Another Simple Proof for the

Existence of the Small Witt Design

Hans Havlicek∗ Hanfried Lenz
arXiv:1304.0096v1 [math.CO] 30 Mar 2013

Abstract
We give a short proof for the existence of the small Witt design which is based on the
projective plane of order three with one point deleted.

1 Introduction
The Swiss geometer J. Steiner posed the following question (“Combinatorische Aufgabe”)
in 1853:
“Welche Zahl, N , von Elementen hat die Eigenschaft, dass sich die Elemente so zu dreien
ordnen lassen, dass je zwei in einer, aber nur in einer Verbindung vorkommen?”
If we write v, k, and t instead of N , 3, and 2, respectively, then we arrive at the following
contemporary definition: A Steiner system S(t, k, v) is a finite set V of elements (called
points) with a distinguished family of subsets (called blocks) such that the following holds
true:

1. There are exactly v points in V.

2. Each block has exactly k elements.

3. Any t distinct points belong to a unique block.

In order to avoid trivialities it is usually assumed that 2 ≤ t < k < v.

So Steiner asked for S(2, 3, v) systems. As a matter of fact, T.P. Kirkman proved already in
1847 that an S(2, 3, v) exists if, and only if, v ≡ 1, 3 (mod 6).
In this short communication we present another proof for the existence of a Steiner system
S(5, 6, 12) which is also called small Witt design W12 . See [1] or [2], in particular Chapter
IV. There the reader will also find the definition of a t-design (which is more general than
that of a Steiner system) and references on other results mentioned in this section.
In an S(5, 6, 12) there are twelve points, each block has exactly six elements, and any five
distinct points are contained in a unique block. There is a unique S(5, 6, 12) up to isomor-
phism. The same uniqueness property holds true for an S(5, 8, 24) which carries the name
∗
Partially supported by the City of Vienna (Hochschuljubiläumsstiftung der Stadt Wien, Projekt H-39/98.)

1
large Witt design W24 . The Steiner systems W12 and W24 are due to E. Witt (1938) and
R.D. Carmichel (1937). For many decades W12 and W24 were the only known Steiner systems
with parameter t = 5. Even today only finitely many Steiner systems S(t, k, v) with t > 3
and none with t > 5 seem to be known [3, 67], [7].
Another remarkable property of the two Witt designs concerns their automorphism group.
Recall that a group G of permutations acts (sharply) t-transitively, if for two ordered t-tuples
of elements there is a (unique) permutation in G taking the first to the second t-tuple. The
automorphism groups of the Witt designs W12 and W24 act 5-transitively on their sets of
points; for the small Witt design the action is even sharply 5-transitive. These automor-
phism groups are the Mathieu groups M12 and M24 , respectively. They were discovered by
E. Mathieu in 1861 and 1873, and they are early examples of sporadic finite simple groups.
The only finite t-transitive permutation groups with t > 3 other than symmetric and al-
ternating groups that seem to be known are the two Mathieu groups mentioned above and
two of their subgroups (the Mathieu groups M11 and M23 ). So the Witt designs are indeed
remarkable combinatorial structures.
The starting point of our construction of W12 is the projective plane of order three with
point set P. It is a Steiner system S(2, 4, 13), but its blocks are called lines.
The first step is to discuss the 6-sets of points in P. They fall into four classes which can be
described in various ways, but the crucial observation is that two 6-sets are in the same class
if and only if they have the same number of trisecants (i.e. lines meeting the set in exactly
three points).
Next we choose one point of P, say U . The twelve points of W := P \ {U } will be the points
of the Witt design W12 . We introduce three kinds of 6-subsets of W and call them blocks.
Each block together with the distinguished point U has a complement in P with exactly six
elements. So properties of 6-sets in P carry over to properties of blocks.
Finally, we show that W, together with the set of all blocks, is a Steiner system S(5, 6, 12).
Again, the results on 6-sets of points turn out useful when showing that any 5-set M ⊂ W
is contained in a block, since M ∪ {U } is a 6-set of points in the projective plane.
The proof presented in this paper is closely related to a projective representation, in the five-
dimensional projective space of order three, of the small Witt design due to H.S.M. Coxeter
[4]; see [5] and the references given there. Furthermore, we refer to [6] for an alternative
description of the present construction of W12 using completely different methods.

2 Construction
Let P be the set of points of the projective plane of order three or, in other words, the
Steiner system S(2, 4, 13) [1, 19]. There are exactly 4 lines (blocks) through each point of
P. The unique line joining distinct points A and B will be written as AB.
First we introduce four types of sets S ⊂ P, each consisting of exactly six points.

1. S is the union of a line and two further points (fig. 1).

2. S is the symmetric difference of two different lines (fig. 2).

2
 3. S consists of a triangle and an inscribed triangle, i.e. each point of the second triangle
lies on exactly one line of the first triangle (fig. 3).

4. S is the set of vertices of a quadrilateral, i.e. the set of points where two distinct lines
of the quadrilateral meet (fig. 4).

Fig. 1. Fig. 2. Fig. 3. Fig. 4.

A set of type 1 contains a unique line. So there are exactly 13 · 92 = 13 · 36 sets of type 1.

A set S of type 2 can be written as symmetric difference of two lines in one way only. Hence
there are exactly 13·12
2!
= 13 · 6 sets of type 2.
If S is of type 3 then each vertex of the “basic” triangle is on exactly two trisecants of S,
whereas each point of the “inscribed” triangle is on one trisecant only. So the role of the
two triangles is not the same. Since two distinct vertices of the inscribed triangle determine
the remaining one uniquely, the number of 6-sets of type 3 is 13·12·9
3!
· 2 · 2 = 13 · 72.
If S is of type 4 then the defining quadrilateral can be recovered from S as the set its four
trisecants. So the number of sets of type 4 equals 13·12·9·4
4!
= 13 · 18.
We observe that a 6-set of type i ∈ {1, 2, 3, 4} has exactly i trisecants. So the four types of
6-sets do not overlap. Finally, from
 
13
13 · (36 + 6 + 72 + 18) = 13 · 132 = ,
6

our list from above comprises all 6-sets of points.

Let U ∈ P be a fixed point and put W := P \ {U }. A block, say B, is defined to be a subset
of W satisfying one of the following conditions:

A. B is the symmetric difference of two distinct lines, each not incident with U .

B. B ∪ {U } is the union of two distinct lines.

C. B consists of a quadrangle together with two of its diagonal points; moreover, U is the
remaining diagonal point.

3
If a block B is of type A, B, or C, then P \ (B ∪ {U }) is easily seen to be a 6-set of type 1, 2
or 4, respectively. Thus the blocks fall into classes A, B, and C. Also, let us remark that the
complement in W of a block of type A, B, or C is a block of type B, A, or C, respectively.
The number of blocks of type A is equal to the number of 2-sets of lines, both not running
through U . So it is 9·8
2
= 36.
Blocks of type B are of the form (a ∪ b) \ {U } with lines a 6= b and U ∈ a ∪ b. Counting the
possibilities for a and b, and taking

into account whether U is on both lines or not, shows
4
that there are precisely 4 · 9 + 2 = 42 blocks of type B.
We obtain all quadrangles with diagonal point U by drawing two distinct lines, say a and
b, through U and choosing two distinct
points
on a \ {U } and b \ {U }, respectively. So the
number of blocks of type C equals 42 · 32 · 32 = 54.

Summing up shows that there are exactly 132 blocks.

Here is our main result:

Theorem 1 The set W, together with the set of all blocks, is a Steiner system S(5, 6, 12).

Proof: (a) By definition, all blocks have exactly 6 elements and #W = 12.
(b) We show that each 5-set M in W belongs to at least one block. There are four cases,
depending on the type of the 6-set S := M ∪ {U }:

B C B C=U B C

Q Q X X
P P E

A X A A D

R=U R F =U

Fig. 5. Fig. 6. Fig. 7.

1. Suppose that S consists of a line a and two further points; let b be the line joining
those points. Then (a ∪ b) \ {U } is a block of type B containing M.

2. Let S be the symmetric difference of distinct lines a and b. Then (a ∪ b) \ {U } is a

block of type B with the required property.

3. Let S be the union of a triangle {A, B, C} and an inscribed triangle {P, Q, R} such
that P ∈ BC, Q ∈ CA, and R ∈ AB. There are two subcases:
If U ∈ {P, Q, R}, say R = U , then put {X} := AP ∩ BQ. Then {X} ∈ CR and
{A, B, C, X} is a quadrangle with diagonal points P , Q, and R = U which gives rise
to a block of type C containing M (fig. 5).

4
 If U ∈ {A, B, C}, say C = U , then put {X} := P Q ∩ RU . Then U ∈ / AB ∪ P Q. So
the symmetric difference of AB and P Q is a block of type A through M (fig. 6).

4. Let S = {A, B, C, D, E, F } be the set of vertices of a quadrangle. W.l.o.g. let {U } =

{F } = AB ∩ CD. So {A, B, C, D} is a quadrangle with diagonal points E, F = U ,
and X, say. Therefore {A, B, C, D, E, X} ⊃ M is a block of type C (fig. 7).

(c) Given a 5-set M ⊂ W then denote by r(M) the number of blocks passing through it.
Since each of the 132 blocks contains exactly 6 subsets of W with 5 elements, we obtain from
the principle of counting in two ways that
X
r(M) = 132 · 6 = 792.
M⊂W,
#M=5

From (b), r(M) ≥ 1 for each of the 12

5
= 792 sets M appearing in the sum above. So
r(M) = 1 is constant. This completes the proof. 

References
[1] Th. Beth, D. Jungnickel, and H. Lenz, Design Theory, Mannheim Wien Zürich, BI
Wissenschaftsverlag 1985.

[2] Th. Beth, D. Jungnickel, and H. Lenz, Design Theory, 2nd ed., Cambridge, Cambridge
University Press 1999.

[3] C.J. Colbourn and R. Mathon, Steiner Systems, in C.J. Colbourn and J.H. Dinitz (eds.),
The CRC Handbook of Combinatorial Designs, Boca Raton, CRC Press 1996.

[4] H.S.M. Coxeter, Twelve points in PG(5, 3) with 95040 self-transformations, Proc. Royal
Soc. London A 427 (1958), 279–293.

[5] H. Havlicek, Giuseppe Veronese and Ernst Witt — Neighbours in PG(5,3), Aequationes
Math. 58 (1999), 85–92.

[6] H. Havlicek, A Model of the Witt Design W12 Based on Quadrics of PG(2,3), Discrete
Math., submitted.

[7] R. Mathon, Searching for Spreads and Packings, in J.W.P. Hirschfeld, S.S. Magliveras,
and M.J. de Resimini (eds.), Geometry, Combinatorial Designs and Related Structures,
London Math. Soc. Lect. Notes Ser. 245, Cambridge, Cambridge University Press, 1997.

Hans Havlicek, Institut für Geometrie, Technische Universität, Wiedner Hauptstraße 8–10, A–1040
Wien, Austria.

Hanfried Lenz, Mathematisches Institut II (WE2), Freie Universität Berlin, Arnimallee 3, D-14195
Berlin, Germany.