Lines and opposition in Lie incidence geometries of exceptional type Thanks: The first author is funded by the Claussen-Simon-Stiftung and by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy EXC 2044 –390685587, Mathematics Münster: Dynamics–Geometry–Structure. This work is part of the PhD project of the first author. Sira Busch, Hendrik Van Maldeghem Abstract . We characterise sets of points of exceptional Lie incidence geometries, that is, the natural geometries arising from spherical buildings of exceptional types 𝖥 𝟦 \mathsf{F_{4}} , 𝖤 𝟨 \mathsf{E_{6}} , 𝖤 𝟩 \mathsf{E_{7}} , 𝖤 𝟪 \mathsf{E_{8}} and 𝖦 𝟤 \mathsf{G_{2}} , that form a line using the opposition relation. With that, we obtain a classification of so-called “geometric lines” in many of these geometries. Furthermore, our results lead to a characterisation of geometric lines in finite exceptional Lie incidence geometries as minimal blocking sets, that is, point sets of the size of a line admitting no object opposite to all of their members, in most cases, and we classify all exceptions. As a further consequence, we obtain a characterisation of automorphisms of exceptional spherical buildings as certain opposition preserving maps. 1. Introduction The intricate structure of — especially the exceptional — spherical buildings leads to the introduction of seemingly less complex, but certainly more accessible, point-line geometries describing essentially the same object. These geometries are usually called Lie incidence geometries . The procedure to construct such a geometry is nowadays standard: for a spherical building \Delta , say of type \mathsf{X}_{n} , pick a type i , consider all vertices of \Delta of type i and call them points . Pick a chamber C and delete the vertex of type i to obtain a panel of cotype i . Call the set of vertices of type i completing this panel again to a chamber a line . Vary C over all chambers. Then this system of points and lines is a Lie incidence geometry of type \mathsf{X}_{n,i} . It now turns out (as follows from [ 12 , §3] ) that the full automorphism group of this point-line geometry coincides with the full automorphism group of \Delta that preserves the type i . In [ 21 ] , Kasikova and the second author investigate the interaction of (this definition of) lines and the opposition relation in \Delta . The opposition relation is something typical for spherical buildings, and it owes its existence to the strong relation with finite Coxeter groups, in which there is a so-called “longest word”. The main result of [ 21 ] characterises the lines of many Lie incidence geometries in terms of this opposition relation. This led to the introduction of the notion of a “geometric line”, which is a set of points with the property that an arbitrary object of the underlying spherical building is either opposite none of its elements, or not opposite exactly one of its members. In particular, a geometric line does not admit an object opposite all of its members. A set of points with the latter property can be viewed as a blocking set (of points) . Blocking sets are a popular subject in finite geometry, both because they have many applications and because they have a great auxiliary value. A line is always a blocking set, and in [ 10 ] the authors proved that in odd characteristic, geometric lines in finite classical Lie incidence geometries are the only minimal blocking sets. In characteristic 2 they found counterexamples. Their proof, however, makes very little use of the definition of a geometric line; the equivalence in odd characteristic just came out from the classification of minimal blocking sets, which uses a variety of tools from classical finite geometry. In the present paper, the primary aim is to classify minimal blocking sets in many finite Lie incidence geometries of exceptional type. However, unlike in [ 10 ] , in many cases it will turn out to be beneficial to do this by showing directly the equivalence to geometric lines; the exceptions in characteristic 2 become also apparent in this approach. We then either appeal to the classification of geometric lines as provided in [ 21 ] , or we classify them ourselves (if not available in [ 21 ] ). This way, we lay the foundations to study more intensively blocking sets in finite exceptional geometries. More exactly, with the notation introduced in Section 2 , we will show the following theorems. Main Result A . Let \mathsf{X}_{n}\in\{\mathsf{E_{6},E_{7},E_{8},F_{4},G_{2}}\} and 1\leq i\leq n , with i\notin\{2,4,5\} if n=7 and i\in\{7,8\} if n=8 . If in an irreducible thick finite Moufang spherical building \Delta of type \mathsf{X}_{n} , n\geq 2 , the panels of cotype \{i\} are s -thick (that is, every panel of cotype \{i\} is contained in precisely s+1 chambers), then every set of s+1 vertices of type i of \Delta admits a common opposite vertex except precisely in the following five cases. (1) The s+1 vertices form a line in the corresponding Lie incidence geometry of type \mathsf{X}_{n,i} . (2) \Delta is a split building of type \mathsf{F_{4}} , the type i corresponds to a short root in the underlying root system, and the s+1 vertices correspond to a hyperbolic line in a (thick) symp of the corresponding Lie incidence geometry of type \mathsf{F}_{4,i} isomorphic to a symplectic polar space, and i\in\{3,4\} . (3) \Delta has type \mathsf{F_{4}} , has residues isomorphic to Hermitian generalised quadrangles of order (s,\sqrt{s}) , and the s+1 vertices form an ovoid in a (symplectic) subquadrangle of order (\sqrt{s},\sqrt{s}) . (4) \Delta is a split building of type \mathsf{G_{2}} and the s+1 vertices form a hyperbolic (or ideal) line in the corresponding Lie incidence geometry of type \mathsf{G_{2,2}} , which is a split Cayley hexagon. (5) \Delta is a building of type \mathsf{G_{2}} in characteristic 2 and the s+1 vertices form a distance- 3 trace in the corresponding Lie incidence geometry of type \mathsf{G_{2,2}} , which is either a split Cayley hexagon, or a twisted triality hexagon of order (s,\sqrt[3]{s}) . A will follow - from 3.1 for types \mathsf{E_{6,1}} and \mathsf{E_{6,6}} ; - from 3.2 for types \mathsf{E_{6,3}} and \mathsf{E_{6,5}} ; - from 3.7 for type \mathsf{E_{7,7}} ; - from 3.8 for type \mathsf{E_{7,6}} ; - from 4.1 , 4.2 and Theorem 4.4 for types \mathsf{E_{6,2},E_{7,1},E_{8,8},F_{4,1}} and \mathsf{F_{4,4}} ; - from 4.11 , Theorem 4.15 and 4.22 for types \mathsf{E_{6,4},E_{7,2},E_{8,7},F_{4,2}} and \mathsf{F_{4,3}} ; - from 5.3 for types \mathsf{G_{2,1}} and \mathsf{G_{2,2}} . Main Result B . Let \mathsf{X}_{n}\in\{\mathsf{E_{6},E_{7},E_{8},F_{4},G_{2}}\} and 1\leq i\leq n , with i\notin\{2,4,5\} if n=7 and i\in\{7,8\} if n=8 . Then a geometric line of the Lie incidence geometry of type \mathsf{X}_{n,i} associated to an irreducible thick Moufang spherical building \Delta of type \mathsf{X}_{n} , n\geq 2 , is one of the following. (1) A line of the Lie incidence geometry. (2) A hyperbolic line in a symp of the Lie incidence geometry whenever this symp is a symplectic polar space (and this happens (only) in the split case for types \mathsf{F_{4,3}} and \mathsf{F_{4,4}} ). (3) A hyperbolic (or ideal) line of a split Cayley hexagon. (4) A distance- 3 trace of a split Cayley hexagon over a perfect field of characteristic 2 . B follows from [ 21 , Corollary 5.6] for the types \mathsf{E_{6,1},E_{6,6}} and \mathsf{E_{7,7}} . It will follow - from 3.6 for types \mathsf{E_{6,3}} and \mathsf{E_{6,5}} ; - from 3.12 for type \mathsf{E_{7,6}} ; - from 4.2 and Theorem 4.4 for types \mathsf{E_{6,2},E_{7,1},E_{8,8},F_{4,1}} and \mathsf{F_{4,4}} ; - from 4.22 for types \mathsf{E_{6,4},E_{7,2},E_{8,7},F_{4,2}} and \mathsf{F_{4,3}} ; - from 5.9 for types \mathsf{G_{2,1}} and \mathsf{G_{2,2}} . As an application, we deduce that certain opposition preserving maps and transformations of a Lie incidence geometry are actually (bijective) collineations. This characterises collineations using opposition. We refer to Section 6 for the exact statements. As a further motivation, we note that some of the results obtained in the present paper are used in [ 9 ] and [ 11 ] to determine projectivity groups. The paper is structured as follows. We provide preliminary information in Section 2 about the geometries we will work with, and about their well-known properties. We also derive many new properties that are not available in the literature. Towards the end of that section, we state a corollary to a result of Tits that enables one to construct blocking sets in buildings from blocking sets in residues. Finally, we prove some general properties of round-up triples of vertices in general spherical buildings, which we will use to classify such triples in several cases in this paper. We then start the proofs of our main theorems. We chose to prove A and B type-by-type in the same section, so that the properties of the geometry in question are fresh in the memory. Section 3 treats the cases \mathsf{E_{6,1},E_{6,3},E_{7,7}} and \mathsf{E_{7,6}} using the so-called exceptional minuscule geometries. These are the Lie incidence geometries of type \mathsf{E_{6,1}} and \mathsf{E_{7,7}} . Then, in Section 4 , we prove our main theorems for \mathsf{E_{6,2},E_{6,4},E_{7,1},E_{7,2},E_{8,8},E_{8,7}} and \mathsf{F}_{4,i} , i\in\{1,2,3,4\} . The relatively easy cases \mathsf{E_{6,2},E_{7,1},E_{8,8}} and \mathsf{F}_{4,i} , i\in\{1,4\} are proved in Section 4.1 , whereas the other cases are treated in Section 4.2 . This is by far the longest section of the paper. Amongst other things, it classifies all possible mutual positions of two lines in a given hexagonic Lie incidence geometry of exceptional type. This certainly has other potential applications. Doing this enables us to reduce A for lines of exceptional hexagonic Lie incidence geometries to the classification of geometric lines ( B ), except in the case of geometries isomorphic to \mathsf{F_{4,4}}(q,q^{2}) . We treat this case separately in Section 4.2.4 . Then in Section 4.2.5 we prove B uniformly for \mathsf{E_{6,4},E_{7,2},E_{8,7}} and \mathsf{F}_{4,i} , i\in\{2,3\} . Finally, in Section 5 , we prove our main results for type \mathsf{G_{2}} , that is, for generalised hexagons. Section 6 contains the application to certain opposition preserving maps in buildings of exceptional type alluded to above. 2. Preliminaries 2.1. Buildings and Lie incidence geometries We are going to work with the buildings of exceptional type via their well-established point-line geometries, which fit perfectly in the framework of Cooperstein’s theory of parapolar spaces . We will also adopt the corresponding terminology. As a result, we will not spend too much space on pure building-theoretic theory, but instead refer the interested reader to the literature, in particular to [ 1 ] and [ 30 ] . We content ourselves with the following generalities. We view buildings as thick numbered simplicial chamber complexes. The buildings we are interested in are spherical buildings and, as such, there is the notion of opposition of simplices (in particular vertices and chambers), expressing that the two given simplices are at maximal distance apart. There is also the notion of convexity for subcomplexes, and the convex closure of two opposite chambers is a thin finite chamber complex called an apartment and isomorphic to a Coxeter complex. These apartments play a crucial role in building theory. By the definition of a building, every pair of simplices is contained in an apartment and so the mutual position between these two objects can be seen in this finite complex, which is also a triangulation of a sphere. Opposite simplices in an apartment are then just antipodal on the sphere. Also, recall that, since \Delta is numbered, the vertices have types , and a chamber consists of a set of vertices, one of each possible type. The number of types is the rank of the building. A panel of cotype i is a simplex containing vertices of each type except for the type i . Opposition induces a permutation on the types which is an automorphism of the Coxeter diagram. Two simplices are joinable if their union is again a simplex. We now quickly outline how point-line geometries arise from (spherical) buildings, a point-line geometry being a pair (X,\mathscr{L}) consisting of a set X of points , and \mathscr{L} a set of subsets of X , each member of which is called a line . Let \Delta be a spherical building, which we will always assume to be thick and irreducible. We consider the set X of vertices of a given type, say i . The set \mathscr{L} then consists of the subsets of X whose elements each complete a given panel of cotype i to a chamber (hence each panel defines a unique line of (X,\mathscr{L}) , but different panels may define the same line). If the type of \Delta is \mathsf{X}_{n} , where n is the rank and \mathsf{X} one of the Coxeter types \mathsf{A,B,C,D,E,F,G} , then we say that (X,\mathscr{L}) is a Lie incidence geometry of type \mathsf{X}_{n,i} . Usually, we take as type set \{1,2,\ldots,n\} , where the types can be read off the corresponding Coxeter or Dynkin diagram using Bourbaki labelling [ 3 ] . If the diagram is simply laced, then \Delta is completely determined by the underlying skew field \mathbb{K} and we denote the Lie incidence geometry of type \mathsf{X}_{n,i} in this case as \mathsf{X}_{n,i}(\mathbb{K}) . In such geometries, vertices of the building have interpretations as certain subspaces (see below), and we will always call such subspaces opposite when the vertices are opposite in the building. We adopt the notation v\equiv v^{\prime} for opposite vertices (and later, subspaces); the negation is v\not\equiv v^{\prime} , and the set of vertices (not) opposite v is denoted as v^{\equiv} ( v^{\not\equiv} ). When \mathsf{X}_{n} is one of \mathsf{E_{6},E_{7},E_{8},F_{4}} , then a Lie incidence geometry of type \mathsf{X}_{n,i} is always a parapolar space. We provide a brief introduction. First, we introduce some terminology concerning point-line geometries. Let \Gamma=(X,\mathscr{L}) be a point-line geometry. Two points x,y\in X contained in a common line are called collinear and denoted x\perp y . The set of points collinear to x is denoted x^{\perp} and includes x if x is on some line. Sets of points are called collinear if each point of either is collinear to each point of the other. If each line contains exactly two (at least three) points, then \Gamma is called thin ( thick , respectively). A subspace of \Gamma is a set of points with the property that, if two distinct collinear points belong to it, then all points of each line containing both x and y belong to it. We often view a subspace as a point-line subgeometry in the obvious way. A subspace is a (geometric) hyperplane if it intersects each line non-trivially; it is called proper if it does not coincide with X itself, which is a trivial geometric hyperplane. The point-line geometry \Gamma is called a partial linear space if every pair of collinear points is contained in exactly one line. In a partial linear space, we denote the unique line containing two distinct collinear points x and y by xy , or sometimes by \langle x,y\rangle , for clarity. A subspace of a point-line geometry is called singular if every pair of its points is collinear. Trivial examples are the empty set, each singleton, and each line. If there exists a natural number r such that every finite nested sequence of (distinct) singular subspaces (including the empty space) has size at most r+1 , and there exists such a sequence of size r+1 , then we say that \Gamma has singular rank r . A maximal singular subspace is a singular subspace that is not properly contained in another one. The point graph of a point-line geometry \Gamma is the graph with vertices the points of \Gamma , adjacent when (distinct and) collinear. A set of points is called convex if for each pair \{x,y\} of points contained in it, all points of each shortest path between x and y in the point graph are also contained in it. We assume the reader is familiar with projective spaces, which are the Lie incidence geometries \mathsf{A}_{n,1}(\mathbb{K}) , for skew fields \mathbb{K} , also sometimes denoted as \mathsf{PG}(n,\mathbb{K}) , or as \mathsf{PG}(V) , where V is an (n+1) -dimensional vector space over \mathbb{K} . One checks that \mathsf{PG}(n,\mathbb{K}) has singular rank n+1 . As a building, it has rank n , which is also its projective dimension. We extend the definition of \mathsf{PG}(V) to infinite-dimensional vector spaces in the obvious way: the points of \mathsf{PG}(V) are the 1 -spaces of V , the lines are the sets of 1 -spaces contained in given 2 -spaces. The Lie incidence geometries of types \mathsf{B}_{n,1} , n\geq 2 and \mathsf{D}_{n,1} , n\geq 3 , are polar spaces (of rank n ) , that is, thick point-line geometries (X,\mathscr{L}) of singular rank n such that for each point x\in X , the set x^{\perp} is a proper geometric hyperplane. It follows that polar spaces are partial linear spaces (see [ 8 ] ). Also, the non-trivial singular subspaces of any polar space of rank at least 3 are projective spaces. If in a polar space of rank r , every (r-2) -dimensional singular subspace is contained in exactly two (at least three) maximal singular subspaces, then we call the polar space top-thin ( thick , respectively). Buildings of type \mathsf{B}_{n} correspond precisely to the thick polar spaces of rank n , while buildings of type \mathsf{D}_{n} , n\geq 2 (where we identify the type \mathsf{D_{2}} with the reducible type \mathsf{A_{1}\times A_{1}} ) yield top-thin polar spaces. The singular subspaces are also the only subspaces of a polar space that are convex. In fact, every polar space of rank at least 2 is a Lie incidence geometry of type \mathsf{B}_{n,1} or \mathsf{D}_{n,1} , n\geq 2 . Polar spaces of type \mathsf{D}_{n} have a peculiar property: their maximal singular subspaces are divided into two oriflamme classes , where two maximal singular subspaces belong to different classes if, and only if, the parity of the dimension of their intersection coincides with the parity of n . Each polar space \mathsf{D}_{n,1}(\mathbb{K}) , for \mathbb{K} a (commutative) field and n\geq 3 , is isomorphic to the point-line geometry naturally associated to a hyperbolic quadric in \mathsf{PG}(2n-1,\mathbb{K}) , that is, the null set of a quadratic form x_{-n}x_{n}+x_{-n+1}x_{n-1}+\cdots+x_{-1}x_{1} in the coordinates of a point. An oriflamme class of lines in a hyperbolic quadric in \mathsf{PG}(3,\mathbb{K}) (the case n=2 left out in our previous sentence) will be called a regulus . Top-thin polar spaces are also referred to as hyperbolic polar spaces. The Lie incidence geometries of type \mathsf{B}_{n,n} are usually referred to as dual polar spaces . We have defined Lie incidence geometries only for vertices of (spherical) buildings; a straightforward generalisation to simplices is possible, and we will use such generalisation, but only for simplices of buildings of type \mathsf{A}n , where the simplices in question are point-hyperplane pairs of the corresponding projective space. Hence we can define the geometry \mathsf{A}{n,{1,n}}(\mathbb{K}) , \mathbb{K} any skew field, as the point-line geometry with point set the set of incident point-hyperplane pairs of \mathsf{PG}(n,\mathbb{K}) , where the lines are of two types: one type of lines consists of the sets of point-hyperplane pairs with common hyperplane H and point ranging over a given line contained in H ; the other type is the dual. Let us also remark that sometimes a Lie incidence geometry \Delta_{i} of type \mathsf{X}{n,i} can be defined using the one, say \Delta_{j} , of type \mathsf{X}{n,j} by considering the subspaces of \Delta_{i} conforming to the vertices of type j of the associated building, as points, and interpreting then the lines of \Delta_{j} in \Delta_{i} to define them correctly. We give an example. Let \Delta be a building of type \mathsf{D}_{n} , n\geq 4 , and let \Delta_{1}=(X_{1},\mathscr{L}_{1}) be the associated polar space (a Lie incidence geometry of type \mathsf{D}_{n,1} ). Define X_{2} as the set of lines of \Delta_{1} . Now let \Delta_{2} be the point-line geometry with point set X_{2} , and let the set \mathscr{L}_{2} of lines of \Delta_{2} be the set of planar line pencils. Then one checks that (X_{2},\mathscr{L}_{2}) is the Lie incidence geometry of type \mathsf{D}_{n,2} corresponding to \Delta . We say that \Delta_{2} is the line-Grassmannian of \Delta_{1} (and we use this expression also in other situations where a point-line geometry has planes, and hence planar line pencils). All Lie incidence geometries, as we defined them, are parapolar spaces, except for the projective and polar spaces mentioned above, and for the Lie incidence geometries of buildings of rank 2 . Let’s define these objects. Unlike polar spaces, it is not known whether all parapolar spaces of sufficiently high symplectic rank are Lie incidence geometries. A parapolar space , introduced by Cooperstein [ 14 , 15 ] , is a point-line geometry \Gamma=(X,\mathscr{L}) satisfying the following axioms: (i) Each pair of points at distance 2 in the point graph either is collinear to a unique point, or is contained in a convex subspace isomorphic to a polar space (such subspaces are called symplecta , or symps for short). (ii) There exist at least two distinct symps, and each line is contained in a symp. It follows easily that parapolar spaces are partial linear spaces. Now we introduce some specific terminology and notation concerning parapolar spaces. First, a pair of points {x,y} at distance 2 in the point graph, collinear to a unique point z , will be called special , and we denote z=:[x,y] . We also say that x is special to y , or that x and y are special, in symbols x\Join y . The set of points special to a given point x will be denoted as x^{\Join} . A parapolar space without special pairs is called strong . The symp containing two given points x,y at distance 2 in the point graph and which are not special is denoted by \xi(x,y) . The pair {x,y} is called symplectic , and we also say that x is symplectic to y , or that x and y are symplectic, in symbols x\perp\hskip-4.26773pt\perp y . The set of points symplectic to a given point x will be denoted as x^{\perp\hskip-4.26773pt\perp} . The diameter of \Gamma is the diameter of its point graph. We say that \Gamma has symplectic rank at least r if every symp has rank at least r . If every symp has exactly rank r , then we say that \Gamma has uniform rank r . Note that the line-Grassmannian of a polar space contains special pairs. If such a polar space \xi is a symp in a parapolar space, then we speak about \xi -special lines , with the obvious meaning; that is, disjoint lines containing points at distance 2 such that some point of either is collinear to all points of the other. We can interpret residues (or links) of vertices from the theory of buildings in the corresponding Lie incidence geometries as point residuals . Let \Delta=(X,\mathscr{L}) be a parapolar space with the property that all singular subspaces are projective spaces and through each point we have at least one plane. Then, at each point x\in X , we can define the point-residual \mathsf{Res}_{\Delta}(x) , or \mathsf{Res}(x) if no confusion can arise, as the point-line geometry with point set the set of lines of \Delta through x and with as set of lines the planar line pencils of \Delta at x (each line of the line pencil contains x ). As usual, the type of the point residuals can be read off the Coxeter diagram by deleting the node corresponding to the points. Parapolar spaces are so-called gamma spaces , that is, point-line geometries in which each point is collinear to zero, one, or all points of a given line. We will frequently use this property, often without reference. We will be working with specific Lie incidence geometries, mainly of exceptional type. For the classical types, the properties can be derived without much effort from the corresponding projective or polar space. We now review the basic properties of the exceptional Lie incidence geometries we will be working with. Along the way, we also prove some additional properties that we will need. The basic properties reflect the possible mutual positions of certain elements of the geometry, usually points, subspaces, and symps. In the statements of facts, we will occasionally introduce terminology and underline the introduced notions. Subspaces or symps through a common point x will occasionally be called locally opposite (at x ) if they correspond to opposite objects in \mathsf{Res}(x) . We also extend this terminology to all vertices x . In particular, if \xi is a symp, then singular subspaces in \xi that are opposite in \xi as a polar space are called locally opposite at \xi , or briefly \xi -opposite . This extends the terminology for lines of \xi being \xi -special, introduced before. 2.2. Lie incidence geometries of type \mathsf{E_{6,1}} For each field \mathbb{K} there exists a unique Lie incidence geometry isomorphic to \mathsf{E_{6,1}}(\mathbb{K}) . It is a strong parapolar space of diameter 2 and also called a minuscule geometry . The following properties can either be found in [ 28 ] , or can be easily derived from an apartment of the corresponding building (in this case the 1 -skeleton of such an apartment, where vertices are points of \mathsf{E_{6,1}}(\mathbb{K}) and edges are lines, is the Schläfli graph, see [ 6 , §10.3.4] ). One can also use the chain calculus introduced in [ 27 ] , also explained in [ 7 , §4.5.4] . Fact 2.1 . Let p be a point and \xi a symp of \mathsf{E_{6,1}}(\mathbb{K}) . Then p^{\perp}\cap\xi is either empty (and we say that x and \xi are far ; they are also opposite in the corresponding building) or a maximal singular subspace of \mathsf{E_{6,1}}(\mathbb{K}) , which we call a 4^{\prime} -space (and we say that x and \xi are close ). Also, \mathsf{Res}(p) is the Lie incidence geometry \mathsf{D_{5,5}}(\mathbb{K}) . Fact 2.2 . Two symps intersect either in a point, or in a maximal singular subspace of either, in which case we call the subspace a 4 -space and the symps adjacent . The 4 -spaces of a given symp \xi constitute an oriflamme class of the symp, which is a polar space of type \mathsf{D_{5,1}} ; the 4^{\prime} -spaces contained in \xi form the other oriflamme class. The following lemmas can be read off the diagram, checked in an apartment, and is contained in [ 28 , §3.2] , but can also be proved using the above lemmas. Fact 2.3 . Each 3 -space of \mathsf{E_{6,1}}(\mathbb{K}) is the intersection of a unique 4 -space and a unique 5 -space. Fact 2.4 . There are two kinds of maximal singular subspaces in \mathsf{E_{6,1}}(\mathbb{K}) . One kind corresponds to the 4 -spaces, the other to 5 -dimensional projective spaces, called 5 -spaces, which contain 4^{\prime} -spaces. Let p be a point and W a 5 -space. Then either p\in W , or p^{\perp}\cap W is a 3 -dimensional space (called a 3 -space; p and W are called close ), or p is collinear to a unique point of W (and p and W are called far ). Let W and W^{\prime} be two distinct 5 -spaces. Then either W\cap W^{\prime} is a plane (then W and W^{\prime} are called adjacent ), or W\cap W^{\prime} is just a point, or W and W^{\prime} are disjoint and there exists a unique 5 -space intersecting both in a respective plane, or W and W^{\prime} are disjoint and opposite in the building; in the latter case, every point of W is far from W^{\prime} and every point of W^{\prime} is far from W , and collinearity defines a collineation between W and W^{\prime} . . We can now prove some (new) lemmas. Lemma 2.5 . Let x be a point and \xi a symp of \mathsf{E_{6,1}}(\mathbb{K}) . Then x is opposite \xi if, and only if, for some point y\in\xi , the symp \xi(x,y) intersects \xi only in y if, and only if, for all points y\in\xi , the symp \xi(x,y) intersects \xi only in y . Proof. If x is opposite \xi , then \xi\cap\xi(x,y) can not be more than just the point y , for every y\in\xi , because otherwise, \xi\cap\xi(x,y) is a 4 -space by 2.2 and x is collinear to a 3 -space of that 4 -space, contradicting the fact that x is opposite \xi . Now suppose that \xi\cap\xi(x,y)=\{y\} , for some fixed y\in\xi . We want to see that x has to be opposite \xi . Suppose there exists some z\in\xi at distance 2 from x , such that \xi\cap\xi(x,z) is a 4 -space (cf. 2.1 ). Then y and x have to be collinear to 3 -spaces U_{y} and U_{x} , respectively, of that 4 -space, which will necessarily intersect in at least a plane \pi . But then \pi will also be contained in \xi(x,y) , by convexity, and hence \xi\cap\xi(x,y) will be more than just the point y , which is a contradiction. So for every point z\in\xi\setminus x^{\perp} , we have \xi(x,z)=\{z\} . That means that x is not collinear to any point in \xi and thus, x is opposite \xi . ∎ Lemma 2.6 . Let L be a line and \xi a symp of \mathsf{E_{6,1}}(\mathbb{K}) with L\cap\xi=\varnothing . If no point of L is opposite \xi , then L is collinear to a unique plane of \xi . Proof. If no point of L is opposite \xi , then every point of L is collinear to a 4^{\prime} -space of \xi . Two 4^{\prime} -spaces in a symp intersect in either a point or a plane or they coincide, because they belong to the same oriflamme class by 2.4 . If two points of L were collinear to the same 4^{\prime} -space of \xi , then every point of that 4^{\prime} -space would be collinear to every point of L and L and that 4^{\prime} -space would span a 6 -space, which is impossible by 2.4 . Now, let x and y be two points of L , let V be the 4^{\prime} -space that y is collinear to in \xi and let x^{\prime} be some point of \xi\setminus V that x is collinear to. Then x^{\prime} has to be collinear to a 3 -space U\subseteq V . The symp \xi(y,x^{\prime}) contains U and x . With that, x has to be collinear to a plane \pi of U . That means x^{\perp}\cap\xi and y^{\perp}\cap\xi intersect in \pi and since every point of \pi is collinear to x,y\in L , and \mathsf{E_{6,1}}(\mathbb{K}) is a gamma space, every point of \pi has to be collinear to every point of L . With that, L is collinear to a unique plane of \xi . ∎ Lemma 2.7 . Let p be a point and M a 3 -space of \mathsf{E_{6,1}}(\mathbb{K}) , such that no point of M is collinear to p . Then the unique maximal 4 -space containing M does not contain any point collinear to p . Proof. Suppose a point p is collinear to a point q of a 4 -space C containing a 3 -space M which does not contain any point collinear to p . Put C in a symp \xi . If p is in \xi , then p is collinear to a 3 -space of C , which intersects M , a contradiction. If p is not in \xi , then p is collinear to a 4^{\prime} -space W , which already intersects C in q . But the intersection must have odd codimension, hence the intersection C\cap W is either a line or a 3 -space. Both would intersect M , a contradiction. ∎ Lemma 2.8 . Let W,W^{\prime} be two opposite 5 -spaces, and let U be a 4 -space intersecting W in a 3 -space. Then there exists a unique point p\in U\setminus W close to W^{\prime} . Also, p^{\perp}\cap W^{\prime}=\{x^{\prime}\in W^{\prime}\mid(\exists x\in W)(x^{\prime}\perp x)\} . Proof. Putting W,W^{\prime} and U in a common apartment, the existence of p readily follows. 2.7 shows the second assertion, and then uniqueness of p also follows immediately. ∎ Lemma 2.9 . Let L be a line, and let b be a point not collinear to any point of L in \mathsf{E_{6,1}}(\mathbb{K}) . Then \langle b,b^{\perp}\cap L^{\perp}\rangle is a maximal 4 -space. Proof. By convexity of symps. \langle b,b^{\perp}\cap L^{\perp}\rangle is contained in each symp \xi(b,p) , with p\in L . Fix such a symp \xi=\xi(b,p) , for some p\in L . 2.1 implies that L is collinear to a 4^{\prime} -space U^{\prime} of \xi . Then U:=\langle b,b^{\perp}\cap U^{\prime}\rangle is a 4 -space, because U and U^{\prime} belong to different oriflamme classes. By 2.4 , U is a maximal singular subspace and the assertions are proved. ∎ 2.3. Lie incidence geometries of type \mathsf{E_{7,7}} For each field \mathbb{K} there exists a unique Lie incidence geometry isomorphic to \mathsf{E_{7,7}}(\mathbb{K}) . It is a strong parapolar space of diameter 3 and also called a minuscule geometry . The following properties can easily be derived from an apartment of the corresponding building (in this case the 1 -skeleton of such an apartment, where vertices are points of \mathsf{E_{7,7}}(\mathbb{K}) and edges are lines, is the Gosset graph; see [ 6 , §10.3.5] ). Fact 2.10 . Let \Delta be the Lie incidence geometry \mathsf{E_{7,7}}(\mathbb{K}) . Then the following assertions hold. (i) \Delta is strong, has uniform symplectic rank 6 , singular rank 7 and diameter 3 . Points at distance 3 are opposite. (ii) Symps in \Delta are isomorphic to hyperbolic polar spaces \mathsf{D_{6,1}}(\mathbb{K}) . (iii) Point residuals in \Delta are isomorphic to \mathsf{E_{6,1}}(\mathbb{K}) . (iv) The maximal singular subspaces of highest dimension in \Delta are projective spaces of dimension 6 . Like before, we call 5 -dimensional projective subspaces contained in 6 -spaces 5^{\prime} -spaces. (v) Maximal 5 -spaces occur as the intersection of two symps. On the other hand, 5^{\prime} -spaces occur as the intersection of a unique 6 -space and a unique symp. (vi) For each symp \xi , its 5 -spaces form an oriflamme class, and its 5^{\prime} -spaces form the other oriflamme class of \xi . Fact 2.11 . Let p be a point and \xi a symp of \mathsf{E_{7,7}}(\mathbb{K}) , with p\notin\xi . Then precisely one of the following occurs. (i) p is collinear to a 5^{\prime} -space A of \xi , p is symplectic to the points of \xi\setminus A , and we say that p is close to \xi . (ii) p is collinear to a unique point q\in\xi , p is symplectic to the points of \xi\cap(q^{\perp}\setminus\{q\}) , and p is opposite to the points \xi\setminus q^{\perp} . We say that p is far from \xi . This fact implies that, on each line L , there is at least one point symplectic to a given point p (unique when L contains at least one point opposite p ). Fact 2.12 . Let \xi and \xi^{\prime} be two distinct symps of \mathsf{E_{7,7}}(\mathbb{K}) . Then precisely one of the following occurs. (i) \xi\cap\xi^{\prime} is a 5 -space, and we call \xi and \xi^{\prime} adjacent . (ii) \xi\cap\xi^{\prime} is a line L . Then points x\in\xi\setminus L and x^{\prime}\in\xi^{\prime}\setminus L are never collinear. We call \{\xi,\xi^{\prime}\} symplectic . (iii) \xi\cap\xi^{\prime}=\varnothing , and there is a unique symp \xi^{\prime\prime} intersecting both \xi and \xi^{\prime} in respective 5 -spaces A and A^{\prime} , which are opposite in \xi^{\prime\prime} . All points of \xi\setminus A are far from \xi^{\prime} , and each point of A is close to \xi^{\prime} . Each line containing a point of \xi and a point of \xi^{\prime} contains a point of A\cup A^{\prime} . We call \{\xi,\xi^{\prime}\} special . (iv) \xi\cap\xi^{\prime}=\varnothing , and every point of \xi is far from \xi^{\prime} . In this situation, each point of \xi^{\prime} is also far from \xi , and \xi and \xi^{\prime} are opposite. Fact 2.13 . Let \xi_{1} and \xi_{2} be two opposite symps of \mathsf{E_{7,7}}(\mathbb{K}) . Let \mathscr{L} be the set of all lines that contain a point of \xi_{1} and a point of \xi_{2} . Then, for each point p that is contained in a line of \mathscr{L} , there exists a unique symp \xi_{p} that intersects each line L\in\mathscr{L} . We can now show the following lemma. Lemma 2.14 . Let x,y be two points of \mathsf{E_{7,7}}(\mathbb{K}) . Then x and y are opposite if, and only if, they are contained in respective symps intersecting in a line L such that x and y are collinear to unique respective distinct points of L . Proof. First suppose that x and y are opposite. Let \xi_{x} be an arbitrary symp containing x . 2.11 (ii) implies that y is collinear to a unique point z\in\xi_{x} . Let \xi_{y} be an arbitrary symp through y and z . If \xi_{x} and \xi_{y} intersected in more than a line, they would intersect in a 5 -space (cf. 2.12 ), and both x and y would have to be collinear to 4 -spaces of that 5 -space. These would necessarily intersect, meaning that there would exist points collinear to both x and y , contradicting the fact that x and y are opposite. Hence \xi_{x}\cap\xi_{y} is a line L . Now, clearly x and y are not collinear to a common point on L , and the assertion follows. Next suppose, conversely, that x and y are contained in respective symps \xi_{x} and \xi_{y} intersecting in a line L such that x and y are collinear to unique respective distinct points x^{\prime} and y^{\prime} of L . Suppose, for a contradiction, that x is collinear to a 5^{\prime} -space U of \xi_{y} . Then y^{\prime\perp}\cap x^{\perp} contains points of U that do not belong to \xi_{x} , a contradiction. Hence x is far from \xi_{y} , and the assertion follows from 2.11 (ii) . ∎ 2.4. Lie incidence geometries of types \mathsf{E_{6,2},E_{7,1},E_{8,8},F_{4,1}} and \mathsf{F_{4,4}} . These Lie incidence geometries are examples of hexagonic geometries , as defined by Shult [ 26 , Section 13.7] , inspired by his work with Kasikova [ 20 ] . We will not need the formal definition of such geometries; some defining properties will be part of the facts that we state below. Since we are only concerned with exceptional geometries, we will restrict ourselves to these cases. This implies, for instance, that we can assume that the parapolar space in question has uniform symplectic rank (which is at least 3 ). Other examples of hexagonic Lie incidence geometries are the line-Grassmannians of polar spaces, which have symplectic rank at least 3 if the polar space has rank at least 4 , and uniform symplectic rank 3 if, and only if, the polar space has rank 4 . Many facts stated below also hold for these spaces, but can in that case easily be directly checked in the polar space. The following facts can again be easily checked in an apartment (for models of such, see [ 33 ] ), or follow from the diagram. We will refer to the geometries in the title of this section as the exceptional hexagonic (Lie incidence) geometries . The ones of type \mathsf{F_{4,1}} and \mathsf{F_{4,4}} are also known as (thick) metasymplectic spaces . A detailed introduction to the latter is contained in [ 22 ] . We will always assume thickness when mentioning metasymplectic spaces. Fact 2.15 . Let x and y be two distinct non-collinear points of an exceptional hexagonic Lie incidence geometry. Then x and y are either symplectic, special, or opposite. In the latter case, the distance between x and y is 3 . The set x^{\not\equiv} is always a proper geometric hyperplane. The following fact can also be deduced from [ 13 , Lemma 2(v)] . Fact 2.16 . Let x and u be two points of an exceptional hexagonic Lie incidence geometry. Let x\perp y\perp z\perp u . Then x and u are opposite if, and only if, both \{x,z\} and \{y,u\} are special pairs. In particular, if x\perp\hskip-4.26773pt\perp v\perp u for some point v , then x and u are not opposite. Fact 2.17 . Let x be a point and \xi a symp of an exceptional hexagonic Lie incidence geometry. Then exactly one of the following occurs. (i) x\in\xi ; (ii) x^{\perp}\cap\xi is a maximal singular subspace in \xi (this cannot happen in a metasymplectic space); (iii) x^{\perp}\cap\xi is a line L ; (iv) x^{\perp\hskip-4.26773pt\perp}\cap\xi is a maximal singular subspace U ; (v) x^{\perp\hskip-4.26773pt\perp}\cap\xi=\xi (this does not occur in types \mathsf{F_{4},E_{8}} ); (vi) x^{\perp\hskip-4.26773pt\perp}\cap\xi is a unique point y . The following fact follows from the diagrams by taking point residuals. Fact 2.18 . Let x be a point of the parapolar space \Delta isomorphic to either \mathsf{E_{6,2}}(\mathbb{K}),\mathsf{E_{7,1}}(\mathbb{K}),\mathsf{E_{8,8}}(\mathbb{K}) , or a metasymplectic space. Then \mathsf{Res}_{\Delta}(x) is isomorphic to \mathsf{A_{5,3}}(\mathbb{K}),\mathsf{D_{6,6}}(\mathbb{K}),\mathsf{E_{7,7}}(\mathbb{K}) , or a dual polar space of rank 3 , respectively. Consequently, when two symps of \Delta have a line in common, then they have (at least) a plane in common. Two non-disjoint symps either intersect in a point, a plane, or a maximal singular subspace. The singular rank of \Delta is 5,7,8 or 3 , respectively. In general, an i -dimensional singular subspace whose points do not correspond to the set of vertices of the corresponding building contained in a simplex together with another given vertex, will be called an i^{\prime} -space . It usually arises as the intersection of a maximal singular subspace with a symp. We can now be more specific in 2.17 . Lemma 2.19 . Let \Delta be an exceptional hexagonic Lie incidence geometry. Let x be a point of \Delta and \xi a symp of \Delta . Then the following hold. (i) If x^{\perp}\cap\xi is a maximal singular subspace U in \xi , then each point of \xi\setminus U is symplectic to x ; (ii) If x^{\perp}\cap\xi is a line L , then each point y of \xi\setminus L collinear to a unique point of L is special to x (the other points of \xi\setminus L are symplectic to x ); (iii) If x^{\perp\hskip-4.26773pt\perp}\cap\xi is a maximal singular subspace U , then each point of \xi\setminus U is special to x ; (iv) If x^{\perp\hskip-4.26773pt\perp}\cap\xi is a unique point y , then each point of \xi not collinear to y is opposite x (consequently, each other point of \xi\setminus\{y\} is special to x ). Proof. (i) The maximal singular subspace U of \xi has dimension at least 2 , hence for each y\in\xi\setminus U , the set y^{\perp}\cap U has at least three elements, implying, by the definition of parapolar spaces, that x and y are symplectic. (ii) Suppose x and y were symplectic. Then the symps \xi and \xi(x,y) would have a line in common, hence, by 2.18 , they would share a plane \alpha , which has to contain L as x^{\perp}\cap\alpha is a line. Since also y\in\alpha , y is collinear to all points of L , which contradicts the assumptions. (iii) This follows immediately from 2.16 . (iv) No point of \xi is collinear to x , as otherwise it follows from the other cases that y is not unique. Hence all points of \xi collinear to y are special to x . Let z be such a point. Then, by the previous possibilities, z^{\perp}\cap\xi(x,y) is a line K and u:=[x,z]\in K . Suppose, for a contradiction, that u^{\perp}\cap\xi is a maximal singular subspace U of \xi . Then, by 2.17 (ii) and 2.18 , \dim U\geq 3 and any symp \xi(u,w) , with w\in\xi\setminus u^{\perp} , shares a maximal singular subspace W with \xi . It follows from 2.17 that x^{\perp}\cap\xi(u,w) is at least a line M . But then each point of M^{\perp}\cap W , which is at least 1 -dimensional, is symplectic to x , a contradiction. So, u^{\perp}\cap\xi is a line N , and it follows from the previous possibility that u\Join v , for every point v\in\xi\setminus y^{\perp} . Now 2.16 proves the assertion. ∎ Lemma 2.20 . Let x be a point of an exceptional hexagonic Lie incidence geometry, and suppose x is special to y_{1} and y_{2} , with y_{1}\perp y_{2} . Then also the points z_{1}=[x,y_{1}] and z_{2}=[x,y_{2}] are collinear. Proof. By 2.16 , z_{1} and y_{2} are symplectic. The point x is collinear to z_{1} , and hence to a line L of \xi(z_{1},y_{2}) . Then y_{2} has to be collinear to a point of L , but this point can only be z_{2} , since x and y_{2} are special. Note that it is possible that z_{1}=z_{2} . ∎ Lemma 2.21 . Let x and L be a point and line, respectively, of an exceptional hexagonic Lie incidence geometry, and suppose that x is special to each point of L . Then there exists a line M consisting of the points collinear to x and some point of L . Consequently, if x is special to at least two points y_{1},y_{2} of a line K , with [x,y_{1}]=[x,y_{2}] , then it is either collinear or symplectic to a unique point of K , and special to the other points of K . Proof. Let y_{1} and y_{2} be two points on L and set z_{i}:=[x,y_{i}] , i=1,2 . By 2.16 , z_{1} and y_{2} are symplectic, and the symp \xi(z_{1},y_{2})=:\xi contains y_{1} . Since x has to be collinear to a line M of \xi , z_{2} is contained in M , and hence in \xi as well. Suppose z_{1}=z_{2} . Let p be a point on M distinct from z_{1} , and let q be the projection of p onto L . Since z_{1} is collinear to y_{1} and y_{2} , z_{1} is collinear to every point on L , including q . Thus, q is collinear to every point of M , and it follows that x^{\perp}\cap q^{\perp}\supseteq M , and thus, x and q are symplectic, which contradicts the assumptions. Therefore, z_{1} and z_{2} are distinct. By 2.20 it follows that z_{1} and z_{2} are collinear and z_{1}z_{2}=M . Now, for every point b on L , the point [x,b] is the unique point on M collinear to b . ∎ Similarly, we can say something about a point being symplectic to all points of a line. Lemma 2.22 . Let x and L be a point and line, respectively, of an exceptional hexagonic Lie incidence geometry, and suppose that x is symplectic to each point of L . Then there exists a maximal singular subspace W not contained in a symp, containing L , such that x^{\perp}\cap W is complementary to L in W , and each symp containing x and a point of L contains x^{\perp}\cap W . Proof. Pick x_{1},x_{2}\in L and set \xi_{1}:=\xi(x,x_{1}) . Then x_{2}^{\perp}\cap\xi_{1} is either a line or a maximal singular subspace (cf. 2.17 ). If it were a line M , then, by 2.19 , x would be special to x_{2} , as x_{1}\in M and x_{2} is not collinear to x . Hence x_{2}^{\perp}\cap\xi_{1} is a maximal singular subspace U . Defining W as the singular subspace generated by U and x_{2} , the assertions follow. ∎ Lemma 2.23 . Let \xi_{1} and \xi_{2} be two non-disjoint symps of an exceptional hexagonic Lie incidence geometry, and let x_{i}\in\xi_{i} , i=1,2 , be two points. Then x_{1}\equiv x_{2} if, and only if, \xi_{1}\cap\xi_{2} is a point z , the symps \xi_{1} and \xi_{2} are locally opposite at z , and \{x_{i},z\} is a symplectic pair, i=1,2 . Proof. This follows from 2.16 and 2.19 , taking into account that \xi_{1} and \xi_{2} are locally opposite at an intersection point z if, and only if, each point z_{i}\in\xi_{i}\setminus\{z\} is collinear to a unique line of \xi_{j} , \{i,j\}=\{1,2\} (which can easily be seen in \mathsf{Res}(x) ; for instance, for \mathsf{Res}(x)\cong\mathsf{E_{7,7}}(\mathbb{K}) , this is 2.12 (iv) ). ∎ Lemma 2.24 . Let \xi_{1} and \xi_{2} be two opposite symps of an exceptional hexagonic Lie incidence geometry, and let L_{1}\subseteq\xi_{1} be a line. Then the set of points of \xi_{2} symplectic to some point of L_{1} is a line L_{2} of \xi_{2} . All symps having a point on L_{1} and a point on L_{2} share a unique common point x , which is collinear to both L_{1} and L_{2} . Proof. First, we note that, from general building-theoretic considerations, every point in \xi_{1} has an opposite in \xi_{2} ; hence, if x_{1}\in L_{1} and x_{2}\in x_{1}^{\perp\hskip-4.26773pt\perp}\cap\xi_{2} , then 2.23 implies that \xi_{2} and \xi(x_{1},x_{2}) are locally opposite at x_{2} and, by 2.17 , every point of \xi_{2} not collinear to x_{2} is opposite x_{1} . If such a point y_{2} were symplectic to a point y_{1}\in L_{1} , then 2.17 would again imply that x_{1} , being collinear with y_{1} , is special to y_{2} , a contradiction. We conclude that “being symplectic” preserves collinearity in both directions (interchanging the roles of \xi_{1} and \xi_{2} ), and hence is an isomorphism between \xi_{1} and \xi_{2} . Let L_{2}\subseteq\xi_{2} correspond to L_{1} under that isomorphism. Let x_{1}^{\prime}\in L_{1}\setminus\{x_{1}\} . Then there is a unique point [x_{1}^{\prime},x_{2}]=:x\in\xi(x_{1},x_{2}) collinear to x_{2} and x_{1}^{\prime} . Clearly, x\perp L_{1} . Standard arguments switching roles of points on L_{1} and L_{2} imply that x is independent of x_{1} and x_{2} , and so x is collinear to each point of L_{1}\cup L_{2} . It is now easy to see that x is contained in each symp containing a point of L_{1} and a point of L_{2} . Uniqueness of x follows from the fact that x=[x_{1}^{\prime},x_{2}] . ∎ Lemma 2.25 . Let \Delta be an exceptional hexagonic Lie incidence geometry, and let \{x_{1},x_{2}\} be a symplectic pair of points in \Delta . Let x_{2}^{\prime} be another point in \Delta symplectic to x_{1} and collinear to x_{2} . Let x_{1}^{\prime} be a point such that x_{1}\perp x_{1}^{\prime}\perp x_{2}^{\prime} . Then x_{1}^{\prime} and x_{2} cannot be special. Proof. The point x_{1}^{\prime} is collinear to (at least) a line L of \xi(x_{1},x_{2}) , which contains x_{1} and a point y collinear to x_{2} . If x_{1}^{\prime} were special to x_{2} , then y=[x_{1}^{\prime},x_{2}]=x_{2}^{\prime}\perp\hskip-4.26773pt\perp x_{1} , contradicting y\perp x_{1} . ∎ 2.19 implies that, whenever two points x,y of an exceptional hexagonic Lie incidence geometry \Delta are opposite, then every symp through x contains a unique point symplectic to y . This way, one obtains all points x^{\perp\hskip-4.26773pt\perp}\cap y^{\perp\hskip-4.26773pt\perp} . This defines a subspace of \Delta which we call the equator geometry (with poles x and y ) and denote as E(x,y) . We view these as point-line geometries as soon as they contain lines. The latter is the case in the simply laced case (types \mathsf{A}_{n},\mathsf{D}_{n},\mathsf{E_{6},E_{7}} and \mathsf{E_{8}} ). In all those cases, these equator geometries can be defined in exactly the same way for the corresponding hexagonic geometries (here the Lie incidence geometries of types \mathsf{A}_{n,\{1,n\}} , \mathsf{D}_{n,2} , and the exceptional ones not of type \mathsf{F_{4}} ), and we have the following sequences (where \Delta\to\Delta^{\prime} means that \Delta^{\prime} is an equator geometry of \Delta ), which can be deduced from [ 16 ] : \begin{cases}\mathsf{E_{8,8}}(\mathbb{K})\to\mathsf{E_{7,1}}(\mathbb{K})\to\mathsf{D_{6,2}}(\mathbb{K}),\\ \mathsf{E_{6,2}}(\mathbb{K})\to\mathsf{A_{5,\{1,5\}}}(\mathbb{K})\to\mathsf{A_{3,\{1,3\}}}(\mathbb{K}).\end{cases} In the case of type \mathsf{F_{4}} , equator geometries as defined here have no lines. We shall give an alternative definition for that case in the next paragraph. 2.5. Metasymplectic spaces The previous paragraph includes the metasymplectic spaces, that is, the Lie incidence geometries of types \mathsf{F_{4,1}} and \mathsf{F_{4,4}} . We now introduce some notation making apparent the differences between \mathsf{F_{4,1}} and \mathsf{F_{4,4}} , based on the Dynkin diagram of type \mathsf{F_{4}} rather than the Coxeter diagram. Everything in the paragraph can be found in [ 22 ] and is, of course, based on the fundamental work of Tits in [ 30 ] . Given a field \mathbb{K} , a quadratic algebra \mathbb{A} over \mathbb{K} is an algebra that admits a bilinear form b:\mathbb{A}\to\mathbb{K} such that for every x\in\mathbb{A} we have x^{2}-(b(1,x)+b(x,1))x+b(x,x)=0 . The element b(x,x) is called the norm of x and briefly denoted as \mathsf{n}(x) . We assume that \mathbb{A} is alternative , that is, \mathbb{A} satisfies the alternative laws (ab)b=ab^{2} and a(ab)=a^{2}b , and that \mathbb{A} is a unital division algebra, that is, it has an identity and every element has an inverse. We can associate a polar space of rank r\geq 2 with every quadratic alternative unital division algebra as follows. Let V be the vector space isomorphic to the direct sum of \mathbb{A} and 2r copies of \mathbb{K} . Then define the quadratic form \beta:V\to\mathbb{K}:(x_{-r},x_{-r+1},\ldots,x_{-1},x_{0},x_{1},\ldots,x_{r})\mapsto x_{-1}x_{1}+x_{-2}x_{2}+\cdots+x_{-r}x_{r}-\mathsf{n}(x_{0}), where x_{0}\in\mathbb{A} and x_{i}\in\mathbb{K} , for all i\in\{-r,-r+1,\ldots,-1,1,2,\ldots,r\} . Then the null set of \beta defines a quadric of Witt index r , whose natural point-line geometry is a polar space of rank r , which we denote by \mathsf{B}_{r,1}(\mathbb{K},\mathbb{A}) . From the classification of buildings of type \mathsf{F_{4}} in [ 30 ] , we know that such a building is uniquely determined by a field \mathbb{K} and a quadratic alternative unital division algebra \mathbb{A} over \mathbb{K} . We denote that building by \mathsf{F_{4}}(\mathbb{K},\mathbb{A}) , where we usually substitute \mathbb{K} and \mathbb{A} with their sizes if they are finite. Now, we assign the type function to the diagram in such a way that the symps of \mathsf{F_{4,1}}(\mathbb{K},\mathbb{A}) are isomorphic to the polar space \mathsf{B_{3,1}}(\mathbb{K},\mathbb{A}) . The building \mathsf{F_{4}}(\mathbb{K},\mathbb{K}) will sometimes be referred to as “split”. It is also characterised by the fact that the residues of type 4 correspond to symplectic polar spaces, that is, polar spaces defined by a non-degenerate alternating bilinear form, or equivalently, a null polarity. If we define a hyperbolic line of a polar space as the set (x^{\perp}\cap y^{\perp})^{\perp}=\{x,y\}^{\perp\perp} , for two non-collinear points x and y , then, in a symplectic polar space, a hyperbolic line is an ordinary line of the ambient projective space which is not a line of the polar space. We now define equator geometries. Let p,q be two opposite points of \mathsf{F}_{4,1}(\mathbb{K},\mathbb{A}) , i\in\{1,4\} . The equator E(p,q) is the set of points that are symplectic simultaneously to p and q . The intersection of E(p,q) with the union of the symps through a given plane containing either p or q is, by definition, a line of the equator geometry . One checks that, replacing “plane” with “maximal singular subspace contained in a symp”, this definition applied to the simply laced case provides the same equator geometries as defined earlier (this is proved explicitly in many cases in [ 16 ] ). We are going to briefly need the extended equator geometry, but only for the split case \mathbb{A}=\mathbb{K} . Let E(p,q) be an equator of \mathsf{F_{4,4}}(\mathbb{K},\mathbb{K}) and define \widehat{E}(p,q) as the set of points symplectic to at least two opposite points of E(p,q) . Endow \widehat{E}(p,q) with all lines of each equator geometry included in it. Then we obtain the extended equator geometry , also denoted by \widehat{E}(p,q) . This time, p,q\in\widehat{E}(p,q) . It is always isomorphic to \mathsf{B_{4,1}}(\mathbb{K},\mathbb{K}) , see [ 22 ] . If \mathbb{K} is perfect of characteristic 2 , then notice that \mathsf{B_{4,1}}(\mathbb{K},\mathbb{K}) is a symplectic polar space. 2.6. Generalised hexagons Our results include buildings of type \mathsf{G_{2}} ; the associated point-line geometries are better known as generalised hexagons, introduced by Tits [ 29 ] . For n\geq 3 , a generalised n -gon , or generalised polygon if we do not want to specify n , is a point-line geometry \Gamma=(X,\mathscr{L}) such that the (bipartite) graph on X\cup\mathscr{L} , with x\in X adjacent to L\in\mathscr{L} if x\in L , has diameter n and girth 2n (we call this graph the incidence graph of \Gamma ). We also assume that every line has at least three points and every point is contained in at least three lines (thickness of the associated building). Generalised 3 -gons are the same things as projective planes, and generalised 4 -gons, also known as generalised quadrangles, are polar spaces of rank 2 . For more general background and results on generalised n -gons, see [ 32 ] . Finite generalised quadrangles are studied in detail in [ 23 ] . We recall the following definitions. The order of a generalised n -gon is the pair (s,t) such that each line contains precisely s+1 points and each point is contained in precisely t+1 lines. A spread in a generalised quadrangle \Gamma=(X,\mathscr{L}) is a partition of X into members of \mathscr{L} . If \Gamma has order (s,t) , then a spread contains exactly 1+st lines. A subpolygon of a generalised polygon is the generalised polygon induced on a convex subgraph of the incidence graph of \Gamma . Here, we are particularly interested in generalised hexagons, and more specifically in those that satisfy the Moufang condition , as these are the counterparts of type \mathsf{G_{2}} of the spherical buildings of rank at least 3 (since these automatically satisfy such condition) and are the natural geometries for the simple algebraic groups of that type. Recall from [ 31 ] that a Moufang hexagon is determined by a field \mathbb{K} and a quadratic Jordan division algebra \mathbb{J} over \mathbb{K} ; using the Bourbaki labelling of nodes for Dynkin diagrams, we define \mathsf{G_{2,2}}(\mathbb{K},\mathbb{J}) as the hexagon where the point rows are parametrised by \mathbb{J} and the line pencils by \mathbb{K} . There is perhaps ambiguity when \mathbb{J}=\mathbb{K} , but we take that ambiguity away by mentioning that, in this case, \mathsf{G_{2,2}}(\mathbb{K},\mathbb{K}) is the hexagon arising from a triality of type I id , as defined by Tits [ 29 ] . Such a hexagon shall be called the split Cayley hexagon (as in [ 32 ] ), and the corresponding building is referred to as being split . The triality hexagon is \mathsf{G_{2,2}}(\mathbb{K},\mathbb{J}) , where \mathbb{J} is a cubic Galois field extension of \mathbb{K} . We will only need this hexagon in the finite case. The finite triality hexagon has order (q^{3},q) , for some prime power q . In the finite case, we replace the field and the Jordan algebra with their sizes in the notation, just like we did for finite metasymplectic spaces. Let \Gamma=(X,\mathscr{L}) be a generalised hexagon. We use the notation and terminology of parapolar spaces to express the mutual position of two points, with the obvious meaning. In particular, points are special if they are not collinear but they are collinear to a unique common point. With that, a hyperbolic line H in \Gamma is a set of mutual special points collinear to a common given point p such that, for each point q opposite p , we have q^{\Join}\cap p^{\perp}=H as soon as |q^{\Join}\cap H|\geq 2 . If we call a point of \Gamma close to a line if it is collinear to a unique point of that line, then a distance- 3 trace in \Gamma is the set of points close to two given opposite lines L and M . We denote it by [L,M]_{3} . It is called regular if [N,M]_{3}=[L,M]_{3} whenever N is opposite M and |[N,M]_{3}\cap[L,M]_{3}|\geq 2 . There are reasons to call a distance- 3 trace in a split Cayley hexagon over a perfect field in characteristic 2 an imaginary line , as we shall explain in the proof of 5.9 . We have adopted this terminology in A and B above, too. The terminology between parentheses, namely ideal line , is Ronan’s terminology [ 25 ] . 2.7. Opposition and projections In this final paragraph of the preliminaries, we invoke some general theory about opposition, define projections, and note down some consequences which are rather interesting in our context because they unify some arguments across all types. Let F and F^{\prime} be opposite simplices in a spherical building \Omega . For us, F and F^{\prime} will almost always be vertices, but we choose to state and define things slightly more generally. Then, by [ 30 , look up] , for every chamber C\supseteq F there exists a unique chamber C^{\prime}\supseteq F^{\prime} at nearest (gallery) distance from C . By [ 30 , look up] , the map C\mapsto C^{\prime} induces an isomorphism from \mathsf{Res}_{\Omega}(F) to \mathsf{Res}_{\Omega}(F^{\prime}) . This means that for every vertex v joinable to F , there exists a unique vertex v^{\prime} joinable to F^{\prime} and closest to v ; we denote v^{\prime}=\mathsf{proj}^{F}_{F^{\prime}}(v) and we call v^{\prime} the projection of v onto F^{\prime} . If F is obvious or not important, we sometimes write \mathsf{proj}_{F^{\prime}} instead of \mathsf{proj}_{F^{\prime}}^{F} . For example, the map in the proof of 2.24 between the two opposite symps \xi_{1} and \xi_{2} of an exceptional hexagonic Lie incidence geometry \Delta , given on the points by “being symplectic”, coincides with the pair of projections \mathsf{proj}_{\xi^{\prime}}^{\xi} and \mathsf{proj}_{\xi}^{\xi^{\prime}} . As a second example, the projection \mathsf{proj}^{y}_{x} from a point x to an opposite point y in an exceptional hexagonic Lie incidence geometry maps a line L through x to the unique line through y containing a point collinear to some point of L . The following notion will be very convenient to classify geometric lines. Let \Omega be any spherical building and let v_{1},v_{2},v_{3} be three vertices of the same type. Then, using the terminology of [ 21 ] , we call \{v_{1},v_{2},v_{3}\} a round-up triple if no vertex of \Omega is opposite exactly one of v_{1},v_{2},v_{3} . Then, by definition, round-up triples in Lie incidence geometries correspond to round-up triples in the corresponding building. We will make extensive use of round-up triples when classifying geometric lines in this paper. We have the following connection between global and local opposition. Proposition 2.26 (Proposition 3.29 of [ 30 ] ) . Let F and F^{\prime} be opposite simplices of a spherical building \Omega . Let v be a vertex of \Omega adjacent to each vertex of F , and let i be the type of v . Then the type i^{\prime} of the vertex \mathsf{proj}^{F}_{F^{\prime}}(v) is the opposite in \mathsf{Res}(F^{\prime}) of the opposite type of i in \Omega . Also, vertices v\sim F and v^{\prime}\sim F^{\prime} are opposite in \Omega if, and only if, v^{\prime} is opposite \mathsf{proj}^{F}_{F^{\prime}}(v) in \mathsf{Res}_{\Omega}(F^{\prime}) . Corollary 2.27 . Let F be a simplex of a spherical building \Omega . Then a collection T of vertices in \mathsf{Res}(F) admits an opposite in \mathsf{Res}(F) (viewed as a spherical building on its own) if and only if it admits an opposite vertex in \Omega . Also, a set of vertices in \mathsf{Res}(F) is a geometric line of \mathsf{Res}(F) if, and only if, it is a geometric line of \Omega . A triplet of vertices is a round-up triple in \mathsf{Res}(F) if, and only if, it is a round-up triple in \Omega . Proof. If v is a vertex in \mathsf{Res}(F) opposite each vertex of T in \mathsf{Res}(F) , then, for any simplex F^{\prime} opposite F , the vertex \mathsf{proj}^{F}_{F^{\prime}}(v) is opposite every member of T by 2.26 . Conversely, let v^{\prime} be a vertex of \Omega opposite each member of T . Select t\in T . Applying 2.26 , we find a simplex F^{\prime} in \mathsf{proj}^{t}_{v^{\prime}}(F) opposite F . Then, again by 2.26 , \mathsf{proj}^{F^{\prime}}_{F}(v^{\prime}) is, in \mathsf{Res}(F) , opposite every member of T . The second and third assertions now also follow. ∎ If we want to use 2.27 , then we have to know something about blocking sets, geometric lines, or round-up triples in residues; if these are classical, then from [ 4 ] , [ 10 ] , and [ 21 ] , we infer: Proposition 2.28 . (1) Let T be a set of q+1 vertices of type j in either \mathsf{A_{n}}(q) , 1\leq j\leq n , n\geq 2 , or \mathsf{D_{n}}(q) , 1\leq j\leq n , n\geq 4 , such that no vertex is opposite all of them. Then T is a line in the corresponding Lie incidence geometry \mathsf{A_{n,j}}(q) , or \mathsf{D}_{n,j}(\mathbb{K}) , respectively. (2) Every geometric line of \mathsf{A_{n,j}}(\mathbb{K}) , 1\leq j\leq n , n\geq 2 , for \mathbb{K} an arbitrary skew field, is an ordinary line. Hence, every round-up triple of points in such a geometry is contained in an ordinary line. The same holds in \mathsf{D_{n,j}}(\mathbb{K}) , 1\leq j\leq n , n\geq 4 , for \mathbb{K} an arbitrary field. Proposition 2.29 . (1) Let L be a geometric line of a Lie incidence geometry \Gamma of type \mathsf{B}_{n,j} , 1\leq j\leq n . Then L is either a line of \Gamma , or j\neq n , \Gamma is the j -Grassmannian of a symplectic polar space \Delta , there exists a singular subspace U of dimension j-2 of \Delta , and L corresponds to a hyperbolic line of \mathsf{Res}_{\Delta}(U) , or j=n , all symps of \Gamma are symplectic polar spaces of rank 2 , and L is a hyperbolic line in a symp. (2) Let T be a round-up triple of points of a Lie incidence geometry \Gamma of type \mathsf{B}_{n,j} , 1\leq j\leq n , and let \Delta be the associated polar space. Then L is either contained in a line of \Gamma , or j\neq n , there exists a singular subspace U of \Delta of dimension j-2 , and T is contained in a hyperbolic line of \mathsf{Res}_{\Delta}(U) , or j=n , and L is contained in a hyperbolic line of a symp of \Gamma . (3) Let T be a set of at most q+1 vertices of type j in a finite (thick) building \Delta of type \mathsf{B}_{n} , j\leq n-1 , n\geq 3 , such that each panel of cotype j is contained in exactly q+1 chambers. Suppose no vertex of \Delta is opposite all members of T . Then T is either a geometric line in the corresponding Lie incidence geometry of type \mathsf{B}_{n,j} , or \Delta has a rank 2 residue corresponding to a generalised quadrangle with order (\sqrt{q},q) , and T is a spread in a subquadrangle \Gamma^{\prime} of order (\sqrt{q},\sqrt{q}) . In the latter case, \Gamma^{\prime} is a symplectic polar space, and q is a power of 2 . In particular, if |T|\leq q , then it always admits a vertex opposite all its members. Also, let us quote the following property, which we will use regularly. Proposition 2.30 (Proposition 8.5 of [ 9 ] ) . If every panel of a spherical building is contained in at least s+1 chambers, then every set of s chambers admits an opposite chamber. In particular, every set of s vertices admits an opposite vertex. Noting that in any apartment each vertex has a unique opposite, the following assertion is immediate by considering an apartment through S_{1} and S_{2} . Lemma 2.31 . If S_{1} and S_{2} are two distinct simplices of a spherical building, then there exists at least one simplex opposite S_{1} , but not opposite S_{2} . One more property we will use frequently is the following. Proposition 2.32 . Suppose every round-up triple of points in a Lie incidence geometry \Delta is contained in a line. Then each geometric line of \Delta is an ordinary line. Proof. Let L be a geometric line of \Delta . By 2.30 , L contains at least three elements. Pick x_{1},x_{2}\in L . Clearly, every triple of elements of L containing x_{1},x_{2} is a round-up triple and hence contained in a line M , which coincides with the unique line through x_{1} and x_{2} . Hence L\subseteq M . If L\neq M , then each point opposite x_{1} and not opposite some point of M\setminus L would be opposite all members of L , a contradiction. The proposition is proved. ∎ In the present paper, we will have to classify round-up triples in many geometries. The following general properties will come in handy. Lemma 2.33 . Let \{v_{1},v_{2},v_{3}\} be a round-up triple of vertices of some common type in a spherical building \Delta . If v_{1} and v_{2} are joinable to a common vertex v , then also v_{3} is joinable to v . Hence, in this case, \{v_{1},v_{2},v_{3}\} is a round-up triple in the residue of v . Proof. Consider an apartment \Sigma of \Delta containing v and v_{3} , and let w be the unique vertex of \Sigma opposite v_{3} . By the definition of round-up triple, we may assume that w is opposite v_{1} . Let \Sigma^{\prime} be an apartment containing w and the simplex \{v,v_{1}\} . Let \varphi:\Sigma^{\prime}\to\Sigma be an isomorphism of complexes fixing both w and v (as required to exist by the definition of a building). Then \varphi(v_{1})=v_{3} as opposition is preserved. Hence v is joinable to v_{3} , as \varphi preserves the simplicial structure. ∎ Lemma 2.34 . Let \{v_{1},v_{2},v_{3}\} be a round-up triple of vertices of some common type in a spherical building \Delta . Suppose there are simplices \{v,w_{1}\} and \{v,w_{2}\} such that w_{i} is joinable to v_{i} , i=1,2 . Suppose also that v_{1} and v_{2} are not joinable to a common vertex. Then there exists a vertex w_{3} joinable to both v and v_{3} , and the type of w_{3} can be chosen equal to the type of either w_{1} or w_{2} . Proof. Consider an apartment \Sigma of \Delta containing \{v,w_{1}\} and v_{3} , and let w be the unique vertex of \Sigma opposite v_{3} . Assume, for a contradiction, that v_{1} is opposite w . Let \Sigma^{*} be an apartment containing \{v_{1},w_{1}\} and w , and let \varphi^{*}:\Sigma^{*}\to\Sigma be an isomorphism fixing w and w_{1} . Then \varphi^{*}(v_{1})=v_{3} by uniqueness of opposites in apartments. Hence w_{1} is joinable to v_{3} , and 2.33 implies that it is also joinable to v_{2} , contradicting our assumptions. We conclude that w is opposite v_{2} . Let \Sigma^{\prime} be an apartment containing \{v_{2},w_{2}\} and w , and let \Sigma^{\prime\prime} be an apartment containing \{v,w_{2}\} and w . Let \varphi^{\prime\prime}:\Sigma^{\prime\prime}\to\Sigma be an isomorphism fixing w and v , and denote \varphi^{\prime\prime}(w_{2}) briefly as w_{3} . Note that, by the definition of buildings, we may assume that \varphi^{\prime\prime} is type-preserving, implying that w_{3} has the same type as w_{2} . Also, w_{3} is joinable to v , as w_{2} is. Let \varphi^{\prime}:\Sigma^{\prime}\to\Sigma^{\prime\prime} be an isomorphism fixing w_{2} and w . Then \varphi=\varphi^{\prime\prime}\circ\varphi^{\prime}:\Sigma^{\prime}\to\Sigma is an isomorphism fixing w , hence mapping v_{2} to v_{3} . But \varphi(w_{2})=w_{3} . Hence w_{3} is joinable to both v_{3} and v . Interchanging the roles of v_{1} and v_{2} , we can also choose the type of w_{3} to be the same as that of w_{1} . This completes the proof of the lemma. ∎ Setting v_{2}=w_{2} in the previous lemma, noting that the only vertex of the same type as v_{2} joinable to v_{3} is v_{3} itself, and then also noting the symmetry of the situation, we obtain the following consequence. Corollary 2.35 . Let \{v_{1},v_{2},v_{3}\} be a round-up triple of vertices of some common type in a spherical building \Delta . Suppose there exists a simplex \{w_{1},w_{2}\} such that w_{i} is joinable to v_{i} , i=1,2 . Suppose also that v_{1} and v_{2} are not joinable to a common vertex. Then v_{3} is joinable to both w_{1} and w_{2} . Finally, we note that we called certain objects “far” from each other. It was always the case that these objects referred to vertices of the corresponding building, and either they were opposite (for example, a point and a symp in \mathsf{E_{6,1}}(\mathbb{K}) ), or one vertex was joinable to a vertex opposite the other. We will extend now this notion of far to all such situations. Hence two objects in a Lie incidence geometry will be called far (from each other) if they correspond to vertices in the building and one vertex is joinable to a vertex opposite the other. 3. Points and lines in the exceptional minuscule geometries 3.1. Points of Lie incidence geometries of type \mathsf{E_{6,1}} 3.1.1. Blocking sets Proposition 3.1 . If every line of a parapolar space \Gamma of type \mathsf{E_{6,1}} contains exactly s+1 points, then there exists a symp opposite each point of an arbitrary set S of s+1 (distinct) points, except if these points are contained in a single line. Proof. Let S be a set of s+1 distinct points p_{0},\dots,p_{s} in \Gamma that are not all on a common line. Assume first that all points of S are mutually collinear. Then p_{0},\dots,p_{s} are either contained in a common 4 -space or in a common 5 -space. Since both are residues in the corresponding building, 2.27 and 2.28 imply that S is a line, proving the assertion. So, we may assume that at least two points of S have distance 2 . Let p_{0} and p_{1} be two non-collinear points of S . Consider the symp \xi_{0}=\xi(p_{0},p_{1}) . For each point of S close to \xi_{0} , we choose an arbitrary point in \xi_{0} collinear with it. Let S^{\prime} be the set of points thus obtained, complemented with the points of S contained in \xi_{0} . By the main result of [ 10 ] , in particular Lemma 3.5 therein, we find a point b in \xi_{0} non-collinear to each member of S^{\prime} , and hence non-collinear to each member of S . Let \Xi be the set of symps \xi(b,p_{i}) , for i\in\{0,1,\ldots,s\} . There are at most s such different symps, as \xi(b,p_{0})=\xi(b,p_{1}) . Applying 2.30 in \mathsf{Res}(b) , we find a symp \xi through b intersecting each member of \Xi in exactly b . Then, by 2.5 , the symp \xi is far from all points of S . The proof is complete. ∎ 3.1.2. Geometric lines Corollary 5.6 of [ 21 ] states that geometric lines of Lie incidence geometries of type \mathsf{E_{6,1}} are the same things as lines. 3.2. Lines of Lie incidence geometries of type \mathsf{E_{6,1}} We now look at the lines of Lie incidence geometries of type \mathsf{E_{6,1}} . 3.2.1. Blocking sets Proposition 3.2 . Let \Delta be Lie incidence geometry of type \mathsf{E_{6,1}} such that every line has s+1 points. Let T=\{L_{0},\dots,L_{s}\} be a set of s+1 different lines in \Delta such that they do not form a line pencil in a plane. Then there exists a 4 -space opposite all lines L_{0},\dots,L_{s} in \Delta . Proof. We may put L_{0},\dots,L_{s} in respective symps not containing a common 4 -space. Then (the dual of) 3.1 yields a point b opposite all of those symps. Then 2.9 yields 3 -spaces U_{i}=b^{\perp}\cap L_{i}^{\perp} . In \mathsf{Res}(b)\cong\mathsf{D_{5,5}}(\mathbb{K}) , the \langle b,U_{i}\rangle are 3 -spaces. Viewed as \mathsf{D_{5,1}}(\mathbb{K}) , they become lines. If they do not form a planar line pencil, then 2.28 yields a line opposite all of them, which translates into a 4 -space W through b locally opposite each \langle p,U_{i}\rangle . 2.26 implies that W is opposite each member of T . So, we may suppose that the \langle b,U_{i}\rangle form a planar line pencil in \mathsf{Res}(b) , viewed as a polar space. Since points and planes of that polar space correspond to symps and lines of \mathsf{Res}(b) , viewed as \mathsf{D_{5,5}}(\mathbb{K}) , we may assume that the 4 -spaces \langle b,U_{i}\rangle form the set of 4 -spaces through a given plane \beta of \Delta contained in a given symp \xi . All U_{i} intersect \beta in lines N_{i} not containing b . Set W_{i}=\langle U_{i},L_{i}\rangle . We can find a 5 -space opposite all W_{i} unless all W_{i} intersect in a common plane \alpha . Case 1: Suppose first that all W_{i} do not intersect in a common plane and denote by B a 5 -space opposite all of them. The last assertion of 2.4 yields lines L_{i}^{\prime} in B such that each point of L_{i} is collinear to a unique point of L_{i}^{\prime} , but no point of B-\bigcup_{i=0}^{s}L_{i}^{\prime} is collinear to any point of any L_{i} . If the L_{i}^{\prime} do not form a planar line pencil in B , then we can find a 3 -space M in B opposite each L_{i}^{\prime} . By 2.3 , that 3 -space is contained in a unique maximal 4 -space that we will denote by C , and, by 2.7 , C is opposite all L_{i} . So suppose the L_{i}^{\prime} form a planar line pencil in B . We will denote the corresponding plane as \gamma and the intersection L_{0}^{\prime}\cap L_{1}^{\prime}\cap\cdots L_{s}^{\prime} as d . 2.8 yields a point q_{0}\in\langle b,U_{0}\rangle\setminus U_{0} collinear to a 3 -space of B disjoint from L_{0} , and hence intersecting \gamma in a unique point q_{0}^{\prime} . Without loss of generality, we may assume that q_{0}^{\prime}\in L_{1}^{\prime} . Hence there is a unique point q_{1}\in L_{1} collinear to q_{0}^{\prime} . Since q_{0}\notin U_{1} , the symp \xi(q_{0},q_{1}) is well-defined and contains both q_{1}^{\prime} and N_{1} . But then N_{1} , which is contained \langle L_{1},U_{1}\rangle , contains a point collinear to q_{1}^{\prime} . Since b is collinear to all points of N_{1} and to no point of L_{1} , these lines are disjoint. This contradicts the last assertion of 2.4 . We conclude that Case 1 cannot occur. Case 2: Now suppose that all W_{i} intersect in a common plane \alpha . Let i\in\{0,1,\ldots,s\} . Since U_{i}\subseteq\xi , the 5 -space W_{i} intersects \xi in a 4^{\prime} -space V_{i} . Since, by 2.1 , W_{i} is determined by \xi and any point of W_{i}\setminus V_{i} , the plane \alpha is contained in \xi . Hence the V_{i} form the set of 4^{\prime} -spaces through \alpha . Since U_{i}=b^{\perp}\cap V_{i} , each line N_{i} coincides with L:=\alpha\cap b^{\perp}\subseteq U_{i} . By the choice of b , the line L_{i} is not contained in \xi , and a fortiori neither in \alpha . Let A_{0} be a 5 -space containing L_{0} that does not contain the plane \alpha . Then we can find a 5 -space B opposite all 5 -spaces A_{0} , W_{1},\dots,W_{s} . Again, since \langle b,U_{i}\rangle is a 4 -space and B a 5 -space, we can find a point q_{1} in \langle b,U_{1}\rangle that is collinear to a 3 -space of B . As in Case 1, q_{1}\notin U_{1} . Since every 3 -space of B intersects \gamma , q_{1} is collinear to a point q_{1}^{\prime} in \gamma that is collinear to some point q_{k} of L_{k} , for some k\in\{0,2,3\ldots,s\} . As in Case 1, k\neq 0 leads again to a contradiction. Hence, considering \xi(q_{0},q_{1}) , we see that q^{\prime}_{1} is collinear to some line q_{0}q , with q\in L . We now make some observations and define planes \epsilon and \epsilon^{\prime} . • Since W_{0} and A_{0} have the line L_{0} in common, A_{0}\cap W_{0} is a plane and hence they are adjacent. Since \delta is in A_{0} and A_{0} is opposite B , every point of \delta is far from B . • Since W_{0} and B cannot be opposite, but W_{0} is adjacent to the opposite 0 to B , 2.16 yields a 5 -space W intersecting both W_{0} and B in planes denoted by \epsilon and \epsilon^{\prime} , respectively. • Every point of \epsilon is close to B , and every point of \alpha is far from B . Therefore, the planes \alpha and \epsilon cannot intersect. As L_{0}\subseteq A_{0} , no point of L_{0} is in \epsilon . But since each point of L_{0} has to be collinear to a point of \epsilon^{\prime} , and every point of L_{0} is collinear to a unique point of B on L_{0}^{\prime} , it follows that L_{0}^{\prime} is contained in \epsilon^{\prime} . In particular d\in\epsilon^{\prime} . It follows that d^{\perp}\cap W_{0} is a 3 -space containing \epsilon , hence intersecting \alpha in a unique point e . Since d\in L_{i}^{\prime} , i=1,2,\ldots,s , it is collinear to a unique point of W_{i} , and that point is on L_{i} . hence e\in L_{1}\cap L_{2}\cap\cdots\cap L_{s} . We conclude \{e\}=\xi\cap L_{i} . Switching the roles of L_{0} and L_{1} , we obtain L_{0}\cap\xi=L_{2}\cap\xi=\{e\} . Now T belongs to \mathsf{Res}_{\Delta}(e) and the assertion follows from 2.27 . ∎ 3.2.2. Geometric lines Now we classify geometric lines in Lie incidence geometries of type \mathsf{E_{6,3}} . In order to do so, we classify round-up triples of lines in \mathsf{E_{6,1}}(\mathbb{K}) , for an arbitrary field \mathbb{K} . Lemma 3.3 . Let \{L_{1},L_{2},L_{3}\} be a round-up triple of lines in the exceptional Lie incidence geometry \mathsf{E_{6,1}}(\mathbb{K}) such that L_{1} and L_{2} have at least one point in common. Then exactly one of the following holds. (i) L_{1}=L_{2}=L_{3} ; (ii) L_{1},L_{2},L_{3} are three lines in a common planar line pencil. Proof. If L_{1}=L_{2}\neq L_{3} , then a 4 -space opposite L_{1} but not opposite L_{3} (which exists by 2.31 ) violates the defining property of a round-up triple. Hence, if L_{1}=L_{2} , then (i) holds. Now assume L_{1}\cap L_{2}=\{x\} . By 2.33 , x\in L_{3} . Then 2.27 and 2.28 imply that L_{1},L_{2},L_{3} are contained in a plane, and (ii) follows. ∎ Lemma 3.4 . Let \{L_{1},L_{2},L_{3}\} be a round-up triple of pairwise disjoint lines in the exceptional Lie incidence geometry \mathsf{E_{6,1}}(\mathbb{K}) . Let y_{1} be an arbitrary point of L_{1} . If y_{1} is symplectic to some point of L_{2} , then it is collinear to some point of L_{3} . Proof. Suppose y_{i}\in L_{i} , i=1,2 , with y_{1}\perp\hskip-4.26773pt\perp y_{2} . We claim that there exists a point u\in y_{1}^{\perp}\cap y_{2}^{\perp} not collinear to any point of L_{3} . Indeed, suppose each point of y_{1}^{\perp}\cap y_{2}^{\perp} is collinear to some point of L_{3} . If that point of L_{3} is unique, then, since \{y_{1},y_{2}\}^{\perp\perp}=\{y_{1},y_{2}\} in each hyperbolic quadric, either y_{1} or y_{2} belongs to L_{3} , a contradiction. If L_{3} were contained in \xi(y_{1},y_{2}) , then y_{1} would be collinear to a point of L_{3} , and the assertion would follow. If L_{3} intersected \xi(y_{1},y_{2}) in a point z , then for each point z_{3}\in L_{3} we would have z_{3}^{\perp}\cap\xi(y_{1},y_{2})\subseteq z^{\perp}\cap\xi(y_{1},y_{2}) , and this only contains y_{1}^{\perp}\cap y_{2}^{\perp} if z\in\{y_{1},y_{2}\} , a contradiction again. Hence L_{3} is disjoint from \xi(y_{1},y_{2}) . Now 2.6 implies that L_{3} is collinear to a plane \pi of \xi(y_{1},y_{2}) , and all points of \xi(y_{1},y_{2}) collinear to some point of L_{3} are collinear to all points of \pi , a contradiction since y_{1}^{\perp}\cap y_{2}^{\perp} is only collinear to \{y_{1},y_{2}\} . The claim is proved. Now, by 2.26 , a 4 -space U through u locally opposite the projection of L_{3} onto u is opposite L_{3} , but not opposite either L_{1} or L_{2} , as u is collinear to both y_{1}\in L_{1} and y_{2}\in L_{2} . This final contradiction proves the lemma. ∎ Lemma 3.5 . Let \{L_{1},L_{2},L_{3}\} be a round-up triple of lines in the exceptional Lie incidence geometry \mathsf{E_{6,1}}(\mathbb{K}) . Then exactly one of the following holds. (i) L_{1}=L_{2}=L_{3} ; (ii) L_{1},L_{2},L_{3} are three lines in a common planar line pencil. Proof. By 3.3 , the statement is true if, for some i,j\in\{1,2,3\} , i\neq j , the lines L_{i} and L_{j} have a point in common. So we may assume that L_{1},L_{2},L_{3} are pairwise disjoint. By 3.4 , we may assume that some point x_{1}\in L_{1} is collinear to some point x_{2}\in L_{2} . Set M:=x_{1}x_{2} . 2.34 implies that M also intersects L_{3} . So M\cap L_{3} is also a point x_{3} . Now select y_{1}\in L_{1}\setminus\{x_{1}\} . By 3.4 , we may again assume that y_{1} is collinear to some point y_{2} of L_{2} . Then, as in the previous paragraph, the line L_{3} has a point y_{3} in common with y_{1}y_{2} . Clearly y_{3}\neq x_{3} , as otherwise L_{1},L_{2} are contained in the plane generated by the lines M and y_{1}y_{2} , contradicting our assumption that L_{1}\cap L_{2}=\varnothing . Similarly, y_{2}\neq x_{2} . If all of x_{1},x_{2},x_{3} are collinear to all of y_{1},y_{2},y_{3} , then all of L_{1},L_{2},L_{3} are contained in a singular 3 -space, which is a residue of a simplex of type \{2,5,6\} in the corresponding building. Then 2.27 and 2.28 lead to the contradiction that L_{1},L_{2},L_{3} are coplanar and hence not disjoint. So we may assume that x_{1} is not collinear to y_{2} . But then all of L_{1},L_{2},L_{3} are contained in the symp \xi(x_{1},y_{2}) , which is again a residue, and so 2.27 and 2.28 again lead to a contradiction. The lemma is proved. ∎ Proposition 3.6 . Every geometric line of \mathsf{E_{6,3}}(\mathbb{K}) is an ordinary line. Proof. This follows directly from 2.32 and 3.5 . ∎ 3.3. Points of Lie incidence geometries of type \mathsf{E_{7,7}} 3.3.1. Blocking sets Proposition 3.7 . If every line of a parapolar space \Gamma of type \mathsf{E_{7,7}} contains exactly s+1 points, then there exists a point at distance 3 from each point of an arbitrary set S of s+1 (distinct) points, except if these points are contained in a single line. Proof. Suppose S is not a line. Set S=\{p_{0},\ldots,p_{s}\} . We distinguish several cases. (i) Suppose all points in S are pairwise collinear. Then S is contained in a (maximal) singular subspace. As this corresponds to a residue in the corresponding building, 2.27 and 2.28 lead to a contradiction. (ii) Suppose some pair of points from S is symplectic. Suppose p_{0} and p_{1} are symplectic and let \xi be the symp containing them. Let the set T of points of \Delta consist of the points of S in \xi , the set of points of \xi collinear to a point of T far from \xi , and, for each point p_{i} of S close to \xi , an arbitrary point of \xi collinear to p_{i} . Then T is a set of at most s+1 points not forming a line (as p_{0} and p_{1} belong to T and are not collinear). Hence, by 2.28 (or more precisely, [ 10 , Lemma 3.5] ), we find a point b\in\xi not collinear to any member of T . Hence, for each point p_{i}\in S not far from \xi , there is a unique symp \xi_{i} containing b and p_{i} . Since \xi=\xi_{0}=\xi_{1} , there are at most s such symps. 2.30 yields a line L through b locally opposite each such symp, which means that each point of L\setminus\{b\} is far from each such symp. This implies that each point of L\setminus\{b\} is opposite each point of S that is not far from \xi . But since b is opposite every point of S that is far from \xi by construction, the line L contains points opposite each member of S . Hence, each point p_{i} , i=0,1,\ldots,s , has a unique projection p_{i}^{\prime} onto L . Since b=p_{0}^{\prime}=p_{1}^{\prime} , there is at least one point of L opposite each point of S . (iii) Some pair of points from S is opposite and there are no symplectic pairs in S . It follows that collinearity is an equivalence relation in S . Let C\subseteq S be a corresponding equivalence class of minimal size (then certainly |C|<1+s/2 and |S\setminus C|\geq 2 ). Let \xi be a symp either containing C (if the latter is contained in a singular 5 -space or a singular 5^{\prime} -space; we assume p_{0}\in C ), or containing at least one point, say again p_{0} , of C and intersecting the 6 -space spanned by C in a 5^{\prime} -space (if C generates a 6 -space; 2.10 allows for this). Either way, let W be the singular subspace of \xi obtained by intersecting \xi with the singular subspace generated by C . Let S^{\prime}\neq\varnothing be the set of points of \xi collinear to a point of S\setminus C . Select q\in S^{\prime} arbitrarily and suppose q\perp p_{1}\in S\setminus C . As |S\setminus C|\geq 2 , we can select a line M in \xi through q disjoint from W and not containing S^{\prime} . With that, M contains at least one point b\perp q such that b\notin S^{\prime} and b is not collinear to a member of C . It follows that b is not collinear to any point of S . So we can consider the set \Xi of symps containing b and some point of S . Since p_{0} and p_{1} are opposite, the symps \xi=\xi(b,p_{0}) and \xi(b,p_{1}) only intersect in the line bq . Hence, in the residue of b , the set \Xi does not correspond to the set of symps containing a given maximal singular subspace. Hence, the dual of 3.1 yields a line L through b locally opposite each member of \Xi . As in (i) above, the line L is far from each member of S , but p_{0} and p_{1} project to the same point of L , implying that there is at least one point on L opposite every member of S . ∎ 3.3.2. Geometric lines Corollary 5.6 of [ 21 ] states that a geometric line of a Lie incidence geometry of type \mathsf{E_{7,7}} is the same thing as a line. 3.4. Lines of Lie incidence geometries of type \mathsf{E_{7,7}} We now look at the lines of Lie incidence geometries of type \mathsf{E_{7,7}} . 3.4.1. Blocking sets Proposition 3.8 . If every line of a parapolar space \Gamma of type \mathsf{E_{7,7}} contains exactly s+1 points, then there exists a line opposite each member of an arbitrary set T of s+1 (distinct) lines, except if these lines form a planar line pencil. Proof. Since symps are vertices of type 1 in the corresponding building, 4.11 and 4.22 , proved independently, allow us to consider symps \widehat{\xi_{0}},\widehat{\xi_{1}},\dots,\widehat{\xi_{s}} , such that L_{i}\in\widehat{\xi_{i}} and such that there exists a symp \xi opposite every \widehat{\xi_{i}} , for i\in\{0,1,\dots,s\} . Then, in view of 2.12 , every point of any \widehat{\xi_{i}} is collinear to a unique point of \xi . The points of \xi , which are in bijection with the points of L_{i}\in\widehat{\xi_{i}} , form a line again that we will denote by L_{i}^{\prime} . Suppose the lines L_{i}^{\prime} do not form a planar line pencil. Then, by Theorem A of [ 10 ] , or more in particular Lemma 3.15 therein (see also 2.28 ), we can find a line M in \xi that is locally opposite all L_{i}^{\prime} at \xi (viewed as a residue). With 2.26 , it follows that M is opposite all L_{i} . Now suppose the L_{i}^{\prime} do form a planar line pencil in \xi and denote the plane by \pi . Let p be the point in which all L_{i}^{\prime} intersect and let q be a point in \xi\setminus\pi , not collinear to p . Then \mathsf{proj}_{\pi}(q) (in \xi ) is a line K , which intersects each L_{i}^{\prime} in some point p_{i}^{\prime} . Set p_{i}=\mathsf{proj}_{L_{i}}p_{i}^{\prime} . Since p_{i} and q are both collinear to p_{i}^{\prime} , they are symplectic and we denote the symp spanned by p_{i} and q by \xi_{i} . If the \xi_{i} do not all intersect in a common 5 -space, then we can find a line M through q locally opposite all \xi_{i} and, by 2.26 , M is opposite all L_{i} . So we may assume from now on that all \xi_{i} intersect in a common 5 -space Q . Note that they are pairwise distinct, as every \xi_{i} intersects \xi in exactly the line qp_{i}^{\prime} . Let q_{0} be a point on qp_{0}^{\prime} distinct from q and p_{0}^{\prime} . The point q_{0} is in the plane \alpha:=\langle q,K\rangle and with that \mathsf{proj}_{\pi}(q_{0})=\mathsf{proj}_{\pi}(q)=K . Hence, q_{0} is symplectic to each p_{i} as well. We will denote the symps spanned by p_{i} and q_{0} by \xi_{i}^{0} . Since q_{0} is on qp_{0}^{\prime} , \xi_{0}^{0} coincides with \xi_{0} . Similarly to before, we may assume that all \xi_{i}^{0} intersect in a common 5 -space Q^{0} . For each j\in\{1,2,\dots,s\} , we have that \xi_{j}\neq\xi_{j}^{0} and \xi_{j}^{0} intersects \xi in exactly the line q_{0}p_{j} . The 5 -spaces Q and Q^{0} are distinct, but both contained in \xi_{0} . Since symps in \Delta are polar spaces of type \mathsf{D_{6,1}} , two 5 -spaces in \xi_{0} intersect in even codimension and hence either in nothing, a line, a 3 -space, or they coincide. Note that p_{j} , for j\in\{1,2,\dots,s\} , cannot be contained in \xi_{0} , since otherwise \xi_{0}=\xi(q,p_{j})=\xi_{j} , a contradiction. We have that \mathsf{proj}_{\xi_{0}}(p_{j}) (in \Delta ) is a 5^{\prime} -space A_{j} . We know that \mathsf{proj}_{Q}(p_{j}) is a 4 -space U_{j}\subseteq A_{j} and \mathsf{proj}_{Q^{0}}(p_{j}) is a 4 -space U_{j}^{0}\subseteq A_{j} . Since U_{j}\cap U_{j}^{0}\subseteq Q\cap Q^{0} , it follows that Q\cap Q^{0}=U_{j}\cap U_{j}^{0}=:V is a 3 -space independent of j\in\{1,2,\ldots,s\} . Let q_{1} be a point of qp_{1}^{\prime}\setminus\{q,p_{1}^{\prime}\} . Without loss of generality, assume p_{2}^{\prime}\in q_{0}q_{1} . Then the symps \xi(q_{1},p_{1}) and \xi(q_{1},p_{2}) contain both V , and it follows again that all symps \xi(q_{1},p_{i}) contain V . Since they also contain q_{1} , we find q_{1}\perp V . Since q_{1} was essentially arbitrary, we conclude that V is the intersection of all symps defined by p_{i} and some point of \alpha\setminus K . Moreover, \alpha\perp V and W:=\langle\alpha,V\rangle is a (maximal) 6 -space. Since both p_{0} and V are in the intersection of \xi(q,p_{0}) and \xi(q_{1},p_{0}) , we also deduce p_{0}\perp V . We observe that A_{j}=\mathsf{proj}_{\xi_{0}}(p_{j})=\langle\mathsf{proj}_{Q}(p_{j}),\mathsf{proj}_{Q^{0}}(p_{j})\rangle=\langle U_{j},U_{j}^{0}\rangle and that all A_{j} contain V . We define U_{0} and U_{0}^{0} as the 4 -spaces \mathsf{proj}_{Q}(p_{0}) and \mathsf{proj}_{Q^{0}}(p_{0}) , respectively. By the foregoing, V=U_{0}^{0}\cap U_{0} . The point q is contained in Q and the point q_{0} is contained in Q^{0} . In a 5 -space, there can only be s+1 different 4 -spaces which contain a given 3 -space. Since none of the U_{i} ( U_{i}^{0} respectively) can contain q ( q_{0} respectively), at least two of the U_{i} ( U_{i}^{0} respectively) have to coincide. If U_{m}=U_{n} for some m,n\in\{0,1,\dots,s\} , then it follows that U_{m}^{0}=U_{n}^{0} , since otherwise A_{m} and A_{n} would intersect in a 4 -space. We conclude that two A_{i} , for i\in\{0,1,\dots,s\} , have to coincide. Let m,n\in\{0,1,\dots,s\} be such that A_{m}=A_{n} . Since, by 2.10 , 6 -spaces cannot intersect in 5^{\prime} -spaces, \langle p_{m},A_{m}\rangle and \langle p_{n},A_{m}\rangle have to be equal. Hence p_{m} and p_{n} are collinear. We now claim that all p_{i} are contained in a common symp. Indeed, by 2.10 , there is a unique symp \zeta containing the 5^{\prime} -space \langle V,K\rangle . Now \xi_{0} and \zeta have the 4 -space \langle p_{0}^{\prime},V\rangle in common, and hence they intersect in the 5 -space \langle p_{0},p_{0}^{\prime},V\rangle , since this is the unique 5 -space of \xi_{0} containing \langle p_{0}^{\prime},V\rangle . We get p_{0}\in\zeta . Similarly, p_{i}\in\zeta , for all i\in\{1,2,\ldots,s\} , and the claim is proved. We now vary K over all lines of \pi not containing p . We obtain s^{2} distinct pairs of collinear points, where each pair of points is contained in two different lines L_{i},L_{j} , i\neq j , and no point of such pair is collinear to p . Since we only have \binom{s+1}{2}=\frac{1}{2}s(s+1)2 , this is again a contradiction to |b^{\not\equiv}\cap L|\in\{0,2\} . Now assume that \mathbb{K} is perfect of characteristic 2 . We may embed every equator geometry in an extended equator geometry \widehat{E} , which is then isomorphic to a symplectic polar space of rank 4 (see Section 2.5 ). By [ 17 , Corollary 5.38] , there exists a point b of \Gamma such that H:=b^{\not\equiv}\cap\widehat{E} is a (hyperbolic) polar subspace of type \mathsf{D_{4,1}} . Note that we still may assume that every hyperbolic line is a geometric line. Then H is also a geometric hyperplane of the ambient projective space \mathsf{PG}(7,\mathbb{K}) of \widehat{E} , and hence coincides with p^{\perp} for some point p\in\widehat{E} (where the perp \perp is now with respect to the symplectic polar space). This is clearly a contradiction. The theorem is proved. ∎ 4.2. Lines of exceptional hexagonic Lie incidence geometries 4.2.1. Two lemmas in the residues We begin with two results in the point residuals of hexagonic geometries. The first one summarises earlier findings. Lemma 4.5 . Let \Delta be either a Lie incidence geometry of type \mathsf{A_{5,3}} , \mathsf{D_{6,6}} or \mathsf{E_{7,7}} , or a dual polar space of rank 3 . Suppose each line has exactly s+1 points. Suppose also that \Delta is not isomorphic to \mathsf{B_{3,3}}(\sqrt{s},s) . Then a set of at most s+1 points of \Delta admits no common opposite point if, and only if, the points form a geometric line. In particular, if there are at most s points, or if there exists a point opposite at least one point of the set, and not opposite at least two points of the set, then the set admits an opposite point. Proof. This follows from Main Results A and B of [ 10 ] for types \mathsf{A_{5,3}} , \mathsf{D_{6,6}} , and for dual polar spaces, and from 3.7 and [ 21 , Corollary 5.6] for type \mathsf{E_{7,7}} . ∎ Lemma 4.6 . Let \Delta be either a Lie incidence geometry of type \mathsf{A_{5,3}} , \mathsf{D_{6,6}} or \mathsf{E_{7,7}} , or a dual polar space of rank 3 . Suppose each line contains precisely s+1 points and let p be a point. Let Q:=\{q_{1},\ldots,q_{\ell}\} be a set of points not containing p , \ell\leq s . Then there exists a point p^{\prime} opposite each member of Q , and not opposite p . Proof. We first construct a symp \xi through p far from each member of Q . If q\in Q is opposite p , each symp through p qualifies. If q\in Q is symplectic to p , let K_{q} be an arbitrary line through p in the symp containing p and q ; if q\perp p then let K_{q} be the line containing p and q . Since \ell\leq s , we infer from 2.30 and 2.29 that there exists a symp \xi through p locally opposite all K_{q} , q\in Q . Then \xi is far from each member of Q . Now the same references yield a point p^{\prime} in \xi (locally) opposite in \xi each intersection \xi\cap q^{\perp} , for q\in Q . The point p^{\prime} is opposite each member of Q and not opposite p . ∎ 4.2.2. Description of mutual positions We will describe the mutual position \delta(L,M) of two lines L and M of an exceptional hexagonic Lie incidence geometry with four parameters (a,b,c,d) , chosen in the set \{=,\perp,\perp\hskip-4.26773pt\perp,\Join,\equiv\} , where a,b,c,d are defined as follows. If there is a unique point on L closest to M , we call it x ; otherwise, x is an arbitrary point on L . Similarly, if there is a point on M nearest to L , call it y ; if not, but there is a point on M nearest to x , call this y . If not, then y is an arbitrary point. Let x^{\prime} be any point on L distinct from x . If there is a point on M nearest to x^{\prime} , and it is different from y , call it y^{\prime} ; otherwise, y^{\prime} is any point on the second line distinct from y . Then a is the relation between x and y , while b is the relation between x and y^{\prime} . Also, c is the relation between x^{\prime} and y , whereas d is the relation between x^{\prime} and y^{\prime} . It will turn out that such a 4-tuple unambiguously determines the mutual position. In a shorthand alternative notation, we write 0 for = , 1 for \perp , \frac{3}{2} for \perp\hskip-4.26773pt\perp , 2 for \Join and 3 for \equiv . The inverse of \delta(L,M) is \delta(M,L) . The dual of \delta(L,M) is \delta(M,K) , for the line K opposite L in any apartment containing L and M . On the level of point distances, 0 is dual to 3 , 1 to 2 , and \frac{3}{2} is self-dual. We sometimes call \delta(L,M) the distance between L and M . There are basically four classes of positions \delta(L,M) , if one takes into account the homogeneity with respect to the points of the lines L and M . The classes are the following (where we use the notation of the previous paragraphs). Class I — Completely homogeneous In this class, every point of either line has the same distance to each point of the other line. All positions here are equal to their inverse. The cases are: (1111) (\perp,\perp,\perp,\perp) : the two lines L,M span a singular 3 -space. (\frac{3}{2}\frac{3}{2}\frac{3}{2}\frac{3}{2}) (\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp) : each point of each line is symplectic to each point of the other line. The dual of (1111) is the following: (2222) (\Join,\Join,\Join,\Join) : each point of each line is special to each point of the other line. Class II — Projection homogeneous In this class, each point of each line has a unique projection onto the other line; the distances between corresponding points are constant, and the other distances as well. All distances are their own inverse. The cases are: (0110) (=,\perp,\perp,=) : the lines are equal, that is, L=M . (1\frac{3}{2}\frac{3}{2}1) (\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\perp) : the lines are opposite in a symp. (\frac{3}{2}22\frac{3}{2}) (\perp\hskip-4.26773pt\perp,\Join,\Join,\perp\hskip-4.26773pt\perp) : the lines are each other’s projection from a symp to an opposite symp. (2332) (\Join,\equiv,\equiv,\Join) : the lines are opposite, that is, L\equiv M . Class III — Symmetric non-homogeneous In this class, both lines contain a unique projection point with respect to the other line. Moreover, all the positions are their own inverse again. The cases are: (0111) (=,\perp,\perp,\perp) : the lines are coplanar. (011\frac{3}{2}) (=,\perp,\perp,\perp\hskip-4.26773pt\perp) : the lines meet and determine a unique symp. (0112) (=,\perp,\perp,\Join) : the lines meet and are locally opposite. (111\frac{3}{2}) (\perp,\perp,\perp,\perp\hskip-4.26773pt\perp) : the lines are special in (the line-Grassmannian of) a symp. (1\frac{3}{2}\frac{3}{2}\frac{3}{2}) (\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp) : the projection point x of M onto L is contained in a symp \xi with M , while L is locally close to \xi at x ; the same holds with the roles of L and M interchanged. (1\frac{3}{2}\frac{3}{2}2) (\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\Join) : the projection point x on L is contained in a symp \xi with M , while L is locally opposite the line connecting the two projection points x and y at x , and hence locally far from \xi at x ; the same holds with the roles of L and M interchanged. (\frac{3}{2}\frac{3}{2}\frac{3}{2}2) (\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\Join) : the projection points x and y are symplectic, the lines L and M are locally close to the symp \xi(x,y) , and the projections of the lines L,M onto \xi(x,y) (which are maximal singular subspaces of \xi(x,y) ) intersect in a unique point. (\frac{3}{2}222) (\perp\hskip-4.26773pt\perp,\Join,\Join,\Join) : the points x and y are symplectic, the lines L and M are locally far from \xi(x,y) , and the projections of the lines L,M onto \xi(x,y) , which are lines themselves, are \xi(x,y) -special. (1223) (\perp,\Join,\Join,\equiv) : the lines L,M are each other’s projection from a point x^{\prime} or y^{\prime} to an opposite point y^{\prime} or x^{\prime} , respectively. (\frac{3}{2}223) (\perp\hskip-4.26773pt\perp,\Join,\Join,\equiv) : the lines are locally far from the symp \xi determined by the projection points x and y (which are symplectic), and the projections of the lines onto \xi are \xi -opposite lines. (2223) (\Join,\Join,\Join,\equiv) : There is a pair of opposite points x^{\prime}\in L and y^{\prime}\in M , and the projection of L (or M ) onto y^{\prime} (or x^{\prime} ) is locally symplectic to M (or L ) at y^{\prime} (or x^{\prime} , respectively); equivalently, L and M lie in opposite symps and the projection of L (or M ) onto the symp through M (or L ) is special to M (or L , respectively) in (the line Grassmannian of) that symp. Class IV — Asymmetric positions Up to now, in all 4 -tuples, the second and third entry coincided. This is going to change now. The — final — class that we consider in this paragraph contains the asymmetric mutual positions, that is, those that do not coincide with their inverse. There are four cases with two projection points. These cases are: (1\frac{3}{2}12) (\perp,\perp\hskip-4.26773pt\perp,\perp,\Join) : the line L is coplanar with y ; the line M and the plane \langle L,y\rangle are locally far at y . (11\frac{3}{2}2) (\perp,\perp,\perp\hskip-4.26773pt\perp,\Join) : the line M is coplanar with x ; the line L and the plane \langle M,x\rangle are locally far at x . (1\frac{3}{2}22) (\perp,\perp\hskip-4.26773pt\perp,\Join,\Join) : the line M and the point x are in a unique symp \xi ; the line L is locally far from \xi at x and locally opposite the line xy at x (and x and y are collinear). (12\frac{3}{2}2) (\perp,\Join,\perp\hskip-4.26773pt\perp,\Join) : the line L and the point y are in a unique symp \xi ; the line M is locally far from \xi at y and locally opposite the line xy at y (and x and y are collinear). Finally, there are four cases where only one line has a projection point. Hence, there will be only two distinct distances around, and the corresponding 4 -tuples have the shape (a,a,b,b) or (a,b,a,b) (where these are each other’s inverse). In a Lie incidence geometry of type \mathsf{A_{5,3}} , \mathsf{D_{6,6}} or \mathsf{E_{7,7}} , or a dual polar space of rank 3 , we call a point and a line almost far if every point of the line is symplectic to the point (however, this does not exist in dual polar spaces). (1\frac{3}{2}1\frac{3}{2}) (\perp,\perp\hskip-4.26773pt\perp,\perp,\perp\hskip-4.26773pt\perp) : the line L is coplanar with y (which is the projection point of L on M ); the line M and the plane \langle L,y\rangle are locally almost far at y . (11\frac{3}{2}\frac{3}{2}) (\perp,\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp) : the line M is coplanar with x (which is the projection point of M on L ); the line L and the plane \langle M,x\rangle are locally almost far at x . (\frac{3}{2}\frac{3}{2}22) (\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\Join,\Join) : the line M is contained in a symp \xi close to the projection point x of M onto L , and x is collinear to a line of \xi that is \xi -opposite M . (\frac{3}{2}2\frac{3}{2}2) (\perp\hskip-4.26773pt\perp,\Join,\perp\hskip-4.26773pt\perp,\Join) : the line L is contained in a symp \xi close to the projection point y of L onto M , and y is collinear to a line of \xi that is \xi -opposite L We note that dual distances are obtained from each other by interchanging and dualising the first and fourth entry, and dualising the second and third entry. We now have the following result. Lemma 4.7 . Let L and M be two arbitrary lines of an exceptional hexagonic Lie incidence geometry \Delta of uniform symplectic rank r . Then \delta(L,M) is one of the 4 -tuples enumerated above in Class I up to Class IV . All cases occur, except for metasymplectic spaces, where the following positions cannot occur: (1111) , (\frac{3}{2}\frac{3}{2}\frac{3}{2}\frac{3}{2}) , (2222) , (1\frac{3}{2}\frac{3}{2}\frac{3}{2}) , (\frac{3}{2}\frac{3}{2}\frac{3}{2}2) , (1\frac{3}{2}1\frac{3}{2}) , (11\frac{3}{2}\frac{3}{2}) , (\frac{3}{2}\frac{3}{2}22) and (\frac{3}{2}2\frac{3}{2}2) . Proof. We note that existence of a given mutual position is equivalent to the existence of its dual. Part I. It is convenient to first consider the case where all points of L have the same distance to all points of M . Then clearly we have one of the three completely homogeneous cases. The existence of (1111) is easy: consider two lines in a common singular subspace of dimension at least 3 . Conversely, clearly, if \delta(L,M)=(1111) , then L and M span a singular subspace of dimension 3. Since these do not exist in metasymplectic spaces, this mutual distance occurs if and only if \Delta is not a metasymplectic space. By duality, the same holds for (2222) . Let \delta(L,M)=(\frac{3}{2}\frac{3}{2}\frac{3}{2}\frac{3}{2}) . Pick x,x^{\prime}\in L and y,y^{\prime}\in M , x\neq x^{\prime} , y\neq y^{\prime} . If y^{\prime} were collinear to only a line of \xi(x,y) , then, by 2.19 (ii) , x and y^{\prime} would be special, a contradiction (and it follows that \Delta does not have type \mathsf{F_{4}} ). It follows that x^{\perp}\cap M^{\perp} is a singular subspace U of dimension r-2 . We claim that x^{\prime\perp}\cap U is (r-5) -dimensional. To fix the ideas, suppose \Delta has type \mathsf{E_{8,8}} . Since x^{\prime\perp}\cap\xi(x,y) is a 6^{\prime} -space and \langle x,U\rangle is a 6 -space, we have that x^{\prime\perp}\cap\langle x,U\rangle contains a line K . In \mathsf{Res}_{\Delta}(K) , we have a point z^{\prime} (corresponding to x^{\prime} ) close to each symp through a given 4 -space W (corresponding to \langle x,U\rangle ). If z^{\prime\perp}\cap W=\varnothing , then there exist non-collinear points a and b in different symps through W , both collinear to z^{\prime} . The symp \xi(a,b) contains z^{\prime} and a plane \alpha of W ; hence z is collinear to a line of W after all, a contradiction. So z^{\prime\perp}\cap W\neq\varnothing and, by parity, it is a line. The claim follows. It is now easy to see that, (i) in case of \mathsf{E_{8,8}} , L and M arise from opposite lines in the residue of a plane (which is a parapolar space of type \mathsf{D_{5,5}} ); (ii) in case of \mathsf{E_{7,1}} , L and M arise from opposite lines in a given para of type \mathsf{D_{5,5}} in a given point residual; (iii) in case of \mathsf{E_{6,2}} , L and M are opposite lines in a given para of type \mathsf{D_{5,5}} . The last claim might be the least straightforward, as in this case L^{\perp}\cap M^{\perp}=\varnothing . So let us prove this case as an example (the other cases are then easier because the parapolar spaces in the respective residues are simpler; they have types \mathsf{D_{5,5}} and \mathsf{D_{6,6}} , respectively). Translated to type \mathsf{E_{6,1}} , we have to prove that, if each 5 -space through a given plane \alpha intersects each 5 -space through another given plane \beta in exactly a point, then either \alpha meets \beta in a point at which they are locally opposite, or \alpha and \beta are contained in a symp in which they are (locally) opposite. So suppose \alpha and \beta do not intersect. Then one checks that, if U_{1},U_{2} are two distinct 5 -spaces through \alpha and W_{1},W_{2} two distinct 5 -spaces through \beta , the points p_{ij}=U_{i}\cap W_{j} , i,j\in\{1,2\} , form a quadrangle. That quadrangle is contained in a unique symp \xi that contains both \alpha and \beta . If the latter are not \xi -opposite, then, arguing in the polar space \mathsf{D_{5,1}}(\mathbb{K}) , we find 4^{\prime} -spaces through them that intersect in a plane; hence this yields adjacent 5 -spaces through them, a contradiction. This concludes the completely homogeneous case. We obtain all members of Class I. Part II. Next we consider the case where each point of L\cup M has a unique nearest point on the other line. It is easy to deduce that the distances between such nearest points are always the same. This distance can be equal, collinear, symplectic or special, in which case the other pairs are collinear, symplectic, special or opposite, respectively (use 2.16 for instance). Then it is easy to see that the lines are equal, opposite in a symp, the projection of each other from opposite symps, or opposite, respectively. Hence we obtain precisely all cases of Class II. Part III. Having done the more homogeneous cases separately, we can proceed to consider the smallest distance that can occur between points of L and M . In order to do so, we let p\in L and q\in M be points at minimal distance. Case 1: p=q . Here the lines L and M meet in p=q , and we clearly have only the three possibilities (0111) , (011\frac{3}{2}) and (0112) of Class III. Existence is trivial in these cases. Case 2: p\perp q . We set K equal to the line through p and q . We now consider the different possible mutual positions of L and K , and of K and M . First suppose that K and M are coplanar; say they span the plane \alpha . Then \alpha and L are contained in a common symp if and only if one of the following two possibilities occurs: (1) L\perp M ; then we have case (1111) , (2) |L^{\perp}\cap M|=1 ; then case (111\frac{3}{2}) occurs. So we may assume that no point of L\setminus\{p\} is collinear to any point of \alpha\setminus\{p\} . Let \xi be a symp containing \alpha . There are again two possibilities. (1) L^{\perp}\cap\xi is a line N . Then N is not contained in \alpha , so that there is a unique point q^{\prime}\in M collinear to all points of N . Then q^{\prime}\perp\hskip-4.26773pt\perp p^{\prime} , for all p^{\prime}\in L\setminus\{p\} , and q^{\prime\prime}\Join p^{\prime} , for all q^{\prime\prime}\in M\setminus\{q^{\prime}\} and all p^{\prime}\in L\setminus\{p\} . We get (11\frac{3}{2}2) . (2) L^{\perp}\cap\xi is a maximal singular subspace U . Then U\cap M=\varnothing , and each point of M is symplectic to each point of L\setminus\{p\} . We obtain (11\frac{3}{2}\frac{3}{2}) . If L and K are coplanar, then we obtain the inverse distances (1111) , (111\frac{3}{2}) , (1\frac{3}{2}12) and (1\frac{3}{2}1\frac{3}{2}) . Hence we may assume that K is not coplanar with either L or M . Pick p^{\prime}\in L\setminus\{p\} and q^{\prime}\in M\setminus\{q\} . If both pairs \{p^{\prime},q\} and \{p,q^{\prime}\} are special, then we have (1223) . So we may assume \{p,q^{\prime}\} is symplectic. Again, there are some possibilities. (1) L^{\perp}\cap\xi(p,q^{\prime}) is a line N not \xi(p,q^{\prime}) -opposite M . Then our assumptions imply that q\perp N , and so we obtain (1\frac{3}{2}\frac{3}{2}2) . (2) L^{\perp}\cap\xi(p,q^{\prime}) is a line N which is \xi(p,q^{\prime}) -opposite M . Clearly, this gives rise to (1\frac{3}{2}22) . (3) L^{\perp}\cap\xi(p,q^{\prime}) is a maximal singular subspace U . Then we clearly have (1\frac{3}{2}\frac{3}{2}\frac{3}{2}) . The case where L and K are contained in a common symp gives additionally rise to the inverse (12\frac{3}{2}2) of (1\frac{3}{2}22) . This takes care of all distances beginning with “collinear”. Case 3: p\perp\hskip-4.26773pt\perp q . Let \xi be the symp containing both p and q . Note that neither L nor M is contained in \xi , as otherwise we are back in the previous case. There are a few possibilities. (1) Both L^{\perp}\cap\xi and M^{\perp}\cap\xi are maximal singular subspaces, and they intersect in a subspace of dimension at least 1 . Then we are in the homogeneous case (\frac{3}{2}\frac{3}{2}\frac{3}{2}\frac{3}{2}) , which we already discussed in detail in Part I. (2) Both L^{\perp}\cap\xi and M^{\perp}\cap\xi are maximal singular subspaces, and they intersect in exactly a point. By Part I, we are not in case (\frac{3}{2}\frac{3}{2}\frac{3}{2}\frac{3}{2}) ; hence we have distance (\frac{3}{2}\frac{3}{2}\frac{3}{2}2) . (3) Both L^{\perp}\cap\xi and M^{\perp}\cap\xi are maximal singular subspaces, and they are disjoint. Select p^{\prime}\in L\setminus\{p\} and q^{\prime}\in M\setminus\{q\} . We claim that p^{\prime}\perp\hskip-4.26773pt\perp q^{\prime} . Indeed, the only alternative is p^{\prime}\Join q^{\prime} . If so, let p^{\prime}\perp r\perp q^{\prime} . Then, by considering symps through points of U:=L^{\perp}\cap\xi and q^{\prime} , we see that r\perp U ; likewise, r is collinear to each point of M^{\perp}\cap\xi , a contradiction. Hence this leads to (\frac{3}{2}\frac{3}{2}\frac{3}{2}\frac{3}{2}) , which we discussed in Part I. (4) Suppose L^{\perp}\cap\xi is a maximal singular subspace U of \xi and M^{\perp}\cap\xi is a line K . If K\cap U is a point x , then \langle x,L\rangle and \langle x,M\rangle are planes intersecting in x , and it follows from 2.21 that each point of L is symplectic to either a unique point of M , or all points of M . This now clearly leads to (\frac{3}{2}\frac{3}{2}\frac{3}{2}2) again (and replacing p with the unique point on L symplectic to all points of M , we are back to (2)). Hence we may assume that K\cap U=\varnothing . If some point p^{\prime} of L\setminus\{p\} were symplectic to some point of M\setminus\{q\} , then we could replace p by p^{\prime} and are back to situation (2). If no point of L\setminus\{p\} is symplectic to any point of M\setminus\{q\} , then we have distance (\frac{3}{2}2\frac{3}{2}2) . (5) Similarly, L^{\perp}\cap\xi a line and M^{\perp}\cap\xi a maximal singular subspace lead to (\frac{3}{2}\frac{3}{2}\frac{3}{2}2) or (\frac{3}{2}\frac{3}{2}22) . (6) Finally, suppose L^{\perp}\cap\xi is a line K and M^{\perp}\cap\xi is a line N . If K and N intersect, then we are back to a previous case already handled, namely (\frac{3}{2}22\frac{3}{2}) . If K and N are \xi -special, then we claim that we have distance (\frac{3}{2}222) . Indeed, the alternative would be that some point q^{\prime}\in M\setminus\{q\} is symplectic to some point p^{\prime}\in L\setminus\{p\} . This would imply that p^{\prime} is locally close to the symp determined by q^{\prime} and N^{\perp}\cap K . But this implies that q is symplectic to p^{\prime} , which contradicts the fact that q is not collinear to all points of K . If K and N are \xi -opposite, then, using similar arguments, we have distance (\frac{3}{2}223) . This takes care of the case p\perp\hskip-4.26773pt\perp q . Case 4: p\Join q . Since we may assume we are not in the “Completely homogeneous” case, there are opposite pairs of points on L\cup M . Since we may also assume that we are not in the “Projection homogeneous” case, we may assume that no point of L\setminus\{p\} is special to any point of M\setminus\{q\} . But then every point of L is special to q and every point of M is special to p , whereas each point of L\setminus\{p\} is opposite each point of M\setminus\{q\} . This is distance (2223) . One checks that the cases involving a singular subspace of dimension at least 3 (this includes each case where some point is collinear to a maximal singular subspace of a symp) are precisely the positions (1111) , (\frac{3}{2}\frac{3}{2}\frac{3}{2}\frac{3}{2}) , (2222) , (1\frac{3}{2}\frac{3}{2}\frac{3}{2}) , (\frac{3}{2}\frac{3}{2}\frac{3}{2}2) , (1\frac{3}{2}1\frac{3}{2}) , (11\frac{3}{2}\frac{3}{2}) , (\frac{3}{2}\frac{3}{2}22) and (\frac{3}{2}2\frac{3}{2}2) . The lemma is completely proved. ∎ For any ordered pair of lines (L,M) , we call each point of L distinct from the projection point, if it exists, a free point (for (L,M) ) ; hence if there is no projection point, every point is free. We define the following order 0<1<\frac{3}{2}<2<3 . We have the following lemma. Lemma 4.8 . Let L,M\in\mathscr{L} be two lines of an exceptional hexagonic Lie incidence geometry \Delta=(X,\mathscr{L}) . Let x\in L be a free point for (L,M) . Then there exists a line K through x , not locally opposite L at x , such that, if \delta(L,M)=(abcd) , then whenever L^{\prime}\ni x is locally opposite K at x , then \delta(L^{\prime},M)=(cdef) , for some e,f , if c\leq d , otherwise \delta(L^{\prime},M)=(dcef) , where (cdef) or (dcef) is determined according to Table 1 . Applying this assertion to (L^{\prime},M) again and again, we eventually arrive at an opposite pair. In Table 1 , we list the consecutive distances when we apply this algorithm. The penultimate column lists the local mutual position of L and K (with respect to the first arrow in the row), and the last column mentions when K is not unique. \{1\} (0110) \to (0112) \to (1223) \to (2332) equal \{2\} (0111) \to (11\frac{3}{2}2) \to (\frac{3}{2}223) \to (2332) eq or coll not unique \{3\} (011\frac{3}{2}) \to (1\frac{3}{2}22) \to (2223) \to (2332) equal \{4\} (111\frac{3}{2}) \to (1\frac{3}{2}22) \to (2223) \to (2332) collinear \{5\} (1\frac{3}{2}\frac{3}{2}1) \to (1\frac{3}{2}22) \to (2223) \to (2332) symplectic \{6\} (0112) \to (1223) \to (2332) equal \{7\} (1\frac{3}{2}12) \to (1223) \to (2332) collinear \{8\} (11\frac{3}{2}2) \to (\frac{3}{2}223) \to (2332) equal \{9\} (1\frac{3}{2}\frac{3}{2}2) \to (\frac{3}{2}223) \to (2332) collinear \{10\} (12\frac{3}{2}2) \to (\frac{3}{2}223) \to (2332) symplectic \{11\} (\frac{3}{2}22\frac{3}{2}) \to (\frac{3}{2}223) \to (2332) collinear \{12\} (1223) \to (2332) equal \{13\} (1\frac{3}{2}22) \to (2223) \to (2332) eq or coll not unique \{14\} (\frac{3}{2}222) \to (2223) \to (2332) coll or sympl not unique \{15\} (\frac{3}{2}223) \to (2332) collinear \{16\} (2223) \to (2332) symplectic \{17\} (2332) \to (2332) special \{18\} (1111) \to (11\frac{3}{2}2) \to (\frac{3}{2}223) \to (2332) collinear not unique \{19\} (\frac{3}{2}\frac{3}{2}\frac{3}{2}\frac{3}{2}) \to (\frac{3}{2}\frac{3}{2}22) \to (2223) \to (2332) collinear not unique \{20\} (2222) \to (2223) \to (2332) symplectic not unique \{21\} (1\frac{3}{2}1\frac{3}{2}) \to (1\frac{3}{2}22) \to (2223) \to (2332) collinear \{22\} (11\frac{3}{2}\frac{3}{2}) \to (\frac{3}{2}\frac{3}{2}22) \to (2223) \to (2332) equal \{23\} (1\frac{3}{2}\frac{3}{2}\frac{3}{2}) \to (\frac{3}{2}\frac{3}{2}22) \to (2223) \to (2332) collinear not unique \{24\} (\frac{3}{2}\frac{3}{2}22) \to (2223) \to (2332) collinear not unique \{25\} (\frac{3}{2}2\frac{3}{2}2) \to (\frac{3}{2}223) \to (2332) symplectic \{26\} (\frac{3}{2}\frac{3}{2}\frac{3}{2}2) \to (\frac{3}{2}223) \to (2332) collinear Table 1. Combing distances between lines Proof. We have to treat the 26 cases one by one. However, some cases are immediate or at least easy, and we skip those. It concerns many of the cases where K=L , namely \{1\} , \{3\} , \{6\} , \{12\} . Other cases are easy once one knows K , and we only give that information below. Basically, K can always be thought of as a kind of projection of M onto x . In cases \{12\} , \{15\} , \{16\} and \{17\} , the point x is opposite some point of M , and then K is really that projection, and so these cases are straightforward and we skip them, too. More tricky cases are treated in full detail. It concerns in particular the cases that cannot occur in type \mathsf{F_{4}} . \{2\} (=,\perp,\perp,\perp)\to(\perp,\perp,\perp\hskip-4.26773pt\perp,\Join) . The line K is any line through x in the plane spanned by L and M . \{4\} (\perp,\perp,\perp,\perp\hskip-4.26773pt\perp)\to(\perp,\perp\hskip-4.26773pt\perp,\Join,\Join) . The line K is the line joining x with the unique point of M collinear to each point of L . \{5\} (\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\perp)\to(\perp,\perp\hskip-4.26773pt\perp,\Join,\Join) . Here, K is the line joining x with the unique point y\in M collinear to x . Then let N be a line through x locally opposite K and pick z\in N\setminus\{x\} . Let \xi be the symp containing L and M and set z^{\perp}\cap\xi=M^{\prime} . Then M^{\prime} and M are \xi -opposite, since if they were not, either x would be collinear to M (which contradicts the fact that L and M are \xi -opposite), or y would be collinear to all points of M^{\prime} (which contradicts the fact that z\Join y by the choice of N locally opposite K ). \{7\} (\perp,\perp\hskip-4.26773pt\perp,\perp,\Join)\to(\perp,\Join,\Join,\equiv) . Let y\in M be collinear to each point of L . Then K=xy , and the rest follows from 2.16 . \{8\} (\perp,\perp,\perp\hskip-4.26773pt\perp,\Join)\to(\perp\hskip-4.26773pt\perp,\Join,\Join,\equiv) . Here K=L . Set p\in L the unique point collinear to each point of M and q\in M the unique point of M symplectic to x . By 2.16 , every other point of M is opposite each point of L^{\prime}\setminus\{x\} . It follows that the distance between L^{\prime} to M is (\frac{3}{2}223) . \{9\} (\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\Join)\to(\perp\hskip-4.26773pt\perp,\Join,\Join,\equiv) . Let p\in L and q\in M be collinear. Let \xi be the symp through M and p and let z be the unique point of \xi collinear to M and collinear to x . Set K=xz . Then, precisely like in the previous case \{8\} , we conclude that the distance between L^{\prime} to M is (\frac{3}{2}223) . \{10\} (\perp,\Join,\perp\hskip-4.26773pt\perp,\Join)\to(\perp\hskip-4.26773pt\perp,\Join,\Join,\equiv) . Let \xi be the symp containing L and a unique point q\in M . Then there is a unique point z\in\xi collinear to M and x . Putting K=xz , the rest of the argument is the same as for the previous cases \{8\} and \{9\} . \{11\} (\perp\hskip-4.26773pt\perp,\Join,\Join,\perp\hskip-4.26773pt\perp)\rightarrow(\perp\hskip-4.26773pt\perp,\Join,\Join,\equiv) . 2.24 yields a point z collinear to each point of L\cup M . Then K=xz , and the same arguments as in the three previous cases imply that the distance between L^{\prime} to M is (\frac{3}{2}223) . \{13\} (\perp,\perp\hskip-4.26773pt\perp,\Join,\Join)\rightarrow(\Join,\Join,\Join,\equiv) . Let p\in L be the point of L contained in a common symp \xi with M . Then x\neq p is collinear to a line N of \xi . This line N is \xi -opposite M . Then K is any line through x in the plane spanned by x and N . Set K\cap N=\{p^{\prime}\} and p^{\prime\perp}\cap M=\{q^{\prime}\} . Then the position (2223) between L^{\prime} and M follows from the facts that L^{\prime} is locally opposite K at x ; K is locally opposite p^{\prime}q^{\prime} at p^{\prime} , and p^{\prime}q^{\prime} locally symplectic to M at q^{\prime} . \{14\} (\perp\hskip-4.26773pt\perp,\Join,\Join,\Join)\rightarrow(\Join,\Join,\Join,\equiv) . Let \xi be the symp through the unique points p\in L and q\in M . Set N:=x^{\perp}\cap\xi , and let z be the point on N collinear to q . 2.21 yields a line N^{\prime}\ni z consisting of all points collinear to a point of M and x . Then K is any line through x in the plane \langle x,N^{\prime}\rangle . Similar arguments as in the previous case \{13\} show that the mutual position of L^{\prime} and M is (2223) . \{18\} (\perp,\perp,\perp,\perp)\rightarrow(\perp,\perp,\perp\hskip-4.26773pt\perp,\Join) . Let K be any line through x intersecting M . Then, by 2.21 , there is a unique point on M symplectic to the points of L^{\prime}\setminus\{x\} . This implies that the mutual position between L^{\prime} and M is (11\frac{3}{2}2) . \{19\} (\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp)\rightarrow(\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\Join,\Join) . Pick two points y_{1},y_{2} on M . By 2.19 , the point y_{2} is collinear to a maximal singular subspace U of \xi(x,y_{1}) . Then x is collinear to a hyperplane W of U , and K is any line joining x with a point z of W . Let u\in L^{\prime}\setminus\{x\} be arbitrary. Then u and z are special. Suppose u were symplectic to some point of M , and we may without loss of generality assume that point is y_{1} . Then u would have to be collinear to a line N of \xi(x,y_{1}) including x and y_{1} . But x is not collinear to any point of M . It follows that u is special to every point of M . With that, L^{\prime} contains exactly one point, which is x , symplectic to all points of L_{2} ; and otherwise, all remaining points of L^{\prime} are special to all points of L_{2} , since u\in M was arbitrary. \{20\} (\Join,\Join,\Join,\Join)\rightarrow(\Join,\Join,\Join,\equiv) . 2.21 yields a line N consisting of all points collinear to x and some point of M . Then K is any line through x in the plane \langle x,N\rangle . Note that K is locally symplectic to L at x . Let u\in L^{\prime}\setminus\{x\} . Then there is a unique point on N symplectic to u ; all other points of N are special to u . 2.16 implies that u is opposite all but exactly one point of M . Since x is special to all points of M , the mutual distance between L^{\prime} and M is (2223) . \{21\} (\perp,\perp\hskip-4.26773pt\perp,\perp,\perp\hskip-4.26773pt\perp)\rightarrow(\perp,\perp\hskip-4.26773pt\perp,\Join,\Join) . Let q be the unique point of M collinear to all points of L . Then K is the line qx . Let q^{\prime} be any point of M\setminus\{q\} . Then M is contained in the symp \xi(q^{\prime},x) . Let u be an arbitrary point of L^{\prime} not equal to x . Then u and q are special, and consequently u can only be collinear to a line of the symp \xi(q^{\prime},x) through x , which implies that u and q^{\prime} are special. In summary, q is collinear to x and special to every other point of L^{\prime} , and every point in M\setminus\{q\} is symplectic to x and special to every point in L^{\prime}\setminus\{x_{1}\} . \{22\} (\perp,\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp)\rightarrow(\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\Join,\Join) . Let p be the point of L collinear to every point of M . Then x\neq p . Here, K=L . Let u be some arbitrary point of L^{\prime} not equal to x . Then u and p are special. The point x is symplectic to every point of M . Considering the symp \xi containing x and an arbitrary point y\in M , we see that u^{\perp}\cap\xi is a line N through x (indeed, p\in\xi and u\Join p ). Now, y is only collinear to a unique point of L , since y is not collinear to x . It follows, with 2.19 , that u is special to y . We obtain (\frac{3}{2}\frac{3}{2}22) . \{23\} (\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp)\rightarrow(\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\Join,\Join) . Let \xi be the symp through M and the unique point p of L collinear to some point q of M . It is easy to see that x^{\perp}\cap\xi is a maximal singular subspace U of \xi . Then K is any line through x and a point of M^{\perp}\cap U . Considering the respective symps through x and the points of M , we see that the points of M and L^{\prime}\setminus\{x\} are special. Hence we obtain \left(\frac{3}{2}\frac{3}{2}22\right) . \{24\} (\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\Join,\Join)\rightarrow(\Join,\Join,\Join,\equiv) . Let p\in L be symplectic to all points of M . 2.22 yields a singular subspace U=p^{\perp}\cap M^{\perp} contained in each symplecton containing p and a point of M , and such that \langle p,U\rangle is a maximal singular subspace of each such symp. Combining this with 2.21 , we find a line N\subseteq U consisting of all points collinear to x and some point of M . Then K is an arbitrary line through x in the plane \langle x,N\rangle . Taking into account that N and M are contained in a common symp \xi in which they are \xi -opposite, we arrive at (2223) for the distance between L^{\prime} to M . \{25\} (\perp\hskip-4.26773pt\perp,\Join,\perp\hskip-4.26773pt\perp,\Join)\rightarrow(\perp\hskip-4.26773pt\perp,\Join,\Join,\equiv) . Here, there is a unique point q\in M symplectic to each point of L , in particular to x . The line M is collinear to a unique line N of \xi(q,x) , and x is collinear to a unique point z of N . Then K is the line xz , and it is immediate that L^{\prime} and M are at mutual distance (\frac{3}{2}223) . \{26\} (\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\perp\hskip-4.26773pt\perp,\Join)\rightarrow(\perp\hskip-4.26773pt\perp,\Join,\Join,\equiv) . Let p\in L be the point symplectic to all points of M , and let q\in M be the point symplectic to all points of L . Then M is collinear to a unique line N of \xi(q,x) , and x is collinear to a unique point z of N . We define K=xz . Then 2.16 implies that each point of L^{\prime}\setminus\{x\} is opposite each point of M\setminus\{q\} . Since x\perp\hskip-4.26773pt\perp q and x is special to all points of M\setminus\{q\} , we conclude that the mutual distance of L^{\prime} and M is given by (\frac{3}{2}223) . This completes the proof of the lemma. ∎ The length of the sequence in the previous lemma is called the level of the corresponding line M (with respect to L ), except that when M is opposite L , we say it has level 0 . 4.2.3. Algorithms and end of the proof Let \Delta=(X,\mathscr{L}) be a finite exceptional hexagonic Lie incidence geometry whose lines carry exactly s+1 points. We introduce two algorithms, that we will call combing algorithms . They require that certain conditions are met, and we will also introduce these. Naturally, we will only run them when all conditions are satisfied. They are defined as follows. Definition 4.9 (The combing algorithms) . Let L_{0},L_{1},\ldots,L_{s}\in\mathscr{L} be s+1 lines of \Delta and let L be another arbitrary line. Suppose (ALG1) There exists a point x\in L which is free for every pair (L,L_{i}) , i=0,1,\ldots,s . Condition (ALG1) just means that the set of projection points on L with respect to the lines L_{i} , i=0,1,\ldots,s , does not cover L . For each L_{i} , i\in\{0,1,\ldots,s\} , and each point x on L that is free with respect to each line L_{j} , j\in\{0,1,\ldots,s\} , we define a line M_{i} as follows. If L_{i} is opposite L , then M_{i} is the unique line through x containing a point collinear to L_{i} . The pair \{L,M_{i}\} has distance (0112) , or is, in other words, locally opposite at x . If L_{i} is not opposite L , then we set M_{i} equal to the line K , as (perhaps not uniquely) defined in 4.8 . If the line K is not uniquely defined there, then we arbitrarily choose one (and one may think of taking the closest to L , if this exists). We set \mathscr{M}=\{M_{i}\mid i\in\{0,1,\ldots,s\}\} . (ALG2) There exists a line L^{\prime} through x locally opposite each member of \mathscr{M} at x . The line L^{\prime} is the outcome of the first combing algorithm. We then replace L with L^{\prime} . (ALG3) There exists a line M\in\mathscr{M} locally opposite L at x and there exists a line L^{\prime\prime} through x locally opposite each member of \mathscr{M}\setminus\{M\} (where we view \mathscr{M} as a set and not as a multiset) and not opposite M . The line L^{\prime\prime} is the outcome of the second combing algorithm (combing back at M ). We then replace L with L^{\prime\prime} . We observe: Lemma 4.10 . Under the second combing algorithm, level 0 always goes to level at most 1 . We will always use 4.5 to be able to perform the first combing algorithm, whereas 4.6 will allow us to perform the second combing algorithm. This is roughly the content of the proof of the next result. Proposition 4.11 . Every set T=\{L_{0},L_{1},\ldots,L_{s}\} of s+1 lines in a metasymplectic space, not isomorphic to \mathsf{F_{4,4}}(\sqrt{s},s) , or in an exceptional long root subgroup geometry of type \mathsf{E_{6}} , \mathsf{E_{7}} or \mathsf{E_{8}} , where every line has exactly s+1 points, such that every other line is not opposite at least one member of T , is a geometric line in the line-Grassmannian geometry, that is, has the property that every other line is either not opposite a unique member of T , or opposite no member of T . Proof. Obviously, a geometric line has the stated property. So assume now that T is not a geometric line, but every other line is not opposite at least one member of T . The only way in which we can violate the defining property of a geometric line is to assume the existence of a line L not opposite at least 2 members of T and opposite at least one member of T . We prove that this leads to a contradiction. The rough idea is to apply the combing algorithms to L and T until we find a line opposite all members of T . Since our proof will be inductive in some sense, it is important that after each application of the combing algorithm, the new line L satisfies the same assumption, that is, the new line L is not opposite at least two members of T and opposite at least one member of T , or the proof (locally) ends and L is opposite each member of T . (We say in these cases that the new line L is legal .) This little condition implies that we cannot blindly run the combing algorithms, but we have to choose the right one. The way we do this goes as follows. We start by noting that a line M_{i}\in\mathscr{M} is locally opposite L if and only if L_{i}\equiv L . Hence, since at least two lines of T are not opposite L , and at least one line of T is opposite L , 4.5 implies that (ALG2) is satisfied. Also, since opposite lines do not define projection points, and there is at least one line in T opposite L , (ALG1) is satisfied. Moreover, 4.6 allows us to run the second combing algorithm since T contains at least one line opposite L . Now we combine the two combing algorithms in one overarching algorithm that proves the theorem. That algorithm goes as follows. \bullet If T contains at least two members at level at least 2 , then we apply the first combing algorithm. Note that elements of T opposite L remain opposite L^{\prime} , and elements of T at level k\geq 1 with respect to L are at level k-1 with respect to L^{\prime} . Hence L^{\prime} is legal, and the maximum level decreases. \bullet If T contains exactly one member L_{0} at level at least 2 (and hence at least one member L_{1} of level 1 ), then we apply the second combing algorithm combing back at an arbitrary M_{2}\in\mathscr{M} locally opposite L . Then L_{0} comes at level at least 1 , L_{1} at level 0 , and L_{2} at level 1 . Hence L^{\prime\prime} is legal. Moreover, 4.10 guarantees that the maximum level again decreases. \bullet We apply the previous two steps as long as the maximal level is at least 2 . If the maximal level is or becomes 1 , then we apply the first combing algorithm and obtain a line L^{\prime} opposite each member of T , a contradiction that proves the assertion. ∎ 4.2.4. The exceptional case \mathsf{F_{4,4}}(q,q^{2}) The above does not work for lines of metasymplectic spaces \Delta isomorphic to \mathsf{F_{4,4}}(q,q^{2}) , because we cannot apply 4.5 since in the polar space \mathsf{B_{3,1}}(q,q^{2}) , corresponding to the residue of a point p , there are sets of q^{2}+1 planes admitting no common opposite plane, and yet not isomorphic to a geometric line (pencil of planes). The examples are sets of planes through a common point b forming a spread in a subquadrangle of order (q,q) of the residue at b . We will call such an example an OBS (ovoidal blocking set) . So we have to provide a different proof. Note that, viewed in \Delta , the point b is a symp, and the elements of an OBS are lines through a common point forming an ovoid in a subquadrangle of order (q,q) of the residue at that common point. Also, such a set will be called an OBS. Also, note that, viewed in \mathsf{F_{4,1}}(q,q^{2}) , the point p is a symp, and the elements of an OBS are planes through a common point b of the symp p forming a spread in a subquadrangle of order (q,q) of the residue at b . We first observe that it is really an example of a blocking set. Lemma 4.12 . Let T be a set of q^{2}+1 lines of \Delta=\mathsf{F_{4,4}}(q,q^{2}) incident with a common point b and forming an ovoid in a subquadrangle of the residue \mathsf{Res}_{\xi}(b) at b of some symp \xi through b . Then no line of \Delta is opposite each member of T . Proof. This follows directly from 2.27 . ∎ We now show a converse to 4.12 , that is, any set T of q^{2}+1 lines of \Delta=\mathsf{F_{4,4}}(q,q^{2}) with the property that no line of \Delta is opposite each member of T is either a planar line pencil, or an OBS. Let T=\{L_{0},\ldots,L_{t}\} , t=q^{2} , be a set of lines of \Delta\cong\mathsf{F_{4,4}}(q,q^{2}) admitting no common opposite line. Note that each point of a singular subspace S is opposite some point of a given symp \xi if, and only if, S and \xi are far. Indeed, if there is a symp through S opposite \xi , then clearly, each point of S is opposite some point of \xi . Now suppose each point of S is opposite some point of \xi . Pick x\in S and let \zeta be the unique symp through x intersecting \xi . Our assumption implies that S and \zeta intersect just in x . Hence we can find a symp \zeta^{\prime} through S locally opposite \zeta in x . Then \zeta^{\prime} is opposite \xi by 2.26 . Lemma 4.13 . There exist a point b and a symp \xi in \Delta , with b\in\xi , such that both b and \xi are far from each member of T . For each such b and \xi we have that the projections of the members of T onto b and \xi , respectively, form either both a planar line pencil, or both an OBS. Proof. We can choose points b_{i} contained in L_{i} such that the b_{i} do not form a geometric line in \Delta . Then 4.1 , 4.2 and Theorem 4.4 yield a point b opposite all b_{i} , i\in\{0,1,\ldots,s\} . So, b is far from each member of T . Now set T^{\prime}=\{L_{i}^{\prime}=\mathsf{proj}^{b_{i}}_{b}(L_{i})\mid i\in\{0,1,\ldots,s\}\} . If T^{\prime} is not an OBS and not a planar line pencil, then we can find a line L through b locally opposite each member of T^{\prime} , and so, by 2.26 , L is opposite each member of T , a contradiction. We conclude that T^{\prime} is contained in a symp \zeta through b . Now let \xi be a symp locally opposite \zeta at b . Then, again by 2.26 , the projection \xi_{i} of \zeta onto b_{i} is opposite \xi . However, \xi_{i} contains L_{i} as \zeta contains L_{i}^{\prime} , i=0,1,\ldots,s . Hence \xi is far from each member of T . Now let L_{i}^{\prime\prime} be the projection of L_{i} onto \xi and let L_{i}^{\prime\prime\prime} be the unique line of \xi collinear to L_{i}^{\prime} . We claim that L_{i}^{\prime\prime} intersects L_{i}^{\prime\prime\prime} , which then shows that the projection of T^{\prime} onto \xi coincides with the projection onto b of the projection of T onto \xi , and hence T^{\prime} is isomorphic to the projection of T onto \xi and the lemma follows. Let M^{\prime\prime\prime}_{i} be the unique line through b intersecting L_{i}^{\prime\prime} , say in the point x_{i}^{\prime\prime} . Since x_{i}^{\prime\prime}\in L_{i}^{\prime\prime} , there is a unique point x_{i}\in L_{i} symplectic to x_{i}^{\prime\prime} . Then b is collinear to a unique line K_{i} of \xi(x_{i},x_{i}^{\prime\prime}) through x_{i}^{\prime\prime} , and x_{i} is collinear to a unique point x_{i}^{\prime} of K_{i} . Now x_{i}\perp x_{i}^{\prime}\perp b defines a path of length 2 from x_{i}\in L_{i} to b , hence bx_{i}^{\prime}=L_{i}^{\prime} and M_{i}^{\prime\prime}=L_{i}^{\prime\prime\prime} and the claim follows. ∎ Lemma 4.14 . Each pair of members of T is either contained in a symp, or has a point in common. Proof. It is convenient to consider the dual situation, that is, T corresponds to a set T^{*} of planes \{\alpha_{0},\ldots,\alpha_{s}\} , s=q^{2} , of \mathsf{F_{4,1}}(q,q^{2}) . By 4.13 we can find a symp \xi far from each member of T^{*} . Hence we can project all planes \alpha_{i} onto \xi and obtain planes \alpha_{i}^{\prime} . By 2.27 and 2.29 , the \alpha_{i}^{\prime} form a full plane pencil or an OBS. In particular, all planes \alpha_{i}^{\prime} contain a common point q , and for each line L_{0} of \alpha_{0}^{\prime} through q , except for the possible intersection line with \alpha_{1}^{\prime} , there exist q^{2} lines L_{1} of \alpha_{1}^{\prime} through q not coplanar with L_{0} . Let z_{0} and z_{1} be two arbitrary points on L_{0} and L_{1} , respectively. Select a point p in \xi not collinear to q , but collinear to both z_{0} and z_{1} . Let p_{i} and x_{i} be the unique points in \alpha_{i} symplectic to z_{i} and q , respectively, i=0,1 . Since p is collinear to a unique line of \xi(p_{i},z_{i}) , there is a unique point y_{i} in \xi(p_{i},z_{i}) collinear to p_{i},z_{i} and p , i=0,1 . The line L_{i}^{\prime\prime}=py_{i} is the projection of L_{i} from x_{i} onto p . By the “dual” of 4.13 , the points y_{0} and y_{1} are either collinear or symplectic. But since \xi(p_{i},z_{i}) is symplectic to \xi , and z_{0} is symplectic to z_{1} , the symps \xi(p_{0},z_{0}) and \xi(p_{1},z_{1}) are opposite (use 2.26 together with the observation that symplectic symps are locally opposite at their intersection point). Hence y_{0} and y_{1} are symplectic. Let q_{i} be the unique point of \xi(p_{i},z_{i}) collinear to q,p_{i} and z_{i} , i=0,1 . Then, varying p over all points of \xi not collinear to q , but collinear to both z_{0} and z_{1} , we deduce that p_{0}^{\perp}\cap z_{0}^{\perp}\setminus q_{0}^{\perp} corresponds to p_{1}^{\perp}\cap z_{1}^{\perp}\setminus q_{1}^{\perp} under the projection map from \xi(p_{0},z_{0}) to \xi(p_{1},z_{1}) given on the points by “being symplectic”. It easily follows that p_{0}^{\perp}\cap z_{0}^{\perp} corresponds to p_{1}^{\perp}\cap z_{1}^{\perp} . Hence (p_{0}^{\perp}\cap z_{0}^{\perp})^{\perp} corresponds to (p_{1}^{\perp}\cap z_{1}^{\perp})^{\perp} . Since symps are isomorphic to quadrics Q^{-}(7,q) , which are embedded in non-degenerate (symplectic) polarities, we have (p_{i}^{\perp}\cap z_{i}^{\perp})^{\perp}=\{p_{i},z_{i}\} , i=0,1 . Since z_{0} corresponds to z_{1} , we conclude that p_{0} and p_{1} are symplectic. We have shown that p_{0} is symplectic to all points of \alpha_{1} , except possibly the points of a unique line. It then easily follows that p_{0} is collinear or symplectic to any given point of \alpha_{1} . By the arbitrariness of p_{0} in \alpha_{0}\setminus\{x_{0}\} , we deduce that any pair of points in \alpha_{0}\cup\alpha_{1} is symplectic, collinear or identical. Consider any symp \xi_{1} through \alpha_{1} . If p_{0}\in\xi_{1} , then p_{0} is collinear to at least a line of \alpha_{1} . If p_{0}\notin\xi_{1} , then it must be close to it and the line p_{0}^{\perp}\cap\xi_{1} must be contained in \alpha_{1} . Hence in any case, there is a line of \alpha_{1} collinear to p_{0} , and so we can assume that \xi_{1} contains p_{0} . Suppose some point r_{0}\in\alpha_{0} does not belong to \xi_{1} . Then r_{0}^{\perp}\cap\xi_{1}\subseteq\alpha_{1} , as before, showing p_{0}\in\alpha_{1} . Hence we have shown that either \alpha_{0} and \alpha_{1} are contained in a symp, or they have a point in common. This means that, if \alpha_{i} corresponds to L_{i} , then L_{0} and L_{1} either intersect in a point, or are contained in a common symp. The assertion follows by the arbitrariness of L_{0} and L_{1} in T . ∎ We can now classify the blocking sets of lines of size q^{2}+1 in \mathsf{F_{4,4}}(q,q^{2}) . Theorem 4.15 . Let T be a set of q^{2}+1 lines of \Delta=\mathsf{F_{4,4}}(q,q^{2}) . Then all members of T are incident with a common point b and form either a planar line pencil, or an ovoid in a subquadrangle of the residue \mathsf{Res}_{\xi}(b) at b of some symp \xi through b if, and only if, no line of \Delta is opposite each member of T . Proof. The “only if” part is 4.12 . We now show the “if” part. We first claim that each pair of members of T intersect nontrivially. Indeed, we may assume for a contradiction that L_{0} and L_{1} do not intersect. Then by 4.14 they are contained in a common symp \zeta . 4.13 yields a symp \xi far from each member of T . Also, the same 4.13 implies that the projection of L_{0},L_{1} from \zeta onto \xi is a pair of intersecting lines. Since the projection from \zeta to \xi is an isomorphism of polar spaces, this implies that L_{0} and L_{1} also intersect. The claim follows. We next claim that all members of T are either contained in a plane, or contain a common point. Suppose the latter does not hold. Then there are three lines L_{0},L_{1},L_{2} forming a triangle in a plane. Clearly, all other members of T have to be contained in that plane. The claim is proved. Since we now have that T belongs to a point residual, or the residue of a plane, the theorem follows from 2.27 , 2.28 and 2.29 . ∎ 4.2.5. Geometric lines We now classify geometric lines in the line-Grassmannian of hexagonic Lie incidence geometries. This will follow from the classification of round-up triples of lines. Lemma 4.16 . Let \{L_{1},L_{2},L_{3}\} be a round-up triple of lines in an exceptional hexagonic Lie incidence geometry \Delta of rank at least 3 , such that L_{1} and L_{2} intersect. Then exactly one of the following holds. (i) L_{1}=L_{2}=L_{3} ; (ii) L_{1},L_{2},L_{3} are three lines in a common planar line pencil; (iii) L_{1},L_{2},L_{3} are three lines in a common symp \xi containing a common point p and contained in a common hyperbolic line of \mathsf{Res}_{\xi}(p) . This only happens if \Delta corresponds to a building of type \mathsf{F_{4}} . Proof. Clearly, if L_{1}=L_{2} , then also L_{3}=L_{1} since otherwise there exists a line opposite L_{3} and not opposite L_{1} . So we may assume L_{1}\cap L_{2}=\{x\} . By 2.33 , also x\in L_{3} . By 2.27 , \{L_{1},L_{2},L_{3}\} is a round-up triple in \mathsf{Res}(x) . The result now follows from 2.28 for types \mathsf{E_{6}} and \mathsf{E_{7}} , from 2.29 for type \mathsf{F_{4}} , and from [ 21 , Corollary 5.5] for type \mathsf{E_{8}} . ∎ Lemma 4.17 . Let \{L_{1},L_{2},L_{3}\} be a round-up triple of disjoint lines in an exceptional hexagonic Lie incidence geometry of rank at least 3 . Then no point of L_{2} is collinear to any point of L_{1} . Proof. Let, for a contradiction, M be a line joining a point x_{1}\in L_{1} to a point x_{2}\in L_{2} . Note that L_{1}\neq M\neq L_{2} . 2.34 shows that M intersects L_{3} , say in the point x_{3} . Assume first that M and L_{i} are locally opposite at x_{i} , for every i\in\{1,2,3\} . Let \pi be any plane containing M . Let K_{i} be the line in \pi through x_{i} not locally opposite L_{i} at x_{i} , guaranteed to exist by 2.21 . Suppose first that z:=K_{1}\cap K_{2} does not belong to K_{3} . Let N be a line locally opposite zx_{3} at z . Then any point u\in N\setminus\{z\} is opposite some point of L_{3} , but is not opposite any point of L_{1}\cup L_{2} . It follows that there exists a line through u opposite L_{3} , but not opposite either L_{1} or L_{2} , a contradiction. Hence we may assume that there exists some line K_{3}^{\prime}\subseteq\pi through x_{3} intersecting K_{2} in some point y_{2}\notin K_{1} , with y_{2} special to every point of L_{3}\setminus\{x_{3}\} . Then we pick a line N^{\prime} through y_{2} locally opposite K_{3}^{\prime} at y_{2} , but not locally opposite x_{1}y_{2} at y_{2} . Then no point w on N^{\prime} is opposite some point of L_{1}\cup L_{2} since the pair \{w,x_{1}\} is collinear or symplectic, and the pair \{y_{2},y_{2}^{\prime}\} , with y_{2}^{\prime}\in L_{2}\setminus\{x_{2}\} , is symplectic. As above, there exists a line through w opposite L_{3} , but not opposite either L_{1} or L_{2} . So, we may assume without loss of generality that M and L_{3} are contained in a symp \xi . Then 2.35 implies x_{1}\in L_{2}\subseteq\xi , contrary to our assumptions. The lemma is proved. ∎ Lemma 4.18 . Let \{L_{1},L_{2},L_{3}\} be a round-up triple of lines in an exceptional hexagonic Lie incidence geometry \Delta of rank at least 3 , such that no point of L_{i} is collinear to any point of L_{j} , for all i,j\in\{1,2,3\} , i\neq j . Then no point of L_{1} is symplectic to any point of L_{2} . Proof. Suppose for a contradiction that some point x_{1}\in L_{1} is symplectic to some point x_{2}\in L_{2} . Let \xi be the corresponding symp. By 2.34 , L_{3} shares a point x_{3} with \xi . We claim that L_{3} is collinear to a maximal singular subspace of \xi . Indeed, suppose not. Then L_{3}^{\perp}\cap\xi is a line L_{3}^{*} . There are two cases. (1) Suppose every point of x_{1}^{\perp}\cap x_{2}^{\perp} is collinear to x_{3} . Then x_{3}\in\{x_{1},x_{2}\}^{\perp\perp} and \Delta corresponds to type \mathsf{F_{4}} . Let \xi_{3} be an arbitrary symp not containing L_{3} and locally opposite \xi at x_{3} . Select z_{3}\in\xi_{3}\setminus x_{3}^{\perp} . Since x_{1}\equiv z_{3}\equiv x_{2} , we can define M_{i}:=\mathsf{proj}_{z_{3}}^{x_{i}}(L_{i}) , i=1,2 . If M_{1}\neq M_{2} , we can take a line K through z_{3} locally opposite M_{1} at z_{3} , but not locally opposite M_{2} at z_{3} , and then K is opposite L_{1} , and not opposite either L_{2} or L_{3} (the latter because z_{3} is symplectic to every point of L_{3} ), a contradiction. Hence, we may assume that M_{1}=M_{2} . Set u_{i}:=M_{i}\cap x_{i}^{\Join} , i=1,2 . If u_{1}\neq u_{2} , then we may replace z_{3} with any point in (M_{1}^{\perp}\cap\xi_{3})\setminus\{z_{3}\} and apply the previous argument. So, we may assume u_{1}=u_{2} . Let N_{i} be the line through u_{i} intersecting L_{i} , say in the point w_{i} , i=1,2 . Then, since by 2.16 , x_{1} and w_{2} are not opposite, the same 2.16 implies that N_{1} and N_{2} are not locally opposite at u_{1} . Hence w_{1} and w_{2} are symplectic (as we may assume that they are not collinear by 4.17 ). Hence, by 2.34 , the line L_{3} intersects \xi(w_{1},w_{2}) in a point w_{3} , which we may assume to belong to \{w_{1},w_{2}\}^{\perp\perp} (as otherwise we are in case (2) below). Hence u_{1}\perp w_{3} and so u_{1}=[w_{3},u_{1}] , which, however, is contained in \xi_{3} and coincides with L_{3}^{\perp}\cap\xi_{3}\cap u_{1}^{\perp} . Hence u_{1}\perp x_{3} . But then, similarly, w_{1}=[u_{1},x_{1}]\in\xi , implying w_{1}=x_{1} , clearly a contradiction. (2) Suppose some point y\in x_{1}^{\perp}\cap x_{2}^{\perp} is not collinear to x_{3} . Select y_{3}\in L_{3}\setminus\{x_{3}\} . Then, y_{3}\Join y and u=[y,y_{3}]\in L_{3}^{*}\setminus\{x_{3}\} . Let M be some line through y locally opposite yu at y . Let m be some arbitrary point on M not equal to y . Let L_{3}^{\prime} be the projection of L_{3} onto m . Note that L_{3}^{\prime}\neq M as L_{3}\neq uy_{3} . Hence there exists a line K through m locally opposite L_{3}^{\prime} but not locally opposite M . Then K is opposite L_{3} but not opposite L_{1} and L_{2} , because, by 2.16 , no point of K is opposite x_{1} or x_{2} . Since both cases lead to contradictions, we conclude that L_{3} is collinear to a maximal singular subspace U_{3} of \xi . Likewise, L_{1} and L_{2} are also collinear to respective maximal singular subspaces U_{1} and U_{2} of \xi . Note that this implies that \xi is top-thin (or hyperbolic). It follows that, since x_{1} is not collinear to x_{2}\in U_{2} , the set x_{3}^{\perp}\cap U_{2} contains some point z_{2} that is not collinear to x_{1} . Let y_{3} be an arbitrary point of L_{3}\setminus\{x_{3}\} . If y_{3}\perp\hskip-4.26773pt\perp z_{2} , let \xi_{3} be the symp through y_{3} and z_{2} . If y_{3}\perp z_{2} , then let \xi_{3} be a symp containing L_{3} and z_{2} . Let \xi_{2} be a symp through z_{2} locally opposite \xi but not locally opposite \xi_{3} . Let w_{2}\in\xi_{2} be symplectic to z_{2} . Then, since z_{2} is collinear to each point of L_{2} , 2.16 implies that w_{2} is not opposite any point of L_{2} . Also, since \xi_{2} is not locally opposite \xi_{3} , the point w_{2} is not opposite any point of L_{3} . But w_{2} is opposite x_{1} and so there is a line K through w_{2} opposite L_{1} , and K is not opposite either L_{2} or L_{3} , a contradiction. This completes the proof of the lemma. ∎ Lemma 4.19 . Let \{L_{1},L_{2},L_{3}\} be a round-up triple of lines in an exceptional hexagonic Lie incidence geometry of rank at least 3 . Let x_{1}\in L_{1} and x_{2}\in L_{2} be collinear to a common point y . Then y is collinear to a point of L_{3} . Proof. Suppose y is not collinear to any point of L_{3} . Let \Sigma be an apartment containing L_{3} and y . Since y is not collinear to any point of L_{3} , it is not special to and not opposite at least two points of the line L_{3}^{*} that is opposite L_{3} in \Sigma . But then y is equal, collinear or symplectic with each point of L_{3}^{*} , implying that no point of L_{3}^{*} is opposite either x_{1} or x_{2} . Hence L_{3}^{*} is opposite L_{3} , but not opposite either L_{1} or L_{2} , a contradiction. ∎ Lemma 4.20 . Let \{L_{1},L_{2},L_{3}\} be a round-up triple of lines in an exceptional hexagonic Lie incidence geometry \Delta of rank at least 3 , such that each point of L_{1} is special to each point of L_{2} . Then some point of L_{3} is symplectic, collinear or equal to some point of L_{1}\cup L_{2} . Proof. Using 2.21 , we see that the set of points collinear to a point of L_{1} and to a point of L_{2} is a hyperbolic quadric Q_{12} (of rank 2). Each line of Q_{12} is collinear to a unique point of L_{1}\cup L_{2} . We claim that non-collinear points on Q_{12} are also non-collinear in \Delta . Indeed, one checks that in that case Q_{12} generates a 3 -space U . Picking non-collinear respective points in Q_{12} and L_{1} , we see that they are symplectic and the corresponding symps contain L_{1} and at least a plane of U . The intersection of two such symps (with different planes in U ) contains a 3 -space. Hence the symps coincide. Now the symp through L_{1} and U has a 3 -space in common with the symp through L_{2} and U and hence L_{1} and L_{2} are contained in a common symp, a contradiction. By 4.19 , each point of Q_{12} is also collinear to a unique point of L_{3} (unique indeed since, if not, then 2.21 would yield a point of L_{3} collinear or symplectic to some point of L_{1} , contradicting 4.17 and 4.18 ). It follows that, for any line L in Q_{12} , collinearity defines either a bijection between L and L_{3} , or a constant transformation from L to L_{3} . In the latter case, the unique points of L_{3} and L_{1}\cup L_{2} collinear with all points of L are symplectic, contradicting 4.18 . In the former case, pick p\in L and let L^{\prime} be the unique line of Q_{12} through p distinct from L . Then again, collinearity defines a bijection between L^{\prime} and L_{3} . Hence there is a point of L_{3} collinear to two non-collinear points of Q_{12} , and since these points are also non-collinear in \Delta , this contradicts 2.21 . The lemma is proved. ∎ Lemma 4.21 . Let L_{1},L_{2},L_{3} be three lines of an exceptional hexagonic Lie incidence geometry of rank at least 3 , such that each point of L_{1} is special to or opposite each point of L_{2} . Then \{L_{1},L_{2},L_{3}\} is not a round-up triple. Proof. By Lemmas 4.16 , 4.17 and 4.18 , we may assume that no point of L_{3} is equal, collinear or symplectic to any point of L_{1}\cup L_{2} . Moreover, by 4.20 , we may also assume that some point of L_{3} is opposite L_{1} and some point of L_{3} is opposite some point of L_{2} . This implies that the mutual positions of L_{i} and L_{j} , i,j\in\{1,2,3\} , i\neq j , are given by either (2223) or (2332) . We may assume that \{L_{1},L_{2},L_{3}\} is a round-up triple. By the nature of (2223) and (2332) , there exists at most one point of L_{3} that is not opposite all points of L_{1} , and at most one point of L_{3} that is not opposite all points of L_{2} . Hence we find a point x_{3}\in L_{3} that is opposite at least one point x_{1} of L_{1} and at least one point x_{2} of L_{2} . We can then project L_{i} , i=1,2 , from x_{i} onto x_{3} and obtain lines L_{i}^{\prime} and points y_{i}\in L_{i}^{\prime} collinear to a point of L_{i} . 4.19 and the uniqueness of the projections yield L_{1}^{\prime}=L_{2}^{\prime}=:M_{3} and y_{1}=y_{2}=:y . Let M_{i} , i=1,2 , be the lines through y intersecting L_{i} . Our assumptions imply that these lines are pairwise locally opposite at y . By 2.28 , 2.29 and [ 21 , Corollary 5.6] , \{M_{1},M_{2},M_{3}\} is not a round-up triple in the point residual \mathsf{Res}(y) . Hence, up to renumbering, we find a line M\ni y locally opposite M_{1} , and not locally opposite either M_{2} or M_{3} at y . Pick z\in M\setminus\{y\} . Then 2.16 implies that z is opposite x_{1} , but not opposite any point of L_{2}\cup L_{3} . It follows that each line K through z opposite L_{1} (which exists) is not opposite either L_{2} or L_{3} , a contradiction. This proves the lemma completely. ∎ Proposition 4.22 . Let T be a geometric line of the line-Grassmannian of an exceptional hexagonic Lie incidence geometry \Delta of rank at least 3 . Then exactly one of the following holds. (i) T is an ordinary line of the corresponding line-Grassmannian parapolar space, that is, a planar line pencil of \Delta ; (ii) \Delta is \mathsf{F_{4,4}}(\mathbb{K},\mathbb{K}) and T is a cone over a hyperbolic line in a symplectic symp. Proof. If \Delta is not \mathsf{F_{4,4}}(\mathbb{K},\mathbb{K}) , then 2.32 , together with Lemmas 4.16 , 4.17 , 4.18 , 4.20 and 4.21 imply (i) . Suppose now \Delta\cong\mathsf{F_{4,4}}(\mathbb{K},\mathbb{K}) . If some pair of elements of T is contained in an ordinary line K of the line-Grassmannian of \Delta , then, again by the previous lemmas and the fact that every triple of members of T is a round-up triple, all elements are contained in that line, hence all triples are and 2.32 implies that T=K . Next suppose that two elements L,M of T are not coplanar. Then, again by Lemmas 4.16 , 4.17 , 4.18 , 4.20 and 4.21 , they are contained in a hyperbolic line H of the point residual of a symp \xi . The same lemmas now imply that T\subseteq H and 2.29 (ii) yields T=H and \xi is a symplectic polar space. The proposition is proved. ∎ 5. Generalised hexagons 5.1. Blocking sets We start with a nonexistence result of a class of hexagons with certain parameters. Lemma 5.1 . Let t be a natural number at least 2 . Then there does not exist a generalised hexagon of order (s,t) , with s=t+t^{2} . Proof. Since by [ 18 ] the number st is a perfect square, we have that t^{2}+t^{3} is a perfect square. Hence t+1 is a perfect square, say t=a^{2}-1 . Then s=a^{2}(a^{2}-1) . Now, by [ 18 ] , we know that the rational number \frac{st(1+s+t+st)(1+\sqrt{st}+st)}{2(s+t+\sqrt{st})} is an integer. The denominator of that expression is equal to 2(a-1)(a+1)(a^{2}+a+1)=2t(a^{2}+a+1) . Hence a^{2}+a+1 divides the numerator divided by t . We now observe, taking into account that a^{2}+a+1 is odd, the following facts. • Clearly \mathsf{gcd}(a^{2},a^{2}+a+1)=1 . • Since a^{2}-1=(a^{2}+a+1)-(a+2) and a^{2}+a+1=(a+2)^{2}-3(a+2)+3 , we find \mathsf{gcd}(a^{2}-1,a^{2}+a+1)\in\{1,3\} . • We have 1+s+t+st=a^{2}(a^{4}-a^{2}+1) . Since a^{4}-a^{2}+1=a^{2}(a^{2}+a+1)-a(a^{2}+a+1)-(a^{2}+a+1)+2(a+1), we find \mathsf{gcd}(a^{4}-a^{2}+1,a^{2}+a+1)=\mathsf{gcd}(a+1,a^{2}+a+1)=1 . • We have 1+\sqrt{st}+st=a^{6}-2a^{4}+a^{3}+a^{2}-a+1 . Since a^{6}-2a^{4}+a^{3}+a^{2}-a+1=(a^{4}-a^{3}-2a^{2}+4a-1)(a^{2}+a+1)-4a+2, we find \mathsf{gcd}(1+\sqrt{st}+st,a^{2}+a+1)=\mathsf{gcd}(2a-1,a^{2}+a+1), which, in view of 4a^{2}+4a+4=(2a-1)^{2}+4(2a-1)+7 , is either 1 or 7 . In the latter case, 2a-1 is divisible by 7 , implying in particular a\geq 4 . We conclude that the greatest common divisor of s(1+s+t+st)(1+\sqrt{st}+st) and a^{2}+a+1 is one of 1,3,7,21 . It follows that a^{2}+a+1\in\{3,7,21\} , hence a\in\{2,4\} (remember a>1 ). But then, by the last bullet point above, a=4 . But in this case, one calculates that the number \frac{st(1+s+t+st)(1-\sqrt{st}+st)}{2(s+t-\sqrt{st})} is not an integer (the denominator is divisible by 13 , whereas this is not the case for the numerator), as is required by [ 18 ] . ∎ Every point x of a generalised hexagon has a projection onto a given line L , which is x itself if x\in L , which is the unique point of L collinear to x if x is close to L , and which is special to x if x is far from L . We call this projection occasionally the nearest point to x on L . Also, recall that, if the nearest point to x on L is collinear to but distinct from x , then we called x and L close (as we also did above). Proposition 5.2 . If a set of s+1 points S=\{p_{0},p_{1},\ldots,p_{s}\} of a generalised hexagon \Delta of finite order (s,t) , s,t>1 , admits no opposite point, then it is either (i) a line, or (ii) a hyperbolic line (and then s=t ), or (iii) a regular distance- 3 trace (and then s\geq t ). Proof. Suppose S=\{p_{0},p_{1},\ldots,p_{s}\} is a set of s+1 points in \Delta such that no point of \Delta is opposite every point of S . We proceed with proving some claims. Claim 1. If p_{0}\perp x\perp p_{1} , with p_{0} not collinear to p_{1} , and x\notin S , then every line through x contains at least one member of S . Indeed, suppose the line L through x is disjoint from S . Since p_{0} and p_{1} project onto the same point x of L , there exists some point y\in L not collinear to any member of S . Consider a line M\neq L through y . No point of S is contained in M or is close to M (since this would lead to a 4 -gon or 5 -gon containing y ). Hence they are all far from M . Since p_{0} and p_{1} project onto the same point y of M , there is a point z\in M opposite each member of S , a contradiction. Claim 1 is proved. Claim 2. If p_{0} and p_{1} are collinear, then S is a line of \Delta . Indeed, suppose first that there exists a point x\in L:=p_{0}p_{1} that does not coincide with a projection of some member of S onto L . Consider a line M\neq L containing x . Then the only points of S close to M are on L . Since p_{0} and p_{1} project onto the same point x of M , there exists some point y\in M which is not the projection of any member of S onto M . We deduce that every line K\neq M through y is far from every member of S . Since p_{0} and p_{1} project onto the same point y of such a line K , there exists a point opposite every member of S on each such line K , a contradiction. Hence every point q_{i} on L is the projection of a unique point p_{i} of S . Suppose S\neq L . Then we may assume that p_{2}\notin L and so p_{2} is either special to or opposite p_{0} . If p_{2} is special to p_{0} , then by Claim 1 each line through q_{2} contains a point, say p_{3} , of S , implying that q_{2}=q_{3} , contradicting the uniqueness of p_{2} . So p_{2} is far from L . Select a line M^{\prime} through q_{2} distinct from L and far from p_{2} ( M^{\prime} exists since t\geq 2 ). Then, since q_{2}\neq q_{i} , the point p_{i} is either on L or opposite q_{2} , for each i\in\{3,4,\ldots,s\} . It follows that, with M^{\prime} in the role of M in the previous paragraph, we again reach the same contradiction. Claim 2 is proved. From now on, we may assume that S does not contain two collinear points. Claim 3. If p_{0} and p_{1} are opposite, and some line L close to both contains no point of S , then L is close to each point of S and each point of L is collinear to a unique point of S . Let x_{i} be the nearest point to p_{i} on L , i\in\{0,1,\ldots,s\} , and note that x_{i}\neq p_{i} by assumption. Suppose there exists a point x\in L\setminus\{x_{0},x_{1},\ldots,x_{s}\} . Let M\neq L be any line through x . Then M is far from each point of S . Since x is special to at least two points p_{0},p_{1} of S , there is some point of M opposite each point of S , a contradiction. Hence L=\{x_{0},x_{1},\ldots,x_{s}\} . Suppose some point, say p_{2} , of S is not collinear to its projection x_{2} onto L . Let M_{2} be a line through x_{2} not close to x_{2} and distinct from L (which exists as t>1 ). Then M_{2} is far from each point of S , but p_{0} , p_{1} and p_{2} have the same projection x_{2} , yielding a point y_{2}\in M_{2} opposite each point of S , a contradiction. Hence p_{i}\perp x_{i} , for all i\in\{0,1,\ldots,s\} . Claim 4. If S only contains pairwise opposite points, then s\geq t and S is a regular distance- 3 trace. The assumptions of Step 3 are satisfied for each line close to both p_{0} and p_{1} . Hence every point of S is collinear to some point of each line that is close to both p_{0} and p_{1} . We conclude that S is a regular distance- 3 trace. We now show that s\geq t . Let z be any point special to both p_{0} and p_{1} , and not on a line close to p_{0} and p_{1} . Note that z is not collinear to any point of S , as z\perp p_{2} would imply, by interchanging the roles of p_{1} and p_{2} , that the line through z close to p_{0} is also close to p_{1} , which is not the case by the assumptions on z . Then, similarly as before, every line through z is close to some point of S , but not to two such points, as this would mean that z is already collinear to some point of S (by the definition of distance- 3 trace), contradicting our note above. We conclude t\leq s . Claim 5. If S only contains pairwise special points, then s=t and S is a hyperbolic line. Indeed, set x=[p_{0},p_{1}] . By Claim 1, every line through x contains a point of S . Hence t\leq s . If t=s , then let y be a point opposite x , but not opposite either p_{0} or p_{1} . We claim that p_{i} is not opposite y , for every i\in{2,3,\ldots,s} . Indeed, suppose p_{2} is opposite y and let L_{y} be the unique line through y not opposite xp_{2} . No point p_{i} , i\in\{0,1,\ldots,s\} , is collinear to some point q_{i} of L_{y} , as this would induce a 5 -gon containing x,p_{i},q_{i} and the lines L_{y} and xp_{i} . Hence all points p_{i} have a unique point on L_{y} to which they are not opposite. But p_{0} and p_{1} are not opposite the same point, yielding a point on L_{y} opposite all members of S , a contradiction. The claim is proved. This now implies that S is a hyperbolic line. Suppose now that t1 . Now, by the definition of \mathscr{L} , every member of \mathscr{L} contains a unique member of S . A standard count reveals that |S|=1+s^{\prime}t+(s^{\prime}t)^{2} points. It follows that s=s^{\prime}t+(s^{\prime}t)^{2} . Now, both st and s^{\prime}t are perfect squares by [ 18 ] . It follows that s/t=s^{\prime}(1+s^{\prime}t) is a perfect square. Since s^{\prime} and 1+s^{\prime}t are relatively prime, both s^{\prime} and 1+s^{\prime}t are perfect squares, which contradicts the fact that s^{\prime}t is a perfect square. This completes the proof of Claim 5. This also completes the proof of the proposition. ∎ The following result classifies very explicitly all sets of size s+1 admitting no global opposite point in finite Moufang hexagons of order (s,t) . Corollary 5.3 . A set S of s+1 points of a Moufang generalised hexagon of finite order (s,t) , s,t>1 , admits no opposite point if, and only if, it is either (i) a line, or (ii) a hyperbolic line in the split Cayley hexagon \mathsf{G_{2,2}}(s,s) , or (iii) a distance- 3 trace in the split Cayley hexagon \mathsf{G_{2,2}}(s,s) with s even, or in the twisted triality hexagon \mathsf{G_{2,2}}(t,s) , with t^{3}=s even. Proof. In view of 5.2 , (ii) follows from [ 32 , Remark 6.3.5] and (iii) follows from [ 19 , Theorem 1] . ∎ Note that the previous corollary implies that not every regular distance- 3 trace is a set of s+1 points such that no point is opposite each point of that set. Indeed, the split Cayley hexagons and the twisted triality hexagons in odd characteristic are counterexamples. Remark 5.4 . In Claim 4 of the proof of 5.2 , hexagons of order (t^{2}+t,t) appear as possible counterexamples, but, as we assume thickness, they are killed by 5.1 . If we drop the thickness assumption, it is curious to note that the arguments of that step give rise to a rather exceptional example of a set F of s+1 point-line flags in a projective plane of order s such that no point-line flag is opposite all members of S . Indeed, putting t=1 , we obtain a hexagon of order (2,1) , which arises from \mathsf{PG}(2,2) , and S consists of three flags \{p_{0},p_{0}p_{1}\},\{p_{1},p_{1}p_{2}\},\{p_{2},p_{0}p_{2}\} from a triangle \{p_{0},p_{1},p_{2}\} . The points of these flags do not form a line, and the lines of these flags do not form a line pencil. We conjecture that this is the only example of size s+1 in any projective plane of order s with that property and such that no flag of the plane is opposite all of its members. 5.2. Geometric lines We now classify geometric lines in Moufang hexagons. We first consider the general case and then specify further. As usual, we deal with round-up triples. Lemma 5.5 . Let \Gamma=(X,\mathscr{L}) be a generalised hexagon and \{x_{1},x_{2},x_{3}\} a round-up triple of points. Suppose x_{1}\perp x_{2} . Then x_{1},x_{2},x_{3} are contained in a common line. Proof. Let L be the line containing x_{1} and x_{2} . If x_{3}\notin L , then some point u of L is special to x_{3} . Let z be a point collinear to u but special to [x_{3},u] . Then z\equiv x_{3} , whereas x_{1}\not\equiv z\not\equiv x_{2} , a contradiction. ∎ Lemma 5.6 . Let \Gamma=(X,\mathscr{L}) be a generalised hexagon and \{x_{1},x_{2},x_{3}\} a round-up triple of points. Suppose x_{1}\Join x_{2} . Then x_{1},x_{2},x_{3} are contained in [x_{1},x_{2}]^{\perp}\cap y^{\Join} , for every point y\in[x_{1},x_{2}]^{\equiv}\cap x_{1}^{\Join}\cap x_{2}^{\Join} . Proof. By 5.5 we have x_{3}\neq[x_{1},x_{2}] . If x_{3}\perp[x_{1},x_{2}] , then the assertion follows directly from the definition of a round-up triple. If x_{3}\Join[x_{1},x_{2}] , then any point z\in[x_{1},x_{2}]^{\perp}\setminus\langle[x_{1},x_{2}],[[x_{1},x_{2}],x_{3}]\rangle is opposite x_{3} , but not opposite either x_{1} or x_{2} , a contradiction. Finally, if x_{3}\equiv[x_{1},x_{2}] , then [x_{1},x_{2}] obviously violates the defining property of \{x_{1},x_{2},x_{3}\} being a round-up triple. ∎ Lemma 5.7 . Let \Gamma=(X,\mathscr{L}) be a generalised hexagon and \{x_{1},x_{2},x_{3}\} a round-up triple of points. Suppose x_{1}\equiv x_{2} . Then x_{1},x_{2},x_{3} are contained in every distance- 3 trace containing at least two of them. Proof. Let L be an arbitrary line of \Gamma containing a point x_{1}^{\prime} collinear to x_{1} and also a point x_{2}^{\prime} collinear to x_{2} . By 5.6 , x_{3}\notin L . Assume x_{3} is special to some point y\in L with [x_{3},y]\notin L . Then any point of L\setminus\{y\} is opposite x_{3} and not opposite both x_{1},x_{2} . So x_{3} is collinear to some point of L , distinct from both x_{1}^{\prime} and x_{2}^{\prime} (use 5.6 again). Now it is clear that the assertion follows. ∎ Proposition 5.8 . Let \Gamma=(X,\mathscr{L}) be a generalised hexagon and let T be a geometric line of \Gamma . Then T is either a line, a hyperbolic line, or a regular distance- 3 trace. Proof. Since every triple of points of a geometric line is a round-up triple, the previous three lemmas imply that T is contained in either a line, or a hyperbolic line, or a distance- 3 trace. But if T were not equal to one of these objects, then, in each case, it is easy to find a point opposite every member of T , a contradiction. ∎ We can now prove B for type \mathsf{G_{2}} . Proposition 5.9 . Let \Gamma=(X,\mathscr{L}) be a Moufang generalised hexagon and let T be a geometric line. Then T is either (1) an ordinary line, or (2) a hyperbolic line in a split Cayley hexagon, or (3) a distance- 3 trace in a split Cayley hexagon over a perfect field in characteristic 2 . Proof. By 5.8 there are three possibilities for T . The first one is a line, which leads to (1) . The second one is a hyperbolic line. Let T be collinear to the unique point c . Since T is a geometric line, every hyperbolic line in c^{\perp} intersects T in exactly one point. By transitivity of the automorphism group on paths x_{1}\perp x_{2}\perp x_{3} , with x_{1}\Join x_{3} , which follows readily from the Moufang condition, we see that every pair of hyperbolic lines in c^{\perp} intersects. Then [ 32 , Corollary 5.14] implies that \Gamma is a split Cayley hexagon. The third possibility is that T is a regular distance- 3 trace. Let x,y\in T . Then we obviously can write T=\{x,y\}^{\not\equiv\not\equiv} . Let L\in\mathscr{L} be arbitrary but such that it contains unique points x^{\prime} and y^{\prime} special to x and y , respectively, with x^{\prime}\neq y^{\prime} . That at least one such line exists is easily seen. Now every point of L is special to precisely one point of T , since T is a geometric line. This means that, in the terminology of [ 32 , Definitions 6.5.5] , the set T is a long imaginary line, and [ 32 , Theorem 6.5.6] now implies that \Gamma is a split Cayley hexagon over a perfect field in characteristic 2 . ∎ Remark 5.10 . For every natural number n\geq 5 , there exists an (obvious) analogue of 5.8 for the class of (thick) generalised n -gons. This requires defining a “regular” distance- i trace, 2\leq i\leq\frac{n}{2} , similarly to a hyperbolic line (which would be a regular distance- 2 trace) and a regular distance 3 -trace for a generalised hexagon. Proofs are straightforward generalisations of the above proofs for hexagons. Restricting to Moufang octagons (the only class of exceptional Moufang buildings not yet considered in this paper), one obtains that, using the results in [ 2 ] (see also [ 32 , Section 6.5] ), the only geometric lines in Moufang octagons are the ordinary lines. However, the classification of minimal blocking sets in finite Moufang octagons is still open; however, see also Remark 5.11 . Remark 5.11 . S. Petit and G. Van de Voorde [ 24 , Theorem 6] prove that, if s\leq t , then every blocking set of s+1 points in a finite generalised polygon of order (s,t) is either a line or a regular distance- i trace. Together with Remark 5.10 , this leads to a classification of blocking sets of size s+1 in the Moufang octagons of order (s,s^{2}) : only lines occur. The case of Moufang octagons of order (s,\sqrt{s}) is hence the only open case for finite Moufang polygons. Note that 5.2 extends [ 24 , Theorem 6] for generalised hexagons to arbitrary order. 6. An application Proposition 6.1 . Let \Delta and \Delta^{\prime} be two buildings of the same exceptional type \mathsf{F_{4},E_{6},E_{7}} or \mathsf{E_{8}} . Let, with Bourbaki labelling, T_{i} and T_{i}^{\prime} be the set of vertices of type i of \Delta and \Delta^{\prime} , respectively, where \bullet i\in\{1,2,3,4\} is arbitrary if \Delta has type \mathsf{F_{4}} ; \bullet i\in\{1,2,3,4,5,6\} is arbitrary if \Delta has type \mathsf{E_{6}} ; \bullet i\in\{1,2,6,7\} if \Delta has type \mathsf{E_{7}} ; \bullet i\in\{7,8\} if \Delta has type \mathsf{E_{8}} . Then any surjective map \varphi\colon T_{i}\to T_{i}^{\prime} preserving opposition and non-opposition is induced by an isomorphism of buildings. Proof. It suffices to show that \varphi is a bijective collineation between the corresponding i -Grassmannian geometries. We first show that \varphi is bijective. Suppose, for a contradiction, that two vertices v,u\in T_{i} are mapped onto the same vertex. 2.31 yields a vertex w\in T_{i} opposite v but not opposite u . Then our assumptions imply \varphi(v)\equiv\varphi(w)\not\equiv\varphi(u)=\varphi(v) , a contradiction. Hence \varphi is a bijection. Since opposition and non-opposition are preserved, one deduces that geometric lines are mapped onto geometric lines. If the only geometric lines are the ordinary lines, then this concludes the proof of the proposition. By [ 21 , Corollary 6.6] and 4.22 , we may assume i=3 and \Delta has type \mathsf{F_{4}} . It suffices to recognise the planar line pencils of \mathsf{F_{4,4}}(\mathbb{K},\mathbb{K}) among all geometric lines of \mathsf{F_{4,3}}(\mathbb{K},\mathbb{K}) . Let \Gamma be the point-line geometry with point set the points of \mathsf{F_{4,3}}(\mathbb{K},\mathbb{K}) and line set the set of ordinary and geometric lines of \mathsf{F_{4,3}}(\mathbb{K},\mathbb{K}) . We claim that no geometric line different from an ordinary line of \mathsf{F_{4,3}}(\mathbb{K},\mathbb{K}) is contained in a maximal subspace of \Gamma isomorphic to a projective plane. Suppose, for a contradiction, that the geometric line Z is contained in a maximal singular subspace \alpha of \Gamma isomorphic to a projective plane, and that Z is not an ordinary line of \mathsf{F_{4,3}}(\mathbb{K},\mathbb{K}) . We argue in \mathsf{F_{4,4}}(\mathbb{K},\mathbb{K}) , where Z is a cone in a symp with vertex p over a hyperbolic line h . Suppose first that \alpha does not contain any ordinary line of \mathsf{F_{4,3}}(\mathbb{K},\mathbb{K}) . So, the point set of \alpha corresponds to a set \Pi of lines through p , and the cones over hyperbolic lines correspond to the lines of \alpha . Consequently, any two points on distinct lines of \Pi are symplectic, and the unique hyperbolic line through them is contained in the union of all lines of \Pi . We select a point q opposite p . Then every line L\in\Pi contains a unique point p_{L} special to q . We claim that the set \beta=\{p_{L}\mid L\in\Pi\} , endowed with the hyperbolic lines contained in it, is a projective plane. Indeed, in view of the fact that \Pi is the point set of a projective plane whose lines are geometric lines of \mathsf{F_{4,3}}(\mathbb{K},\mathbb{K}) , it suffices to prove that \beta is closed under taking hyperbolic lines through two arbitrary distinct points y_{1} and y_{2} of \beta . Since p\in\xi(y_{1},y_{2}) , q is far from \xi(y_{1},y_{2}) . Since y_{1} and y_{2} are collinear to the unique point q^{\prime} of \xi(y_{1},y_{2}) symplectic to q , all points of the hyperbolic line h(y_{1},y_{2}) defined by y_{1},y_{2} are collinear to q^{\prime} , by the very definition of hyperbolic line. The claim is proved. Now [ 17 , Lemma 5.21] implies that \beta is contained in an extended equator geometry \widehat{E} . Then [ 17 , Proposition 5.24] implies that p is collinear to a set \gamma of points of \widehat{E} that forms a 3 -dimensional projective space when endowed with the hyperbolic lines it contains. Hence the line set \{px\mid x\in\gamma\} forms a projective 3 -space in \Gamma . So, \alpha is not maximal, a contradiction. Consequently, we may suppose that \alpha contains at least one ordinary line of \mathsf{F_{4,3}}(\mathbb{K},\mathbb{K}) . Then we have a plane \pi of \mathsf{F_{4,4}}(\mathbb{K},\mathbb{K}) through p intersecting h in some point x . Select y\in h\setminus\{x\} and z\in\pi\setminus px . Let \xi be the symp containing h ; we have p\in\xi . Suppose, for a contradiction, that z is not contained in \xi . Since z\perp px and y\notin px , we deduce z\perp\hskip-4.26773pt\perp y . Hence y\perp px , a contradiction. Consequently z\in\xi and so \pi\subseteq\xi . Now the set of lines of \xi through p forms a 3 -dimensional projective space of \Gamma , contradicting the maximality of \alpha . Hence Z is not contained in a maximal singular subspace of \Gamma isomorphic to a projective plane. Evidently, any line of \mathsf{F_{4,4}}(\mathbb{K},\mathbb{K}) is contained in an ordinary projective plane of \mathsf{F_{4,4}}(\mathbb{K},\mathbb{K}) , which gives rise to a maximal singular subspace of \Gamma of dimension 2 . Hence we can recognise the ordinary lines of \mathsf{F_{4,3}}(\mathbb{K},\mathbb{K}) , and the proof of the proposition is complete. ∎ The following is an immediate consequence. Corollary 6.2 . Let \Delta be a finite building of exceptional type \mathsf{F_{4},E_{6},E_{7}} or \mathsf{E_{8}} . Let, with Bourbaki labelling, T_{i} be the set of vertices of type i of \Delta , where \bullet i\in\{1,2,3,4\} is arbitrary if \Delta has type \mathsf{F_{4}} ; \bullet i\in\{1,2,3,4,5,6\} is arbitrary if \Delta has type \mathsf{E_{6}} ; \bullet i\in\{1,2,6,7\} if \Delta has type \mathsf{E_{7}} ; \bullet i\in\{7,8\} if \Delta has type \mathsf{E_{8}} . Then any map \varphi:T_{i}\to T_{i}^{\prime} preserving opposition and non-opposition is induced by an automorphism of \Delta . Proof. We only have to establish the surjectivity of \varphi in order to be able to apply 6.1 . Therefore, we note that the injectivity of \varphi is proved in a completely similar way as in the first paragraph of the proof of 6.1 . Now the assertion follows from the trivial fact that an injective transformation of a finite set is always surjective. ∎