ConceptComplete

Ruled Surfaces

A ruled surface is a surface that admits a fibration over a curve with every fiber isomorphic to P1\mathbb{P}^1. These surfaces are among the best-understood in algebraic geometry: their Picard group, intersection theory, and canonical class can all be computed explicitly. They play a central role in the Enriques--Kodaira classification as the surfaces of Kodaira dimension βˆ’βˆž-\infty.


Ruled and geometrically ruled surfaces

Definition5.7Ruled surface

A smooth projective surface XX is ruled if it is birational to CΓ—P1C \times \mathbb{P}^1 for some smooth projective curve CC. Equivalently, XX contains a pencil of rational curves covering XX.

A ruled surface over CC is a smooth projective surface XX together with a surjective morphism π ⁣:Xβ†’C\pi \colon X \to C to a smooth curve CC such that every fiber Ο€βˆ’1(p)\pi^{-1}(p) is isomorphic to P1\mathbb{P}^1.

Definition5.8Geometrically ruled surface

A geometrically ruled surface over a smooth curve CC is a surface of the form

X=P(E)X = \mathbb{P}(\mathcal{E})

where E\mathcal{E} is a rank-22 locally free sheaf on CC, and P(E)\mathbb{P}(\mathcal{E}) denotes the projectivization Proj⁑(Sym⁑(E))\operatorname{Proj}(\operatorname{Sym}(\mathcal{E})).

The natural projection π ⁣:P(E)β†’C\pi \colon \mathbb{P}(\mathcal{E}) \to C makes XX into a P1\mathbb{P}^1-bundle over CC. Note that P(E)β‰…P(EβŠ—L)\mathbb{P}(\mathcal{E}) \cong \mathbb{P}(\mathcal{E} \otimes \mathcal{L}) for any line bundle L\mathcal{L} on CC, so one may always normalize E\mathcal{E} so that H0(C,E)β‰ 0H^0(C, \mathcal{E}) \neq 0 but H0(C,EβŠ—Lβˆ’1)=0H^0(C, \mathcal{E} \otimes \mathcal{L}^{-1}) = 0 for every line bundle L\mathcal{L} of positive degree.

RemarkRuled vs. geometrically ruled

Every geometrically ruled surface is ruled, but the converse is slightly subtle. A ruled surface need not admit a P1\mathbb{P}^1-bundle structure globally -- it may have singular fibers after performing a minimal model reduction. However, every minimal ruled surface is geometrically ruled: it is isomorphic to P(E)\mathbb{P}(\mathcal{E}) for some rank-22 bundle E\mathcal{E} on a curve CC.

ExampleThe trivial ruled surface

When E=OCβŠ•OC\mathcal{E} = \mathcal{O}_C \oplus \mathcal{O}_C, we get P(E)β‰…CΓ—P1\mathbb{P}(\mathcal{E}) \cong C \times \mathbb{P}^1, the trivial product. This is the simplest geometrically ruled surface over CC.

For C=P1C = \mathbb{P}^1: P1Γ—P1\mathbb{P}^1 \times \mathbb{P}^1 is the smooth quadric QβŠ‚P3Q \subset \mathbb{P}^3, with two rulings giving two P1\mathbb{P}^1-bundle structures.


Hirzebruch surfaces

Definition5.9Hirzebruch surfaces

For each integer nβ‰₯0n \geq 0, the nn-th Hirzebruch surface is

Fn=P(OP1βŠ•OP1(βˆ’n)).\mathbb{F}_n = \mathbb{P}(\mathcal{O}_{\mathbb{P}^1} \oplus \mathcal{O}_{\mathbb{P}^1}(-n)).

This is a geometrically ruled surface over P1\mathbb{P}^1 with projection π ⁣:Fnβ†’P1\pi \colon \mathbb{F}_n \to \mathbb{P}^1. The Hirzebruch surfaces are (up to isomorphism) exactly the geometrically ruled surfaces over P1\mathbb{P}^1.

ExampleFβ‚€ = β„™ΒΉ Γ— β„™ΒΉ

The surface F0=P(OβŠ•O)\mathbb{F}_0 = \mathbb{P}(\mathcal{O} \oplus \mathcal{O}) is the product P1Γ—P1\mathbb{P}^1 \times \mathbb{P}^1. It is the smooth quadric surface QβŠ‚P3Q \subset \mathbb{P}^3 defined by xw=yzxw = yz. It has two distinct rulings:

  • The projection to the first factor, and
  • The projection to the second factor.

The Picard group is Pic⁑(F0)β‰…ZF1βŠ•ZF2\operatorname{Pic}(\mathbb{F}_0) \cong \mathbb{Z} F_1 \oplus \mathbb{Z} F_2 with F12=F22=0F_1^2 = F_2^2 = 0 and F1β‹…F2=1F_1 \cdot F_2 = 1.

ExampleF₁ = Blowup of β„™Β² at a point

The Hirzebruch surface F1\mathbb{F}_1 is isomorphic to the blowup Bl⁑P(P2)\operatorname{Bl}_P(\mathbb{P}^2) of P2\mathbb{P}^2 at one point PP. The ruling π ⁣:F1β†’P1\pi \colon \mathbb{F}_1 \to \mathbb{P}^1 corresponds to the pencil of lines in P2\mathbb{P}^2 through PP:

  • Each line through PP becomes a fiber Fβ‰…P1F \cong \mathbb{P}^1 of Ο€\pi.
  • The exceptional divisor EE of the blowup is the unique section with E2=βˆ’1E^2 = -1.
  • The strict transform of a line not through PP gives a section C0C_0 with C02=1C_0^2 = 1.

Note: F1\mathbb{F}_1 is not minimal because it contains the (βˆ’1)(-1)-curve EE. Contracting EE recovers P2\mathbb{P}^2.

ExampleFβ‚‚ and the rational normal cone

The surface F2=P(OβŠ•O(βˆ’2))\mathbb{F}_2 = \mathbb{P}(\mathcal{O} \oplus \mathcal{O}(-2)) can be realized as the minimal resolution of the quadric cone V(xzβˆ’y2)βŠ‚P3V(xz - y^2) \subset \mathbb{P}^3. More precisely:

  • The quadric cone has a singular point at the vertex [0:0:0:1][0:0:0:1].
  • Blowing up the vertex resolves the singularity, and the resulting surface is F2\mathbb{F}_2.
  • The exceptional curve of the resolution is the unique section C0C_0 with C02=βˆ’2C_0^2 = -2.

Alternatively, F2\mathbb{F}_2 is the image of the 22-uple embedding of P1\mathbb{P}^1 scrolled over P1\mathbb{P}^1: the cone over the rational normal curve of degree 22 in P2\mathbb{P}^2 (a conic).

ExampleF_n for large n

The surface Fn\mathbb{F}_n for nβ‰₯1n \geq 1 can be realized as the projectivized cone over the rational normal curve of degree nn in Pn\mathbb{P}^n:

  • Embed P1β†ͺPn\mathbb{P}^1 \hookrightarrow \mathbb{P}^n via [s:t]↦[sn:snβˆ’1t:β‹―:tn][s:t] \mapsto [s^n : s^{n-1}t : \cdots : t^n] (the nn-uple Veronese).
  • Take the cone over this curve in Pn+1\mathbb{P}^{n+1}.
  • Resolve the vertex of the cone to get Fn\mathbb{F}_n.

The negative section C0C_0 with C02=βˆ’nC_0^2 = -n is the exceptional curve of this resolution. For nβ‰₯2n \geq 2, C0C_0 is the unique irreducible curve with negative self-intersection on Fn\mathbb{F}_n.


Sections, fibers, and the Picard group

Definition5.10The invariant e and the Picard group

Let π ⁣:X=P(E)β†’C\pi \colon X = \mathbb{P}(\mathcal{E}) \to C be a geometrically ruled surface with E\mathcal{E} normalized (i.e., H0(E)β‰ 0H^0(\mathcal{E}) \neq 0 but H0(EβŠ—Lβˆ’1)=0H^0(\mathcal{E} \otimes \mathcal{L}^{-1}) = 0 for all L\mathcal{L} of positive degree). Define:

  • Fiber class FF: a fiber Ο€βˆ’1(p)β‰…P1\pi^{-1}(p) \cong \mathbb{P}^1.
  • Tautological section C0C_0: the section corresponding to the surjection Eβ† OP(E)(1)\mathcal{E} \twoheadrightarrow \mathcal{O}_{\mathbb{P}(\mathcal{E})}(1).
  • The invariant e=βˆ’deg⁑(E)=βˆ’deg⁑(det⁑E)e = -\deg(\mathcal{E}) = -\deg(\det \mathcal{E}) when E\mathcal{E} is normalized. Equivalently, e=βˆ’C02e = -C_0^2.

Then Pic⁑(X)β‰…ZC0βŠ•Ο€βˆ—Pic⁑(C)\operatorname{Pic}(X) \cong \mathbb{Z} C_0 \oplus \pi^*\operatorname{Pic}(C), and numerically, NS⁑(X)β‰…ZC0βŠ•ZF\operatorname{NS}(X) \cong \mathbb{Z} C_0 \oplus \mathbb{Z} F.

ExamplePicard group of Hirzebruch surfaces

For Fn\mathbb{F}_n, the Picard group is Pic⁑(Fn)β‰…ZC0βŠ•ZF\operatorname{Pic}(\mathbb{F}_n) \cong \mathbb{Z} C_0 \oplus \mathbb{Z} F where:

  • C0C_0 is the section with C02=βˆ’nC_0^2 = -n,
  • FF is a fiber of π ⁣:Fnβ†’P1\pi \colon \mathbb{F}_n \to \mathbb{P}^1.
  • C0β‹…F=1C_0 \cdot F = 1 (a section meets each fiber in one point).
  • F2=0F^2 = 0 (two distinct fibers are disjoint).

The intersection matrix in the basis (C0,F)(C_0, F) has entries C02=βˆ’nC_0^2 = -n, C0β‹…F=Fβ‹…C0=1C_0 \cdot F = F \cdot C_0 = 1, F2=0F^2 = 0, with determinant βˆ’1-1, confirming NS⁑(Fn)\operatorname{NS}(\mathbb{F}_n) is unimodular.

Any section σ ⁣:P1β†’Fn\sigma \colon \mathbb{P}^1 \to \mathbb{F}_n defines a divisor Σ∼C0+mF\Sigma \sim C_0 + mF for some integer mβ‰₯0m \geq 0, with Ξ£2=βˆ’n+2mβ‰₯βˆ’n\Sigma^2 = -n + 2m \geq -n. The section C0C_0 achieves the minimum self-intersection βˆ’n-n.


Intersection theory on ruled surfaces

TheoremIntersection calculus on ruled surfaces

Let π ⁣:X=P(E)β†’C\pi \colon X = \mathbb{P}(\mathcal{E}) \to C be a geometrically ruled surface with invariant ee, section C0C_0, and fiber FF. Any divisor DD on XX is numerically equivalent to aC0+bFaC_0 + bF for a,b∈Za, b \in \mathbb{Z}. The intersection numbers are:

  • C02=βˆ’eC_0^2 = -e,
  • C0β‹…F=1C_0 \cdot F = 1,
  • F2=0F^2 = 0.

Hence for D∼aC0+bFD \sim aC_0 + bF and Dβ€²βˆΌaβ€²C0+bβ€²FD' \sim a'C_0 + b'F:

Dβ‹…Dβ€²=βˆ’eaaβ€²+abβ€²+aβ€²b.D \cdot D' = -eaa' + ab' + a'b.

In particular, D2=βˆ’ea2+2abD^2 = -ea^2 + 2ab.

ExampleEffective divisors on F_n

On Fn\mathbb{F}_n (with nβ‰₯1n \geq 1), a divisor D∼aC0+bFD \sim aC_0 + bF is effective if and only if aβ‰₯0a \geq 0 and bβ‰₯nab \geq na (when a>0a > 0), or a=0a = 0 and bβ‰₯0b \geq 0. To see this:

  • If DD is effective and irreducible, either D=C0D = C_0, D=FD = F, or DD is a section/multisection disjoint from C0C_0.
  • Dβ‹…C0=a(βˆ’n)+b+0=bβˆ’naβ‰₯0D \cdot C_0 = a(-n) + b + 0 = b - na \geq 0 for effective DD not equal to C0C_0.
  • Dβ‹…F=aβ‰₯0D \cdot F = a \geq 0 for effective DD.

The nef cone of Fn\mathbb{F}_n is generated by C0+nFC_0 + nF and FF: a divisor aC0+bFaC_0 + bF is nef iff aβ‰₯0a \geq 0 and bβ‰₯nab \geq na.

ExampleAmple divisors on F_n

On Fn\mathbb{F}_n (with nβ‰₯1n \geq 1), the divisor D∼aC0+bFD \sim aC_0 + bF is ample if and only if a>0a > 0 and b>nab > na. Checking via Nakai--Moishezon:

  • D2=βˆ’na2+2ab=a(2bβˆ’na)>0D^2 = -na^2 + 2ab = a(2b - na) > 0 requires b>na/2b > na/2. But we also need:
  • Dβ‹…C0=bβˆ’na>0D \cdot C_0 = b - na > 0, i.e., b>nab > na.
  • Dβ‹…F=a>0D \cdot F = a > 0.

So ampleness requires a>0a > 0 and b>nab > na. For example, on F2\mathbb{F}_2:

  • C0+3FC_0 + 3F is ample (a=1,b=3>2a = 1, b = 3 > 2).
  • C0+2FC_0 + 2F is nef but not ample (Dβ‹…C0=0D \cdot C_0 = 0).
  • C0+FC_0 + F is not nef (Dβ‹…C0=βˆ’1<0D \cdot C_0 = -1 < 0).

Canonical class of ruled surfaces

TheoremCanonical class of a geometrically ruled surface

Let π ⁣:X=P(E)β†’C\pi \colon X = \mathbb{P}(\mathcal{E}) \to C be a geometrically ruled surface over a smooth curve CC of genus gg, with invariant ee. Then:

KXβˆΌβˆ’2C0+(deg⁑(det⁑E)+2gβˆ’2)F=βˆ’2C0+(2gβˆ’2βˆ’e)F.K_X \sim -2C_0 + (\deg(\det \mathcal{E}) + 2g - 2)F = -2C_0 + (2g - 2 - e)F.

In particular:

  • KXβ‹…C0=2(βˆ’e)+(2gβˆ’2βˆ’e)(βˆ’1)β‹…0+…K_X \cdot C_0 = 2(-e) + (2g - 2 - e)(-1) \cdot 0 + \ldots; more directly, KXβ‹…C0=eβˆ’2+2gK_X \cdot C_0 = e - 2 + 2g and KXβ‹…F=βˆ’2K_X \cdot F = -2.
  • KX2=8(1βˆ’g)K_X^2 = 8(1 - g).
  • The Kodaira dimension is ΞΊ(X)=βˆ’βˆž\kappa(X) = -\infty since ∣mKX∣=βˆ…|mK_X| = \emptyset for all mβ‰₯1m \geq 1.
ExampleCanonical class of Hirzebruch surfaces

For Fn\mathbb{F}_n (base P1\mathbb{P}^1, genus g=0g = 0, invariant e=ne = n):

KFnβˆΌβˆ’2C0+(nβˆ’2)F.K_{\mathbb{F}_n} \sim -2C_0 + (n - 2)F.

Checking self-intersection: K2=4(βˆ’n)+2β‹…(βˆ’2)(nβˆ’2)+0=βˆ’4nβˆ’4n+8=8βˆ’8nK^2 = 4(-n) + 2 \cdot (-2)(n - 2) + 0 = -4n - 4n + 8 = 8 - 8n... let us use the formula directly: K2=8(1βˆ’0)=8K^2 = 8(1 - 0) = 8. This is consistent for all Fn\mathbb{F}_n.

Concrete cases:

  • F0=P1Γ—P1\mathbb{F}_0 = \mathbb{P}^1 \times \mathbb{P}^1: KβˆΌβˆ’2C0βˆ’2FK \sim -2C_0 - 2F, i.e., KβˆΌβˆ’2F1βˆ’2F2K \sim -2F_1 - 2F_2 (bidegree (βˆ’2,βˆ’2)(-2, -2)). Then K2=8K^2 = 8.
  • F1=Bl⁑P(P2)\mathbb{F}_1 = \operatorname{Bl}_P(\mathbb{P}^2): KβˆΌβˆ’2C0βˆ’FK \sim -2C_0 - F. Using C0=Hβˆ’EC_0 = H - E and F=HF = H... or directly, K2=8K^2 = 8.
  • F2\mathbb{F}_2: KβˆΌβˆ’2C0K \sim -2C_0. The anticanonical class is βˆ’K∼2C0-K \sim 2C_0, which is effective but not ample.
ExampleAdjunction on fibers and sections

On a ruled surface π ⁣:Xβ†’C\pi \colon X \to C with canonical class KXK_X:

Fibers: For a fiber Fβ‰…P1F \cong \mathbb{P}^1 (with g=0g = 0), adjunction gives βˆ’2=F2+KXβ‹…F=0+(βˆ’2)=βˆ’2-2 = F^2 + K_X \cdot F = 0 + (-2) = -2. Consistent.

Sections: For the section C0C_0 on Fn\mathbb{F}_n, adjunction gives:

2g(C0)βˆ’2=C02+KXβ‹…C0=βˆ’n+(βˆ’2(βˆ’n)+(nβˆ’2))=βˆ’n+(nβˆ’2)=βˆ’2.2g(C_0) - 2 = C_0^2 + K_X \cdot C_0 = -n + (-2(-n) + (n-2)) = -n + (n - 2) = -2.

So g(C0)=0g(C_0) = 0, confirming C0β‰…P1C_0 \cong \mathbb{P}^1. More generally, a section Σ∼C0+mF\Sigma \sim C_0 + mF on Fn\mathbb{F}_n has genus:

2g(Ξ£)βˆ’2=Ξ£2+Kβ‹…Ξ£=(2mβˆ’n)+(βˆ’2(2mβˆ’n)+(nβˆ’2)β‹…1)2g(\Sigma) - 2 = \Sigma^2 + K \cdot \Sigma = (2m - n) + (-2(2m - n) + (n-2) \cdot 1)

which simplifies (using Kβ‹…Ξ£=βˆ’2(mβˆ’n+0)+(nβˆ’2)=βˆ’2m+nβˆ’2K \cdot \Sigma = -2(m - n + 0) + (n-2) = -2m + n - 2... let us compute directly): Kβ‹…Ξ£=(βˆ’2C0+(nβˆ’2)F)β‹…(C0+mF)=βˆ’2(βˆ’n)+(βˆ’2)(m)+(nβˆ’2)(1)+0=2nβˆ’2m+nβˆ’2=3nβˆ’2mβˆ’2K \cdot \Sigma = (-2C_0 + (n-2)F) \cdot (C_0 + mF) = -2(-n) + (-2)(m) + (n-2)(1) + 0 = 2n - 2m + n - 2 = 3n - 2m - 2.

So 2gβˆ’2=(2mβˆ’n)+(3nβˆ’2mβˆ’2)=2nβˆ’22g - 2 = (2m - n) + (3n - 2m - 2) = 2n - 2, giving g(Ξ£)=ng(\Sigma) = n... wait, that cannot be right for all sections. Let me redo: Ξ£=C0+mF\Sigma = C_0 + mF, so Ξ£2=βˆ’n+2m\Sigma^2 = -n + 2m and Kβ‹…Ξ£=(βˆ’2C0+(nβˆ’2)F)β‹…(C0+mF)=2nβˆ’2m+nβˆ’2=3nβˆ’2mβˆ’2K \cdot \Sigma = (-2C_0 + (n-2)F) \cdot (C_0 + mF) = 2n - 2m + n - 2 = 3n - 2m - 2... hmm, this gives g=ng = n independent of mm, which is wrong. The correct computation (being careful with the fiber): Kβ‹…Ξ£=βˆ’2(C0β‹…C0)βˆ’2m(C0β‹…F)+(nβˆ’2)(Fβ‹…C0)+m(nβˆ’2)(Fβ‹…F)K \cdot \Sigma = -2(C_0 \cdot C_0) - 2m(C_0 \cdot F) + (n-2)(F \cdot C_0) + m(n-2)(F \cdot F) =βˆ’2(βˆ’n)βˆ’2m+(nβˆ’2)+0=2nβˆ’2m+nβˆ’2= -2(-n) - 2m + (n-2) + 0 = 2n - 2m + n - 2. So indeed 2gβˆ’2=(2mβˆ’n)+(3nβˆ’2mβˆ’2)=2nβˆ’22g - 2 = (2m - n) + (3n - 2m - 2) = 2n - 2, giving g=ng = n. But C0C_0 itself has m=0m = 0 and g=0g = 0 from n=0n = 0 case... The issue is that Σ∼C0+mF\Sigma \sim C_0 + mF for mβ‰₯nm \geq n gives a smooth section only when Ξ£\Sigma actually is smooth. For F0\mathbb{F}_0, n=0n = 0, so g=0g = 0 for all sections -- correct since sections of P1Γ—P1β†’P1\mathbb{P}^1 \times \mathbb{P}^1 \to \mathbb{P}^1 are rational.


Minimal ruled surfaces and Nagata's theorem

TheoremNagata's theorem on minimal ruled surfaces

Let π ⁣:X=P(E)β†’C\pi \colon X = \mathbb{P}(\mathcal{E}) \to C be a geometrically ruled surface over a curve CC of genus gg, with normalized bundle E\mathcal{E} and invariant ee. Then:

  • eβ‰₯βˆ’ge \geq -g always.
  • If E\mathcal{E} is decomposable, i.e., Eβ‰…OCβŠ•L\mathcal{E} \cong \mathcal{O}_C \oplus \mathcal{L} for a line bundle L\mathcal{L} with deg⁑L=βˆ’e≀0\deg \mathcal{L} = -e \leq 0, then eβ‰₯0e \geq 0.
  • For every value eβ‰₯βˆ’ge \geq -g, there exists a ruled surface over CC with that invariant.
  • Over C=P1C = \mathbb{P}^1, every vector bundle splits (Grothendieck's theorem), so Eβ‰…OβŠ•O(βˆ’n)\mathcal{E} \cong \mathcal{O} \oplus \mathcal{O}(-n) with nβ‰₯0n \geq 0. Thus the Hirzebruch surfaces Fn\mathbb{F}_n (nβ‰₯0n \geq 0) are exactly the minimal ruled surfaces over P1\mathbb{P}^1.
RemarkMinimality of ruled surfaces

A geometrically ruled surface P(E)\mathbb{P}(\mathcal{E}) is minimal (contains no (βˆ’1)(-1)-curves in a fiber) if and only if no fiber contains an exceptional curve. Equivalently:

  • Over P1\mathbb{P}^1: Fn\mathbb{F}_n is minimal for nβ‰ 1n \neq 1. The surface F1\mathbb{F}_1 is not minimal because the section C0C_0 with C02=βˆ’1C_0^2 = -1 is an exceptional curve that can be contracted to yield P2\mathbb{P}^2.
  • Over a curve of genus gβ‰₯1g \geq 1: P(E)\mathbb{P}(\mathcal{E}) is always minimal (a ruled surface over a curve of positive genus has no (βˆ’1)(-1)-curves mapping to a point on the base).
ExampleBlowing down F₁ to β„™Β²

The surface F1\mathbb{F}_1 has a unique (βˆ’1)(-1)-curve, namely the section C0C_0 with C02=βˆ’1C_0^2 = -1. Contracting C0C_0 gives P2\mathbb{P}^2:

F1=Bl⁑P(P2)β†’contractΒ C0P2.\mathbb{F}_1 = \operatorname{Bl}_P(\mathbb{P}^2) \xrightarrow{\text{contract } C_0} \mathbb{P}^2.

After contraction, the fibers of the ruling become lines through PP in P2\mathbb{P}^2. Conversely, blowing up any point P∈P2P \in \mathbb{P}^2 gives F1\mathbb{F}_1 with the ruling by proper transforms of lines through PP.


Rational surfaces

Definition5.11Rational surface

A smooth projective surface XX is rational if it is birational to P2\mathbb{P}^2 (equivalently, birational to P1Γ—P1\mathbb{P}^1 \times \mathbb{P}^1, or equivalently, ruled over P1\mathbb{P}^1).

A rational surface satisfies:

  • q(X)=h1(OX)=0q(X) = h^1(\mathcal{O}_X) = 0 (irregularity zero),
  • pg(X)=h0(KX)=0p_g(X) = h^0(K_X) = 0 (geometric genus zero),
  • ΞΊ(X)=βˆ’βˆž\kappa(X) = -\infty (Kodaira dimension negative infinity).
TheoremCastelnuovo's rationality criterion

A smooth projective surface XX is rational if and only if q(X)=P2(X)=0q(X) = P_2(X) = 0, where q=h1(OX)q = h^1(\mathcal{O}_X) is the irregularity and P2=h0(2KX)P_2 = h^0(2K_X) is the second plurigenus.

This is a deep result: the vanishing of two numerical invariants suffices to guarantee birationality to P2\mathbb{P}^2.

TheoremClassification of minimal rational surfaces

Every minimal rational surface is isomorphic to exactly one of:

  • P2\mathbb{P}^2, or
  • Fn\mathbb{F}_n for n=0n = 0 or nβ‰₯2n \geq 2.

The surface F1\mathbb{F}_1 is excluded because it is not minimal (it is the blowup of P2\mathbb{P}^2 at one point). No two surfaces in this list are isomorphic to each other.

Every rational surface is obtained from one of these minimal models by a finite sequence of blowups.

ExampleHirzebruch surfaces are pairwise non-isomorphic

For n≠mn \neq m (with n,m≠1n, m \neq 1), Fn≇Fm\mathbb{F}_n \not\cong \mathbb{F}_m. One way to distinguish them:

  • If nβ‰₯2n \geq 2, then Fn\mathbb{F}_n contains a unique irreducible curve with negative self-intersection, namely C0C_0 with C02=βˆ’nC_0^2 = -n. Different values of nn give different self-intersection numbers.
  • F0=P1Γ—P1\mathbb{F}_0 = \mathbb{P}^1 \times \mathbb{P}^1 has no curves of negative self-intersection at all, distinguishing it from Fn\mathbb{F}_n for nβ‰₯2n \geq 2.
  • P2\mathbb{P}^2 is distinguished from all Fn\mathbb{F}_n by its Picard number: ρ(P2)=1\rho(\mathbb{P}^2) = 1 while ρ(Fn)=2\rho(\mathbb{F}_n) = 2.
ExampleBirational but not isomorphic rational surfaces

All Hirzebruch surfaces are birational to each other (all rational), but not isomorphic:

  • F0\mathbb{F}_0 and F2\mathbb{F}_2 are both minimal rational surfaces, birational but not isomorphic.
  • To go from F0\mathbb{F}_0 to F2\mathbb{F}_2: blow up a point P∈F0P \in \mathbb{F}_0 to get a surface with two (βˆ’1)(-1)-curves (the exceptional divisor and the proper transform of the fiber through PP), then contract the proper transform of the fiber to get F2\mathbb{F}_2.
  • This is an elementary transformation (or elm): F0β‡’F2\mathbb{F}_0 \dashrightarrow \mathbb{F}_2 is a birational map that is not an isomorphism.

More generally, an elementary transformation centered at a point on Fn\mathbb{F}_n gives either Fnβˆ’1\mathbb{F}_{n-1} or Fn+1\mathbb{F}_{n+1}, depending on whether the point lies on C0C_0 or not.


Scrolls in projective space

Definition5.12Scroll

A rational normal scroll S(a,b)S(a, b) with 1≀a≀b1 \leq a \leq b is the image of the map Fbβˆ’aβ†’Pa+b+1\mathbb{F}_{b-a} \to \mathbb{P}^{a+b+1} given by the linear system ∣C0+bF∣|C_0 + bF|, where C0C_0 is the section with C02=βˆ’(bβˆ’a)C_0^2 = -(b-a) and FF is the fiber class.

Equivalently, S(a,b)S(a, b) is the union of lines joining corresponding points on two rational normal curves of degrees aa and bb in complementary linear subspaces of Pa+b+1\mathbb{P}^{a+b+1}.

The scroll S(a,b)S(a, b) has degree a+ba + b in Pa+b+1\mathbb{P}^{a+b+1} and is a surface of minimal degree (degree =codim⁑+1= \operatorname{codim} + 1).

ExampleThe scroll S(1,1)

S(1,1)S(1,1) is the image of F0=P1Γ—P1\mathbb{F}_0 = \mathbb{P}^1 \times \mathbb{P}^1 under the Segre embedding P1Γ—P1β†ͺP3\mathbb{P}^1 \times \mathbb{P}^1 \hookrightarrow \mathbb{P}^3. This is the smooth quadric surface QβŠ‚P3Q \subset \mathbb{P}^3, a surface of degree 22 in P3\mathbb{P}^3 (minimal degree: 2=3βˆ’2+1=codim⁑+12 = 3 - 2 + 1 = \operatorname{codim} + 1).

ExampleThe scroll S(1,2)

S(1,2)S(1,2) is the image of F1\mathbb{F}_1 in P4\mathbb{P}^4 via the linear system ∣C0+2F∣|C_0 + 2F|. This is a cubic scroll: a surface of degree 33 in P4\mathbb{P}^4. It can be described as the union of lines joining a point and a conic in complementary subspaces P0βŠ‚P4\mathbb{P}^0 \subset \mathbb{P}^4 and P2βŠ‚P4\mathbb{P}^2 \subset \mathbb{P}^4... but actually a=1,b=2a = 1, b = 2, so it is the join of a line and a conic in disjoint subspaces P1\mathbb{P}^1 and P2\mathbb{P}^2 of P4\mathbb{P}^4.

It is a rational surface of minimal degree, ruled by lines.

ExampleThe scroll S(1,3) and the rational normal scroll of degree 4

S(1,3)S(1,3) is a surface of degree 44 in P5\mathbb{P}^5, obtained from F2\mathbb{F}_2 via ∣C0+3F∣|C_0 + 3F|. It is the join of a line and a twisted cubic in disjoint linear subspaces P1\mathbb{P}^1 and P3\mathbb{P}^3 of P5\mathbb{P}^5.

The "degenerate" case S(0,n)S(0, n) (where a=0a = 0) gives the cone over the rational normal curve of degree nn: this is a surface of degree nn in Pn+1\mathbb{P}^{n+1} with a singular vertex point.


Ruled surfaces over curves of higher genus

ExampleRuled surfaces over an elliptic curve

Let CC be an elliptic curve. Then:

  • Decomposable bundles: E=OCβŠ•L\mathcal{E} = \mathcal{O}_C \oplus \mathcal{L} with deg⁑L≀0\deg \mathcal{L} \leq 0 gives e=βˆ’deg⁑Lβ‰₯0e = -\deg \mathcal{L} \geq 0. For L=OC\mathcal{L} = \mathcal{O}_C, we get CΓ—P1C \times \mathbb{P}^1 with e=0e = 0.

  • Atiyah's bundle: There is a unique non-split extension 0β†’OCβ†’E0β†’OCβ†’00 \to \mathcal{O}_C \to \mathcal{E}_0 \to \mathcal{O}_C \to 0 (the Atiyah bundle). Then P(E0)\mathbb{P}(\mathcal{E}_0) is a ruled surface over CC with e=0e = 0 but not isomorphic to CΓ—P1C \times \mathbb{P}^1. This surface has an interesting property: it has no section Ξ£\Sigma with Ξ£2=0\Sigma^2 = 0 (unlike CΓ—P1C \times \mathbb{P}^1, which has many).

  • Indecomposable bundles with e=βˆ’1e = -1: Nagata's theorem allows e=βˆ’1e = -1 (since g=1g = 1), and these correspond to indecomposable rank-22 bundles of odd degree.

ExampleRuled surfaces over a genus-2 curve

Let CC be a curve of genus 22. Nagata's theorem gives eβ‰₯βˆ’2e \geq -2. So the possible invariants are e∈{βˆ’2,βˆ’1,0,1,2,…}e \in \{-2, -1, 0, 1, 2, \ldots\}.

  • e=βˆ’2e = -2 (the extreme case): These arise from stable rank-22 bundles. The moduli space of stable bundles of rank 22 and degree 11 on CC is isomorphic to P3\mathbb{P}^3 (Newstead's theorem), so there is a 33-parameter family of such ruled surfaces.
  • e=0e = 0, decomposable: E=OβŠ•O\mathcal{E} = \mathcal{O} \oplus \mathcal{O} gives CΓ—P1C \times \mathbb{P}^1.
  • e=0e = 0, indecomposable: Non-split extensions 0β†’Oβ†’Eβ†’Oβ†’00 \to \mathcal{O} \to \mathcal{E} \to \mathcal{O} \to 0 exist and give ruled surfaces not isomorphic to CΓ—P1C \times \mathbb{P}^1.

Numerical invariants

TheoremNumerical invariants of ruled surfaces

Let X=P(E)X = \mathbb{P}(\mathcal{E}) be a geometrically ruled surface over a curve CC of genus gg. Then:

  • q(X)=gq(X) = g (irregularity equals genus of the base),
  • pg(X)=0p_g(X) = 0 (geometric genus is zero),
  • Ο‡(OX)=1βˆ’g\chi(\mathcal{O}_X) = 1 - g,
  • KX2=8(1βˆ’g)K_X^2 = 8(1 - g),
  • e(X)=4(1βˆ’g)e(X) = 4(1 - g) (topological Euler characteristic, by Noether's formula: 12Ο‡(OX)=K2+e(X)12\chi(\mathcal{O}_X) = K^2 + e(X), so e(X)=12(1βˆ’g)βˆ’8(1βˆ’g)=4(1βˆ’g)e(X) = 12(1 - g) - 8(1 - g) = 4(1 - g)),
  • b1(X)=2gb_1(X) = 2g, b2(X)=2b_2(X) = 2 (Betti numbers, over C\mathbb{C}),
  • ΞΊ(X)=βˆ’βˆž\kappa(X) = -\infty.

In particular, all ruled surfaces over the same base curve CC share the same numerical invariants -- they differ only in their algebraic structure.

ExampleInvariants for small genus

Here are the invariants for ruled surfaces P(E)β†’C\mathbb{P}(\mathcal{E}) \to C:

For g=0g = 0 (base P1\mathbb{P}^1, i.e., Hirzebruch surfaces): q=0q = 0, pg=0p_g = 0, Ο‡(O)=1\chi(\mathcal{O}) = 1, K2=8K^2 = 8, e=4e = 4, b2=2b_2 = 2.

For g=1g = 1 (base an elliptic curve): q=1q = 1, pg=0p_g = 0, Ο‡(O)=0\chi(\mathcal{O}) = 0, K2=0K^2 = 0, e=0e = 0, b2=2b_2 = 2.

For g=2g = 2: q=2q = 2, pg=0p_g = 0, Ο‡(O)=βˆ’1\chi(\mathcal{O}) = -1, K2=βˆ’8K^2 = -8, e=βˆ’4e = -4, b2=2b_2 = 2.

For g=gg = g: q=gq = g, pg=0p_g = 0, Ο‡(O)=1βˆ’g\chi(\mathcal{O}) = 1 - g, K2=8(1βˆ’g)K^2 = 8(1-g), e=4(1βˆ’g)e = 4(1-g), b2=2b_2 = 2.

Note the curious feature: K2<0K^2 < 0 for gβ‰₯2g \geq 2. The canonical class is "very negative" on a ruled surface, reflecting the rational fibers.


Elementary transformations

Definition5.13Elementary transformation

Let π ⁣:Xβ†’C\pi \colon X \to C be a geometrically ruled surface, and let P∈XP \in X be a point lying on the fiber F0=Ο€βˆ’1(Ο€(P))F_0 = \pi^{-1}(\pi(P)). The elementary transformation (or elm) at PP is the birational map:

elm⁑P ⁣:Xβ‡’Xβ€²\operatorname{elm}_P \colon X \dashrightarrow X'

defined by: blow up PP to get X~\tilde{X} with exceptional divisor EE, then contract the proper transform F~0\tilde{F}_0 of the fiber F0F_0 (which is now a (βˆ’1)(-1)-curve on X~\tilde{X}) to get a new ruled surface Xβ€²X' over CC.

If X=FnX = \mathbb{F}_n over P1\mathbb{P}^1 and P∈C0P \in C_0 (the negative section), then Xβ€²β‰…Fnβˆ’1X' \cong \mathbb{F}_{n-1}. If Pβˆ‰C0P \notin C_0, then Xβ€²β‰…Fn+1X' \cong \mathbb{F}_{n+1}.

ExampleElementary transformation from Fβ‚€ to Fβ‚‚

Start with F0=P1Γ—P1\mathbb{F}_0 = \mathbb{P}^1 \times \mathbb{P}^1. Pick any point P∈F0P \in \mathbb{F}_0.

  1. Blow up PP: get X~\tilde{X} with exceptional curve EE (E2=βˆ’1E^2 = -1), and the proper transform F~\tilde{F} of the fiber through PP has F~2=βˆ’1\tilde{F}^2 = -1.
  2. Contract F~\tilde{F}: the image surface is F2\mathbb{F}_2.

The resulting birational map F0β‡’F2\mathbb{F}_0 \dashrightarrow \mathbb{F}_2 is not regular (it is undefined at PP and contracts a curve). This shows that F0\mathbb{F}_0 and F2\mathbb{F}_2 are birational but not isomorphic.

Repeating: F0β‡’F2β‡’F4β‡’β‹―\mathbb{F}_0 \dashrightarrow \mathbb{F}_2 \dashrightarrow \mathbb{F}_4 \dashrightarrow \cdots by choosing points off the negative section, or F2β‡’F1β‡’F0\mathbb{F}_2 \dashrightarrow \mathbb{F}_1 \dashrightarrow \mathbb{F}_0 by choosing points on C0C_0.


The Tsen--Castelnuovo theorem

TheoremEvery ruled surface over β„™ΒΉ is rational

If XX is ruled over P1\mathbb{P}^1, then XX is rational. More precisely, every P1\mathbb{P}^1-bundle over P1\mathbb{P}^1 is isomorphic to some Fn\mathbb{F}_n (by Grothendieck's theorem on vector bundles over P1\mathbb{P}^1), and every Fn\mathbb{F}_n is rational.

Conversely, if XX is a ruled surface over a curve CC of genus gβ‰₯1g \geq 1, then XX is irrational: it is not birational to P2\mathbb{P}^2, since q(X)=g>0q(X) = g > 0 while q(P2)=0q(\mathbb{P}^2) = 0, and the irregularity is a birational invariant.


Ruled surfaces in the Enriques--Kodaira classification

RemarkRole in the classification of surfaces

In the Enriques--Kodaira classification of algebraic surfaces by Kodaira dimension ΞΊ\kappa:

  • ΞΊ=βˆ’βˆž\kappa = -\infty: All such surfaces are ruled. More precisely, a surface has ΞΊ=βˆ’βˆž\kappa = -\infty if and only if it is birational to CΓ—P1C \times \mathbb{P}^1 for some curve CC (this is a deep theorem, relying on the abundance conjecture in dimension 2, proved by the Italian school and Kodaira).
  • ΞΊ=0\kappa = 0: K3 surfaces, abelian surfaces, Enriques surfaces, hyperelliptic surfaces.
  • ΞΊ=1\kappa = 1: Properly elliptic surfaces.
  • ΞΊ=2\kappa = 2: Surfaces of general type.

Thus ruled surfaces form the entire class ΞΊ=βˆ’βˆž\kappa = -\infty, and their complete classification reduces to understanding rank-22 vector bundles on curves.

ExampleDetecting ruled surfaces via invariants

To check if a surface XX is ruled, one can use the following criteria:

  • Sufficient: If P12(X)=h0(12KX)=0P_{12}(X) = h^0(12K_X) = 0, then XX is ruled (Enriques).
  • Necessary and sufficient: ΞΊ(X)=βˆ’βˆž\kappa(X) = -\infty, equivalently Pn(X)=0P_n(X) = 0 for all nβ‰₯1n \geq 1.
  • Over the base curve CC: If XX is ruled over CC, then q(X)=g(C)q(X) = g(C). So the Albanese variety Alb⁑(X)β‰…Jac⁑(C)\operatorname{Alb}(X) \cong \operatorname{Jac}(C).

For example, a surface with q=0,pg=0,K2=8q = 0, p_g = 0, K^2 = 8 is either P2\mathbb{P}^2 (if ρ=1\rho = 1), F0\mathbb{F}_0 or Fn\mathbb{F}_n (if ρ=2\rho = 2 and minimal), or a non-minimal rational surface.


Summary

RemarkKey facts about ruled surfaces

The theory of ruled surfaces rests on the following pillars:

  • Structure: Every geometrically ruled surface is P(E)\mathbb{P}(\mathcal{E}) for a rank-22 bundle E\mathcal{E} on a curve CC, classified (up to twist) by the invariant ee and the isomorphism class of E\mathcal{E}.
  • Picard group: NS⁑(X)β‰…ZC0βŠ•ZF\operatorname{NS}(X) \cong \mathbb{Z} C_0 \oplus \mathbb{Z} F with C02=βˆ’eC_0^2 = -e, C0β‹…F=1C_0 \cdot F = 1, F2=0F^2 = 0.
  • Canonical class: KXβˆΌβˆ’2C0+(2gβˆ’2βˆ’e)FK_X \sim -2C_0 + (2g - 2 - e)F, giving ΞΊ=βˆ’βˆž\kappa = -\infty.
  • Minimality over P1\mathbb{P}^1: The minimal models are P2\mathbb{P}^2 and Fn\mathbb{F}_n (nβ‰ 1n \neq 1).
  • Birational geometry: Elementary transformations connect all ruled surfaces over the same base curve.
  • Scrolls: Geometrically ruled surfaces embed as scrolls -- surfaces of minimal degree in projective space.