Binary cubics and the complete fiber description#
The map has an explicit three-valued algebraic inverse. This chapter makes the projective-root interpretation precise.
Throughout this chapter the base field is \(\mathbb C\) (the same arguments work over any algebraically closed field of characteristic zero).
Finite roots#
For a target \((a,b,c)\), dehomogenize its binary cubic at \(V=1\):
Every preimage with \(x\ne0\) has
Conversely, let \(w\) be any simple root and put
Then \(h\ne0\), and the unique corresponding preimage is
Substitution into the rational identities of the previous chapter returns \((A,B,C)=(a,b,c)\).
The root at infinity#
The projective point \([U:V]=[1:0]\) is a root precisely when \(c=0\). It is always simple because \(Q_V(1,0)=-2\). Its corresponding source point is
Indeed, \(F(0,y,z)=(z+4y^2,y,0)\).
Why every simple projective root gives one point#
At any simple root \([u_0:v_0]\), Euler’s identity for the homogeneous cubic implies that its gradient is a nonzero scalar multiple of \((v_0,-u_0)\). There is a unique rescaling \((U,V)=s(u_0,v_0)\) for which
When \(V\ne0\), the preceding affine formula reconstructs \((x,y,z)\). When \(V=0\), the explicit infinity formula does. This is inverse to the marked-root construction from the source.
More precisely, let
Then
is an isomorphism, not merely a bijection on complex points. The finite-root formula gives its regular inverse on \(V\ne0\). On \(U\ne0\), put
Simplicity is exactly \(k\ne0\), and a regular inverse on this chart is
At \(s=0\) this reduces to the root-at-infinity formula.
Fiber cardinalities and image#
A nonzero binary cubic over \(\mathbb C\) has one of three root patterns:
three simple projective roots;
one double root and one simple root;
one triple root.
Only simple roots produce source points. Therefore the corresponding fiber cardinalities are exactly \(3\), \(1\), and \(0\).
The triple-root cubics are parametrized by
Consequently the set-theoretic image on complex points is
Because \(F\) is dominant, its scheme-theoretic image—and the Zariski closure of its point-set image—is still all of \(\mathbb A^3\). The omitted curve is also the reduced singular locus of the discriminant hypersurface.
For example \((1/3,2,2/3)\) is omitted because
Independently, a Gröbner basis gives
Generic degree and monodromy#
The generic cubic has three distinct roots, so the map has geometric and function-field degree \(3\):
There is also a direct function-field proof. Over \(k=\mathbb C(b,c)\), solve the root equation for the remaining coefficient:
The rational map \(h:\mathbb P^1\to\mathbb P^1\) has degree three, so
The inverse formulas show that \(w\) generates the full source function field over \(\mathbb C(a,b,c)\); hence its cubic is the minimal polynomial and the map has geometric degree three.
The discriminant factor \(\Delta\) is not a square. Viewed as a quadratic in \(a\), its own discriminant is
which is nonsquare in \(\mathbb C(b,c)\). Thus \(\Delta\) is irreducible and has odd valuation along its divisor. The degree-three extension is generically nonnormal (and has trivial automorphism group), while its degree-six normal closure—and the geometric monodromy—has group \(S_3\).
“Generically three-to-one” is the correct phrase. The exact fiber sizes vary on the discriminant and omitted loci.