Reading guide

This book is organized by mathematical constructions. The original CLD calculus, its 2025 expansion, and the subsequent work of Gubler and his collaborators are read together at the point where they address the same problem. The two CLD versions retain separate names throughout. Their definitions of psh, their conventions for the involution, and the scope of several positivity assertions differ. The source concordance at the end records the fixed versions and the qualifications needed when using them.

The intended reader knows algebraic geometry, basic valuation theory and elementary differential forms. Familiarity with Berkovich spaces is helpful but not assumed for the first chapters. The analytic foundations are developed to the extent needed for the later constructions: completed residue fields, reduction, affinoid domains, tropical multiplicities and skeleta. Deeper existence and comparison theorems are quoted with their hypotheses. This permits a proof to remain readable without hiding its geometric input.

Reading the analytic foundations

The first four chapters now form a continuous introduction to the analytic geometry used throughout the book. They should be read in order on a first encounter. Each construction is followed far enough to identify its role in a later proof. The repeated examples are the closed disc, the annulus, the radius-rr field and a formal blow-up of a disc.

Chapter Central construction and its later use
1 Convergent algebras, Weierstrass division, seminorm points and completed residue fields. The nilpotent example separates points from multiplicity; adjoining a radius prepares the non-strict theory.
2 Domains and gluing, local rings, analytic fibres, boundary, flatness and analytification. The Galois product and the boundary obstruction to naive flatness explain why algebraic-looking formulas need analytic hypotheses.
3 Canonical and graded reduction, Shilov boundary, models, Abhyankar dimension and germ reduction. Formal frames then produce model metrics; residual geometry will later supply intersection degrees.
4 The four types of points, annular homotopy, residual curves, semistable skeleta and the slope formula. This gives the local geometric meaning of tropical directions before general tropical geometry begins.

For a first pass, follow the explicit proofs of Gauss multiplicativity, division, compactness, Laurent gluing, the finite-fibre length formula, the norm/model example, the Gauss residue-dimension formula and the annular retraction. Tate acyclicity, general analytic flatness, GAGA, germ reduction and semistable reduction are stated with precise source inputs. Reading their statements is enough to continue; their full foundational proofs belong to the cited monographs.

Readers returning from version 0.5 should note the numbering change. The former first chapter is replaced by Chapters 1–4; the former Chapters 2–21 are now Chapters 5–24. The Gauss residue-dimension proof and the singleton Shilov criterion have moved from the former Chapter 3 into Chapter 3. Their links and statement identifiers have been retained. Old theorem numbers should therefore be cited with their old version, not silently transferred to this edition.

The order of the book

Chapters 1–6 explain how analytic geometry supplies a tropical chart with integral tangent directions and multiplicities. Chapters 7–10 pass from these charts to forms, currents, integration and divisor identities. Chapters 11–14 study potentials, approximation and continuous products. The dependency is essential: the topology and positive cone of the test space must be fixed before one can assign a meaning to a positive Hessian current.

Chapters 15–18 develop and compare the delta and harmonic calculi. The former retains polyhedral curvature in an intersection algebra; the latter enlarges the functions allowed as smooth coordinates. Chapters 19–21 connect these local objects with model metrics, residual degrees, canonical measures and toric Monge–Ampère equations. Chapters 22–24 treat higher-codimension Green objects, degree normalizations, descent and cohomological limits.

Further source comparisons

For the source repairs, read Sections 11.6 and 19.17; the residual proof obligations are separated in Section 19.18. The delta conventions and graph examples are in Sections 15.9–15.10. Sections 16.7, 16.8 and 19.19 compare polyhedral and non-Archimedean envelopes, including a complete orthogonality counterexample and the relative-volume proof mechanism. Accessible skeleta and weighted measure germs are treated in Sections 6.12 and 21.28. For research involving cycle classes or Green objects, read Sections 24.15, 24.16 and 22.11 together. Section 23.6 supplies the global arithmetic completion framework; Section 24.17 states the remaining comparisons.

Routes for a first reading

A first reading can follow one of the routes below. The references point to the arguments on which the later material depends; they are also useful when returning to the book for a specific calculation.

Question Route
How does a tropical chart acquire a weight? Gauss norms and lattices in Chapter 5; initial degenerations in Section 5.3; monomial finite-free degrees in Section 5.7; Shilov points and faithful charts in Sections 6.6 and 3.4.
Which positivity cone is needed? Chapter 7, followed by coefficient measures and order zero in Chapter 9. Then read the Plücker counterexample in Appendix A and the separate Hessian/corner criterion in Section 10.6.
What does the continuous BT construction prove? Potential classes in Chapter 11; regularized maxima in Chapter 12; fixed-test estimates and mixed products in Chapter 13; the kernel calculation in Chapter 14.
When does model positivity agree with analytic positivity? The comparison table in Chapter 18; numerical curve detection and local model approximation in Chapter 19; residual curvature and Chern measures in Chapter 20.
How are concrete measures and potentials computed? Metric graphs in Chapter 20; real tori, toric roofs and discrete Monge–Ampère inversion in Chapter 21; weighted trace under graph maps in Chapter 23.
What is required for an actual Green form? The divisor sign in Chapter 10; regular sequences, finite-flat multiplicities and all delta-tests in Chapter 22; the distinction between current descent and representative descent in Chapter 23.
Can tropical Hodge theory be passed to an analytic limit? The affine inverse limit in Chapter 6; the superform resolution in Chapter 7; filtered cohomology, tropical duality and the explicit flat-torus calculation in Chapter 24.

Recurring distinctions

Several questions recur across the literature because the same formula can describe objects with different information. A tropical set forgets its lattice multiplicities. A current remembers the action on its chosen tests, while a form also has a representative and relations among presentations. A curvature current forgets metric constants. A top-degree measure comparison need not identify lower-degree forms. The examples throughout the book make these losses of information explicit.

The two kinds of independence in the BT construction deserve particular attention. Approximation independence means that two admitted approximations to the same continuous function define the same product. Potential independence means that two different functions with the same curvature act identically on a background current. The first follows from a limit theorem. The second asks how the Hessian kernel acts on that background. The tropical cross shows that these are separate assertions. On a convex Euclidean domain the kernel is affine and the second assertion has a short proof; on singular supports the kernel can be larger.

There are also two dimensions at a skeletal point. The real dimension of a simplex describes its visible tropical directions. The completed residue field retains the remaining algebraic directions. Section 3.8 computes their sum directly. The same division reappears in the canonical-measure formula: a simplex volume is multiplied by a residual intersection degree. It is the reason that mixed reduction cannot be treated merely as the totally degenerate case with fewer coordinates.

Using the source corrections

Appendix A separates three situations. A false literal statement has a counterexample and a replacement with its own hypotheses. A local formula or index error is corrected by degree, sign or domain checks. A missing proof step remains a stated obligation unless an independent argument supplies it. The conclusions of a paper are not all assigned the status of one problematic lemma. In particular, a counterexample on a reducible abstract tropical support is not automatically a counterexample on a smooth Berkovich space.

For the most consequential issues, the main text includes the repair rather than leaving it in an error list: ordinary-positive coefficient bounds, the convex representative theorem, fixed-neighbourhood approximation, the two parts of a piecewise smooth Hessian, regularized maximum normalization, and the finite-flat comparison at Abhyankar points. The source concordance points back to those proofs. Statements imported from the literature retain their source attribution even when their proof has been reorganized here.

For the current comparison repair, read Section 11.11.1 for harmonic lower contacts, Section 11.11.2 for relative supports through zeros, and Section 11.11.3 for the independent Q-PL identification on unit charts. Theorem 11.11.6 then supplies nef residues, and Corollary 11.11.7 returns to original smooth approximation on one neighbourhood. Section 11.11.5 repairs the interval-support and strict-localization calculations in APW. For vanishing coordinates continue with Section 11.12: its recession argument completes the relative rational PL identification, and Corollary 11.12.6 gives the forward smooth approximation. Section 11.12.6 repairs the Gaussian boundary calculation, and Section 11.12.7 proves fixed-fan coefficient density while exhibiting the obstruction to freezing a general subdivision. The remaining reverse-comparison problems are separated in Section 11.12.8.

Section 11.13 treats a different coefficient problem: an already strict R-PL germ is rationalized on one residue tube by a controlled ample perturbation. Theorem 11.13.6 gives the resulting comparison only within this PL category. Read Section 11.14 alongside the projective transfer: it separates a genuine fixed-model counterexample to GM Lemma 4.12 from the sufficient cofinal ample-model construction. The latter preserves the approximation input of Boucksom–Eriksson Theorem 7.8.

Consulting and citing this version

Continue with Section 11.15 for the passage from formal to algebraic domination and the corrected reduced cofinality theorem. Its norm examples distinguish reducedness, normality, equality of unit balls and equality after scalar extension. The final determinant estimate explains why a bounded discrepancy of norms disappears in relative volume.

Section 11.16 then realizes the fixed-chart obstruction by analytic units and gives a smooth psh example in every positive dimension. The example itself admits a different convex presentation, so it isolates a chart-dependent shortcut and does not disprove the full local comparison. Section 11.17 completes this test case in the original chart: it constructs a decreasing relative rational PL approximation even though ambient convex approximation fails. Read the midpoint defect and the coefficient-rounded formula together. The final monotone-convex composition result gives a higher-dimensional class for which the relative-to-original-smooth passage is available, with all domains fixed before taking limits. Section 11.18 gives a second, polyhedral route: extend the earlier fixed-fan argument to a finite polytopal complex whose vertices share one value-group scale. Its final examples explain why rational normals and value-group-compatible intercepts must be checked separately.

Section 11.19 separates the arithmetic continuity obstruction from the relative-support condition. Its closure formula and fixed-refinement theorem isolate the remaining arithmetic obstruction; the moving rectangle shows how changing subdivisions can overcome it. The final boundary-extension construction gives cellwise convexity in arbitrary dimension, with the missing cross-cell support condition stated separately. The rectangle example explains why a rational simplicial structure is an additional hypothesis, even when all facet normals are integral.

Section 11.20 completes the moving-subdivision argument for relatively supported real PL functions on coefficient-compatible compact complexes. The proof first controls the normal slopes in convex boundary extension, then preserves supports by a quantitative strict barrier. Read its final scope paragraph before applying the result to real-intercept source approximants or to original smooth functions.

Section 11.21 treats real-intercept PL germs by moving their centres within a stratum. For sequences, read the common-domain hypothesis in Corollary 11.21.6 together with Example 11.21.7. Corollary 11.21.8 states the original-smooth consequence with its analytic hypotheses. Proposition 11.21.10 distinguishes uniform approximation from Lipschitz-small error. Section 11.21.4 separates real-polyhedral, rational-polyhedral and coefficient-compatible PL conventions before using the source regularization.

For the general continuous relative-support input on a fixed compact image, start instead with Section 11.22. Its rational local bumps remove the earlier assumptions on the input subdivision. Theorem 11.22.3 controls coefficients before gluing; Theorem 11.22.8 passes through local maxima of unit logarithms and the fixed projective transfer. Read Example 11.22.11 to distinguish approximation of a limit from preservation of the support definition by that limit. Theorem 11.22.12 and Corollary 11.22.13 extend this construction through zeros, using admissible boundary supports and an exponential strictifier. Proposition 11.22.14 also treats a single original smooth psh input. Its final paragraph explains why the neighbourhood may depend on the input and how arbitrary sequences require a separate argument, now supplied by Section 11.24. The earlier geometric constructions remain useful for stronger quantitative control.

For varying local presentations, read Section 11.24. Exact absolute values on Laurent domains permit one finite extended presentation on a prescribed affinoid. This yields fixed-domain rational replacement and equality of four precisely defined first decreasing stages. Lemma 11.24.5 handles a second decrease when the rows already share one domain. Its final paragraph isolates the remaining sheafification issue.

The English and Chinese editions are separate texts with corresponding numbered statements. The Chinese edition keeps technical terminology in English. There are no exercises. The subject index gives entry points by mathematical term, and the notation table precedes the main text. References to a chapter number of CLD12 or CLD25 are explicitly marked as source references; cross-references within this book are hyperlinks.

This is the full working manuscript, version 0.38. A citation to one of its statements should retain the version and theorem or section number; when the statement is an imported theorem, cite the original paper as well. Complete proofs, proof reconstructions using specified external inputs, and remaining comparison questions are distinguished in the text. The manuscript has undergone internal mathematical and editorial checks, but no independent mathematical certification is claimed. The unresolved comparisons are part of the account of the subject, not assertions that all literature gaps have been closed.

Proof completeness and permanent references.

The proof architecture distinguishes a complete internal proof from a theorem used as an external input. It also describes the intended dependency graph and permanent identifiers. In particular, a source citation, a reconstruction under named inputs, and a mathematically closed proof are three different states. The separate project charter records the long-term plan; the reference text remains organized by mathematical constructions.

Fixed-affinoid replacement and first stages.

Sections 11.23–11.24 separate common unit charts from the stronger method using extended presentations. Read the exact-norm fraction lemma, the saturated-cover globalization, and the fixed-affinoid approximation theorem in that order. Corollary 11.24.4 identifies four explicitly defined first-stage germ classes in the nontrivially valued integral-projective setting. The common-row diagonal handles a further iteration only when all rows already share a domain. Full CLD25/Wu comparisons and general sheaf-level stability remain separate. Example 11.23.5 shows, in every positive dimension, why global unit presentations cannot simply be assumed.

Projective approximation and its generator class.

Section 11.9 proves a fixed-neighbourhood approximation theorem for decreasing sequences of explicitly specified convex rational PL generators on an integral projective variety over a nontrivially valued field. Its barrier is fixed before the sequence; its scales may vary with the terms. The proof uses the cited model/PL and model-nefness criteria, known uniform Fubini–Study approximation, and the diagonal lemma. No envelope hypothesis is used for this inclusion. Section 11.10 treats ambient analytic generators at the extended tropical boundary, including zeros. Read its boundary factorization and finite-selection proofs, then the fixed-domain corollary in Section 11.10.5. Theorem 11.10.11 gives the broader transfer for locally semipositive rational PL generators, using the maximum theorem and local-to-global model semipositivity. The fixed-affinoid and first-stage comparison is now supplied by Section 11.24; source-category identification, further sheaf-level stability and trivial valuation remain separate. Section A.8 repairs the touching-support formula in Wu’s convexity definition.

Common domains and source corrections.

Section 11.8 supplies a quantitative diagonal lemma, a uniform tropical localization formula and projective Fubini–Study barriers. Example 11.8.5 refutes only the whole-domain majorization added to APW26 Proposition 5.19; Proposition 11.8.6 preserves its local extension conclusion. Theorem 11.8.9 constructs the geometric data needed in Section 11.7.3. Section 11.9 subsequently supplies a common domain in one direction for the stated rational PL class. The full Wu comparison still requires identification of its precise source conventions and field scope; see Section 11.8.5 and the source table in Section A.7.

Envelope localization.

Section 11.7 incorporates Wu’s recent local theory and separates its first-stage classes from the generated CLD sheaf. Read the fixed bounded extension in Proposition 11.7.3, then Theorem 11.7.5: bounded decreasing stability is sufficient to identify every continuous ordinal stage with the first stage. Corollary 11.7.6 gives the projective application using explicitly imported comparison and envelope theorems. Its source dependencies and the general analytic question are kept separate in Remark 11.7.7.

The compact-barrier argument.

Sections 11.6.1–11.6.3 separate decreasing iteration on a fixed compact set from the local sheaf construction. The compact upper-barrier theorem gives first-stage approximation whenever a common compact representation is available. The indexed inclusion and the locally integrable hierarchy are proved separately. The general continuous replacement in CLD25 Proposition 8.1.11 remains open here; Proposition 8.1.12 is retained with its valid proof.

The integral PL comparison.

Read Sections 19.18.8–19.18.13 after the residual criterion in Theorem 19.18.12. The corrected weighted maximum gives semiample first-stage approximation. The new prescribed graded premodel construction then proves the integral nef implication over arbitrary complete fields, including trivial valuation, on one fixed reduction tube. Corollary 19.18.23 compares original-current positivity, residual nefness and local C25\mathcal C_{25} membership. The general continuous approximation and predecessor-replacement questions remain open in this manuscript.