Riemann-Roch 01: Roadmap
Riemann-Roch notes / I
A first reading of Riemann-Roch
The theorem is best approached as a counting theorem for meromorphic functions with prescribed poles, not as a formula to memorize.
Riemann-Roch is often quoted before its terms have had time to become familiar. That is a bad way to meet the theorem. The useful version is not a slogan about curves, but a rule for a particular counting problem: given a compact Riemann surface \(X\) and a prescription of allowed poles, determine the dimension of the resulting space of meromorphic functions.
The prescription is a divisor
\[D=\sum_P n_P P.\]The vector space being counted is
\[L(D)=\{f\in \mathbb C(X): (f)+D\ge 0\}.\]Here \((f)\) is the divisor of zeros and poles of \(f\). The inequality says that the poles of \(f\) are no worse than the positive coefficients of \(D\), and that negative coefficients of \(D\) impose zeros.
The theorem in one line
For a canonical divisor \(K\) and genus \(g\),
\(\ell(D)-\ell(K-D)=\deg(D)+1-g.\)
The formula is short, but each term has a different origin. The degree is arithmetic. The genus is topology. The term \(\ell(K-D)\) is the part that comes from differentials and duality.
The order of ideas
The notes are arranged so that the formula is not doing all the work at once.
| Step | Question |
|---|---|
| Divisors | How are zeros and poles recorded? |
| \(L(D)\) | Which functions obey a divisor bound? |
| Canonical divisors | Where does the differential term come from? |
| The theorem | How do degree, genus, and the correction fit together? |
| Examples | What changes between \(\mathbb P^1\) and an elliptic curve? |
| Cohomology and codes | Why does the same theorem reappear in modern language? |
| Applications | Why compact Riemann surfaces are algebraic, and why Goppa codes use the same divisor logic. |
A good mental model is the case \(X=\mathbb P^1\) and \(D=n\infty\). Then \(L(D)\) is spanned by
\[1,z,z^2,\ldots,z^n,\]so \(\ell(D)=n+1\). Riemann-Roch is the theorem saying what survives of this elementary count on an arbitrary compact Riemann surface.
A warning about intuition
The degree of \(D\) is only the first approximation. On curves of positive genus, divisors of the same degree may behave differently. The correction term is not decoration; it is exactly where that extra geometry enters.