Axiomatising The Modal Logic of Affine Planes: Ian Hodkinson Joint Work With Altaf Hussain Thanks To Nick For Inviting Me

introduction
Projective and affine planes incidence systems of points and lines
Axiomatising the modal logic of affine planes

Ian Hodkinson
joint work with Altaf Hussain
Thanks to Nick for inviting me
axiomatic abstractions of real plane geometry

no quantitative information (distances)
qualitative directional information (collinearity, parallelism)
Heavily studied in mathematics as a core part of geometry.
Studied in modal logic by Stebletsova, Venema, Balbiani, Goranko,
Vakarelov, . . . Spatial logic is currently of some interest.
Venema (1999) proposed 2-sorted treatment, and axiomatised
projective planes.
BalbianiGoranko (2001) axiomatised weak affine planes.
This talk: the true affine planes are not finitely axiomatisable!
1
definition projective plane

A projective plane is a triple (P, L, E), where P, L are the sets of
points and lines (resp.), E P L, and
outline
definitions, examples
existing work (BruckRyser, Venema, BalbianiGoranko)
proof of non-finite axiomatisability
conclusion, problems
P1. any two distinct points lie on a unique line

x, y P (x 6= y ! l L(x E l y E l))
P2. any two distinct lines meet in a unique point
l, m L(l 6= m ! x P (x E l x E m))
P3. there exist four points,
^ no three of which are collinear
x0 x1 x2 x3 P
l L(xi E l xj E l xk E l)
i<j<k<4
examples
definition affine plane

An affine plane is a triple (P, L, E, ||), where P, L are the sets of
points and lines (resp.), E P L, || L L, and
A0. two lines are parallel iff they are equal or disjoint
l, m L(l || m l = m x P (x E l x E m))
A1. any two distinct points lie on a unique line
x, y P (x 6= y ! l L(x E l y E l))
A2. there is a unique line through any point parallel to any given line
x P l L ! m L(x E m m || l)
A3. there exist three non-collinear points
x0 x1 x2 P l L(x0 E l x1 E l x2 E l)
(equivalently: L 6= , and for any l L, there is a point not on l)
The real plane R2 naturally yields an affine plane.

Replace R by a finite field you get a finite affine plane.
It can be completed to a finite projective plane.
completion: affine plane projective plane
In an affine plane, || is an equivalence relation. Take its equivalence
classes to be the points of a new line (the line at infinity). This gives
a projective plane the completion of the affine plane.
projective plane affine plane
Rip a line and its points out of a projective plane get affine plane.
v
J

v Jv
projective plane
b
v"J
b"
"
b J
"
b

"
v
v b
Jv
JJ
affine plane
v J
"b J
"
b
"
b
Jv
v
order
The order of a projective plane is (the number of points on a line) 1.
This is well-defined.
modal logic of projective planes
The order of an affine plane is the number of points on a line.

This is (the number of parallel classes) 1.
Yde Venema (1999) proposed treating projective plane (P, L, E) as

2-sorted Kripke frame (P, L, E, ), where is converse of E.
Modal formulas:
BruckRyser theorem, 1949: If a finite projective plane has order n,

and n 1 or 2 mod 4, then n is the sum of two squares.
Exercise: For all n 0, the number 2 32n+1 is 2 mod 4 and is not
the sum of two squares.
So theres no projective (or affine) plane of order 2 32n+1 , for any
n 0.
No affine plane has 2 32n+1 + 1 parallel classes, for any n 0.
6
point formulas
::= p0 | | | [01]
line formulas
::= p1 | | | [10]
The boxes [01], [10] have accessibility relations E, (resp.).

h01i, h10i are the corresponding diamonds (the usual abbreviations).
Venemas axioms and rules
what about affine planes?
(Sahlqvist) axioms:
all propositional tautologies
normality of [01], [10]
[01] and [10] are mutually converse
h01i, h10i
transitivity of hi and hi, where hi = h01ih10i, hi = h10ih01i.
(In a projective plane, these are universal modalities for their sorts.)
Regard affine planes as 2-sorted Kripke frames (P, L, E, , ||).
Rules: modus ponens, generalisation for [01], [10], substitution.
This system is (strongly) sound and complete for projective planes.

Venema also proved that satisfiability for the logic of projective
planes is decidable and NE XPTIME-complete.
point formulas
line formulas
::= p0 | | | [01]
::= p1 | | | [10] | [||]
BalbianiGoranko (2002) proposed axioms for affine planes.

Completeness was left open (though it was proved for a wider class
of structures called weak affine models).
IHHussain (2005): the 2-sorted modal logic of affine planes is not

finitely axiomatisable.
So BGs axioms are not complete.
details configurations (affine-like frames)

proof of non-finite axiomatisability
Idea: construct finite affine-like frames Cn (n < ) such that:
1. each Cn is not the bounded morphic image of any affine plane,
2. their limit C (e.g., an ultraproduct) is a bounded morphic
image of an affine plane.
If the logic of affine planes were axiomatisable by a finite set of
formulas, then
Let 1 be a cardinal.
A -configuration is a frame C = (P, L, E, , ||) such that
C0. two lines are parallel iff they are equal or disjoint
C1. For all x, y P , there are at least lines l L with x, y
l.
C2. there is a unique line through any point parallel to any given line
C3. L 6= , and for any l L, there is a point not on l.
1. each Cn would satisfy at least one of { : },

2. C would validate .
With suitable construction of limit, this is impossible!
C0, C2, C3 are equivalent to affine axioms A0, A2, A3.

C1 is different from A1 (any two distinct points lie on a unique line).
Any affine plane is a 1-configuration.
10
11
making the Cn
configurations and affine planes
Lemma 1. For every finite k 1, there is a finite k-configuration Ck

with exactly c parallel classes, where c = 2 32n+1 + 1 for some n.
Lemma 2. Let C be a -configuration, A an affine plane, and

f : A C a surjective homomorphism. Then A and C have the
same number of ||-classes.
Proof. Pick c, d, n with k 2d2 , 4kd2 2 32n+1 + 1 = c, c 2d1 .

Take any set P with |P | = 2d. Put S := {l P : |l| = d}.
Choose L S with
l LP \l L
x, y P k l L(x, y l)
d2
k.
Possible since |{l S : x, y l}| 2d2
d2 2
2
Need to pick total of (2d) k sets l (plus their complements).
|L|/2 = c.
So far, |L|/2 4kd2 c. Just add more l S to L until |L|/2 = c.

d1
d
c.
Possible since |S| = 2d
d 2 so |S|/2 2
Check Ck = (P, L, , , ||) (where || is defined by C0) is as required.
Proof. Show f induces a bijection on ||-classes. Take lines l, m of A.

l || m f (l) || f (m) trivial.
Assume (l || m) but f (l) || f (m), and get contradiction:
'
'
$
$
A

u
l
f
(l)
``` xu

```

` m f
f (x)
f (m)
(
(
(

u
((((

n
n

C&
&
%
%
By BruckRyser theorem, no Ck from Lemma 1 is the bounded
morphic image of an affine plane.
12
13
limit
coherence (soundness) conditions and defects
Let C be a limit (e.g., non-principal ultraproduct) of the Ck .

It is a |C |-configuration.
S1. |Nk | < |C | for each k
Lemma 3. C is the bounded morphic image of an affine plane.

Proof. Build affine plane A and surjective bounded morphism
f : A C by a step-by-step construction similar to Venema (1999).
'
$
yXX
X
XXX
C
XXX
f surjective bounded morphism
yX
X
&
%
XXXf2 . . . XXXX
XXX
XXX
YH
H
f1
XX
f0 6
X
'
$
H

t
affine plane
...
N0 t AAt
N1
N2
A

&
%
chain of networks Nk (k < |C |)
14
S2. each fk is a homomorphism

S3. l || m fk (l) || fk (m), for all lines l, m of Nk
S4. any two distinct points have at most one line through them
S5. no two distinct parallel lines intersect
S6. there exist three non-collinear points
D1. back-defects for fk for each of h01i, h10i, h||i.
D2. two points with no line joining them
D3. parallel-axiom defects
D4. non-parallel lines with no point of intersection
15
D2 defects (the main case)

'
$
C
need to add a line l through x, y

$
'
C
&
Nk
6fk
'
x
&
&
Nk
6fk
'
x
&
$
l?
16
. . . but x, y already mapped by fk . . .

'
$
v
fk (y)
v
C fk (x)
&
Nk
6fk
'
x
&
. . . so need to define the fk+1 -image of l carefully

'
$
C
17
$
l?
&
Nk
v
fk (y)
v
fk (x)
6fk
'
x
&
18
fk+1 (l)?
$
l?
%
19
danger. . .
'
C
v
fk (x)
&
v fk (m)
fk (y)
C
%
6fk
$
'
PP
v PPvP
l?
Nk x
y P
m
&
escape
'
v
fk (x)
fk+1 (l)
v fk (m)
fk (y)
&
%
k fk+1
Q
Q
6fk
Q'
$
$
'
PP
PP
PPv l
v PPvP
v
P
Nk x
y P
y Pm Nk+1
x
m
&
% &
21
20
summary
complexity of satisfiability problem
Each Ck is not the bounded morphic image of an affine plane

The limit C is the bounded morphic image of an affine plane
So the modal logic of affine planes is not finitely axiomatisable.
This is a surprising contrast with projective planes.

Can view as a positive result explicit axioms may have interesting
geometrical content.
22
IHHussain
hybrid projective
Z
}
Z
Z

(ess.) BalbianiGoranko
Z
Z
trivial Z
Z
Z
hybrid affine
6
trivial
modal affine
:

IHHussain
Z

Z

modal projective (NE XPTIME-complete)
23
some references
1. P. Balbiani and V. Goranko, Modal logics for parallelism,
orthogonality, and affine geometries., Journal of Applied
Non-Classical Logics 12 (2002), no. 3-4, 365398.
open questions
1. decidability, complexity of modal logic of affine planes (etc)?
2. find explicit axioms for modal logic of affine planes
3. what if we use hybrid logic?
4. what if we use strict (irreflexive) parallelism?
BalbianiGoranko (2002) axiomatised affine planes using
non-standard irreflexivity rules.
decidability and complexity open
2. R H Bruck and H J Ryser, The nonexistence of certain finite

projective planes, Canad. J. Math. 1 (1949), 8893.
3. A. Hussain, Applications of extended modal logics, Ph.D. thesis,
University of London, 2006, submitted.
4. V Stebletsova, Algebras, relations and geometries, Ph.D. thesis,
Zeno, LeidenUtrecht Research Institute of Philosophy,
Heidelberglaan 8, 3584 CS Utretcht, Netherlands, 2000, Number
XXXII in Qustiones Infinit.
5. Y. Venema, Points, lines and diamonds: a two-sorted modal logic
for projective planes, J. Logic Computat. 9 (1999), 601621.
24
25

Axiomatising The Modal Logic of Affine Planes: Ian Hodkinson Joint Work With Altaf Hussain Thanks To Nick For Inviting Me

Uploaded by

Copyright:

Available Formats

Axiomatising The Modal Logic of Affine Planes: Ian Hodkinson Joint Work With Altaf Hussain Thanks To Nick For Inviting Me

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Axiomatising The Modal Logic of Affine Planes: Ian Hodkinson Joint Work With Altaf Hussain Thanks To Nick For Inviting Me

Uploaded by

Copyright:

Available Formats

introduction

Projective and affine planes incidence systems of points and lines

Axiomatising the modal logic of affine planes

axiomatic abstractions of real plane geometry

definition projective plane

P1. any two distinct points lie on a unique line

definition affine plane

The real plane R2 naturally yields an affine plane.

modal logic of projective planes

The order of an affine plane is the number of points on a line.

Yde Venema (1999) proposed treating projective plane (P, L, E) as

BruckRyser theorem, 1949: If a finite projective plane has order n,

The boxes [01], [10] have accessibility relations E, (resp.).

Venemas axioms and rules

what about affine planes?

Regard affine planes as 2-sorted Kripke frames (P, L, E, , ||).

Rules: modus ponens, generalisation for [01], [10], substitution.

This system is (strongly) sound and complete for projective planes.

BalbianiGoranko (2002) proposed axioms for affine planes.

IHHussain (2005): the 2-sorted modal logic of affine planes is not

details configurations (affine-like frames)

1. each Cn would satisfy at least one of { : },

C0, C2, C3 are equivalent to affine axioms A0, A2, A3.

configurations and affine planes

Lemma 1. For every finite k 1, there is a finite k-configuration Ck

Lemma 2. Let C be a -configuration, A an affine plane, and

Proof. Pick c, d, n with k 2d2 , 4kd2 2 32n+1 + 1 = c, c 2d1 .

Proof. Show f induces a bijection on ||-classes. Take lines l, m of A.

coherence (soundness) conditions and defects

Let C be a limit (e.g., non-principal ultraproduct) of the Ck .

S1. |Nk | < |C | for each k

Lemma 3. C is the bounded morphic image of an affine plane.

S2. each fk is a homomorphism

D2 defects (the main case)

need to add a line l through x, y

. . . but x, y already mapped by fk . . .

. . . so need to define the fk+1 -image of l carefully

complexity of satisfiability problem

Each Ck is not the bounded morphic image of an affine plane

This is a surprising contrast with projective planes.

2. R H Bruck and H J Ryser, The nonexistence of certain finite

You might also like