Modular curves and congruence subgroups

The principal subgroup Γ(N) consists of matrices congruent to I modulo N. The subgroup Γ₁(N) requires a and d congruent to one and c congruent to zero, while Γ₀(N) requires only c congruent to zero. Thus Γ(N) lies inside Γ₁(N), which lies inside Γ₀(N). A congruence subgroup is one containing some Γ(N). Reduction modulo N gives finite index, but finite-index subgroups need not be congruence. Their analytic quotient constructions still exist even when the arithmetic theory differs.

Over the complex numbers, Γ₀(N) describes an elliptic curve together with a cyclic subgroup of order N, while Γ₁(N) remembers a point of exact order N. Full level remembers a basis of N-torsion, with a Weil-pairing choice when one fixes a connected component. These are coarse moduli descriptions unless automorphisms and fine-moduli hypotheses are dealt with. The central sign belongs to Γ₀(N), but not Γ₁(N) for N>2, so an SL₂ index may differ from the effective geometric index in other families.

For a finite-index group Γ, the quotient YΓ of H by Γ is a noncompact Riemann surface. Its cusp classes are Γ-orbits in the rational boundary together with infinity; there are finitely many because the index is finite. Adding one point for each class with the exponential cusp topology gives the compact Riemann surface XΓ. Merely adjoining every real boundary point would produce the wrong space. At elliptic points, local power coordinates account for finite stabilizers and supply the coarse surface's complex charts.

If σ sends infinity to a selected cusp, its width h is the least positive integer such that the effective image of σTʰσ⁻¹ belongs to Γ. For even trivial-character forms, f|ₖσ is periodic with period h and uses local coordinate exp(2πiτ/h). For Γ₀(N), a cusp represented by a/c in lowest terms has width N/gcd(c²,N). This includes infinity with c=0. It makes the cusp at infinity width one and the cusp zero width N, even though both are ends of the same modular curve.

A weight-two form makes f(τ)dτ invariant because d(γτ)=(cτ+d)⁻²dτ. Near a cusp of width h, this differential is h/(2πi) times f(q)dq/q. It extends holomorphically exactly when the form vanishes at that cusp. Thus weight-two cusp forms identify with holomorphic differentials on the compact curve, whose dimension is its genus. A nonzero constant term would create a simple differential pole. This is a geometric reason the cusp condition matters beyond boundedness of a function.

Every compact Riemann surface is algebraic over C, but an arithmetic application needs a distinguished model over a smaller field. The standard curve X₀(N) has a canonical model over Q, compatible with its elliptic-curve-and-cyclic-subgroup moduli interpretation. A connected full-level curve with fixed Weil pairing naturally has a model over a cyclotomic field; one should not declare every chosen level component canonically over Q without qualification. Rational models and rational maps are extra arithmetic structure, which the modularity theorem will use in a way not visible from a complex tessellation alone.