Tropical Geometry

⚡ Key Takeaways

Tropical geometry is the study of geometry over the tropical semiring $(\mathbb{R}\cup\{\infty\}, \min, +)$ , where classical algebraic varieties degenerate into piecewise-linear polyhedral complexes. Key result: algebraic varieties over a valued field have tropicalizations that are balanced rational polyhedral complexes of the same dimension, turning algebraic geometry into combinatorics. Applies to: enumerative geometry, mirror symmetry, phylogenetics, optimization, and computational algebra. Origins: named after Brazilian computer scientist Imre Simon (1980s); developed into geometry by Sturmfels, Mikhalkin, Maslov, and others in the 2000s.

💡 The Core Insight

Take the field of Puiseux series and look only at the lowest-order exponent (the valuation). Under this lens, multiplication $x\cdot y$ becomes addition of valuations $\mathrm{val}(x)+\mathrm{val}(y)$ , and addition $x+y$ becomes $\min(\mathrm{val}(x),\mathrm{val}(y))$ — because the smallest exponent dominates, unless cancellation occurs. The "unless cancellation" is exactly where the geometry lives: the tropical variety is the locus where the minimum is achieved at least twice. Everything piecewise-linear and combinatorial flows from this single dequantization.

DefinitionThe Tropical Semiring

The min-plus tropical semiring is $\mathbb{T}=(\mathbb{R}\cup\{+\infty\},\oplus,\odot)$ with

1

\quad x\oplus y := \min(x,y), \qquad x\odot y := x+y.

Derivation hint: Replace $(+,\times)$ on $\mathbb{R}_{>0}$ by their images under $\log_t$ as $t\to\infty$ ; the additive identity is $+\infty$ (the "tropical zero") and the multiplicative identity is $0$ (the "tropical one"). It is a commutative idempotent semiring ( $x\oplus x = x$ ) with no additive inverses. The max-plus convention $(\max,+)$ is isomorphic via $x\mapsto -x$ ; both appear in the literature.

DefinitionTropical Polynomials and Hypersurfaces

A tropical polynomial in variables $x_1,\dots,x_n$ is a finite tropical sum of monomials:

2

\quad p(\mathbf{x}) = \bigoplus_{\mathbf{a}\in A} c_{\mathbf{a}}\odot \mathbf{x}^{\odot \mathbf{a}} = \min_{\mathbf{a}\in A}\left( c_{\mathbf{a}} + \langle \mathbf{a},\mathbf{x}\rangle \right),

where $A\subset\mathbb{Z}^n$ is finite, $c_{\mathbf{a}}\in\mathbb{R}$ , and $\langle\mathbf{a},\mathbf{x}\rangle=\sum a_i x_i$ . Derivation hint: Expand each tropical monomial $c_{\mathbf a}\odot x_1^{\odot a_1}\odot\cdots = c_{\mathbf a}+a_1x_1+\cdots+a_nx_n$ using [1] repeatedly; the tropical sum collapses to the pointwise minimum of these affine-linear functions. The function $p$ is the lower envelope of finitely many affine functions, hence piecewise-linear, concave (in min-plus). The tropical hypersurface $V(p)$ is the non-smooth locus:

3

\quad V(p) = \left\{ \mathbf{x}\in\mathbb{R}^n : \text{the minimum in } [2] \text{ is attained at least twice} \right\}.

Derivation hint: This is the corner locus where the graph of $p$ bends; at such points two or more linear pieces meet, mirroring where the classical "vanishing" $f(\mathbf x)=0$ occurs after dequantization, since cancellation of leading terms requires equal lowest exponents.

DefinitionValuation and Tropicalization

Let $K$ be a field with a nonarchimedean valuation $\mathrm{val}:K^*\to\mathbb{R}$ (e.g. $K=\mathbb{C}\{\!\{t\}\!\}$ , Puiseux series, with $\mathrm{val}(\sum a_q t^q)=\min\{q : a_q\neq 0\}$ ). For $f=\sum_{\mathbf a} c_{\mathbf a}\,\mathbf x^{\mathbf a}\in K[x_1^{\pm},\dots,x_n^{\pm}]$ , the tropicalization is

4

\quad \mathrm{trop}(f)(\mathbf w) = \min_{\mathbf a}\left( \mathrm{val}(c_{\mathbf a}) + \langle \mathbf a, \mathbf w\rangle\right).

Derivation hint: Substitute a point $\mathbf x$ with $\mathrm{val}(x_i)=w_i$ into $f$ ; each monomial $c_{\mathbf a}\mathbf x^{\mathbf a}$ has valuation $\mathrm{val}(c_{\mathbf a})+\langle\mathbf a,\mathbf w\rangle$ , and the valuation of a sum is the minimum of the valuations (generically), so the whole expression tropicalizes to [4]. The tropicalization of a variety $X=V(I)\subset (K^*)^n$ is

5

\quad \mathrm{Trop}(X) = \overline{\{ (\mathrm{val}(x_1),\dots,\mathrm{val}(x_n)) : \mathbf x\in X\}} \subset \mathbb{R}^n.

Derivation hint: Apply the coordinatewise valuation map to the algebraic set and take the Euclidean closure; the Fundamental Theorem identifies this image with the intersection of tropical hypersurfaces $\bigcap_{f\in I} V(\mathrm{trop}(f))$ .

TheoremFundamental Theorem of Tropical Geometry (Kapranov / Speyer–Sturmfels)

Let $K$ be an algebraically closed field with a nontrivial valuation whose value group is dense in $\mathbb{R}$ , and let $X\subset(K^*)^n$ be a closed subvariety with ideal $I$ . Then the following three sets coincide: 1. $\mathrm{Trop}(X)$ , the closure of the image of $X$ under the valuation map (eq. [5]); 2. $\bigcap_{f\in I} V(\mathrm{trop}(f))$ , the intersection of all tropical hypersurfaces; 3. $\{\mathbf w : \mathrm{in}_{\mathbf w}(I) \text{ contains no monomial}\}$ , the locus of initial-ideal degenerations that are not "everything." Moreover, if $X$ is irreducible of dimension $d$ , then $\mathrm{Trop}(X)$ is the support of a pure $d$ -dimensional weighted balanced rational polyhedral complex, connected in codimension one.

Proof

Step 1 (Hypersurface case — Kapranov). For a single $f$ , show $\mathrm{trop}(f)$ vanishes (min attained twice) at $\mathbf w$ iff there is $\mathbf x\in V(f)$ with $\mathrm{val}(\mathbf x)=\mathbf w$ . The forward direction: if the min is attained uniquely by monomial $\mathbf a_0$ , then $f(\mathbf x)$ has a strictly dominant lowest-order term, so $f(\mathbf x)\neq 0$ , hence $\mathbf w\notin\mathrm{Trop}$ .

Step 2 (Reverse / lifting). If the min is attained at $\ge 2$ monomials, the initial form $\mathrm{in}_{\mathbf w}(f)$ over the residue field is a nonzero non-monomial polynomial; by Hensel's lemma / algebraic closedness one lifts a residue solution to an honest point $\mathbf x\in V(f)$ with the prescribed valuation. This uses density of the value group to realize arbitrary $\mathbf w$ .

Step 3 (Initial ideals). Define $\mathrm{in}_{\mathbf w}(I)=\langle \mathrm{in}_{\mathbf w}(f): f\in I\rangle$ . Show $\mathbf w\in\mathrm{Trop}(X)$ iff $\mathrm{in}_{\mathbf w}(I)$ is monomial-free; a monomial in the initial ideal forces some coordinate function to be invertible in the degeneration, contradicting nonemptiness.

Step 4 (Polyhedral structure). The Gröbner fan of $I$ partitions $\mathbb{R}^n$ into cones on which $\mathrm{in}_{\mathbf w}(I)$ is constant; $\mathrm{Trop}(X)$ is a union of (closures of) such cones, hence a rational polyhedral complex. Dimension equals $\dim X$ by upper-semicontinuity of fiber dimensions under the degeneration.

Step 5 (Balancing). Assign each maximal cone $\sigma$ a weight $m_\sigma$ (lattice multiplicity of $\mathrm{in}_{\mathbf w}(I)$ ). Around each codimension-one cone $\tau$ , the primitive integer normal vectors $u_i$ of adjacent maximal cones satisfy $\sum_i m_{\sigma_i} u_i \equiv 0 \pmod{L_\tau}$ , the balancing condition — proved by reduction to the one-dimensional Newton polygon / tropical Bézout case. $\blacksquare$

∎

GM

Grigory Mikhalkin

b. 1970

Mikhalkin transformed tropical geometry from a curiosity into a powerful enumerative tool. His 2005 Correspondence Theorem (Enumerative tropical algebraic geometry in $\mathbb{R}^2$ , Journal of the AMS) proved that Gromov–Witten / Severi-degree counts of complex plane curves equal weighted counts of tropical curves, reducing hard enumerative questions to lattice-path combinatorics. He also developed the theory of tropical amoebas, the amoeba–spine limit, and tropical curves as metric graphs satisfying a balancing condition, anchoring the field's connection to symplectic topology and the work of Maxim Kontsevich on mirror symmetry.

🧠 Intuition

Picture the amoeba of a complex curve $V(f)\subset(\mathbb{C}^*)^2$ : apply $\mathrm{Log}_t(z,w)=(\log_t|z|,\log_t|w|)$ to get a tentacled region in $\mathbb{R}^2$ . As the base $t\to\infty$ , the amoeba's fat tentacles shrink to their spine — a piecewise-linear graph with straight rays going to infinity in the directions dictated by the Newton polygon. That spine is the tropical curve. So tropicalization is a "logarithmic photograph taken from infinitely far away": curvature and area disappear, only the asymptotic skeleton survives. The balancing condition is conservation of momentum at each vertex — the weighted edge-directions sum to zero, exactly as the tentacles of an amoeba must balance.

Figure

Tropical line in the plane:

\mathrm{trop}(p)=\min(x,\,y,\,0)

, with its three rays meeting at the origin and the balancing condition

u_1+u_2+u_3=0

.

```svg

```

ExampleTropical line and stable intersection

Given: Two tropical lines $L_1=V(\min(x,y,0))$ (vertex at origin) and $L_2=V(\min(x+1,\,y-2,\,3))$ in $\mathbb{R}^2$ .

Step 1 — Vertex of $L_2$ . The three affine pieces $x+1,\;y-2,\;3$ pairwise tie at: $x+1=3\Rightarrow x=2$ , and $y-2=3\Rightarrow y=5$ ; check $x+1=y-2\Rightarrow y=x+3=5$ . ✓ Vertex of $L_2$ is $(2,5)$ .

Step 2 — Each tropical line has three rays in directions $(-1,0)$ won't appear; standard tropical line $V(\min(x,y,0))$ has rays in directions $(1,0),(0,1),(-1,-1)$ from its vertex. $L_1$ : vertex $(0,0)$ ; $L_2$ : vertex $(2,5)$ , same ray directions (translation does not change directions).

Step 3 — Stable intersection. Two general tropical lines meet in exactly one point, with multiplicity given by the lattice index $|\det[u\,|\,v]|$ of the crossing edge-directions. A horizontal ray of $L_1$ (direction $(1,0)$ ) crosses a downward $(-1,-1)$ ray of $L_2$ ; determinant $\left|\det\begin{pmatrix}1&-1\\0&-1\end{pmatrix}\right| = 1$ .

Boxed answer:

\boxed{\;L_1 \cdot_{\mathrm{st}} L_2 = 1 \text{ point, multiplicity } 1 \;\Longrightarrow\; \text{Tropical Bézout: } \deg L_1\cdot\deg L_2 = 1\cdot 1 = 1.\;}

ExampleTropicalizing a curve over Puiseux series

Given: $f = t\,x^2 + xy + t^3 y^2 + x + t\,y + t^5 \in K[x,y]$ , $K=\mathbb{C}\{\!\{t\}\!\}$ , $\mathrm{val}(t)=1$ .

Step 1 — Valuations of coefficients. $\mathrm{val}(t)=1,\ \mathrm{val}(1)=0,\ \mathrm{val}(t^3)=3,\ \mathrm{val}(t)=1,\ \mathrm{val}(t^5)=5$ . Monomials with exponents $\mathbf a$ : $(2,0)\!\to\!1$ , $(1,1)\!\to\!0$ , $(0,2)\!\to\!3$ , $(1,0)\!\to\!0$ , $(0,1)\!\to\!1$ , $(0,0)\!\to\!5$ .

Step 2 — Tropical polynomial (eq. [4]):

\mathrm{trop}(f)(w_1,w_2)=\min\big(1+2w_1,\; w_1+w_2,\; 3+2w_2,\; w_1,\; 1+w_2,\; 5\big).

Step 3 — Find the tropical curve (where min attained twice). Compare dominant terms in regions. For instance, the term $w_1$ ties with $w_1+w_2$ when $w_2=0$ (and both beat others); it ties with $5$ when $w_1=5$ . Computing the full subdivision of the Newton polygon $\mathrm{conv}\{(2,0),(1,1),(0,2),(1,0),(0,1),(0,0)\}$ dual to $\mathrm{trop}(f)$ gives the trivalent graph.

Step 4 — Genus from the dual subdivision. The number of interior lattice points of the regular subdivision equals the genus of the tropical curve. Here the Newton polygon (a triangle of size 2) has interior lattice point $(1,1)$ ? — it is on the boundary edges; with the induced subdivision the first Betti number (number of bounded loops) is computed from cycles in the graph.

Boxed answer:

\boxed{\;\mathrm{Trop}(V(f)) = V\big(\min(1{+}2w_1,\,w_1{+}w_2,\,3{+}2w_2,\,w_1,\,1{+}w_2,\,5)\big),\ \text{a balanced graph dual to the regular subdivision of }\mathrm{Newt}(f).\;}

⚠️ Common mistake

The tropical hypersurface is NOT the image of a "set of solutions." There is no equation $p(\mathbf x)=0$ in $\mathbb{T}$ in the naive sense — $\mathbb{T}$ has no additive inverses, so " $=0$ " is meaningless. The correct definition of vanishing is "the minimum is attained at least twice" (eq. [3]). Beginners who try to "solve $p(\mathbf x)=\infty$ " or " $p(\mathbf x)=0$ " get nonsense.

⚠️ Common mistake

Min-plus vs max-plus is a genuine sign convention. Whether monomials are summed by $\min$ or $\max$ depends on whether your valuation increases or decreases. Curves bend "downward" in one and "upward" in the other; balancing vectors flip sign. Always state the convention. A theorem proven in min-plus must have its inequalities reversed for max-plus.

Counterexample

Set-theoretic intersection $\neq$ stable intersection. Two tropical lines $L_1,L_2$ in $\mathbb{R}^2$ can overlap along an entire ray (positive-dimensional intersection) when arranged non-generically — e.g. both with vertex at the origin and a shared $(1,0)$ ray. The set-theoretic intersection is then a half-line, violating the naive expectation of one point. The fix is the stable intersection $L_1\cdot_{\mathrm{st}}L_2 = \lim_{\epsilon\to 0}(L_1\cap (L_2+\epsilon v))$ , perturbing generically; this restores Bézout's count $=1$ . The lesson: tropical varieties must be intersected stably, not set-theoretically.

connection

Tropical geometry sits at a crossroads. To Algebraic Geometry it is a degeneration / toric-limit theory: the tropical variety records the structure of a toric degeneration and the dual subdivision of a Newton polytope. To Polyhedral Combinatorics and Matroid Theory it provides geometric models — the Bergman fan of a matroid is precisely the tropicalization of a linear space, linking to the Hodge theory of matroids (Adiprasito–Huh–Katz). To Convex Geometry it connects via the Legendre transform (tropical polynomials are support functions). To Nonarchimedean Geometry the work of Payne shows $\mathrm{Trop}(X)$ is a limit of Berkovich analytifications: $X^{\mathrm{an}}=\varprojlim \mathrm{Trop}(X)$ . And to <a class="wiki-link" href="/wiki/symplectic-geometry">Symplectic Geometry</a> and mirror symmetry, tropical curves model holomorphic disks counted in Gromov–Witten theory.

🔬 Where this appears

⏳ History

- 1979–1980s: Imre Simon uses the min-plus semiring to solve decision problems in automata theory; the adjective "tropical" is coined by French colleagues honoring Simon's Brazilian origin.

- 1971 (precursor): George Bergman introduces the "logarithmic limit set" of an algebraic variety — essentially

\mathrm{Trop}(X)

avant la lettre.

- 1990s: Victor Maslov and the Russian school develop "idempotent analysis" / "dequantization," viewing

(\max,+)

as the

\hbar\to 0

classical limit of quantum mechanics (Maslov dequantization).

- 2000: Mikael Passare and Hans Rullgård study amoebas and their spines, formalizing the amoeba

\to

tropical limit.

- 2002: Mike Develin, Bernd Sturmfels and collaborators define tropical convexity and tropical polytopes; Speyer–Sturmfels prove tropical Grassmannian results.

- 2004 (Kapranov): the hypersurface case of the Fundamental Theorem is established (often attributed to Kapranov via Einsiedler–Kapranov–Lind and Speyer–Sturmfels).

- 2005: Grigory Mikhalkin proves the Correspondence Theorem, the field's defining result, linking tropical and complex enumerative geometry.

- 2006–2008: Andreas Gathmann, Hannah Markwig build tropical intersection theory and moduli of tropical curves rigorously.

- 2015–2018: Karim Adiprasito, June Huh, Eric Katz use the tropical/Bergman-fan model of matroids to prove the Heron–Rota–Welsh conjecture (log-concavity), a landmark application earning Huh the 2022 Fields Medal.

open-problem

Several deep questions remain open or only partially resolved. (1) Realizability: not every abstract balanced weighted polyhedral complex is the tropicalization of an actual algebraic variety; classifying which tropical curves/surfaces lift over a valued field is subtle, governed by obstructions in the cohomology of the Berkovich space and remains open in dimension $\ge 2$ . (2) Tropical Hodge theory: building a satisfactory theory of tropical $(p,q)$ -forms, Dolbeault cohomology, and a tropical Lefschetz/Hodge–Riemann package (Itenberg–Katzarkov–Mikhalkin–Zharkov, Babaee–Huh) — the Hodge conjecture for tropical cycles is open. (3) Higher-genus correspondence: extending Mikhalkin's theorem to arbitrary target varieties and to refined (real, $q$ -deformed Block–Göttsche) invariants is active. (4) Tropical scheme theory: finding the "right" foundational category — semiring schemes, blueprints (Lorscheid), or the bend relations of Giansiracusa–Giansiracusa — so that tropicalization is a functor with a Grothendieck topology, is still being settled.

Formal Statement

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also