• Keine Ergebnisse gefunden

Example 5.25. Let C be the category from Figure 5.3.

x

f1

xx

f2

h11

h12

&&

y

h1

h2

&&

y1

f21

f11

xxz

Figure 5.3: An acyclic category

The unlabeled morphisms are given by

γ11h1f1f11 h11 h2f1f11h12γ12 γ22h2f2f21 h12 h1f2f21h11γ21.

Figure 5.4 shows the realizations ofC with respect to Gmin andGmax. On the left hand side we have the realization with respect toGmin, where we have only labeled 0– and1–cells. On the right hand side we have the realization with respect toGmax, where the morphisms in the picture represent their respective closures. The reader may convince himself that these yield indeed all objects inDpC,Gmaxq.

f1

h11 h12

f1

f2

h12 h11

f2

h1

f11

h2

f21 γ11 γ12

γ22 γ21

idy

idx idy1

idx

idy

idx idy1

idx

f1

h11 h12

f1

f2

h12 h11

f2

h1

f11

h2

f21 idz

idy

idx idy1

idx

idy

idx idy1

idx

xγ11y xγ12y

xγ22y

xγ21y

Figure 5.4: The realizations of the category from Figure 5.3 with respect toGmin

(left) andGmax (right)

Proof. Elementary calculations show that pairs of factorizations, for which the pairs of first and second coordinates satisfy the properties of Definition 5.8, do satisfy those properties for the product.

Since trivial factorizations of the product category correspond to pairs of trivial factorizations of the factor categories, products of decomposition sets contain all trivial decompositions of the product.

Remark 5.27. The product of maximal decomposition sets is the maximal decom-position set of the product category.

Proposition 5.28. The product of minimal decomposition sets is the minimal decomposition set of the product category if and only if the factors are discrete categories.

Proof. Pairs of trivial decompositions correspond to trivial decompositions of the product if and only if either both first or both second coordinates are identities.

If and only if both factor categories have non-identity morphisms this is not the case for any such pair.

Theorem 5.29. For decomposition sets G1 of C1 and G2 of C2 and a chain C inC1C2

xCyG1G21pCqyG12pCqyG2

holds, whereπ1, π2 denote the projection functors of the product category.

Proof. Let us show the first inclusion by induction on the size ofG1 andG2. For the start we considerGminpC1qGminpC2q. Decompositions corresponding to pairs of trivial decompositions indeed generate new morphisms, but they change neither π1pCqnor π2pCq. Thus even xCyGminpC1qGminpC2q π1pCq π2pCqis true. For the induction itself set G11 : G1Y tpf1, h1qu. If xCyG11G2 xCyG1G2, by the induction hypothesis, this equals xπ1pCqyG12pCqyG2, which is a subcategory ofxπ1pCqyG112pCqyG2. So w.l.o.g. assume, for some morphismf2 ofxπ2pCqyG2

the pair pf1, f2q is a morphism of xCyG11G2 which does not exist in xCyG1G2. So there has to be a decomposition (for which w.l.o.g. f1, f2 are both the front morphisms)ppf1, f2q,ph1, h2qq PG11G2zG1G2such thatph1f1, h2f2qdoes exist inxCyG1G21pCqyG12pCqyG2. Thus, in particular,h1f1is a morphism of xπ1pCqyG1. This gives the existence off1 inxπ1pCqyG11. Sopf1, f2qis a morphism ofxπ1pCqyG112pCqyG2 which proves the first inclusion.

For the second inclusion note that for a decomposition pf1, h1qof G1, the pair ppf1,idq,ph1,idqqis a decomposition of the product G1G2. Thus any morphism f1 of xπ1pCqyG1 gives rise to a morphism pf1,idq of xCyG1G2, no matter which identity is meant. Similarly, we get morphisms pid, f2q of xCyG1G2 from mor-phismsf2ofxπ2pCqyG2. Therefore arbitrary morphismspf1, f2q pf1,idq pid, f2q ofxπ1pCqyG12pCqyG2 belong toxCyG1G2.

So the set of morphisms of the two subcategories is identical.

Corollary 5.30. Let G1,G2 be decomposition sets of C1 resp. C2, then

DpC1C2,G1G2q DpC1,G1q DpC2,G2q

via the canonical isomorphisms of taking products and projecting to coordinates.

Proof. Clearly the functors are inverse to each other. So it is enough to check that they are well defined. For chains pf1, . . . , fnq of C1 and ph1, . . . , hmq of C2, Theorem 5.29 applied to the chainC ppf1, h1q, . . . ,pfn, h1q, . . . ,pfn, hmqq imme-diately shows that taking products is well defined. By Theorem 5.29 we obtain π1pxCyG1G2q xπ1pCqyG1 andπ2pxCyG1G2q xπ2pCqyG2 So the projection func-tors are well defined, too.

Example 5.31. Consider Example 5.25. The category shown in Figure 5.4 is isomorphic to a product of the (up to ismorphism unique) acyclic category with two objects and two non-identity morphisms with itself. Since for this category there are no proper decompositions, there is only one decomposition set and real-izations of the decomposition complex are subdivided spheresS1. Thus, indeed, the decomposition complex of the product is a subdivided torusS1S1, see Figure 5.4.

Example 5.32. Consider Example 5.14. Realizations of decomposition complexes of the factors are subdivided spheresS1resp. the interval D1. Thus the decompo-sition complexes of the product is a subdivided cylinderS1D1. Figure 5.5 shows realizations of the decomposition complexes with respect to the maximal (left) and the minimal (right) decomposition set.

Theorem 5.33. LetG1be a decomposition set ofC1 and letG2be a decomposition set ofC2. Then

DpC1

ºC2,G1YG2q DpC1,G1q º

DpC2,G2q.

h1 f11

h f21

f2

f1

idx

idy1 idz

idy

h1 f11

h f21

f2 γ2

f1 γ1

idx

idy1 idz

idy

Figure 5.5: Realizations of decomposition complexes of Example 5.14 with respect to the maximal(left) and the minimal(right) decomposition set.

Proof. Since decompositions of the coproduct correspond to decompositions of the summands and chains are either contained in C1 or in C2, the statement is true, like in the case of nerves.

Remark 5.34. The properties of symmetry and downwards closedness are pre-served under taking products. The same is true for realizations. Coordinate-wise products of G– andG1–realizations are GG1-realizations.

We end with a theorem from Kozlov [23], which becomes a corollary of our theory.

Corollary 5.35. [23, Thm. 10.21] For arbitrary acyclic categoriesC andD

|DpCD,GminpCDqq| subdivides |DpC,GminpCqq| |DpD,GminpDqq|.

Proof. Since GminpCDq „ GminpCq GminpDq, we can apply Corollary 5.24.

Then it is just left to note that as sets

|DpC,GminpCqq| |DpD,GminpDqq| |DpCD,GminpCq GminpDqq|

holds by Corollary 5.30.

6 Summary

We understand how different subdivisions of Bergman complexes are related to each other. The unification of the different languages of chains, nested sets and matroid types enables us to generalize these concepts to arbitrary posets and even acyclic categories. Since all comes down to product decompositions of intervals, the main advantage of decomposition complexes over order complexes is their closedness under products.

Whenever chains of specific posets have combinatorial interpretations and/or there are suitable product structures, it is reasonable to consider decomposition complexes to gain deeper insights. Potentially, chains that are glued together can be characterized in terms of these combinatorial interpretations - like combinatorial types of phylogenetic trees.

7 Bibliography

[1] Federico Ardila and Caroline J. Klivans. The Bergman complex of a matroid and phylogenetic trees. J. Combin. Theory Ser. B, 96(1):38–49, 2006.

[2] George M. Bergman. The logarithmic limit-set of an algebraic variety. Trans.

Amer. Math. Soc., 157:459–469, 1971.

[3] Robert Bieri and John R. J. Groves. The geometry of the set of characters induced by valuations. J. Reine Angew. Math., 347:168–195, 1984.

[4] Anders Bj¨orner. On the homology of geometric lattices. Algebra Universalis, 14(1):107–128, 1982.

[5] Joseph Bonin, Anna de Mier, and Marc Noy. Lattice path matroids: enumera-tive aspects and Tutte polynomials.J. Combin. Theory Ser. A, 104(1):63–94, 2003.

[6] Joseph E. Bonin. Lattice path matroids: the excluded minors. J. Combin.

Theory Ser. B, 100(6):585–599, 2010.

[7] Joseph E. Bonin and Anna de Mier. Lattice path matroids: structural prop-erties. European J. Combin., 27(5):701–738, 2006.

[8] Roman Bruckner and Martin Dlugosch. Decomposition complexes for acyclic categories. preprint in preparation.

[9] James A. Carlson. Polyhedral resolutions of algebraic varieties. Trans. Amer.

Math. Soc., 292(2):595–612, 1985.

[10] Ralph L. Cohen, John D. S. Jones, and Graeme B. Segal. Stability for holo-morphic spheres and Morse theory. InGeometry and topology: Aarhus (1998), volume 258 ofContemp. Math., pages 87–106. Amer. Math. Soc., Providence, RI, 2000.

[11] Giacomo d’Antonio and Emanuele Delucchi. A Salvetti complex for toric arrangements and its fundamental group. Int. Math. Res. Not. IMRN, (15):3535–3566, 2012.

[12] Emanuele Delucchi and Martin Dlugosch. Bergman complexes of lattice path matroids. preprint, July 2012, arxiv:1207.4700v1.

[13] Martin Dlugosch. Decomposition complexes. preprint, January 2013, arXiv:0707.3168v1.

[14] Martin Dlugosch. New light on Bergman complexes by decomposing matroid types. Order, 2013, doi:10.1007/s11083-013-9290-8.

[15] Eva M. Feichtner. De Concini-Procesi wonderful arrangement models: a dis-crete geometer’s point of view. InCombinatorial and computational geometry, volume 52 of Math. Sci. Res. Inst. Publ., pages 333–360. Cambridge Univ.

Press, Cambridge, 2005.

[16] Eva M. Feichtner and Dmitry N. Kozlov. Incidence combinatorics of resolu-tions. Selecta Math. (N.S.), 10(1):37–60, 2004.

[17] Eva M. Feichtner and Irene M¨uller. On the topology of nested set complexes.

Proc. Amer. Math. Soc., 133(4):999–1006 (electronic), 2005.

[18] Eva M. Feichtner and Bernd Sturmfels. Matroid polytopes, nested sets and Bergman fans. Port. Math. (N.S.), 62(4):437–468, 2005.

[19] Eva M. Feichtner and Sergey Yuzvinsky. Chow rings of toric varieties defined by atomic lattices. Invent. Math., 155(3):515–536, 2004.

[20] Jon Folkman. The homology groups of a lattice.J. Math. Mech., 15:631–636, 1966.

[21] Israel M. Gel1fand, Mark Goresky, Robert D. MacPherson, and Vera V.

Serganova. Combinatorial geometries, convex polyhedra, and Schubert cells.

Adv. in Math., 63(3):301–316, 1987.

[22] Mark Goresky and Robert D. MacPherson.Stratified Morse theory, volume 14 ofErgebnisse der Mathematik und ihrer Grenzgebiete (3) [Results in Mathe-matics and Related Areas (3)]. Springer-Verlag, Berlin, 1988.