# CM curves, the invariant derivation and divisibility

The chapter [CM curves, the invariant derivation and divisibility](/proof/cm-divisibility/) compares four steps of the timeline with the code. The code has a CM elliptic curve, a formula for the $p$-th power of an invariant derivation, $l$-adic bounds on derivatives of Eisenstein sums over periods, a norm bound from the Chowla-Selberg formula, and a gcd of depths. The pieces sit in different branches.

## Read first

- [Integers and congruences](/learn/algebra/#integers-and-congruences): the greatest common divisor and Bezout, for Step 8.
- [Polynomials and power series](/learn/algebra/#polynomials-and-power-series). Step 2 computes with formal power series.
- [Ideals and the class group](/learn/number-fields/#ideals-and-the-class-group). Step 6 averages over the ideal class group.
- [The splitting of primes](/learn/number-fields/#the-splitting-of-primes): the timeline wants $p$ inert, but in the code $p$ has ramification.
- [Zeta functions of number fields](/learn/l-functions/#zeta-functions-of-number-fields) covers the partial zeta function, the class number formula and the Kronecker limit formula.
- [Tate's thesis and Hecke L-functions](/learn/class-field-theory/#tates-thesis-and-hecke-l-functions). The number $\beta$ of Step 7 comes from a Hecke L-value.
- [Lattices and the Weierstrass function](/learn/elliptic-curves/#lattices-and-the-weierstrass-function): $g_2$, $g_3$, the discriminant $\Delta(\Lambda)$ and the Eisenstein sum of Step 4.
- [Reduction and minimal models](/learn/elliptic-curves/#reduction-and-minimal-models) and [quadratic twists](/learn/elliptic-curves/#quadratic-twists). Step 1 twists a curve to reach the minimal discriminant $-p^3$.
- [Complex multiplication](/learn/elliptic-curves/#complex-multiplication) gives $j(L)$ and Shimura reciprocity.
- [The Chowla-Selberg formula](/learn/elliptic-curves/#the-chowla-selberg-formula) is Step 6.
- [Gross's curves](/learn/elliptic-curves/#grosss-curves) are the curves of Step 1.
- [The invariant derivation and the logarithm](/learn/formal-groups/#the-invariant-derivation-and-the-logarithm), for $D_u$ and a quasi-logarithm.
- [Prismatic Dieudonné theory](/learn/formal-groups/#prismatic-dieudonn-theory), the deepest: Step 3 uses the formula for a p-divisible group $G$ and modules over $A_{\mathrm{inf}}$.

## The new objects

- CMCurveGoodOutsidePrimeSt says: for each prime $p > 3$ with $p \equiv 3 \pmod 4$ there is an elliptic curve over $\mathbb{Q}(j)$ with complex multiplication by the integers of $\mathbb{Q}(\sqrt{-p})$, good reduction outside $p$ and minimal discriminant $-p^3$. It enters through [published_results_eigen](/step/published-results-eigen/).
- For a lift $\mu$ of a formal group law, $D_u$ is the invariant derivation in the direction $u$. The formula of Step 2: the constant term of $D_u^{\,p}(X_j)$ is $\sum_i c_i u_i^{\,p}$ modulo $\ker\theta + (p)$, with $c_i$ read from the $X_i^p$ terms of a quasi-logarithm.
- JetInt says that derivatives at $0$ of an Eisenstein lattice sum for a CM field and a CM type, divided by powers of periods, have bounded $l$-adic size.
- LatIntRam and LatIntUnr bound $v_l(\beta)$ from below, where $\beta = (C_0 T)^{12} / \Delta(\Lambda)^{\mathrm{wt}}$ for a value $T$ of a lattice sum. ToricDenominator adds bounds on denominators of partial zeta values.
- BetaArch bounds the log of the norm of $\beta$, read off its minimal polynomial.
- DepthOK says that a product of local test functions vanishes unless $Kx$ and $Ky$ lie in given lattices, and has period $K$. Periods $k$ and $k'$ give the period $\gcd(k, k')$.

## The path through the chapter

1. Gross's curve gives the field data of the host field (Step 1).
2. The $D^p$ formula makes a dual Frobenius nilpotent, inside the proof of a prismatic Dieudonné statement (Steps 2 and 3).
3. JetInt, proved from the functional equation of partial zeta functions, gives the bounds on $v_l(\beta)$ (Steps 4 and 5).
4. Shimura reciprocity gives the conjugates of $\beta$, and Chowla-Selberg averages their sizes to bound the norm (Steps 6 and 7).
5. Bezout gives the gcd of depths, so the least depth divides each valid depth (Step 8). Both bounds feed the [unitary hypotheses](/proof/unitary-hypotheses/).
