ConceptComplete

Blowing Up

Blowing up replaces a point on a surface by an entire projective line, "separating" tangent directions at that point. It is the most fundamental operation in the birational geometry of surfaces: every birational morphism between smooth surfaces factors as a sequence of blowups, and every surface singularity can be resolved by iterated blowups.


Definition of the blowup

Definition5.4Blowup of a surface at a point

Let XX be a smooth algebraic surface and P∈XP \in X a closed point. Choose local coordinates (x,y)(x, y) at PP. The blowup of XX at PP is the surface

X~=Bl⁑P(X)\widetilde{X} = \operatorname{Bl}_P(X)

together with a proper birational morphism Ο€:X~β†’X\pi : \widetilde{X} \to X such that:

  • Ο€\pi is an isomorphism over Xβˆ–{P}X \setminus \{P\},
  • the fiber E=Ο€βˆ’1(P)β‰…P1E = \pi^{-1}(P) \cong \mathbb{P}^1, called the exceptional divisor,
  • X~\widetilde{X} is smooth if XX is smooth.

Concretely, in local coordinates the blowup is the closure of the graph of the rational map (x,y)↦[x:y]∈P1(x,y) \mapsto [x:y] \in \mathbb{P}^1:

Bl⁑P(X)={(q,[s:t])∈XΓ—P1∣x(q)β‹…t=y(q)β‹…s}β€ΎβŠ‚XΓ—P1.\operatorname{Bl}_P(X) = \overline{\{(q, [s:t]) \in X \times \mathbb{P}^1 \mid x(q) \cdot t = y(q) \cdot s\}} \subset X \times \mathbb{P}^1.

Points of EE correspond to tangent directions at PP: each point [s:t]∈E[s:t] \in E remembers "the direction of approach" to PP.

ExampleBlowup of AΒ² at the origin

Let X=A2X = \mathbb{A}^2 with coordinates (x,y)(x, y) and P=(0,0)P = (0, 0). The blowup is:

Bl⁑0(A2)={((x,y),[s:t])∈A2Γ—P1∣xt=ys}.\operatorname{Bl}_0(\mathbb{A}^2) = \{((x, y), [s:t]) \in \mathbb{A}^2 \times \mathbb{P}^1 \mid xt = ys\}.

In the chart sβ‰ 0s \neq 0 (set s=1s = 1, so t=y/xt = y/x): the blowup is A2\mathbb{A}^2 with coordinates (x,t)(x, t) where y=xty = xt. The map Ο€\pi sends (x,t)↦(x,xt)(x, t) \mapsto (x, xt). The exceptional divisor in this chart is x=0x = 0 (a copy of A1\mathbb{A}^1 with coordinate tt).

In the chart t≠0t \neq 0 (set t=1t = 1, so s=x/ys = x/y): coordinates (s,y)(s, y) with x=syx = sy. The exceptional divisor is y=0y = 0.

The two charts glue to give Eβ‰…P1E \cong \mathbb{P}^1, and Bl⁑0(A2)\operatorname{Bl}_0(\mathbb{A}^2) is again a smooth surface.


The exceptional divisor

Definition5.5Exceptional divisor

The exceptional divisor of Ο€:Bl⁑P(X)β†’X\pi: \operatorname{Bl}_P(X) \to X is the fiber E=Ο€βˆ’1(P)E = \pi^{-1}(P). It has the following fundamental properties:

  • Eβ‰…P1E \cong \mathbb{P}^1 (as a variety),
  • E2=βˆ’1E^2 = -1 (self-intersection on X~\widetilde{X}),
  • OE(E)β‰…OP1(βˆ’1)\mathcal{O}_E(E) \cong \mathcal{O}_{\mathbb{P}^1}(-1) (the tautological bundle).

A smooth rational curve Cβ‰…P1C \cong \mathbb{P}^1 with C2=βˆ’1C^2 = -1 on a smooth surface is called a (βˆ’1)(-1)-curve or exceptional curve of the first kind.

ExampleComputing EΒ² = -1

On Bl⁑0(A2)\operatorname{Bl}_0(\mathbb{A}^2), in the chart (x,t)(x, t) with y=xty = xt, the exceptional divisor is E=V(x)E = V(x). To compute E2E^2, we use the adjunction formula. Since Eβ‰…P1E \cong \mathbb{P}^1 has genus 00:

2g(E)βˆ’2=E2+KX~β‹…E.2g(E) - 2 = E^2 + K_{\widetilde{X}} \cdot E.

We will show KX~β‹…E=βˆ’1K_{\widetilde{X}} \cdot E = -1 (from KX~=Ο€βˆ—KX+EK_{\widetilde{X}} = \pi^*K_X + E), giving βˆ’2=E2+(βˆ’1)-2 = E^2 + (-1), hence E2=βˆ’1E^2 = -1.

Alternatively, compute directly: OE(E)=NE/X~\mathcal{O}_E(E) = N_{E/\widetilde{X}}, the normal bundle. The normal bundle of Eβ‰…P1E \cong \mathbb{P}^1 embedded via the tautological sub-bundle in the trivial rank-22 bundle is OP1(βˆ’1)\mathcal{O}_{\mathbb{P}^1}(-1). Therefore E2=deg⁑OP1(βˆ’1)=βˆ’1E^2 = \deg \mathcal{O}_{\mathbb{P}^1}(-1) = -1.


Universal property

TheoremUniversal property of blowup

The blowup Ο€:Bl⁑P(X)β†’X\pi: \operatorname{Bl}_P(X) \to X is the universal morphism that turns the ideal sheaf mPβŠ‚OX\mathfrak{m}_P \subset \mathcal{O}_X of the point PP into an invertible sheaf. That is, mPβ‹…OX~=OX~(βˆ’E)\mathfrak{m}_P \cdot \mathcal{O}_{\widetilde{X}} = \mathcal{O}_{\widetilde{X}}(-E) is an invertible ideal, and any morphism f:Yβ†’Xf: Y \to X such that mPβ‹…OY\mathfrak{m}_P \cdot \mathcal{O}_Y is invertible factors uniquely through Ο€\pi.

RemarkBlowup as Proj

The blowup can also be constructed as:

Bl⁑P(X)=Proj⁑(⨁nβ‰₯0mPn),\operatorname{Bl}_P(X) = \operatorname{Proj}\left(\bigoplus_{n \geq 0} \mathfrak{m}_P^n\right),

the Proj of the Rees algebra of mP\mathfrak{m}_P. This construction generalizes to blowing up any ideal sheaf, not just the maximal ideal of a point.


Strict transform vs. total transform

Definition5.6Total and strict transforms

Let Ο€:X~β†’X\pi: \widetilde{X} \to X be the blowup of XX at PP, and let CβŠ‚XC \subset X be a curve.

  • The total transform is Ο€βˆ—C\pi^*C, the divisorial pullback: Ο€βˆ—C=C~+mβ‹…E\pi^*C = \widetilde{C} + m \cdot E where m=mult⁑P(C)m = \operatorname{mult}_P(C) is the multiplicity of CC at PP.

  • The strict transform (or proper transform) C~\widetilde{C} is the closure of Ο€βˆ’1(Cβˆ–{P})\pi^{-1}(C \setminus \{P\}) in X~\widetilde{X}.

Thus Ο€βˆ—C=C~+mult⁑P(C)β‹…E\pi^*C = \widetilde{C} + \operatorname{mult}_P(C) \cdot E.

ExampleStrict transform of a line

Consider C=V(y)βŠ‚A2C = V(y) \subset \mathbb{A}^2 (the xx-axis) and the blowup Ο€:Bl⁑0(A2)β†’A2\pi: \operatorname{Bl}_0(\mathbb{A}^2) \to \mathbb{A}^2. In the chart (x,t)(x, t) with y=xty = xt:

Ο€βˆ—C:y=0β€…β€ŠβŸΊβ€…β€Šxt=0β€…β€ŠβŸΊβ€…β€Š(x=0)βˆͺ(t=0).\pi^*C : y = 0 \iff xt = 0 \iff (x = 0) \cup (t = 0).

So Ο€βˆ—C=E+C~\pi^*C = E + \widetilde{C} where E=V(x)E = V(x) and C~=V(t)\widetilde{C} = V(t). Here mult⁑0(C)=1\operatorname{mult}_0(C) = 1, consistent with Ο€βˆ—C=C~+1β‹…E\pi^*C = \widetilde{C} + 1 \cdot E.

The strict transform C~=V(t)\widetilde{C} = V(t) meets E=V(x)E = V(x) at the single point (x,t)=(0,0)(x, t) = (0, 0), which corresponds to the tangent direction [1:0][1:0] of the line y=0y = 0 at the origin.

ExampleStrict transform of a nodal curve

Let C=V(y2βˆ’x2βˆ’x3)C = V(y^2 - x^2 - x^3), a node at the origin with two branches tangent to y=Β±xy = \pm x. In the chart (x,t)(x, t) with y=xty = xt:

Ο€βˆ—C:(xt)2βˆ’x2βˆ’x3=0β€…β€ŠβŸΊβ€…β€Šx2(t2βˆ’1βˆ’x)=0.\pi^*C : (xt)^2 - x^2 - x^3 = 0 \iff x^2(t^2 - 1 - x) = 0.

So Ο€βˆ—C=2E+C~\pi^*C = 2E + \widetilde{C} where C~=V(t2βˆ’1βˆ’x)\widetilde{C} = V(t^2 - 1 - x). Here mult⁑0(C)=2\operatorname{mult}_0(C) = 2.

The strict transform C~\widetilde{C} meets EE at x=0,t2=1x = 0, t^2 = 1, i.e., at t=Β±1t = \pm 1. These correspond to the two tangent directions y=Β±xy = \pm x at the node. The strict transform C~\widetilde{C} is smooth: the node has been resolved by one blowup.

ExampleStrict transform of a cuspidal curve

Let C=V(y2βˆ’x3)C = V(y^2 - x^3), a cusp at the origin. In the chart (x,t)(x, t) with y=xty = xt:

Ο€βˆ—C:x2t2βˆ’x3=0β€…β€ŠβŸΊβ€…β€Šx2(t2βˆ’x)=0.\pi^*C: x^2 t^2 - x^3 = 0 \iff x^2(t^2 - x) = 0.

So Ο€βˆ—C=2E+C~\pi^*C = 2E + \widetilde{C} with C~=V(t2βˆ’x)\widetilde{C} = V(t^2 - x). The strict transform meets EE at (0,0)(0, 0) only (a single point with t=0t = 0). Moreover, C~\widetilde{C} is a smooth parabola: the cusp has been resolved by a single blowup.


Intersection numbers under blowup

TheoremIntersection calculus on the blowup

Let Ο€:X~=Bl⁑P(X)β†’X\pi: \widetilde{X} = \operatorname{Bl}_P(X) \to X be the blowup of a smooth surface at PP, with exceptional divisor EE. Then:

  • Pic⁑(X~)=Ο€βˆ—Pic⁑(X)βŠ•Zβ‹…E\operatorname{Pic}(\widetilde{X}) = \pi^*\operatorname{Pic}(X) \oplus \mathbb{Z} \cdot E.
  • Ο€βˆ—Dβ‹…E=0\pi^*D \cdot E = 0 for all D∈Pic⁑(X)D \in \operatorname{Pic}(X).
  • E2=βˆ’1E^2 = -1.
  • Ο€βˆ—Cβ‹…Ο€βˆ—D=Cβ‹…D\pi^*C \cdot \pi^*D = C \cdot D for all C,D∈Pic⁑(X)C, D \in \operatorname{Pic}(X).

For the strict transform C~=Ο€βˆ—Cβˆ’mult⁑P(C)β‹…E\widetilde{C} = \pi^*C - \operatorname{mult}_P(C) \cdot E:

C~2=C2βˆ’mult⁑P(C)2.\widetilde{C}^2 = C^2 - \operatorname{mult}_P(C)^2.

C~β‹…D~=Cβ‹…Dβˆ’mult⁑P(C)β‹…mult⁑P(D).\widetilde{C} \cdot \widetilde{D} = C \cdot D - \operatorname{mult}_P(C) \cdot \operatorname{mult}_P(D).

ExampleIntersection numbers on Bl_P(β„™Β²)

Let X~=Bl⁑P(P2)\widetilde{X} = \operatorname{Bl}_P(\mathbb{P}^2), H=Ο€βˆ—β„“H = \pi^*\ell (pullback of a line), EE the exceptional divisor.

Basis: Pic⁑(X~)=ZHβŠ•ZE\operatorname{Pic}(\widetilde{X}) = \mathbb{Z} H \oplus \mathbb{Z} E with H2=1H^2 = 1, E2=βˆ’1E^2 = -1, Hβ‹…E=0H \cdot E = 0.

Strict transform of a line through PP: β„“~=Hβˆ’E\widetilde{\ell} = H - E (since mult⁑P(β„“)=1\operatorname{mult}_P(\ell) = 1). Then β„“~2=1βˆ’1=0\widetilde{\ell}^2 = 1 - 1 = 0. This makes sense: two lines through PP become disjoint after blowup (their intersection point has been separated).

Strict transform of a conic through PP: C~=2Hβˆ’E\widetilde{C} = 2H - E. Then C~2=4βˆ’1=3\widetilde{C}^2 = 4 - 1 = 3.

Strict transform of a conic tangent to a line at PP: the conic has mult⁑P=1\operatorname{mult}_P = 1, so C~=2Hβˆ’E\widetilde{C} = 2H - E. The line has β„“~=Hβˆ’E\widetilde{\ell} = H - E. Then C~β‹…β„“~=2βˆ’1=1\widetilde{C} \cdot \widetilde{\ell} = 2 - 1 = 1. The tangency has been "separated" by the blowup.


Canonical class of a blowup

TheoremCanonical class formula

Let Ο€:X~=Bl⁑P(X)β†’X\pi: \widetilde{X} = \operatorname{Bl}_P(X) \to X. Then the canonical class of the blowup is:

KX~=Ο€βˆ—KX+E.K_{\widetilde{X}} = \pi^* K_X + E.

More generally, for the blowup at nn distinct points P1,…,PnP_1, \ldots, P_n:

KX~=Ο€βˆ—KX+E1+E2+β‹―+En.K_{\widetilde{X}} = \pi^* K_X + E_1 + E_2 + \cdots + E_n.

ExampleCanonical class of Bl_P(β„™Β²)

For X=P2X = \mathbb{P}^2, KX=βˆ’3HK_X = -3H. So:

KBl⁑P(P2)=βˆ’3H+E.K_{\operatorname{Bl}_P(\mathbb{P}^2)} = -3H + E.

Check with adjunction on EE: Kβ‹…E=(βˆ’3H+E)β‹…E=0+(βˆ’1)=βˆ’1K \cdot E = (-3H + E) \cdot E = 0 + (-1) = -1. Then 2g(E)βˆ’2=E2+Kβ‹…E=βˆ’1+(βˆ’1)=βˆ’22g(E) - 2 = E^2 + K \cdot E = -1 + (-1) = -2, giving g(E)=0g(E) = 0.

Check K2K^2: (βˆ’3H+E)2=9βˆ’1=8(-3H + E)^2 = 9 - 1 = 8. Since KP22=9K_{\mathbb{P}^2}^2 = 9, blowing up one point decreases K2K^2 by 11.

For the blowup of P2\mathbb{P}^2 at nn points: K2=9βˆ’nK^2 = 9 - n.


Resolution of singularities

RemarkResolution by blowups

Every surface singularity over a field of characteristic 00 (and many in positive characteristic) can be resolved by a finite sequence of blowups. The key examples:

  • Node (A1A_1: y2=x2+x3y^2 = x^2 + x^3 or xy=0xy = 0): resolved by one blowup, exceptional divisor is a (βˆ’2)(-2)-curve after normalizing.
  • Cusp (y2=x3y^2 = x^3): resolved by one blowup (strict transform becomes smooth).
  • AnA_n singularity (y2=x2+xn+2y^2 = x^2 + x^{n+2} or x2+y2+zn+1=0x^2 + y^2 + z^{n+1} = 0): resolved by nn successive blowups, producing a chain of nn rational (βˆ’2)(-2)-curves.
  • DnD_n singularity (x2+y2z+znβˆ’1=0x^2 + y^2 z + z^{n-1} = 0): resolved by blowups, dual graph is a DnD_n Dynkin diagram.
  • E6,E7,E8E_6, E_7, E_8 singularities: dual graph of the exceptional divisor is the corresponding Dynkin diagram.
ExampleResolving a node

Let S=V(xyβˆ’z2)βŠ‚A3S = V(xy - z^2) \subset \mathbb{A}^3, which has an A1A_1 (node) singularity at the origin. Blow up the origin in A3\mathbb{A}^3 and take the strict transform.

In the chart [1:u:v][1:u:v] (i.e., y=ux,z=vxy = ux, z = vx): the equation becomes x2(uβˆ’v2)=0x^2(u - v^2) = 0. The strict transform S~=V(uβˆ’v2)\widetilde{S} = V(u - v^2) is smooth, isomorphic to A2\mathbb{A}^2 with coordinates (x,v)(x, v).

The exceptional curve on S~\widetilde{S} is E=S~∩V(x)=V(x,uβˆ’v2)E = \widetilde{S} \cap V(x) = V(x, u - v^2), parametrized by vv. So Eβ‰…P1E \cong \mathbb{P}^1. One computes E2=βˆ’2E^2 = -2 on S~\widetilde{S}: this is the minimal resolution of the A1A_1 singularity.

ExampleResolving A_n singularities

The AnA_n singularity is (locally) x2+y2+zn+1=0x^2 + y^2 + z^{n+1} = 0 (over C\mathbb{C}, equivalently uv=zn+1uv = z^{n+1}). The minimal resolution has exceptional divisor consisting of a chain of nn smooth rational curves E1,E2,…,EnE_1, E_2, \ldots, E_n, each with Ei2=βˆ’2E_i^2 = -2, and:

  • Eiβ‹…Ei+1=1E_i \cdot E_{i+1} = 1 for i=1,…,nβˆ’1i = 1, \ldots, n - 1,
  • Eiβ‹…Ej=0E_i \cdot E_j = 0 for ∣iβˆ’j∣β‰₯2|i - j| \geq 2.

The intersection matrix is minus the Cartan matrix of the AnA_n Dynkin diagram. For A2A_2 (uv=z3uv = z^3): two (βˆ’2)(-2)-curves meeting in one point. For A3A_3: three (βˆ’2)(-2)-curves in a chain.


Blowup of β„™Β² at one point: the Hirzebruch surface F₁

ExampleBl_P(β„™Β²) = F₁

The blowup Bl⁑P(P2)\operatorname{Bl}_P(\mathbb{P}^2) is isomorphic to the Hirzebruch surface F1=P(OP1βŠ•OP1(1))\mathbb{F}_1 = \mathbb{P}(\mathcal{O}_{\mathbb{P}^1} \oplus \mathcal{O}_{\mathbb{P}^1}(1)).

To see this, the linear system ∣Hβˆ’E∣|H - E| (strict transforms of lines through PP) defines a morphism Bl⁑P(P2)β†’P1\operatorname{Bl}_P(\mathbb{P}^2) \to \mathbb{P}^1. The fibers are strict transforms of lines through PP, each isomorphic to P1\mathbb{P}^1 with self-intersection 00. This exhibits Bl⁑P(P2)\operatorname{Bl}_P(\mathbb{P}^2) as a ruled surface over P1\mathbb{P}^1.

The exceptional divisor EE is a section with E2=βˆ’1E^2 = -1, and the fibers F=Hβˆ’EF = H - E satisfy F2=0F^2 = 0, Fβ‹…E=1F \cdot E = 1. This matches F1\mathbb{F}_1: the unique Hirzebruch surface with a section of self-intersection βˆ’1-1.

Numerical invariants:

  • Pic⁑=ZHβŠ•ZE\operatorname{Pic} = \mathbb{Z} H \oplus \mathbb{Z} E, with H2=1,E2=βˆ’1,Hβ‹…E=0H^2 = 1, E^2 = -1, H \cdot E = 0.
  • K=βˆ’3H+EK = -3H + E, K2=8K^2 = 8.
  • Ο‡(O)=1\chi(\mathcal{O}) = 1, pg=q=0p_g = q = 0.

Blowup of β„™Β² at multiple points

ExampleBl_n(β„™Β²): blowup at n points

Let X~=Bl⁑P1,…,Pn(P2)\widetilde{X} = \operatorname{Bl}_{P_1, \ldots, P_n}(\mathbb{P}^2) be the blowup of P2\mathbb{P}^2 at nn points in general position.

Picard group: Pic⁑(X~)=ZHβŠ•ZE1βŠ•β‹―βŠ•ZEn\operatorname{Pic}(\widetilde{X}) = \mathbb{Z} H \oplus \mathbb{Z} E_1 \oplus \cdots \oplus \mathbb{Z} E_n with:

  • H2=1H^2 = 1, Ei2=βˆ’1E_i^2 = -1, Hβ‹…Ei=0H \cdot E_i = 0, Eiβ‹…Ej=0E_i \cdot E_j = 0 for iβ‰ ji \neq j.

Canonical class: K=βˆ’3H+E1+β‹―+EnK = -3H + E_1 + \cdots + E_n, K2=9βˆ’nK^2 = 9 - n.

Anticanonical class: βˆ’K=3Hβˆ’E1βˆ’β‹―βˆ’En-K = 3H - E_1 - \cdots - E_n, (βˆ’K)2=9βˆ’n(-K)^2 = 9 - n.

For the anticanonical class to be effective, we need n≀9n \leq 9 (curves of degree 33 through nn points exist iff n≀9n \leq 9).

Key classes of divisors:

  • Lines through PiP_i: class Hβˆ’EiH - E_i, self-intersection 00.
  • Lines through Pi,PjP_i, P_j: class Hβˆ’Eiβˆ’EjH - E_i - E_j, self-intersection βˆ’1-1 (a (βˆ’1)(-1)-curve!).
  • Conics through P1,…,P5P_1, \ldots, P_5: class 2Hβˆ’E1βˆ’β‹―βˆ’E52H - E_1 - \cdots - E_5, self-intersection βˆ’1-1.
  • Cubics through P1,…,P6P_1, \ldots, P_6 with a double point at P1P_1: class 3Hβˆ’2E1βˆ’E2βˆ’β‹―βˆ’E63H - 2E_1 - E_2 - \cdots - E_6, self-intersection βˆ’1-1.

Castelnuovo's contraction theorem

TheoremCastelnuovo's contraction theorem

Let XX be a smooth projective surface and EβŠ‚XE \subset X a (βˆ’1)(-1)-curve (i.e., Eβ‰…P1E \cong \mathbb{P}^1 and E2=βˆ’1E^2 = -1). Then there exists a smooth projective surface YY and a point P∈YP \in Y such that Xβ‰…Bl⁑P(Y)X \cong \operatorname{Bl}_P(Y) with EE as the exceptional divisor.

In other words: every (βˆ’1)(-1)-curve can be blown down (contracted) to a smooth point.

ExampleContracting the (-1)-curve on F₁

On F1=Bl⁑P(P2)\mathbb{F}_1 = \operatorname{Bl}_P(\mathbb{P}^2), the exceptional curve EE with E2=βˆ’1E^2 = -1 can be blown down to recover P2\mathbb{P}^2. But F1\mathbb{F}_1 has another family of (βˆ’1)(-1)-curves: the strict transforms of lines through PP, which have class Hβˆ’EH - E and self-intersection 00... wait, (Hβˆ’E)2=0(H-E)^2 = 0, not βˆ’1-1. Those are fibers, not (βˆ’1)(-1)-curves.

Actually, on F1\mathbb{F}_1 the only (βˆ’1)(-1)-curve is EE itself. Blowing down EE recovers P2\mathbb{P}^2. On F1\mathbb{F}_1 there is no other contraction to a smooth surface.

RemarkMinimal surfaces

A smooth projective surface is minimal if it contains no (βˆ’1)(-1)-curves. By Castelnuovo's theorem, any non-minimal surface can be blown down. Iterating:

Every smooth projective surface XX can be obtained from a minimal surface Xmin⁑X_{\min} by a finite sequence of blowups. The minimal model Xmin⁑X_{\min} is unique in the birational equivalence class (for surfaces not birational to P2\mathbb{P}^2 or ruled surfaces).

For rational surfaces: the minimal models are P2\mathbb{P}^2 and the Hirzebruch surfaces Fn\mathbb{F}_n (n≠1n \neq 1).


Del Pezzo surfaces

Definition5.7Del Pezzo surface

A del Pezzo surface is a smooth projective surface SS with ample anticanonical class βˆ’KS-K_S. The degree of SS is d=KS2d = K_S^2.

Every del Pezzo surface of degree d≀7d \leq 7 is isomorphic to the blowup of P2\mathbb{P}^2 at n=9βˆ’dn = 9 - d points in general position (no 33 collinear, no 66 on a conic, no 88 on a singular cubic with one being the singular point).

  • d=9d = 9 (n=0n = 0): P2\mathbb{P}^2 itself.
  • d=8d = 8 (n=1n = 1): Bl⁑P(P2)β‰…F1\operatorname{Bl}_P(\mathbb{P}^2) \cong \mathbb{F}_1, or P1Γ—P1\mathbb{P}^1 \times \mathbb{P}^1.
  • d=7d = 7 (n=2n = 2): Bl⁑P1,P2(P2)\operatorname{Bl}_{P_1, P_2}(\mathbb{P}^2).
  • d=6d = 6 (n=3n = 3): Bl⁑P1,P2,P3(P2)\operatorname{Bl}_{P_1, P_2, P_3}(\mathbb{P}^2), a cubic in P2Γ—P2\mathbb{P}^2 \times \mathbb{P}^2 (a toric variety).
  • d=5d = 5 (n=4n = 4): Bl⁑P1,…,P4(P2)\operatorname{Bl}_{P_1, \ldots, P_4}(\mathbb{P}^2), the intersection of 55 hyperplanes in Gr⁑(2,5)\operatorname{Gr}(2, 5).
  • d=4d = 4 (n=5n = 5): Bl⁑P1,…,P5(P2)\operatorname{Bl}_{P_1, \ldots, P_5}(\mathbb{P}^2), a complete intersection of two quadrics in P4\mathbb{P}^4.
  • d=3d = 3 (n=6n = 6): Bl⁑P1,…,P6(P2)\operatorname{Bl}_{P_1, \ldots, P_6}(\mathbb{P}^2), a smooth cubic surface in P3\mathbb{P}^3.
  • d=2d = 2 (n=7n = 7): Bl⁑P1,…,P7(P2)\operatorname{Bl}_{P_1, \ldots, P_7}(\mathbb{P}^2), a double cover of P2\mathbb{P}^2 branched over a quartic.
  • d=1d = 1 (n=8n = 8): Bl⁑P1,…,P8(P2)\operatorname{Bl}_{P_1, \ldots, P_8}(\mathbb{P}^2), a double cover of a quadric cone branched over a sextic.
ExampleAmpleness of -K on del Pezzo surfaces

On S=Bl⁑P1,…,Pn(P2)S = \operatorname{Bl}_{P_1, \ldots, P_n}(\mathbb{P}^2) with n≀8n \leq 8: βˆ’K=3Hβˆ’E1βˆ’β‹―βˆ’En-K = 3H - E_1 - \cdots - E_n.

Check ampleness via Nakai--Moishezon:

  • (βˆ’K)2=9βˆ’n>0(-K)^2 = 9 - n > 0 for n≀8n \leq 8. Good.
  • (βˆ’K)β‹…Ei=1>0(-K) \cdot E_i = 1 > 0. Good.
  • (βˆ’K)β‹…(Hβˆ’Ei)=2>0(-K) \cdot (H - E_i) = 2 > 0 (strict transform of a line through one point). Good.
  • (βˆ’K)β‹…(Hβˆ’Eiβˆ’Ej)=1>0(-K) \cdot (H - E_i - E_j) = 1 > 0 (strict transform of a line through two points). Good.
  • (βˆ’K)β‹…(2Hβˆ’E1βˆ’β‹―βˆ’E5)=1>0(-K) \cdot (2H - E_1 - \cdots - E_5) = 1 > 0 (conic through five points). Good.

For all irreducible curves CC on SS, one verifies (βˆ’K)β‹…C>0(-K) \cdot C > 0 provided the points are in general position. Hence βˆ’K-K is ample.


The cubic surface and 27 lines

ExampleThe cubic surface as Bl₆(β„™Β²)

A smooth cubic surface SβŠ‚P3S \subset \mathbb{P}^3 is isomorphic to Bl⁑P1,…,P6(P2)\operatorname{Bl}_{P_1, \ldots, P_6}(\mathbb{P}^2) where P1,…,P6P_1, \ldots, P_6 are in general position.

The anticanonical embedding βˆ£βˆ’KS∣=∣3Hβˆ’E1βˆ’β‹―βˆ’E6∣|-K_S| = |3H - E_1 - \cdots - E_6| maps SS into P3\mathbb{P}^3 as a cubic surface. Here (βˆ’K)2=3(-K)^2 = 3, confirming the degree.

The 27 lines on SS: A line β„“βŠ‚SβŠ‚P3\ell \subset S \subset \mathbb{P}^3 is a curve with deg⁑ℓ=1\deg \ell = 1, i.e., (βˆ’K)β‹…β„“=1(-K) \cdot \ell = 1, and β„“β‰…P1\ell \cong \mathbb{P}^1. The 27 lines are:

  • The 66 exceptional divisors: E1,…,E6E_1, \ldots, E_6. Each has (βˆ’K)β‹…Ei=1(-K) \cdot E_i = 1.
  • The 1515 strict transforms of lines PiPjβ€Ύ\overline{P_iP_j}: class Hβˆ’Eiβˆ’EjH - E_i - E_j with (βˆ’K)β‹…(Hβˆ’Eiβˆ’Ej)=3βˆ’1βˆ’1=1(-K) \cdot (H - E_i - E_j) = 3 - 1 - 1 = 1. There are (62)=15\binom{6}{2} = 15 of these.
  • The 66 strict transforms of conics through 55 of the 66 points: class 2Hβˆ’E1βˆ’β‹―βˆ’Ek^βˆ’β‹―βˆ’E6=2Hβˆ’βˆ‘iβ‰ kEi2H - E_1 - \cdots - \widehat{E_k} - \cdots - E_6 = 2H - \sum_{i \neq k} E_i. Each has (βˆ’K)β‹…(2Hβˆ’βˆ‘iβ‰ kEi)=6βˆ’5=1(-K) \cdot (2H - \sum_{i \neq k} E_i) = 6 - 5 = 1. There are (65)=6\binom{6}{5} = 6 of these.

Total: 6+15+6=276 + 15 + 6 = 27 lines. Each is a (βˆ’1)(-1)-curve on SS.

ExampleIncidence of the 27 lines

Each of the 27 lines on a cubic surface meets exactly 1010 of the other 2626 lines. The intersection pattern:

Type EiE_i (6 lines): Eiβ‹…Ej=0E_i \cdot E_j = 0 for jβ‰ ij \neq i. Eiβ‹…(Hβˆ’Ejβˆ’Ek)=1E_i \cdot (H - E_j - E_k) = 1 if i∈{j,k}i \in \{j, k\}, 00 otherwise. Eiβ‹…(2Hβˆ’βˆ‘jβ‰ kEj)=1E_i \cdot (2H - \sum_{j \neq k} E_j) = 1 if i=ki = k, 00 otherwise. So EiE_i meets: the 55 lines Hβˆ’Eiβˆ’EjH - E_i - E_j (jβ‰ ij \neq i) plus the 55 conics 2Hβˆ’βˆ‘jβ‰ kEj2H - \sum_{j \neq k} E_j (kβ‰ ik \neq i). Total: 1010.

Type Hβˆ’Eiβˆ’EjH - E_i - E_j (15 lines): meets EiE_i, EjE_j, and 44 lines of type Hβˆ’Ekβˆ’ElH - E_k - E_l with {k,l}∩{i,j}=βˆ…\{k,l\} \cap \{i,j\} = \varnothing (since (Hβˆ’Eiβˆ’Ej)(Hβˆ’Ekβˆ’El)=1(H-E_i-E_j)(H-E_k-E_l) = 1 iff {i,j}∩{k,l}=βˆ…\{i,j\} \cap \{k,l\} = \varnothing, giving (42)=6\binom{4}{2} = 6... let me recount). We get (Hβˆ’Eiβˆ’Ej)β‹…(Hβˆ’Ekβˆ’El)=1βˆ’Ξ΄ikβˆ’Ξ΄ilβˆ’Ξ΄jkβˆ’Ξ΄jl(H - E_i - E_j) \cdot (H - E_k - E_l) = 1 - \delta_{ik} - \delta_{il} - \delta_{jk} - \delta_{jl}, which is 11 when the sets are disjoint and 00 when they share one element. With 66 points, the number of disjoint pairs is (42)=6\binom{4}{2} = 6. Adding the 22 exceptional curves Ei,EjE_i, E_j, and 22 conics, the count reaches 1010.

This incidence structure is the SchlΓ€fli graph, which has deep connections to the E6E_6 root system and the Weyl group W(E6)W(E_6) of order 5184051840, which acts as the symmetry group of the 2727 lines.


Blowup and Cremona transformations

ExampleThe standard Cremona transformation

The standard Cremona transformation Οƒ:P2β‡’P2\sigma: \mathbb{P}^2 \dashrightarrow \mathbb{P}^2 is defined by [x:y:z]↦[1/x:1/y:1/z]=[yz:xz:xy][x:y:z] \mapsto [1/x : 1/y : 1/z] = [yz : xz : xy]. It is a birational involution, undefined at the three coordinate points P1=[1:0:0]P_1 = [1:0:0], P2=[0:1:0]P_2 = [0:1:0], P3=[0:0:1]P_3 = [0:0:1].

The Cremona transformation factors through blowups:

P2←π1Bl⁑P1,P2,P3(P2)β†’Ο€2P2.\mathbb{P}^2 \xleftarrow{\pi_1} \operatorname{Bl}_{P_1, P_2, P_3}(\mathbb{P}^2) \xrightarrow{\pi_2} \mathbb{P}^2.

Here Ο€1\pi_1 blows up P1,P2,P3P_1, P_2, P_3, and Ο€2\pi_2 blows down the strict transforms of the three coordinate lines P1P2β€Ύ\overline{P_1P_2}, P1P3β€Ύ\overline{P_1P_3}, P2P3β€Ύ\overline{P_2P_3} (each of which is a (βˆ’1)(-1)-curve on the blowup).

On the blowup, the map Ο€2\pi_2 is the anticanonical map βˆ£βˆ’K∣=∣3Hβˆ’E1βˆ’E2βˆ’E3∣|{-K}| = |3H - E_1 - E_2 - E_3|, which has (βˆ’K)2=9βˆ’3=6(-K)^2 = 9 - 3 = 6... actually Οƒβˆ—Hβ€²=2Hβˆ’E1βˆ’E2βˆ’E3\sigma^*H' = 2H - E_1 - E_2 - E_3 (pulling back a line in the target gives a conic through the 33 points). So Οƒ\sigma acts on Pic⁑\operatorname{Pic} by H′↦2Hβˆ’E1βˆ’E2βˆ’E3H' \mapsto 2H - E_1 - E_2 - E_3.


Blowup and the Euler characteristic

ExampleTopological invariants of blowups

Over C\mathbb{C}, the blowup Bl⁑P(X)\operatorname{Bl}_P(X) has:

  • e(Bl⁑P(X))=e(X)+1e(\operatorname{Bl}_P(X)) = e(X) + 1, where ee denotes the topological Euler characteristic. This is because we replace a point (e=1e = 1) by P1\mathbb{P}^1 (e=2e = 2).

  • b2(Bl⁑P(X))=b2(X)+1b_2(\operatorname{Bl}_P(X)) = b_2(X) + 1 (one extra generator [E][E] in H2H_2).

  • H2(Bl⁑P(X),Z)β‰…H2(X,Z)βŠ•Zβ‹…[E]H^2(\operatorname{Bl}_P(X), \mathbb{Z}) \cong H^2(X, \mathbb{Z}) \oplus \mathbb{Z} \cdot [E], and the intersection form gains a (βˆ’1)(-1) summand.

Concrete values for Bl⁑n(P2)\operatorname{Bl}_n(\mathbb{P}^2):

  • e(P2)=3e(\mathbb{P}^2) = 3, so e(Bl⁑n(P2))=3+ne(\operatorname{Bl}_n(\mathbb{P}^2)) = 3 + n.
  • b2=1+nb_2 = 1 + n, b1=0b_1 = 0, b0=b4=1b_0 = b_4 = 1.
  • Intersection form: diag⁑(1,βˆ’1,βˆ’1,…,βˆ’1)\operatorname{diag}(1, -1, -1, \ldots, -1) of signature (1,n)(1, n).
  • Ο‡(O)=1\chi(\mathcal{O}) = 1 (birational invariant for rational surfaces).

Blowups and minimal model program

RemarkRole of blowups in classification

The classification of surfaces proceeds by first finding the minimal model (no (βˆ’1)(-1)-curves), then classifying minimal surfaces by Kodaira dimension ΞΊ\kappa:

  • ΞΊ=βˆ’βˆž\kappa = -\infty: rational or ruled surfaces. Minimal models: P2\mathbb{P}^2, Fn\mathbb{F}_n (nβ‰ 1n \neq 1), or ruled surfaces over curves of genus gβ‰₯1g \geq 1.
  • ΞΊ=0\kappa = 0: K3, Enriques, abelian, or bielliptic surfaces. Already minimal (they have no (βˆ’1)(-1)-curves since Kβ‹…E=βˆ’1K \cdot E = -1 would force KK to be nonzero on a nontrivial curve).
  • ΞΊ=1\kappa = 1: properly elliptic surfaces. Minimal model admits an elliptic fibration.
  • ΞΊ=2\kappa = 2: surfaces of general type. The minimal model has KK nef.

The key theorem (Castelnuovo--Enriques): the process of blowing down (βˆ’1)(-1)-curves terminates. Every birational morphism f:Xβ†’Yf: X \to Y of smooth surfaces is a composition of blowups. The factorization theorem: if f:Xβ‡’Yf: X \dashrightarrow Y is a birational map of smooth surfaces, then there exists a smooth surface ZZ with morphisms Zβ†’XZ \to X and Zβ†’YZ \to Y, each a composition of blowups.

ExampleBlowups and Kodaira dimension

The Kodaira dimension is a birational invariant, so blowing up does not change it:

  • Bl⁑n(P2)\operatorname{Bl}_n(\mathbb{P}^2) has ΞΊ=βˆ’βˆž\kappa = -\infty for all nn: it is rational.
  • Blowing up a point on a K3 surface gives ΞΊ=0\kappa = 0, but now the surface has a (βˆ’1)(-1)-curve, so it is not minimal. Blowing down recovers the K3.
  • Blowing up a point on a surface of general type with K2>0K^2 > 0: the new surface has K2βˆ’1K^2 - 1, still ΞΊ=2\kappa = 2 as long as KK is still big (which it always is for surfaces of general type).

Summary

RemarkKey formulas for blowup

Let Ο€:X~=Bl⁑P(X)β†’X\pi: \widetilde{X} = \operatorname{Bl}_P(X) \to X with exceptional divisor EE. The essential identities:

  • Picard group: Pic⁑(X~)=Ο€βˆ—Pic⁑(X)βŠ•ZE\operatorname{Pic}(\widetilde{X}) = \pi^*\operatorname{Pic}(X) \oplus \mathbb{Z} E.
  • Self-intersection of EE: E2=βˆ’1E^2 = -1.
  • Pullback orthogonality: Ο€βˆ—Dβ‹…E=0\pi^*D \cdot E = 0.
  • Strict transform: C~=Ο€βˆ—Cβˆ’mult⁑P(C)β‹…E\widetilde{C} = \pi^*C - \operatorname{mult}_P(C) \cdot E, C~2=C2βˆ’mult⁑P(C)2\widetilde{C}^2 = C^2 - \operatorname{mult}_P(C)^2.
  • Canonical class: KX~=Ο€βˆ—KX+EK_{\widetilde{X}} = \pi^*K_X + E.
  • K2K^2 drop: KX~2=KX2βˆ’1K_{\widetilde{X}}^2 = K_X^2 - 1.
  • Euler characteristic: e(X~)=e(X)+1e(\widetilde{X}) = e(X) + 1, Ο‡(OX~)=Ο‡(OX)\chi(\mathcal{O}_{\widetilde{X}}) = \chi(\mathcal{O}_X).
  • Contraction criterion: a (βˆ’1)(-1)-curve can always be blown down to a smooth point.