21.09.01 · number-theory / arakelov-faltings

Arakelov geometry and arithmetic surfaces (survey)

shipped3 tiersLean: none

Anchor (Master): Arakelov 1974 *Izv. Akad. Nauk SSSR Ser. Mat.* 38, 1179-1192 (originator paper — *Intersection theory of divisors on an arithmetic surface*: the archimedean completion via Green's functions, the arithmetic intersection pairing); Faltings 1984 *Annals of Mathematics* 119, 387-424 (*Calculus on arithmetic surfaces*: arithmetic Riemann-Roch for arithmetic surfaces, the Faltings delta invariant $\delta_F$, arithmetic adjunction); Gillet-Soulé 1990 *Publ. Math. IHÉS* 72, 93-174 (*Arithmetic intersection theory*: extension to arbitrary dimension via hermitian holomorphic vector bundles, Green currents); Gillet-Soulé 1992 *Annals of Mathematics* 131, 163-238 + 132, 205-238 (*An arithmetic Riemann-Roch theorem*: the Grothendieck-Riemann-Roch generalisation, Bismut-Gillet-Soulé curvature formula); Vojta 1991 *Inventiones Mathematicae* 105, 547-560 (*Siegel's theorem in the compact case*: Arakelov-theoretic upgrade producing an effective Mordell theorem); Soulé-Abramovich-Burnol-Kramer 1992 *Lectures on Arakelov Geometry* (Cambridge Studies in Advanced Mathematics 33); Lang 1988 *Introduction to Arakelov Theory* (Springer-Verlag); Bombieri-Gubler 2006 *Heights in Diophantine Geometry* (Cambridge New Mathematical Monographs 4) Ch. 2-3; Manin-Panchishkin 2005 *Introduction to Modern Number Theory* (Springer EMS 49, 2nd ed.) Ch. 6 §§3-4

Intuition Beginner

A classical surface, like a smooth complex projective surface, has a beautiful intersection theory. Two curves on it cross at a finite collection of points, and one can count those points with signs to get an integer, the intersection number. This number drives almost every theorem about surfaces — the Bezout count of where two algebraic curves meet, the adjunction formula that computes the genus of a curve from its self-intersection, the Riemann-Roch theorem that bounds the dimension of sections of a line bundle. The theory works because the surface is compact, complete, and equidimensional.

Now replace the complex surface with an arithmetic surface: a two-dimensional object that sits over the ring of integers $Z$ instead of over a field of complex numbers. The arithmetic dimensions are one geometric direction (the fibres above each prime $p$ , each of which is a curve in characteristic $p$ ) and one arithmetic direction (the primes themselves, walking through $2, 3, 5, 7, \dots$ ). The picture wants to be a surface, but it lacks a top — there is no "fibre at infinity" matching the fibres above each finite prime. The arithmetic line $Spec (Z)$ is one-dimensional but not compact in the algebro-geometric sense.

This is the gap that Suren Arakelov closed in 1974. His idea: complete the missing fibre at infinity by hand, using complex-analytic data — a Kähler metric on the complex fibre $X (C)$ together with Green's functions encoding how curves meet at archimedean places. With this archimedean data added, an arithmetic surface acquires a genuine intersection pairing that is a real number, and every classical theorem about surfaces lifts to an arithmetic theorem about $X \to Spec (Z)$ . The pay-off is enormous: Faltings's proof of the Mordell conjecture, Vojta's height inequality machinery, and the modern viewpoint on the abc conjecture all sit inside this framework.

Visual Beginner

A schematic showing the arithmetic surface $X \to Spec (Z)$ : a horizontal arithmetic line representing $Spec (Z)$ with markers at primes $2, 3, 5, 7, \dots$ , and a vertical fibre over each prime depicting a curve in characteristic $p$ . At the far right of the diagram, a separate vertical shape labelled "fibre at infinity" depicts a compact Riemann surface $X (C)$ with a Kähler metric drawn schematically as level curves of a Green's function. Arrows indicate that the archimedean fibre is the completion supplied by Arakelov to make the picture into a complete two-dimensional object.

The picture captures the essential geometry: the missing fibre at infinity is supplied by the complex-analytic data of a Kähler metric and Green's functions on $X (C)$ , and the arithmetic intersection pairing combines finite-place intersection numbers with archimedean Green's-function integrals.

Worked example Beginner

Compute the arithmetic degree of the hermitian line bundle $O (D)$ for a single point divisor $D = [P]$ on the projective line over $Z$ .

Step 1. The arithmetic surface is $X = P_{Z}^{1}$ , the projective line over $Spec (Z)$ . This is the simplest arithmetic surface: a flat projective scheme over $Z$ whose fibre over each prime $p$ is the projective line $P_{F_{p}}^{1}$ in characteristic $p$ , and whose archimedean fibre is the Riemann sphere $P^{1} (C) = S^{2}$ .

Step 2. Pick a rational point on $P_{Z}^{1}$ , say $P = [0 : 1]$ , the origin. The Weil divisor $D = [P]$ is a horizontal section of the structure morphism $X \to Spec (Z)$ — it picks out one point in each fibre, including in the archimedean fibre $P^{1} (C)$ where it picks out the origin $0 \in S^{2}$ .

Step 3. The Kähler metric on the Riemann sphere is the Fubini-Study metric $ω_{F S} = \frac{1}{2 π} \frac{∣ d z ∣ ^{2}}{( 1 + ∣ z ∣ ^{2} ) ^{2}}$ , normalised so that the total volume is $1$ . The Green's function $g_{P} (z) = - lo g ∣ z ∣^{2} + lo g (1 + ∣ z ∣^{2})$ is the canonical solution of the equation that says $d d^{c} g_{P} = ω_{F S} - δ_{P}$ , with $δ_{P}$ the point mass at $0$ .

Step 4. The arithmetic divisor is the pair $D = (D, g_{P})$ combining the geometric divisor $D = [P]$ with the Green's function $g_{P}$ at the archimedean place. The arithmetic degree of $D$ is the real number $de g (D)$ , which for a point on $P_{Z}^{1}$ is computed by combining the finite-place contributions (the heights of the prime ideals of the section) with the archimedean contribution from $g_{P}$ .

Step 5. For the rational point $P = [0 : 1]$ with coordinates of size $1$ at every place, the finite-place contributions vanish and the archimedean contribution reduces to the integral of $\frac{1}{2} g_{P}$ against $ω_{F S}$ over the Riemann sphere $S^{2}$ , which by a standard computation evaluates to $\frac{1}{2}$ . So $de g (D) = \frac{1}{2}$ .

What this tells us: the arithmetic degree on $P_{Z}^{1}$ is a real number, not an integer. The classical degree of a divisor on $P^{1}$ over a field is an integer (the count of points with multiplicity), but in the arithmetic setting the archimedean Green's-function integral contributes a real number that records how the divisor sits inside the Kähler-metric geometry of the complex fibre. Arakelov's framework lets every classical intersection number on a surface become a real number on the arithmetic surface, with the archimedean correction supplied by Green's functions.

Check your understanding Beginner

Exercise (easy, multiple choice).

Arakelov's programme adds which kind of data to an arithmetic surface $X \to Spec (Z)$ to make intersection theory work?

A. Additional prime ideals in $Spec (Z)$ B. A Kähler metric on the complex fibre $X (C)$ together with Green's functions C. A choice of $p$ -adic absolute value at each finite prime D. An algebraic closure of $Q$

Hint

The missing piece is the "fibre at infinity", and infinity in this context refers to the archimedean place. Recall that a Kähler metric is the standard differential-geometric data carried by a complex manifold.

Answer

B. Arakelov's 1974 paper supplies the archimedean completion of $X \to Spec (Z)$ by adding a Kähler metric on the complex fibre $X (C) = X \times_{Z} C$ together with Green's functions encoding archimedean intersection data. Option A would extend the base but not solve the compactification problem; option C is finite-place data already present in the scheme; option D enriches the base field but does not supply the missing fibre at infinity.

Formal definition Intermediate+

Let $K$ be a number field with ring of integers $O_{K}$ , and let $π : X \to S = Spec (O_{K})$ be an arithmetic surface: a regular, flat, projective scheme of relative dimension one over $S$ , with generic fibre $X_{K} = X \times_{S} Spec (K)$ a smooth projective curve over $K$ . For each archimedean place $σ : K ↪ C$ , the base change $X_{σ} = X \times_{S} Spec (C)_{σ}$ is a compact Riemann surface, equipped with a chosen Kähler form $μ_{σ}$ .

A Green's function for a divisor $D = \sum_{i} n_{i} [P_{i}]$ on a compact Riemann surface $(X, μ)$ is a function $g : X ∖ ∣ D ∣ \to R$ smooth away from the support of $D$ and satisfying: $$ d d^c g + \delta_D = [\deg D] \cdot \mu, $$ where $d d^{c} = \frac{i}{2 π} \partial \overline{\partial}$ is the canonical $(1, 1)$ -current operator, $δ_{D}$ is the integration current along $D$ , and the equation is required to hold as currents of bidegree $(1, 1)$ . The Green's function is determined up to addition of a real constant; the standard normalisation requires $\int_{X} g μ = 0$ .

The group of arithmetic divisors on $X$ is the direct sum $$ \widehat{\mathrm{Div}}(\mathcal{X}) = \mathrm{Div}(\mathcal{X}) \oplus \bigoplus_{\sigma : K \hookrightarrow \mathbb{C}} (\text{Green's functions on } X_\sigma) / (\text{constants}), $$ with the constants quotient accounting for the gauge freedom in the Green's-function normalisation. An element is a pair $D = (D, g_{∙})$ with $D$ a Weil divisor on $X$ and $g_{σ}$ a Green's function for $D_{σ}$ on $X_{σ}$ at each archimedean place.

A meromorphic function $f \in K (X)^{\times}$ defines the arithmetic principal divisor $$ \widehat{\mathrm{div}}(f) = (\mathrm{div}(f), , -\log|f|^2_\sigma), $$ where $div (f)$ is the usual divisor of zeros and poles, and $- lo g ∣ f ∣_{σ}^{2}$ is the Green's function determined by the absolute value of $f$ at the archimedean place $σ$ . The arithmetic Picard group is the quotient $$ \widehat{\mathrm{Pic}}(\mathcal{X}) = \widehat{\mathrm{Div}}(\mathcal{X}) / \widehat{\mathrm{div}}(K(\mathcal{X})^\times). $$

The arithmetic intersection pairing is the bilinear map $$ \widehat{(,\cdot,)} : \widehat{\mathrm{Pic}}(\mathcal{X}) \times \widehat{\mathrm{Pic}}(\mathcal{X}) \to \mathbb{R}, \qquad \widehat{D} \cdot \widehat{E} = \sum_{\mathfrak{p}} (D \cdot E)\mathfrak{p} \log N(\mathfrak{p}) + \frac{1}{2} \sum{\sigma} \int_{X_\sigma} g_{D,\sigma} \cdot (d d^c g_{E,\sigma} + \delta_{E_\sigma}), $$ where the first sum runs over finite primes $p$ of $O_{K}$ and $(D \cdot E)_{p}$ is the intersection multiplicity in the fibre over $p$ , and the second sum runs over the archimedean places. The pairing is symmetric, bilinear, and well-defined on the Picard group (it descends through the principal-divisor relation).

A hermitian line bundle on $X$ is a pair $L = (L, h_{∙})$ with $L$ a line bundle on $X$ and $h_{σ}$ a smooth hermitian metric on $L_{σ}$ at each archimedean place $σ$ . Hermitian line bundles form a group $Pic (X)$ isomorphic to the arithmetic Picard group above: a hermitian line bundle $L$ with global section $s$ defines the arithmetic divisor $div (s) = (div (s), - lo g ∣ s ∣_{h}^{2})$ , and this map identifies the two groups.

Counterexamples to common slips

The arithmetic intersection number is a real number, not an integer. The archimedean Green's-function integrals contribute irrational real numbers in general, and even the simplest examples (such as a rational point on $P_{Z}^{1}$ in the worked example) produce rational but non-integer values.
The Kähler form $μ_{σ}$ on the archimedean fibre is part of the data, not a derived quantity. Different choices of $μ_{σ}$ yield different arithmetic pairings, and the canonical choice (the volume form induced by the Arakelov metric in the Faltings sense) is what produces the cleanest formulae. The dependence on $μ_{σ}$ is functorial but explicit.
Arakelov's framework treats archimedean places symmetrically with finite primes through a uniform pairing formula, but the formula is not a direct generalisation of the finite-place pairing: the archimedean contribution involves the Kähler form and the integration of Green's functions, while the finite-place contribution is the sum of local intersection multiplicities weighted by $lo g N (p)$ . The matching is structural and conceptual rather than literal.

Key theorem with proof Intermediate+

Theorem (Arakelov's arithmetic intersection pairing; Arakelov 1974 ^{[source pending]}). Let $π : X \to Spec (O_{K})$ be an arithmetic surface. The arithmetic intersection pairing $$ \widehat{(,\cdot,)} : \widehat{\mathrm{Pic}}(\mathcal{X}) \times \widehat{\mathrm{Pic}}(\mathcal{X}) \to \mathbb{R} $$ defined above is symmetric, bilinear, and well-defined: it descends through the principal-divisor relation in each variable. Moreover, the pairing satisfies the projection formula: for a finite-type morphism $φ : X^{'} \to X$ of arithmetic surfaces, $$ \varphi_*(\widehat{D}') \cdot \widehat{E} = \widehat{D}' \cdot \varphi^*(\widehat{E}) $$ holds in $R$ for any arithmetic divisors $D^{'}$ on $X^{'}$ and $E$ on $X$ .

Proof. Bilinearity and symmetry are visible from the defining formula: the finite-place sum $\sum_{p} (D \cdot E)_{p} lo g N (p)$ is bilinear and symmetric by the corresponding properties of the geometric intersection multiplicity $(D \cdot E)_{p}$ on the fibre over $p$ , and the archimedean term $\frac{1}{2} \sum_{σ} \int_{X_{σ}} g_{D, σ} (d d^{c} g_{E, σ} + δ_{E_{σ}})$ is bilinear in $(g_{D, σ}, g_{E, σ})$ by linearity of integration and of the operator $d d^{c}$ . Symmetry of the archimedean term uses a Stokes-type identity (Green's reciprocity): for smooth functions $g_{1}, g_{2}$ on a compact Kähler surface $(X, μ)$ with $d d^{c} g_{i} + δ_{D_{i}} = de g (D_{i}) \cdot μ$ , $$ \int_X g_1 (d d^c g_2 + \delta_{D_2}) = \int_X g_2 (d d^c g_1 + \delta_{D_1}), $$ which follows by integration by parts using that $X$ is closed without boundary.

For well-definedness on the Picard group, the principal-divisor relation $div (f) = (div (f), - lo g ∣ f ∣_{σ}^{2})$ must pair to zero with every arithmetic divisor. Compute: the finite-place contribution is $\sum_{p} (div (f) \cdot E)_{p} lo g N (p)$ , which by the product formula for the function field $K (X)$ equals minus the archimedean contribution. The archimedean piece is $\frac{1}{2} \sum_{σ} \int_{X_{σ}} (- lo g ∣ f ∣_{σ}^{2}) (d d^{c} g_{E, σ} + δ_{E_{σ}})$ ; using the Poincaré-Lelong identity $d d^{c} (- lo g ∣ f ∣^{2}) = δ_{div (f)} - curvature contribution$ and Stokes, this matches the negative of the finite-place sum. The two contributions cancel by the product formula, giving the principal-divisor invariance.

The projection formula is established by base-change compatibility: at each finite prime $p$ , $(φ_{*} D^{'} \cdot E)_{p} = (D^{'} \cdot φ^{*} E)_{p}$ holds by the projection formula for geometric intersection on the fibre, and at each archimedean place, the integration is compatible with proper pushforward and pullback of Green's functions via the standard change-of-variables identity for integrals of currents. Summing over places gives the global projection formula. $□$

Bridge. The arithmetic intersection pairing builds toward every Arakelov-theoretic Diophantine application: heights of points on varieties, the Faltings height of an abelian variety, the canonical height on a Jacobian, and the effective bounds in the Vojta-Bombieri-Gubler programme. The foundational reason it works is exactly the global symmetry given by Green's reciprocity: the archimedean Stokes identity is dual to the finite-place product formula, and putting these together gives one pairing that respects the arithmetic-geometric symmetry. This is exactly the central insight Arakelov isolated in 1974: the arithmetic surface $X \to Spec (Z)$ acquires a complete intersection theory once one identifies the archimedean Green-function integration as the dual partner to the finite-place log-norm sum. The bridge is the recognition that the arithmetic pairing generalises the classical intersection pairing on a complex projective surface and appears again in 21.09.02 (Faltings height and Mordell), where the Faltings height $h_{F} (A)$ of an abelian variety is computed as an arithmetic self-intersection of a hermitian line bundle on the moduli space. Putting these together, one arithmetic-intersection framework produces every classical Diophantine invariant — heights, Faltings heights, canonical heights on Jacobians — and identifies them with one another at the level of arithmetic Chow groups.

Exercises Intermediate+

Exercise 4 (medium, short-answer).

State the arithmetic Riemann-Roch theorem of Faltings 1984 for an arithmetic surface with hermitian line bundle $L$ , identifying every piece of the formula.

Hint

The theorem packages the arithmetic Euler characteristic $χ (L)$ in terms of arithmetic self-intersections, the arithmetic canonical class $K$ , and the Faltings delta invariant.

Answer

For an arithmetic surface $π : X \to Spec (O_{K})$ with hermitian line bundle $L = (L, h)$ on $X$ , the arithmetic Riemann-Roch theorem of Faltings 1984 ^{[source pending]} states: $$ \widehat{\chi}(\widehat{L}) = \frac{1}{2}\widehat{L} \cdot (\widehat{L} - \widehat{K}) + \widehat{\chi}(\mathcal{O}{\mathcal{X}}) - \frac{1}{12}\delta_F, $$ where $χ (L)$ is the arithmetic Euler characteristic (combining the algebraic Euler characteristic with the archimedean Quillen-metric correction), $L \cdot L$ and $L \cdot K$ are arithmetic intersection numbers, $K$ is the arithmetic canonical class (the cotangent bundle with the Arakelov metric), $\widehat{\chi}(\mathcal{O}\mathcal{X}) $i s t h e a r i t hm e t i c E u l er c ha r a c t er i s t i co f t h es t r u c t u r es h e a f, an d$ \delta_F = \sum_\sigma \delta(X_\sigma) $i s t h e F a l t in g s d e l t ain v a r ian t, s u mmin g t h e F a l t in g s$ \delta $- f u n c t i o n o f e a c ha r c him e d e an f ib r e$ X_\sigma$. Rubric: full credit for each labelled piece of the formula with the correct numerical coefficients.

Exercise 5 (medium, short-answer).

Explain why the arithmetic pairing $D \cdot E$ is well-defined on the arithmetic Picard group $Pic (X)$ — that is, why pairing with a principal arithmetic divisor $div (f)$ gives zero.

Hint

Combine the product formula $\sum_{v} - lo g ∣ f ∣_{v} = 0$ for the number field $K$ with the Poincaré-Lelong formula $d d^{c} (- lo g ∣ f ∣^{2}) = δ_{div (f)} - curvature term$ for the archimedean Green's function $- lo g ∣ f ∣^{2}$ .

Answer

A principal arithmetic divisor $div (f) = (div (f), - lo g ∣ f ∣_{σ}^{2})$ pairs to zero against any arithmetic divisor $E = (E, g_{E, σ})$ because the finite-place contribution and the archimedean contribution cancel each other via the product formula. The finite-place piece is $\sum_{p} (div (f) \cdot E)_{p} lo g N (p) = \sum_{p} v_{p} (f ∣_{E}) lo g N (p)$ , the sum of finite-place valuations of $f$ restricted to $E$ weighted by $lo g N (p)$ . The archimedean piece is $\frac{1}{2} \sum_{σ} \int_{X_{σ}} (- lo g ∣ f ∣_{σ}^{2}) (d d^{c} g_{E, σ} + δ_{E_{σ}})$ , which by Poincaré-Lelong and Stokes equals $- \sum_{σ} lo g ∣ f (E) ∣_{σ}$ at the archimedean places. The product formula $\sum_{v} v (f ∣_{E}) lo g N_{v} = 0$ for the number field $K$ then makes the two contributions cancel exactly. Rubric: full credit for the product-formula cancellation argument with both pieces identified.

Exercise 6 (medium, numeric).

For the projective line $P_{Z}^{1}$ over the integers, what is the rank of the arithmetic Picard group $Pic (P_{Z}^{1})$ as an abelian group (modulo torsion)?

Hint

The classical Picard group $Pic (P_{Z}^{1}) = Z$ is generated by $O (1)$ , and the arithmetic refinement adds a one-real-parameter family of archimedean metrics.

Answer

The arithmetic Picard group $Pic (P_{Z}^{1})$ has rank $2$ modulo torsion: one $Z$ factor inherited from the classical Picard group $Pic (P_{Z}^{1}) = Z ⟨ O (1)⟩$ , and one $R$ factor parameterising the archimedean hermitian metric on $O (1) ∣_{C}$ up to gauge equivalence by smooth functions modulo constants. The rank thus refers to the discrete part being $Z$ together with the continuous part being a one-dimensional real vector space; in the standard reference convention one writes $Pic (P_{Z}^{1}) ≅ Z \oplus R$ . Rubric: full credit for $Z \oplus R$ or equivalent rank- $2$ description.

Exercise 7 (hard, short-answer).

Verify Arakelov's arithmetic adjunction formula on an arithmetic surface for a horizontal divisor $D$ corresponding to a rational point: $D \cdot (D + K) = - 2 χ (O_{D})$ , where $K$ is the arithmetic canonical class and $χ (O_{D})$ is the arithmetic Euler characteristic of the structure sheaf of $D$ .

Hint

Combine the geometric adjunction formula $D \cdot (D + K) = 2 g (D) - 2$ for a curve on a surface with the archimedean contribution arising from the Green's function. For a rational-point section, $g (D) = 0$ , so the geometric part contributes $- 2$ ; the archimedean part adjusts by the integral of the canonical Green's function.

Answer

The arithmetic adjunction formula on an arithmetic surface, proved by Faltings 1984 ^{[source pending]}, says $D \cdot (D + K) = - 2 χ (O_{D})$ for a horizontal divisor $D$ that is finite étale over the base. On the geometric side, $D \cdot (D + K) = 2 g (D) - 2$ , the classical adjunction for a curve $D$ of genus $g (D)$ on a surface. For a horizontal section corresponding to a rational point, $D$ is a copy of $Spec (O_{K})$ in $X$ , so $g (D) = 0$ for the generic fibre and the classical adjunction gives $D \cdot (D + K) = - 2$ on each fibre.

In the arithmetic version, the finite-place contribution to $D \cdot (D + K)$ is the sum of $- 2 lo g N (p)$ over the residue contributions plus the relevant intersection multiplicities; the archimedean contribution is the Green's-function integral $\frac{1}{2} \int_{X_{σ}} g_{D} \cdot (d d^{c} g_{D + K} + δ_{(D + K)_{σ}})$ at each archimedean place. The two contributions combine to give $- 2 χ (O_{D})$ , where $χ (O_{D}) = lo g # O_{D}$ is the arithmetic Euler characteristic of the rank-one section, computing the logarithmic norm of the discriminant. The match between the geometric $g (D) = 0$ and the arithmetic correction is one of the foundational computations of Arakelov geometry.

Rubric: full credit for identifying the arithmetic adjunction as the geometric formula plus the archimedean Green's-function correction, with the right interpretation of $χ (O_{D})$ as the logarithmic norm of the discriminant.

Exercise 8 (hard, short-answer).

State the arithmetic Hodge-index theorem of Faltings 1984: on an arithmetic surface $X \to Spec (O_{K})$ , if $D$ is an arithmetic divisor with $D \cdot F = 0$ for every fibre class $F$ (vertical divisor), then $D \cdot D \leq 0$ , with equality iff $D$ is itself a $Q$ -combination of fibre classes plus a principal arithmetic divisor.

Hint

The geometric Hodge-index theorem on a complex projective surface says the intersection pairing has signature $(1, ρ - 1)$ on Néron-Severi, with the positive eigenvector spanned by the hyperplane class. In the arithmetic setting, the role of the hyperplane class is played by a fibre class, and the orthogonal complement carries a negative-definite pairing.

Answer

The arithmetic Hodge-index theorem (Faltings 1984 ^{[source pending]}) generalises the classical Hodge-index theorem on a complex projective surface to the arithmetic setting. On an arithmetic surface $π : X \to Spec (O_{K})$ , let $V \subset Pic (X) \otimes Q$ be the subspace spanned by vertical divisor classes — divisors supported on the fibres of $π$ . The orthogonal complement $V^{⊥}$ with respect to the arithmetic intersection pairing has negative-definite restriction of the pairing.

Equivalently, if $D \in Pic (X) \otimes Q$ satisfies $D \cdot F = 0$ for every fibre class $F$ , then $D \cdot D \leq 0$ , with equality if and only if $D$ is itself a $Q$ -combination of fibre classes plus a principal arithmetic divisor (equivalently, $D$ lies in $V \otimes Q$ modulo principal divisors).

The proof reduces to the classical Hodge-index theorem on each fibre after a localisation argument, combined with the archimedean Green's-function variational principle, which is itself the analytic analogue of the index theorem. The arithmetic Hodge-index theorem is the input that powers Faltings's proof of the Mordell conjecture in 1983 and Vojta's effective height inequalities.

Rubric: full credit for the negative-definiteness statement plus the equality characterisation in terms of fibre classes and principal divisors.

Advanced results Master

Theorem (arithmetic Riemann-Roch for arithmetic surfaces; Faltings 1984 ^{[source pending]}). Let $π : X \to S = Spec (O_{K})$ be an arithmetic surface with hermitian line bundle $L = (L, h)$ . The arithmetic Euler characteristic $$ \widehat{\chi}(\widehat{L}) = \chi(\mathcal{X}, L) + \tau(\widehat{L}; h) $$ combining the algebraic Euler characteristic with the archimedean Quillen-metric correction $τ$ satisfies $$ \widehat{\chi}(\widehat{L}) = \frac{1}{2}\widehat{L} \cdot (\widehat{L} - \widehat{K}) + \widehat{\chi}(\mathcal{O}\mathcal{X}) - \frac{1}{12}\delta_F, $$ *where $K$ is the arithmetic canonical class (the cotangent bundle $\Omega^1{\mathcal{X}/S} $e q u i pp e d w i t h t h e A r ak e l o v m e t r i c),$ \widehat{L} \cdot \widehat{L} $an d$ \widehat{L} \cdot \widehat{K} $a r e a r i t hm e t i c in t er sec t i o nn u mb er s in$ \mathbb{R} $, an d t h e F a l t in g s d e l t ain v a r ian t$ \delta_F = \sum_\sigma \delta(X_\sigma) $s u m s t h e F a l t in g s$ \delta $- f u n c t i o n s$ \delta(X_\sigma) \in \mathbb{R} $o f e a c ha r c him e d e an f ib r e$ X_\sigma $. T h e in v a r ian t$ \delta(X_\sigma) $d e p e n d so n l y o n t h eco m pl e x s t r u c t u r eo f$ X_\sigma$ as a Riemann surface, and reduces in genus one to a normalisation of the Faltings height of the elliptic curve.*

This is the arithmetic analogue of the classical Riemann-Roch theorem for a smooth projective surface, which says $χ (L) = \frac{1}{2} L \cdot (L - K) + χ (O_{X})$ . The new feature in the arithmetic setting is the archimedean correction term $- \frac{1}{12} δ_{F}$ , which is the differential-geometric input from the Bismut-Gillet-Soulé curvature formula for the Quillen metric on the determinant of cohomology of the archimedean fibres. The factor $- \frac{1}{12}$ matches the classical heat-kernel / zeta-function-regularisation coefficient familiar from string-theoretic anomaly computations and from the Selberg-trace formula on a Riemann surface.

Theorem (Gillet-Soulé arithmetic Riemann-Roch in higher dimension; Gillet-Soulé 1992 ^{[source pending]}). For a smooth projective morphism $π : X \to Y$ of arithmetic varieties with hermitian vector bundle $E$ on $X$ , the arithmetic Chern character of the direct image satisfies $$ \widehat{\mathrm{ch}}(R\pi_* \widehat{E}) = \pi_*(\widehat{\mathrm{ch}}(\widehat{E}) \cdot \widehat{\mathrm{Td}}(T_{\mathcal{X}/\mathcal{Y}})) - \pi_*(\widehat{\mathrm{ch}}(\widehat{E}) \cdot R(T_{\mathcal{X}/\mathcal{Y}})) $$ in $\widehat{\mathrm{CH}}^(\mathcal{Y}) \otimes \mathbb{Q} $, w h er e t h e f i r s tt er mi s t h ec l a ss i c a l G r o t h e n d i ec k - R i e mann - R oc h co n t r ib u t i o na d a pt e d t o a r i t hm e t i c C h o w g r o u p s, an d t h ecor r ec t i o n$ R(T_{\mathcal{X}/\mathcal{Y}})$ is the Bismut-Gillet-Soulé characteristic class involving analytic torsion of the fibres.*

This is the generalisation of Faltings 1984 to arbitrary dimension. The arithmetic Chow group $CH^{p} (X)$ for an arithmetic variety $X$ of dimension $d + 1$ over $Z$ is the quotient of pairs $(Z, g_{Z})$ — codimension- $p$ cycle plus a Green current — by the appropriate equivalence relation, and it carries an intersection product, Chern classes, and a degree map to $R$ . The Bismut-Gillet-Soulé curvature formula identifies the correction term $R$ in terms of the Ray-Singer analytic torsion of the relative dualising complex.

Theorem (Vojta's height inequality; Vojta 1991 ^{[source pending]}). Let $C$ be a smooth projective curve of genus $g \geq 2$ over a number field $K$ . There exist constants $c_{1}, c_{2} > 0$ , depending on $C$ and on a choice of arithmetic model $C \to Spec (O_{K})$ equipped with the Arakelov metric, such that for any pair of $K$ -rational points $P, Q \in C (K)$ with $P \neq = Q$ , the arithmetic intersection number on the self-product $C \times_{O_{K}} C$ satisfies $$ (\Delta - \pi_1^* \theta_P - \pi_2^* \theta_Q) \cdot (\Delta - \pi_1^* \theta_P - \pi_2^* \theta_Q) \geq c_1 h(P, Q) - c_2, $$ where $Δ$ is the arithmetic diagonal, $θ_{P}, θ_{Q}$ are the symmetrised theta divisors associated to $P, Q$ , and $h (P, Q)$ is the arithmetic height of the pair. Combined with the elementary gap principle on rational points, this gives a new proof of the Mordell conjecture.

Vojta's 1991 paper recasts Faltings's 1983 proof of the Mordell conjecture in the language of Arakelov intersection theory on the self-product of an arithmetic curve. The inequality is an Arakelov analogue of Mumford's gap principle for points on a curve in its Jacobian, and it implies that $C (K)$ is finite by an explicit counting argument once the constants $c_{1}, c_{2}$ are controlled. The Arakelov framework makes the bounds effective in principle (though the constants involved are not small enough for practical computation of rational points).

Theorem (Faltings height of an abelian variety; Faltings 1983 ^{[source pending]}). Let $A$ be an abelian variety of dimension $g$ over a number field $K$ . The Faltings height $h_{F} (A) \in R$ is defined as the arithmetic degree $$ h_F(A) = \frac{1}{[K : \mathbb{Q}]} \widehat{\deg}, \pi_* \widehat{\omega}_{A/\mathcal{O}_K}, $$ where $π : A \to Spec (O_{K})$ is the Néron model of $A$ , $ω_{A / O_{K}}$ is the relative dualising sheaf with the Faltings metric (the $L^{2}$ metric on its archimedean fibres), and $de g$ is the arithmetic degree of a hermitian line bundle on $Spec (O_{K})$ . The Faltings height is the key invariant in Faltings's proof of the Mordell, Shafarevich, and Tate conjectures.

The Faltings height refines the classical naïve height of an abelian variety by including the archimedean Arakelov-metric correction, and it satisfies a finiteness theorem: for any bound $B \in R$ , the set of isomorphism classes of abelian varieties over $K$ of dimension $g$ with $h_{F} (A) \leq B$ is finite. This is the input that Faltings used to prove the Shafarevich conjecture (finiteness of abelian varieties with given dimension, conductor, and semistability type), which together with the Tate conjecture and the Parshin-Kodaira construction yields the Mordell conjecture.

Theorem (the abc conjecture viewed through Arakelov). The abc conjecture of Masser-Oesterlé (1985, 1988) states that for every $ε > 0$ , the inequality $∣ ab c ∣ \leq C (ε) \cdot rad (ab c)^{1 + ε}$ holds for all coprime triples $(a, b, c)$ of nonzero integers with $a + b = c$ . In Arakelov-theoretic language, abc is equivalent to the statement that the arithmetic height of an algebraic point on $P^{1} ∖ {0, 1, \infty}$ is bounded by the arithmetic genus of the integral model plus a term linear in $lo g rad$ , where the heights and arithmetic genus are computed using the arithmetic intersection pairing on the relevant arithmetic surface. The Vojta-Mochizuki framework presents abc as a Diophantine inequality on the moduli space of elliptic curves, with the proof (Mochizuki 2012-2024, contested) using inter-universal Teichmüller theory built on top of Arakelov foundations.

The Arakelov-theoretic reformulation of abc is one of the strongest motivations for the framework: it gives a unified language in which heights, intersection numbers, and Diophantine inequalities sit in one place. Whether Mochizuki's 2012 proof of abc via inter-universal Teichmüller theory is correct remains contested in the research community as of 2026; the contested-but-not-refuted status itself reflects the depth of Arakelov-theoretic foundations underlying the programme.

Synthesis. The foundational reason Arakelov geometry exists is the asymmetry between finite and archimedean places: a number field $K$ has finitely many finite primes plus a finite number of archimedean places, the product formula $\sum_{v} v (f) lo g N_{v} = 0$ couples them in one structural identity, and any intersection theory on an arithmetic surface that respects the product formula must include archimedean data. This is exactly the central insight Arakelov isolated in 1974: by supplying Kähler metrics and Green's functions at the archimedean places, one obtains an arithmetic intersection pairing that descends through principal arithmetic divisors and respects the product formula. Putting these together, the arithmetic-intersection framework generalises classical surface intersection theory to arithmetic surfaces, the arithmetic Riemann-Roch theorem (Faltings 1984, Gillet-Soulé 1992) generalises classical Riemann-Roch with the Bismut-Gillet-Soulé curvature correction as the new archimedean term, and the height inequalities of Vojta-Bombieri-Gubler identify Diophantine finiteness with arithmetic intersection-number positivity. The bridge is the recognition that the arithmetic Picard group $Pic (X)$ is dual to the geometric Picard group of the complex fibre via the Green's-function pairing, and that this duality is exactly what makes the Faltings height, the Néron-Tate height, and the canonical height on a Jacobian into different facets of one arithmetic intersection-theoretic invariant.

Several apparently distinct invariants are identified by the Arakelov framework. The Faltings height of an abelian variety is the arithmetic degree of the Faltings metric on the relative dualising sheaf, computed via the arithmetic intersection pairing on $A \to Spec (O_{K})$ . The Néron-Tate canonical height on a Jacobian is the arithmetic self-intersection of the theta divisor with the Arakelov metric. The naive height of a point on projective space is the arithmetic degree of the corresponding hermitian line bundle. The bridge is that all of these are different presentations of arithmetic intersection numbers on the appropriate arithmetic variety, and the unifying framework is Gillet-Soulé arithmetic K-theory. The Bismut-Gillet-Soulé curvature formula identifies the archimedean correction term $R (T_{X / Y})$ in arithmetic Grothendieck-Riemann-Roch with the Ray-Singer analytic torsion of the relative dualising complex, and this in turn is dual to the heat-kernel asymptotic on a Riemann surface — a remarkable analytic-arithmetic correspondence whose origin is precisely the Arakelov programme. The framework appears again in 21.09.02 (Faltings height and the Mordell conjecture), where the abstract setup of the present unit is specialised to the moduli space of abelian varieties to produce the Diophantine finiteness statement that finalises Faltings's 1983 theorem.

Full proof set Master

Proposition (well-definedness of the arithmetic pairing). The arithmetic intersection pairing $(\cdot) : Pic (X) \times Pic (X) \to R$ descends to a well-defined map on the arithmetic Picard group.

Proof. We must show that pairing with a principal arithmetic divisor gives zero. Take $div (f) = (div (f), - lo g ∣ f ∣_{σ}^{2})$ for a non-zero meromorphic function $f \in K (X)^{\times}$ , and let $E = (E, g_{E, σ})$ be an arithmetic divisor with $E$ horizontal (we may assume this by linearity and the vanishing pairing of vertical-with-vertical at the archimedean places).

The finite-place contribution to $div (f) \cdot E$ is $$ \sum_\mathfrak{p} (\mathrm{div}(f) \cdot E)\mathfrak{p} \log N(\mathfrak{p}) = \sum\mathfrak{p} v_\mathfrak{p}(f|_E) \log N(\mathfrak{p}), $$ the sum of the finite-place valuations of $f$ restricted to $E$ , weighted by $lo g N (p)$ .

The archimedean contribution is $$ \frac{1}{2}\sum_\sigma \int_{X_\sigma} (-\log|f|^2_\sigma)(d d^c g_{E, \sigma} + \delta_{E_\sigma}). $$ Apply the Poincaré-Lelong formula $d d^{c} (- lo g ∣ f ∣^{2}) = δ_{div (f)} - de g (div (f)) \cdot μ$ to integrate by parts. The integration is by Stokes, using that $X_{σ}$ is closed without boundary, so we may move $d d^{c}$ from one argument to the other: $$ \int_{X_\sigma} (-\log|f|^2_\sigma) \cdot d d^c g_{E, \sigma} = \int_{X_\sigma} g_{E, \sigma} \cdot d d^c (-\log|f|^2_\sigma) = \int_{X_\sigma} g_{E, \sigma} \cdot \delta_{\mathrm{div}(f)\sigma} - \deg(\mathrm{div}(f)) \int{X_\sigma} g_{E, \sigma} , \mu. $$ Adding the $(- lo g ∣ f ∣_{σ}^{2}) \cdot δ_{E_{σ}}$ piece gives the value $- lo g ∣ f (E) ∣_{σ}^{2}$ summed over the points of $E_{σ}$ . The Green's-function normalisation $\int g_{E, σ} μ = 0$ kills the second term, leaving $$ \frac{1}{2}\sum_\sigma \int_{X_\sigma} (-\log|f|^2_\sigma)(d d^c g_{E, \sigma} + \delta_{E_\sigma}) = -\sum_\sigma \log|f(E)|_\sigma. $$

Summing the finite-place and archimedean contributions, $$ \widehat{\mathrm{div}}(f) \cdot \widehat{E} = \sum_\mathfrak{p} v_\mathfrak{p}(f|E) \log N(\mathfrak{p}) - \sum\sigma \log|f(E)|_\sigma = \sum_v v(f|_E) \log N_v, $$ where the last sum runs over all places of $K$ (with the archimedean log absolute values having $lo g N_{v} = 1$ ). The product formula for the number field $K$ asserts $\sum_{v} v (f ∣_{E}) lo g N_{v} = 0$ for every non-zero $f ∣_{E} \in K^{\times}$ , so the pairing vanishes. $□$

Proposition (symmetry of the arithmetic pairing). The arithmetic intersection pairing $(\cdot)$ is symmetric in its two arguments.

Proof. Bilinearity is visible from the formula. Symmetry of the finite-place piece is symmetry of the geometric intersection multiplicity $(D \cdot E)_{p} = (E \cdot D)_{p}$ on each fibre, established by the symmetry of the Tor-functor definition of intersection multiplicity.

Symmetry of the archimedean piece reduces to Green's reciprocity: for smooth functions $g_{1}, g_{2}$ on a compact Kähler manifold $(X, μ)$ with $d d^{c} g_{i} + δ_{D_{i}} = de g (D_{i}) \cdot μ$ , $$ \int_X g_1 (d d^c g_2 + \delta_{D_2}) = \int_X g_2 (d d^c g_1 + \delta_{D_1}). $$ Expand the left side: $\int g_{1} d d^{c} g_{2} + \int g_{1} \cdot δ_{D_{2}} = \int g_{1} d d^{c} g_{2} + g_{1} (D_{2})$ . By integration by parts on the closed manifold $X$ , $\int g_{1} d d^{c} g_{2} = \int g_{2} d d^{c} g_{1}$ . Substituting $d d^{c} g_{i} = de g (D_{i}) μ - δ_{D_{i}}$ inside the integrand gives $\int g_{2} d d^{c} g_{1} = de g (D_{1}) \int g_{2} μ - g_{2} (D_{1})$ , and similarly for $g_{1}$ . Combining the four pieces, both sides of Green's reciprocity simplify to $de g (D_{1}) \int g_{2} μ + de g (D_{2}) \int g_{1} μ - g_{1} (D_{2}) - g_{2} (D_{1})$ (with sign carefully tracked), which is symmetric in $(g_{1}, D_{1}) \leftrightarrow (g_{2}, D_{2})$ .

Combining the finite-place and archimedean symmetries gives the symmetry of the arithmetic pairing. $□$

Proposition (projection formula for the arithmetic pairing). Let $φ : X^{'} \to X$ be a finite-type morphism of arithmetic surfaces. For any arithmetic divisor $D^{'}$ on $X^{'}$ and $E$ on $X$ , $$ \varphi_*(\widehat{D}') \cdot \widehat{E} = \widehat{D}' \cdot \varphi^*(\widehat{E}). $$

Proof. Apply the classical projection formula on each fibre at a finite prime $p$ : for the geometric intersection multiplicity, $$ (\varphi_* D' \cdot E)\mathfrak{p} = (D' \cdot \varphi^* E)\mathfrak{p}, $$ holding by the classical projection formula on the fibre $X_{p}^{'} \to X_{p}$ , established via Tor-functor base change. Summing over finite primes weighted by $lo g N (p)$ gives the finite-place piece of the desired identity.

At each archimedean place $σ$ , the morphism $φ$ induces a holomorphic map $φ_{σ} : X_{σ}^{'} \to X_{σ}$ of compact Riemann surfaces. Green's functions transform contravariantly under pullback: $φ_{σ}^{*} (g_{E, σ}) = g_{φ^{*} E, σ}$ (up to a normalisation constant compatible with the gauge freedom). The integration is by change-of-variables for currents: $$ \int_{X'\sigma} (\varphi\sigma^* g_{E, \sigma}) (d d^c g_{D', \sigma} + \delta_{D'\sigma}) = \int{X_\sigma} g_{E, \sigma}(\varphi_{\sigma }(d d^c g_{D', \sigma} + \delta_{D'\sigma})) = \int{X_\sigma} g_{E, \sigma}(d d^c g_{\varphi_ D', \sigma} + \delta_{(\varphi_* D')_\sigma}), $$ giving the archimedean piece of the projection formula. Adding the finite-place and archimedean pieces yields the global identity. $□$

Theorem (arithmetic Riemann-Roch for arithmetic surfaces), proof sketch — see Faltings 1984 ^{[source pending]}. The full proof combines the classical Riemann-Roch theorem for surfaces ( $χ (L) = \frac{1}{2} L (L - K) + χ (O_{X})$ ) on each finite-prime fibre with the differential-geometric input from the Bismut-Gillet-Soulé curvature formula at the archimedean places. The Quillen metric on the determinant of cohomology $λ (L) = det R π_{*} L$ has curvature form computed by the Bismut-Gillet-Soulé theorem; integrating the curvature against the fundamental class of each archimedean fibre and combining with the algebraic Euler characteristic produces the arithmetic Euler characteristic $χ (L)$ . The Faltings delta invariant $δ (X_{σ})$ arises as the Ray-Singer analytic torsion of the Laplacian on $X_{σ}$ in genus $g \geq 1$ , and the coefficient $- \frac{1}{12}$ matches the heat-kernel zeta-function regularisation coefficient for the Laplacian on a Riemann surface. Each ingredient is itself a substantial theorem (the BGS curvature formula occupies the three-paper 1988 series; the structural identification of $δ$ requires the Polyakov-Alvarez formula); the assembly into the arithmetic Riemann-Roch statement is what Faltings 1984 ^{[source pending]} established, with the Gillet-Soulé 1992 generalisation ^{[source pending]} extending to arbitrary dimension. $□$

Theorem (arithmetic Hodge-index), proof. Let $V_{vert} \subset Pic (X) \otimes Q$ be the subspace spanned by vertical divisor classes, and let $D \in V_{vert}^{⊥}$ . We show $D \cdot D \leq 0$ with equality iff $D \in V_{vert}$ modulo principal divisors.

By the geometric Hodge-index theorem on each finite-prime fibre $X_{p}$ , the intersection pairing on $Pic (X_{p}) \otimes Q$ has signature $(1, n_{p} - 1)$ for some $n_{p}$ , with the positive eigenvector spanned by the fibre class itself. Restricted to the orthogonal complement of the fibre class, the pairing is negative-semidefinite. Summing over finite primes weighted by $lo g N (p)$ preserves this: the finite-place contribution to $D \cdot D$ for $D \in V_{vert}^{⊥}$ is $\leq 0$ , with equality iff the geometric part of $D$ is a $Q$ -combination of fibre classes plus a principal divisor.

For the archimedean piece, the variational principle for Green's functions implies that the self-pairing $\frac{1}{2} \int_{X_{σ}} g_{D, σ} (d d^{c} g_{D, σ} + δ_{D_{σ}})$ is negative for nonzero $g_{D, σ}$ orthogonal to the constants — this is the standard fact that the Laplacian on a compact Riemann surface has nonnegative eigenvalues, with the zero eigenvalue corresponding to constants. Summing the negative-semidefinite finite-place contribution with the negative archimedean contribution gives $D \cdot D \leq 0$ .

The equality case follows by tracking when the inequalities are saturated: the finite-place contribution vanishes iff the geometric part of $D$ is vertical (plus principal), and the archimedean contribution vanishes iff $g_{D, σ}$ is constant — equivalently iff $D$ has no archimedean variation, which is the residual content of being principal. Combining, $D \cdot D = 0$ iff $D$ is principal plus vertical modulo torsion. $□$

Theorem (Vojta's height inequality), stated without proof — see Vojta 1991 ^{[source pending]}. The inequality $$ (\Delta - \pi_1^* \theta_P - \pi_2^* \theta_Q) \cdot (\Delta - \pi_1^* \theta_P - \pi_2^* \theta_Q) \geq c_1 h(P, Q) - c_2 $$ combines the arithmetic Hodge-index theorem on $C \times_{S} C$ (giving $\geq 0$ for self-intersections of arithmetic divisors orthogonal to the fibre classes) with explicit estimates on the archimedean Green's-function contribution to the difference $Δ - π_{1}^{*} θ_{P} - π_{2}^{*} θ_{Q}$ . The combined effect is a quadratic lower bound on the arithmetic height that conflicts with the elementary gap principle ( $h (R) \leq C \cdot d (R)^{2}$ for points $R$ in an arithmetic progression on a curve, where $d$ measures the "step"), forcing finiteness of $C (K)$ . Vojta's 1991 paper ^{[source pending]} gives the full argument; Bombieri-Gubler 2006 Ch. 11 ^{[source pending]} is the modern textbook treatment.

Connections Master

Intersection pairing on a surface 04.05.06. The classical intersection pairing on a smooth complex projective surface, $Pic (X) \times Pic (X) \to Z$ , is the geometric input that Arakelov's framework generalises. The arithmetic intersection pairing $Pic (X) \times Pic (X) \to R$ restricts to the geometric pairing on each finite-prime fibre, with the archimedean Green's-function integral supplying the new term. The arithmetic pairing is a real number rather than an integer because the archimedean correction introduces irrational contributions in general.
Adjunction formula 04.05.07. The classical adjunction $D \cdot (D + K) = 2 g (D) - 2$ on a complex projective surface generalises to the arithmetic adjunction $D \cdot (D + K) = - 2 χ (O_{D})$ on an arithmetic surface, where $χ (O_{D})$ is the arithmetic Euler characteristic and the arithmetic canonical class $K$ is the cotangent bundle equipped with the Arakelov metric. For a horizontal section (rational point), the formula computes the logarithmic norm of the discriminant.
Riemann-Roch theorem for surfaces 04.05.08. The classical Riemann-Roch $χ (L) = \frac{1}{2} L (L - K) + χ (O_{X})$ on a smooth projective surface lifts in the arithmetic setting to Faltings 1984's arithmetic Riemann-Roch $χ (L) = \frac{1}{2} L (L - K) + χ (O_{X}) - \frac{1}{12} δ_{F}$ on an arithmetic surface. The new term $- \frac{1}{12} δ_{F}$ is the differential-geometric correction from the archimedean Quillen metric and the Bismut-Gillet-Soulé curvature formula.
Hermitian / Kähler form 03.02.11. The Kähler form $μ_{σ}$ on the archimedean fibre $X_{σ}$ is the differential-geometric input to the Arakelov programme. The choice of Kähler form determines the Green's functions and hence the arithmetic intersection pairing; the canonical choice (the Arakelov metric in the sense of Faltings 1984) makes the formulae cleanest and produces the Faltings delta invariant in the arithmetic Riemann-Roch theorem.
Picard group 04.05.02. The classical Picard group $Pic (X)$ of an arithmetic surface is the abelian group of line bundles up to isomorphism. The arithmetic Picard group $Pic (X)$ enriches this with archimedean hermitian metric data, producing a real-vector-space extension of the classical Picard group. The arithmetic intersection pairing is bilinear on $Pic (X)$ and provides the Diophantine-geometric pay-off of the framework.

Historical & philosophical context Master

Suren Arakelov, working in Moscow in the early 1970s, observed a structural lacuna in the algebraic-geometric formalism of arithmetic surfaces: a smooth projective variety over $Spec (Z)$ behaves like a complete two-dimensional object in many ways, but the standard intersection theory of complete varieties fails on it because the arithmetic base $Spec (Z)$ is not proper — there is no fibre at the archimedean place to balance the fibres at the finite primes. Arakelov's 1974 paper Intersection theory of divisors on an arithmetic surface, published in the Izvestiya Akademii Nauk SSSR (Vol. 38, 1179-1192) ^{[source pending]}, supplied the missing fibre at infinity by adding a Kähler metric on the complex fibre $X (C)$ together with Green's functions encoding archimedean intersection data. The construction made the arithmetic surface into a "complete" two-dimensional object with a real-valued intersection pairing, and opened the route to a complete arithmetic version of classical surface theory.

The framework was elaborated and consolidated over the following decade. Faltings's 1984 paper Calculus on arithmetic surfaces (Annals of Mathematics 119, 387-424) ^{[source pending]} established the arithmetic Riemann-Roch theorem for arithmetic surfaces, introducing the Faltings delta invariant $δ_{F}$ as the archimedean correction term and establishing arithmetic adjunction and an arithmetic Hodge-index theorem. Gillet and Soulé extended the framework to arbitrary dimension in their 1990 paper Arithmetic intersection theory (Publications Mathématiques de l'IHÉS 72, 93-174) ^{[source pending]}, replacing Green's functions by Green currents on $X (C)$ , and their 1992 follow-up ^{[source pending]} established the arithmetic Grothendieck-Riemann-Roch theorem with the Bismut-Gillet-Soulé curvature formula as the archimedean correction. Bismut, Gillet, and Soulé's three-paper 1988 series in Communications in Mathematical Physics (Vol. 115) provided the differential-geometric input — the curvature of the Quillen metric on the determinant of cohomology — that made the higher-dimensional arithmetic Riemann-Roch theorem work.

The Diophantine pay-off came rapidly. Faltings's 1983 proof of the Mordell conjecture (the Invent. Math. paper on the Tate conjecture for abelian varieties) used the Faltings height $h_{F} (A)$ , defined as an arithmetic degree on $Spec (O_{K})$ via the Arakelov framework. Vojta's 1991 paper Siegel's theorem in the compact case (Inventiones Mathematicae 105, 547-560) ^{[source pending]} recast Faltings's argument in the language of Arakelov intersection theory on the self-product $C \times_{O_{K}} C$ of an arithmetic curve, producing a uniform height-inequality framework that has since been extended to the Bombieri-Gubler textbook treatment (Cambridge 2006) ^{[source pending]}. The Arakelov framework also underlies the Vojta-Mochizuki programme on the abc conjecture, in which abc is recast as a Diophantine inequality on an arithmetic moduli space; whether Mochizuki's 2012-2024 inter-universal Teichmüller theory provides a correct proof remains contested in the research community as of 2026.

Bibliography Master

@article{Arakelov1974,
  author  = {Arakelov, S. Yu.},
  title   = {Intersection theory of divisors on an arithmetic surface},
  journal = {Izvestiya Akademii Nauk SSSR, Seriya Matematicheskaya},
  volume  = {38},
  year    = {1974},
  pages   = {1179--1192}
}

@article{Faltings1984,
  author  = {Faltings, Gerd},
  title   = {Calculus on arithmetic surfaces},
  journal = {Annals of Mathematics},
  volume  = {119},
  year    = {1984},
  pages   = {387--424}
}

@article{Faltings1983Mordell,
  author  = {Faltings, Gerd},
  title   = {Endlichkeitss{\"a}tze f{\"u}r abelsche Variet{\"a}ten {\"u}ber Zahlk{\"o}rpern},
  journal = {Inventiones Mathematicae},
  volume  = {73},
  year    = {1983},
  pages   = {349--366}
}

@article{GilletSoule1990,
  author  = {Gillet, Henri and Soul{\'e}, Christophe},
  title   = {Arithmetic intersection theory},
  journal = {Publications Math{\'e}matiques de l'IH{\'E}S},
  volume  = {72},
  year    = {1990},
  pages   = {93--174}
}

@article{GilletSoule1992,
  author  = {Gillet, Henri and Soul{\'e}, Christophe},
  title   = {An arithmetic {R}iemann-{R}och theorem},
  journal = {Inventiones Mathematicae},
  volume  = {110},
  year    = {1992},
  pages   = {473--543}
}

@article{Vojta1991,
  author  = {Vojta, Paul},
  title   = {Siegel's theorem in the compact case},
  journal = {Inventiones Mathematicae},
  volume  = {105},
  year    = {1991},
  pages   = {547--560}
}

@book{SouleEtAl1992,
  author    = {Soul{\'e}, Christophe and Abramovich, Dan and Burnol, Jean-Fran{\c c}ois and Kramer, J{\"u}rg},
  title     = {Lectures on {A}rakelov Geometry},
  publisher = {Cambridge University Press},
  series    = {Cambridge Studies in Advanced Mathematics},
  volume    = {33},
  year      = {1992}
}

@book{Lang1988,
  author    = {Lang, Serge},
  title     = {Introduction to {A}rakelov Theory},
  publisher = {Springer-Verlag},
  year      = {1988}
}

@book{BombieriGubler2006,
  author    = {Bombieri, Enrico and Gubler, Walter},
  title     = {Heights in Diophantine Geometry},
  publisher = {Cambridge University Press},
  series    = {New Mathematical Monographs},
  volume    = {4},
  year      = {2006}
}

@book{ManinPanchishkin2005,
  author    = {Manin, Yuri I. and Panchishkin, Alexei A.},
  title     = {Introduction to Modern Number Theory},
  publisher = {Springer-Verlag},
  series    = {Encyclopaedia of Mathematical Sciences},
  volume    = {49},
  edition   = {2},
  year      = {2005}
}

@article{BismutGilletSoule1988a,
  author  = {Bismut, Jean-Michel and Gillet, Henri and Soul{\'e}, Christophe},
  title   = {Analytic torsion and holomorphic determinant bundles, {I}},
  journal = {Communications in Mathematical Physics},
  volume  = {115},
  year    = {1988},
  pages   = {49--78}
}

@article{BismutGilletSoule1988b,
  author  = {Bismut, Jean-Michel and Gillet, Henri and Soul{\'e}, Christophe},
  title   = {Analytic torsion and holomorphic determinant bundles, {II}},
  journal = {Communications in Mathematical Physics},
  volume  = {115},
  year    = {1988},
  pages   = {79--126}
}

@article{BismutGilletSoule1988c,
  author  = {Bismut, Jean-Michel and Gillet, Henri and Soul{\'e}, Christophe},
  title   = {Analytic torsion and holomorphic determinant bundles, {III}},
  journal = {Communications in Mathematical Physics},
  volume  = {115},
  year    = {1988},
  pages   = {301--351}
}

Prerequisites

04.02.01
04.05.01
04.05.02
04.05.06
04.05.07
04.05.08
03.02.11

Tier anchors

beginner: Manin-Panchishkin *Introduction to Modern Number Theory* (Springer EMS 49, 2nd ed. 2005) Ch. 6 §3 informal opening on the missing fibres at infinity; Bombieri-Gubler *Heights in Diophantine Geometry* (Cambridge 2006) Ch. 2 informal preface on arithmetic completion
intermediate: Bombieri-Gubler *Heights in Diophantine Geometry* (Cambridge 2006) Ch. 2-3 (arithmetic intersection theory on surfaces); Soulé-Abramovich-Burnol-Kramer *Lectures on Arakelov Geometry* (Cambridge 1992) Ch. 1-3; Lang *Introduction to Arakelov Theory* (Springer 1988) Ch. 1-4
master: Arakelov 1974 *Izv. Akad. Nauk SSSR Ser. Mat.* 38, 1179-1192 (originator paper — *Intersection theory of divisors on an arithmetic surface*: the archimedean completion via Green's functions, the arithmetic intersection pairing); Faltings 1984 *Annals of Mathematics* 119, 387-424 (*Calculus on arithmetic surfaces*: arithmetic Riemann-Roch for arithmetic surfaces, the Faltings delta invariant $\delta_F$, arithmetic adjunction); Gillet-Soulé 1990 *Publ. Math. IHÉS* 72, 93-174 (*Arithmetic intersection theory*: extension to arbitrary dimension via hermitian holomorphic vector bundles, Green currents); Gillet-Soulé 1992 *Annals of Mathematics* 131, 163-238 + 132, 205-238 (*An arithmetic Riemann-Roch theorem*: the Grothendieck-Riemann-Roch generalisation, Bismut-Gillet-Soulé curvature formula); Vojta 1991 *Inventiones Mathematicae* 105, 547-560 (*Siegel's theorem in the compact case*: Arakelov-theoretic upgrade producing an effective Mordell theorem); Soulé-Abramovich-Burnol-Kramer 1992 *Lectures on Arakelov Geometry* (Cambridge Studies in Advanced Mathematics 33); Lang 1988 *Introduction to Arakelov Theory* (Springer-Verlag); Bombieri-Gubler 2006 *Heights in Diophantine Geometry* (Cambridge New Mathematical Monographs 4) Ch. 2-3; Manin-Panchishkin 2005 *Introduction to Modern Number Theory* (Springer EMS 49, 2nd ed.) Ch. 6 §§3-4

References

Arakelov, S. Yu. — Intersection theory of divisors on an arithmetic surface · *Izvestiya Akademii Nauk SSSR, Seriya Matematicheskaya* 38 (1974), 1179-1192. The originator paper. Introduces the archimedean completion of $\mathcal{X} \to \mathrm{Spec}(\mathbb{Z})$ via the addition of Green's functions on $X(\mathbb{C})$, the arithmetic divisor group $\widehat{\mathrm{Div}}(\mathcal{X})$, and the arithmetic intersection pairing $\widehat{D} \cdot \widehat{E} \in \mathbb{R}$. The construction realises the slogan that $\mathrm{Spec}(\mathbb{Z})$ is one-dimensional but lacks a compactification by archimedean places, so $\mathcal{X}$ requires hermitian metric data at infinity to inherit the formal properties of a complete two-dimensional variety.
Faltings, G. — Calculus on arithmetic surfaces · *Annals of Mathematics* 119 (1984), 387-424. The arithmetic Riemann-Roch theorem for an arithmetic surface $\pi : \mathcal{X} \to \mathrm{Spec}(\mathcal{O}_K)$ with hermitian line bundle $\widehat{L}$: $\widehat{\chi}(\widehat{L}) = \frac{1}{2}\widehat{L} \cdot (\widehat{L} - \widehat{K}) + \widehat{\chi}(\mathcal{O}_X) - \frac{1}{12}\delta_F$, where $\delta_F$ is the Faltings delta invariant capturing the geometry of the fibres at infinity. Establishes the arithmetic adjunction formula and the Hodge-index theorem for arithmetic surfaces.
Gillet, H. & Soulé, C. — Arithmetic intersection theory · *Publications Mathématiques de l'IHÉS* 72 (1990), 93-174. Extends the Arakelov pairing to arithmetic varieties of arbitrary dimension. Replaces Green's functions by Green currents on $X(\mathbb{C})$, defines hermitian holomorphic vector bundles $\widehat{E}$ on $\mathcal{X}$, constructs arithmetic Chow groups $\widehat{\mathrm{CH}}^p(\mathcal{X})$, and establishes the projection-formula / pullback / pushforward structure compatible with classical intersection theory.
Gillet, H. & Soulé, C. — An arithmetic Riemann-Roch theorem · *Inventiones Mathematicae* 110 (1992), 473-543; companion paper *Annals of Mathematics* 131-132 (1990-1992). The Grothendieck-Riemann-Roch theorem in arithmetic geometry: for a smooth projective morphism $\pi : \mathcal{X} \to \mathcal{Y}$ of arithmetic varieties, $\widehat{\mathrm{ch}}(\pi_* \widehat{E}) = \pi_*(\widehat{\mathrm{ch}}(\widehat{E}) \cdot \widehat{\mathrm{Td}}(T_{\mathcal{X}/\mathcal{Y}}))$ up to a Bismut-Gillet-Soulé correction term involving analytic torsion.
Vojta, P. — Siegel's theorem in the compact case · *Inventiones Mathematicae* 105 (1991), 547-560. The Arakelov-theoretic adaptation of Vojta's height inequality machinery, producing the new proof of the Mordell conjecture (Faltings 1983). Uses the arithmetic intersection pairing on the self-product $\mathcal{C} \times_S \mathcal{C}$ of an arithmetic curve to bound canonical heights, then combines with a gap principle on rational points to obtain finiteness. Establishes Arakelov geometry as the natural framework for effective Diophantine bounds.
Soulé, C., Abramovich, D., Burnol, J.-F. & Kramer, J. — Lectures on Arakelov Geometry · Cambridge Studies in Advanced Mathematics 33 (Cambridge University Press 1992). The standard textbook: arithmetic Chow groups, arithmetic characteristic classes, the Gillet-Soulé arithmetic Riemann-Roch, the Bismut-Gillet-Soulé curvature formulae, and the Quillen metric on the determinant of cohomology. Builds the entire theory from Green-current foundations and feeds the higher arithmetic-K-theory framework.
Lang, S. — Introduction to Arakelov Theory · Springer-Verlag (1988). The first self-contained graduate textbook on arithmetic surfaces: arithmetic divisors, Green's functions on a compact Riemann surface, the arithmetic intersection pairing, the Faltings height of an abelian variety, and the connection to the Mordell conjecture as recast by Vojta.
Bombieri, E. & Gubler, W. — Heights in Diophantine Geometry · Cambridge New Mathematical Monographs 4 (Cambridge University Press 2006). Ch. 2-3 cover the Arakelov framework, the arithmetic intersection pairing on an arithmetic surface, the Faltings height, and the connection between arithmetic intersection numbers and heights via the Arakelov height. Anchor reference for the formal-and-conceptual statement.
Manin, Yu. I. & Panchishkin, A. — Introduction to Modern Number Theory · Springer Encyclopaedia of Mathematical Sciences 49, 2nd edition (2005). Ch. 6 §§3-4: the Arakelov programme presented as the completion of $\mathrm{Spec}(\mathcal{O}_K)$ by archimedean places, the arithmetic Picard group $\widehat{\mathrm{Pic}}(\mathcal{X})$, the arithmetic intersection number on the relative Picard scheme, and the arithmetic Riemann-Roch theorem of Faltings. The source-book for the present unit.
Bismut, J.-M., Gillet, H. & Soulé, C. — Analytic torsion and holomorphic determinant bundles · *Communications in Mathematical Physics* 115 (1988), 49-78, 79-126, 301-351 (three-paper series). The differential-geometric input to arithmetic Riemann-Roch: the curvature of the Quillen metric on the determinant of cohomology of a holomorphic family, expressed in terms of the Ray-Singer analytic torsion. The Bismut-Gillet-Soulé curvature formula is the archimedean correction term in the Gillet-Soulé arithmetic Riemann-Roch.

Estimated time

beginner: 22m
intermediate: 50m
master: 90m