/^creative ^commor ARS MATHEMATICA CONTEMPORANEA ISSN 1855-3966 (printed edn.), ISSN 1855-3974 (electronic edn.) ARS MATHEMATICA CONTEMPORANEA 16 (2019) 257-276 https://doi.org/10.26493/1855-3974.1619.a03 (Also available at http://amc-journal.eu) Pappus's Theorem in Grassmannian Gr (3, Cn) Sumire Sawada, Simona Settepanella *, So Yamagata Department of Mathematics, Hokkaido University, Japan Received 26 February 2018, accepted 13 August 2018, published online 6 January 2019 Abstract In this paper we study intersections of quadrics, components of the hypersurface in the Grassmannian Gr(3, Cn) introduced by S. Sawada, S. Settepanella and S. Yamagata in 2017. This lead to an alternative statement and proof of Pappus's Theorem retrieving Pappus's and Hesse configurations of lines as special points in the complex projective Grassmannian. This new connection is obtained through a third purely combinatorial object, the intersection lattice of Discriminantal arrangement. Keywords: Discriminantal arrangements, intersection lattice, Grassmannian, Pappus's Theorem. Math. Subj. Class.: 52C35, 05B35, 14M15 1 Introduction Pappus's hexagon Theorem, proved by Pappus of Alexandria in the fourth century A.D., began a long development in algebraic geometry. In its changing expressions one can see reflected the changing concerns of the field, from synthetic geometry to projective plane curves to Riemann surfaces to the modern development of schemes and duality. (D. Eisenbud, M. Green and J. Harris [4]) There are several knowns proofs of Pappus's Theorem including its generalizations such as Cayley Bacharach Theorem (see Chapter 1 of [9] for a collection of proofs of Pappus's Theorem and [4] for proofs and conjectures in higher dimension). In this paper, by mean of recent results in [6] and [10], we connect Pappus's hexagon configuration to intersections of well defined quadrics in the Grassmannian providing a new statement and proof of Pappus's Theorem as an original result on dependency conditions for defining polynomials of those quadrics. This result enlightens a new connection *The second named author was supported by JSPS Kakenhi Grant Number 26610001. E-mail addresses: b.lemon329@gmail.com (Sumire Sawada), s.settepanella@math.sci.hokudai.ac.jp (Simona Settepanella), so.yamagata.math@gmail.com (So Yamagata) ©® This work is licensed under https://creativecommons.org/licenses/by/4.0/ 258 Ars Math. Contemp. 16 (2019) 245-255 between special configurations of points (lines) in the projective plane and hypersurfaces in the projective Grassmannian Gr(3, Cn). This connection is made through a third combinatorial object, the intersection lattice of the Discriminantal arrangement. Introduced by Manin and Schechtman in 1989, it is an arrangement of hyperplanes generalizing classical braid arrangement (cf. [7, p. 209]). Fixed a generic arrangement A = {H0,..., H} in Ck, the Discriminantal arrangement B(n, k, A), n, k G N for k > 2 (k = 1 corresponds to Braid arrangement), consists of parallel translates H^1,..., Htn, (ti,..., tn) G Cn, of A which fail to form a generic arrangement in Ck. The combinatorics of B(n, k, A) is known in the case of very generic arrangements, i.e. A belongs to an open Zariski set Z in the space of generic arrangements H0, i = 1,..., n (see [7], [1] and [2]), but still almost unknown for A G Z .In 2016, Libgober and Settepanella (cf. [6]) gave a sufficient geometric condition for an arrangement A not to be very generic, i.e. A G Z .In particular in the case k = 3, their result shows that multiplicity 3 codimension 2 intersections of hyperplanes in B(n, 3, A) appears if and only if collinearity conditions for points at infinity of lines, intersections of certain planes in A, are satisfied (Theorem 3.8 in [6]). More recently (see [10]) authors applied this result to show that points in a specific degree 2 hypersurface in the Grassmannian Gr(3, Cn) correspond to generic arrangements of n hyperplanes in C3 with associated discriminantal arrangement having intersections of multiplicity 3 in codi-mension 2 (Theorem 5.4 in [10]). In this paper we look at Pappus's configuration (see Figure 1) as a generic arrangement of 6 lines in P2 which intersection points satisfy certain collinearity conditions (see Figure 2). This allows us to apply results on [6] and [10] to restate and re-prove Pappus's Theorem. More in details, let A be a generic arrangement in C3 and the arrangement of lines in ~ P2 directions at infinity of planes in A. The space of generic arrangements of n lines in (P2)n is Zariski open set U in the space of all arrangements of n lines in (P2)n. On the other hand in Gr(3, Cn) there is open set U' consisting of 3-spaces intersecting each coordinate hyperplane transversally (i.e. having dimension of intersection equal 2). One has also one set U in Hom(C3, Cn) consisting of embeddings with image transversal to coordinate hyperplanes and U/ GL(3) = U' and U/(C*)n = U. Hence generic arrangements in C3 can be regarded as points in Gr(3, Cn). Let {si < • • • < s6} C {1,..., n} be a set of indices of a generic arrangement A = {H0,..., H} in C3, ai the normal vectors of H0's and = det(a^ aj, a;). For any permutation a G S6 denote by [a] = {{ii,i2}, {i3,i4}, {i5,i6}}, ij = sCT(j), and by QCT the quadric in Gr(3,Cn) of equation ^i2i5i6 - Pi2i3i4A^ie = 0. The following theorem, equivalent to the Pappus's hexagon Theorem, holds. Theorem 5.3 (Pappus's Theorem). For any disjoint classes [ai] and [a2], there exists a unique class [a3] disjoint from [ai] and [a2] such that {QCT1, QCT2, QCT3} is a Pappus configuration, i.e. 3 Qai1 n Qffi2 = p| Qffi i=i for any {ii, i2} C [3]. In the rest of the paper, we retrieve the Hesse configuration of lines studying intersections of six quadrics of the form Qa for opportunely chosen [a]. This lead to a better understanding of differences in the combinatorics of Discriminantal arrangement in the complex and real case. Indeed it turns out that this difference is connected with existence of the Hesse arrangement (see [8]) in P2 (C), but not in P2 (R). S. Sawada et al.: Pappus's Theorem in Grassmannian Gr(3, Cn) 259 From above results it seems very likely that a deeper understanding of combinatorics of Discriminantal arrangements arising from non very generic arrangements of hyperplanes in Ck (i.e. A 2 and n > 3s, a good 3s-partition (see [10]) is a set T = {L i, L2, L3}, with Li subsets of [n] such that |Li| = 2s, |Li n Lj | = s (i = j), L i n L2 n L3 = 0 (in particular | U Li| = 3s), i.e. L i = {i i,..., i2S}, L2 = {i i,..., is,i2S+i,... ,i3s}, L3 = 260 Ars Math. Contemp. 16 (2019) 245-255 {¿s + 1, . . . , ¿3s}. Notice that given a generic arrangement A in C2s-1, subsets L define hyperplanes DLi in the Discriminantal arrangement B(n, 2s - 1, ATO). In this paper we are mainly interested in the case s = 2 corresponding to generic arrangements in C3. 2.3 Matrices A () and At () Let a = (a^,..., ajk) be the normal vectors of hyperplanes ffj, 1 < i < n, in the generic arrangement A in Ck. Normal here is intended with respect to the usual dot product (ai, .. ., afc) • (vi,. .., vfc) = ajVj. j Then the normal vectors to hyperplanes DL, L = {s1 < ••• < sfc+1} c [n] in S ~ Cn are nonzero vectors of the form k+1 «L = 53(-1)j det(asi,. .. ,aSi,.. ., aSfc+i )eSi, (2.1) i=1 where {ej}1 if j = ii e I, 0 otherwise. Plucker relations, i.e. conditions for dim(ker y>x) = k, are vanishing conditions of all (n — k + 1) x (n — k +1) minors of Mx. It is well known (see for instance [5]) that Plucker relations are degree 2 relations and they can also be written as k E(—1)1 ..pk-iqi ftqo...qi...qk = 0 (2.2) 1=0 for any 2k-tuple (pi,... ,Pk-1, qo,..., qk). Remark 2.1. Notice that vectors in the equation (2.1) normal to hyperplanes correspond to rows indexed by L in the Plucker matrix Mx, that is A(ATO ) = Mx, up to permutation of rows. Notice that, in particular, det(asi,..., ofSi,..., aSfc+1) is the Plucker coordinate ,0/, I = {s1, s2,..., sk+1} \ {«»}. 2.5 Relation between intersections of lines in and quadrics in Gr (3, Cn) Let A = {H0,..., H} be a generic arrangement in C3. If there exist L1, L2, L3 c [n] subsets of indices of cardinality 4, such that codimension of DLl n DL2 n DL3 is 2 then A is non very generic arrangement (see [2]). Let T = {L1, L2, L3} be a good 6-partition of indices {s1,...,s6} C [n]. In [6], authors proved that the codimension of DLl n DL2 n DL3 is 2 if and only if points f1ieL1nL2 f1ieL1nL3 and PlieL2nL3 are collinear in ([6, Lemma 3.1]). Since is vector normal to , the codimension of DLl n DL2 n DL3 is 2 if and only if rank AT(ATO) = 2, i.e. all 3 x 3 minors of AT(ATO) vanish. In [10] authors proved the following Lemma. Lemma 2.2 ([10, Lemma 5.3]). Let A be an arrangement of n hyperplanes in C3 and O.T = {{i1, i2, i3, i4}, {i1,i2, i5,i6}, {i3, i4, i5, i6}} a good 6-partition of indices s1 < • • • < s6 e [n] such that j = sCT(j), o permutation in S6. Then rank ACT.T(ATO) = 2 if and only if A is a point in the quadric of Grassmannian Gr (3, Cn) of equation £¿2 ¿5 ¿6 ^¿2«3i4 ^¿1i5«6 0. (2.3) 262 ArsMath. Contemp. 16(2019)203-213 As consequence of above results, we obtain correspondence between points x = ^ Piej, pj = 0, IC[n] | I | =3 in the quadric of equation (2.3) and generic arrangements of n hyperplanes A in C3 such that n HOTji2, HOTjia n and HOTji5 n HOTji6 are collinear in HOT. Notice that condition pI = 0 is direct consequence of A being generic arrangement. 3 Motivating example of Pappus's Theorem for quadrics in Gr (3, Cn) In classical projective geometry the following theorem is known as Pappus's theorem or Pappus's hexagon theorem. Theorem 3.1 (Pappus). On a projective plane, consider two lines l1 and l2, and a couple of triple points A, B, C and A', B', C' which are on 11 and l2 respectively. Let X, Y, Z be points of AB' n A'B, AC' n A'C and BC' n B'C respectively. Then there exists a line l3 passing through the three points X, Y, Z (see Figure 1). Figure 1: Original Pappus's Theorem. This theorem was originally stated by Pappus of Alexandria around 290-350 A.D. In this section, we restate this classical theorem in terms of quadrics in the Grassman-nian. Indeed the six lines AB', A'B, BC', B'C, AC', A'C e P2 (C) correspond to lines in the trace at infinity of a generic arrangement A in C3 and lines l1, l2 and l3 correspond to collinearity conditions for intersection points of lines in AOT. Consider a generic arrangement A = {H1,..., H6} of 6 hyperplanes in C3, its trace at infinity and T = {L1,L2,L3} the good 6-partition defined by L1 = {1,2,3,4}, L2 = {1,2,5,6}, L3 = {3,4,5,6}. By Lemma 2.2 we get that the triple points nieLinL2 Hi n H~, fWLa Hi n H~, nieL2nL3 Hi n H~ S. Sawada et al.: Pappus's Theorem in Grassmannian Gr(3, Cn) 263 are collinear if and only if A is a point of the quadric Ql : ^134^256 — ^234^156 = 0 in Gr(3, C6). Analogously if T' = {Ll, L2, ¿3}, Ll = {4, 6, 2, 5}, ¿2 = I4, 6,1, 3}, ¿3 = {2, 5,1, 3} and T'' = {¿1', ¿2', ¿3'}, ¿1' = {2,4,1, 6}, ¿2' = {2, 4, 3, 5}, ¿3' = {1, 6, 3, 5} are different good 6-partitions then triple points fliGLJnLi. Hi n nieiinL3 Hi n HieL^nL^ ^ n and PlieL'/nL^' H n PlieL'/nL^' Hi n PlieL2'nL3' Hi n are collinear if and only if A is, respectively, a point of quadrics Q2: ^425^613 — ^625^413 = 0 and Q3 : ^216^435 — ^416^235 = 0. With above remarks and notations we can restate Pappus's Theorem as follows (see Figure 2). Theorem 3.2 (Pappus's Theorem). Let A = {H1,..., H6} be a generic arrangement of hyperplanes in C3. If A is a point of two of three quadrics Q1, Q2 and Q3 in the Grassmannian Gr (3, C6), then A is also a point of the third. In other words 3 Qi ' n Qi2 = ff Qi, {¿1,i2}c [3]. i=1 We develop this argument in the following sections providing in Theorem 5.3 a general statement on quadrics in the Grassmannian which implies Pappus hexagon Theorem in the projective plane. 4 Pappus Variety In this section, we consider a generic arrangement {H1,..., Hn} in C3 (n > 6). Let's introduce basic notations that we will use in the rest of the paper. Notation. Let {s1,..., s6} be a subset of indices {1,..., n} and T = {¿1, ¿2, ¿3} be the good 6-partition given by ¿1 = {s1, S2, S3, S4}, ¿2 = {s1, S2, S5, S6} and ¿3 = {S3, S4, S5, S6}. Then for any permutation a G S6 we denote by a.T = , aX3} the good 6- partition given by subsets = {¿1, ¿2, ¿3, ¿4}, a^2 = {¿1, ¿2, ¿5, ¿6} and 0X3 = {¿3, ¿4, ¿5, ¿6} with j = sCT(j). Accordingly, we denote by QCT the quadric in Gr (3, Cn) of equation Q : A 1 i3 i4 ^i2i5i6 ^i2i3i4 Ali5i6 0. 264 ArsMath. Contemp. 16(2019)203-213 The following lemma holds. Lemma 4.1. Let ct, ct' g S6 be distinct permutations, then QCT = if and only if there exists t g S3 such that ct.Lj n CT.Lj = ct'.Lt(j) n ct'.Ltj (1 < i < j < 3). Proof. By definition of good 6-partition we have that Li = (Li n L2) U (Li n L3), L2 = (L2 n Li) u (L2 n L3), L3 = (L3 n Li) u (L3 n L2). Then there exists t g S3 such that ct and ct' satisfy a.Lj n a.Lj = ct'.Lt(^ n ct'.Ltj (1 < i < j < 3) if and only if ct.L; = ct'.Lt(() for l = 1,2,3, that is ACT/.T(AOT) is obtained by permuting rows of ACT.T(AOT). It follows that rank ACT.T(AOT) = 2 if and only if rank ACT/.T(AOT) = 2 and hence by Lemma 2.2 this is equivalent to QCT n NS1... S6 = Qct' n NS1i...jS6, where NS1I...IS6 = {x = E ei I = 0 for any I c {si,... ,se}}. IC[n] |I|=3 Since NS1 i...jS6 is dense open set in 7(Gr(3,Cn)), Q- n NS1i...iS6 = Q- n NS1i...jS6 if and only if QCT = . Vice versa if QCT n Ns1... s6 = n Ns1... s6, then any generic arrangement A corresponding to a point in QCT n Ns1. . . s6 corresponds to a point in n NS1 i...jS6, that is rank ACT.T(AOT) = 2 if and only if rank ACT/.T(AOT) = 2. It follows that Act T(Aot) and ACT/.T(AOT) are submatrices of A(AOT) defined by the same three rows, i.e. ct.L; = ct'.Ltfor l = 1,2,3. □ Definition 4.2. For any 6 fixed indices T = {si,..., s6} C [n] the Pappus Variety is the hypersurface in Gr(3, Cn) given by pt = U Q-. CTGS6 S. Sawada et al.: Pappus's Theorem in Grassmannian Gr(3, Cn) 265 Notice that all the content of this section and the following section is based on the choice of six indices {si < • • • < s6} C [n]. This is related to result in Theorem 3.8 in [6] and, consequently, Lemma 5.3 in [10] (Lemma 2.2 in this paper). Indeed Theorem 3.8 in [6] states that in order to study special configurations of n lines in P2, that is non very generic arrangements of n lines in P2, it is sufficient to study subsets of six lines out of n. On the other hand since Pappus Variety can be defined inside Gr(3, Cn), we decided to keep the discussion more general picking six indices {s1 < ••• as following: a.T - a'.T t g S3 such that a.L, n a.Lj = a'.LT(i) n a'.LT(1 < i < j < 3). We denote by [a] the equivalence class containing a.T and by QCT the corresponding quadric (notice that a in the notation QCT can be any representative of [a]). By Lemma 4.1 [a] only depends on couples Lj n Lj hence for each class [a] we can choice a representative a.To = {{ji, j2, j3, j'4}, {ji, j2, j5, je}, {j3,j4,j5, je}} such that j < j2, j3 < j4, j5 < j6 and j < j'3 < j5 and we can equivalently define [a] = {{j1, j2}, {j3, j4}, {j5,j6}}. (6)(4)(2) Since the number of choices of [a] is V2y v32/V2y =15, Pappus Variety is composed by 15 quadrics. Finally remark that [a] = {{ji, j'2}, {j'3, j'4 }, {j5, j'e}} and [a'] = {{ji,j2}, {j'3,j'4}, {j'5,j'6}} are disjoint, i.e. [a] n [a'] = 0, if and only if {j2;-i, j'2;} = {j'2i'-i,j'2v} for any 1 < /,/' < 3. Definition 4.3 (Pappus configuration). Let [ai], [a2] and [a3] be disjoint classes, a Pappus configuration is a set {QCT1, QCT2, QCT3} of quadrics in Gr (3, Cn) such that 3 Q4 It follows that in the real case, for any choice of indices {s1,..., s6} C [n], we have at most 4 collinearity conditions (see Figure 7) corresponding to 15 hyperplanes in the Dis-criminantal arrangement with 4 multiplicity 3 intersections in codimension 2 (see Figure 8). While in the complex case Hesse configuration (see Figure 6) gives rise to a Discriminantal arrangement containing 15 hyperplanes intersecting in 6 multiplicity 3 spaces in codimen-sion 2. This remark allows a better understanding of differences in the combinatorics of Dis-criminantal arrangement in the real and complex cases. Indeed the existence of a discrimi-nantal arrangement of 15 hyperplanes intersecting in 6 multiplicity 3 spaces in codimension 2 in C but not in R implies that there exist combinatorics of Discriminantal arrangements that cannot be realised in any field. This is especially interesting since in the case known until now, i.e. in the case of very generic arrangements A, the combinatorics of Discriminantal arrangement B(n, k, A) is independent from the field (see [1]). S. Sawada et al.: Pappus's Theorem in Grassmannian Gr(3, Cn) 275 Q, Figure 7: Generic arrangement A in R3 containing 6 lines satisfying 4 collinearity conditions. A " " Figure 8: Codimension 2 intersections of 15 hyperplanes in B(n, 3, ATO) indexed in {si,...,s6} C [n] with 4 multiplicity 3 points ▲ corresponding to intersections n3=i DCT.Li, f|3=i DCT/.Li, n3=i Dff//.Li and P|3=i DCT///.Li, ct, ct', a", a"' as in Figure 7. 234 Ars Math. Contemp. 16(2019)215-276 Remark 6.7. Finally Theorem 6.5 implies that the maximum number of intersections of multiplicity 3 in codimension 2 in the complex case is strictly higher than the one in the real case. This agrees with results on maximum number of triple points in an arrangement of lines in P2 (see [3] for a discussion on line arrangements with maximal number of triple points over arbitrary fields). Those observations suggest that special configurations of lines in the projective plane intersecting in a big number of triple points could be understood by studying Discriminantal arrangements with maximum number of multiplicity 3 intersections in codimension 2. Indeed each multiplicity 3 intersection in codimension 2 of B(n, 3, ) corresponds to a collinearity condition for lines in which is equivalent to the possibility to add a line that gives rise to "higher" number of triple points. It seems hence interesting to study exact number of intersections of type (1) and (2) in Theorem 6.5 in the Grassmannian Gr(3, Cn). This will be object of further studies. References [1] C. A. Athanasiadis, The largest intersection lattice of a discriminantal arrangement, Beiträge Algebra Geom. 40 (1999), 283-289, https://www.emis.de/journals/BAG/vol. 40/no.2/1.html. [2] M. M. Bayer and K. A. Brandt, Discriminantal arrangements, fiber polytopes and formality, J. Algebraic Combin. 6 (1997), 229-246, doi:10.1023/a:1008601810383. [3] M. Dumnicki, L. Farnik, A. Glöwka, M. Lampa-Baczynska, G. Malara, T. Szemberg, J. Szpond and H. Tutaj-Gasinska, Line arrangements with the maximal number of triple points, Geom. Dedicata 180 (2016), 69-83, doi:10.1007/s10711-015-0091-7. [4] D. Eisenbud, M. Green and J. Harris, Cayley-Bacharach theorems and conjectures, Bull. Amer. Math. Soc. (N. S.) 33 (1996), 295-324, doi:10.1090/s0273-0979-96-00666-0. [5] J. Harris, Algebraic Geometry: A First Course, volume 133 of Graduate Texts in Mathematics, Springer-Verlag, New York, 1992, doi:10.1007/978-1-4757-2189-8. [6] A. Libgober and S. Settepanella, Strata of discriminantal arrangements, J. Singul. 18 (2018), 440-454, doi:10.5427/jsing.2018.18w. [7] Yu. I. Manin and V. V. Schechtman, Arrangements of hyperplanes, higher braid groups and higher Bruhat orders, in: J. Coates, R. Greenberg, B. Mazur and I. Satake (eds.), Algebraic Number Theory, Academic Press, Boston, MA, volume 17 of Advanced Studies in Pure Mathematics, pp. 289-308, 1989, in honor of K. Iwasawa on the occasion of his 70th birthday on September 11, 1987. [8] P. Orlik and H. Terao, Arrangements of Hyperplanes, volume 300 of Grundlehren der Mathematischen Wissenschaften, Springer-Verlag, Berlin, 1992, doi:10.1007/978-3-662-02772-1. [9] J. Richter-Gebert, Perspectives on Projective Geometry: A Guided Tour Through Real and Complex Geometry, Springer, Heidelberg, 2011, doi:10.1007/978-3-642-17286-1. [10] S. Sawada, S. Settepanella and S. Yamagata, Discriminantal arrangement, 3 x 3 minors of Plücker matrix and hypersurfaces in Grassmannian Gr(3, n), C. R. Acad. Sci. Paris Ser. 1355 (2017), 1111-1120, doi:10.1016/j.crma.2017.10.011.