In mathematics, tropical geometry is the study of polynomials and their geometric properties when addition is replaced with minimization and multiplication is replaced with ordinary addition:
,
.
So for example, the classical polynomial would become . Such polynomials and their solutions have important applications in optimization problems, for example the problem of optimizing departure times for a network of trains.
Tropical geometry is a variant of algebraic geometry in which polynomial graphs resemble piecewise linear meshes, and in which numbers belong to the tropical semiring instead of a field. Because classical and tropical geometry are closely related, results and methods can be converted between them. Algebraic varieties can be mapped to a tropical counterpart and, since this process still retains some geometric information about the original variety, it can be used to help prove and generalize classical results from algebraic geometry, such as the Brill–Noether theorem or computing Gromov Witten inavriants, using the tools of tropical geometry.[1]
History
The basic ideas of tropical analysis were developed independently using the same notation by mathematicians working in various fields.[2] The central ideas of tropical geometry appeared in different forms in a number of earlier works. For example, Victor Pavlovich Maslov introduced a tropical version of the process of integration. He also noticed that the Legendre transformation and solutions of the Hamilton–Jacobi equation are linear operations in the tropical sense.[3] However, only since the late 1990s has an effort been made to consolidate the basic definitions of the theory. This was motivated by its application to enumerative algebraic geometry, with ideas from Maxim Kontsevich[4] and works by Grigory Mikhalkin[5] among others.
Tropical geometry is based on the tropical semiring. This is defined in two ways, depending on max or min convention.
The min tropical semiring is the semiring, with the operations:
,
.
The operations and are referred to as tropical addition and tropical multiplication respectively. The identity element for is , and the identity element for is 0.
Similarly, the max tropical semiring is the semiring , with operations:
,
.
The identity element for is , and the identity element for is 0.
These semirings are isomorphic, under negation , and generally one of these is chosen and referred to simply as the tropical semiring. Conventions differ between authors and subfields: some use the min convention, some use the max convention.
The tropical semiring operations model how valuations behave under addition and multiplication in a valued field.
Some common valued fields encountered in tropical geometry (with min convention) are:
or with the trivial valuation, for all .
or its extensions with the p-adic valuation, for a and b coprime to p.
The field of Laurent series (integer powers), or the field of (complex) Puiseux series, with valuation returning the smallest exponent of t appearing in the series.
Tropical polynomials
A tropical polynomial is a function that can be expressed as the tropical sum of a finite number of monomial terms. A monomial term is a tropical product (and/or quotient) of a constant and variables from . Thus a tropical polynomial F is the minimum of a finite collection of affine-linear functions in which the variables have integer coefficients, so it is concave, continuous, and piecewise linear.[8]
Given a polynomial f in the Laurent polynomial ring where K is a valued field, the tropicalization of f, denoted , is the tropical polynomial obtained from f by replacing multiplication and addition by their tropical counterparts and each constant in K by its valuation. That is, if
,
then
.
The set of points where a tropical polynomial F is non-differentiable is called its associated tropical hypersurface, denoted (in analogy to the vanishing set of a polynomial). Equivalently, is the set of points where the minimum among the terms of F is achieved at least twice. When for a Laurent polynomial f, this latter characterization of reflects the fact that at any solution to , the minimum valuation of the terms of f must be achieved at least twice in order for them all to cancel.[9]
Tropical varieties
Definitions
For X an algebraic variety in the algebraic torus, the tropical variety of X or tropicalization of X, denoted , is a subset of that can be defined in several ways. The equivalence of these definitions is referred to as the Fundamental Theorem of Tropical Geometry.[9]
Intersection of tropical hypersurfaces
Let be the ideal of Laurent polynomials that vanish on X in . Define
When X is a hypersurface, its vanishing ideal is a principal ideal generated by a Laurent polynomial f, and the tropical variety is precisely the tropical hypersurface .
Every tropical variety is the intersection of a finite number of tropical hypersurfaces. A finite set of polynomials is called a tropical basis for X if is the intersection of the tropical hypersurfaces of . In general, a generating set of is not sufficient to form a tropical basis. The intersection of a finite number of a tropical hypersurfaces is called a tropical prevariety and in general is not a tropical variety.[9]
Initial ideals
Choosing a vector in defines a map from the monomial terms of to by sending the term m to . For a Laurent polynomial , define the initial form of f to be the sum of the terms of f for which is minimal. For the ideal , define its initial ideal with respect to to be
.
Then define
.
Since we are working in the Laurent ring, this is the same as the set of weight vectors for which does not contain a monomial.
When K has trivial valuation, is precisely the initial ideal of with respect to the monomial order given by a weight vector . It follows that is a subfan of the Gröbner fan of .
Image of the valuation map
Suppose that X is a variety over a field K with valuation v whose image is dense in (for example a field of Puiseux series). By acting coordinate-wise, v defines a map from the algebraic torus to . Then define
,
where the overline indicates the closure in the Euclidean topology. If the valuation of K is not dense in , then the above definition can be adapted by extending scalars to larger field which does have a dense valuation.
If X is a variety over , can be considered as the limiting object of the amoeba as the base t of the logarithm map goes to infinity.[11]
Polyhedral complex
The following characterization describes tropical varieties intrinsically without reference to algebraic varieties and tropicalization. A set V in is an irreducible tropical variety if it is the support of a weighted polyhedral complex of pure dimension d that satisfies the zero-tension condition and is connected in codimension one. When d is one, the zero-tension condition means that around each vertex, the weighted-sum of the out-going directions of edges equals zero. For higher dimension, sums are taken instead around each cell of dimension after quotienting out the affine span of the cell.[8] The property that V is connected in codimension one means for any two points lying on dimension d cells, there is a path connecting them that does not pass through any cells of dimension less than .[12]
Tropical curves
The study of tropical curves (tropical varieties of dimension one) is particularly well developed and is strongly related to graph theory. For instance, the theory of divisors of tropical curves are related to chip-firing games on graphs associated to the tropical curves.[13]
Many classical theorems of algebraic geometry have counterparts in tropical geometry, including:
Oleg Viro used tropical curves to classify real curves of degree 7 in the plane up to isotopy. His method of patchworking gives a procedure to build a real curve of a given isotopy class from its tropical curve.
Applications
A tropical line appeared in Paul Klemperer's design of auctions used by the Bank of England during the financial crisis in 2007.[17] Yoshinori Shiozawa defined subtropical algebra as max-times or min-times semiring (instead of max-plus and min-plus). He found that Ricardian trade theory (international trade without input trade) can be interpreted as subtropical convex algebra.[18][non-primary source needed] Tropical geometry has also been used for analyzing the complexity of feedforward neural networks with ReLU activation.[19]
Moreover, several optimization problems arising for instance in job scheduling, location analysis, transportation networks, decision making and discrete event dynamical systems can be formulated and solved in the framework of tropical geometry.[20] A tropical counterpart of the Abel–Jacobi map can be applied to a crystal design.[21] The weights in a weighted finite-state transducer are often required to be a tropical semiring. Tropical geometry can show self-organized criticality.[22]
Tropical geometry has also found applications in several topics within theoretical high energy physics. In particular, tropical geometry has been used to drastically simplify string theory amplitudes to their field-theoretical limits [23] and has found connections to constructions such as the Amplituhedron[24] and tropological (Carrollian) sigma models.[25]
^See Cuninghame-Green, Raymond A. (1979). Minimax algebra. Lecture Notes in Economics and Mathematical Sciences. Vol. 166. Springer. ISBN978-3-540-09113-4 and references therein.
^Zhang, Liwen; Naitzat, Gregory; Lim, Lek-Heng (2018). "Tropical Geometry of Deep Neural Networks". Proceedings of the 35th International Conference on Machine Learning. 35th International Conference on Machine Learning. pp. 5824–5832.
^Krivulin, Nikolai (2014). "Tropical optimization problems". In Leon A. Petrosyan; David W. K. Yeung; Joseph V. Romanovsky (eds.). Advances in Economics and Optimization: Collected Scientific Studies Dedicated to the Memory of L. V. Kantorovich. New York: Nova Science Publishers. pp. 195–214. arXiv:1408.0313. ISBN978-1-63117-073-7.
^Sunada, T. (2012). Topological Crystallography: With a View Towards Discrete Geometric Analysis. Surveys and Tutorials in the Applied Mathematical Sciences. Vol. 6. Springer Japan. ISBN9784431541769.
Maslov, Victor (1986). "New superposition principle for optimization problems", Séminaire sur les Équations aux Dérivées Partielles 1985/6, Centre de Mathématiques de l’École Polytechnique, Palaiseau, exposé 24.
Maslov, Victor (1987). "Méthodes Opératorielles". Moscou, Mir, 707 p. (See Chapter 8, Théorie linéaire sur semi moduli, pp. 652–701).
Einsiedler, Manfred; Kapranov, Mikhail; Lind, Douglas (2006). "Non-archimedean amoebas and tropical varieties". J. Reine Angew. Math. 601: 139–157. arXiv:math/0408311. Bibcode:2004math......8311E.
Gathmann, Andreas (2006). "Tropical algebraic geometry". arXiv:math/0601322v1.
Gross, Mark (2010). Tropical geometry and mirror symmetry. Providence, R.I.: Published for the Conference Board of the Mathematical Sciences by the American Mathematical Society with support from the National Science Foundation. ISBN9780821852323.
Theobald, Thorsten (2003). "First steps in tropical geometry". arXiv:math/0306366v2.
Further reading
Amini, Omid; Baker, Matthew; Faber, Xander, eds. (2013). Tropical and non-Archimedean geometry. Bellairs workshop in number theory, tropical and non-Archimedean geometry, Bellairs Research Institute, Holetown, Barbados, USA, May 6–13, 2011. Contemporary Mathematics. Vol. 605. Providence, RI: American Mathematical Society. ISBN978-1-4704-1021-6. Zbl1281.14002.