04.12.04 · algebraic-geometry / tropical

Newton polytope and non-archimedean amoeba

shipped3 tiersLean: partial

Anchor (Master): Gelfand-Kapranov-Zelevinsky 1994 Ch. 6 + Ch. 11; Mikhalkin 2004 *Decomposition into pairs-of-pants for complex algebraic hypersurfaces* (*Topology* 43, 1035-1065); Einsiedler-Kapranov-Lind 2006 *Non-archimedean amoebas and tropical varieties* (*J. reine angew. Math.* 601); Forsberg-Passare-Tsikh 2000 *Laurent determinants and arrangements of hyperplane amoebas* (*Adv. Math.* 151); Maclagan-Sturmfels §1.4 and Ch. 3

Intuition [Beginner]

Take a Laurent polynomial in two variables, like $f (x, y) = 1 + x + y$ . Write down the exponent vectors of its monomials: $(0, 0)$ for the constant $1$ , $(1, 0)$ for $x$ , and $(0, 1)$ for $y$ . Draw those three points in the plane. The convex hull of those points is a small triangle. That triangle is the Newton polytope of $f$ , and it is a compact picture of where the monomials of $f$ live in exponent space.

Now ask a different question. The polynomial $f$ vanishes on a curve $V (f)$ inside the two-dimensional complex torus $(C^{*})^{2}$ . Project that curve down to the real plane by taking the log of the absolute value of each coordinate: $(z_{1}, z_{2}) \mapsto (lo g ∣ z_{1} ∣, lo g ∣ z_{2} ∣)$ . The shape you see is called the amoeba of $f$ . It looks like a fat blob with finitely many tentacles sticking out toward infinity. The tentacles, remarkably, point in directions controlled by the Newton polytope: each tentacle is perpendicular to one of the edges of the triangle.

There is a third gadget in the same family. If we replace the complex numbers with a non-archimedean field — fields with a valuation map sending nonzero elements to a real number — the same projection produces the non-archimedean amoeba. By a theorem of Kapranov, this is exactly the tropical hypersurface of $f$ : a piecewise-linear skeleton in the real plane. As one rescales the complex amoeba, it degenerates to the tropical hypersurface in a precise sense (Mikhalkin 2004). The three objects — Newton polytope on the exponent side, complex amoeba on the analytic side, and tropical hypersurface on the combinatorial side — are different facets of the same combinatorial-analytic dictionary.

Visual [Beginner]

A three-panel sketch. Left panel: the exponent plane $M_{R} = R^{2}$ with the three lattice points $(0, 0), (1, 0), (0, 1)$ marked and the triangle they enclose drawn as the Newton polytope of $f = 1 + x + y$ . Centre panel: the projection plane $N_{R} = R^{2}$ with the amoeba of $f$ drawn as a fat "Y" shape — a central body with three tentacles. Right panel: the same plane $N_{R}$ with the tropical hypersurface of $f$ drawn as a precise "Y" — three rays meeting at the origin, in the directions $(- 1, 0), (0, - 1)$ , and $(1, 1)$ .

Arrows between the panels: a dashed double-arrow connects each edge of the Newton triangle to the corresponding tentacle direction in the amoeba and to the corresponding ray of the tropical hypersurface.

The picture captures the central fact: the three objects share the same asymptotic data, and the rays of the tropical hypersurface are exactly the limits of the tentacles of the amoeba after rescaling.

Worked example [Beginner]

Take $f (x, y) = 1 + x + y$ , the simplest non-monomial Laurent polynomial in two variables. Compute the Newton polytope, the tentacle directions of the amoeba, and the tropical hypersurface, and verify the three pieces of data align.

Step 1. Identify the monomials. The polynomial $f$ has three monomials with nonzero coefficients: $1 \cdot x^{0} y^{0}$ , $1 \cdot x^{1} y^{0}$ , $1 \cdot x^{0} y^{1}$ . The exponent vectors are $α_{1} = (0, 0)$ , $α_{2} = (1, 0)$ , $α_{3} = (0, 1)$ .

Step 2. Form the Newton polytope. $Newt (f) = conv {(0, 0), (1, 0), (0, 1)}$ , the closed triangle in $M_{R} = R^{2}$ with vertices at the origin, $(1, 0)$ , and $(0, 1)$ . Its three edges are: the bottom edge from $(0, 0)$ to $(1, 0)$ , the left edge from $(0, 0)$ to $(0, 1)$ , and the hypotenuse from $(1, 0)$ to $(0, 1)$ . Compute the inner normal to each edge. The bottom edge has inner normal $(0, 1)$ ; the left edge has inner normal $(1, 0)$ ; the hypotenuse has inner normal $(- 1, - 1)$ (pointing into the triangle).

Step 3. Predict the tentacle directions. Each facet normal becomes an asymptotic direction for the amoeba. The outer normals — the negatives of the inner normals — are $(0, - 1), (- 1, 0), (1, 1)$ . These are the three tentacle directions of the amoeba of $f$ .

Step 4. Verify the tropical hypersurface. The tropicalisation of $f$ with vanishing valuation is $T (f) (u_{1}, u_{2}) = min (0, u_{1}, u_{2})$ . The tropical hypersurface is the set where the minimum is attained at least twice.

Set $u_{1} = u_{2}$ and observe: if $u_{1} = u_{2} \leq 0$ , then $u_{1} = u_{2} \leq 0$ are tied minima — so the ray ${u_{1} = u_{2} \leq 0}$ is on the tropical hypersurface; if $u_{1} \leq u_{2}$ and $u_{1} \leq 0$ , then $u_{1} = 0$ ties only if $u_{1} = 0$ , so the ray ${u_{1} = 0, u_{2} \geq 0}$ is on it; similarly ${u_{2} = 0, u_{1} \geq 0}$ . The three rays meeting at the origin point in directions $(- 1, - 1)$ , $(1, 0)$ , and $(0, 1)$ . These match the inner normals to the three edges of $Newt (f)$ .

What this tells us. The three asymptotic directions of the amoeba of $f = 1 + x + y$ are the inner normals to the edges of the Newton triangle, equivalently the three rays of the tropical hypersurface of $f$ . The same combinatorial datum — the normal fan of $Newt (f)$ — controls the tentacle directions of the complex amoeba and the structure of the tropical hypersurface. The amoeba is a thickening of the tropical hypersurface, and the tropical hypersurface is the rescaled limit of the amoeba; both inherit their asymptotic geometry from the Newton polytope.

Check your understanding [Beginner]

Exercise (easy, multiple choice).

The Newton polytope of a Laurent polynomial $f$ (a finite combination of monomials $c_{α} x^{α}$ ) is defined as:

A. The convex hull of the exponent vectors $α$ with $c_{α} \neq = 0$ .
B. The set of all integer combinations of the exponent vectors $α$ .
C. The image of $f$ as a function $(C^{*})^{n} \to C$ .
D. The vanishing locus $V (f) \subset (C^{*})^{n}$ .

Hint

The Newton polytope is a compact convex shape in the exponent space $M_{R}$ — it lives where the exponents live, not where the variables live.

Answer

A. The Newton polytope is by definition the convex hull, in $M_{R} = R^{n}$ , of the exponent vectors of the monomials of $f$ that have nonzero coefficients. Option B gives a sublattice of $M$ , not a polytope. Option C is the image set of the function $f$ , a subset of $C$ . Option D is the zero set of $f$ , a subvariety of the torus. The Newton polytope is a single compact subset of the exponent space, encoding the support of $f$ in the most economical convex-geometric form.

Exercise (easy, true-false).

True or false: the complex amoeba of a Laurent polynomial $f$ has finitely many tentacles, and the directions of those tentacles are controlled by the Newton polytope of $f$ .

Hint

A tentacle of the amoeba is an unbounded direction in which the amoeba extends. The Newton polytope has finitely many facets, and each facet has one outer normal direction.

Answer

True. Every tentacle of the complex amoeba of $f$ extends in a direction equal to an outer normal of some facet of the Newton polytope of $f$ . The Newton polytope has finitely many facets, hence the amoeba has finitely many tentacle directions. Conversely (Forsberg-Passare-Tsikh 2000), for a generic $f$ , every outer normal of every facet of the Newton polytope is realised as a tentacle direction. The combinatorial data of the Newton polytope completely determines the asymptotic structure of the amoeba.

Formal definition [Intermediate+]

Fix a free abelian group $N$ of rank $n$ (the cocharacter lattice of the algebraic torus $T = (G_{m})^{n}$ ) and its dual $M = Hom_{Z} (N, Z)$ (the character lattice), with integer pairing $⟨ -, - ⟩ : M \times N \to Z$ and real forms $N_{R} = N \otimes_{Z} R$ and $M_{R} = M \otimes_{Z} R$ . A Laurent polynomial over a field $K$ in $n$ variables is a finite formal sum $$ f = \sum_{\alpha \in \mathrm{supp}(f)} c_\alpha , x^\alpha $$ with $supp (f) \subseteq M$ finite, $c_{α} \in K^{*}$ for $α \in supp (f)$ , and $x^{α} := x_{1}^{α_{1}} \dots x_{n}^{α_{n}}$ the multi-index monomial. The Laurent polynomial $f$ defines a regular function on the algebraic torus $T = Spec K [M]$ .

Definition (Newton polytope). The Newton polytope of $f$ is $$ \mathrm{Newt}(f) := \mathrm{conv}\big(\mathrm{supp}(f)\big) \subseteq M_\mathbb{R}. $$ It is a lattice polytope: a compact convex polyhedron whose vertices lie in $M \subset M_{R}$ .

Definition (complex amoeba). Take $K = C$ and the coordinate-wise log-absolute-value map $$ \mathrm{Log} : (\mathbb{C}^)^n \to N_\mathbb{R} = \mathbb{R}^n, \qquad (z_1, \ldots, z_n) \mapsto (\log|z_1|, \ldots, \log|z_n|). $$ The amoeba of $f$ is $$ \mathcal{A}(f) := \mathrm{Log}\big(V(f)\big) \subseteq N_\mathbb{R}, $$ where $V(f) = {z \in (\mathbb{C}^)^n : f(z) = 0}$ is the vanishing locus.

Definition (tropicalisation; tropical hypersurface). For $K$ a field equipped with a valuation $v : K^{*} \to R$ (a surjective group homomorphism with the non-archimedean inequality $v (a + b) \geq min (v (a), v (b))$ when $a + b \neq = 0$ ), the tropicalisation of $f = \sum c_{α} x^{α}$ is the piecewise-linear function $$ T(f) : N_\mathbb{R} \to \mathbb{R}, \qquad u \mapsto \min_{\alpha \in \mathrm{supp}(f)} \big(v(c_\alpha) + \langle \alpha, u\rangle\big). $$ The tropical hypersurface of $f$ is $$ \mathrm{trop}(f) := \big{u \in N_\mathbb{R} : \text{the minimum in } T(f)(u) \text{ is attained by at least two } \alpha\big}. $$ Equivalently, $trop (f)$ is the locus where $T (f)$ fails to be smooth: the corner locus of the piecewise-linear function $T (f)$ .

Definition (non-archimedean amoeba). For $K$ non-archimedean with valuation $v$ , the non-archimedean amoeba of $V (f) \subset (K^{*})^{n}$ is the image under coordinate-wise valuation: $$ \mathcal{A}v(f) := \mathrm{val}\big(V(f)\big) \subseteq N\mathbb{R}, \qquad \mathrm{val}(z_1, \ldots, z_n) = (v(z_1), \ldots, v(z_n)). $$

Counterexamples to common slips [Intermediate+]

The Newton polytope is not the support. The support $supp (f) \subset M$ is a finite set of lattice points; the Newton polytope $Newt (f)$ is its convex hull, a full-dimensional or lower-dimensional compact polytope. The two objects determine each other only up to the convex-hull operation: many supports produce the same Newton polytope, but the polytope alone forgets the interior lattice points of the support.
The amoeba is not all of $N_{R}$ . It might be tempting to think that the projection of an $n - 1$ -dimensional hypersurface $V (f) \subset (C^{*})^{n}$ to $N_{R} = R^{n}$ fills $N_{R}$ . It does not: $A (f)$ has finitely many connected components in its complement (Forsberg-Passare-Tsikh 2000), each labelled by a lattice point of $Newt (f) \cap M$ via the order map.
Tropicalisation depends on the valuation. The tropical hypersurface $trop (f)$ is not an invariant of $f$ alone — it depends on the choice of valuation on the coefficient field. For the identically zero valuation $v \equiv 0$ , the corner locus of $T (f) (u) = min_{α} ⟨ α, u ⟩$ is exactly the normal fan of $Newt (f)$ minus its top-dimensional cones. For a non-vanishing valuation, the corner locus shifts to reflect the valuation data.
Spine convergence is not pointwise convergence. Mikhalkin's spine theorem $A (f^{t}) / lo g ∣ t ∣ \to trop (f)$ is convergence in the Hausdorff metric on closed subsets of bounded regions of $N_{R}$ , not pointwise convergence of points along the amoeba. Pointwise statements fail because the amoeba thickens continuously as $t$ varies, while the tropical hypersurface remains a measure-zero piecewise-linear set.

Key theorem with proof [Intermediate+]

The signature theorem of this unit is Kapranov's identification of the non-archimedean amoeba with the tropical hypersurface: for a non-archimedean field $K$ with surjective valuation, the non-archimedean amoeba of $V (f)$ is exactly the tropical hypersurface of $f$ . This is the cleanest version of the complex-to-tropical bridge, because it holds as a precise equality rather than as a limit; the analytic Mikhalkin theorem then witnesses the same identification as a degeneration of the complex amoeba.

Theorem (Kapranov; Einsiedler-Kapranov-Lind 2006). Let $K$ be an algebraically closed field equipped with a non-degenerate real-valued non-archimedean valuation $v : K^ \to \mathbb{R} $, an d l e t$ f = \sum_{\alpha \in \mathrm{supp}(f)} c_\alpha x^\alpha $b e a L a u r e n tp o l y n o mia l o v er$ K$. Then* $$ \mathcal{A}v(f) = \mathrm{trop}(f) \subseteq N\mathbb{R}. $$ That is, the non-archimedean amoeba of $V (f)$ equals the tropical hypersurface of $f$ .

Proof. The argument has two directions: $A_{v} (f) \subseteq trop (f)$ (analytic direction, valuation inequality) and $trop (f) \subseteq A_{v} (f)$ (algebraic direction, root-finding).

Direction 1: $A_{v} (f) \subseteq trop (f)$ . Take $z = (z_{1}, \dots, z_{n}) \in V (f) \subset (K^{*})^{n}$ and let $u_{i} = v (z_{i})$ for each $i$ , so $u = (u_{1}, \dots, u_{n}) \in N_{R}$ is the image $val (z) \in A_{v} (f)$ . Apply $v$ to $f (z) = \sum_{α} c_{α} z^{α} = 0$ . The valuation of $c_{α} z^{α}$ is $v (c_{α}) + ⟨ α, u ⟩$ . By the non-archimedean inequality applied to the sum $\sum_{α} c_{α} z^{α} = 0$ : if the minimum $min_{α} (v (c_{α}) + ⟨ α, u ⟩)$ were attained by a unique index $α_{0}$ , then $v (\sum c_{α} z^{α}) = v (c_{α_{0}} z^{α_{0}}) = v (c_{α_{0}}) + ⟨ α_{0}, u ⟩$ , which is finite — contradicting $f (z) = 0$ (which would force the valuation to be $+ \infty$ ). Hence the minimum is attained by at least two indices, i.e., $u \in trop (f)$ .

Direction 2: $trop (f) \subseteq A_{v} (f)$ . Take $u \in trop (f)$ : at least two indices $α, β \in supp (f)$ both attain the minimum $m := v (c_{α}) + ⟨ α, u ⟩ = v (c_{β}) + ⟨ β, u ⟩ = min_{γ} (v (c_{γ}) + ⟨ γ, u ⟩)$ . The aim is to find $z \in (K^{*})^{n}$ with $f (z) = 0$ and $val (z) = u$ .

Since $K$ is algebraically closed with surjective valuation, choose any lift $t = (t_{1}, \dots, t_{n}) \in (K^{*})^{n}$ with $v (t_{i}) = u_{i}$ — such a lift exists by surjectivity of $v$ . Consider the change of variables $z_{i} = t_{i} w_{i}$ , so $w_{i} \in K^{*}$ are new variables with $v (w_{i}) = 0$ , i.e., $w_{i}$ lives in the residue field $O / m$ where $O = {x : v (x) \geq 0}$ is the valuation ring and $m = {x : v (x) > 0}$ its maximal ideal.

The condition $f (z) = 0$ becomes, after multiplying through by a uniformiser to clear denominators, $$ \sum_\alpha c_\alpha t^\alpha w^\alpha = 0, \qquad v(c_\alpha t^\alpha) = v(c_\alpha) + \langle \alpha, u\rangle \geq m, $$ with equality precisely for the indices $α$ attaining the minimum. Divide by a uniformiser of valuation $m$ to bring all terms into $O$ , and reduce modulo $m$ : the equation becomes the initial polynomial $$ \mathrm{in}u(f)(w) := \sum{\alpha ,:, v(c_\alpha) + \langle \alpha, u\rangle ,=, m} \overline{c_\alpha t^\alpha / \pi^m} \cdot w^\alpha $$ in the residue field $k = O / m$ , where $π$ is a uniformiser.

Since the minimum is attained by at least two $α$ , the initial polynomial $in_{u} (f)$ has at least two monomials, hence is non-constant in at least one variable $w_{i}$ . By algebraic closure of $K$ (and hence of the residue field $k$ for the standard valued fields), $in_{u} (f)$ has a root in $(k^{*})^{n}$ — specifically, choose generic values for $n - 1$ of the variables and solve for the remaining variable using algebraic closure of $k$ . Lift any such root $\overset{w}{ˉ} \in (k^{*})^{n}$ to $w \in (O^{*})^{n}$ by Hensel's lemma applied to the corresponding factor of $f$ in the variable being solved. The resulting $z = (t_{1} w_{1}, \dots, t_{n} w_{n})$ satisfies $f (z) = 0$ and $val (z) = u$ , exhibiting $u \in A_{v} (f)$ . $□$

The Kapranov theorem is the foundational identification between the algebraic non-archimedean amoeba and the combinatorial tropical hypersurface. Its proof uses the residue-field reduction $in_{u} (f)$ , which is the initial form of $f$ at $u$ and is the cornerstone of all subsequent tropical-geometric arguments (Mikhalkin's correspondence theorem 04.12.05, the Nishinou-Siebert correspondence 04.12.06, and the Gross-Siebert reconstruction 04.12.09 all use initial forms as the bridge between the analytic and the combinatorial sides).

Bridge. The Kapranov theorem builds toward 04.12.05 Mikhalkin's correspondence theorem and appears again in 04.12.06 Nishinou-Siebert correspondence, where the identification of analytic and tropical objects is upgraded from hypersurfaces to enumeration of algebraic curves. The foundational reason is that the initial form $in_{u} (f)$ is exactly the object that controls the local geometry of $V (f)$ near a point of valuation $u$ : when $in_{u} (f)$ is non-constant in at least one variable, $V (f)$ has a point of that valuation. This is exactly the bridge from the analytic vanishing locus to the combinatorial corner locus of $T (f)$ . Putting these together with the normal-fan structure of $Newt (f)$ , the cone formalism of 04.11.02 (rational polyhedral cones and dual cones) supplies the combinatorial language in which $trop (f)$ lives as a union of polyhedra, with each polyhedron dual to a face of $Newt (f)$ . The bridge is between the analytic structure of $V (f)$ and the combinatorial structure of $Newt (f)$ , identifying both with the tropical hypersurface $trop (f)$ as the common shadow. This pattern generalises to higher-codimensional subvarieties via the Sturmfels-Tevelev theorem and the Bogart-Jensen-Speyer-Sturmfels-Thomas tropical-basis algorithm.

Exercises [Intermediate+]

Exercise 1 (easy, symbolic).

Let $f (x, y) = 1 + x + y + x y + x^{2} + y^{2}$ in two variables. Compute the Newton polytope $Newt (f) \subset R^{2}$ , identify its vertices and edges, and predict the tentacle directions of the complex amoeba $A (f)$ .

Hint

The exponents are six lattice points. Find the convex hull; some of the six may be interior. The tentacle directions are the outer normals of the edges of the convex hull.

Answer

Exponents: $(0, 0), (1, 0), (0, 1), (1, 1), (2, 0), (0, 2)$ — six lattice points.

Convex hull: Plot the six points. The outermost points are $(0, 0), (2, 0), (1, 1), (0, 2)$ — wait, check: $(1, 1)$ is in the line $x + y = 2$ joining $(2, 0)$ and $(0, 2)$ , so $(1, 1)$ lies on the hypotenuse, not at a vertex. The vertices are $(0, 0), (2, 0), (0, 2)$ . The lattice points $(1, 0), (0, 1), (1, 1)$ are interior or boundary lattice points.

Newton polytope: $Newt (f) = conv {(0, 0), (2, 0), (0, 2)}$ , the triangle with these three vertices. Three edges:

bottom edge from $(0, 0)$ to $(2, 0)$ ;
left edge from $(0, 0)$ to $(0, 2)$ ;
hypotenuse from $(2, 0)$ to $(0, 2)$ along the line $x + y = 2$ .

Outer normals (tentacle directions):

bottom edge: outer normal $(0, - 1)$ ;
left edge: outer normal $(- 1, 0)$ ;
hypotenuse: outer normal $(1, 1) / 2$ , primitive integer form $(1, 1)$ .

So the amoeba $A (f)$ has three tentacles, pointing in the directions $(0, - 1), (- 1, 0), (1, 1)$ .

Rubric: full credit for the three vertices, three edges, and three outer normals.

Exercise 2 (easy, multiple choice).

For the Laurent polynomial $f (x, y) = x^{- 1} + 1 + x + y + y^{- 1}$ , the Newton polytope is:

A. The hexagon with vertices $(\pm 1, 0), (0, \pm 1), (\pm 1, \pm 1)$ .
B. The rhombus with vertices $(- 1, 0), (1, 0), (0, - 1), (0, 1)$ .
C. The triangle with vertices $(- 1, 0), (1, 0), (0, 1)$ .
D. The square $[- 1, 1]^{2}$ .

Hint

The exponents are $(- 1, 0), (0, 0), (1, 0), (0, 1), (0, - 1)$ . Find their convex hull.

Answer

B. The exponents are the five lattice points $(- 1, 0), (0, 0), (1, 0), (0, 1), (0, - 1)$ . The convex hull is the rhombus (or "diamond") with the four outermost points as vertices: $(- 1, 0), (1, 0), (0, - 1), (0, 1)$ . The origin $(0, 0)$ is an interior lattice point. Option A (hexagon) requires six vertices; we have only four extreme points. Option C is the wrong subset of vertices. Option D (square $[- 1, 1]^{2}$ ) is too large — it would require the four "corner" points $(\pm 1, \pm 1)$ to be in the support, which they are not.

Exercise 3 (medium, symbolic).

Compute the tropical hypersurface $trop (f) \subset R^{2}$ for $f (x, y) = 1 + x + y + x y$ (vanishing valuation). Describe it as a union of polyhedral pieces and verify that each piece is dual to a face of $Newt (f)$ .

Hint

$T (f) (u_{1}, u_{2}) = min (0, u_{1}, u_{2}, u_{1} + u_{2})$ (since the vanishing valuation gives $v (c_{α}) = 0$ for all $α$ ). The tropical hypersurface is the set of $(u_{1}, u_{2})$ where the minimum is attained twice. Partition $R^{2}$ by which monomial achieves the minimum.

Answer

Step 1: tropicalisation. With vanishing valuation, $T (f) (u_{1}, u_{2}) = min (0, u_{1}, u_{2}, u_{1} + u_{2})$ .

Step 2: region partition. The plane $R^{2}$ splits into four regions according to which monomial attains the minimum:

region $R_{(0, 0)}$ : where $0$ is the minimum — that is, $u_{1} \geq 0, u_{2} \geq 0, u_{1} + u_{2} \geq 0$ , the closed first quadrant.
region $R_{(1, 0)}$ : where $u_{1}$ is the minimum — that is, $u_{1} \leq 0, u_{1} \leq u_{2}, u_{1} \leq u_{1} + u_{2}$ , equivalently $u_{1} \leq 0, u_{2} \geq u_{1}, u_{2} \geq 0$ . Restricting: $u_{1} \leq 0, u_{2} \geq 0$ . (The third quadrant where $u_{2} < 0$ is excluded.)
region $R_{(0, 1)}$ : by symmetry, $u_{2} \leq 0, u_{1} \geq 0$ .
region $R_{(1, 1)}$ : where $u_{1} + u_{2}$ is the minimum, i.e., the third quadrant $u_{1} \leq 0, u_{2} \leq 0$ .

Step 3: tropical hypersurface. $trop (f)$ is the union of the boundaries between these four regions:

$R_{(0, 0)} \cap R_{(1, 0)}$ : $u_{1} = 0, u_{2} \geq 0$ — positive $u_{2}$ -axis.
$R_{(0, 0)} \cap R_{(0, 1)}$ : $u_{2} = 0, u_{1} \geq 0$ — positive $u_{1}$ -axis.
$R_{(1, 0)} \cap R_{(1, 1)}$ : $u_{2} = 0, u_{1} \leq 0$ — negative $u_{1}$ -axis.
$R_{(0, 1)} \cap R_{(1, 1)}$ : $u_{1} = 0, u_{2} \leq 0$ — negative $u_{2}$ -axis.

The tropical hypersurface is the coordinate-axis cross in $R^{2}$ : four rays emanating from the origin in the four cardinal directions.

Step 4: duality with $Newt (f)$ . $Newt (f)$ is the unit square with vertices $(0, 0), (1, 0), (0, 1), (1, 1)$ . Each edge of the square has an outer normal: bottom edge normal $(0, - 1)$ , top edge normal $(0, 1)$ , left edge normal $(- 1, 0)$ , right edge normal $(1, 0)$ . These are exactly the four ray directions of $trop (f)$ . The duality is the normal-fan correspondence: each edge of $Newt (f)$ corresponds to a ray of $trop (f)$ pointing in the inner normal direction.

Rubric: full credit for the four rays, the union form, and the explicit duality with the four edges of the square.

Exercise 4 (medium, short-answer).

Prove that the tropical hypersurface $trop (f) \subseteq N_{R}$ is a rational polyhedral complex of pure dimension $n - 1$ : a finite union of $(n - 1)$ -dimensional polyhedra glued along faces, each polyhedron defined by linear inequalities with lattice-rational coefficients.

Hint

Each top-dimensional cell of $trop (f)$ corresponds to a pair ${α, β} \subset supp (f)$ attaining the minimum simultaneously, while no other index $γ$ does. The cell is cut out by linear equations and inequalities in $α - β$ (and the other $γ$ 's).

Answer

For each pair ${α, β} \subseteq supp (f)$ with $α \neq = β$ , let $$ C_{{\alpha, \beta}} := {u \in N_\mathbb{R} : \langle \alpha - \beta, u\rangle = v(c_\beta) - v(c_\alpha), \text{ and } v(c_\alpha) + \langle \alpha, u\rangle \leq v(c_\gamma) + \langle \gamma, u\rangle \text{ for all } \gamma \in \mathrm{supp}(f)}. $$ This is the locus where $α$ and $β$ tie for the minimum, and no $γ$ beats them. It is defined by one linear equation (the equation $⟨ α - β, u ⟩ = v (c_{β}) - v (c_{α})$ , with lattice-rational coefficient $α - β \in M$ ) and finitely many linear inequalities ( $⟨ α - γ, u ⟩ \leq v (c_{γ}) - v (c_{α})$ , again with lattice-rational coefficients), so $C_{{α, β}}$ is a rational polyhedron of dimension at most $n - 1$ .

The tropical hypersurface decomposes as the union $trop (f) = ⋃_{{α, β}} C_{{α, β}}$ over all pairs of distinct indices in the support. The full union is finite (at most $(2 ∣ supp ( f ) ∣)$ cells), and each cell is a rational polyhedron. The top-dimensional cells are exactly those of dimension $n - 1$ , corresponding to pairs ${α, β}$ such that no third index $γ$ ties: the polyhedron is non-degenerate iff its defining linear equation has full rank within the constraint set, which happens when $α - β$ generates a non-vanishing direction in $M_{R}$ .

Lower-dimensional cells correspond to triple, quadruple, etc., simultaneous ties of three or more support indices; they are intersections of multiple cells $C_{{α, β}}$ and form the codimension-2 and higher skeleton of the polyhedral complex. The dimension stratification, combined with rationality of every defining equation/inequality, identifies $trop (f)$ as a rational polyhedral complex of pure dimension $n - 1$ .

Rubric: full credit for the cell description, the polyhedron structure, and the dimension count.

Exercise 5 (medium, symbolic).

Apply the Kapranov theorem to the Laurent polynomial $f (x, y) = 1 + t x + t y$ over the field $K = C {{t}}$ of formal Puiseux series with the $t$ -adic valuation $v (\sum a_{q} t^{q}) = min {q : a_{q} \neq = 0}$ . Compute the non-archimedean amoeba $A_{v} (f) \subset R^{2}$ directly via Kapranov, then verify by finding an explicit point.

Hint

The valuation data is $v (1) = 0, v (t) = 1, v (t) = 1$ . Tropicalise: $T (f) (u_{1}, u_{2}) = min (0, 1 + u_{1}, 1 + u_{2})$ . Find the locus where two of these tie.

Answer

Tropicalisation: $T (f) (u_{1}, u_{2}) = min (0, 1 + u_{1}, 1 + u_{2})$ , the minimum of three affine functions on $R^{2}$ .

Three regions:

$0$ is min: $u_{1} \geq - 1, u_{2} \geq - 1$ ;
$1 + u_{1}$ is min: $u_{1} \leq - 1, u_{1} \leq u_{2}$ ;
$1 + u_{2}$ is min: $u_{2} \leq - 1, u_{2} \leq u_{1}$ .

Tropical hypersurface: the boundaries:

$0 = 1 + u_{1}$ at $u_{2} \geq - 1$ , i.e., the half-line $u_{1} = - 1, u_{2} \geq - 1$ ;
$0 = 1 + u_{2}$ at $u_{1} \geq - 1$ , i.e., $u_{2} = - 1, u_{1} \geq - 1$ ;
$1 + u_{1} = 1 + u_{2}$ at both $\leq 0$ , i.e., $u_{1} = u_{2} \leq - 1$ .

So $trop (f)$ is a piecewise-linear "Y" with vertex at $(- 1, - 1)$ and three rays in the directions $(0, 1), (1, 0), (- 1, - 1)$ . By Kapranov, this equals $A_{v} (f)$ .

Explicit point verification: Take $u = (- 1, - 1)$ , the vertex of the tropical "Y". We need $z = (z_{1}, z_{2}) \in V (f) \subset (K^{*})^{2}$ with $v (z_{1}) = v (z_{2}) = - 1$ . Try $z_{1} = t^{- 1}, z_{2} = t^{- 1} w$ for some $w \in K^{*}$ with $v (w) = 0$ . Substitute into $f (z) = 1 + t z_{1} + t z_{2} = 1 + 1 + w = 0$ , so $w = - 2 \in C^{*} \subset K^{*}$ with $v (- 2) = 0$ . The point $z = (t^{- 1}, - 2 t^{- 1}) \in V (f)$ has valuation $u = (- 1, - 1)$ , confirming $(- 1, - 1) \in A_{v} (f)$ .

Rubric: full credit for the three-ray description, the explicit point at the vertex, and the confirmation that the point matches Kapranov.

Exercise 6 (medium, multiple choice).

Let $f$ be a Laurent polynomial whose Newton polytope $Newt (f)$ is a regular hexagon centred at the origin in $R^{2}$ . Without further information about $f$ , how many tentacles does the complex amoeba $A (f)$ have?

A. $3$ (one per pair of opposite edges).
B. $6$ (one per edge of the hexagon).
C. $12$ (one per edge and one per vertex).
D. Could be anywhere from $0$ to $6$ depending on the specific coefficients.

Hint

Forsberg-Passare-Tsikh 2000: tentacles correspond to facets of the Newton polytope, and every outer-normal direction of a facet is realised as a tentacle direction for generic coefficients. Coefficient choices can occasionally make a tentacle "swallow" into the body, but for generic $f$ , all six are present.

Answer

B. A regular hexagon has six edges, each with a distinct outer normal direction (pointing to the six cardinal-ish directions at $60°$ intervals). For generic coefficients, every outer normal becomes a tentacle: the amoeba has six tentacles. Forsberg-Passare-Tsikh 2000 Theorem 2.6 establishes a bijection between the unbounded connected components of $N_{R} ∖ A (f)$ and the lattice points of $Newt (f) \cap M$ via the order map; the outer ones correspond to vertices of the polytope and contribute tentacles between adjacent vertex-cones, which is one tentacle per edge.

Option D is morally right for non-generic coefficients (specifically tuned families can collapse tentacles), but the question asked for the generic count, which is the number of edges.

Exercise 7 (hard, short-answer).

Prove the tentacle-alignment theorem for the complex amoeba: every unbounded direction in which $A (f)$ extends — every recession direction $v \in N_{R}$ for $A (f)$ — lies in the outer normal cone of some facet of $Newt (f)$ .

Hint

Suppose a sequence $z^{(k)} \in V (f)$ has $Log (z^{(k)}) \to \infty$ in direction $v$ , i.e., $lo g ∣ z^{(k)} ∣/∥ lo g ∣ z^{(k)} ∣∥ \to v$ . Apply the cancellation argument: the monomial $c_{α} (z^{(k)})^{α}$ with the largest modulus must cancel against another monomial $c_{β} (z^{(k)})^{β}$ of comparable modulus. Compare $∣ α \cdot v ∣$ to $∣ β \cdot v ∣$ .

Answer

Let $v \in N_{R}$ be a recession direction of $A (f)$ : there exists a sequence $z^{(k)} \in V (f) \subset (C^{*})^{n}$ with $u^{(k)} := Log (z^{(k)})$ satisfying $∥ u^{(k)} ∥ \to \infty$ and $u^{(k)} /∥ u^{(k)} ∥ \to v$ .

Step 1: monomial magnitudes. For each $α \in supp (f)$ , the monomial $c_{α} (z^{(k)})^{α}$ has log-modulus $$ \log|c_\alpha (z^{(k)})^\alpha| = \log|c_\alpha| + \langle \alpha, u^{(k)}\rangle. $$ As $k \to \infty$ , the dominant term is $⟨ α, u^{(k)} ⟩ \sim ⟨ α, v ⟩ \cdot ∥ u^{(k)} ∥$ . The maximum over $α \in supp (f)$ is achieved by indices $α$ maximising $⟨ α, v ⟩$ .

Step 2: maximising indices form a face. The set $F_{v} := argmax_{α \in supp (f)} ⟨ α, v ⟩$ is precisely the set of vertices of $Newt (f)$ where the linear functional $⟨ -, v ⟩$ attains its maximum. By the supporting-hyperplane theorem (a consequence of the cone-duality theorem of 04.11.02 applied to the polytope $Newt (f)$ ), $F_{v}$ is the vertex set of a face of $Newt (f)$ — the face exposed by the linear functional $v$ .

Step 3: cancellation requires a face of positive dimension. Suppose $∣ F_{v} ∣ = 1$ , i.e., a unique $α_{0}$ maximises $⟨ α, v ⟩$ . Then the monomial $c_{α_{0}} (z^{(k)})^{α_{0}}$ has log-modulus exceeding every other monomial by a margin growing linearly in $∥ u^{(k)} ∥$ : $$ \log|c_{\alpha_0} (z^{(k)})^{\alpha_0}| - \log\Big(\sum_{\alpha \neq \alpha_0} |c_\alpha (z^{(k)})^\alpha|\Big) \to +\infty. $$ This contradicts $f (z^{(k)}) = 0$ , which requires $∣ c_{α_{0}} (z^{(k)})^{α_{0}} ∣ \leq \sum_{α \neq = α_{0}} ∣ c_{α} (z^{(k)})^{α} ∣$ by the triangle inequality. So $∣ F_{v} ∣ \geq 2$ : at least two indices share the maximum.

Step 4: $F_{v}$ is the vertex set of a facet (or higher-dimensional face). Since $F_{v}$ has $\geq 2$ vertices, the face it spans has dimension $\geq 1$ . The linear functional $v$ is constant on this face; equivalently, $v$ lies in the normal cone to this face in the normal-fan-of- $Newt (f)$ structure on $N_{R}$ . The "tentacle direction" $v$ is therefore in the relative interior of the normal cone to a face of $Newt (f)$ of dimension $\geq 1$ . The maximal-dimension case — when $v$ is in the relative interior of the normal cone to a facet — is the generic case; lower-dimensional faces correspond to "tentacle-tip" directions where multiple tentacles converge.

Conclusion. Every recession direction $v$ of $A (f)$ lies in the normal cone of some face of $Newt (f)$ of dimension $\geq 1$ , and in particular in the union of the normal cones to facets of $Newt (f)$ . The tentacle directions are exhausted by facet outer normals.

Rubric: full credit for the cancellation argument, the use of the supporting-hyperplane theorem to identify $F_{v}$ , and the dimension count showing $∣ F_{v} ∣ \geq 2$ .

Exercise 8 (hard, short-answer).

Prove Mikhalkin's spine convergence theorem (sketched form): let $f$ be a Laurent polynomial over the Puiseux series field $K = C {{t}}$ with $t$ -adic valuation, and let $f^{t} \in C [x_{1}^{\pm 1}, \dots, x_{n}^{\pm 1}]$ be the family obtained by specialising $t$ to a positive real. Then $$ \frac{1}{\log t} \cdot \mathcal{A}(f^t) \longrightarrow \mathrm{trop}(f) \quad \text{as } t \to \infty $$ in the Hausdorff metric on closed subsets of bounded regions of $N_{R}$ .

Hint

Express the amoeba $A (f^{t})$ as the image of $V (f^{t}) \subset (C^{*})^{n}$ under $Log_{t} : (z_{1}, \dots, z_{n}) \mapsto (lo g_{t} ∣ z_{1} ∣, \dots, lo g_{t} ∣ z_{n} ∣)$ , where $lo g_{t} = lo g / lo g t$ . Then $Log_{t} (z) = Log (z) / lo g t$ . Compare to the Kapranov theorem for the limit at $t = \infty$ .

Answer

Step 1: log-base- $t$ rephrasing. The rescaled amoeba $\frac{1}{l o g t} A (f^{t})$ equals the image of $V (f^{t})$ under the base- $t$ logarithm map $$ \mathrm{Log}t : (\mathbb{C}^*)^n \to N\mathbb{R}, \qquad (z_1, \ldots, z_n) \mapsto (\log_t|z_1|, \ldots, \log_t|z_n|). $$

Step 2: $t \to \infty$ asymptotic. A point $u \in N_{R}$ lies in $\frac{1}{l o g t} A (f^{t})$ iff there exists $z \in V (f^{t}) \subset (C^{*})^{n}$ with $lo g_{t} ∣ z_{i} ∣ = u_{i}$ , i.e., $∣ z_{i} ∣ = t^{u_{i}}$ . As $t \to \infty$ , the magnitude $∣ z_{i} ∣ = t^{u_{i}}$ scales geometrically in $t$ , and the coefficients of $f^{t}$ scale according to the Puiseux-series exponents — exactly the data of the valuation $v$ .

Step 3: comparison with Kapranov. Fix $u \in trop (f)$ and find $z$ approximately satisfying $f^{t} (z) = 0$ and $lo g_{t} ∣ z ∣ \approx u$ . By Kapranov, $u \in A_{v} (f)$ : there exists $\tilde{z} \in V (f) \subset (K^{*})^{n}$ with $val (\tilde{z}) = u$ . Approximate $\tilde{z}$ by a truncated Puiseux series and substitute $t \mapsto$ a positive real: the result is a $z \in V (f^{t})$ up to higher-order corrections in $1/ lo g t$ , with $lo g_{t} ∣ z ∣ = u + O (1/ lo g t)$ . The Hausdorff distance between $\frac{1}{l o g t} A (f^{t})$ and $trop (f)$ is then bounded by $O (1/ lo g t) \to 0$ .

Step 4: the reverse inclusion. Conversely, fix $u \in / trop (f)$ : in the tropicalisation $T (f) (u) = v (c_{α_{0}}) + ⟨ α_{0}, u ⟩$ is attained at a unique $α_{0}$ . For large $t$ and $z$ with $lo g_{t} ∣ z ∣ \approx u$ , the monomial $c_{α_{0}}^{t} z^{α_{0}}$ in $f^{t} (z)$ has $lo g_{t}$ -modulus exceeding every other monomial by a fixed positive margin (depending on the gap between the second-smallest and smallest $v (c_{α}) + ⟨ α, u ⟩$ ). Hence $∣ f^{t} (z) ∣ ≍ ∣ c_{α_{0}}^{t} z^{α_{0}} ∣ \neq = 0$ for such $z$ , contradicting $z \in V (f^{t})$ . So $u \in / \frac{1}{l o g t} A (f^{t})$ for large $t$ , completing the Hausdorff-convergence statement.

Step 5: quantitative bound. The Hausdorff distance is bounded by $C / lo g t$ on any compact region of $N_{R}$ , with $C$ depending on $f$ and the region but not on $t$ . This is the quantitative form of Mikhalkin 2004 Theorem 4.

Conclusion. $\frac{1}{l o g t} A (f^{t})$ converges to $trop (f)$ in Hausdorff metric on bounded regions, at rate $O (1/ lo g t)$ . The geometric content: the analytic complex amoeba degenerates (after the natural rescaling $1/ lo g t$ ) to its combinatorial tropical shadow, witnessing the analytic-to-tropical bridge as a precise limit.

Rubric: full credit for the log-base- $t$ rephrasing, the use of Kapranov for the forward inclusion, the cancellation argument for the reverse inclusion, and the Hausdorff-convergence statement.

Lean formalization [Intermediate+]

The companion file Codex.AlgGeom.Tropical.NewtonPolytopeAmoeba declares the central objects of this unit together with the key alignment statement.

import Mathlib.Algebra.Tropical.Basic
import Mathlib.Analysis.Convex.Hull
import Mathlib.Analysis.Convex.Cone.Basic

namespace Codex.AlgGeom.Tropical

variable {n : ℕ}

/-- Schematic Laurent polynomial: a finite map from exponent vectors
in `Fin n → ℤ` to coefficients in a field `K`. -/
structure LaurentPoly (K : Type*) [Field K] (n : ℕ) where
  support : Finset (Fin n → ℤ)
  coeff : (Fin n → ℤ) → K
  coeff_support : ∀ α ∈ support, coeff α ≠ 0
  coeff_zero : ∀ α, α ∉ support → coeff α = 0

/-- The Newton polytope of a Laurent polynomial. -/
noncomputable def newtonPolytope {K : Type*} [Field K]
    (f : LaurentPoly K n) : Set (MR n) :=
  convexHull ℝ ((fun α : Fin n → ℤ => (fun i => (α i : ℝ)) : Fin n → ℤ → MR n) '' f.supp)

/-- The complex amoeba `A(f) := Log(V(f)) ⊂ N_R`. -/
noncomputable def amoeba (f : LaurentPoly ℂ n) : Set (NR n) := sorry

/-- The tropical hypersurface: the set of `u ∈ N_R` where the
tropicalisation `T(f)(u) := min_α (val(c_α) + α·u)` is attained
by at least two `α`. -/
noncomputable def tropicalHypersurface {K : Type*} [Field K]
    (val : K → ℝ) (f : LaurentPoly K n) : Set (NR n) := sorry

/-- Tentacle-alignment theorem (Forsberg-Passare-Tsikh 2000): every
recession direction of the complex amoeba `A(f)` is normal to a
facet of the Newton polytope `Newt(f)`. -/
theorem tentacle_alignment
    (f : LaurentPoly ℂ n) (v : NR n)
    (hv : IsRecessionDirection (amoeba f) v) :
    IsFacetNormal f v := sorry

/-- Mikhalkin spine convergence (Mikhalkin 2004): as `t → ∞`, the
rescaled amoeba `A(f^t) / log t` converges to the tropical
hypersurface `trop(f)` in Hausdorff metric on bounded regions. -/
theorem spine_convergence
    (f : LaurentPoly ℂ n) (val_data : (Fin n → ℤ) → ℝ) :
    True := sorry

end Codex.AlgGeom.Tropical

Mathlib already has the tropical-semiring infrastructure (Mathlib.Algebra.Tropical.Basic), the convex-hull and convex-cone development (Mathlib.Analysis.Convex.Hull, Mathlib.Analysis.Convex.Cone.Basic), and a partial Laurent-polynomial API. What it lacks is the named objects of amoeba theory: the explicit Newton-polytope construction, the complex amoeba A(f) as a subset of $N_{R}$ , the non-archimedean amoeba via valuations, the tropical hypersurface as the corner locus of $T (f)$ , and the headline theorems (Kapranov's identification, Forsberg-Passare-Tsikh tentacle alignment, Mikhalkin's spine convergence). The sorry-stubbed file records the API and the named statements so a future Mathlib contribution has a target to fill in. The expected route is a Mathlib.AlgebraicGeometry.Tropical namespace housing Laurent polynomials, the Newton-polytope functor, and the analytic / non-archimedean amoeba constructions, with the key alignment theorem reducing (in the algebraic-geometric direction) to the Kapranov theorem and (in the analytic direction) to the Mikhalkin spine theorem.

Advanced results [Master]

Newton polytope and its normal-fan structure

The Newton polytope $Newt (f)$ of a Laurent polynomial inherits the full combinatorial machinery of lattice polytopes from 04.11.02. Its normal fan $Σ (Newt (f))$ — the fan in $N_{R}$ whose maximal cones are the inner-normal cones of the vertices of $Newt (f)$ — is the combinatorial datum that organises both the asymptotic structure of the complex amoeba $A (f)$ and the affine cells of the tropical hypersurface $trop (f)$ . The toric variety $X_{Σ (Newt (f))}$ associated to this fan (constructed in 04.11.04) is the natural ambient space in which the closure of $V (f) \subset (C^{*})^{n}$ lives, and the boundary divisor of $X_{Σ (Newt (f))} ∖ T$ corresponds combinatorially to the rays of the normal fan, equivalently the facets of $Newt (f)$ .

Theorem (Bernstein-Kushnirenko, 1975). For Laurent polynomials $f_{1}, \dots, f_{n}$ in $n$ variables over $C$ with generic coefficients, the number of common zeros in $(\mathbb{C}^)^n $e q u a l s t h e * * mi x e d v o l u m e * *$ \mathrm{MV}(\mathrm{Newt}(f_1), \ldots, \mathrm{Newt}(f_n))$ of the Newton polytopes.*

The Bernstein-Kushnirenko theorem is the foundational quantitative result on the Newton polytope: it bounds the number of solutions of a system of Laurent polynomials by a combinatorial invariant of the Newton polytopes. For a single polynomial $f$ in two variables with generic coefficients, the number of complex zeros in $(C^{*})^{2}$ on a generic line is bounded by the lattice width of $Newt (f)$ in the direction of the line. The same convex-geometric bound governs the topology of $V (f)$ via the Khovanskii-Newton theorem on the genus of $V (f)$ .

Complex amoeba A(f) and its tentacle directions

Theorem (Forsberg-Passare-Tsikh, 2000). The complement $N_{R} ∖ A (f)$ of the complex amoeba is a disjoint union of finitely many open convex regions, and the connected components are in bijection with the lattice points of $Newt (f) \cap M$ via the order map $ord : (N_{R} ∖ A (f)) \to Newt (f) \cap M$ . The unbounded components correspond to the vertices of $Newt (f)$ , and the tentacles of $A (f)$ are the asymptotic structures separating these unbounded components.

The order map is defined as follows: for $u \in / A (f)$ , the Ronkin function $N_{f} (u) = \frac{1}{( 2 π i ) ^{n}} \int_{{∣ z_{i} ∣ = e^{u_{i}}}} lo g ∣ f (z) ∣ \frac{d z _{1}}{z _{1}} \dots \frac{d z _{n}}{z _{n}}$ is real-analytic and convex on $N_{R} ∖ A (f)$ . Its gradient $\nabla N_{f} (u)$ is constant on each component of the complement and takes lattice-point values in $Newt (f) \cap M$ (Passare-Rullgård 2004); this gradient is the order map. The order map gives the precise statement that the complement of $A (f)$ is combinatorially labelled by lattice points of $Newt (f)$ .

Theorem (Mikhalkin, 2000). The amoeba $A (f)$ admits a spine $S (f) \subset A (f)$ — a deformation retract that is a piecewise-linear $(n - 1)$ -dimensional polyhedral complex. The spine $S (f)$ has the same homotopy type as $A (f)$ , and its asymptotic structure matches the tropical hypersurface of $f$ for a suitable valuation.

Mikhalkin's spine is the analytic precursor to the tropical hypersurface; for generic $f$ , the spine $S (f)$ approaches the tropical hypersurface $trop (f)$ after a rescaling, leading to the spine-convergence theorem below.

Non-archimedean amoeba = tropical hypersurface

Theorem (Kapranov; Einsiedler-Kapranov-Lind 2006). Let $K$ be an algebraically closed field with a non-degenerate real-valued non-archimedean valuation $v$ . For any Laurent polynomial $f \in K [x_{1}^{\pm 1}, \dots, x_{n}^{\pm 1}]$ , the non-archimedean amoeba $A_{v} (f) := val (V (f)) \subseteq N_{R}$ coincides with the tropical hypersurface $trop (f)$ defined by the corner-locus of the tropicalisation $T (f)$ .

Kapranov's theorem (proved in the Intermediate-tier section) is the cleanest version of the complex-to-tropical bridge — it identifies the algebraic non-archimedean amoeba with the combinatorial tropical hypersurface as an exact equality, not as a limit. The non-archimedean setting bypasses the analytic complications of the complex amoeba: the residue-field reduction $in_{u} (f) = \sum_{α \in F_{u}} \overline{c_{α} t^{α} / π^{m}} \cdot w^{α}$ — where $F_{u}$ is the set of indices attaining the minimum in $T (f) (u)$ and $π$ is a uniformiser — supplies a polynomial in the residue field whose zeros lift, by Hensel's lemma, to points of $V (f)$ with the desired valuation. The proof structure generalises to higher-codimensional subvarieties (Bieri-Groves 1984, Sturmfels-Tevelev 2005), giving the full structure theorem of tropical algebraic geometry: the tropicalisation $trop (X)$ of any subvariety $X \subseteq T$ is a rational polyhedral complex of dimension equal to $dim X$ , equipped with positive-integer multiplicities satisfying a balancing condition at every codimension-1 face.

Mikhalkin spine convergence — amoebas degenerate to tropical varieties

Theorem (Mikhalkin, 2004). Let $K = C {{t}}$ be the field of Puiseux series with the $t$ -adic valuation, and let $f \in K [x_{1}^{\pm 1}, \dots, x_{n}^{\pm 1}]$ be a Laurent polynomial. For each $t > 1$ , let $f^{t} \in C [x_{1}^{\pm 1}, \dots, x_{n}^{\pm 1}]$ be the Laurent polynomial obtained by specialising the formal parameter to a positive real number $t$ . Then as $t \to \infty$ , $$ \frac{1}{\log t} \cdot \mathcal{A}(f^t) \longrightarrow \mathrm{trop}(f) $$ in the Hausdorff metric on closed subsets of any compact region of $N_{R}$ .

The Mikhalkin spine theorem is the analytic witness of the same identification Kapranov's theorem produces algebraically. The rescaling $A (f^{t}) / lo g t$ is the natural way to compare the analytic amoeba (which thickens as $t \to \infty$ ) to the tropical hypersurface (which remains a measure-zero set). The proof passes through the Forsberg-Passare-Tsikh order map and the Ronkin-function analysis: the components of $N_{R} ∖ A (f^{t})$ converge to components of $N_{R} ∖ trop (f)$ , and the amoeba $A (f^{t})$ shrinks onto its spine, which converges to $trop (f)$ .

The historical predecessor of Mikhalkin's theorem is Bergman's 1971 logarithmic limit set $Log_{\infty} (V)$ , defined as the limit (in the Hausdorff metric) of the rescaled amoebas $A (V) / r$ as $r \to \infty$ . Bergman conjectured that $Log_{\infty} (V)$ is a finite union of rational polyhedral cones; Bieri-Groves 1984 proved this conjecture by showing $Log_{\infty} (V) = trop (V)$ for an irreducible $V$ — the same identification Kapranov-Mikhalkin would later put in non-archimedean and analytic-degeneration forms.

Newton-polytope subdivisions and tropical curves

Theorem (regular subdivision correspondence; Maclagan-Sturmfels §1.4). Let $f = \sum c_{α} x^{α}$ be a Laurent polynomial in two variables. The polyhedral complex $trop (f) \subset R^{2}$ is dual to a regular subdivision $R_{f}$ of the Newton polytope $Newt (f) \subset R^{2}$ : each vertex of $trop (f)$ corresponds to a cell of $R_{f}$ , each bounded edge of $trop (f)$ corresponds to an internal edge of $R_{f}$ , and each unbounded edge of $trop (f)$ corresponds to a boundary edge of $R_{f}$ . The subdivision $R_{f}$ is the projection of the lower hull of the lifted points ${(α, v (c_{α})) : α \in supp (f)} \subset R^{3}$ down to $R^{2}$ .

The regular-subdivision correspondence is the combinatorial heart of tropical curve theory: a tropical curve in $R^{2}$ is dual (in a precise sense) to a regular subdivision of its Newton polytope. The correspondence specialises to the lattice triangulation when all subdivision cells are lattice triangles, which is the trivalent case for the tropical curve. This duality is the foundation of Mikhalkin's correspondence theorem 04.12.05, where the combinatorial type of a tropical curve — equivalently the regular subdivision of its Newton polytope — controls the enumerative geometry of the corresponding complex algebraic curves in toric surfaces.

Bergman fan and the universal-construction

Theorem (Bergman-Bieri-Groves; Maclagan-Sturmfels Ch. 3). For any subvariety $V \subseteq T = (\mathbb{C}^)^n $o f p u r e d im e n s i o n$ d $, t h e t r o p i c a l i s a t i o n$ \mathrm{trop}(V) := \overline{\mathrm{Log}(V)/r} $(a s$ r \to \infty $, e q u i v a l e n tl y t h e n o n - a r c him e d e anam oe ba o f$ V $o v er an o n - a r c him e d e an f i e l d$ K $) i s a r a t i o na l p o l y h e d r a l co m pl e x o f p u r e d im e n s i o n$ d $in$ N_\mathbb{R} $. T h eco m pl e x c a r r i es p os i t i v e - in t e g er m u l t i pl i c i t i es$ \mu_\sigma $o ni t s$ d $- d im e n s i o na l co n ess a t i s f y in g aba l an c in g co n d i t i o n : a t e v er y co d im e n s i o n - 1 co n e$ \tau $o f$ \mathrm{trop}(V)$,* $$ \sum_{\sigma \supset \tau} \mu_\sigma \cdot v_{\sigma/\tau} = 0 \in N_\mathbb{R}/\mathrm{span}(\tau), $$ where $v_{σ / τ}$ is the primitive lattice vector in $N / span (τ)$ pointing from $τ$ into $σ$ .

The Bergman-Bieri-Groves theorem is the general structure theorem: the tropical variety of any algebraic variety is a balanced rational polyhedral complex. For a hypersurface $V = V (f)$ , the balancing condition specialises to the data of the Newton polytope subdivision: the multiplicity $μ_{σ}$ on a top-dimensional cone $σ$ equals the lattice length of the corresponding edge in the dual subdivision $R_{f}$ , and balancing reduces to the closing condition on this subdivision (each interior lattice point lies on an even number of subdivision edges, weighted by their lattice lengths).

The Sturmfels-Tevelev geometric tropicalisation

Theorem (Sturmfels-Tevelev, 2005). Let $X \subseteq T$ be an irreducible subvariety of the algebraic torus, and let $\overset{ˉ}{X} \subseteq X_{Σ}$ be its closure in a toric variety $X_{Σ}$ with fan $Σ$ refining the recession fan of $trop (X)$ . The tropicalisation $trop (X)$ is then the union of the recession cones (in $N_{R}$ ) of the orbit-strata of $\overset{ˉ}{X}$ in $X_{Σ}$ : for each cone $σ \in Σ$ such that $\overset{ˉ}{X}$ meets the orbit $O_{σ}$ , the cone $σ$ contributes to $trop (X)$ with multiplicity equal to the intersection multiplicity of $\overset{ˉ}{X}$ with $O_{σ}$ .

The Sturmfels-Tevelev theorem is the geometric description of tropicalisation: it identifies $trop (X)$ with the recession data of the closure of $X$ in a suitable toric compactification. This is the bridge from the abstract algebraic / valuative definition of tropicalisation to the explicit toric-geometric computation, and is what makes tropical methods computationally accessible — to compute $trop (X)$ , find a toric compactification of $T$ in which $\overset{ˉ}{X}$ is well-positioned, then read off the recession cones of the orbit-strata.

Synthesis. The three objects — Newton polytope on the exponent side, complex amoeba on the analytic side, tropical hypersurface on the combinatorial side — are different faces of the same combinatorial-analytic structure, and the central insight is that the normal fan of the Newton polytope is exactly the asymptotic skeleton of both the complex amoeba and the tropical hypersurface. The foundational reason is the cancellation principle in the vanishing equation $f (z) = 0$ : a sum of monomials can vanish only when at least two monomials achieve the maximum modulus (complex case) or the minimum valuation (non-archimedean case), and this two-monomial condition is exactly the corner-locus condition on the tropicalisation $T (f)$ . The Kapranov theorem captures this on the algebraic side, the Mikhalkin spine theorem on the analytic side, and the regular-subdivision correspondence on the combinatorial side. The bridge is between the Newton polytope (an object in exponent space $M_{R}$ ) and the tropical hypersurface (an object in valuation / coordinate space $N_{R}$ ), via the normal-fan duality of 04.11.02: every facet of $Newt (f)$ produces a ray of $trop (f)$ , every edge of $Newt (f)$ in dimension $2$ produces a vertex of $trop (f)$ , and the duality identifies cells of one with cells of the other in inclusion-reversing fashion.

Putting these together with the toric-variety construction $X_{Σ (Newt (f))}$ from 04.11.04, the cone-and-fan formalism of 04.11.02 becomes the combinatorial backbone of the whole tropical-degeneration story: the Newton polytope's normal fan supplies a toric compactification in which $V (f)$ degenerates to a union of orbit-closures, and the tropicalisation of this degeneration is exactly the tropical hypersurface $trop (f)$ . This identifies the tropicalisation procedure with the special-fibre limit in a toric degeneration, generalising to the Sturmfels-Tevelev geometric tropicalisation and the Nishinou-Siebert correspondence 04.12.06, which extends from hypersurfaces to enumeration of algebraic curves in toric varieties.

The pattern recurs in mirror symmetry: in the Gross-Siebert program 04.12.09, a Calabi-Yau variety $X$ near a maximal degeneration point in moduli admits a toric degeneration whose dual intersection complex is a tropical manifold $B$ — the "tropical Calabi-Yau" — and the mirror $\overset{ˇ}{X}$ is reconstructed from $B$ via a slab-function / wall-crossing procedure. The tropical hypersurface developed here is the simplest case of the dual intersection complex: when $X$ is the hypersurface $V (f) \subset X_{Σ (Newt (f))}$ , the dual complex of its toric degeneration is exactly $trop (f)$ , and the mirror reconstruction generalises the cone-amoeba duality of this unit to the full Gross-Siebert framework. The Newton-polytope formalism is therefore the entry point to the entire tropical-geometry-and-mirror-symmetry program — every subsequent unit in this chapter reads a piece of geometry out of the combinatorics developed here.

Full proof set [Master]

Theorem (Kapranov; non-archimedean amoeba = tropical hypersurface), proof. Given in the Intermediate-tier section. Direction $A_{v} (f) \subseteq trop (f)$ uses the non-archimedean inequality: if $z \in V (f)$ has $val (z) = u$ and the minimum in $T (f) (u)$ is attained by a unique $α_{0}$ , then $v (f (z)) = v (c_{α_{0}}) + ⟨ α_{0}, u ⟩$ is finite, contradicting $f (z) = 0$ . Direction $trop (f) \subseteq A_{v} (f)$ uses the initial-form reduction: for $u \in trop (f)$ , the initial polynomial $in_{u} (f)$ in the residue field has at least two monomials and hence (by algebraic closure of the residue field) a root in $(k^{*})^{n}$ , which lifts by Hensel's lemma to $z \in V (f)$ with $val (z) = u$ . $□$

Proposition (tentacle alignment; Forsberg-Passare-Tsikh), proof. Given in Exercise 7. Let $v \in N_{R}$ be a recession direction of $A (f)$ , and let $z^{(k)} \in V (f)$ be a sequence with $Log (z^{(k)}) /∥ Log (z^{(k)}) ∥ \to v$ . The dominant monomials at $z^{(k)}$ are those $α$ maximising $⟨ α, v ⟩$ , forming the face $F_{v}$ of $Newt (f)$ exposed by $v$ . The cancellation $f (z^{(k)}) = 0$ requires $∣ F_{v} ∣ \geq 2$ : at least two indices share the maximum, so $F_{v}$ spans a positive-dimensional face of $Newt (f)$ , and $v$ lies in the normal cone of that face. The maximum-dimensional case is when $v$ is normal to a facet. $□$

Proposition (regular-subdivision correspondence), proof. Let $f = \sum c_{α} x^{α}$ in two variables. The tropical curve $trop (f) \subset R^{2}$ is dual to the regular subdivision $R_{f}$ of $Newt (f) \subset R^{2}$ obtained by lifting each support point $α$ to $(α, v (c_{α})) \in R^{3}$ and projecting the lower hull of the lifted points to $R^{2}$ .

Proof. The cells of $R_{f}$ are indexed by faces of the lower hull of ${(α, v (c_{α})) : α \in supp (f)}$ in $R^{3}$ . A vertex $u^{*}$ of $trop (f)$ corresponds to a $u \in N_{R}$ where the minimum in $T (f) (u) = min_{α} (v (c_{α}) + ⟨ α, u ⟩)$ is attained by $\geq 3$ indices. The set of indices attaining the minimum is the upper face of $Newt (f) \times {0}$ exposed by the affine functional $(α, h) \mapsto h - ⟨ α, u ⟩$ (lifted to $R^{3}$ via the valuation data) — equivalently the lower face of the lift. So vertices of $trop (f)$ correspond bijectively to top-dimensional cells of $R_{f}$ . Similarly, edges of $trop (f)$ correspond to edges of $R_{f}$ (regions where two indices tie but not three), and unbounded edges of $trop (f)$ correspond to boundary edges of $R_{f}$ (regions where the tying involves a vertex of $Newt (f)$ ). The duality is dimension-reversing within $Newt (f)$ : a $k$ -cell of $R_{f}$ corresponds to a $(2 - k)$ -cell of $trop (f)$ . $□$

Proposition (Bernstein-Kushnirenko in dimension 2), proof. For two Laurent polynomials $f_{1}, f_{2}$ in $C [x^{\pm 1}, y^{\pm 1}]$ with generic coefficients, the number of common zeros in $(\mathbb{C}^)^2 $e q u a l s t h e mi x e d v o l u m e$ \mathrm{MV}(\mathrm{Newt}(f_1), \mathrm{Newt}(f_2)) = \mathrm{vol}(\mathrm{Newt}(f_1) + \mathrm{Newt}(f_2)) - \mathrm{vol}(\mathrm{Newt}(f_1)) - \mathrm{vol}(\mathrm{Newt}(f_2))$.*

Proof sketch. Construct the toric surface $X_{Σ}$ where $Σ$ refines both normal fans of $Newt (f_{1})$ and $Newt (f_{2})$ . Each $f_{i}$ extends to a section $\overset{ˉ}{f}_{i}$ of a line bundle $L_{i}$ on $X_{Σ}$ ; the line bundle $L_{i}$ has top self-intersection number $L_{i}^{2} = vol (2 Newt (f_{i})) = 4 vol (Newt (f_{i}))$ and intersection $L_{1} \cdot L_{2} = 2 MV (Newt (f_{1}), Newt (f_{2}))$ (the toric intersection formula). The number of common zeros of $f_{1}, f_{2}$ in $(C^{*})^{2}$ equals the intersection number $L_{1} \cdot L_{2}$ on $X_{Σ}$ minus contributions from boundary divisors, which (by genericity of coefficients) vanish. So the count is $MV (Newt (f_{1}), Newt (f_{2}))$ . The general $n$ -dimensional case is analogous, using Kushnirenko's theorem on the mixed-volume intersection formula on toric varieties. $□$

Proposition (Mikhalkin spine convergence; sketched), proof. Given in Exercise 8. The base- $t$ -logarithm rephrasing turns $\frac{1}{l o g t} A (f^{t})$ into the image of $V (f^{t})$ under $Log_{t}$ , and the comparison with the Puiseux-series specialisation gives a $O (1/ lo g t)$ Hausdorff bound between the rescaled amoeba and $trop (f)$ . The forward inclusion uses Kapranov's theorem applied to the Puiseux-series Laurent polynomial; the reverse inclusion uses the cancellation argument for $u \in / trop (f)$ that the dominant monomial cannot be cancelled at large $t$ . $□$

Proposition (Bergman-Bieri-Groves for hypersurfaces), proof sketch. For a hypersurface $V = V (f) \subseteq T$ defined by a Laurent polynomial $f$ , the tropical variety $trop (V) = trop (f)$ is a rational polyhedral complex of pure dimension $n - 1$ , balanced with multiplicities $μ_{σ}$ on its top-dimensional cones equal to the lattice lengths of the corresponding edges in the regular subdivision $R_{f}$ of $Newt (f)$ .

Proof. By Exercise 4, $trop (f)$ is a rational polyhedral complex of dimension $\leq n - 1$ , decomposed as the union of cells $C_{{α, β}}$ for pairs ${α, β} \subseteq supp (f)$ . For generic $f$ , all top-dimensional cells have dimension exactly $n - 1$ (corresponding to pairs of indices whose difference $α - β$ spans a one-dimensional subspace of $M_{R}$ modulo the other constraints). Set the multiplicity $μ_{C_{{α, β}}}$ to the lattice length of the segment from $α$ to $β$ in $M$ , i.e., the GCD of the components of $β - α \in M$ .

Balancing at codimension-1 faces. A codimension-1 face $τ$ of $trop (f)$ corresponds to a triple ${α, β, γ} \subseteq supp (f)$ of indices simultaneously tying for the minimum. The three top-dimensional cells of $trop (f)$ adjacent to $τ$ have multiplicities equal to lattice lengths of $β - α$ , $γ - α$ , $γ - β$ . The balancing condition $\sum_{σ} μ_{σ} v_{σ / τ} = 0$ in $N_{R} / span (τ)$ becomes the lattice identity $(β - α) + (γ - β) + (α - γ) = 0$ in $M$ , which holds tautologically. So the multiplicities satisfy balancing at codimension-1, completing the structure theorem for hypersurfaces. The proof generalises to higher-codimensional subvarieties via the Sturmfels-Tevelev theorem on the geometric description of tropicalisation. $□$

Connections [Master]

Rational polyhedral cone and dual cone 04.11.02. The normal fan of the Newton polytope $Newt (f)$ — the combinatorial structure organising the tentacle directions of the complex amoeba and the cells of the tropical hypersurface — is built out of the rational polyhedral cones developed in the prerequisite unit. The dual cone $σ^{\lor}$ of a cone $σ \in Σ (Newt (f))$ is the cone of $u \in N_{R}$ for which the linear functional $⟨ -, u ⟩$ achieves its minimum on a specified face of $Newt (f)$ ; this is the precise sense in which the cone-and-fan formalism of 04.11.02 specialises to the tropical-hypersurface formalism of this unit.
Algebraic torus and character lattices 04.11.01. The Laurent polynomial $f = \sum c_{α} x^{α}$ defines a regular function on the algebraic torus $T = (G_{m})^{n}$ , and its support $supp (f) \subseteq M$ lives in the character lattice. The Log map $Log : T (C) \to N_{R}$ used to define the amoeba is the moment map for the maximal compact torus $T_{S^{1}} \subset T (C)$ , and the tentacle structure of $A (f)$ reflects the $T_{S^{1}}$ -equivariant geometry of the vanishing locus $V (f)$ .
Affine toric variety $U_{σ}$ 04.11.03. The closure $\overline{V (f)} \subseteq X_{Σ (Newt (f))}$ of the hypersurface in the toric compactification associated to the normal fan of $Newt (f)$ is the natural ambient space in which the tropical-hypersurface formalism takes its geometric form. The orbit-cone correspondence of 04.11.03 identifies the boundary strata of $X_{Σ (Newt (f))}$ with the faces of $Newt (f)$ in the dual fashion, providing the geometric backbone of the Sturmfels-Tevelev geometric tropicalisation.
Fan and toric variety $X_{Σ}$ 04.11.04. The construction of the toric compactification $X_{Σ (Newt (f))}$ from the fan $Σ (Newt (f))$ is the bridge between the convex-geometric Newton polytope and the algebraic-geometric ambient variety in which the tropical-degeneration story plays out. The closure $\overline{V (f)} \subseteq X_{Σ (Newt (f))}$ meets each boundary stratum in a divisor whose tropicalisation is a cell of $trop (f)$ .
Tropical semiring and tropical polynomial 04.12.01. The tropicalisation $T (f) : N_{R} \to R$ defined here is the canonical example of a tropical polynomial: a piecewise-linear function obtained by replacing $+$ with $min$ and $\times$ with $+$ in the formula for $f$ . The tropical-semiring formalism (pending unit) is the algebraic substrate of the corner-locus definition of $trop (f)$ , and the balancing-and-multiplicity structure on $trop (f)$ from the Bergman-Bieri-Groves theorem is the analogue of the algebraic-multiplicity structure on the vanishing locus of an ordinary polynomial.
Mikhalkin's correspondence theorem 04.12.05. The Newton-polytope and amoeba formalism of this unit supplies the input data for Mikhalkin's correspondence theorem, which counts algebraic curves in toric surfaces in terms of tropical curves. A tropical curve in $R^{2}$ — equivalently, a balanced piecewise-linear graph dual to a regular subdivision of a Newton polytope — corresponds to a finite collection of complex algebraic curves in the toric surface $X_{Σ (Newt (f))}$ , with the count given by a combinatorial formula on the tropical side. The dictionary developed here is the lower-dimensional precursor.
Kapranov's theorem 04.12.03. The Kapranov theorem stated and proved here (the identification $A_{v} (f) = trop (f)$ for non-archimedean $K$ ) is the prerequisite combinatorial-algebraic input to the more general Kapranov theorem on higher-codimensional tropical varieties, which extends the hypersurface case via the Sturmfels-Tevelev geometric description of tropicalisation. The prerequisite unit develops Kapranov's theorem in its more abstract form; this unit specialises to the hypersurface case with explicit Newton-polytope combinatorics.
Toric divisor and support function 04.11.08. The Newton polytope $Newt (f)$ is the moment polytope of the toric divisor on $X_{Σ (Newt (f))}$ obtained by lifting $f$ to a section of the corresponding line bundle. The support function of this divisor — a piecewise-linear function on $N_{R}$ extending the data of $Newt (f)$ to global linearity on each cone of $Σ (Newt (f))$ — is, after the natural identification, the tropicalisation $T (f)$ specialised to the vanishing valuation. The bridge between toric-divisor support functions and tropical polynomials is one of the most useful translations in the toric-tropical dictionary.
Polytope-fan dictionary; the line bundle $L_{P}$ 04.11.10. The Newton polytope $Newt (f) \subseteq M_{R}$ of a Laurent polynomial is exactly the polytope side of the polytope-fan dictionary applied to the toric compactification $X_{Σ (Newt (f))}$ : the normal fan $Σ (Newt (f))$ in $N_{R}$ is the fan of the toric variety, and the polytope itself supplies the polarisation $L_{Newt (f)}$ whose global sections $H^{0} (X_{Σ (Newt (f))}, L_{Newt (f)}) = ⨁_{m \in Newt (f) \cap M} C χ^{m}$ are indexed by the lattice points of $Newt (f)$ — exactly the support of $f$ . The Laurent polynomial $f$ is, in this sense, a specific global section of the polarisation $L_{Newt (f)}$ , and the tropicalisation $T (f)$ is the support function of $L_{Newt (f)}$ read through the lattice point data of $f$ . The polytope-fan dictionary is the foundational reason Newton-polytope data and tropical-hypersurface data are dual on the same lattice pair $(M, N)$ .
Tropical curve as balanced rational metric graph 04.12.02. For $f$ a Laurent polynomial in two variables, the tropical hypersurface $trop (f) \subseteq R^{2}$ is the embedded tropical curve of the cluster sub-chapter unit, and the Newton polygon $Newt (f) \subseteq M_{R}$ supplies its combinatorial type via the regular subdivision $Ξ$ induced by the lift $α \mapsto (α, c_{α})$ . The vertices of the tropical curve correspond to triangles of $Ξ$ , the bounded edges correspond to interior edges of $Ξ$ , and the unbounded rays correspond to boundary edges of $Ξ$ — with primitive directions the inner normals to the corresponding edges of $Newt (f)$ . The Newton-polygon duality of this unit is the foundational reason embedded tropical curves carry a balanced rational metric-graph structure with integer edge weights.
Nishinou-Siebert correspondence 04.12.06. The downstream higher-dimensional correspondence theorem uses the Newton-polytope-to-fan construction of this unit as the input data to a toric degeneration $X \to Spec C {{t}}$ : the polyhedral subdivision $Ξ$ of a polytope of an ample line bundle on the toric variety $X_{Σ (Δ)}$ produces the dual intersection complex $B (Ξ)$ on which the Nishinou-Siebert tropical curves live. The Newton-polytope formalism of this unit is the foundational input; the toric-degeneration mechanism converts the polytope subdivision into the algebraic-geometric degeneration along whose special fibre the tropical-to-complex lift takes place.
Toric degeneration of a Calabi-Yau variety 04.12.07. The Mumford construction $χ^{m} = t^{φ (m)} χ^{m}$ that defines the toric degeneration of a Calabi-Yau in [04.12.07] is, by construction, the one-parameter algebraic-geometric incarnation of the non-archimedean amoeba developed in the present unit: as $t = e^{- r} \to 0$ , the family $X_{t}$ tropicalises to the non-archimedean amoeba over the Puiseux-series field $C {{t}}$ , and the polyhedral subdivision $P$ is the secondary-polytope-vertex realisation of the corresponding tropical hypersurface. The Newton-polytope formalism of the present unit is the foundational combinatorial input on which the toric-degeneration setup rests.
Dual intersection complex; tropical manifold $B$ 04.12.08. The dual intersection complex $(B, P)$ of [04.12.08] starts from a lattice polytope $Δ \subseteq M_{R}$ associated to an ample line bundle on $X_{Σ}$ — exactly the Newton-polytope formalism of the present unit — and refines it by a polyhedral subdivision $Ξ$ ; the dual intersection complex is the cone-complex dual of $Ξ$ . The Newton-polytope and amoeba formalism developed here is the foundational input from which the tropical manifold is constructed.
Gross-Siebert reconstruction theorem 04.12.09. The reconstruction theorem of [04.12.09] assembles a smooth Calabi-Yau from a Newton-polytope-style input plus structural data: the polytope $Δ$ and its polyhedral subdivision $P$ are the polyhedral primitive, the slab functions on codim- $1$ cells are the gluing data, and the smooth mirror is reconstructed order-by-order in the smoothing parameter $t$ . The Newton-polytope-and-amoeba dictionary of the present unit is the foundational combinatorial substrate on which the reconstruction theorem is built; the polytope is the input from which $P$ refines and the smoothing emerges.
Strominger-Yau-Zaslow conjecture 04.12.10. The SYZ base of a Calabi-Yau toric degeneration is the dual intersection complex of a Newton-polytope subdivision in the present unit's formalism: the integer-affine structure on $B$ is built from local Newton-polytope charts glued along singularities, and the non-archimedean amoeba is the simplest case (smooth toric central fibre) of the SYZ-base construction in [04.12.10]. The Newton-polytope formalism of the present unit is the entry-level input to the SYZ-and-mirror-symmetry programme.
Slab function and structure of a tropical manifold 04.12.11. The slab functions on codim- $1$ cells of $P$ in [04.12.11] are Laurent polynomials in the toric coordinates of the adjacent chambers, with the Newton polytope of each slab function controlled by the Newton-polytope formalism of the present unit. The wall functions on higher-codimension scattering walls are Newton-polytope-decorated tropical-hypersurface pieces. The present unit's Newton-polytope and amoeba dictionary is the foundational toric-combinatorial substrate of the slab-and-wall apparatus.
Theta function of a polarised tropical manifold 04.12.12. The polarised tropical manifold $(B, P, φ)$ on which theta functions live in [04.12.12] is, in the rigid-toric specialisation, the Newton polytope of the polarisation; in the Calabi-Yau setting it is a global polytope-like object built from Newton-polytope local charts glued along singularities. The non-archimedean amoeba framework developed here supplies the analytic underpinning: theta functions correspond, on the non-archimedean side, to canonical analytic functions on the Berkovich skeleton of $X_{t}$ , with broken lines being the Berkovich-side tropicalisation of the smoothing's local sections.

Historical & philosophical context [Master]

The Newton polytope was implicit in Isaac Newton's Method of Series and Fluxions (1671, published 1736), where Newton used the "diagram" of a power-series equation to compute fractional power-series solutions — what is now called the Newton-Puiseux expansion. The systematic theory of lattice polytopes and their convex-geometric invariants is a nineteenth-century product of Minkowski's Geometrie der Zahlen (1896) ^{[Minkowski 1896]} and Gordan's work on linear inequalities (1873) ^{[Gordan 1873]}, reaching its modern form in the early-twentieth-century work of Hermann Weyl 1935 and the mid-twentieth-century systematic toric-variety treatments of Demazure 1970, Kempf-Knudsen-Mumford-Saint-Donat 1973, and Fulton 1993.

The amoeba was introduced by Gelfand, Kapranov, and Zelevinsky in their landmark Discriminants, Resultants, and Multidimensional Determinants (Birkhauser 1994) ^{[Gelfand-Kapranov-Zelevinsky 1994]}, Chapter 11, as a tool to study the topology of complex hypersurfaces in algebraic tori. The terminology "amoeba" was coined by Bernd Sturmfels in correspondence with Gelfand; the visual analogy is to the amoeba-like shape of the projection of a hypersurface. The systematic study of complex amoebas was developed by Mikhail Kapranov, Mikhail Passare, August Tsikh, and Grigory Mikhalkin in the late 1990s and early 2000s, culminating in Forsberg-Passare-Tsikh's 2000 Adv. Math. paper ^{[Forsberg-Passare-Tsikh 2000]} on the order map and the lattice-point labelling of complement components.

The non-archimedean amoeba and its identification with the tropical hypersurface (Kapranov's theorem) was first proved in unpublished form by Mikhail Kapranov around 1995 ^{[Kapranov 1995]}; the first published proof and the systematic theory appeared in Einsiedler-Kapranov-Lind 2006 J. reine angew. Math. ^{[Einsiedler-Kapranov-Lind 2006]}. The bridge between the analytic complex amoeba and the combinatorial tropical hypersurface — the spine-convergence theorem $A (f^{t}) / lo g t \to trop (f)$ — was Mikhalkin's contribution in his 2004 Topology paper Decomposition into pairs-of-pants for complex algebraic hypersurfaces ^{[Mikhalkin 2004]}, where the amoeba spine theorem was used as the analytic foundation for the topological pairs-of-pants decomposition of complex hypersurfaces in $(C^{*})^{n}$ .

The historical predecessor of all these results is George Bergman's 1971 Trans. AMS paper on the logarithmic limit set of an algebraic variety ^{[Bergman 1971]}, where the rescaled-amoeba limit $Log_{\infty} (V) = lim_{r \to \infty} A (V) / r$ was introduced. Bergman conjectured that $Log_{\infty} (V)$ is a finite union of rational polyhedral cones for irreducible $V$ ; the conjecture was proved by Bieri-Groves in 1984. The Bergman-Bieri-Groves theorem in its modern form — that $trop (V)$ is a balanced rational polyhedral complex of dimension $dim V$ — is in Maclagan-Sturmfels 2015 Introduction to Tropical Geometry (Graduate Studies in Mathematics 161, AMS) ^{[Maclagan-Sturmfels 2015]}, Chapter 3, which is the canonical textbook reference for the entire tropical-geometric framework developed in this unit.

The mirror-symmetry application of tropical geometry was opened by Strominger-Yau-Zaslow's 1996 Nuclear Physics B conjecture ^{[Strominger-Yau-Zaslow 1996]} (the SYZ conjecture, see 04.12.10) on torus-fibration interpretations of mirror symmetry, and developed into a precise reconstruction program by Mark Gross and Bernd Siebert in a sequence of papers from 2003 onward, culminating in the Gross-Siebert reconstruction theorem 04.12.09. The Newton-polytope and amoeba formalism of this unit is the simplest non-degenerate case of the Gross-Siebert framework: when the Calabi-Yau in question is a hypersurface in a toric variety, the dual intersection complex of its toric degeneration is exactly the tropical hypersurface of the defining Laurent polynomial.

Bibliography [Master]

@book{GelfandKapranovZelevinsky1994,
  author    = {Gelfand, Israel M. and Kapranov, Mikhail M. and Zelevinsky, Andrei V.},
  title     = {Discriminants, Resultants, and Multidimensional Determinants},
  publisher = {Birkh{\"a}user},
  address   = {Boston},
  year      = {1994},
  series    = {Mathematics: Theory and Applications}
}

@article{Mikhalkin2004,
  author  = {Mikhalkin, Grigory},
  title   = {Decomposition into pairs-of-pants for complex algebraic hypersurfaces},
  journal = {Topology},
  volume  = {43},
  year    = {2004},
  pages   = {1035--1065}
}

@book{MaclaganSturmfels2015,
  author    = {Maclagan, Diane and Sturmfels, Bernd},
  title     = {Introduction to Tropical Geometry},
  publisher = {American Mathematical Society},
  series    = {Graduate Studies in Mathematics},
  volume    = {161},
  year      = {2015}
}

@article{EinsiedlerKapranovLind2006,
  author  = {Einsiedler, Manfred and Kapranov, Mikhail and Lind, Douglas},
  title   = {Non-archimedean amoebas and tropical varieties},
  journal = {Journal f\"ur die reine und angewandte Mathematik},
  volume  = {601},
  year    = {2006},
  pages   = {139--157}
}

@article{ForsbergPassareTsikh2000,
  author  = {Forsberg, Mikael and Passare, Mikael and Tsikh, August},
  title   = {Laurent determinants and arrangements of hyperplane amoebas},
  journal = {Advances in Mathematics},
  volume  = {151},
  year    = {2000},
  pages   = {45--70}
}

@article{Bergman1971,
  author  = {Bergman, George M.},
  title   = {The logarithmic limit-set of an algebraic variety},
  journal = {Transactions of the American Mathematical Society},
  volume  = {157},
  year    = {1971},
  pages   = {459--469}
}

@article{BieriGroves1984,
  author  = {Bieri, Robert and Groves, J. R. J.},
  title   = {The geometry of the set of characters induced by valuations},
  journal = {Journal f\"ur die reine und angewandte Mathematik},
  volume  = {347},
  year    = {1984},
  pages   = {168--195}
}

@article{SturmfelsTevelev2008,
  author  = {Sturmfels, Bernd and Tevelev, Jenia},
  title   = {Elimination theory for tropical varieties},
  journal = {Mathematical Research Letters},
  volume  = {15},
  year    = {2008},
  pages   = {543--562}
}

@article{PassareRullgard2004,
  author  = {Passare, Mikael and Rullg{\aa}rd, Hans},
  title   = {Amoebas, {Monge}-{Amp{\`e}re} measures, and triangulations of the {Newton} polytope},
  journal = {Duke Mathematical Journal},
  volume  = {121},
  year    = {2004},
  pages   = {481--507}
}

@book{ItenbergMikhalkinShustin2009,
  author    = {Itenberg, Ilia and Mikhalkin, Grigory and Shustin, Eugenii},
  title     = {Tropical Algebraic Geometry},
  publisher = {Birkh{\"a}user},
  series    = {Oberwolfach Seminars},
  volume    = {35},
  year      = {2009}
}

@book{FultonToric,
  author    = {Fulton, William},
  title     = {Introduction to Toric Varieties},
  publisher = {Princeton University Press},
  series    = {Annals of Mathematics Studies},
  volume    = {131},
  year      = {1993}
}

@article{Kapranov1995,
  author  = {Kapranov, Mikhail M.},
  title   = {Amoebas over non-archimedean fields},
  journal = {Preprint},
  year    = {1995}
}

@article{StromingerYauZaslow1996,
  author  = {Strominger, Andrew and Yau, Shing-Tung and Zaslow, Eric},
  title   = {Mirror symmetry is {T}-duality},
  journal = {Nuclear Physics B},
  volume  = {479},
  year    = {1996},
  pages   = {243--259}
}

Prerequisites

04.11.02

Tier anchors

beginner: Maclagan-Sturmfels *Introduction to Tropical Geometry* §1.4 informal — Newton polytope as the 'shadow' of a Laurent polynomial; the amoeba as the projection of a hypersurface to log-modulus coordinates
intermediate: Maclagan-Sturmfels §1.4 + Ch. 3 (Newton polytope, complex amoeba, tropical hypersurface); Gelfand-Kapranov-Zelevinsky *Discriminants, Resultants, and Multidimensional Determinants* Ch. 6 §1 + Ch. 11 §1
master: Gelfand-Kapranov-Zelevinsky 1994 Ch. 6 + Ch. 11; Mikhalkin 2004 *Decomposition into pairs-of-pants for complex algebraic hypersurfaces* (*Topology* 43, 1035-1065); Einsiedler-Kapranov-Lind 2006 *Non-archimedean amoebas and tropical varieties* (*J. reine angew. Math.* 601); Forsberg-Passare-Tsikh 2000 *Laurent determinants and arrangements of hyperplane amoebas* (*Adv. Math.* 151); Maclagan-Sturmfels §1.4 and Ch. 3

References

TODO_REF
Gelfand-Kapranov-Zelevinsky — Discriminants, Resultants, and Multidimensional Determinants (Birkhauser 1994) · Ch. 6 §1 (Newton polytope of a Laurent polynomial; lattice polytopes; the normal fan); Ch. 11 §1 (amoebas of hypersurfaces; the order map and the Ronkin function); Ch. 11 §2 (tentacles aligned to facet normals of the Newton polytope)
TODO_REF
Mikhalkin — Decomposition into pairs-of-pants for complex algebraic hypersurfaces · *Topology* 43 (2004), 1035-1065; Theorem 1 (spine of the complex amoeba); Theorem 4 (Hausdorff convergence of A(f^t)/log|t| to the tropical hypersurface as t → ∞)
TODO_REF
Maclagan-Sturmfels — Introduction to Tropical Geometry · Graduate Studies in Mathematics 161, AMS 2015; §1.4 (Newton polytope, amoebas, tropical hypersurface); Theorem 3.1.3 (Kapranov: non-archimedean amoeba = tropical hypersurface); Ch. 3 (tropical varieties)
TODO_REF
Einsiedler-Kapranov-Lind — Non-archimedean amoebas and tropical varieties · *J. reine angew. Math.* 601 (2006), 139-157; Theorem A (the non-archimedean amoeba of V(f) is exactly the tropical hypersurface of f)
TODO_REF
Forsberg-Passare-Tsikh — Laurent determinants and arrangements of hyperplane amoebas · *Adv. Math.* 151 (2000), 45-70; Theorem 2.6 (the order map ord : complement of A(f) → ℤ^n ∩ Newt(f) and the convex-geometry interpretation of tentacle directions)
TODO_REF
Bergman 1971 — The logarithmic limit-set of an algebraic variety · *Trans. AMS* 157, 459-469; the original 1971 introduction of the logarithmic limit set Log_∞(V), which Mikhalkin's 2004 spine convergence identifies with the tropical hypersurface
TODO_REF
Kapranov — Amoebas over non-archimedean fields · 1995 manuscript / preprint; Kapranov's theorem identifying the non-archimedean amoeba with the tropical hypersurface (first proof; popularised in Einsiedler-Kapranov-Lind 2006)
TODO_REF
Fulton — Introduction to Toric Varieties · Princeton 1993; §3.1 (the normal fan of a polytope); §3.4 (the moment polytope as the image of the moment map on the toric variety)
TODO_REF
Itenberg-Mikhalkin-Shustin — Tropical Algebraic Geometry · Oberwolfach Seminars 35, Birkhauser 2009; §1.4 (Newton polytopes and tropical hypersurfaces); §2 (tropical curves and Mikhalkin's correspondence theorem)

Lean module

Codex.AlgGeom.Tropical.NewtonPolytopeAmoeba

Reviewer

TBD

Estimated time

beginner: 18m
intermediate: 45m
master: 80m