OpenMath

Proof source

Global first-energy Green calculus and relative Aubin–Mabuchi geometry

Complete proof text in the unsigned reading edition.

Download sourceBrowse research records

Reading edition of the complete source note. Original labels and explicit equation tags are retained; references use source titles where compiled numbering is unavailable. Current result status is recorded separately in the project register.

Contents of this source
  1. Global first-energy Green calculus and relative Aubin–Mabuchi geometry

Global first-energy Green calculus and relative Aubin–Mabuchi geometry

Throughout this section the domain and background satisfy precisely (TC1)–(TC2): VV is the strict original analytic Laurent box compactly contained in the interior of one original affinoid domain, and the arbitrary original symmetric middle-positive separately closed (p,p)(p,p)-current SS is defined on a neighborhood of its closed box. Put q=n−p≥1q=n-p\geq1, retain the negative bounded barrier ρ\rho of Lemma “An explicit bounded boundary barrier”, and abbreviate E1=ES,∂1(V)\mathcal E^1=\mathcal E^1_{S,\partial}(V). The background is not assumed strongly positive. The positive energy matrices used below come from rank-one original (1,1)(1,1) squares wedged with the same middle-positive background.

The compact bounded Green identity is the frozen R4 input eq:relative-green-energy in relative_energy_agent.tex, with the analytic local scope made explicit in Section “Analytic Laurent collars and bounded relative quasicontinuity”. The bounded common-collar Aubin–Mabuchi variation is already frozen in R3 energy_convex_blocks.tex, lem:r3-energy-midpoint. The new assertions here remove that fixed compact collar by the proved unbounded zero-boundary energy class. The complete argument keeps track of the global boundary term; it does not assume that it vanishes merely because a potential has pointwise zero boundary values.

Lemma (Global bounded Green exchange from finite curvature mass). Let QQ be any original symmetric middle-positive separately closed (n−1,n−1)(n-1,n-1)-current on VV. Let a,b≤0a,b\leq0 be bounded classical psh functions satisfying BG(a,0)\mathrm{BG}(a,0) and BG(b,0)\mathrm{BG}(b,0). Suppose the two positive measures D(aQ)D(aQ) and D(bQ)D(bQ) have finite total mass on VV. Then ∫V(−a)D(bQ)=∫V(−b)D(aQ),∫VΓQ(a,b)=∫V(−a)D(bQ),∫VΓQ(a)=∫V(−a)D(aQ)<∞.(GE1)\begin{split} \int_V(-a)D(bQ)&=\int_V(-b)D(aQ),\\ \int_V\Gamma_Q(a,b)&=\int_V(-a)D(bQ),\qquad \int_V\Gamma_Q(a)=\int_V(-a)D(aQ)<\infty. \end{split} \tag{GE1} The cross energy has finite total variation. These are identities of the actual original Borel multipliers and energy measures.

Status: NEW_PROVED as a global zero-boundary extension of compact bounded Green exchange.

Proof. For ϵ>0\epsilon>0 put aϵ=max⁡(a,−ϵ)a^\epsilon=\max(a,-\epsilon). The bounded difference a−aϵa-a^\epsilon is supported in the compact set Kϵ={a<−ϵ}‾⋐VK_\epsilon=\overline{\{a<-\epsilon\}}\Subset V. Compact Green exchange therefore gives ∫(a−aϵ)D(bQ)=∫bD(aQ)−∫bD(aϵQ).(GE2)\int(a-a^\epsilon)D(bQ) =\int bD(aQ)-\int bD(a^\epsilon Q). \tag{GE2} The measure D(aϵQ)D(a^\epsilon Q) has the same finite total mass as D(aQ)D(aQ). Indeed their difference is D((aϵ−a)Q)D((a^\epsilon-a)Q), with compactly supported multiplier. A compact original smooth cutoff equal to one on a neighborhood of that support has zero pairing with this derivative. Outside the compact support the two measures agree, by ordinary locality. This proves both finiteness and equality of their total masses.

Fix δ>0\delta>0. The set Lδ={b<−δ}‾L_\delta=\overline{\{b<-\delta\}} is compact in VV. Relative CLN for the fixed QQ and the normalized classical function (aϵ+ϵ)/ϵ∈[0,1](a^\epsilon+\epsilon)/\epsilon\in[0,1] gives D(aϵQ)(Lδ)≤Cδ,QϵD(a^\epsilon Q)(L_\delta)\leq C_{\delta,Q}\epsilon. Consequently ∣∫bD(aϵQ)∣≤∥b∥∞Cδ,Qϵ+δD(aQ)(V).(GE3)\left|\int bD(a^\epsilon Q)\right| \leq \|b\|_\infty C_{\delta,Q}\epsilon +\delta D(aQ)(V). \tag{GE3} First let ϵ↓0\epsilon\downarrow0, then δ↓0\delta\downarrow0. The last term in (GE2) tends to zero. Bounded convergence on its left proves the first equality in (GE1).

The original bounded clipping rule, including the zero energy of an exact scalar level, gives ΓQ(a−aϵ,a)=1{a<−ϵ}ΓQ(a)\Gamma_Q(a-a^\epsilon,a) =\mathbf1_{\{a<-\epsilon\}}\Gamma_Q(a). Apply compact Green to this difference and aa: ∫{a<−ϵ}ΓQ(a)=∫(aϵ−a)D(aQ).(GE4)\int_{\{a<-\epsilon\}}\Gamma_Q(a) =\int(a^\epsilon-a)D(aQ). \tag{GE4} As ϵ↓0\epsilon\downarrow0 the right side increases to the finite integral ∫(−a)D(aQ)\int(-a)D(aQ). The energy gives no mass to {a=0}\{a=0\}, so monotone convergence gives the diagonal energy identity. Apply it also to bb and a+ba+b; the latter has the same boundary property and finite curvature mass. Polarization and the first equality of (GE1) identify the cross integral. The measure Cauchy–Schwarz inequality bounds its total variation by the product of the square roots of the two finite diagonal masses. ◻

For u0,…,uq∈E1u_0,\ldots,u_q\in\mathcal E^1 define IS(u0,…,uq)=∫V(−u0)M(u1,…,uq).(GE5)\mathcal I_S(u_0,\ldots,u_q) =\int_V(-u_0)M(u_1,\ldots,u_q). \tag{GE5} It is finite by the already proved mixed estimate (PE11). Its interpretation uses the actual potentials; all mixed measures annihilate their joint pole union.

Theorem (Full symmetry of the global first mixed energy). The nonnegative functional (GE5) is symmetric in all q+1q+1 slots and separately additive and positively homogeneous on E1\mathcal E^1. In particular the undifferentiated slot may be exchanged with any differentiated slot without a boundary term. Every integral is over the whole ordinary domain VV.

Status: NEW_PROVED for the unbounded first-power boundary class and arbitrary original SS.

Proof. Use the decreasing finite-mass approximants from Theorem “Bounded finite-mass approximation of the power class”: bi,j=max⁡(ui,jρ)b_{i,j}=\max(u_i,j\rho). For a fixed jj they all lie in [−j,0][-j,0] and satisfy the compact-sublevel boundary condition. Let Aj=∑i=0qbi,jA_j=\sum_{i=0}^q b_{i,j}. Since Aj≥(q+1)jρA_j\geq(q+1)j\rho and ρ<0\rho<0, comparison with c(q+1)jρc(q+1)j\rho for c>1c>1 has strict gap all of VV and gives μAj(V)≤[(q+1)j]qμρ(V)<∞(GE6)\mu_{A_j}(V) \leq[(q+1)j]^q\mu_\rho(V)<\infty \tag{GE6} after c↓1c\downarrow1. The positive mixed expansion of μAj\mu_{A_j} shows that every mixed top measure formed from these bounded slots has finite total mass.

Fix the last q−1q-1 slots and use their bounded mixed product as QQ in Lemma “Global bounded Green exchange from finite curvature mass”. Its two curvature measures are finite by (GE6), so the lemma exchanges the first two slots. Symmetry among the differentiated slots is already part of bounded BT; hence the bounded mixed energy is fully symmetric.

The approximants in each slot have the fixed lower barrier ui∈E1u_i\in\mathcal E^1. The global critical mixed-energy continuity of Theorem “Global critical mixed-energy continuity above fixed lower barriers” therefore passes the bounded identity to (GE5), on the whole domain and with no escaped boundary mass. This proves full symmetry. Additivity and positive homogeneity in the first slot are ordinary integration identities; in the other slots they follow from the unbounded mixed-current calculus. The same symmetry transfers all of them to every slot. ◻

Theorem (Globally finite unbounded energy matrices). Let a,b,c2,…,cq∈E1a,b,c_2,\ldots,c_q\in\mathcal E^1 and put Q=N(c2,…,cq)Q=N(c_2,\ldots,c_q), with Q=SQ=S when q=1q=1. The graded energy currents at the endpoint σ=1\sigma=1 are finite measures on all of VV and satisfy ∫VΓQ(a,b)=IS(a,b,c2,…,cq),∣∫VΓQ(a,b)∣2≤IS(a,a,c2,…,cq)IS(b,b,c2,…,cq).(GE7)\begin{split} \int_V\Gamma_Q(a,b) &=\mathcal I_S(a,b,c_2,\ldots,c_q),\\ \left|\int_V\Gamma_Q(a,b)\right|^2 &\leq\mathcal I_S(a,a,c_2,\ldots,c_q) \mathcal I_S(b,b,c_2,\ldots,c_q). \end{split} \tag{GE7} Every cross energy has finite total variation, and the same matrix positivity holds for finite real linear combinations of the displayed potentials. Thus it applies to differences even though a difference need not be psh.

Status: NEW_PROVED as global finiteness and Green identification; local endpoint existence is the separate proved theorem (PG10)–(PG12).

Proof. Fix ϵ>0\epsilon>0 and put aϵ=max⁡(a,−ϵ)a^\epsilon=\max(a,-\epsilon). This bounded function belongs to E1\mathcal E^1 by solidity. For integers j>ϵj>\epsilon, write aj=max⁡(a,−j)a_j=\max(a,-j) and let QjQ_j use the same canonical level jj in its other slots. The difference aj−aϵa_j-a^\epsilon is supported in the one compact set Kϵ={a<−ϵ}‾K_\epsilon=\overline{\{a<-\epsilon\}}. The bounded compact Green formula and clipping locality give ∫{a<−ϵ}ΓQj(aj)=∫(aϵ−aj)D(ajQj).(GE8)\int_{\{a<-\epsilon\}}\Gamma_{Q_j}(a_j) =\int(a^\epsilon-a_j)D(a_jQ_j). \tag{GE8} Insert a fixed original compact cutoff equal to one near KϵK_\epsilon when reading this identity. The endpoint graded energy convergence (PG11) passes its left side in local total variation. Its indicator is independent of j>ϵj>\epsilon. The weighted curvature convergence (PG4) with a=1,k=qa=1,k=q passes its right side, since a+k=q+1=q+σa+k=q+1=q+\sigma for σ=1\sigma=1. Include the bounded slot aϵa^\epsilon in the fixed family; its canonical truncation is itself for large jj. Thus ∫{a<−ϵ}ΓQ(a)=∫(aϵ−a)M(a,c2,…,cq).(GE9)\int_{\{a<-\epsilon\}}\Gamma_Q(a) =\int(a^\epsilon-a)M(a,c_2,\ldots,c_q). \tag{GE9} The energy gives zero mass to the joint pole set by (PG12). It also gives no mass to {a=0}\{a=0\}: the corresponding bounded energies vanish there, and (PG11) is local total-variation convergence on the identical set {aj=0}={a=0}\{a_j=0\}=\{a=0\}. Let ϵ↓0\epsilon\downarrow0 in (GE9). Both sides increase, and the right side has finite limit by (PE11). This proves the global diagonal identity in (GE7).

Apply it to a+ba+b as well. The full symmetry from the preceding theorem and polarization give the cross identity. The positive local energy matrix and an increasing compact exhaustion give the global Cauchy–Schwarz inequality and the total-variation bound. No convergence of global unweighted energy masses was assumed in order to obtain their finiteness. ◻

Theorem (The constant-one first-energy Holder inequality). For all u0,…,uq∈E1u_0,\ldots,u_q\in\mathcal E^1, IS(u0,…,uq)≤∏i=0qe1(ui)1/(q+1).(GE10)\boxed{\quad \mathcal I_S(u_0,\ldots,u_q) \leq\prod_{i=0}^q e_1(u_i)^{1/(q+1)}. \quad} \tag{GE10} In particular e1(u+v)1/(q+1)≤e1(u)1/(q+1)+e1(v)1/(q+1).(GE11)e_1(u+v)^{1/(q+1)} \leq e_1(u)^{1/(q+1)}+e_1(v)^{1/(q+1)}. \tag{GE11} The constant one in (GE10) is optimal whenever a potential of positive first energy exists. This is an integral inequality; it makes no pointwise Alexandrov–Fenchel or hyperbolicity claim for the arbitrary middle-positive background.

Status: NEW_PROVED. The earlier (PE11) supplied a finite constant for every power and is used to justify integrability before this sharper first-power proof.

Proof. The zero-energy case already follows from (PE11), including its proved zero-factor limiting argument. Otherwise rescale each potential so that e1(ui)=1e_1(u_i)=1. Put d=q+1d=q+1. Among the finitely many mixed values Bα=IS(u0,…,u0⏟α0,…,uq,…,uq⏟αq),αi∈N,∑iαi=d,B_\alpha=\mathcal I_S( \underbrace{u_0,\ldots,u_0}_{\alpha_0},\ldots, \underbrace{u_q,\ldots,u_q}_{\alpha_q}), \qquad \alpha_i\in\mathbf N,\quad\sum_i\alpha_i=d, choose a largest value MM, and among its indices choose one maximizing ∑iαi2\sum_i\alpha_i^2. If this index has two positive entries αi,αl\alpha_i,\alpha_l, the two-slot energy Cauchy–Schwarz inequality (GE7), with all remaining slots fixed, gives M2=Bα2≤Bα+ei−elBα−ei+el≤M2.M^2=B_\alpha^2 \leq B_{\alpha+e_i-e_l}B_{\alpha-e_i+e_l} \leq M^2. Since a diagonal value is one, M≥1>0M\geq1>0, and both neighboring values must equal MM. At least one neighboring index has a strictly larger sum of squares: the two changes are 2(αi−αl)+22(\alpha_i-\alpha_l)+2 and its opposite plus 22. This contradicts the choice. Thus a maximizing index is pure, and M=e1(ui)=1M=e_1(u_i)=1. Every mixed value is at most one, proving (GE10) after undoing the normalization.

Expand e1(u+v)e_1(u+v) by the fully symmetric multilinearity of (GE5), and apply (GE10) term by term. The binomial theorem gives (GE11). If all slots equal one potential with positive energy, equality holds in (GE10), proving optimality. ◻

Proposition (The relative capacity quotient determines every original product). Let ui,vi∈FS,∂(V)u_i,v_i\in\mathcal F_{S,\partial}(V) for a finite family of indices, and suppose CS({ui≠vi};V)=0for every i.(GE12)C_S(\{u_i\ne v_i\};V)=0\quad\hbox{for every }i. \tag{GE12} Then every mixed original product of the uiu_i is equal to the corresponding product of the viv_i, in every degree 0≤k≤q0\leq k\leq q, against every original complementary test and after every actual moment refinement. For one pair, μu=μv⟺CS({u≠v};V)=0⟺CS∗({u≠v};V)=0.(GE13)\mu_u=\mu_v \quad\Longleftrightarrow\quad C_S(\{u\ne v\};V)=0 \quad\Longleftrightarrow\quad C_S^*(\{u\ne v\};V)=0. \tag{GE13} If one member of a pair in (GE13) belongs to ES,∂σ\mathcal E^\sigma_{S,\partial}, so does the other, with the same power energy. All defined graded energy currents and integrable weighted products agree after these replacements.

Status: NEW_PROVED as a representative quotient within the specified actual-potential classes. It is not the analytic PS/PL relation quotient GQ.

Proof. Let EE be the finite union of the exceptional Borel sets in (GE12); it has zero raw capacity. At each canonical level jj, the bounded differences ui,j−vi,ju_{i,j}-v_{i,j} vanish outside EE. Every bounded mixed background QjQ_j over SS gives EE zero variation in each original coefficient by the global coefficient bound (DG3), including degree zero. Therefore (ui,j−vi,j)Qj=0(u_{i,j}-v_{i,j})Q_j=0 as an original current. The bounded telescoping formula ∏i=1kDui,j∧S−∏i=1kDvi,j∧S=∑i=1kD((ui,j−vi,j)Qi,j)(GE14)\prod_{i=1}^k Du_{i,j}\wedge S- \prod_{i=1}^k Dv_{i,j}\wedge S =\sum_{i=1}^k D\bigl((u_{i,j}-v_{i,j})Q_{i,j}\bigr) \tag{GE14} is zero; Qi,jQ_{i,j} uses the already replaced slots before ii and the unreplaced slots after it. Pass the uncut local coefficient total-variation limits from the all-degree theorem. This proves product equality without assuming a locality principle on an arbitrary equality Borel set.

For a single pair, equality of the top measures gives both one-sided outer-capacity domination statements from Corollary cor:r11-boundary-outer-domination. Their boundary-gap hypotheses hold because both potentials are negative and have compact negative sublevels. Countable subadditivity gives outer capacity zero of their inequality set. Outer zero implies raw zero, and the proved product equality gives the remaining implication in (GE13).

All products annihilate EE in their coefficient variations, again by (DG3) and the increasing cut construction. Thus the actual multiplier values of the two families agree almost everywhere for each such coefficient. This preserves integrability, power energies and weighted products. The graded energy formula (PG10) then proves energy-current equality wherever its degree hypotheses are satisfied. ◻

Theorem (Global Aubin–Mabuchi variation and energy rigidity). For u∈E1u\in\mathcal E^1 put AS(u)=1q+1∫Vu dμu=−e1(u)q+1.(GE15)\mathscr A_S(u)=\frac1{q+1}\int_V u\,d\mu_u =-\frac{e_1(u)}{q+1}. \tag{GE15} For u,v∈E1u,v\in\mathcal E^1, let h=v−uh=v-u and ut=(1−t)u+tvu_t=(1-t)u+tv, 0≤t≤10\leq t\leq1. Then AS(ut)\mathscr A_S(u_t) is a polynomial and ddtAS(ut)=∫Vh dμut,d2dt2AS(ut)=−q∫VΓN(utq−1)(h)≤0.(GE16)\begin{split} \frac d{dt}\mathscr A_S(u_t)&=\int_V h\,d\mu_{u_t},\\ \frac {d^2}{dt^2}\mathscr A_S(u_t) &=-q\int_V\Gamma_{N(u_t^{q-1})}(h)\leq0. \end{split} \tag{GE16} All displayed signed integrals are absolutely convergent. The functional is increasing under pointwise order, and ∫(v−u) dμv≤AS(v)−AS(u)≤∫(v−u) dμu.(GE17)\int(v-u)\,d\mu_v \leq\mathscr A_S(v)-\mathscr A_S(u) \leq\int(v-u)\,d\mu_u. \tag{GE17} The symmetric discrepancy IS(u,v):=∫V(u−v)(dμv−dμu)=∑l=0q−1∫VΓN(ul,vq−1−l)(u−v)≥0(GE18)\begin{split} \mathscr I_S(u,v) &:=\int_V(u-v)(d\mu_v-d\mu_u)\\ &=\sum_{l=0}^{q-1} \int_V\Gamma_{N(u^l,v^{q-1-l})}(u-v)\geq0 \end{split} \tag{GE18} vanishes exactly when the equivalent conditions (GE13) hold. Consequently AS\mathscr A_S is strictly concave on the relative capacity quotient: equality with its chord at one strict interior parameter holds exactly for equivalent potentials. No pointwise separation or positivity on every ordinary open set is assumed for the background.

Status: NEW_PROVED as the global unbounded energy extension and quotient rigidity. The bounded common-collar variation formula itself is an INTERNAL_INPUT, as specified above.

Proof. Full symmetric multilinearity of (GE5) makes the energy a polynomial along the segment. Differentiating its q+1q+1 equal slots gives the first formula in (GE16). Differentiating once more gives q∫hD(hN(utq−1))q\int hD(hN(u_t^{q-1})). The global energy matrix identity (GE7), extended by bilinearity to the difference hh, is exactly its negative in (GE16). Finiteness follows from (GE10) in each expanded mixed term. Concavity follows from matrix positivity, and the tangent inequalities (GE17) follow by integrating the decreasing first derivative. If u≤vu\leq v the first derivative is nonnegative, proving monotonicity.

The telescoping identity for μv−μu\mu_v-\mu_u and the same global Green identity give (GE18). Suppose its sum is zero. Every summand is nonnegative, so for each Ql=N(ul,vq−1−l)Q_l=N(u^l,v^{q-1-l}) one has ΓQl(h)(V)=0\Gamma_{Q_l}(h)(V)=0. For an original compact smooth scalar test φ\varphi, choose an original compact cutoff χ\chi equal to one on its differential support, and use canonical bounded approximants hj=vj−ujh_j=v_j-u_j and Ql,jQ_{l,j}. Compact integration by parts and energy Cauchy–Schwarz give ∣⟨D(hjQl,j),φ⟩∣2≤(∫χΓQl,j(hj))(∫χΓQl,j(φ)).(GE19)|\langle D(h_jQ_{l,j}),\varphi\rangle|^2 \leq\left(\int\chi\Gamma_{Q_{l,j}}(h_j)\right) \left(\int\chi\Gamma_{Q_{l,j}}(\varphi)\right). \tag{GE19} The first factor tends to zero by the endpoint local energy total-variation convergence (PG11). The second is bounded: its scalar differentials are fixed original smooth forms, and Ql,jQ_{l,j} has local coefficient total-variation convergence. The weighted convergence (PG4) also gives hjQl,j→hQlh_jQ_{l,j}\to hQ_l, so (GE19) proves D(hQl)=0D(hQ_l)=0 against every original test. Summing the recursive telescoping formula yields μu=μv\mu_u=\mu_v. Conversely this equality makes the defining expression in (GE18) zero. Proposition “The relative capacity quotient determines every original product” gives precisely the other equivalences.

If the energy agrees with its chord at one interior point, concavity forces it to be affine on the whole segment. Its endpoint derivatives are equal, so (GE16) gives IS(u,v)=AS′(0)−AS′(1)=0\mathscr I_S(u,v)=\mathscr A_S'(0)-\mathscr A_S'(1)=0. Conversely equivalent potentials have identical mixed products and energies by the preceding proposition, and the energy is constant on their segment. This proves the claimed strict concavity on the quotient. ◻

Corollary (The variational necessary condition retains its domain gate). Let μ\mu be a positive Radon datum such that ∫(−v) dμ<∞\int(-v)\,d\mu<\infty for every v∈E1v\in\mathcal E^1. If u∈E1u\in\mathcal E^1 solves μu=μ\mu_u=\mu, then it maximizes v⟼AS(v)−∫v dμ(GE20)v\longmapsto\mathscr A_S(v)-\int v\,d\mu \tag{GE20} on E1\mathcal E^1. Conversely, a maximizer satisfies the current equation provided that for every original compact smooth scalar φ\varphi there is ϵφ>0\epsilon_\varphi>0 such that u+tφ∈E1u+t\varphi\in\mathcal E^1 for all ∣t∣<ϵφ|t|<\epsilon_\varphi. This last variation hypothesis is explicit and is not asserted for a general classical psh potential or arbitrary datum.

Status: DIRECT_COROLLARY for necessity and CONDITIONAL for the stated converse.

Proof. The upper tangent inequality in (GE17) gives necessity. Under the printed two-sided variation hypothesis, apply the first derivative identity to the admissible segment u+tφu+t\varphi. Maximality for both signs gives ∫φ dμu=∫φ dμ\int\varphi\,d\mu_u=\int\varphi\,d\mu for every original compact smooth scalar test. This is exactly the original current equation. No envelope differentiability or unstated admissibility of arbitrary smooth perturbations is used. ◻

Scope and rejected shortcut.

The tempting inference from first-energy dual boundedness to an unrestricted ambient classical Dirichlet solution remains FALSE, by Corollary “Finite dual energy does not imply ambient classical existence”. Concavity alone supplies no compactness or representability theorem for a maximizing family. The two-sided variation gate in the last corollary cannot be discarded by analogy with a smooth strictly elliptic problem. The relative capacity quotient in (GE12)–(GE14) identifies actual potentials in the proved boundary classes; it does not identify analytic PS/PL presentations, does not enlarge the class of delta tests, and does not close GQ or low-real-degree CTO.

One possible alternate approximation is max⁡(ui−ϵj,jρ)\max(u_i-\epsilon_j,j\rho), which imposes a common ordinary collar. That route requires separate uniform control of the small-value energy near the moving collar. The proof above instead uses (GE3) and the already proved decreasing finite-mass approximation; no claim about the alternate route is needed.

References requiring source context

Some references in this note depend on its original surrounding manuscript. Their source labels are retained in the reading edition; its TeX is available above.

Source text and edition identity

This reading copy omits personal attribution and places existing mathematical question marks over their relation symbols. The mathematical content and source status are retained.

Reading-copy SHA-256 92f6ec7ecf9274f2698e5048465c6cf3b5949a0a7c6e770a659a5ee68b8d53a3

Frozen-original SHA-256 92f6ec7ecf9274f2698e5048465c6cf3b5949a0a7c6e770a659a5ee68b8d53a3

\section{Global first-energy Green calculus and relative Aubin--Mabuchi geometry}
\label{sec:r11-global-first-energy}

Throughout this section the domain and background satisfy precisely
(TC1)--(TC2): $V$ is the strict original analytic Laurent box compactly
contained in the interior of one original affinoid domain, and the
arbitrary original symmetric middle-positive separately closed
$(p,p)$-current $S$ is defined on a neighborhood of its closed box.
Put $q=n-p\geq1$, retain the negative bounded barrier $\rho$ of
Lemma~\ref{lem:r11-strict-Laurent-barrier}, and abbreviate
$\mathcal E^1=\mathcal E^1_{S,\partial}(V)$.
The background is not assumed strongly positive. The positive energy
matrices used below come from rank-one original $(1,1)$ squares
wedged with the same middle-positive background.

The compact bounded Green identity is the frozen R4 input
\texttt{eq:relative-green-energy} in
\texttt{relative\_energy\_agent.tex}, with the analytic local
scope made explicit in Section~\ref{sec:r11-analytic-Laurent-collars}.
The bounded common-collar Aubin--Mabuchi variation is already frozen
in R3 \texttt{energy\_convex\_blocks.tex},
\texttt{lem:r3-energy-midpoint}. The new assertions here remove that
fixed compact collar by the proved unbounded zero-boundary energy
class. The complete argument keeps track of the global boundary
term; it does not assume that it vanishes merely because a potential
has pointwise zero boundary values.

\begin{lemma}[Global bounded Green exchange from finite curvature mass]
\label{lem:r11-global-bounded-Green}
Let $Q$ be any original symmetric middle-positive separately closed
$(n-1,n-1)$-current on $V$. Let $a,b\leq0$ be bounded classical
psh functions satisfying $\mathrm{BG}(a,0)$ and
$\mathrm{BG}(b,0)$. Suppose the two positive measures
$D(aQ)$ and $D(bQ)$ have finite total mass on $V$. Then
\[
 \begin{split}
  \int_V(-a)D(bQ)&=\int_V(-b)D(aQ),\\
  \int_V\Gamma_Q(a,b)&=\int_V(-a)D(bQ),\qquad
  \int_V\Gamma_Q(a)=\int_V(-a)D(aQ)<\infty.
 \end{split}
 \tag{GE1}
\]
The cross energy has finite total variation. These are identities
of the actual original Borel multipliers and energy measures.
\end{lemma}

\noindent\textbf{Status: \texttt{NEW\_PROVED}} as a global
zero-boundary extension of compact bounded Green exchange.

\begin{proof}
For $\epsilon>0$ put $a^\epsilon=\max(a,-\epsilon)$.
The bounded difference $a-a^\epsilon$ is supported in the compact
set $K_\epsilon=\overline{\{a<-\epsilon\}}\Subset V$.
Compact Green exchange therefore gives
\[
 \int(a-a^\epsilon)D(bQ)
     =\int bD(aQ)-\int bD(a^\epsilon Q).
 \tag{GE2}
\]
The measure $D(a^\epsilon Q)$ has the same finite total mass as
$D(aQ)$. Indeed their difference is $D((a^\epsilon-a)Q)$,
with compactly supported multiplier. A compact original smooth
cutoff equal to one on a neighborhood of that support has zero
pairing with this derivative. Outside the compact support the
two measures agree, by ordinary locality. This proves both
finiteness and equality of their total masses.

Fix $\delta>0$. The set
$L_\delta=\overline{\{b<-\delta\}}$ is compact in $V$.
Relative CLN for the fixed $Q$ and the normalized classical
function $(a^\epsilon+\epsilon)/\epsilon\in[0,1]$ gives
$D(a^\epsilon Q)(L_\delta)\leq C_{\delta,Q}\epsilon$.
Consequently
\[
 \left|\int bD(a^\epsilon Q)\right|
 \leq \|b\|_\infty C_{\delta,Q}\epsilon
                     +\delta D(aQ)(V).
 \tag{GE3}
\]
First let $\epsilon\downarrow0$, then $\delta\downarrow0$.
The last term in (GE2) tends to zero. Bounded convergence on
its left proves the first equality in (GE1).

The original bounded clipping rule, including the zero energy
of an exact scalar level, gives
$\Gamma_Q(a-a^\epsilon,a)
 =\mathbf1_{\{a<-\epsilon\}}\Gamma_Q(a)$.
Apply compact Green to this difference and $a$:
\[
 \int_{\{a<-\epsilon\}}\Gamma_Q(a)
                 =\int(a^\epsilon-a)D(aQ).
 \tag{GE4}
\]
As $\epsilon\downarrow0$ the right side increases to the
finite integral $\int(-a)D(aQ)$. The energy gives no mass
to $\{a=0\}$, so monotone convergence gives the diagonal
energy identity. Apply it also to $b$ and $a+b$; the latter
has the same boundary property and finite curvature mass.
Polarization and the first equality of (GE1) identify the
cross integral. The measure Cauchy--Schwarz inequality bounds
its total variation by the product of the square roots of the
two finite diagonal masses.
\end{proof}

For $u_0,\ldots,u_q\in\mathcal E^1$ define
\[
 \mathcal I_S(u_0,\ldots,u_q)
       =\int_V(-u_0)M(u_1,\ldots,u_q).
 \tag{GE5}
\]
It is finite by the already proved mixed estimate (PE11).
Its interpretation uses the actual potentials; all mixed
measures annihilate their joint pole union.

\begin{theorem}[Full symmetry of the global first mixed energy]
\label{thm:r11-global-first-mixed-symmetry}
The nonnegative functional (GE5) is symmetric in all $q+1$
slots and separately additive and positively homogeneous on
$\mathcal E^1$. In particular the undifferentiated slot may
be exchanged with any differentiated slot without a boundary
term. Every integral is over the whole ordinary domain $V$.
\end{theorem}

\noindent\textbf{Status: \texttt{NEW\_PROVED}} for the
unbounded first-power boundary class and arbitrary original $S$.

\begin{proof}
Use the decreasing finite-mass approximants from
Theorem~\ref{thm:r11-finite-mass-power-approximation}:
$b_{i,j}=\max(u_i,j\rho)$. For a fixed $j$ they all lie
in $[-j,0]$ and satisfy the compact-sublevel boundary condition.
Let $A_j=\sum_{i=0}^q b_{i,j}$. Since
$A_j\geq(q+1)j\rho$ and $\rho<0$, comparison with
$c(q+1)j\rho$ for $c>1$ has strict gap all of $V$ and gives
\[
          \mu_{A_j}(V)
              \leq[(q+1)j]^q\mu_\rho(V)<\infty
 \tag{GE6}
\]
after $c\downarrow1$. The positive mixed expansion of
$\mu_{A_j}$ shows that every mixed top measure formed from
these bounded slots has finite total mass.

Fix the last $q-1$ slots and use their bounded mixed product
as $Q$ in Lemma~\ref{lem:r11-global-bounded-Green}. Its two
curvature measures are finite by (GE6), so the lemma exchanges
the first two slots. Symmetry among the differentiated slots
is already part of bounded BT; hence the bounded mixed energy
is fully symmetric.

The approximants in each slot have the fixed lower barrier
$u_i\in\mathcal E^1$. The global critical mixed-energy
continuity of Theorem~\ref{thm:r11-critical-energy-barrier-continuity}
therefore passes the bounded identity to (GE5), on the whole
domain and with no escaped boundary mass. This proves full
symmetry. Additivity and positive homogeneity in the first
slot are ordinary integration identities; in the other slots
they follow from the unbounded mixed-current calculus. The
same symmetry transfers all of them to every slot.
\end{proof}

\begin{theorem}[Globally finite unbounded energy matrices]
\label{thm:r11-global-first-energy-matrix}
Let $a,b,c_2,\ldots,c_q\in\mathcal E^1$ and put
$Q=N(c_2,\ldots,c_q)$, with $Q=S$ when $q=1$.
The graded energy currents at the endpoint $\sigma=1$ are
finite measures on all of $V$ and satisfy
\[
 \begin{split}
 \int_V\Gamma_Q(a,b)
       &=\mathcal I_S(a,b,c_2,\ldots,c_q),\\
 \left|\int_V\Gamma_Q(a,b)\right|^2
       &\leq\mathcal I_S(a,a,c_2,\ldots,c_q)
                   \mathcal I_S(b,b,c_2,\ldots,c_q).
 \end{split}
 \tag{GE7}
\]
Every cross energy has finite total variation, and the same
matrix positivity holds for finite real linear combinations
of the displayed potentials. Thus it applies to differences
even though a difference need not be psh.
\end{theorem}

\noindent\textbf{Status: \texttt{NEW\_PROVED}} as global
finiteness and Green identification; local endpoint existence
is the separate proved theorem (PG10)--(PG12).

\begin{proof}
Fix $\epsilon>0$ and put $a^\epsilon=\max(a,-\epsilon)$.
This bounded function belongs to $\mathcal E^1$ by solidity.
For integers $j>\epsilon$, write $a_j=\max(a,-j)$ and
let $Q_j$ use the same canonical level $j$ in its other slots.
The difference $a_j-a^\epsilon$ is supported in the one
compact set $K_\epsilon=\overline{\{a<-\epsilon\}}$.
The bounded compact Green formula and clipping locality give
\[
 \int_{\{a<-\epsilon\}}\Gamma_{Q_j}(a_j)
                =\int(a^\epsilon-a_j)D(a_jQ_j).
 \tag{GE8}
\]
Insert a fixed original compact cutoff equal to one near
$K_\epsilon$ when reading this identity. The endpoint graded
energy convergence (PG11) passes its left side in local total
variation. Its indicator is independent of $j>\epsilon$.
The weighted curvature convergence (PG4) with $a=1,k=q$
passes its right side, since $a+k=q+1=q+\sigma$ for
$\sigma=1$. Include the bounded slot $a^\epsilon$ in the
fixed family; its canonical truncation is itself for large $j$.
Thus
\[
 \int_{\{a<-\epsilon\}}\Gamma_Q(a)
           =\int(a^\epsilon-a)M(a,c_2,\ldots,c_q).
 \tag{GE9}
\]
The energy gives zero mass to the joint pole set by (PG12).
It also gives no mass to $\{a=0\}$: the corresponding bounded
energies vanish there, and (PG11) is local total-variation
convergence on the identical set $\{a_j=0\}=\{a=0\}$.
Let $\epsilon\downarrow0$ in (GE9). Both sides increase,
and the right side has finite limit by (PE11). This proves
the global diagonal identity in (GE7).

Apply it to $a+b$ as well. The full symmetry from the
preceding theorem and polarization give the cross identity.
The positive local energy matrix and an increasing compact
exhaustion give the global Cauchy--Schwarz inequality and the
total-variation bound. No convergence of global unweighted
energy masses was assumed in order to obtain their finiteness.
\end{proof}

\begin{theorem}[The constant-one first-energy Holder inequality]
\label{thm:r11-optimal-first-energy-Holder}
For all $u_0,\ldots,u_q\in\mathcal E^1$,
\[
 \boxed{\quad
 \mathcal I_S(u_0,\ldots,u_q)
                    \leq\prod_{i=0}^q e_1(u_i)^{1/(q+1)}.
 \quad}
 \tag{GE10}
\]
In particular
\[
 e_1(u+v)^{1/(q+1)}
           \leq e_1(u)^{1/(q+1)}+e_1(v)^{1/(q+1)}.
 \tag{GE11}
\]
The constant one in (GE10) is optimal whenever a potential
of positive first energy exists. This is an integral inequality;
it makes no pointwise Alexandrov--Fenchel or hyperbolicity
claim for the arbitrary middle-positive background.
\end{theorem}

\noindent\textbf{Status: \texttt{NEW\_PROVED}}. The earlier
(PE11) supplied a finite constant for every power and is used
to justify integrability before this sharper first-power proof.

\begin{proof}
The zero-energy case already follows from (PE11), including
its proved zero-factor limiting argument. Otherwise rescale
each potential so that $e_1(u_i)=1$. Put $d=q+1$.
Among the finitely many mixed values
\[
 B_\alpha=\mathcal I_S(
             \underbrace{u_0,\ldots,u_0}_{\alpha_0},\ldots,
             \underbrace{u_q,\ldots,u_q}_{\alpha_q}),
       \qquad \alpha_i\in\mathbf N,\quad\sum_i\alpha_i=d,
\]
choose a largest value $M$, and among its indices choose one
maximizing $\sum_i\alpha_i^2$. If this index has two positive
entries $\alpha_i,\alpha_l$, the two-slot energy
Cauchy--Schwarz inequality (GE7), with all remaining slots
fixed, gives
\[
       M^2=B_\alpha^2
           \leq B_{\alpha+e_i-e_l}B_{\alpha-e_i+e_l}
           \leq M^2.
\]
Since a diagonal value is one, $M\geq1>0$, and both
neighboring values must equal $M$. At least one neighboring
index has a strictly larger sum of squares: the two changes
are $2(\alpha_i-\alpha_l)+2$ and its opposite plus $2$.
This contradicts the choice. Thus a maximizing index is
pure, and $M=e_1(u_i)=1$. Every mixed value is at most one,
proving (GE10) after undoing the normalization.

Expand $e_1(u+v)$ by the fully symmetric multilinearity of
(GE5), and apply (GE10) term by term. The binomial theorem
gives (GE11). If all slots equal one potential with positive
energy, equality holds in (GE10), proving optimality.
\end{proof}

\begin{proposition}[The relative capacity quotient determines every original product]
\label{prop:r11-capacity-representative-quotient}
Let $u_i,v_i\in\mathcal F_{S,\partial}(V)$ for a finite
family of indices, and suppose
\[
       C_S(\{u_i\ne v_i\};V)=0\quad\hbox{for every }i.
 \tag{GE12}
\]
Then every mixed original product of the $u_i$ is equal to
the corresponding product of the $v_i$, in every degree
$0\leq k\leq q$, against every original complementary test
and after every actual moment refinement. For one pair,
\[
 \mu_u=\mu_v
  \quad\Longleftrightarrow\quad C_S(\{u\ne v\};V)=0
  \quad\Longleftrightarrow\quad C_S^*(\{u\ne v\};V)=0.
 \tag{GE13}
\]
If one member of a pair in (GE13) belongs to
$\mathcal E^\sigma_{S,\partial}$, so does the other, with
the same power energy. All defined graded energy currents
and integrable weighted products agree after these replacements.
\end{proposition}

\noindent\textbf{Status: \texttt{NEW\_PROVED}} as a
representative quotient within the specified actual-potential
classes. It is not the analytic PS/PL relation quotient GQ.

\begin{proof}
Let $E$ be the finite union of the exceptional Borel sets in
(GE12); it has zero raw capacity. At each canonical level $j$,
the bounded differences $u_{i,j}-v_{i,j}$ vanish outside $E$.
Every bounded mixed background $Q_j$ over $S$ gives $E$ zero
variation in each original coefficient by the global
coefficient bound (DG3), including degree zero. Therefore
$(u_{i,j}-v_{i,j})Q_j=0$ as an original current. The bounded
telescoping formula
\[
 \prod_{i=1}^k Du_{i,j}\wedge S-
 \prod_{i=1}^k Dv_{i,j}\wedge S
       =\sum_{i=1}^k D\bigl((u_{i,j}-v_{i,j})Q_{i,j}\bigr)
 \tag{GE14}
\]
is zero; $Q_{i,j}$ uses the already replaced slots before
$i$ and the unreplaced slots after it. Pass the uncut local
coefficient total-variation limits from the all-degree theorem.
This proves product equality without assuming a locality
principle on an arbitrary equality Borel set.

For a single pair, equality of the top measures gives both
one-sided outer-capacity domination statements from
Corollary~\ref{cor:r11-boundary-outer-domination}. Their
boundary-gap hypotheses hold because both potentials are
negative and have compact negative sublevels. Countable
subadditivity gives outer capacity zero of their inequality
set. Outer zero implies raw zero, and the proved product
equality gives the remaining implication in (GE13).

All products annihilate $E$ in their coefficient variations,
again by (DG3) and the increasing cut construction. Thus
the actual multiplier values of the two families agree
almost everywhere for each such coefficient. This preserves
integrability, power energies and weighted products.
The graded energy formula (PG10) then proves energy-current
equality wherever its degree hypotheses are satisfied.
\end{proof}

\begin{theorem}[Global Aubin--Mabuchi variation and energy rigidity]
\label{thm:r11-global-Aubin-Mabuchi}
For $u\in\mathcal E^1$ put
\[
             \mathscr A_S(u)=\frac1{q+1}\int_V u\,d\mu_u
                              =-\frac{e_1(u)}{q+1}.
 \tag{GE15}
\]
For $u,v\in\mathcal E^1$, let $h=v-u$ and
$u_t=(1-t)u+tv$, $0\leq t\leq1$. Then $\mathscr A_S(u_t)$
is a polynomial and
\[
 \begin{split}
 \frac d{dt}\mathscr A_S(u_t)&=\int_V h\,d\mu_{u_t},\\
 \frac {d^2}{dt^2}\mathscr A_S(u_t)
              &=-q\int_V\Gamma_{N(u_t^{q-1})}(h)\leq0.
 \end{split}
 \tag{GE16}
\]
All displayed signed integrals are absolutely convergent.
The functional is increasing under pointwise order, and
\[
 \int(v-u)\,d\mu_v
    \leq\mathscr A_S(v)-\mathscr A_S(u)
    \leq\int(v-u)\,d\mu_u.
 \tag{GE17}
\]
The symmetric discrepancy
\[
 \begin{split}
 \mathscr I_S(u,v)
    &:=\int_V(u-v)(d\mu_v-d\mu_u)\\
    &=\sum_{l=0}^{q-1}
       \int_V\Gamma_{N(u^l,v^{q-1-l})}(u-v)\geq0
 \end{split}
 \tag{GE18}
\]
vanishes exactly when the equivalent conditions (GE13) hold.
Consequently $\mathscr A_S$ is strictly concave on the
relative capacity quotient: equality with its chord at one
strict interior parameter holds exactly for equivalent
potentials. No pointwise separation or positivity on every
ordinary open set is assumed for the background.
\end{theorem}

\noindent\textbf{Status: \texttt{NEW\_PROVED}} as the
global unbounded energy extension and quotient rigidity.
The bounded common-collar variation formula itself is an
\texttt{INTERNAL\_INPUT}, as specified above.

\begin{proof}
Full symmetric multilinearity of (GE5) makes the energy a
polynomial along the segment. Differentiating its $q+1$
equal slots gives the first formula in (GE16). Differentiating
once more gives
$q\int hD(hN(u_t^{q-1}))$. The global energy matrix identity
(GE7), extended by bilinearity to the difference $h$, is
exactly its negative in (GE16). Finiteness follows from
(GE10) in each expanded mixed term. Concavity follows from
matrix positivity, and the tangent inequalities (GE17) follow
by integrating the decreasing first derivative. If $u\leq v$
the first derivative is nonnegative, proving monotonicity.

The telescoping identity for $\mu_v-\mu_u$ and the same
global Green identity give (GE18). Suppose its sum is zero.
Every summand is nonnegative, so for each
$Q_l=N(u^l,v^{q-1-l})$ one has
$\Gamma_{Q_l}(h)(V)=0$.
For an original compact smooth scalar test $\varphi$, choose
an original compact cutoff $\chi$ equal to one on its
differential support, and use canonical bounded approximants
$h_j=v_j-u_j$ and $Q_{l,j}$. Compact integration by parts
and energy Cauchy--Schwarz give
\[
 |\langle D(h_jQ_{l,j}),\varphi\rangle|^2
       \leq\left(\int\chi\Gamma_{Q_{l,j}}(h_j)\right)
             \left(\int\chi\Gamma_{Q_{l,j}}(\varphi)\right).
 \tag{GE19}
\]
The first factor tends to zero by the endpoint local
energy total-variation convergence (PG11). The second is
bounded: its scalar differentials are fixed original smooth
forms, and $Q_{l,j}$ has local coefficient total-variation
convergence. The weighted convergence (PG4) also gives
$h_jQ_{l,j}\to hQ_l$, so (GE19) proves $D(hQ_l)=0$
against every original test. Summing the recursive telescoping
formula yields $\mu_u=\mu_v$. Conversely this equality
makes the defining expression in (GE18) zero. Proposition
\ref{prop:r11-capacity-representative-quotient} gives precisely
the other equivalences.

If the energy agrees with its chord at one interior point,
concavity forces it to be affine on the whole segment.
Its endpoint derivatives are equal, so (GE16) gives
$\mathscr I_S(u,v)=\mathscr A_S'(0)-\mathscr A_S'(1)=0$.
Conversely equivalent potentials have identical mixed products
and energies by the preceding proposition, and the energy is
constant on their segment. This proves the claimed strict
concavity on the quotient.
\end{proof}

\begin{corollary}[The variational necessary condition retains its domain gate]
\label{cor:r11-variational-necessary-condition}
Let $\mu$ be a positive Radon datum such that
$\int(-v)\,d\mu<\infty$ for every $v\in\mathcal E^1$.
If $u\in\mathcal E^1$ solves $\mu_u=\mu$, then it maximizes
\[
                   v\longmapsto\mathscr A_S(v)-\int v\,d\mu
 \tag{GE20}
\]
on $\mathcal E^1$. Conversely, a maximizer satisfies the
current equation provided that for every original compact
smooth scalar $\varphi$ there is $\epsilon_\varphi>0$
such that $u+t\varphi\in\mathcal E^1$ for all
$|t|<\epsilon_\varphi$. This last variation hypothesis is
explicit and is not asserted for a general classical psh
potential or arbitrary datum.
\end{corollary}

\noindent\textbf{Status: \texttt{DIRECT\_COROLLARY}} for
necessity and \texttt{CONDITIONAL} for the stated converse.

\begin{proof}
The upper tangent inequality in (GE17) gives necessity.
Under the printed two-sided variation hypothesis, apply
the first derivative identity to the admissible segment
$u+t\varphi$. Maximality for both signs gives
$\int\varphi\,d\mu_u=\int\varphi\,d\mu$ for every
original compact smooth scalar test. This is exactly the
original current equation. No envelope differentiability or
unstated admissibility of arbitrary smooth perturbations is
used.
\end{proof}

\paragraph{Scope and rejected shortcut.}
The tempting inference from first-energy dual boundedness to
an unrestricted ambient classical Dirichlet solution remains
\texttt{FALSE}, by Corollary~\ref{cor:r11-dual-energy-insufficient}.
Concavity alone supplies no compactness or representability
theorem for a maximizing family. The two-sided variation gate
in the last corollary cannot be discarded by analogy with a
smooth strictly elliptic problem. The relative capacity quotient
in (GE12)--(GE14) identifies actual potentials in the proved
boundary classes; it does not identify analytic PS/PL
presentations, does not enlarge the class of delta tests, and
does not close GQ or low-real-degree CTO.

One possible alternate approximation is
$\max(u_i-\epsilon_j,j\rho)$, which imposes a common ordinary
collar. That route requires separate uniform control of the
small-value energy near the moving collar. The proof above
instead uses (GE3) and the already proved decreasing finite-mass
approximation; no claim about the alternate route is needed.