In this section, we will introduce and outline the Fano variety and Fano polytope. For more information see [5], [22], [7], [8] and [14].
Definition 3.4.1. For all r in Z, we define the Hirzebruch surface
Hr:= {([x0 : x1], [y0 : y1 : y2])∈ CP1× CP2 | xr0y0 = xr1y1}.
Since Hr is isomorphic to H−r for all r in Z, we sometimes assume Z≥0.
Theorem 3.4.2. Let P ba a polytope in NR. If all of facets of P are the convex hull of a basis of N if and only if XP is a smooth Fano variety.
Proof. See [14] Proposition 3.6.7 and [7] Lemma 8.5.
The Fano varieties in two-dimension are also called a del Pezzo surface.
Theorem 3.4.3. There exist five distinct toric Fano varieties of two-dimension up to isomorphism,
1. CP2,
2. CP1× CP1,
3. the equivariant blowing-up of CP2 at one point (i.e. the Hirzebruch surface H1),
4. the equivariant blowing-up of CP2 at two point, 5. the equivariant blowing-up of CP2 at three point.
Proof. See [22] Propsition 2.21.
Chapter 4
Tropical Toric Variety
4.1 K(G, R, M )
Definition 4.1.1. For all real number x, we define
x+ := max(x, 0),
x− := max(−x, 0),
and called positive part and negative part of x, respectively.
Remark 4.1.2. The x+ and x− are non-negative and x = x+− x−.
Definition 4.1.3. For all extended real-valued function f , the positive part
of f is defined by f+(x) := max{f(x), 0}, and the negaive part of f is defined by f−(x) := max{−f(x), 0}. So we have f = f+− f−.
Remark 4.1.4. Let f := (f1,· · · , fn) ∈ R1×n, then f+ = (f1+,· · · , fn+), f− = (f1−,· · · , fn−), and f = f+− f−.
Proposition 4.1.5. Let f := (f1,· · · , fn) ∈ R1×n and x := (x1,· · · , xn) ∈ Rn, then the equations f · x = 0 if and only if f+· x = f−· x.
Proof. Suppose that f · x = 0 where f := (f1,· · · , fn) ∈ R1×n and x :=
(x1,· · · , xn) ∈ Rn. Then (f1,· · · , fn)· (x1,· · · , xn) = 0 implies f1 × x1 +
· · · fn× xn = 0. Since fi = fi+− fi− for all i = 1,· · · , n, we have (f1+− f1−)× x1+· · · + (fn+− fn−)× xn= 0. Therefore, f1+× x1+· · · + fn+× xn= f1−× x1 +· · · + fn−× xn. Hence f+· x = f−· x.
Conversely, assume that f+ · x = f− · x where f+, f− ∈ R1×n and x :=
(x1,· · · , xn)∈ Rn, then f1+× x1+· · · + fn+× xn= f1−× x1+· · · + fn−× xn. So we have (f1+− f1−)× x1+· · · + (fn+− fn−)× xn = 0. Hence f· x = 0 since fi = fi+− fi− for all i = 1,· · · , n and x := (x1,· · · , xn)∈ Rn.
Definition 4.1.6. Let S be a semigroup in Zn and let G = {g1,· · · , gm} be a finite set of generators of S. Let R = {r1,· · · , rk} ⊆ Zm generate the integer relation between a set of G, that is, SpanZ(R) = {z ∈ Zm | g1z1 +· · · + gmzm = 0}. Let M be another commutative semigroup. We
define
K(G, R, M ) := �
x∈ M|G|| r+· x = r−· x ∀r ∈ R�
We will discuss about that r· x = 0 is different from r+· x = r−· x in the tropical semifield T.
Example 4.1.7. Let S ⊆ Z2 and let G = {(1, 1), (4, 4)} be the generating set of S. Then |G| = 2, and we have (1, 1)z1+ (4, 4)z2 = 0 for all z1, z2 ∈ Z.
This implies R = {(−4, 1)} and SpanZ(R) = {(z1, z2) ∈ Z2 | z1 + 4z2 = 0} Let M = T and let x = (x1, x2) ∈ T2. Then r · x = 0T where 0T is the tropical additive identity. This implies (−4 ⊙ x1)⊕ (1 ⊙ x2) = 0T, so max{−4 + x1, 1 + x2} = −∞. Hence (x1, x2) = (−∞, −∞).
However, r+· x = (0, 1) · (x1, x2) = (0⊙ x1)⊕ (1 ⊙ x2) = max{0 + x1, 1 + x2}.
Similarly r−·x = max{4+x1, 0+x2}. So {(x1, x2)∈ T2 | r+· x = r−· x ∀r ∈ R} = {−∞, x1+ 3 = x2}. Then K(G, R, T) = {−∞, x1+ 3 = x2}
Hence r· x = 0 is different from r+· x = r−· x in T.
Proposition 4.1.8. Given G = {g1, . . . , gm}. Let R = {r1,· · · , rk} ⊆ Zm Let M be a tropical semifield T. Suppose that K = K(G, R, T) = {x ∈ Tm | r+· x = r−· x ∀r ∈ R}. We define two operations ⊕K : K× K → K and ⊗K : T× K → K by x ⊕Ky = (x1⊕ y1, . . . , xm⊕ ym) and t⊗Kx = (t⊙x1, . . . , t⊙xm), respectively. Then (K,⊕K,⊗K) is a tropical vector space.
Proof. Let x = (x1, . . . , xm), y = (y1, . . . , ym), and v = (v1, . . . , vm) be in K, and let t, t1, and t2 be in T. Given r = (z1, . . . , zm) is in R. Then r+ · x = r− · x and r+ · y = r− · y. We claim that K is closed under an operations ⊕K.
r+· (x ⊕Ky) = (z1+, . . . , zm+)· (x1 ⊕ y1, . . . , xm⊕ ym)
= (z1+⊙ (x1⊕ y1))⊕ · · · ⊕ (zm+⊙ (xm⊕ ym))
= (z1++ max{x1, y1}) ⊕ · · · ⊕ (zm++ max{xm, ym})
= max{z1++ x1, z1++ y1} ⊕ · · · ⊕ max{zm++ xm, zm++ ym}
= ((z1+⊙ x1)⊕ (z1+⊙ y1))⊕ · · · ⊕ ((zm+⊙ xm)⊕ (zm+⊙ ym))
= ((z1+⊙ x1)⊕ · · · ⊕ (zm+⊙ xm))⊕ ((z1+⊙ y1)⊕ · · · ⊕ (z+m⊙ ym)) (since (T,⊕) is a commutative monoid.)
= ((z1+, . . . , zm+)· (x1, . . . , xm))⊕ ((z1+, . . . , zm+)· (y1, . . . , ym))
= (r+· x) ⊕ (r+· y)
= (r−· x) ⊕ (r−· y)
= ((z1−, . . . , zm−)· (x1, . . . , xm))⊕ ((z1−, . . . , zm−)· (y1, . . . , ym))
= ((z1−⊙ x1)⊕ · · · ⊕ (zm−⊙ xm))⊕ ((z1−⊙ y1)⊕ · · · ⊕ (z−m⊙ ym))
= ((z1−⊙ x1)⊕ (z1−⊙ y1))⊕ · · · ⊕ ((zm−⊙ xm)⊕ (zm−⊙ ym)) (since (T,⊕) is a commutative monoid.)
= max{z1−+ x1, z1−+ y1} ⊕ · · · ⊕ max{zm−+ xm, zm−+ ym}
= (z1−+ max{x1, y1}) ⊕ · · · ⊕ (zm−+ max{xm, ym})
= (z1−⊙ (x1⊕ y1))⊕ · · · ⊕ (zm−⊙ (xm⊕ ym))
= (z1−, . . . , zm−)· (x1 ⊕ y1, . . . , xm⊕ ym)
= r−· (x ⊕Ky)
Hence x⊕K y is in K.
We claim that K is closed under an operations ⊗K.
r+· (t ⊗K x) = (z1+, . . . , zm+)· (t ⊙ x1, . . . , t⊙ xm)
= (z1+⊙ (t ⊙ x1))⊕ · · · ⊕ (zm+⊙ (t ⊙ xm))
= ((z1+⊙ t) ⊙ x1)⊕ · · · ⊕ ((zm+⊙ t) ⊙ xm) (since (T,⊙) satisfies associative.)
= ((t⊙ z1+)⊙ x1)⊕ · · · ⊕ ((t ⊙ zm+)⊙ xm) (since (T\ {0T}, ⊙) is abelian.)
= (t⊙ (z1+⊙ x1))⊕ · · · ⊕ (t ⊙ (zm+⊙ xm))
= t⊙ ((z1+⊙ x1)⊕ · · · ⊕ (zm+⊙ xm)) ( since (T,⊕, ⊙) is a semifield.)
= t⊙ (r+· x)
= t⊙ (r−· x)
= t⊙ ((z1−⊙ x1)⊕ · · · ⊕ (zm−⊙ xm))
= (t⊙ (z1−⊙ x1))⊕ · · · ⊕ (t ⊙ (zm−⊙ xm)) (since (T,⊕, ⊙) is a semifield.)
= ((t⊙ z1−)⊙ x1)⊕ · · · ⊕ ((t ⊙ zm−)⊙ xm) (since (T\ {0T}, ⊙) is abelian.)
= ((z1−⊙ t) ⊙ x1)⊕ · · · ⊕ ((zm−⊙ t) ⊙ xm)
= (z1−⊙ (t ⊙ x1))⊕ · · · ⊕ (zm−⊙ (t ⊙ xm)) (since (T,⊙) satisfies associative.)
= (z1−, . . . , zm−)· (t ⊙ x1, . . . , t⊙ xm)
= r−· (t ⊗Kx)
Hence t⊗Kx is in K.
x⊕K(y⊕Kv) = x⊕K(y1⊕ v1, . . . , ym⊕ vm)
= (x1⊕ (y1⊕ v1), . . . , xm⊕ (ym⊕ vm))
= ((x1⊕ y1)⊕ v1), . . . , (xm⊕ ym)⊕ vm)
= (x1⊕ y1, . . . , xm⊕ ym)⊕Kv
= (x⊕K y)⊕Kv
(0T, . . . , 0T)⊕Kx = (−∞, . . . , −∞) ⊕K (x1, . . . , xm)
= ((−∞) ⊕ x1, . . . , (−∞) ⊕ xm)
= (x1, . . . , xm)
(t1⊙ t2)⊗Kx = ((t1 ⊙ t2)⊙ x1, . . . , (t1⊙ t2)⊙ xm)
= (t1⊙ (t2⊙ x1), . . . , t1(⊙t2⊙ xm))
= t1⊗K(t2⊙ x1, . . . , t2⊙ xm)
= t1⊗K(t2⊗K x)
(t1⊕ t2)⊗Kx = ((t1⊕ t2)⊙ x1, . . . , (t1⊕ t2)⊙ xm)
= ((t1⊙ x1)⊕ (t2⊙ x1), . . . , (t1⊙ xm)⊕ (t2⊙ xm))
= (t1⊙ x1, . . . , t1⊙ xm)⊕K(t2⊙ x2, . . . , t1⊙ xm)
= (t1⊗Kx)⊕K(t2⊗Kx)
t⊗K(x⊕Ky) = t⊗K(x1⊕ y1, . . . , xm⊕ ym)
= (t⊙ (x1⊕ y1), . . . , t⊙ (xm⊕ ym))
= ((t⊙ x1)⊕ (t ⊙ y1), . . . , (t⊙ xm)⊕ (t ⊙ ym))
= (t⊙ x1, . . . , t⊕ xm)⊕K(t⊙ y1, . . . , t⊕ ym)
= (t⊗Kx)⊕K(t⊗Ky)
Hence (K,⊕K,⊗K) is a tropical vector space.
Proposition 4.1.9. If M is a abelian group, then K(G, R, M ) is a abelian
group.
Proof. Let M be an abelian group with an operation ∗ : M × M → M by ∗(a, b) = a ∗ b. Then M|G| is also an abelian group. Because G is finite set, so we can consider |G| = k. We want to show that K(G, R, M) is a subgroup of Mk. Let x = (x1, . . . , xk) and y = (y1, . . . , yk) in K(G, R, M ).
We claim that x∗ y−1 = (x1∗ y−11 , . . . , xk∗ yk−1) in K(G, R, M ) where y−1 is the inverse for y. Since M is an abelian group, we have xi∗ y−1i in M for all i = 1, . . . , k, thus x∗ y−1 in Mk. Let r = (z1, . . . , zk) in SpanZR. Since M is an abelian, M is a Z-module, thus (zi+· (xi∗ y−1i )) = ((z+i · xi)∗ (z+i · yi−1)) and (zi−· (xi∗ yi−1)) = ((zi−· xi)∗ (z−i · yi−1)) for all i = 1, . . . , k. Moreover, r+· x = r− · x and r+· y−1 = r−· y−1, because x = (x1, . . . , xk) and y = (y1, . . . , yk) in K(G, R, M ). So we have
r+· (x ∗ y−1) = (z1+, . . . , zk+)· (x1∗ y−11 , . . . , xk∗ y−1k )
= (z1+· (x1∗ y−11 ))∗ · · · ∗ (zk+· (xk∗ yk−1))
= ((z1+· x1)∗ (z1+· y1−1))∗ · · · ∗ ((zk+· xk)∗ (z+k · y−1k ))
= ((z1+· x1)∗ · · · ∗ (zk+· xk))∗ ((z1+· y1−1)∗ · · · ∗ (z+k · yk−1))
= r+· x ∗ r+· y−1
= r−· x ∗ r−· y−1
= ((z1−· x1)∗ · · · ∗ (zk−· xk))∗ ((z1−· y1−1)∗ · · · ∗ (z−k · yk−1))
= ((z1−· x1)∗ (z1−· y1−1))∗ · · · ∗ ((zk−· xk)∗ (z−k · y−1k ))
= (z1−· (x1∗ y−11 ))∗ · · · ∗ (zk−· (xk∗ yk−1))
= (z1−, . . . , zk+)· (x1∗ y−11 , . . . , xk∗ y−1k )
= r−· (x ∗ y−1).
Hence we get x∗ y−1 in K(G, R, M ), i.e. K(G, R, M ) is a group.
Next we claim that K(G, R, M ) is an abelian. Suppose that x = (x1, . . . , xk) and y = (y1, . . . , yk) in K(G, R, M ). Since M is an abelian, xi ∗ yi = yi∗ xi for all i = 1, . . . , k. Then
x∗ y = (x1, . . . , xk)∗ (y1, . . . , yk)
= (x1∗ y1, . . . , xk∗ yk)
= (y1∗ x1, . . . , yk∗ xk)
= y∗ x
Hence K(G, R, M ) is an abelian group.
By the above proposition, K(G, R, T\ {−∞}) is an abelian group since (T\ {−∞}, ⊙) is an abelian group.
Example 4.1.10. Let S ⊆ Z2 and let G ={(1, 2), (2, 4)} be the generating set of S. Then |G| = 2, and we have (1, 2)z1+ (2, 4)z2 = 0 for all z1, z2 ∈ Z.
This implies R ={r = (−2, 1)}. So r+= (0, 1) and r− = (2, 0).
Given an abelian group M = (Z4, +). Let x = (x1, x2)∈ Z4× Z4. If r+· x = r−· x, then x2 = 2x1, and so K(G, R, Z4) = {(x1, x2)∈ Z4× Z4 | x2 = x1} is subset of Z4× Z4.
We claim that K(G, R, Z4) is a subgroup of Z24. Suppose that (x1, x2) and (y1, y2) are in K(G, R, Z4), then x2 = 2x1and y2 = 2y1. Then 2(x1+(−y1)) = 2x1 + 2(−y1) = 2x1+ (−2y1) = x2 + (−y2) where (−y1,−y2) is the inverse element of (y1, y2), so (x1+ (−y1), x2+ (−y2)) is in K(G, R, Z4).
Since xi+ yi = yi+ xi for all i = 1, 2, (x1, x2) + (y1, y2) = (x1+ y1, x2+ y2) = (y1+ x1, y2+ x2) = (y1, y2) + (x1, x2)
Hence K(G, R, Z4) is abelian group.
Proposition 4.1.11. If M is a ring, then K(G, R, M ) is a M -module.
Proof. Suppose that M is a ring with binary operation ∗, together with a second binary operation ⊗. Because (M, ∗, ⊗) is a ring, so (M, ∗) is an abelian group, thus K(G, R, M ) is also an abelian group.
Define a operation � : K(G, R, M) × M → K(G, R, M) via x � m = (x1⊗ m, . . . , xk ⊗ m). To check that it is well-defined. Suppose that r is in R and |G| = k. Let m, n be in M and let x = (x1, . . . , xk), y = (y1, . . . , yk) in
K(G, R, M ). Then r+· x = r−· x.
r+· (x � m) = (z1+, . . . , z+k)· (x1⊗ m, . . . , xk⊗ m)
= z1+⊗ (x1⊗ m) ∗ · · · ∗ zk+⊗ (xk⊗ m)
= (z1+⊗ x1)⊗ m ∗ · · · ∗ (zk+⊗ xk)⊗ m
= (z1+⊗ x1, . . . , z+k ⊗ xk)� m
= (r+· x) � m
= (r−· x) � m
= (z1−⊗ x1, . . . , z−k ⊗ xk)� m
= (z1−⊗ x1)⊗ m ∗ · · · ∗ (zk−⊗ xk)⊗ m
= z1−⊗ (x1⊗ m) ∗ · · · ∗ zk−⊗ (xk⊗ m)
= (z1−, . . . , z−k)· (x1⊗ m, . . . , xk⊗ m)
= r−· (x � m)
So x� m is in K(G, R, M).
Suppose that x = y (i.e. xi = yi for all i = 1, . . . , k) and m = n, then x� m = (x1 ⊗ m, . . . , xk⊗ m) = (y1 ⊗ n, . . . , yk⊗ n) = y � n. Hence it is well-defined.
(x∗ y) � m = (x1∗ y1, . . . , xk∗ yk)� m
= ((x1∗ y1)⊗ m, . . . , (xk∗ yk)⊗ m)
= ((x1⊗ m) ∗ (y1⊗ m), . . . , (xk⊗ m) ∗ (yk⊗ m))
= (x1⊗ m, . . . , xk⊗ m) ∗ (y1⊗ m, . . . , yk⊗ m)
= (x� m) ∗ (y � m)
x� (m ∗ n) = (x1, . . . , xk)� (m ∗ n)
= (x1⊗ (m ∗ n), . . . , xk⊗ (m ∗ n))
= ((x1⊗ m) ∗ (x1⊗ n), . . . , (xk⊗ m) ∗ (xk⊗ n))
= (x1⊗ m, . . . , xk⊗ m) ∗ (x1⊗ n, . . . , xk⊗ n)
= (x� m) ∗ (x � n)
(x� m) � n = (x1⊗ m, . . . , xk⊗ m) � n
= ((x1⊗ m) ⊗ n, . . . , (xk⊗ m) ⊗ n)
= (x1⊗ (m ⊗ n), . . . , xk⊗ (m ⊗ n))
= x� (m ⊗ n)
If M has an identity 1M, that is m ⊗ 1M = m for all m ∈ M. Then x� 1M = (x1⊗ 1M, . . . , xk⊗ 1M) = (x1, . . . , xk) = x.
Remark 4.1.12. If M is a field, then K(G, R, M ) is a vector space.
Example 4.1.13. Let M = GF (4) be a Galois field. Let S ⊆ Z2 and let G = {(1, 2), (4, 8)} be the generating set of S. Then |G| = 2, and we have (1, 2)z1+ (4, 8)z2 = 0 for all z1, z2 ∈ Z. This implies R = {r = (−4, 1)}. So r+= (0, 1) and r− = (4, 0). If r+· x = r−· x where x = (x1, x2) is in GF (4)2, then x2 = 4x1 = 0, and so K(G, R, GF (4)) ={(x1, x2)∈ GF (4)2 | x2 = 0} is a subset of GF (4)2. We claim that K(G, R, GF (4)) is a vector space, in fact, we just need to show that K(G, R, GF (4)) is a subspace of GF (4)2. To start with, it is clearly that (0, 0) is in K(G, R, GF (4)). Next, suppose that x = (x1, x2) and y = (y1, y2) are in K(G, R, GF (4)), then x2 = 0 and y2 = 0, then x2+y2 = 0, and so x+y is in K(G, R, GF (4)). Finally, let c be in GF (4), and let x = (x1, x2) be in K(G, R, GF (4)), then x2 = 0 and cx = (cx1, cx2), then cx2 = 0, and so cx is in K(G, R, GF (4)). Hence K(G, R, GF (4)) is a subspace of GF (4)2.
Note that the Galois field GF (4) is isomorphic to GF (2)�
(x2+ x + 1), in fact GF (2) � Z2 since GF (p)� Zp for all prime p.
Theorem 4.1.14. Let S be a finitely generated semigroup on Zn. Let G = {g1,· · · , gl} be a set of generaters with relations generated by R = {r1,· · · , rk}. Let M be an additive semigroup. Then there is a bijection between hom(S, M ) and K(G, R, M ).
Proof. We define a map φ : hom(S, M ) → K(G, R, M). Define a function
K(G, R, M ). To check that φ is surjective, and ψ is injective, i.e. φ ◦ ψ identity function.
φ◦ ψ(x) = φ(ψ(x)) = (f(g1), . . . , f (gl)) = (x1, . . . , xl) = x.
Proposition 4.1.15. Let S be a finitely generated semigroup on Zn. Let G = {g1,· · · , gl} be a set of generaters with relations generated by R = {r1,· · · , rk}. Then there are bijective between Spec(C[S]), hom(S, C) and K(G, R, C).
Proof. The correspondence between Spec(C[Sσ]) and hom(S, C) is imme-diate from Proposition 3.3.11. The correspondence between hom(S, C) and K(G, R, C) is immediate from the Theorem 4.1.14.
Theorem 4.1.16. Let S be a finitely generated semigroup on Zn. Let G = {g1,· · · , gl} and H = {h1,· · · , hd} be the different sets of the generators of S with relations generated by R = {r1,· · · , rk}and P = {p1,· · · , pm}, re-spectively. Then there is a linear isomorphism φ : K(G, R, R)→ K(H, P, R).
Proof. Since G = {g1,· · · , gl} and H = {h1,· · · , hd} are the different sets of the generators of S, let gi = �d
j=1λijhj with nonnegative integer λij for all i = 1, . . . , l, and let hi = �l
i=1µjigi with nonnegative integer µji for all j = 1, . . . , d. Since hom(S, R) is in bijection with K(G, R, R), let xi = f (gi)
for all i = 1, . . . , l where f is in hom(S, R). Then �l
=
i=1µdixi). Then the function is well-defined by the above discussion.
−x1}, thus S = {(1, −2, −1}. So we have
K(H, P, R) ={(x1, x2, x3)∈ R3 | (1, 0, 0) · (x1, x2, x3) = (0, 2, 1)· (x1, x2, x3)}
={(x1, x2, x3)∈ R3 | x1 = 2x2+ x3}.
Then the function φ : K(G, R, R) → K(H, P, R) via φ(x1, x2) = (x1 + 2x2, x2, x1) is a linear isomophism.
4.2 Tropical Toric Variety
Definition 4.2.1. Let T be a tropical semifield. Let σ be a rational poly-hedral cone with the semigroup Sσ. Then the affine toric variety is Uσ :=
hom(Sσ, T) where hom(Sσ, T) is the semigroup homomorphisms Sσ → T.
Note that we set T∗ = T\ {−∞}.
Example 4.2.2. Let N � Z2 be a lattice with associated vector space NR � R2, and let M be a dual lattice of N with associated vector space MR � R2. Given a cone σ = pos{0} in NR, then its dual cone {0}∨ = pos{(1, 0), (−1, 0), (0, 1), (0, −1)}, then S{0}={0}∨∩M = Z≥0(1, 0)⊕Z≥0(−1, 0)⊕
Z≥0(0, 1)⊕ Z≥0(0,−1). Because U{0} = hom(S{0}, T). Define a homomor-phism f : S{0} → T by f(1, 0) = x1, f (−1, 0) = x2, f (0, 1) = x3, and
f (0,−1) = x4. Then x1 + x2 = 0 and x3 + x4 = 0, and so x1, x2, x3, and x4 don’t equal −∞. So U{0} = hom(S{0}, T) � (T∗)2. Hence (T∗)2 is the 2-dimensional algebraic torus over T.
Example 4.2.3. Let σ = pos{(0, −1)} in NR � R2, then the dual cone σ∨ = pos{(1, 0), (−1, 0), (0, −1)}, and the corresponding semigroup Sσ = Z≥0(1, 0) ⊕ Z≥0(−1, 0) ⊕ Z≥0(0,−1). Let f be in hom(Sσ, T). Suppose that f (1, 0) = x1, f (−1, 0) = x2, and f (0,−1) = x3, then 0 = f (0, 0) = f (0, 1) + f (0,−1) = f(0, 1) ⊗ f(0, −1) = x1 ⊗ x2 = x1 + x2. Hence the corresponding affine toric variety is Uσ = hom(Sσ, T) = R× R.
Proposition 4.2.4. If τ ⊂ σ is a face of a cone σ, then we obtain the embedding hom(Sτ, T) = Uτ �→ Uσ = hom(Sσ, T).
Proof. Because τ ⊂ σ is a face of a cone σ, so σ∨ is a subset of τ∨. Since Sσ = σ∨∩ M is a subset of Sτ = τ∨∩ M, we have the embedding Sσ �→ Sτ. Because Uτ = hom(Sτ, T) and Uσ = hom(Sσ, T), so hom(Sτ, T) = Uτ �→ Uσ = hom(Sσ, T). Note that we have the following commutative diagram:
Sσ� � h ��
f◦h
��Sτ
f ��T
Definition 4.2.5. Let F be a rational fan. Then the tropical toric variety
XF(T) is defined as the quotient
Note that we have the following commutative diagram:
Uτ∩σ⊂h2> Uσ
Definition 4.2.7. Let XF(T) be a tropical toric variety. A point p is said to be regular or smooth in XF(T) if there is a unimodular cone σ in F such that p is in Uσ. A tropical toric variety XF(T) is called regular or smooth if all points of XF(T) are regular or smooth. A tropical toric variety XF(T) is called singular if XF(T) is not regular.
Definition 4.2.8. Let F be fan in NR. A set ∆ is called a subfan of F if ∆ is a subset of F and is also a fan.
Definition 4.2.9. Let the lattice N = Ze1 ⊕ · · · ⊕ Zen, and let another lattice N� = Ze1⊕ · · · ⊕ Zem. Let F be a fan in NR, and let ∆ be another fan in N�. Let φ : NR → NR� be a linear map such that φ(N ) ⊆ N� and there is a cone σ� in ∆ containing φ(σ) for all cone σ in F . Then the map φ : F → ∆ is said a map of fans.
Theorem 4.2.10. Let N � Zn and N� � Zm be two different lattice, and let M = hom(N, Z) and M� = hom(N�, Z) be the dual lattice. Let F be a fan in NR and let ∆ be another fan in NR� . Let φ : F → ∆ be a map of fans.
Then φ extends to a continuous map φ : XF(T)→ X∆(T).
Proof. Since φ : F → ∆ is a map of fans, there is a cone σ� in ∆ such that φ(σ)⊂ σ�. Let u be in Sσ�. Then u is in hom(N�, Z) and is in σ�∨, and thus we have the map u : N� → Z via u(n�) =< u, n� >. Since φ : F → ∆ is a
map of fans, φ(N ) ⊂ N� u ��Z . Let µ : N → Z via µ(n) =< u, φ(n) >, then µ is in hom(N, Z). Since φ(σ) is a subset of σ�, φ(n) is in σ� for all n in σ. Since u is in σ�∨, < u, φ(n) >≥ 0, and this implies µ is in σ∨. So µ is in σ∨∩ hom(N, Z) = Sσ. Therefore, we obtain a map Sσ� → Sσ, and thus have a map Uσ → Uσ�. And we have the following commutative diagram:
Uσ1 < ⊃Uσ1∩σ2 ⊂ > Uσ2
Uσ1�
∨ < ⊃Uσ�1∩σ�2
∨
⊂ > Uσ2�
∨
Hence XF(T)→ X∆(T).
Definition 4.2.11. For all r in Z, we define the tropical Hirzebruch surface Hr to be
THr:= {([x0 : x1], [y0 : y1 : y2])∈ TP1 × TP2 | rx0+ y0 = rx1+ y1}.
For more information on tropical Hirzebruch surfaces, see [3] § 2.
Figure 4.1: the fan ∆
Figure 4.2: the polytope P
4.3 Smooth two-dimensional tropical toric Fano varieties
In this section, we know that there are only five smooth Fano polytopes in R2 up to the action of GL(2, Z), so I will calculate these cases of smooth two-dimensional tropical toric Fano varieties.
Example 4.3.1. Given the lattice N = Z(1, 0)⊕ Z(0, 1) � Z2, then NR = N ⊗ R � R2, the dual lattice M � Z2 and MR= M ⊗ R.
Let the fan ∆ in NR. Suppose that the fan ∆ (the figure 4.1 ) has
σ1 = pos{(1, 0), (0, 1)}, σ2 = pos{(−1, 0), (0, 1)},
σ3 = pos{(−1, 0), (0, −1)}, σ4 = pos{(1, 0), (0, −1)},
together with
τ1 = σ1∩ σ2 = pos{(0, 1)}, τ2 = σ2∩ σ3 = pos{(−1, 0)},
τ3 = σ3∩ σ4 = pos{(0, −1)}, τ4 = σ4 ∩ σ1 = pos{(1, 0)},
and the origin. Then the dual cones
σ1∨ = pos{(1, 0), (0, 1)}, σ2∨ = pos{(−1, 0), (0, 1)},
σ∨3 = pos{(−1, 0), (0, −1)}, σ4∨ = pos{(1, 0), (0, −1)}.
Moreover, the corresponding semigroups
Sσ1 = σ1∨∩ M = Z≥0(1, 0)⊕ Z≥0(0, 1),
Sσ2 = σ2∨∩ M = Z≥0(−1, 0) ⊕ Z≥0(0, 1),
Sσ3 = σ3∨∩ M = Z≥0(−1, 0) ⊕ Z≥0(0,−1),
Sσ4 = σ4∨∩ M = Z≥0(1, 0)⊕ Z≥0(0,−1),
together with
Sτ1 = Sσ1 + Sσ2 = Z≥0(1, 0)⊕ Z≥0(−1, 0) ⊕ Z≥0(0, 1),
Sτ2 = Sσ2+ Sσ3 = Z≥0(0, 1)⊕ Z≥0(0,−1) ⊕ Z≥0(−1, 0),
Sτ3 = Sσ3+ Sσ4 = Z≥0(1, 0)⊕ Z≥0(−1, 0) ⊕ Z≥0(0,−1),
Sτ4 = Sσ4 + Sσ1 = Z≥0(0, 1)⊕ Z≥0(0,−1) ⊕ Z≥0(1, 0),
S{0}= Z≥0(1, 0)⊕ Z≥0(0, 1)⊕ Z≥0(−1, 0) ⊕ Z≥0(0,−1).
Let fi be in Uσi = hom(Sσi, T) for all i = 1, 2, 3, 4, then we have some maps
f1 : Sσ1 → T via f1(1, 0) = x and f1(0, 1) = y,
f2 : Sσ2 → T via f2(−1, 0) = −x and f2(0, 1) = y,
f3 : Sσ3 → T via f3(−1, 0) = −x and f3(0,−1) = −y,
f4 : Sσ4 → T via f4(1, 0) = x and f4(0,−1) = −y.
Therefore, the affine toric varieties
Uσ1 = hom(Sσ1, T) = T2, Uσ2 = hom(Sσ2, T) = T2,
Uσ3 = hom(Sσ3, T) = T2, Uσ4 = hom(Sσ4, T) = T2,
together with
Uτ1 = hom(Sτ1, T) = R× T, Uτ2 = hom(Sτ2, T) = R× T,
Uτ3 = hom(Sτ3, T) = R× T, Uτ4 = hom(Sτ4, T) = R× T,
U{0}= hom(S{0}, T) = R2.
The gluing of the affine toric varieties Uσ1 and Uσ2 along their common subset Uτ1 gives TP1 × T with coordinates ((x0 : x1), y) where x = x1− x0. The gluing of the affine toric varieties Uσ2 and Uσ3 along their common subset Uτ2 gives T× TP1 with coordinates (−x, (y0 : y1)) where y = y1 − y0. The gluing of the affine toric varieties Uσ3 and Uσ4 along their common subset Uτ3 gives TP1 × T with coordinates ((x0 : x1),−y) where x = x1− x0. The gluing of the affine toric varieties Uσ4 and Uσ1 along their common subset Uτ4 gives T× TP1 with coordinates (x, (y0 : y1)) where y = y1 − y0. The
following commutative diagram:
Hence the gluing of these two gives the tropical toric variety
X∆(T) = (� basis of N , XP is a smooth Fano polytope (by Theorem 3.4.2).
Example 4.3.2. Given the lattice N � Z2, then NR = N ⊗ R � R2, the dual lattice M � Z2 and MR = M⊗ R.
Let the fan ∆ in NR. Suppose that the fan ∆ (the figure 4.3) has
σ1 = pos{(1, 0), (0, 1)}, σ2 = pos{(−1, −1), (0, 1)}, σ3 = pos{(1, 0), (−1, −1)},
Figure 4.3: the fan ∆
Figure 4.4: the polytope P
together with
τ1 = σ1∩ σ2 = pos{(0, 1)}, τ2 = σ2∩ σ3 = pos{(−1, −1)},
τ3 = σ3∩ σ1 = pos{(1, 0)}, and the origin.
Then the dual cones
σ1∨ = pos{(1, 0), (0, 1)}, σ2∨ = pos{(−1, 0), (−1, 1)}, σ∨3 = pos{(1, −1), (0, −1)}.
Moreover, the corresponding semigroups
Sσ1 = σ1∨∩ M = Z≥0(1, 0)⊕ Z≥0(0, 1),
Sσ2 = σ2∨∩ M = Z≥0(−1, 0) ⊕ Z≥0(−1, 1),
Sσ3 = σ3∨∩ M = Z≥0(1,−1) ⊕ Z≥0(0,−1),
together with
Sτ1 = Sσ1 + Sσ2 = Z≥0(1, 0)⊕ Z≥0(−1, 0) ⊕ Z≥0(0, 1),
Sτ2 = Sσ2 + Sσ3 = Z≥0(1,−1) ⊕ Z≥0(−1, 1) ⊕ Z≥0(−1, −1),
Sτ3 = Sσ3 + Sσ1 = Z≥0(1, 0)⊕ Z≥0(0, 1)⊕ Z≥0(0,−1),
S{0}= Z≥0(1, 0)⊕ Z≥0(0, 1)⊕ Z≥0(−1, 0) ⊕ Z≥0(0,−1).
Let fi be in Uσi = hom(Sσi, T) for all i = 1, 2, 3, then we have some maps
f1 : Sσ1 → T via f1(1, 0) = x and f1(0, 1) = y,
f2 : Sσ2 → T via f2(−1, 0) = −x and f2(−1, 1) = −x + y,
f3 : Sσ3 → T via f3(1,−1) = x − y and f3(0,−1) = −y,
Therefore, the affine toric variety
Uσ1 = hom(Sσ1, T) = T2, Uσ2 = hom(Sσ2, T) = T2,
Uσ3 = hom(Sσ3, T) = T2,
together with
Uτ1 = hom(Sτ1, T) = R× T, Uτ2 = hom(Sτ2, T) = R× T,
Uτ3 = hom(Sτ3, T) = R× T, U{0} = hom(S{0}, T) = R2.
The gluing of the affine toric varieties Uσ1 and Uσ2 along their common subset Uτ1 gives TP2 with coordinates (x0 : x1 : x2) where x = x1 − x0
Hence the gluing of these two gives the tropical toric variety
X∆(T) = (�
σ∈∆
Uσ)/∼= TP2.
The polytopes P = conv{0, (1, 0), (0, 1), (−1, −1)} (the figure 4.4) in NR.Since conv{(1, 0), (0, 1)}, conv{(0, 1), (−1, −1)}, conv{(−1, −1), (1, 0)}
are the facets of P , and they are the convex hull of a basis of N , XP is a smooth Fano polytope (by Theorem 3.4.2).
Example 4.3.3. Given the lattice N � Z2, then NR = N ⊗ R � R2, the dual lattice M � Z2 and MR = M⊗ R.
Let the fan ∆ in NR. Suppose that the fan ∆ (the figure 4.5) has
σ1 = pos{(1, 0), (1, 1)}, σ2 = pos{(1, 1), (0, 1)},
σ3 = pos{(0, 1), (−1, −1)}, σ4 = pos{(−1, −1), (1, 0)},
together with
τ1 = σ1∩ σ2 = pos{(1, 1)}, τ2 = σ2 ∩ σ3 = pos{(0, 1)},
τ3 = σ3∩ σ4 = pos{(−1, −1)}, τ4 = σ4 ∩ σ1 = pos{(1, 0)},
Figure 4.5: the fan ∆
Figure 4.6: the polytope P
and the origin. Then the dual cones
σ∨1 = pos{(−1, −1), (0, 1)}, σ2∨ = pos{(1, 0), (−1, 1)},
σ∨3 = pos{(−1, 0), (−1, 1)}, σ4∨ = pos{(0, −1), (1, −1)}.
Moreover, the corresponding semigroups
Sσ1 = σ1∨∩ M = Z≥0(0, 1)⊕ Z≥0(1,−1),
Sσ2 = σ2∨∩ M = Z≥0(1, 0)⊕ Z≥0(−1, 1),
Sσ3 = σ3∨∩ M = Z≥0(−1, 0) ⊕ Z≥0(−1, 1),
Sσ4 = σ4∨∩ M = Z≥0(0,−1) ⊕ Z≥0(1,−1),
together with
Sτ1 = Sσ1+ Sσ2 = Z≥0(1, 1)⊕ Z≥0(1,−1) ⊕ Z≥0(−1, 1),
Sτ2 = Sσ2 + Sσ3 = Z≥0(0, 1)⊕ Z≥0(1, 0)⊕ Z≥0(−1, 0),
Sτ3 = Sσ3 + Sσ4 = Z≥0(−1, −1) ⊕ Z≥0(1,−1) ⊕ Z≥0(−1, 1),
Sτ4 = Sσ4 + Sσ1 = Z≥0(1, 0)⊕ Z≥0(0, 1)⊕ Z≥0(0,−1),
S{0}= Z≥0(1, 0)⊕ Z≥0(0, 1)⊕ Z≥0(−1, 0) ⊕ Z≥0(0,−1).
Let fi be in Uσi = hom(Sσi, T) for all i = 1, 2, 3, 4, then we have some maps
f1 : Sσ1 → T via f1(0, 1) = y and f1(1,−1) = x − y,
f2 : Sσ2 → T via f2(1, 0) = x and f2(−1, 1) = −x + y,
f3 : Sσ3 → T via f3(−1, 0) = −x and f3(−1, 1) = −x + y,
f4 : Sσ4 → T via f4(0,−1) = −y and f4(1,−1) = x − y.
Therefore, the affine toric variety
Uσ1 = hom(Sσ1, T) = T2, Uσ2 = hom(Sσ2, T) = T2,
Uσ3 = hom(Sσ3, T) = T2, Uσ4 = hom(Sσ4, T) = T2,
together with
Uτ1 = hom(Sτ1, T) = R× T, Uτ2 = hom(Sτ2, T) = R× T,
Uτ3 = hom(Sτ3, T) = R× T, Uτ4 = hom(Sτ4, T) = R× T,
U{0}= hom(S{0}, T) = R2.
The gluing of the affine toric varieties Uσ2 and Uσ3 along their common subset Uτ2 gives TP1 × T with coordinates ((x0 : x1),−x + y) where x = x0 − x1. The gluing of the affine toric varieties Uσ4 and Uσ1 along their common subset Uτ4 gives TP1 × T with coordinates ((y0 : y1), x− y) where y = y0− y1.
The two copies of TP1×T are glued along their second coordinates gives TP1× TP2 with coordinates ([z0 : z1], [x0− x1 : y0− y1 : 0]) = ([z0 : z1], [x0− x1 : y0− y1 : 0]) where x− y = z0− z1. Since z0− z1 = (x0− x1)− (y0− y1), (x0− x1)− z0 = (y0− y1)− z1. Hence the toric variety
X∆(T) = (�
σ∈∆Uσ)�
∼
={([z0 : z1], [x0− x1 : y0− y1 : 0]) ∈ TP1× TP2 | (x0− x1)− z0 =
(y0− y1)− z1},
={([z0 : z1], [x0+ y1 : y0+ x1 : x1+ y1])∈ TP1× TP2 | (x0− x1)− z0 = (y0− y1)− z1},
= TH−1
where TH−1 is a tropical Hirzebruch surf ace.
The following commutative diagram: basis of N , XP is a smooth Fano polytope (by Theorem 3.4.2).
Example 4.3.4. Given the lattice N � Z2, then NR = N ⊗ R � R2, the dual lattice M � Z2 and MR = M⊗ R.
Let the fan ∆ in NR. Suppose that the fan ∆ (the figure 4.7) has
σ1 = pos{(1, 0), (1, 1)}, σ2 = pos{(1, 1), (0, 1)},
σ3 = pos{(0, 1), (−1, −1)}, σ4 = pos{(−1, −1), (0, −1)},
σ5 = pos{(0, −1), (1, 0)},
together with
τ1 = σ1∩ σ2 = pos{(1, 1)}, τ2 = σ2 ∩ σ3 = pos{(0, 1)},
τ3 = σ3∩ σ4 = pos{(−1, −1)}, τ4 = σ4 ∩ σ5 = pos{(0, −1)},
τ5 = σ5∩ σ1 = pos{(1, 0)},
and the origin. Then the dual cones
σ∨1 = pos{(1, −1), (0, 1)}, σ2∨ = pos{(1, 0), (−1, 1)},
σ∨3 = pos{(−1, 0), (−1, 1)}, σ4∨ = pos{(−1, 0), (1, −1)},
σ∨5 = pos{(0, −1), (1, 0)}.
Moreover, the corresponding semigroups
Sσ1 = σ1∨∩ M = Z≥0(0, 1)⊕ Z≥0(1,−1),
Sσ2 = σ2∨∩ M = Z≥0(1, 0)⊕ Z≥0(−1, 1),
Sσ3 = σ3∨∩ M = Z≥0(−1, 0) ⊕ Z≥0(−1, 1),
Sσ4 = σ4∨∩ M = Z≥0(−1, 0) ⊕ Z≥0(1,−1),
Sσ5 = σ5∨∩ M = Z≥0(0,−1) ⊕ Z≥0(1, 0),
together with
Sτ1 = Sσ1+ Sσ2 = Z≥0(1, 1)⊕ Z≥0(1,−1) ⊕ Z≥0(−1, 1),
Sτ2 = Sσ2 + Sσ3 = Z≥0(0, 1)⊕ Z≥0(1, 0)⊕ Z≥0(−1, 0),
Sτ3 = Sσ3 + Sσ4 = Z≥0(−1, −1) ⊕ Z≥0(1,−1) ⊕ Z≥0(−1, 1),
Sτ4 = Sσ4+ Sσ5 = Z≥0(0,−1) ⊕ Z≥0(1, 0)⊕ Z≥0(−1, 0),
Sτ5 = Sσ5 + Sσ1 = Z≥0(1, 0)⊕ Z≥0(0, 1)⊕ Z≥0(0,−1),
S{0}= Z≥0(1, 0)⊕ Z≥0(0, 1)⊕ Z≥0(−1, 0) ⊕ Z≥0(0,−1).
Therefore, the affine toric variety
Uσ1 = hom(Sσ1, T) = T2, Uσ2 = hom(Sσ2, T) = T2,
Uσ3 = hom(Sσ3, T) = T2, Uσ4 = hom(Sσ4, T) = T2,
Uσ5 = hom(Sσ5, T) = T2,
together with
Uτ1 = hom(Sτ1, T) = R× T, Uτ2 = hom(Sτ2, T) = R× T,
Uτ3 = hom(Sτ3, T) = R× T, Uτ4 = hom(Sτ4, T) = R× T,
Uτ5 = hom(Sτ5, T) = R× T, U{0} = hom(S{0}, T) = R2.
The gluing of the affine toric varieties Uσ2 and Uσ3 along their common subset Uτ2 gives TP1×T with coordinates ([x0 : x1],−x+y) where x = x0−x1. The gluing of the affine toric varieties Uσ3 and Uσ4 along their common subset Uτ3 gives TP1× T with coordinates (−x, [z0 : z1]) where −x + y = z1 − z0. The two copies of TP1×T are glued along their coordinates gives TP1×TP1 with coordinates ([x0 : x1], [z0 : z1]). Since we have some embedding h44 : Uτ4 �→ Uσ4 via h44(x0 − x1, x1− x0, x1− x0+ z0− z1) = (0, x1− x0, z0− z1)
and h45 : Uτ4 �→ Uσ5 via h45(x0 − x1, x1 − x0, x1 − x0 + z0 − z1) = (x0 −
Figure 4.7: the fan ∆
Figure 4.8: the polytope P
conv{(−1, −1), (0, −1)}, conv{(0, −1), (1, 0)} are the facets of P , and they are the convex hull of a basis of N , XP is a smooth Fano polytope (by Theorem 3.4.2).
Example 4.3.5. Given the lattice N � Z2, then NR = N ⊗ R � R2, the dual lattice M � Z2 and MR = M⊗ R.
Let the fan ∆ in NR. Suppose that the fan ∆ (the figure 4.9) has
σ1 = pos{(1, 0), (1, 1)}, σ2 = pos{(1, 1), (0, 1)},
σ3 = pos{(0, 1), (−1, 0)}, σ4 = pos{(−1, 0), (−1, −1)},
σ5 = pos{(−1, −1), (0, −1)}, σ6 = pos{(0, −1), (1, 0)},
Figure 4.9: the fan ∆
Figure 4.10: the polytope P
together with
τ1 = σ1∩ σ2 = pos{(1, 1)}, τ2 = σ2 ∩ σ3 = pos{(0, 1)},
τ3 = σ3∩ σ4 = pos{(−1, 0)}, τ4 = σ4∩ σ5 = pos{(−1, −1)},
τ5 = σ5∩ σ6 = pos{(0, −1)}, τ6 = σ6 ∩ σ1 = pos{(1, 0)},
and the origin. Then the dual cones
σ∨1 = pos{(−1, −1), (0, 1)}, σ2∨ = pos{(1, 0), (−1, 1)},
σ∨3 = pos{(−1, 0), (0, 1)}, σ4∨ = pos{(−1, 1), (0, −1)},
σ∨5 = pos{(−1, 0), (1, −1)}, σ6∨ = pos{(0, −1), (1, 0)}.
Moreover, the corresponding semigroups
Sσ1 = σ1∨∩ M = Z≥0(0, 1)⊕ Z≥0(1,−1),
Sσ2 = σ2∨∩ M = Z≥0(1, 0)⊕ Z≥0(−1, 1),
Sσ3 = σ3∨∩ M = Z≥0(−1, 0) ⊕ Z≥0(0, 1),
Sσ4 = σ4∨∩ M = Z≥0(−1, 1) ⊕ Z≥0(0,−1),
Sσ5 = σ5∨∩ M = Z≥0(−1, 0) ⊕ Z≥0(1,−1),
Sσ6 = σ6∨∩ M = Z≥0(0,−1) ⊕ Z≥0(1, 0),
together with
Sτ1 = Sσ1+ Sσ2 = Z≥0(1, 1)⊕ Z≥0(1,−1) ⊕ Z≥0(−1, 1),
Sτ2 = Sσ2 + Sσ3 = Z≥0(0, 1)⊕ Z≥0(1, 0)⊕ Z≥0(−1, 0),
Sτ3 = Sσ3+ Sσ4 = Z≥0(−1, 0) ⊕ Z≥0(0, 1)⊕ Z≥0(0,−1),
Sτ4 = Sσ4 + Sσ5 = Z≥0(−1, −1) ⊕ Z≥0(1,−1) ⊕ Z≥0(−1, 1),
Sτ5 = Sσ5+ Sσ6 = Z≥0(0,−1) ⊕ Z≥0(1, 0)⊕ Z≥0(−1, 0),
Sτ6 = Sσ6 + Sσ1 = Z≥0(1, 0)⊕ Z≥0(0, 1)⊕ Z≥0(0,−1),
S{0}= Z≥0(1, 0)⊕ Z≥0(0, 1)⊕ Z≥0(−1, 0) ⊕ Z≥0(0,−1).
Therefore, the affine toric variety
Uσ1 = hom(Sσ1, T) = T2, Uσ2 = hom(Sσ2, T) = T2,
Uσ3 = hom(Sσ3, T) = T2, Uσ4 = hom(Sσ4, T) = T2,
Uσ5 = hom(Sσ5, T) = T2, Uσ6 = hom(Sσ6, T) = T2,
together with
Uτ1 = hom(Sτ1, T) = R× T, Uτ2 = hom(Sτ2, T) = R× T,
Uτ3 = hom(Sτ3, T) = R× T, Uτ4 = hom(Sτ4, T) = R× T,
Uτ5 = hom(Sτ5, T) = R× T, Uτ6 = hom(Sτ6, T) = R× T,
U{0}= hom(S{0}, T) = R2.
Let fi be in Uσi = hom(Sσi, T) for all i = 1, 2, 3, 4, 5, 6, then we have
some maps
f1 : Sσ1 → T via f1(0, 1) = y and f1(1,−1) = x − y,
f2 : Sσ2 → T via f2(1, 0) = x and f2(−1, 1) = −x + y,
f3 : Sσ3 → T via f3(−1, 0) = −x and f3(0, 1) = y,
f4 : Sσ4 → T via f4(0,−1) = −y and f4(1,−1) = −x + y,
f5 : Sσ4 → T via f4(−1, 0) = −x and f4(1,−1) = x − y,
f5 : Sσ4 → T via f4(0,−1) = −y and f4(1, 0) = x.
Since we have some embedding h11 : Uτ1 �→ Uσ1 via h11(x + y, x − y,−x + y) = (0, x − y, y) and h12 : Uτ1 �→ Uσ2 via h12(x + y, x− y, −x + y) = (x,−x + y, 0), we have tropical isomorphism h12 ◦ h−111 : Uσ1 → Uσ2 via h12 ◦ h−111(0, x − y, y) = (x, −x + y, 0). Similarly, h23 ◦ h−122 : Uσ2 → Uσ3
via h23 ◦ h−122(x,−x + y, 0) = (−x, 0, y), and h34 ◦ h−133 : Uσ3 → Uσ4 via h34◦h−133(−x, 0, y) = (0, −x+y, −y), h45◦h−144 : Uσ4 → Uσ5 via h45◦h−144(0,−x+
y,−y) = (−x, x − y, 0), h56◦ h−155 : Uσ5 → Uσ6 via h56◦ h−155(−x, x − y, 0) = (x, 0,−y), h61◦h−166 : Uσ6 → Uσ1 via h61◦h−166(x, 0,−y) = (0, x−y, y). Consider the product TP1×TP1×TP1×TP2with homogeneous coordinates [x0 : x1],
[z0 : z1],[y0 : y1] on respective TP1 and homogeneous coordinate [a : b : c] on TP2. We set x = x0 − x1, x− y = z0 − z1, y = y0 − y1. Hence the toric variety
X∆(T) = (
�
σ∈∆Uσ)�
∼
={([x0 : x1], [z0 : z1], [y0, y1], [a : b : c])∈ TP1× TP1× TP1
× TP2 | b + x1 = c + x0, a + z1 = b + z0, c + y1 = a + y0},
that is, X∆(T) is the blow up of TP2 at the three points [0 : −∞ : −∞], [−∞ : 0 : −∞], and [−∞ : −∞ : 0].
The polytopes P = conv{0, (1, 0), (1, 1), (0, 1), (0, −1), (−1, −1), (0, −1)}
(the figure 4.10 ) in NR. Since conv{(1, 0), (1, 1)}, conv{(0, 1), (1, 1)}, conv{(0, 1), (−1, 0)}, conv{(−1, 0), (−1, −1)}, conv{(−1, −1), (0, −1)}, conv{(0, −1), (1, 0)} are the facets of P , and they are the convex hull of a basis of N , XP is a smooth Fano polytope (by Theorem 3.4.2).
Bibliography
[1] Danko Adrovic and Jan Verschelde. Tropical algebraic geometry in maple, a preprocessing algorithm for finding common factors to mul-tivariate polynomials with approximate coefficients. September 2008.
[2] T. Bogart, A. N. Jensen, D. Speyer, B. Sturmfels, and R. R. Thomas.
Computing tropical varieties. J. Symbolic Comput., 42(1-2):54–73, 2007.
[3] E. Brugall´e. Deformation of tropical hirzebruch surfaces. http:
//www.math.jussieu.fr/~brugalle/articles/TropDegHirzebruch/
TropDeg.pdf.
[4] W. Bruns and J. Gubeladze. Polytopes, Rings, and K-Theory. Number 978 0-76352 in Springer Monographs in Mathematics. Springer, 2009.
[5] D.A. Cox, J.B. Little, and H.K. Schenck. Toric Varieties. Graduate Studies in Mathematics. American Mathematical Society, 2011.
[6] David Eisenbud and Joe Harris. The Geometry of Schemes. Springer, November 2001.
[7] G¨unter Ewald. Combinatorial Convexity and Algebraic Geometry (Grad-uate Texts in Mathematics). Springer, October 1996.
[8] William Fulton. Introduction to toric varieties, volume 131 of Annals of Mathematics Studies. Princeton, NJ, 1993.
[9] Andreas Gathmann. Algebraic geometry, 2002.
[10] Andreas Gathmann. Tropical algebraic geometry. Jahresber. Deutsch.
Math.-Verein., 108(1):3–32, 2006.
[11] I.M. Gelfand, M.M. Kapranov, and A.V. Zelevinsky. Discriminants, Resultants, and Multidimensional Determinants. Modern Birkh¨auser Classics. Birkh¨auser, 2008.
[12] Branko Gr¨unbaum. Convex Polytopes : Second Edition Prepared by Volker Kaibel, Victor Klee, and G¨unter Ziegler (Graduate Texts in Math-ematics). Springer, May 2003.
[13] Ilia Itenberg, Grigory Mikhalkin, and Eugenii Shustin. Tropical algebraic geometry, volume 35 of Oberwolfach Seminars. Birkh¨auser Verlag, Basel, second edition, 2009.
[14] A Kasprzyk. Toric Fano Varieties and Convex Polytopes. PhD thesis, 2006.
[15] Diane Maclagan and Bernd Sturmfels. Introduction to Tropical Geome-try. October 9, 2009.
[16] G.G. Magaril-Ilyaev and V.M. Tikhomirov. Convex Analysis: Theory and Applications. Translations of Mathematical Monographs. American Mathematical Society, 2003.
[17] H. Meyer. Intersection Theory on Compact Tropical Toric Varieties.
Sudwestdeutscher Verlag F R Hochschulschriften AG, 2011.
[18] G. Mikhalkin. Amoebas of algebraic varieties and tropical geometry.
ArXiv Mathematics e-prints, February 2004.
[19] Grigory Mikhalkin. Enumerative tropical algebraic geometry in R2. J.
Amer. Math. Soc., 18(2):313–377 (electronic), 2005.
[20] Grigory Mikhalkin. Tropical geometry and its applications. In Interna-tional Congress of Mathematicians. Vol. II, pages 827–852. Eur. Math.
Soc., Z¨urich, 2006.
[21] Grigory Mikhalkin and Ilia Zharkov. Tropical curves, their jacobians and theta functions, November 2007.
[22] T. Oda. Convex bodies and algebraic geometry: an introduction to the theory of toric varieties : with 42 figures. Ergebnisse der Mathematik und ihrer Grenzgebiete. Springer-Verlag, 1988.
[23] J¨urgen Richter-Gebert, Bernd Sturmfels, and Thorsten Theobald. First steps in tropical geometry. In Idempotent mathematics and mathematical physics, volume 377 of Contemp. Math., pages 289–317. Amer. Math.
Soc., Providence, RI, 2005.
[24] R. Tyrrell Rockafellar. Convex Analysis. Princeton University Press, 1970.
[25] K.E. Smith. An Invitation to Algebraic Geometry. Universitext (1979).
Springer, 2000.
[26] David Speyer and Bernd Sturmfels. Tropical mathematics. Math. Mag., 82(3):163–173, 2009.
[27] Thorsten Theobald and Thorsten Theobald. Computing amoebas. Ex-perimental Math, 11:513–526, 2002.
[28] Gunter M. Ziegler. Lectures on Polytopes (Graduate Texts in Mathe-matics). Springer, November 1994.