Translation plane

From HandWiki
Short description: Type of projective plane

In mathematics, a translation plane is a projective plane which admits a certain group of symmetries (described below). Along with the Hughes planes and the Figueroa planes, translation planes are among the most well-studied of the known non-Desarguesian planes, and the vast majority of known non-Desarguesian planes are either translation planes, or can be obtained from a translation plane via successive iterations of dualization and/or derivation.[1]

In a projective plane, let P represent a point, and l represent a line. A central collineation with center P and axis l is a collineation fixing every point on l and every line through P. It is called an elation if P is on l, otherwise it is called a homology. The central collineations with center P and axis l form a group.[2] A line l in a projective plane Π is a translation line if the group of all elations with axis l acts transitively on the points of the affine plane obtained by removing l from the plane Π, Πl (the affine derivative of Π). A projective plane with a translation line is called a translation plane.

The affine plane obtained by removing the translation line is called an affine translation plane. While it is often easier to work with projective planes, in this context several authors use the term translation plane to mean affine translation plane.[3][4]

Algebraic construction with coordinates

Every projective plane can be coordinatized by at least one planar ternary ring.[5] For translation planes, it is always possible to coordinatize with a quasifield.[6] However, some quasifields satisfy additional algebraic properties, and the corresponding planar ternary rings coordinatize translation planes which admit additional symmetries. Some of these special classes are:

  • Nearfield planes - coordinatized by nearfields.
  • Semifield planes - coordinatized by semifields, semifield planes have the property that their dual is also a translation plane.
  • Moufang planes - coordinatized by alternative division rings, Moufang planes are exactly those translation planes that have at least two translation lines. Every finite Moufang plane is Desarguesian and every Desarguesian plane is a Moufang plane, but there are infinite Moufang planes that are not Desarguesian (such as the Cayley plane).

Given a quasifield with operations + (addition) and [math]\displaystyle{ \cdot }[/math] (multiplication), one can define a planar ternary ring to create coordinates for a translation plane. However, it is more typical to create an affine plane directly from the quasifield by defining the points as pairs [math]\displaystyle{ (a,b) }[/math] where [math]\displaystyle{ a }[/math] and [math]\displaystyle{ b }[/math] are elements of the quasifield, and the lines are the sets of points [math]\displaystyle{ (x,y) }[/math] satisfying an equation of the form [math]\displaystyle{ y = m \cdot x + b }[/math] , as [math]\displaystyle{ m }[/math] and [math]\displaystyle{ b }[/math] vary over the elements of the quasifield, together with the sets of points [math]\displaystyle{ (x,y) }[/math] satisfying an equation of the form [math]\displaystyle{ x=a }[/math] , as [math]\displaystyle{ a }[/math] varies over the elements of the quasifield.[7]

Geometric construction with spreads (Bruck/Bose)

Translation planes are related to spreads of odd-dimensional projective spaces by the Bruck-Bose construction.[8] A spread of PG(2n+1, K), where [math]\displaystyle{ n \geq 1 }[/math] is an integer and K a division ring, is a partition of the space into pairwise disjoint n-dimensional subspaces. In the finite case, a spread of PG(2n+1, q) is a set of qn+1 + 1 n-dimensional subspaces, with no two intersecting.

Given a spread S of PG(2n +1, K), the Bruck-Bose construction produces a translation plane as follows: Embed PG(2n+1, K) as a hyperplane [math]\displaystyle{ \Sigma }[/math] of PG(2n+2, K). Define an incidence structure A(S) with "points," the points of PG(2n+2, K) not on [math]\displaystyle{ \Sigma }[/math] and "lines" the (n+1)-dimensional subspaces of PG(2n+2, K) meeting [math]\displaystyle{ \Sigma }[/math] in an element of S. Then A(S) is an affine translation plane. In the finite case, this procedure produces a translation plane of order qn+1.

The converse of this statement is almost always true.[9] Any translation plane which is coordinatized by a quasifield that is finite-dimensional over its kernel K (K is necessarily a division ring) can be generated from a spread of PG(2n+1, K) using the Bruck-Bose construction, where (n+1) is the dimension of the quasifield, considered as a module over its kernel. An instant corollary of this result is that every finite translation plane can be obtained from this construction.

Algebraic construction with spreads (André)

André[10] gave an earlier algebraic representation of (affine) translation planes that is fundamentally the same as Bruck/Bose. Let V be a 2n-dimensional vector space over a field F. A spread of V is a set S of n-dimensional subspaces of V that partition the non-zero vectors of V. The members of S are called the components of the spread and if Vi and Vj are distinct components then ViVj = V. Let A be the incidence structure whose points are the vectors of V and whose lines are the cosets of components, that is, sets of the form v + U where v is a vector of V and U is a component of the spread S. Then:[11]

A is an affine plane and the group of translations xx + w for w in V is an automorphism group acting regularly on the points of this plane.

The finite case

Let F = GF(q) = Fq, the finite field of order q and V the 2n-dimensional vector space over F represented as:

[math]\displaystyle{ V = \{ (x,y) \colon x, y \in F^n \}. }[/math]

Let M0, M1, ..., Mqn - 1 be n × n matrices over F with the property that MiMj is nonsingular whenever ij. For i = 0, 1, ...,qn – 1 define,

[math]\displaystyle{ V_i = \{(x, xM_i) \colon x \in F^n \}, }[/math]

usually referred to as the subspaces "y = xMi". Also define:

[math]\displaystyle{ V_{q^n} = \{ (0,y) \colon y \in F^n \}, }[/math]

the subspace "x = 0".

The set {V0, V1, ..., Vqn} is a spread of V.

The set of matrices Mi used in this construction is called a spread set, and this set of matrices can be used directly in the projective space [math]\displaystyle{ PG(2n-1,q) }[/math] to create a spread in the geometric sense.

Reguli and regular spreads

Let [math]\displaystyle{ \Sigma }[/math] be the projective space PG(2n+1, K) for [math]\displaystyle{ n \geq 1 }[/math] an integer, and K a division ring. A regulus[12] R in [math]\displaystyle{ \Sigma }[/math] is a collection of pairwise disjoint n-dimensional subspaces with the following properties:

  1. R contains at least 3 elements
  2. Every line meeting three elements of R, called a transversal, meets every element of R
  3. Every point of a transversal to R lies on some element of R

Any three pairwise disjoint n-dimensional subspaces in [math]\displaystyle{ \Sigma }[/math] lie in a unique regulus.[13] A spread S of [math]\displaystyle{ \Sigma }[/math] is regular if for any three distinct n-dimensional subspaces of S, all the members of the unique regulus determined by them are contained in S. For any division ring K with more than 2 elements, if a spread S of PG(2n+1, K) is regular, then the translation plane created by that spread via the André/Bruck-Bose construction is a Moufang plane. A slightly weaker converse holds: if a translation plane is Pappian, then it can be generated via the André/Bruck-Bose construction from a regular spread.[14]

In the finite case, K must be a field of order [math]\displaystyle{ q \gt 2 }[/math], and the classes of Moufang, Desarguesian and Pappian planes are all identical, so this theorem can be refined to state that a spread S of PG(2n+1, q) is regular if and only if the translation plane created by that spread via the André/Bruck-Bose construction is Desarguesian.

In the case where K is the field [math]\displaystyle{ GF(2) }[/math], all spreads of PG(2n+1, 2) are trivially regular, since a regulus only contains three elements. While the only translation plane of order 8 is Desarguesian, there are known to be non-Desarguesian translation planes of order 2e for every integer [math]\displaystyle{ e \geq 4 }[/math].[15]

Families of non-Desarguesian translation planes

  • Hall planes - constructed via Bruck/Bose from a regular spread of [math]\displaystyle{ PG(3,q) }[/math] where one regulus has been replaced by the set of transversal lines to that regulus (called the opposite regulus).
  • Subregular planes - constructed via Bruck/Bose from a regular spread of [math]\displaystyle{ PG(3,q) }[/math] where a set of pairwise disjoint reguli have been replaced by their opposite reguli.
  • André planes
  • Nearfield planes
  • Semifield planes

Finite translation planes of small order

It is well known that the only projective planes of order 8 or less are Desarguesian, and there are no known non-Desarguesian planes of prime order.[16] Finite translation planes must have prime power order. There are four projective planes of order 9, of which two are translation planes: the Desarguesian plane, and the Hall plane. The following table details the current state of knowledge:

Order Number of Non-Desarguesian

Translation Planes

9 1
16 7[17][18]
25 20[19][20][21]
27 6[22][23]
32 ≥8[24]
49 1346[25][26]
64 ≥2833[27]

Notes

  1. Eric Moorhouse has performed extensive computer searches to find projective planes. For order 25, Moorhouse has found 193 projective planes, 180 of which can be obtained from a translation plane by iterated derivation and/or dualization. For order 49, the known 1349 translation planes give rise to more than 309,000 planes obtainable from this procedure.
  2. Geometry Translation Plane Retrieved on June 13, 2007
  3. Hughes & Piper 1973, p. 100
  4. Johnson, Jha & Biliotti 2007, p. 5
  5. Hall 1943
  6. There are many ways to coordinatize a translation plane which do not yield a quasifield, since the planar ternary ring depends on the quadrangle on which one chooses to base the coordinates. However, for translation planes there is always some coordinatization which yields a quasifield.
  7. Dembowski 1968, p. 128. Note that quasifields are technically either left or right quasifields, depending on whether multiplication distributes from the left or from the right (semifields satisfy both distributive laws). The definition of a quasifield in Wikipedia is a left quasifield, while Dembowski uses right quasifields. Generally this distinction is elided, since using a chirally "wrong" quasifield simply produces the dual of the translation plane.
  8. Bruck & Bose 1964
  9. Bruck & Bose 1964, p. 97
  10. André 1954
  11. Moorhouse 2007, p. 13
  12. This notion generalizes that of a classical regulus, which is one of the two families of ruling lines on a hyperboloid of one sheet in 3-dimensional space
  13. Bruck & Bose 1966, p. 163
  14. Bruck & Bose 1966, p. 164, Theorem 12.1
  15. Knuth 1965, p. 541
  16. "Projective Planes of Small Order". http://ericmoorhouse.org/pub/planes/. 
  17. "Projective Planes of Order 16". http://ericmoorhouse.org/pub/planes16/. 
  18. Reifart 1984
  19. "Projective Planes of Order 25". http://ericmoorhouse.org/pub/planes25/. 
  20. Dover 2019
  21. Czerwinski & Oakden 1992
  22. "Projective Planes of Order 27". http://ericmoorhouse.org/pub/planes27/. 
  23. Dempwolff 1994
  24. "Projective Planes of Order 32". http://ericmoorhouse.org/pub/planes32/. 
  25. Mathon & Royle 1995
  26. "Projective Planes of Order 49". http://ericmoorhouse.org/pub/planes49/. 
  27. McKay & Royle 2014. This is a complete count of the 2-dimensional non-Desarguesian translation planes; many higher-dimensional planes are known to exist.

References

Further reading

  • Mauro Biliotti, Vikram Jha, Norman L. Johnson (2001) Foundations of Translation Planes, Marcel Dekker ISBN 0-8247-0609-9 .

External links