O_X}u\theta for the frozen compact smooth integral. The primary comparison object is the smooth Gubler–Künnemann pairing JGK(uθ)J_{\rm GK}(u\theta): on a very affine chart of integration its value is the integral of the representing top form on the weighted tropicalization. In our case the single chart TT already contains the support. If μGK\mu_{\rm GK} denotes the signed Radon measure for these smooth metrics, its defining action is ∫u dμGK=JGK(uθ)for smooth u.\int u\,d\mu_{\rm GK}=J_{\rm GK}(u\theta) \quad\text{for smooth }u.

Lemma. NAG-002731 (The full torus chart has weight one). For an interior weight ww of the box BB in (I.2.4), the frozen structure-module weight of QQ with moment coordinates T1,…,TdT_1,\ldots,T_d is one. Its saturated tangent lattice is Zd\mathbf Z^d. The primary algebraic tropicalization of TT has the same lattice and coefficient one. Consequently both chart integrals of aωda\omega_d, for every a∈Cc∞(Int⁡B)a\in C_c^\infty(\operatorname{Int}B), equal ∫Ba(t) dt\int_B a(t)\,dt.

Proof. Write ri=e−wir_i=e^{-w_i}. The entire weight fibre has the actual algebra Tw=K{r−1Z,rZ−1}={∑ν∈ZdaνZν:∣aν∣rν⟶0},∥a∥w=max⁡ν∣aν∣rν.T_w=K\{r^{-1}Z,rZ^{-1}\} =\left\{\sum_{\nu\in\mathbf Z^d}a_\nu Z^\nu: |a_\nu|r^\nu\longrightarrow0\right\}, \qquad\|a\|_w=\max_\nu |a_\nu|r^\nu. This is the relation quotient, with its exact quotient norm, of NAG-001553; the fibre identification is NAG-002081. The Laurent bounds of QQ impose no further relation at this weight. All points of the fibre satisfy the strict upper bounds for both ZiZ_i and Zi−1Z_i^{-1} in the presentation of QQ, hence lie in Int⁡(Q/K)\operatorname{Int}(Q/K). The image germ is open in Rd\mathbf R^d, whose saturated lattice is Zd\mathbf Z^d. Use the identity lattice projection in NAG-002362. The source fibre algebra and target polytorus algebra are both TwT_w. If HwH_w is its Gauss field, multiplicativity of the Gauss norm and the actual tensor identification give Tw⊗^TwHw=Hw,[Zd:Zd]=1.T_w\widehat\otimes_{T_w}H_w=H_w, \qquad [\mathbf Z^d:\mathbf Z^d]=1. Thus the full Gauss-module dimension divided by the lattice index is one. No reduction, radical quotient, separated quotient losing a module, or count of analytic points has entered this calculation.

For the primary torus chart the defining ideal in its own Laurent coordinate ring is zero. Its initial ideal is again zero at every weight. Over a valued extension realizing that weight, its initial scheme is a Laurent torus, with one reduced generic component and length one. Its tangent lattice is exactly Zd\mathbf Z^d. This verifies its tropical coefficient directly, without a general multiplicity or pushforward theorem. Finally NAG-002405 computes the frozen integral with these weights, NAG-002389 gives the covolume-one real integral, and the defining primary chart integral uses the same coefficient and the same positive generator. This proves the claim. ◻

Theorem. NAG-002732 (Identity (M) on every continuous radial test). For the class in Section I.2, Identity (Mtor)(M_{\rm tor}) holds for every F∈SF\in\mathcal S: IK(uFθ)=JGK(uFθ)=∫XanuF dμGK=∫RdF(t)W(t) dt.\begin{equation} I_K(u_F\theta)=J_{\rm GK}(u_F\theta) =\int_{X^\mathrm{an}}u_F\,d\mu_{\rm GK} =\int_{\mathbf R^d}F(t)W(t)\,dt. \end{equation} Both restricted functionals have a unique continuous extension from S\mathcal S to all F∈C(R‾,d)F\in C(\overline{\mathbf R}^{,d}) in the uniform norm, and (I.2.5) holds for every such FF, with IKI_K understood as this extension. These tests belong to the uniform closure of the ordinary smooth moment functions on XanX^\mathrm{an}. The common functional is positive, vanishes if F∣S=0F|_S=0, and satisfies ∣IK(uFθ)∣≤D∥F∥∞,IK(θ)=D.\begin{equation} |I_K(u_F\theta)|\leq D\|F\|_\infty,\qquad I_K(\theta)=D. \end{equation} Here DD is the algebraic mixed degree of L1,…,LdL_1,\ldots,L_d on XX, including the local intersection lengths. If L=L1⊗⋯⊗LdL=L_1\otimes\cdots\otimes L_d has the product metric, then c1(L)d=d!θc_1(L)^d=d!\theta, so its mass and algebraic self-degree are d!Dd!D.

Proof. The local weights and their affine frame transitions give the same curvature representative d′d′′ϕid'd''\phi_i in each of the two smooth chart conventions. Multiplication and wedge product therefore represent uFθu_F\theta on TT by FWωdF W\omega_d. This representative has support in S⊂Int⁡BS\subset\operatorname{Int}B even when FF is not compactly supported. On XanX^\mathrm{an} the whole form has support inside the ordinary interior of QQ. NAG-002425 consequently permits its frozen integral to be computed on this one affinoid neighbourhood. Lemma I.2.1 computes it as ∫FW dt\int F W\,dt. In the primary definition, TT is already a very affine chart of integration for the very same compact representative; the same lemma computes its defining pairing. The equality with the smooth action of μGK\mu_{\rm GK} is its specified definition.

Since each ϕi′′\phi_i'' is nonnegative and compactly supported, ordinary one-variable calculus and (I.2.1) give ∫Rϕi′′(t) dt=ϕi′(+∞)−ϕi′(−∞)=mi.\int_{\mathbf R}\phi_i''(t)\,dt =\phi_i'(+\infty)-\phi_i'(-\infty)=m_i. Iterated integration gives ∫W dt=D\int W\,dt=D. This proves positivity, the support assertion for radial tests, and (I.2.6).

For completeness, the density required for the declared continuous tests is explicit. Identify R‾\overline{\mathbf R} with [−1,1][-1,1] by t↦tanh⁡tt\mapsto\tanh t. For any F∈C(R‾,d)F\in C(\overline{\mathbf R}^{,d}), its restriction to Rd\mathbf R^d is bounded and uniformly continuous. Convolution with a compactly supported smooth approximate identity gives smooth bounded FϵF_\epsilon with ∥Fϵ−F∥∞,Rd→0\|F_\epsilon-F\|_{\infty,\mathbf R^d}\to0. Choose a smooth clamp cR:R→Rc_R:\mathbf R\to\mathbf R equal to the identity on [−R,R][-R,R] and constant on both ends outside [−R−1,R+1][-R-1,R+1], with cR(t)≥Rc_R(t)\geq R for t≥Rt\geq R and cR(t)≤−Rc_R(t)\leq-R for t≤−Rt\leq-R. Such a clamp is obtained by integrating a smooth cutoff of the derivative from one to zero on each end. Let CRC_R be its coordinatewise product. The function Fϵ∘CRF_\epsilon\circ C_R belongs to S\mathcal S. Moreover ∥Fϵ∘CR−F∥∞≤∥Fϵ−F∥∞+∥F∘CR−F∥∞.\|F_\epsilon\circ C_R-F\|_\infty \leq\|F_\epsilon-F\|_\infty+\|F\circ C_R-F\|_\infty. The last term tends to zero as R→∞R\to\infty, because each of tanh⁡(cR(t))\tanh(c_R(t)) and tanh⁡t\tanh t lies in the same end interval of length 1−tanh⁡R1-\tanh R whenever ∣t∣>R|t|>R. This estimate holds at the endpoints as well. Uniform continuity on [−1,1]d[-1,1]^d proves the assertion. Composing with tˉ\bar t now gives uniform approximation on XanX^\mathrm{an} by the specified globally smooth functions.

The bound (I.2.6) gives unique extensions of the two restricted functionals. The right real integral has the same bound, so their extensions still equal it. The action of the given finite signed primary measure is also continuous in the uniform norm, hence it has the same values on these limits. If the previously reviewed bounded extension is subsequently adopted, uniqueness identifies it with this extension on Crad\mathcal C_{\rm rad}; no further density theorem is being assumed.

Finally choose ai∈K×a_i\in K^\times and the effective divisor mi{Ti=ai}m_i\{T_i=a_i\} representing LiL_i. These divisors meet properly at one point. Writing zi=Ti−aiz_i=T_i-a_i, their local intersection quotient is K[z1,…,zd](z1,…,zd)/(z1m1,…,zdmd).K[z_1,\ldots,z_d]_{(z_1,\ldots,z_d)}/(z_1^{m_1},\ldots,z_d^{m_d}). The equations are a regular sequence, since successively the next variable is a polynomial variable over the previous quotient. The residue monomials z1e1⋯zdedz_1^{e_1}\cdots z_d^{e_d} with 0≤ei<mi0\leq e_i<m_i form a KK-basis. The quotient has length DD and residue degree one; this computes the mixed degree in this explicit complete intersection. It retains all the nilpotent layers. Since θi∧θi=0\theta_i\wedge\theta_i=0 and (1,1)(1,1)-forms commute with one another, expansion gives (∑iθi)d=d!θ(\sum_i\theta_i)^d=d!\theta. Independently, the Hilbert polynomial of LL is ∏i(mik+1)\prod_i(m_i k+1), whose leading coefficient is DD; the algebraic self-degree is d!Dd!D. This verifies both normalizations. ◻

Proposition. NAG-002735 (Full degree in the common class). Let q:X′=(PK1)d→Xq:X'=(\mathbf P^1_K)^d\to X be given by Ti=SiniT_i=S_i^{n_i}, where ni≥1n_i\geq1, and put N=∏iniN=\prod_i n_i. The map is finite locally free of rank NN. Its analytic map, with full inverse images and their canonical finite norms, has the same rank. The pulled-back profiles are ϕi(nisi)\phi_i(n_i s_i), where si=−log⁡∣Si∣s_i=-\log|S_i|; they belong to the class of Section I.2 with degrees nimin_i m_i. For every F∈C(R‾,d)F\in C(\overline{\mathbf R}^{,d}), IK(q∗uF q∗θ)=∫X′anq∗uF dμGK,q∗h1,…,q∗hd=N∫RdF(t)W(t) dt.\begin{equation} \begin{split} I_K(q^*u_F\,q^*\theta) &=\int_{X'^\mathrm{an}}q^*u_F\, d\mu_{{\rm GK},q^*h_1,\ldots,q^*h_d}\\ &=N\int_{\mathbf R^d}F(t)W(t)\,dt. \end{split} \end{equation} The common absolute value is at most ND∥F∥∞ND\|F\|_\infty. Here NN includes inseparable degree, without a division by it.

For a line bundle MM on XanX^\mathrm{an} with continuous metric h0h_0, give Nm⁡q(q∗M)\operatorname{Nm}_q(q^*M) the projection-compatible metric from its canonical isomorphism with M⊗NM^{\otimes N}. If h0′=e−uFh0h'_0=e^{-u_F}h_0, the logarithmic ratio of the two norm metrics is NuFNu_F. Its bounded pairing against θ\theta is precisely the value in (I.2.7). This assertion uses the specified pulled-back metric class only.

Proof. On any product of the finite and infinite standard coordinate charts, the algebraic map is a tensor product of maps K[Ti]→K[Si]K[T_i]\to K[S_i], Ti↦SiniT_i\mapsto S_i^{n_i} (using inverse coordinates at infinity). Division of exponents gives the free basis ∏iSiei\prod_i S_i^{e_i}, 0≤ei<ni0\leq e_i<n_i. These charts cover the target and their whole inverse images, proving finite local freeness of rank NN in every characteristic.

The same assertion is analytic, including its norm. On a closed polydisc, or on a closed polyannulus in these coordinates, expand a convergent series upstairs and group its exponent vectors modulo (n1,…,nd)(n_1,\ldots,n_d). This gives a unique sum ∑0≤ei<niSeae(T)\sum_{0\leq e_i<n_i}S^e a_e(T) of NN convergent downstairs series. Convergence follows by grouping the weighted coefficient families; shifting by one of the finitely many exponents ee changes the coefficient norms by fixed positive factors only. Thus both this map and its inverse are bounded. It is the canonical finite-module topology of NAG-001701. Its spectra use the whole inverse domains; for (I.2.4) the inverse box is ∏i[ai′/ni,bi′/ni]\prod_i[a_i'/n_i,b_i'/n_i]. No point or branch is dropped.

For the pulled-back metrics, direct differentiation gives q∗θi=ni2ϕi′′(nisi) d′si∧d′′si.q^*\theta_i=n_i^2\phi_i''(n_i s_i)\,d's_i\wedge d''s_i. Their affine-tail degrees are nimin_i m_i. Theorem I.2.2 applied upstairs gives, for smooth tests first, ∫RdF(n1s1,…,ndsd)∏ini2ϕi′′(nisi) ds.\int_{\mathbf R^d}F(n_1s_1,\ldots,n_ds_d) \prod_i n_i^2\phi_i''(n_i s_i)\,ds. Substitute ti=nisit_i=n_i s_i. The ordinary Jacobian removes one factor ∏ini\prod_i n_i from ∏ini2\prod_i n_i^2, leaving exactly NN. The continuous-test assertion follows either from the same integral or from the already proved uniform approximation and bound. If some nin_i is divisible by the characteristic, this algebraic rank and real substitution are unchanged. In particular T=SpT=S^p has factor pp, although it is purely inseparable and has only one geometric point over each point. Nilpotents in inseparable fibres are retained by the free algebra calculation.

NAG-002539 identifies Nm⁡q(q∗M)\operatorname{Nm}_q(q^*M) with M⊗NM^{\otimes N}. In a local frame the chosen metric is the tensor-power metric, so changing its downstairs weight by uFu_F changes the norm weight by NuFNu_F, exactly as in NAG-002554. Applying the bounded functional of Theorem I.2.2 gives NIK(uFθ)N I_K(u_F\theta), equal to (I.2.7). This uses no determinant section with a zero or pole, and does not construct a norm metric for an arbitrary upstairs metrized line. ◻

Let KK be algebraically closed, complete and nontrivially valued, with its fixed absolute value and v=−log⁡∣⋅∣v=-\log|\cdot|. Throughout, d≥2,X=(PK1)d,T=(Gm,K)d,Li=pr⁡i∗O(mi∞),mi≥1.d\geq2,\qquad X=(\mathbf P^1_K)^d,\qquad T=(\mathbf G_{m,K})^d, \qquad L_i=\operatorname{pr}_i^*\mathcal O(m_i\infty),\quad m_i\geq1. We use the usual good Hausdorff analytification, pure of dimension dd. Put ti=−log⁡∣Ti∣t_i=-\log|T_i| on TanT^\mathrm{an}, and let tˉ:Xan→R‾,d\bar t:X^\mathrm{an}\to\overline{\mathbf R}^{,d} be its continuous extension, where R‾=[−∞,+∞]\overline{\mathbf R}=[-\infty,+\infty]; zero maps to +∞+\infty and infinity to −∞-\infty. The lattices are N=M=ZdN=M=\mathbf Z^d in the coordinate dual bases, and dtdt has covolume one.

Choose ϕi∈C∞(R)\phi_i\in C^\infty(\mathbf R) convex, with exact tails ϕi(t)=−mit+ci−(t≤Ai),ϕi(t)=ci+(t≥Bi),Ai<Bi.\begin{equation} \phi_i(t)=-m_i t+c_i^-\quad(t\leq A_i),\qquad \phi_i(t)=c_i^+\quad(t\geq B_i),\qquad A_i<B_i. \end{equation} The frame ei,0=1e_{i,0}=1 of LiL_i has weight ϕi(ti)\phi_i(t_i), and ei,∞=Timiei,0e_{i,\infty}=T_i^{m_i}e_{i,0} has weight ϕi(ti)+miti\phi_i(t_i)+m_it_i. Both are constant near their respective boundary points. These define global metrics hih_i in both the ordinary moment-smooth convention and the primary algebraic tropical smooth convention. Explicit examples are ϕi(t)=mi∫t∞(s−t)ρi(s) ds\phi_i(t)=m_i\int_t^\infty(s-t)\rho_i(s)\,ds, with ρi∈Cc∞((Ai,Bi))\rho_i\in C_c^\infty((A_i,B_i)), ρi≥0\rho_i\geq0, ∫ρi=1\int\rho_i=1.

Set θi=d′d′′ϕi(ti),θ=⋀i=1dθi,W(t)=∏iϕi′′(ti),D=∏imi,S=∏isupp⁡(ϕi′′).\begin{equation} \theta_i=d'd''\phi_i(t_i),\quad \theta=\bigwedge_{i=1}^d\theta_i, \quad W(t)=\prod_i\phi_i''(t_i),\quad D=\prod_i m_i, \quad S=\prod_i\operatorname{supp}(\phi_i''). \end{equation} Use the positive generator ωd=⋀i(d′ti∧d′′ti)\omega_d=\bigwedge_i(d't_i\wedge d''t_i), so that θ=t∗(Wωd)\theta=t^*(W\omega_d) and ∫aωd=∫a(t)dt\int a\omega_d=\int a(t)dt. There is no d!d! in this mixed wedge and no rescaling of the valuation. The curvature is zero near Xan∖TanX^\mathrm{an}\setminus T^\mathrm{an}. If a finite box BB contains SS in its interior, then the whole compact polyannulus Q=t−1(B)Q=t^{-1}(B) is an ordinary neighbourhood of its support. All frozen chart weights use full fibres and actual canonical Banach algebras. For w∈Rdw\in\mathbf R^d, let ρ(w)\rho(w) be the Gauss point ∣∑m∈ZdamTm∣ρ(w)=max⁡m∣am∣e−⟨m,w⟩on finite Laurent polynomials.\begin{equation} \left|\sum_{m\in\mathbf Z^d}a_mT^m\right|_{\rho(w)} =\max_m |a_m|e^{-\langle m,w\rangle} \quad\text{on finite Laurent polynomials.} \end{equation} Write Γ=ρ(Rd)\Gamma=\rho(\mathbf R^d). Let A\mathcal A be the real algebra of global ordinary smooth moment functions on XanX^\mathrm{an}. Let Acom\mathcal A_{\rm com} be its subalgebra of functions also smooth on primary algebraic tropical charts. Write LK(u)=∫X/K,OXuθ(u∈A),ν=W(t)dt,μΓ=ρ∗ν.L_K(u)=\int_{X/K,\mathcal O_X}u\theta\quad(u\in\mathcal A),\qquad \nu=W(t)dt,\qquad\mu_\Gamma=\rho_*\nu. The last expression is an explicit positive Radon measure: WW has compact support SS, and ρ∣S\rho|_S is continuous into the Hausdorff space XanX^\mathrm{an}. Its Radon regularity also follows directly by transporting regularity from the compact set SS through this homeomorphism onto its image.

For the specified primary smooth measure, take as given its defining action: on a common algebraic tropical chart, it integrates the representing smooth curvature form with its chart weights. No formal-measure comparison enters.

Theorem. NAG-002751 (Identity (M) on all continuous analytic tests). Under the standing hypotheses, for every u∈Au\in\mathcal A, LK(u)=∫Rdu(ρ(t))W(t) dt,∣LK(u)∣≤Dsup⁡ρ(S)∣u∣.\begin{equation} L_K(u)=\int_{\mathbf R^d}u(\rho(t))W(t)\,dt, \qquad |L_K(u)|\leq D\sup_{\rho(S)}|u|. \end{equation} It has a unique continuous extension to all u∈C(Xan)u\in C(X^\mathrm{an}) in the uniform norm. For every such test, including nonradial tests, this extension and the primary smooth Gubler–Künnemann measure satisfy L‾K(u)=∫Xanu dμGK=∫Xanu dμΓ=∫Rdu(ρ(t))W(t) dt.\begin{equation} \overline L_K(u)=\int_{X^\mathrm{an}}u\,d\mu_{\rm GK} =\int_{X^\mathrm{an}}u\,d\mu_\Gamma =\int_{\mathbf R^d}u(\rho(t))W(t)\,dt. \end{equation} The measures have support exactly ρ(S)\rho(S) and mass D=∏imiD=\prod_i m_i, the explicit algebraic mixed degree. The product line L1⊗⋯⊗LdL_1\otimes\cdots\otimes L_d has self-measure d!μΓd!\mu_\Gamma and self-degree d!Dd!D. No logarithmic test with a zero or pole is asserted to be integrable by this theorem.

Proof. First recover the needed radial calculation directly. On a whole interior fibre of the polyannulus QQ, the actual Laurent algebra is Tw=K{ewT,e−wT−1}T_w=K\{e^wT,e^{-w}T^{-1}\}, with its quotient norm (NAG-001553 and NAG-002081). For the identity projection of the saturated lattice Zd\mathbf Z^d, its actual Gauss algebra is Tw⊗^TwHw=HwT_w\widehat\otimes_{T_w}H_w=H_w. Thus the full module dimension and lattice index in NAG-002362 are both one. The frozen chain integral NAG-002405, with NAG-002389’s positive generator, gives LK(F∘tˉ)=∫F(t)W(t)dt(F∈Cc∞(Rd)).\begin{equation} L_K(F\circ\bar t)=\int F(t)W(t)dt\qquad(F\in C_c^\infty(\mathbf R^d)). \end{equation} The same calculation gives LK(1)=DL_K(1)=D, since ∫ϕi′′=ϕi′(+∞)−ϕi′(−∞)=mi\int\phi_i''=\phi_i'(+\infty)-\phi_i'(-\infty)=m_i. The primary torus chart has defining ideal zero, hence initial Laurent torus of generic length one, with the same saturated lattice and positive normalization. Its defining smooth pairing gives the identical formulas. There is no factorial in these mixed calculations.

Let u∈Au\in\mathcal A. The trace u∘ρu\circ\rho is continuous on Rd\mathbf R^d. On a compact neighbourhood of SS, convolution and a smooth compact cutoff give Fϵ∈Cc∞(Rd)F_\epsilon\in C_c^\infty(\mathbf R^d) with sup⁡t∈S∣u(ρ(t))−Fϵ(t)∣<ϵ.\sup_{t\in S}|u(\rho(t))-F_\epsilon(t)|<\epsilon. For example, first multiply the trace by a cutoff equal to one near SS, convolve with a sufficiently small smooth approximate identity, and multiply by a second cutoff equal to one on that neighbourhood. The function v=u−Fϵ∘tˉv=u-F_\epsilon\circ\bar t is globally ordinary smooth. It satisfies ∣v∣<ϵ|v|<\epsilon on ρ(S)\rho(S) and hence on an open neighbourhood of this compact set. By Lemma G.0.22, θ\theta has zero germ outside ρ(S)\rho(S). Therefore both (ϵ+v)θ(\epsilon+v)\theta and (ϵ−v)θ(\epsilon-v)\theta are globally positive sections: on that neighbourhood their scalar coefficients are positive, and on its complement they are zero germs. Positivity of NAG-002425 and LK(1)=DL_K(1)=D imply ∣LK(u)−LK(Fϵ∘tˉ)∣≤ϵD.|L_K(u)-L_K(F_\epsilon\circ\bar t)|\leq\epsilon D. Using (I.2.13) and the same error bound for the real integral proves (I.2.11). This proof uses the trace only in the approximation, without pretending that u∘ρ∘tˉu\circ\rho\circ\bar t is itself smooth.

The ordinary smooth algebra is uniformly dense by Lemma G.0.23; the bound in (I.2.11) is at most D∥u∥∞D\|u\|_\infty. Thus LKL_K has a unique continuous extension. The explicit functional u↦∫u dμΓu\mapsto\int u\,d\mu_\Gamma already extends it, so the first and last expressions in (I.2.12) agree for every continuous uu. This constructs the full measure in the frozen class without invoking a general Riesz existence result for all smooth forms.

For a common smooth function u∈Acomu\in\mathcal A_{\rm com}, apply exactly the same approximation to the primary smooth pairing. The primary zero-germ statement was proved separately in Lemma G.0.22. Its smooth chart integral is nonnegative on a nonnegative smooth top form: after restriction to faces its coefficients are integrated against the nonnegative tropical multiplicities, with the fixed positive normalization. This is the chart definition, and is all the positivity required here. Its radial value and mass were computed above. Hence its value at uθu\theta is ∫u∘ρ Wdt\int u\circ\rho\,Wdt. The defining smooth action of the given finite signed measure μGK\mu_{\rm GK} has these values. The common algebra is uniformly dense, and integration against either finite signed Radon measure is continuous in that norm. Thus the equality extends to every continuous uu, proving (I.2.12) and, in particular, positivity and support of μGK\mu_{\rm GK} in this class. The nonnegative continuous density WW has support exactly SS: every product neighbourhood of a point of ∏isupp⁡(ϕi′′)\prod_i\operatorname{supp}(\phi_i'') contains a product of points where all factors are positive. Thus its measure has support SS, and the section homeomorphism gives measure support exactly ρ(S)\rho(S).

For the degree normalization choose ai∈K×a_i\in K^\times and the divisors mi{Ti=ai}m_i\{T_i=a_i\}. They meet at one point with regular sequence z1m1,…,zdmdz_1^{m_1},\ldots,z_d^{m_d} in the local polynomial ring, where zi=Ti−aiz_i=T_i-a_i. The quotient basis is ∏iziei\prod_i z_i^{e_i} with 0≤ei<mi0\leq e_i<m_i, so its full length is DD, including nilpotent layers; its residue degree is one. This computes the mixed degree. Finally θi∧θi=0\theta_i\wedge\theta_i=0 and the (1,1)(1,1)-factors commute, so (∑iθi)d=d!θ(\sum_i\theta_i)^d=d!\theta. The Hilbert polynomial of the product line is ∏i(mik+1)\prod_i(m_i k+1), confirming self-degree d!Dd!D. ◻

Proposition. NAG-002755 (Power projection and the specified norm metric). Let q:X′=(PK1)d→Xq:X'=(\mathbf P^1_K)^d\to X be Ti=SiniT_i=S_i^{n_i}, with ni≥1n_i\geq1 and N=∏iniN=\prod_i n_i. Give q∗Liq^*L_i the pulled-back metrics and denote their full mixed measure from Theorem I.2.4 by μ′\mu'. For every u∈C(Xan)u\in C(X^\mathrm{an}), ∫X′anq∗u dμ′=N∫Xanu dμΓ,∣∫q∗u dμ′∣≤ND∥u∥∞.\begin{equation} \int_{X'^\mathrm{an}}q^*u\,d\mu'=N\int_{X^\mathrm{an}}u\,d\mu_\Gamma, \qquad\left|\int q^*u\,d\mu'\right|\leq ND\|u\|_\infty. \end{equation} This is the same identity for the extended frozen integrals and the primary smooth measures. The rank NN includes inseparable degree. For a continuously metrized analytic line (M,h0)(M,h_0) on XanX^\mathrm{an}, use only the projection-compatible metric on Nm⁡q(q∗M)≃M⊗N\operatorname{Nm}_q(q^*M)\simeq M^{\otimes N}. Replacing h0h_0 by e−uh0e^{-u}h_0 changes its norm-metric weight by NuNu; its bounded curvature pairing is the value in (I.2.14).

Proof. On the finite and infinite coordinate charts, qq is finite free with basis ∏iSiei\prod_i S_i^{e_i}, 0≤ei<ni0\leq e_i<n_i. On analytic polydiscs and polyannuli, grouping the convergent coefficient families by these exponent classes gives the same basis with bounded inverse coefficient maps. Finite shifts change the weighted norms only by fixed positive factors; thus this is the actual canonical finite topology of NAG-001701. All inverse domains are retained. This proves rank NN in every characteristic, including nilpotents in inseparable fibres.

For si=−log⁡∣Si∣s_i=-\log|S_i|, the upstairs profiles are ϕi(nisi)\phi_i(n_i s_i). Their mixed density and Gauss map satisfy W′(s)=∏ini2ϕi′′(nisi),q(ρ′(s))=ρ(n1s1,…,ndsd).\begin{aligned} W'(s)&=\prod_i n_i^2\phi_i''(n_i s_i),\\ q(\rho'(s))&=\rho(n_1s_1,\ldots,n_ds_d). \end{aligned} The second equality follows by evaluating Laurent polynomials in the Gauss norm: multiplying distinct exponents by (ni)(n_i) does not identify them. Hence the full-measure formula yields ∫q∗u dμ′=∫u(ρ(n1s1,…,ndsd))∏ini2ϕi′′(nisi) ds.\int q^*u\,d\mu' =\int u(\rho(n_1s_1,\ldots,n_ds_d)) \prod_i n_i^2\phi_i''(n_i s_i)\,ds. The real substitution ti=nisit_i=n_i s_i removes one product of the nin_i and leaves NN. This proves (I.2.14) for every continuous uu. The computation still gives pp for T=SpT=S^p in characteristic pp; a point count or separable degree would be wrong. Finally the frozen norm isomorphism NAG-002539 and the specified metric of NAG-002554 identify the norm frame weight with NN times the downstairs weight. Its change is NuNu, giving the stated bounded pairing. ◻

Let KK be complete and nontrivially discretely valued, with valuation ring RR, residue field kk, uniformizer π\pi, and the fixed number λ=−log⁡∣π∣>0\lambda=-\log|\pi|>0. Let XX be an integral proper KK-variety of dimension d≥2d\geq2. A model in this note is proper, flat, integral and normal over RR, of finite type, with generic fibre identified with XX. Retain explicitly the ordinary DVR local rings at all height-one points used below. Its special components have dimension dd. For a special component VV put AV=OX,ηV,bV=ord⁡V(π)>0,∣f(ξV)∣=∣π∣ord⁡V(f)/bV(f∈K(X)×).A_V=\mathcal O_{\mathcal X,\eta_V},\qquad b_V=\operatorname{ord}_V(\pi)>0, \qquad |f(\xi_V)|=|\pi|^{\operatorname{ord}_V(f)/b_V}\quad(f\in K(X)^\times). The valuation norm defines an analytic point by NAG-001707. Its kernel is zero, and its centre on the model is ηV\eta_V. Distinct components have distinct points. No assertion that an ordinary analytic stalk at ξV\xi_V is a DVR is made.

For lines Li\mathcal L_i extending Li⊗eiL_i^{\otimes e_i}, 1≤i≤d1\leq i\leq d, write E=∏i=1dei,δV=deg⁡L1∣V,…,Ld∣V,k(V),μX=∑VbVδVEδξV.\begin{equation} E=\prod_{i=1}^d e_i,\qquad \delta_V=\deg_{\mathcal L_1|_V,\ldots,\mathcal L_d|_V,k}(V),\qquad \mu_{\mathcal X}=\sum_V\frac{b_V\delta_V}{E}\delta_{\xi_V}. \end{equation} This is the specified finite model measure, with the ordinary mixed component degree over kk. All root exponents, nilpotent lengths and full residue-field degrees are retained. It can be signed. The root model metric has ∥s∥=∣a∣1/e\|s\|=|a|^{1/e} when s⊗e=aeMs^{\otimes e}=a e_{\mathcal M} in a model frame of M\mathcal M. The finite measure is the object compared here; its general independence is not assumed in its definition.

A displayed modification is a proper morphism p:Y→Xp:\mathcal Y\to\mathcal X inducing the identity on XX, together with an open U⊆XU\subseteq\mathcal X containing XX and every height-one point of X\mathcal X, such that p−1(U)→Up^{-1}(U)\to U is an isomorphism. This open is part of the hypotheses. We do not infer it merely from the equality of generic fibres. The complement has codimension at least two. A component WW of Yk\mathcal Y_k is called exceptional here if it does not dominate a component of Xk\mathcal X_k.

Lemma. NAG-003046 (Codimension-one coefficients under a modification). Let pp be a displayed modification. For each prime divisor VV of X\mathcal X there is exactly one prime divisor V~\widetilde V of Y\mathcal Y dominating it, and their ordinary local DVRs are identified inside K(X)K(X). For a Cartier divisor DD with rational section presentation, p∗cyc⁡(p∗D)=cyc⁡(D).\begin{equation} p_*\operatorname{cyc}(p^*D)=\operatorname{cyc}(D). \end{equation} The pushforward uses full residue-field degrees on equal-dimensional images and zero on dimension-decreasing images. In particular, for special components, bV~=bVb_{\widetilde V}=b_V and ξV~=ξV\xi_{\widetilde V}=\xi_V. Every exceptional special component has image of dimension at most d−1d-1.

The codimension-one isomorphism hypothesis is checkable for a blowup of a finite-type ideal II with I∣X=OXI|_X=\mathcal O_X: it holds wherever II is locally principal and nonzero, hence on an open containing every height-one point of the normal model. This assertion concerns the blowup itself; any additional normalization or domination needs its own compatibility check.

Proof. Over the generic point of a prime divisor VV, the morphism is an isomorphism. Taking the closure of its inverse-image point gives V~\widetilde V and identifies its local ring and residue field with those of VV. No second divisor can dominate VV, since it would also meet this isomorphic inverse image. Any other prime divisor lies over the complement of UU, so its image has smaller dimension. Properness makes that image closed. The coefficient of p∗Dp^*D at V~\widetilde V is the coefficient of DD at VV, because the same rational function is measured in the same DVR. Its pushforward degree is one, and the other coefficients push forward to zero. This proves (I.2.16). Apply it locally to π\pi and to arbitrary rational functions to obtain the assertions about bb and ξ\xi.

At a height-one point vv, IvI_v is a nonzero finitely generated ideal in a DVR and thus is principal, say (a)(a). After shrinking a neighbourhood, aa is a section of II and every chosen finite generator of II is a multiple of aa; clearing the finitely many localization denominators proves this assertion on an open. On that open, the Rees algebra is ⨁n≥0(an)Tn≃A[t]\bigoplus_{n\geq0}(a^n)T^n\simeq A[t], since aa is a non-zero-divisor. Its Proj is Spec⁡A\operatorname{Spec}A. The union of these opens and the generic fibre gives the required UU for the blowup. ◻

Theorem. NAG-003048 (Exact component-degree criterion for model measures). Use p:Y→Xp:\mathcal Y\to\mathcal X as above and the pulled-back model lines p∗Lip^*\mathcal L_i with the same root exponents. Then the following are equivalent:

  1. The two finite measures on the common XanX^\mathrm{an} are equal.

  2. Their component degrees satisfy δV~=δVfor every V,δW=0for every exceptional W.\begin{equation} \delta_{\widetilde V}=\delta_V\quad\hbox{for every }V, \qquad \delta_W=0\quad\hbox{for every exceptional }W. \end{equation}

Under this criterion all continuous tests, and all Borel tests finite on the common finite support, have equal integrals. In particular, if MM has a model M\mathcal M extending M⊗e0M^{\otimes e_0}, the pullback model metric is the same metric on MM. Every nonzero rational section ss satisfies ∫−log⁡∥s∥ dμY=∫−log⁡∥s∥ dμX.\begin{equation} \int-\log\|s\|\,d\mu_{\mathcal Y} =\int-\log\|s\|\,d\mu_{\mathcal X}. \end{equation} The logarithm is used off its zero/pole support and may be extended arbitrarily across that null set for these atomic measures. Both are finite total-variation integrals. If aV=ord⁡V(s⊗e0)a_V=\operatorname{ord}_V(s^{\otimes e_0}) in a model frame, the common downstairs expression is λe0E∑VaVδV,−log⁡∥s(ξV)∥=λaVe0bV.\begin{equation} \frac{\lambda}{e_0E}\sum_Va_V\delta_V, \qquad -\log\|s(\xi_V)\|=\frac{\lambda a_V}{e_0b_V}. \end{equation} Replacing ss by πs\pi s changes it by +λμX(Xan)+\lambda\mu_{\mathcal X}(X^\mathrm{an}).

Proof. The strict-transform atoms and their bb’s agree by Lemma I.2.6. An exceptional atom cannot equal one of these: its centre on X\mathcal X is in X∖U\mathcal X\setminus U, whereas the strict-transform atom has centre ηV∈U\eta_V\in U. Distinct components of Y\mathcal Y also give distinct atoms. Comparing the coefficients of each Dirac mass in (I.2.15), and using bW>0b_W>0, proves precisely the equivalence with (I.2.17). This argument allows signed degrees; no cancellation between distinct atoms is possible. Equality of the finite measures proves equality for every stated test.

A frame of M\mathcal M pulls back to a frame of p∗Mp^*\mathcal M at every point reducing over that frame. The coefficient of a generic section is the same rational function, evaluated at the same analytic point. The root metric formula therefore proves equality of these metrics on all of XanX^\mathrm{an}. The reduction maps commute with pp, as follows directly by pulling back the inequalities ∣f∣≤1|f|\leq1 and ∣f∣<1|f|<1 defining a centre. Every atom has generic underlying scheme point, so a nonzero rational section is regular and nonvanishing there. The root norm gives the second formula in (I.2.19); multiplication by its mass bVδV/Eb_V\delta_V/E gives the first. This proves finiteness without any singular limiting argument. Finally aVa_V becomes aV+e0bVa_V+e_0b_V for πs\pi s, giving the positive change λ∑VbVδV/E\lambda\sum_Vb_V\delta_V/E. ◻

Proposition. NAG-003052 (Two semipositive systems on the point blowup). The special fibre is V′+FV'+F with both multiplicities one. Its two component DVR norms are ∣f(ξV′)∣=∣f(ξ0)∣,∣∑mamzm∣(ξF)=max⁡m∣am∣ ∣π∣∣m∣,|f(\xi_{V'})|=|f(\xi_0)|,\qquad \left|\sum_ma_mz^m\right|(\xi_F)=\max_m|a_m|\,|\pi|^{|m|}, where zi=Zi/Z0z_i=Z_i/Z_0 and ξ0\xi_0 is the standard unit-radius Gauss point. Put A=H⊗O(−F)A=H\otimes\mathcal O(-F). Both HH and AA are generated by global sections and have semipositive model metrics. For positive integers m1,…,mdm_1,\ldots,m_d and D=∏imiD=\prod_i m_i, their mixed degrees are V′FHm1,…,HmdD0Am1,…,Amd0D\begin{equation} \begin{array}{c|cc} &V'&F\\ \hline H^{m_1},\ldots,H^{m_d}&D&0\\ A^{m_1},\ldots,A^{m_d}&0&D \end{array} \end{equation} Consequently the pullbacks of the standard model lines give μH=Dδξ0\mu_H=D\delta_{\xi_0} on both models, verifying (I.2.17). The other semipositive system gives μA=DδξF\mu_A=D\delta_{\xi_F}. Although HH and AA have the same generic line bundle, AA is not being treated as the pullback of that standard metrized line.

Proof. Near PP, the ideal is (π,z1,…,zd)(\pi,z_1,\ldots,z_d). The π\pi-chart is R[u1,…,ud]R[u_1,\ldots,u_d] with zi=πuiz_i=\pi u_i. Its exceptional generic DVR has uniformizer π\pi and residue field k(u1,…,ud)k(u_1,\ldots,u_d). The zjz_j-chart has equation R[zj,u0,(ui)i≠j]/(π−zju0),zi=zjui (i≠j).R[z_j,u_0,(u_i)_{i\ne j}]/(\pi-z_ju_0), \qquad z_i=z_ju_i\ (i\ne j). These are the usual regular charts of the blowup of the coordinate regular closed point: at a crossing over a codimension-cc point of k[(ui)i≠j]k[(u_i)_{i\ne j}], the local ring has dimension c+2c+2 and its maximal ideal is generated by zj,u0z_j,u_0 and lifts of cc regular parameters. Away from the crossing one of zj,u0z_j,u_0 is a unit, giving a localization of a polynomial chart over RR. Thus the total space is regular, in particular normal. The blowup is projective over the proper model. These charts also show explicitly that the special fibre consists of zj=0z_j=0 and u0=0u_0=0, each with multiplicity one. The component V′V' is Bl⁡PPkd\operatorname{Bl}_P\mathbf P^d_k, while F=PkdF=\mathbf P^d_k. The generic DVR of V′V' is unchanged from Xk\mathcal X_k. In the π\pi-chart, the reduction of a polynomial in the independent uiu_i cannot cancel its distinct monomials at the generic point. Its order is therefore min⁡m(ord⁡Ram+∣m∣)\min_m(\operatorname{ord}_R a_m+|m|), which gives the stated Gauss norm at FF.

The sections Z0,…,ZdZ_0,\ldots,Z_d generate HH. The sections πZ0,Z1,…,Zd\pi Z_0,Z_1,\ldots,Z_d generate AA: their ideal is the pullback of the centre ideal times HH, hence becomes the invertible ideal O(−F)\mathcal O(-F) times HH on the blowup. On FF the line HH is trivial, and the restrictions of those divided generators identify A∣FA|_F with OPkd(1)\mathcal O_{\mathbf P^d_k}(1). On V′V' the first divided generator πZ0\pi Z_0 restricts to zero and the remaining dd generators have no common zero. These descriptions can also be read from the two displayed charts. Global generation gives nonnegative degree on every special-fibre curve: a nonzero generating section on an integral curve has an effective zero divisor. Thus the two metrics are semipositive.

For HH on V′V', use Z0+ZiZ_0+Z_i, 1≤i≤d1\leq i\leq d. Each avoids PP, and their common zero is the single rational point Q=[1:−1:⋯:−1]≠PQ=[1:-1:\cdots:-1]\ne P, where pp is an isomorphism. Raising the iith section to mim_i gives the local algebra k[t1,…,td]/(t1m1,…,tdmd)k[t_1,\ldots,t_d]/(t_1^{m_1},\ldots,t_d^{m_d}), of length DD by its monomial basis. On FF, choose nonzero constant sections of the trivial restrictions of HH; there is no intersection. This proves the first row of (I.2.20).

For AA on F=PkdF=\mathbf P^d_k, dd coordinate hyperplanes meet at a rational point and the same powers give length DD. On V′V', the remaining divided coordinate generators define the resolution of projection from PP to Pkd−1\mathbf P^{d-1}_k. In the zjz_j-chart this is [z1:⋯:zd]=[u1:⋯:1:⋯:ud][z_1:\cdots:z_d]=[u_1:\cdots:1:\cdots:u_d]; its fibre direction is the line through PP. The dd coordinate hyperplanes of Pd−1\mathbf P^{d-1} have empty total intersection. Their pullbacks intersect properly through each nonempty stage (the chart fibres are one-dimensional), and the final intersection is empty. Taking powers preserves that support and gives degree zero. This proves the second row. Substituting the degrees and the two multiplicities into (I.2.15) proves all measure assertions. ◻

Proposition. NAG-003058 (Two ways that forgetting support loses information). In every dimension d≥2d\geq2 the following hold.

  1. On PRd\mathbf P^d_R, a closed rational special-fibre point has class zero in the ordinary Chow group CH0(PRd)CH_0(\mathbf P^d_R), but has vertical degree one. Thus a degree homomorphism on that ordinary Chow group cannot recover the required vertical intersection degree. The support must be retained.

  2. On the blowup of Section 3, p∗[F]=0p_*[F]=0 as a dd-cycle, but deg⁡A∣F,…,A∣F,k(F)=1.\deg_{A|_F,\ldots,A|_F,k}(F)=1. The two semipositive model metrics hH,hAh_H,h_A on the same generic O(1)\mathcal O(1) agree outside the tube of PP. On the standard unit polydisc around PP their difference of weights is u=log⁡(hA/hH)=−log⁡max⁡{∣π∣,∣z1∣,…,∣zd∣}.\begin{equation} u=\log(h_A/h_H) =-\log\max\{|\pi|,|z_1|,\ldots,|z_d|\}. \end{equation} It satisfies u(ξ0)=0u(\xi_0)=0 and u(ξF)=λu(\xi_F)=\lambda. With the semipositive model measure μA=δξF\mu_A=\delta_{\xi_F} one therefore has ∫u dμA=λ>0\int u\,d\mu_A=\lambda>0. Agreement outside that tube and equality of generic line-bundle classes do not imply zero metric correction. If the supported model values of the preceding proposition are available, changing the first model line from AA to H=A(F)H=A(F), with the other dd lines equal to AA, changes that value by +λ+\lambda.

Proof. For (1), choose the horizontal closed section C={Z1=⋯=Zd=0}≃Spec⁡RC=\{Z_1=\cdots=Z_d=0\}\simeq\operatorname{Spec}R. The divisor of π\pi on this integral one-dimensional scheme is its closed point with coefficient length⁡R(R/πR)=1\operatorname{length}_R(R/\pi R)=1. Its pushforward is a principal-divisor relation in CH0(PRd)CH_0(\mathbf P^d_R), so that point has ordinary class zero. Its degree as a zero-cycle on the special fibre is [k:k]=1[k:k]=1. Rational equivalence inside the vertical support cannot use the horizontal curve CC; this is exactly the information lost by forgetting support.

For (2), the image of FF is the point PP, so cycle pushforward in dimension dd is zero. But A∣F=OPkd(1)A|_F=\mathcal O_{\mathbf P^d_k}(1) by Proposition I.2.8, whose top mixed degree is one. This does not contradict projection: the factors AA are not the pulled-back line HH. On the unit polydisc, a generator of the pulled-back centre ideal has absolute value max⁡{∣π∣,∣z1∣,…,∣zd∣}\max\{|\pi|,|z_1|,\ldots,|z_d|\}. The canonical generic section of O(F)\mathcal O(F) has that norm. Since H=A⊗O(F)H=A\otimes\mathcal O(F), the ratio of the metrics gives (I.2.21). Away from the tube {red⁡X(x)=P}\{\operatorname{red}_{\mathcal X}(x)=P\} the centre ideal is generated by a unit of norm one, so the metrics agree there. At ξ0\xi_0 one of the displayed coordinate norms is one; at ξF\xi_F they all equal ∣π∣|\pi|. This proves the two values and the integral. Finally, compatible generic sections of HH and AA have model divisors differing by FF. Additivity of the stated supported interface gives λdeg⁡(c1(A∣F)d[F])=λ\lambda\deg(c_1(A|_F)^d[F])=\lambda, with the asserted positive sign. ◻

Let RR be a DVR, S=Spec⁡RS=\operatorname{Spec}R, η\eta its generic point, and kk its residue field. For analytic assertions assume that RR is the valuation ring of a complete nontrivially discretely valued field KK with its fixed absolute value. Use the dimension function δ(η)=1\delta(\eta)=1, δ(s)=0\delta(s)=0 at the closed point. On a finite-type SS-scheme, dimensions below are δ\delta-dimensions. In particular, a horizontal integral model of relative dimension dd has dimension d+1d+1, and a vertical integral kk-scheme has its usual dimension. For a closed subset TT of a finite-type SS-scheme X\mathcal X set Cr(X;T)=CH⁡r(Tred).C_r(\mathcal X;T)=\operatorname{CH}_r(T_{\mathrm{red}}). Rational equivalences use integral subschemes inside TT. This group is not defined as the image in CH⁡r(X)\operatorname{CH}_r(\mathcal X). Reduction identifies Chow groups: integral subschemes, their rational functions, and their divisor relations are unchanged. This observation does not replace the fundamental cycle of a nonreduced scheme by that of its reduction; its generic lengths remain coefficients of the cycle. We freely use a scheme-theoretic intersection to construct a cycle and the canonical reduction identification to place its class in the group on its closed support.

Use the existing supported Gysin operation G(L,s),T:Cr(X;T)→Cr−1(X;T∩∣Z(s)∣)G_{(\mathcal L,s),T}:C_r(\mathcal X;T)\to C_{r-1}(\mathcal X;T\cap|Z(s)|) for an invertible sheaf L\mathcal L and any global section ss, including the zero section or a zero-divisor. Use the selected signed presentations, with their signed operation defined by (C.1.50). For proper integral W/kW/k, the mixed degree δ(W;M1,…,Md)\delta(W;M_1,\ldots,M_d) is the degree of c1(M1)⋯c1(Md)∩[W]c_1(M_1)\cdots c_1(M_d)\cap[W] as in the selected algebraic component-degree formula. Now let KK be complete and nontrivially discretely valued with ring RR, and put λ=−log⁡∣π∣>0\lambda=-\log|\pi|>0. Let X\mathcal X be an integral normal proper flat model of relative dimension d≥2d\geq2. For 0≤i≤d0\leq i\leq d, let Li\mathcal L_i extend Li⊗eiL_i^{\otimes e_i}, ei≥1e_i\geq1, and let σi=si⊗ei\sigma_i=s_i^{\otimes e_i} be a nonzero rational model section. Specify signed presentations on closed supports ZiZ_i satisfying ⋂i=0dZi∩X=∅,X=XK.\begin{equation} \bigcap_{i=0}^d Z_i\cap X=\varnothing, \qquad X=\mathcal X_K. \end{equation} Write IiI_i for the supported operations, E=∏i=0deiE=\prod_{i=0}^d e_i, and T=⋂iZiT=\bigcap_iZ_i. Define JX=λEdeg⁡k(I0⋯Id[X]),I0⋯Id[X]∈CH⁡0(Tred).\begin{equation} J_{\mathcal X}=\frac{\lambda}{E} \deg_k\bigl(I_0\cdots I_d[\mathcal X]\bigr), \qquad I_0\cdots I_d[\mathcal X]\in\operatorname{CH}_0(T_{\mathrm{red}}). \end{equation} The value is defined on this model data. This construction also applies to the horizontal integral cycles occurring below, using the remaining pulled-back factor presentations, even when a factor vanishes identically.

Theorem. NAG-003112 (Refined operations on every prescribed support). The maps G(L,s),TG_{(\mathcal L,s),T} are well-defined on Chow groups and satisfy all the following assertions.

  1. For an integral W⊂TW\subset T, if s∣Ws|_W is nonzero, the result is the divisor cycle of s∣Ws|_W, with its actual local lengths. If s∣W=0s|_W=0, it is the image of c1(L∣W)∩[W]c_1(\mathcal L|_W)\cap[W] in the stated target, not an instruction to discard WW.

  2. Let p:Y→Xp:\mathcal Y\to\mathcal X be proper, let T′⊂YT'\subset\mathcal Y and T⊂XT\subset\mathcal X be closed with p(T′)⊂Tp(T')\subset T, and pull back the actual line and section. Then p∗G(p∗L,p∗s),T′(α)=G(L,s),T(p∗α)in Cr−1(X;T∩∣D∣).\begin{equation} p_*G_{(p^*\mathcal L,p^*s),T'}(\alpha) =G_{(\mathcal L,s),T}(p_*\alpha) \quad\text{in }C_{r-1}(\mathcal X;T\cap|D|). \end{equation} This includes naturality under enlargement of the prescribed support.

  3. Two such operations commute in the group on T∩∣Z(s)∣∩∣Z(t)∣T\cap|Z(s)|\cap|Z(t)|. No proper-intersection hypothesis is needed.

  4. For (M,t)(\mathcal M,t) a second pair, set B=∣Z(s)∣∪∣Z(t)∣B=|Z(s)|\cup|Z(t)|. After pushing the three targets into T∩BT\cap B, G(L⊗M,s⊗t),T=G(L,s),T+G(M,t),T.\begin{equation} G_{(\mathcal L\otimes\mathcal M,s\otimes t),T} =G_{(\mathcal L,s),T}+G_{(\mathcal M,t),T}. \end{equation} After pushing into TT, the operation is c1(L∣T)∩−c_1(\mathcal L|_T)\cap-. If a coherent F\mathcal F on X\mathcal X is supported in TT, has dim⁡δSupp⁡(F)≤r\dim_\delta\operatorname{Supp}(\mathcal F)\leq r, and s:F⊗L−1→Fs:\mathcal F\otimes\mathcal L^{-1}\to\mathcal F is injective, then G(L,s),T[F]r=[F/s(F⊗L−1)]r−1in Cr−1(X;T∩∣D∣).G_{(\mathcal L,s),T}[\mathcal F]_r =[\mathcal F/s(\mathcal F\otimes\mathcal L^{-1})]_{r-1} \quad\text{in }C_{r-1}(\mathcal X;T\cap|D|). All generic and quotient lengths are retained.

Proof. The definition gives (1), and Stacks, Tag 02TO proves descent through rational equivalence. Apply Stacks, Tag 02TA to the proper map Tred′→TredT'_{\mathrm{red}}\to T_{\mathrm{red}} and the restricted line and section. Its target is the zero scheme of the restricted section on TT, so its pushforward identity is exactly (I.2.24) after reduction. Applying this also to closed immersions proves the support-enlargement statement. It does not pass through the image of CH⁡(T)\operatorname{CH}(T) in the ambient Chow group.

For clarity, the proper-pushforward calculation retains the factor [K(W′):K(p(W′))][K(W'):K(p(W'))] when dimensions agree. If an integral W′W' is contracted, choose a rational section of the relevant line on its integral image WW. Its pullback divisor is supported over a proper closed subset of WW. Every component then has image dimension less than dim⁡δW′−1\dim_\delta W'-1, and its cycle pushforward is zero. This also handles the contained-component Chern term by its divisor representative. Thus dimension-decreasing images do not create an unrecorded coefficient.

Apply Stacks, Tag 0B73 on TredT_{\mathrm{red}}. The rational equivalence in its proof lies on the intersection of both zero schemes; in particular its target is precisely the support in (3). Reduction and the naturality just proved identify the successive targets used here with those of that lemma.

To prove (4), work on an integral cycle [W][W]. If neither restricted section is zero, divisors of their tensor product add as cycles. Locally this is ord⁡(fg)=ord⁡(f)+ord⁡(g)\operatorname{ord}(fg)=\operatorname{ord}(f)+\operatorname{ord}(g); at a one-dimensional local domain one can see the length equality from 0→A/(f)→ g A/(fg)→A/(g)→00\to A/(f)\xrightarrow{\,g\,}A/(fg)\to A/(g)\to0. If one or both sections vanish identically on WW, then W⊂BW\subset B. The product contributes c1((L⊗M)∣W)[W]c_1((\mathcal L\otimes\mathcal M)|_W)[W]; each nonzero restricted section represents its own first Chern class on WW. Additivity of that class, Stacks, Tag 02SP, proves equality already in CH⁡(W)\operatorname{CH}(W), and hence in the required target. Linearity proves (I.2.25). Finally Stacks, Tag 02T9 identifies the push into TT with the first Chern operation, and the coherent assertion uses the actual annihilator scheme Z=V(Ann⁡F)Z=V(\operatorname{Ann}\mathcal F), with ∣Z∣⊂T|Z|\subset T. Keep the full coherent module on ZZ. Tensor the asserted injection with the invertible line so that Stacks, Tag 02TB(2), with k=r−1k=r-1, applies to the actual module and its restriction to the zero scheme. This restriction is exactly F/s(F⊗L−1)\mathcal F/s(\mathcal F\otimes\mathcal L^{-1}). The closed inclusion Zred→ZZ_{\mathrm{red}}\to Z and its base-changed zero scheme give the canonical identifications of the Chow groups: the cycle and rational-function generators are unchanged. Apply Stacks, Tag 02TA to these inclusions and then apply Stacks, Tag 02TA to the proper closed map Zred→TredZ_{\mathrm{red}}\to T_{\mathrm{red}} with its actual zero support. The coherent cycles are computed on the full modules before these reduction identifications, so all lengths are retained. Injectivity remains essential. ◻

Theorem. NAG-003117 (The missing component-degree projection identity). Let q:W→Vq:W\to V be a proper morphism of proper integral kk-schemes, with dim⁡W=d\dim W=d and dim⁡V≤d\dim V\leq d, and let M1,…,MdM_1,\ldots,M_d be invertible sheaves on VV. If dim⁡q(W)<d\dim q(W)<d, then δ(W;q∗M1,…,q∗Md)=0.\delta(W;q^*M_1,\ldots,q^*M_d)=0. If qq is dominant and dim⁡V=d\dim V=d, then δ(W;q∗M1,…,q∗Md)=[k(W):k(V)]δ(V;M1,…,Md).\begin{equation} \delta(W;q^*M_1,\ldots,q^*M_d) =[k(W):k(V)]\delta(V;M_1,\ldots,M_d). \end{equation} In particular, let p:Y→Xp:\mathcal Y\to\mathcal X be a supplied proper map of integral normal proper flat DVR models, inducing the identity on the generic fibre, and an isomorphism over an open containing every height-one point of X\mathcal X. For arbitrary model lines Li\mathcal L_i, each strict-transform special component retains its mixed degree and each exceptional special component has mixed degree zero. The finite model measures with masses bVδV/∏ieib_V\delta_V/\prod_i e_i, where bV=ord⁡V(π)b_V=\operatorname{ord}_V(\pi), are consequently equal on every continuous test and every Borel test finite on their finite support. The conclusion concerns these supplied pullback models.

Proof. Use the global pairs (Mi,0)(M_i,0) on VV. Their Gysin operations are exactly c1(Mi)c_1(M_i), since their zero schemes are all of VV. Iterating (C.1.48) gives q∗(c1(q∗M1)⋯c1(q∗Md)∩[W])=c1(M1)⋯c1(Md)∩q∗[W].q_*\bigl(c_1(q^*M_1)\cdots c_1(q^*M_d)\cap[W]\bigr) =c_1(M_1)\cdots c_1(M_d)\cap q_*[W]. In the contracted case q∗[W]=0q_*[W]=0. In the dominant equal-dimensional case it is [k(W):k(V)][V][k(W):k(V)][V], with the full field degree. Proper pushforward preserves degree over kk, proving both formulas.

For the models, the supplied open identifies the ordinary DVRs at a special component and its unique strict transform. The proper map between these integral special components has degree one. An exceptional component has image dimension less than dd, so the two formulas give respectively the old mixed degree and zero. The identified DVRs have the same bVb_V and the same valuation point on the common generic analytic fibre. The additional exceptional terms have zero weight. Substitution in the specified finite atomic measure proves equality, and hence equality for all the stated tests. Neither existence of the map nor an analytic integral comparison enters this argument. ◻

Theorem. NAG-003121 (Supported projection and horizontal–vertical induction). The value (I.2.23) is independent of the permitted signed presentations and symmetric in all d+1d+1 inputs. Let p:Y→Xp:\mathcal Y\to\mathcal X be a supplied proper dominant map of integral proper flat models of the same relative dimension, and put N=[K(Y):K(X)]N=[K(\mathcal Y):K(\mathcal X)]. With the actual pulled-back factor data and supports on Y\mathcal Y, one has JY=NJX.\begin{equation} J_{\mathcal Y}=N J_{\mathcal X}. \end{equation} Normality of Y\mathcal Y is not needed for this supported equality.

At the ordinary height-one DVRs of X\mathcal X, write div⁡L0(σ0)=∑H horizontalaH[H]+∑V verticalaV[V].\operatorname{div}_{\mathcal L_0}(\sigma_0)=\sum_{H\ \mathrm{horizontal}}a_H[H] +\sum_{V\ \mathrm{vertical}}a_V[V]. Let JHJ_H be the construction (I.2.23) on HH using the remaining dd pulled-back presentations and denominator ∏i=1dei\prod_{i=1}^d e_i. Put δV=δ(V;L1∣V,…,Ld∣V)\delta_V=\delta(V;\mathcal L_1|_V,\ldots,\mathcal L_d|_V). Then JX=1e0∑HaHJH+λe0∏i=1dei∑VaVδV.\begin{equation} J_{\mathcal X}=\frac1{e_0}\sum_Ha_H J_H +\frac{\lambda}{e_0\prod_{i=1}^d e_i}\sum_Va_V\delta_V. \end{equation} Thus the horizontal coefficients are exactly those of div⁡(s0)\operatorname{div}(s_0) on the generic fibre. For the root model metric and the specified finite model measure, the second summand is ∫−log⁡∥s0∥ dμ,−log⁡∥s0(ξV)∥=λaVe0bV,μ=∑VbVδV∏i=1deiδξV.\begin{equation} \int-\log\|s_0\|\,d\mu,\qquad -\log\|s_0(\xi_V)\|=\frac{\lambda a_V}{e_0b_V},\qquad \mu=\sum_V\frac{b_V\delta_V}{\prod_{i=1}^d e_i}\delta_{\xi_V}. \end{equation} It is a finite total-variation integral, regardless of signs of the mixed degrees. At fixed metrics, replacing any sis_i by πsi\pi s_i changes JJ by +λ+\lambda times the corresponding model-measure mass. Only (I.2.22) is required; no pairwise proper intersection or generic nonvanishing on every intermediate cycle is imposed.

Proof. The operations produce a zero-class on TT after d+1d+1 steps. Condition (I.2.22) makes TredT_{\mathrm{red}} a proper kk-scheme, so its degree is defined even if TT has positive dimension. Presentation independence and commutation prove the first assertions. For the supplied proper map, repeated supported projection gives p∗(Ip∗0⋯Ip∗d[Y])=I0⋯Idp∗[Y]=NI0⋯Id[X].p_*(I_{p^*0}\cdots I_{p^*d}[\mathcal Y]) =I_0\cdots I_d p_*[\mathcal Y] =N I_0\cdots I_d[\mathcal X]. The supports remain the displayed inverse images; degree compatibility proves (I.2.27). This proves a statement about a supplied map and compatible data, not existence of a common model.

Apply I0I_0 first. Each factor section in its signed presentation is nonzero on X\mathcal X. Its Gysin image of [X][\mathcal X] is therefore its divisor cycle, and subtraction gives the displayed cycle of σ0\sigma_0 on Z0Z_0. A positive coefficient at a height-one point is the length of the quotient by a local equation in the ordinary DVR; differences give the signed coefficient. Apply the remaining operators to each component, using proper compatibility for its closed immersion. On a horizontal HH, the remaining generic supports have empty total intersection: H⊂Z0H\subset Z_0 and (I.2.22) applies. This contribution is (aH/e0)JH(a_H/e_0)J_H. The value JHJ_H uses the restricted factor presentations, not a purported restriction of a rational section that vanishes identically on HH.

On a vertical VV, push the remaining supported intersection class into CH⁡0(V)\operatorname{CH}_0(V). The last assertion of Theorem C.1.101, its signed extension, and proper compatibility show that the result is c1(L1∣V)⋯c1(Ld∣V)[V]c_1(\mathcal L_1|_V)\cdots c_1(\mathcal L_d|_V)[V]. This push stays inside the proper vertical scheme VV, so it preserves the relevant degree. It gives aVδVa_V\delta_V and proves (I.2.28). Localization at π\pi identifies the horizontal DVR with its generic-fibre DVR. Since σ0=s0e0\sigma_0=s_0^{e_0}, its horizontal coefficient is e0e_0 times the coefficient of div⁡(s0)\operatorname{div}(s_0).

For (I.2.29), the point ξV\xi_V has norm ∣f(ξV)∣=∣π∣ord⁡V(f)/bV|f(\xi_V)|=|\pi|^{\operatorname{ord}_V(f)/b_V} and generic scheme support. A model frame gives the stated root norm. Multiplying its negative logarithm by the mass cancels bVb_V, giving exactly the vertical summand of (I.2.28). All these finitely many values are finite; the zero/pole locus is null for the finite atomic measure. This proves the integral assertion by a finite sum, without a limit.

Finally si↦πsis_i\mapsto\pi s_i changes the model divisor by eidiv⁡(π)e_i\operatorname{div}(\pi) and may enlarge its support by the special fibre. Its first supported operation on [X][\mathcal X] gives ei∑VbV[V]e_i\sum_Vb_V[V]. Commutation and the same vertical calculation give ΔJ=λ∏j≠iej∑VbVδ(V;(Lj∣V)j≠i),\Delta J=\frac{\lambda}{\prod_{j\ne i}e_j} \sum_V b_V\delta(V;(\mathcal L_j|_V)_{j\ne i}), with positive sign. This is the stated model mass; identifying it with a generic-fibre degree would require the separate specialization result. ◻