![Page 1: HodgeTheoryofMatroids - American Mathematical SocietyΒ Β· The Hodge-Riemann relations for the algebra π΄β(π), where πis the matroid attached to as in the section βMatroidsβ](https://reader030.vdocuments.us/reader030/viewer/2022020205/5c1b84ea09d3f2ff0d8c15a3/html5/thumbnails/1.jpg)
Hodge Theory of MatroidsEditorβs Note: Matt Baker is speaking on this topic in the Current Events Bulletin Lecture at the January 2017 JointMathematics Meetings.
Karim Adiprasito, June Huh, and Eric KatzCommunicated by Benjamin Braun
IntroductionLogarithmic concavity is a property of a sequence ofreal numbers, occurring throughout algebraic geometry,convex geometry, and combinatorics. A sequence ofpositive numbers π0,β¦ ,ππ is log-concave if
π2π β₯ ππβ1ππ+1 for all π.
This means that the logarithms, log(ππ), form a concavesequence. The condition implies unimodality of the se-quence (ππ), a property easier to visualize: the sequenceis unimodal if there is an index π such that
π0 β€ β― β€ ππβ1 β€ ππ β₯ ππ+1 β₯ β― β₯ ππ.We will discuss our work on establishing log-concavity
of various combinatorial sequences, such as the coeffi-cients of the chromatic polynomial of graphs and theface numbers of matroid complexes. Our method is moti-vated by complex algebraic geometry, in particular Hodgetheory. From a given combinatorial object π (a matroid),we construct a graded commutative algebra over the realnumbers
π΄β(π) =π
β¨π=0
π΄π(π),
Karim Adiprasito is assistant professor of mathematics at HebrewUniversity of Jerusalem. His e-mail address is [email protected].
June Huh is research fellow at the Clay Mathematics Institute andVeblen Fellow at Princeton University and the Institute for Ad-vanced Study. His e-mail is [email protected].
Eric Katz is assistant professor of mathematics at the Ohio StateUniversity. His e-mail address is [email protected].
For permission to reprint this article, please contact:[email protected]: http://dx.doi.org/10.1090/noti1463
which satisfies analogues of PoincarΓ© duality, the hard Lef-schetz theorem, and the Hodge-Riemann relations for thecohomology of smooth projective varieties. Log-concavitywill be deduced from the Hodge-Riemann relations forπ. We believe that behind any log-concave sequencethat appears in nature there is such a βHodge structureβresponsible for the log-concavity.
Coloring GraphsGeneralizing earlier work of George Birkhoff, in 1932Hassler Whitney introduced the chromatic polynomial ofa connected graph πΊ as the function on β defined by
ππΊ(π) = |{proper colorings of πΊ using π colors}|.In other words, ππΊ(π) is the number of ways to colorthe vertices of πΊ using π colors so that the endpoints ofevery edge have different colors. Whitney noticed that thechromatic polynomial is indeed a polynomial. In fact, wecan write
ππΊ(π)/π = π0(πΊ)ππ βπ1(πΊ)ππβ1 +β―+ (β1)πππ(πΊ)for some positive integers π0(πΊ),β¦ ,ππ(πΊ), where π isone less than the number of vertices of πΊ.
Example 1. The square graph
β’ β’
β’ β’has the chromatic polynomial 1π4 β4π3 + 6π2 β 3π.
The chromatic polynomial was originally devised asa tool for attacking the Four Color Problem, but soonit attracted attention in its own right. Ronald Readconjectured in 1968 that the coefficients of the chromaticpolynomial form a unimodal sequence for any graph.
26 Notices of the AMS Volume 64, Number 1
![Page 2: HodgeTheoryofMatroids - American Mathematical SocietyΒ Β· The Hodge-Riemann relations for the algebra π΄β(π), where πis the matroid attached to as in the section βMatroidsβ](https://reader030.vdocuments.us/reader030/viewer/2022020205/5c1b84ea09d3f2ff0d8c15a3/html5/thumbnails/2.jpg)
A few years later, Stuart Hoggar conjectured that thecoefficients in fact form a log-concave sequence:
ππ(πΊ)2 β₯ ππβ1(πΊ)ππ+1(πΊ) for any π and πΊ.The graph theorist William Tutte quipped, βIn compensa-tion for its failure to settle the FourColourConjecture, [thechromatic polynomial] offers us the Unimodal Conjecturefor our further bafflement.β
The chromatic polynomial can be computed using thedeletion-contraction relation: if πΊ\π is the deletion of anedge π from πΊ and πΊ/π is the contraction of the sameedge, then
ππΊ(π) = ππΊ\π(π) β ππΊ/π(π).The first term counts the proper colorings of πΊ, thesecond term counts the otherwise-proper colorings of πΊwhere the endpoints of π are permitted to have the samecolor, and the third term counts the otherwise-propercolorings of πΊ where the endpoints of π are mandated tohave the same color. Note that, in general, the sum of twolog-concave sequences is not a log-concave sequence.
Example 2. To compute the chromatic polynomial of thesquare graph above, we write
β’ β’
β’ β’=
β’ β’
β’ β’-
β’@@
@
β’ β’and use
ππΊ\π(π) = π(π β 1)3, ππΊ/π(π) = π(π β 1)(π β 2).The Hodge-Riemann relations for the algebra π΄β(π),
where π is the matroid attached to πΊ as in the sectionβMatroidsβ below, imply that the coefficients of thechromatic polynomial of πΊ form a log-concave sequence.This is in contrast to a result of Alan Sokal that the set ofroots of the chromatic polynomials of all graphs is densein the complex plane.
Counting Independent SubsetsLinear independence is a fundamental notion in algebraand geometry: a collection of vectors is linearly indepen-dent if no nontrivial linear combination sums to zero.How many linearly independent collections of π vectorsare there in a given configuration of vectors? Write π΄ fora finite subset of a vector space and ππ(π΄) for the numberof independent subsets of π΄ of size π.Example 3. If π΄ is the set of all nonzero vectors in thethree-dimensional vector space over the field with twoelements, then
π0 = 1, π1 = 7, π2 = 21, π3 = 28.Examples suggest a pattern leading to a conjecture of
Dominic Welsh:ππ(π΄)2 β₯ ππβ1(π΄)ππ+1(π΄) for any π and π΄.
For any small specific case, the conjecture can be verifiedby computing the ππ(π΄)βs by the deletion-contraction rela-tion: if π΄\π£ is the deletion of a nonzero vector π£ from π΄and π΄/π£ is the projection of π΄ in the direction of π£, then
ππ(π΄) = ππ(π΄\π£) + ππβ1(π΄/π£).
The first term counts the number of independent subsetsof size π, the second term counts the independent subsetsof size π not containing π£, and the third term countsthe independent subsets of size π containing π£. As inthe case of graphs, we notice the apparent interferencebetween the log-concavity conjecture and the additivenature of ππ(π΄). The sum of two log-concave sequences is,in general, not log-concave. The conjecture suggests a newdescription for the numbers ππ(π΄) and a correspondingstructure of π΄.
MatroidsIn the 1930s Hassler Whitney observed that severalnotions in graph theory and linear algebra fit together ina common framework, that of matroids. This observationstarted a new subject with applications to a wide rangeof topics such as characteristic classes, optimization, andmoduli spaces, to name a few.
Let πΈ be a finite set. A matroid π on πΈ is a collectionof subsets of πΈ, called flats of π, satisfying the followingaxioms:(1) If πΉ1 and πΉ2 are flats of π, then their intersection is
a flat of π.(2) If πΉ is a flat of π, then any element of πΈ\πΉ is
contained in exactly one flat of π covering πΉ.(3) The ground set πΈ is a flat of π.
Here, a flat of π is said to cover πΉ if it is minimal amongthe flats of π properly containing πΉ. For our purposes,we may assume that π is loopless:
(1) The empty subset of πΈ is a flat of π.Every maximal chain of flats of π has the same length,
Matroids encodea combinatorial
structurecommon tographs and
vectorconfigurations
and this common lengthis called the rank ofπ. We write π\π forthe matroid obtainedby deleting π from theflats of π and π/π forthe matroid obtained bydeleting π from the flatsof π containing π. Whenπ1 is a matroid on πΈ1,π2 is a matroid on πΈ2,and πΈ1 β© πΈ2 is empty,the direct sum π1 β π2is defined to be the ma-troid on πΈ1 βͺ πΈ2 whoseflats are all sets of theform πΉ1 βͺπΉ2, where πΉ1 is a flat of π1 and πΉ2 is a flat of π2.
Matroids encode a combinatorial structure common tographs and vector configurations. If πΈ is the set of edgesof a finite graph πΊ, call a subset πΉ of πΈ a flat when there isno edge in πΈ\πΉ whose endpoints are connected by a pathin πΉ. This defines a graphic matroid on πΈ. If πΈ is a finitesubset of a vector space, call a subset πΉ of πΈ a flat whenthere is no vector in πΈ\πΉ that is contained in the linearspan of πΉ. This defines a linear matroid on πΈ.
January 2017 Notices of the AMS 27
![Page 3: HodgeTheoryofMatroids - American Mathematical SocietyΒ Β· The Hodge-Riemann relations for the algebra π΄β(π), where πis the matroid attached to as in the section βMatroidsβ](https://reader030.vdocuments.us/reader030/viewer/2022020205/5c1b84ea09d3f2ff0d8c15a3/html5/thumbnails/3.jpg)
Example 4. Write πΈ = {0, 1, 2, 3} for the set of edges ofthe square graph πΊ in Example 1. The graphic matroid πon πΈ attached to πΊ has flats
β ,{0}, {1}, {2}, {3}, {0, 1}, {0, 2}, {0, 3},{1, 2}, {1, 3}, {2, 3}, {0, 1, 2, 3}.
Example 5. Write πΈ = {0, 1, 2, 3, 4, 5, 6} for the configura-tion of vectors π΄ in Example 3. The linear matroid π onπΈ attached to π΄ has flats
β ,{0}, {1}, {2}, {3}, {4}, {5}, {6},{1, 2, 3}, {3, 4, 5}, {1, 5, 6}, {0, 1, 4}, {0, 2, 5}, {0, 3, 6},
{2, 4, 6}, {0, 1, 2, 3, 4, 5, 6}.A linear matroid that arises from a subset of a vector
space over a field π is said to be realizable over π.Not surprisingly, this notion is sensitive to the field π.A matroid may arise from a vector configuration overone field, while no such vector configuration exists overanother field. Many matroids are not realizable over anyfield.
Among the rank 3 loopless matroids pictured above,where rank 1 flats are represented by points and rank2 flats containing more than two points are representedby lines, the first is realizable over π if and only if thecharacteristic of π is 2, the second is realizable over π ifand only if the characteristic of π is not 2, and the thirdis not realizable over any field.
The characteristic polynomial ππ(π) of a matroid π isa generalization of the chromatic polynomial ππΊ(π) of agraph πΊ. It can be recursively defined using the followingrules:(1) If π is the direct sum π1 βπ2, then
ππ(π) = ππ1(π) ππ2(π).(2) If π is not a direct sum, then, for any π,
ππ(π) = ππ\π(π) β ππ/π(π).(3) If π is the rank 1 matroid on {π}, then
ππ(π) = πβ 1.(4) If π is the rank 0 matroid on {π}, then
ππ(π) = 0.It is a consequence of the MΓΆbius inversion for partiallyordered sets that the characteristic polynomial of π iswell defined.
We may now state our result in [AHK], which confirmsa conjecture of Gian-Carlo Rota and Dominic Welsh.Theorem 6 ([AHK, Theorem 9.9]). The coefficients of thecharacteristic polynomial form a log-concave sequencefor any matroid π.
This implies the log-concavity of the sequence ππ(πΊ)[Huh12] and the log-concavity of the sequence ππ(π΄)[Len12].
Hodge-Riemann Relations for MatroidsLet π be a mathematical object of βdimensionβ π. Often itis possible to construct from π in a natural way a gradedvector space over the real numbers
π΄β(π) =π
β¨π=0
π΄π(π),
equipped with a graded bilinear pairing
π βΆ π΄β(π) Γπ΄πββ(π) βΆ β,and a graded linear map
πΏ βΆ π΄β(π) βΆ π΄β+1(π), π₯ βΌ Lπ₯(βπβ is for PoincarΓ©, and βπΏβ is for Lefschetz). For example,π΄β(π) may be the cohomology of real (π, π)-forms on acompact KΓ€hler manifold π or the ring of algebraic cyclesmodulo homological equivalence on a smooth projectivevariety π or the combinatorial intersection cohomologyof a convex polytope π [Kar04] or the Soergel bimoduleof an element of a Coxeter group π [EW14] or the Chowring of a matroid π defined below. We expect that, forevery nonnegative integer πβ€ π
2 :(1) the bilinear pairing
π βΆ π΄π(π) Γπ΄πβπ(π) βΆ βis nondegenerate (PoincarΓ© duality for π),
(2) the composition of linear maps
πΏπβ2π βΆ π΄π(π) βΆ π΄πβπ(π)is bijective (the hard Lefschetz theorem for π), and
(3) the bilinear form on π΄π(π) defined by
(π₯1, π₯2) βΌ (β1)π π(π₯1, πΏπβ2ππ₯2)is symmetric and is positive definite on the kernelof
πΏπβ2π+1 βΆ π΄π(π) βΆ π΄πβπ+1(π)(the Hodge-Riemann relations for π).
All threeproperties areknowntohold for theobjects listedabove except one, which is the subject of Grothendieckβsstandard conjectures on algebraic cycles.
For a loopless matroid π on πΈ, the vector space π΄β(π)has the structure of a graded algebra that can be describedexplicitly.
Definition 7. We introduce variables π₯πΉ, one for eachnonempty proper flat πΉ of π, and set
πβ(π) = β[π₯πΉ]πΉβ β ,πΉβ πΈ.The Chow ring π΄β(π) of π is the quotient of πβ(π) bythe ideal generated by the linear forms
βπ1βπΉ
π₯πΉ β βπ2βπΉ
π₯πΉ,
one for each pair of distinct elements π1 and π2 of πΈ, andthe quadratic monomials
π₯πΉ1π₯πΉ2 ,one for each pair of incomparable nonempty proper flatsπΉ1 and πΉ2 of π.
28 Notices of the AMS Volume 64, Number 1
![Page 4: HodgeTheoryofMatroids - American Mathematical SocietyΒ Β· The Hodge-Riemann relations for the algebra π΄β(π), where πis the matroid attached to as in the section βMatroidsβ](https://reader030.vdocuments.us/reader030/viewer/2022020205/5c1b84ea09d3f2ff0d8c15a3/html5/thumbnails/4.jpg)
The Chow ring of π was introduced by Eva MariaFeichtner and Sergey Yuzvinsky. When π is realizableover a field π, it is the Chow ring of the βwonderfulβcompactification of the complement of a hyperplanearrangement defined over π as described by Corrado deConcini and Claudio Procesi.
Let π be the integer one less than the rank of π.
Theorem 8 ([AHK, Proposition 5.10]). There is a linearbijection
degβΆ π΄π(π) βΆ βuniquely determined by the property that
deg(π₯πΉ1π₯πΉ2 β―π₯πΉπ) = 1for every maximal chain of nonempty proper flats
πΉ1 β πΉ2 β β― β πΉπ.In addition, the bilinear pairing
π βΆ π΄π(π) Γπ΄πβπ(π) β β, (π₯, π¦) β¦ deg(π₯π¦)is nondegenerate for every nonnegative π β€ π.
What should be the linear operator πΏ for π? We collectall valid choices of πΏ in a nonempty open convex cone.The cone is an analogue of the KΓ€hler cone in complexgeometry.
Definition 9. A real-valued function π on 2πΈ is said to bestrictly submodular if
πβ = 0, ππΈ = 0,and, for any two incomparable subsets πΌ1, πΌ2 β πΈ,
ππΌ1 + ππΌ2 > ππΌ1β©πΌ2 + ππΌ1βͺπΌ2 .A strictly submodular function π defines an element
πΏ(π) = βπΉππΉπ₯πΉ β π΄1(π)
that acts as a linear operator by multiplication
π΄β(π) βΆ π΄β+1(π), π₯ βΌ πΏ(π)π₯.The set of all such elements is a convex cone in π΄1(π).
The main result of [AHK] states that the triple(π΄β(π),π, πΏ(π)) satisfies the hard Lefschetz theoremand the Hodge-Riemann relations for every strictly sub-modular function π:
Theorem 10. Let π be a nonnegative integer less than π2 .
(1) The multiplication by πΏ(π) defines an isomorphism
π΄π(π) βΆ π΄πβπ(π), π₯ βΌ πΏ(π)πβ2π π₯.(2) The symmetric bilinear form on π΄π(π) defined by
(π₯1, π₯2) βΌ (β1)π π(π₯1, πΏ(π)πβ2ππ₯2)is positive definite on the kernel of πΏ(π)πβ2π+1.
The known proofs of the hard Lefschetz theorem andthe Hodge-Riemann relations for the different types ofobjects listed above have certain structural similarities,but there is no known way of deducing one from theothers.
Sketch of Proof of Log-ConcavityWe now explain why the Hodge-Riemann relations forπ imply log-concavity for ππ(π). The Hodge-Riemannrelations for π, in fact, imply that the sequence (ππ) inthe expression
ππ(π)/(π β 1) = π0ππ βπ1ππβ1 +β―+ (β1)πππ
is log-concave, which is stronger.We define two elements of π΄1(π): for any π β πΈ, set
πΌπ = βπβπΉ
π₯πΉ, π½π = βπβπΉ
π₯πΉ.
The two elements do not depend on the choice of π, andthey are limits of elements of the form πΏ(π) for strictlysubmodular π. A short combinatorial argument showsthat ππ is a mixed degree of πΌπ and π½π:
ππ = deg(πΌππ π½πβπ
π ).Thus, it is enough to prove for every π that
deg(πΌπβπ+1π π½πβ1
π )deg(πΌπβπβ1π π½π+1
π ) β€ deg(πΌπβππ π½π
π)2.This is an analogue of the Teisser-Khovanskii inequal-ity for intersection numbers in algebraic geometry andthe Alexandrov-Fenchel inequality for mixed volumes inconvex geometry. The main case is when π = πβ 1.
By a continuity argument, we may replace π½π byπΏ = πΏ(π) sufficiently close to π½π. The desired inequalityin the main case then becomes
deg(πΌ2ππΏπβ2)deg(πΏπ) β€ deg(πΌππΏπβ1)2.
This follows from the fact that the signature of the bilinearform
π΄1(π) Γπ΄1(π) β β, (π₯1, π₯2) β¦ deg(π₯1πΏπβ2π₯2)restricted to the span of πΌπ and πΏ is semi-indefinite,which, in turn, is a consequence of the Hodge-Riemannrelations for π in the cases π = 0, 1.
This application uses only a small piece of the Hodge-Riemann relations for π. The general Hodge-Riemannrelations for π may be used to extract other interestingcombinatorial information about π.
References[AHK] Karim Adiprasito, June Huh, and Eric Katz, Hodge
theory for combinatorial geometries, arXiv:1511.02888.[EW14] Elias-Williamson, Ben Elias, and Geordie
Williamson, The Hodge theory of Soergel bi-modules, Ann. Math. (2) 180 (2014), 1089β1136.MR3245013
[Huh12] June Huh, Milnor numbers of projective hypersurfacesand the chromatic polynomial of graphs, J. Amer. Math.Soc. 25 (2012), 907β927. MR2904577
[Kar04] Kalle Karu, Hard Lefschetz theorem for nonra-tional polytopes, Invent. Math. 157 (2004), 419β447.MR2076929
[Len12] Matthias Lenz, The f-vector of a representable-matroid complex is log-concave, Adv. in Appl. Math. 51(2013), no. 5, 543β545. MR3118543
January 2017 Notices of the AMS 29
![Page 5: HodgeTheoryofMatroids - American Mathematical SocietyΒ Β· The Hodge-Riemann relations for the algebra π΄β(π), where πis the matroid attached to as in the section βMatroidsβ](https://reader030.vdocuments.us/reader030/viewer/2022020205/5c1b84ea09d3f2ff0d8c15a3/html5/thumbnails/5.jpg)
The creators of MathJobs.Org welcome you to:
MathPrograms.Org
Announcing...
Service is FREE to applicants.
Institutions pay annually for one program or for multiple programs.
Receive, read, rate, and respond to electronic
applications for your mathematical sciences
programs, such as undergraduate summer research programs and travel grant competitions.
Customize your settings and control the application form; also set secure access for the admissions committee.
Enter program announcements for public display.
Download data to personal computers for use in word processing and spreadsheets or as a full permanent storage file.
Photo CreditsPhoto of June Huh is courtesy of Nayoung Kim.Photo of Eric Katz is courtesy of Joseph Rabinoff.
THE AUTHORS
Karim Adiprasito
June Huh
Eric Katz
30 Notices of the AMS Volume 64, Number 1