Arithmetic geometry asks geometric questions over arithmetic bases: not only what shape polynomial equations define, but which of their points have rational, integral, finite-field or local coordinates.
The earlier R24 route built affine varieties, projective varieties, schemes and sheaves. Arithmetic geometry changes the base: equations are studied over Q, number fields, Z, local fields and finite fields, where existence of points becomes a number-theoretic problem.
A curve may have many complex points but no rational point. An equation may have solutions modulo every small prime yet no integer solution. An elliptic curve may have finitely many torsion points but infinitely many rational points generated by a finite-rank group. A model over Z can have smooth fibres for most primes and singular fibres at a finite exceptional set.
Level: advanced undergraduate / beginning graduate enrichment. Prerequisites: elementary number theory, fields, ideals, affine/projective algebraic geometry and basic schemes.
Reading route: rational points → integral models → reduction mod p → local fields → local–global principles → conics → elliptic curves → Mordell–Weil → heights and descent → finiteness theorems → Galois and finite-field bridges → practice → solutions.
1. The base field changes the question
Let X be an algebraic variety defined over a field K. The notation X(K) means its K-rational points.
If X is cut out by polynomial equations with coefficients in K, then X(K) consists of solutions whose coordinates lie in K.
Changing K changes the point set. For x²+y²=−1, there are no real points but many complex points. For x²+y²=3, there may be real points while rational solvability is a stricter arithmetic question.
Arithmetic geometry therefore keeps two questions distinct: what is the geometric object after extending to an algebraic closure, and which points are defined over the arithmetic field of interest?
2. Rational points are morphisms from the base field
Scheme-theoretically, a K-rational point of a K-scheme X is a morphism
Spec K→X
over Spec K.
For an affine scheme X=Spec A over K, such a point corresponds to a K-algebra homomorphism A→K.
This definition survives beyond coordinate tuples. It makes rational points functorial under field extension and fits naturally into the scheme language.
3. Integral points ask for coordinates without denominators
If an affine variety is defined by equations over Z, its integral points are solutions in Zⁿ.
For example, x²+y²=25 has finitely many integer solutions even though its real locus is a full circle.
Integral points depend on the chosen affine model. Removing or adding points at infinity changes which rational functions are required to remain integral.
Over a number field K, one replaces Z by its ring of integers O_K, or by rings of S-integers when selected primes are allowed in denominators.
4. Equations over Z create one family across all primes
An equation with integer coefficients defines a scheme X over Spec Z.
The generic fibre is X_Q=X×_{Spec Z}Spec Q.
For each prime p, the special fibre is X_{F_p}=X×_{Spec Z}Spec F_p.
Thus one integral model packages the rational geometry and all reductions modulo primes into a single morphism X→Spec Z.
This is the geometric meaning of reducing an equation modulo p: it is passing to the fibre over the prime point (p) of Spec Z.
5. Worked reduction example: a conic modulo p
Take X:x²+y²=1 over Z.
Over Q it is a conic with rational point (1,0), hence rationally parametrizable.
Modulo p, the equation becomes x²+y²=1 in F_p.
For p=2, the equation behaves differently because 2=0 and quadratic-form identities can degenerate. For odd p, the fibre remains a plane conic; smoothness is checked using the partial derivatives 2x and 2y together with the defining equation.
The example illustrates why finitely many primes often require special treatment.
6. Good reduction means the geometry survives modulo p
Informally, a smooth projective variety over Q has good reduction at a prime p if it admits a suitable smooth proper model over Z_p whose special fibre remains smooth.
Bad reduction means singularity or another structural defect appears in the special fibre.
For an elliptic curve given by an integral Weierstrass equation, primes dividing the discriminant of a minimal model are the bad-reduction primes.
Good reduction lets arithmetic information move between characteristic zero and finite characteristic in controlled ways.
7. Local fields zoom in on one prime at a time
The p-adic field Q_p is the completion of Q with respect to the p-adic absolute value.
Two rational numbers are p-adically close when their difference is divisible by a high power of p.
The p-adic integers Z_p are the elements of Q_p with nonnegative p-adic valuation.
Q_p is not a subfield of R with a strange coordinate system. It is a different completion of Q with its own topology and convergence.
Arithmetic geometry studies X(Q_p) alongside X(R) and X(Q).
8. Hensel lifting turns modular roots into p-adic roots
A basic form of Hensel’s lemma says that if f(a)≡0 mod p and f'(a) is not divisible by p, then the root a modulo p lifts uniquely to a root in Z_p satisfying the same residue condition.
The exact hypotheses have several equivalent formulations, especially when higher p-adic precision is used.
This makes a nonsingular solution modulo p a local certificate for a p-adic solution.
For the algorithmic side of lifting and valuations, see BTT’s p-Adic Valuations and Hensel Lifting guide. The present article uses Hensel’s lemma as arithmetic-geometric infrastructure rather than duplicating the algorithm.
9. Local solvability is necessary for rational solvability
If X has a rational point, then it has a point over R and over every Q_p, because Q embeds into each completion.
Therefore failure to have a point in even one completion proves X(Q)=∅.
This is the local obstruction strategy: test easier local fields first.
The converse is much deeper. Having points everywhere locally does not imply a rational point for arbitrary varieties.
10. The Hasse principle works in important special cases
The Hasse–Minkowski theorem says a quadratic form over Q represents zero nontrivially over Q exactly when it does so over R and every Q_p.
Equivalently, suitable conic and quadratic-form rational-point questions satisfy a local–global principle.
This success is powerful but should not be extrapolated blindly. Higher-degree equations and higher-genus curves can violate the Hasse principle.
Arithmetic geometry therefore asks both whether local points exist and whether there are further global obstructions.
11. Conics are the first complete local–global laboratory
A smooth projective conic over Q with one rational point is isomorphic to P¹ over Q.
Once a rational point is known, drawing lines of rational slope through that point gives a rational parametrisation.
For x²+y²=z², the point [1:0:1] produces the classical Pythagorean parametrisation.
A conic without a rational point is not isomorphic to P¹ over Q even though it becomes P¹ after extending to an algebraic closure.
This separates geometric type from arithmetic rationality.
12. Worked parametrisation of the unit circle
Take x²+y²=1 and the rational point (−1,0).
A line through it with slope t is y=t(x+1).
Substituting and removing the known intersection yields
x=(1−t²)/(1+t²), y=2t/(1+t²).
For t∈Q these give rational points on the circle, and every rational point other than (−1,0) arises this way.
This is a model example of turning one rational point into a full rational parametrisation on a genus-zero curve.
13. Elliptic curves begin where genus-zero parametrisation stops
An elliptic curve over a field K is a smooth projective curve of genus one together with a specified K-rational point.
Over fields of characteristic not 2 or 3, it can often be written in short Weierstrass form
y²=x³+Ax+B
with discriminant
Δ=−16(4A³+27B²)≠0.
The nonzero discriminant is exactly the nonsingularity condition for this short form.
In characteristics 2 and 3, more general Weierstrass equations are needed.
14. The rational points of an elliptic curve form a group
The chosen rational point serves as the identity, conventionally the point at infinity O in a Weierstrass model.
For a plane cubic, a line through two points P and Q meets the cubic in a third point R counting multiplicity. Reflecting R across the x-axis in short Weierstrass coordinates gives P+Q.
Tangent lines define doubling P+P.
The geometric construction satisfies the axioms of an abelian group.
For explicit finite-field point arithmetic and scalar multiplication, use BTT’s separate Elliptic Curves over Finite Fields computational guide.
15. Mordell–Weil turns infinitely many points into finite generators
The Mordell–Weil theorem states that for an elliptic curve E over a number field K, the group E(K) is finitely generated.
Therefore
E(K)≈E(K)_tors ⊕ Z^r
for a finite torsion subgroup and a nonnegative integer r called the rank.
Rank zero means all K-rational points are torsion and hence finite in number. Positive rank means infinitely many rational points generated from finitely many independent points.
Finding generators and proving a claimed rank can be computationally difficult even though finite generation is guaranteed.
16. Heights measure arithmetic size geometrically
For a rational number x=a/b in lowest terms, a naive logarithmic height records roughly log max(|a|,|b|).
On projective space, heights measure the arithmetic complexity of homogeneous coordinates while respecting common scaling.
On an elliptic curve, the canonical Néron–Tate height ĥ(P) modifies the naive height so that
ĥ(nP)=n²ĥ(P).
It behaves like a positive-definite quadratic form modulo torsion and is central to the proof and computation of Mordell–Weil structure.
17. Descent replaces an infinite search by a finite obstruction problem
Descent studies the image of E(K) under multiplication-by-n and compares it with easier local or cohomological data.
The n-Selmer group is finite and sits between E(K)/nE(K) and the n-torsion part of the Tate–Shafarevich group.
Computing a Selmer group gives an upper bound on the Mordell–Weil rank.
A gap between the Selmer bound and known rational points can reflect hidden global obstruction data rather than a missing elementary search.
This is a cohomological local–global mechanism, not simply a faster enumeration of rational coordinates.
18. The Tate–Shafarevich group records locally soluble torsors without global points
For an elliptic curve E/K, the Tate–Shafarevich group Ш(E/K) consists, roughly, of principal homogeneous spaces for E that have points over every completion of K but may have no K-rational point.
It therefore measures a failure of the Hasse principle in the elliptic-curve setting.
The finiteness of Ш(E/K) is conjectured in general and is not currently known for all elliptic curves.
Any statement that assumes this finiteness should say so explicitly.
19. Good reduction connects rational torsion to finite groups
At a prime of good reduction, an elliptic curve over Q reduces to a smooth elliptic curve over F_p.
There is a reduction map from suitable p-adic or rational points into E(F_p).
For torsion of order prime to p, good reduction gives strong injectivity statements under the standard hypotheses.
This lets finite-field group orders constrain possible rational torsion.
The exact statement depends on the local field, model and torsion order, so “reduction is injective on all rational points” is false and should not be used.
20. Hasse’s bound controls elliptic curves over finite fields
For an elliptic curve E over F_q,
|#E(F_q)−(q+1)|≤2√q.
Thus the number of rational points is close to q+1, with an error controlled by the square root of the field size.
Writing #E(F_q)=q+1−a_q, the integer a_q is the Frobenius trace.
Those traces become inputs to zeta functions, L-functions and Galois representations.
For the representation-theoretic bridge, see Galois Representations | Frobenius Traces, Determinants and Arithmetic Symmetry.
21. Zeta functions organise point counts over finite extensions
For a variety X over F_q, its zeta function packages the numbers #X(F_{q^n}) for all n≥1:
Z(X,t)=exp(Σ_{n≥1} #X(F_{q^n})t^n/n).
For suitable varieties, the Weil conjectures—proved by Dwork, Grothendieck’s school and Deligne—show remarkable rationality, functional-equation and eigenvalue structure.
These results connect finite-field arithmetic to cohomology and topology-like invariants.
This guide uses the zeta function as a bridge rather than attempting the étale-cohomology machinery needed for the full theory.
22. Genus changes the expected arithmetic of curves
For smooth projective curves over a number field, genus gives a useful first division of arithmetic behaviour.
- Genus 0: with a rational point, the curve is rational and usually parametrizable.
- Genus 1: with a rational point, the curve is an elliptic curve and rational points form a finitely generated group.
- Genus at least 2: rational points are finite by Faltings’ theorem.
This classification is structural, not an algorithm that lists the points. Finiteness can be known while finding every point remains difficult.
23. Faltings’ theorem gives a profound finiteness boundary
Faltings’ theorem states that a smooth projective curve of genus at least 2 over a number field K has only finitely many K-rational points.
The theorem does not give a simple universal procedure for listing those points.
It settles the qualitative finiteness statement formerly known as the Mordell conjecture.
Arithmetic geometry repeatedly separates existence, finiteness, effective bounds and actual computation; they are different mathematical jobs.
24. Siegel’s theorem controls integral points on affine curves
A useful form of Siegel’s theorem says that for a smooth affine curve over a number field, S-integral points are finite when the logarithmic Euler characteristic is negative—equivalently, in the standard curve setting, when
2g−2+r>0,
where g is the genus of a smooth projective completion and r is the number of points removed at infinity, after the relevant field and divisor conditions are accounted for.
For genus at least 1, many familiar affine curve models therefore have only finitely many integral points.
The exact S-integral formulation matters; saying simply “every positive-genus curve has finitely many integer points” hides model and infinity data.
25. The Birch and Swinnerton-Dyer conjecture links rank to an L-function
For an elliptic curve E/Q, the Birch and Swinnerton-Dyer conjecture predicts that the Mordell–Weil rank equals the order of vanishing of the L-function L(E,s) at s=1.
Its refined form relates the leading Taylor coefficient to the regulator, torsion, real period, Tamagawa factors and the order of the Tate–Shafarevich group.
BSD remains open in general.
It is a model frontier statement because it links analytic data, algebraic group structure and local arithmetic in one conjectural formula.
26. Galois actions record how points move under field symmetry
Let K̄ be an algebraic closure of K. The absolute Galois group Gal(K̄/K) acts on algebraic points X(K̄).
K-rational points are fixed by this Galois action.
For an elliptic curve, the n-torsion subgroup E[n] carries a Galois representation into GL₂(Z/nZ) under the usual hypotheses.
Passing to ℓ-adic Tate modules gives representations into GL₂(Z_ℓ).
Frobenius elements at good primes encode point-count data through their traces and determinants.
27. Algebraic number fields provide the arithmetic base rings
A number field K is a finite extension of Q. Its ring of integers O_K replaces Z as the natural integral base.
Prime ideals of O_K replace ordinary prime numbers, and residue fields O_K/p are finite.
Splitting, ramification and inertia describe how rational primes factor after extension.
For the algorithmic ideal, norm and basis machinery, use Algebraic Number Fields, Ideals, Norms and Computational Arithmetic.
Arithmetic geometry builds varieties and schemes over these rings rather than treating the number-field computations as an isolated topic.
28. Local–global is a diagnostic architecture, not a universal theorem
A dependable arithmetic workflow tests global claims against several layers:
- real solvability;
- solvability modulo primes;
- p-adic solvability;
- Galois and descent constraints;
- global height or finiteness arguments;
- explicit point search only after structural bounds are understood.
Passing every local test means “no local obstruction has been found”, not “a rational point is guaranteed”.
The gap between those statements is one of the central mathematical spaces in arithmetic geometry.
29. Common misconceptions
If a variety has complex points, it should have rational points. False. Rationality is an arithmetic constraint.
If an equation is soluble modulo every prime, it has an integer solution. False in general.
Points over every Q_p imply a point over Q. True for important quadratic cases, false for general varieties.
Every elliptic curve has infinitely many rational points. False. Rank can be zero.
Good reduction means reduction preserves every rational point injectively. False. Precise statements depend on torsion order and local hypotheses.
BSD is a theorem. No. It remains a major open conjecture in general.
30. Independent practice: twenty questions
- What does X(Q) mean?
- How is a K-rational point described scheme-theoretically?
- What is the difference between a rational point and an integral point on an affine model?
- What is the generic fibre of a scheme over Z?
- What is the fibre over a prime p?
- Why can primes dividing a discriminant require special treatment?
- What does Q_p complete?
- State a simple Hensel-lifting condition.
- Why is local solvability necessary for rational solvability?
- For what major family does Hasse–Minkowski give a local–global theorem?
- Parametrize rational points on x²+y²=1.
- What extra datum turns a genus-one curve into an elliptic curve?
- What does Δ≠0 ensure in short Weierstrass form?
- State the Mordell–Weil structure of E(K).
- What does positive rank imply?
- State the quadratic scaling law for canonical height.
- What does a Selmer computation bound?
- State Hasse’s point-count bound for E/F_q.
- What does Faltings’ theorem say for genus at least 2?
- What does BSD predict about rank?
31. Worked solutions and checks
1. The set of points of X whose coordinates or residue map are defined over Q.
2. As a morphism Spec K→X over Spec K.
3. Rational coordinates may contain denominators; integral points lie in the selected integral model, such as Zⁿ or O_Kⁿ.
4. Base change to Spec Q.
5. Base change to Spec F_p, giving reduction modulo p.
6. The special fibre can become singular or otherwise fail to preserve the smooth structure.
7. Q with respect to the p-adic absolute value.
8. If f(a)≡0 mod p and f'(a) is nonzero mod p, then a lifts uniquely to a p-adic root in the corresponding residue class.
9. A rational point maps into every completion, so absence of a local point rules out a rational one.
10. Quadratic forms, and equivalently the associated conic-type rational-point problems under the standard hypotheses.
11. With slope t through (−1,0): x=(1−t²)/(1+t²), y=2t/(1+t²).
12. A chosen rational point, used as the group identity.
13. Nonsingularity of the short Weierstrass cubic.
14. E(K)≈E(K)_tors⊕Z^r with finite torsion and finite rank r.
15. Infinitely many K-rational points.
16. ĥ(nP)=n²ĥ(P).
17. It gives an upper bound on the Mordell–Weil rank by controlling E(K)/nE(K) through a finite group.
18. |#E(F_q)−(q+1)|≤2√q.
19. A smooth projective curve of genus at least 2 over a number field has finitely many rational points.
20. That rank E(Q) equals the order of vanishing of L(E,s) at s=1; this remains conjectural in general.
32. Where this guide hands off
The last R24 cell turns selected algebraic-geometric tasks into explicit algorithms and counts: Gröbner bases, elimination, resultants, Hilbert functions, singular loci, projective saturation and intersection numbers.
Use the existing computational number-theory estate when the reader’s task is modular arithmetic, finite-field algorithms, factorisation, number-field ideals or explicit elliptic-curve scalar multiplication. Arithmetic geometry owns the larger local–global and rational-point architecture that connects those tools.
Sources and further study
See the Stacks Project for arithmetic schemes, fibres, smoothness and cohomological foundations, and MIT 18.782 Introduction to Arithmetic Geometry for curves, rational points, elliptic curves and number-theoretic methods. The BTT computational-number-theory and Galois-representation guides provide specialist algorithmic and symmetry bridges.