Cyclic Structures and Incidence Theorems - math.tu … · Pappus’s Theorem (123) (159) (186)...
Transcript of Cyclic Structures and Incidence Theorems - math.tu … · Pappus’s Theorem (123) (159) (186)...
Cyclic Structures andIncidence Theorems
Jürgen Richter-GebertTechnical University Munich
Jürgen Richter-GebertTechnical University Munich
Stupid Proofs forInteresting (?) Theorems
Incidence Theorems
Oriented Matroids (Pseudoline arrangements)
Zonotopal Tilings (Penrose like tylings)
Stresses/Liftings/Reciprocal Figures
Bracket Polynomials (Algebra of projective Geometry)
Circle Patterns
DDG
Topics we will meet
Running Example1
32
654
9 8 7
The collinearity of (123), (456), (159), (168), (249), (267), (348), (357)implies the collineartity of (7,8,9).
Pappus‘s Theorem
a
x
b
y
c
z
A Warmup
Area Method
Claim:
area(a,b,c)+ area(x,z,y)+ area(a,x,y,b)+ area(b,y,z,c)+ area(a,x,z,c) = 0
A Warmup
Area Method
Claim:
area(a,b,c)+ area(x,z,y)+ area(a,x,y,b)+ area(b,y,z,c)+ area(a,x,z,c) = 0
a
x
b
y
c
z
A Warmup
Area Method
Claim:
area(a,z,b,w)+ area(a,y,d,z)+ area(b,z,d,x)+ area(a,w,c,y)+ area(b,x,c,w)+ area(c,x,d,y) = 0
Works for any Manifold
w
z
y
x
d
c
b
a
A Warmup
Area Method
Claim:
Decomposing the cube
area(a,z,b,w)+ area(a,y,d,z)+ area(b,z,d,x)+ area(a,w,c,y)+ area(b,x,c,w)+ area(c,x,d,y) = 0
A Warmup
Area MethodDecomposing the cube
==> Discrete Königs NetsBobenko, Suris
A Warmup
Area MethodReciprocal Diagrams
A Warmup
Area MethodReciprocal Diagrams
A Warmup
Area MethodReciprocal Diagrams
You never have a problem with the last edge
Realizability / StrechabilityArrangements of Pseudolines
- Topological lines in RP- Two cross exactly once
Is there anequivalent
line arrangement?
2
Realizability / Strechability
Is there anequivalent
line arrangement?
Arrangements of Pseudolines
- Topological lines in RP- Two cross exactly once
2
Excursion: How to draw pseudoline arrangements?
1st Method:- fix boundary points- make a Tutte embedding- draw smooth splines
ConjectureEvery such pseudolineis „a function graph“
==> no curls allowed!
Realizability / Strechability
Is there anequivalent
line arrangement?
Arrangements of Pseudolines
- Topological lines in RP- Two cross exactly once
2
Excursion: How to draw pseudoline arrangements?
1st Method:- fix boundary points- make a Tutte embedding- draw smooth spines
ConjectureEvery such pseudolineis „a function graph“
==> no curls allowed!
Realizability / Strechability
Is there anequivalent
line arrangement?
Arrangements of Pseudolines
- Topological lines in RP- Two cross exactly once
2
Excursion: How to draw pseudoline arrangements?
2nd Method:- Draw two copies on a sphere- Circle pack- connect the dots
Realizability / Strechability
Is there anequivalent
line arrangement?
Arrangements of Pseudolines
- Topological lines in RP- Two cross exactly once
2
Excursion: How to draw pseudoline arrangements?
2nd Method:- Draw two copies on a sphere- Circle pack- connect the dots
FactUnique (!) representationof the arrangement
QuestionWhat are the special properties?
Realizability / Strechability
Is there anequivalent
line arrangement?
Arrangements of Pseudolines
- Topological lines in RP- Two cross exactly once
2
Realizability / Strechability
Is there anequivalent
line arrangement?
Arrangements of Pseudolines
- Topological lines in RP- Two cross exactly once
2
Realizability / Strechability
Is there anequivalent
line arrangement?
No !!
Pappus’s Theorem1 O Oa b cd e fO 1 Og h ij k lO O 1m n op q r
123456789
Another proof by algebraic cancellation
132
654
9 8 7
Pappus’s Theorem1 O Oa b cd e fO 1 Og h ij k lO O 1m n op q r
123456789
(123)(159)(168)(249)(267)(348)(357)(456)(789)
Another proof by algebraic cancellation
132
654
9 8 7
==> ce=bf==> iq=hr==> ko=ln==> ar=cp==> bj=ak==> fm=do==> dh=eg==> gl=ij==> mq=np
Pappus’s Theorem1 O Oa b cd e fO 1 Og h ij k lO O 1m n op q r
123456789
(123)(159)(168)(249)(267)(348)(357)(456)(789)
Another proof by algebraic cancellation
132
654
9 8 7
==> ce=bf==> iq=hr==> ko=ln==> ar=cp==> bj=ak==> fm=do==> dh=eg==> gl=ij==> mq=np
Pappus’s Theorem1 O Oa b cd e fO 1 Og h ij k lO O 1m n op q r
123456789
(123)(159)(168)(249)(267)(348)(357)(456)(789)
Another proof by algebraic cancellation
132
654
9 8 7
==> ce=bf==> iq=hr==> ko=ln==> ar=cp==> bj=ak==> fm=do==> dh=eg==> gl=ij==> mq=np
Pappus’s Theorem1 O Oa b cd e fO 1 Og h ij k lO O 1m n op q r
123456789
(123)(159)(168)(249)(267)(348)(357)(456)(789)
Another proof by algebraic cancellation
132
654
9 8 7
==> ce=bf==> iq=hr==> ko=ln==> ar=cp==> bj=ak==> fm=do==> dh=eg==> gl=ij<== mq=np
Pappus’s Theorem1 O Oa b cd e fO 1 Og h ij k lO O 1m n op q r
123456789
(123)(159)(168)(249)(267)(348)(357)(456)(789)
132
654
9 8 7
Another proof by algebraic cancellation
Pappus’s Theorem
(123)
(159)
(186)
(429) (726)
(483)
(753) (456)
(789)
Structure of the proof
13
2
654
9 8 7
Pappus’s Theorem
(123)
(159)
(186)
(429) (726)
(483)
(753) (456)
(789)
Structure of the proof
13
2
654
9 8 7
Pappus’s Theorem
(123)
(159)
(186)
(429) (726)
(483)
(753) (456)
(789)
c ab
e
d
f
r
q
p
j
l
k
h ig
m no
e r jb
g
o
Structure of the proof
13
2
654
9 8 7
Pappus’s Theorem
(123)
(159)
(186)
(429) (726)
(483)
(753) (456)
(789)
c ab
e
d
f
r
q
p
j
l
k
h ig
m no
e r jb
g
o
Structure of the proof
13
2
654
9 8 7
Pappus’s Theorem
(123)
(159)
(186)
(429) (726)
(483)
(753) (456)
(789)
c ab
e
d
f
r
q
p
j
l
k
h ig
m no
e r jb
g
o
Structure of the proof ==> a torus
13
2
654
9 8 7
From: Self-Dual Configurations and Regular Graphs Coxeter, 1950
For Desargues’s Theorem there is no such choice of a basis
Generalizing this proof
special choice of basis
2x2 determinants
cancellation pattern
==> ce=bf==> iq=hr==> ko=ln==> ar=cp==> bj=ak==> fm=do==> dh=eg==> gl=ij<== mq=np
(123)(159)(168)(249)(267)(348)(357)(456)(789)
Problems:
Grassmann-Plücker Relations
1
2
3
x
y
[123][1xy]-[12x][13y]+[12y][13x] = OIn every configuration of five points 1,2,3,x,y
is satisfied ( with [abc]=det(a,b,c) ).
(123) collinear ==>
[12x][13y]=[12y][13x]
[12x][13y]=[12y][13x] ==> (123) collinear or (1xy) collinear
Pappos’s Theorem
(123) ==> [124][137]=[127][134](159) ==> [154][197]=[157][194](168) ==> [184][167]=[187][164](249) ==> [427][491]=[421][497](456) ==> [457][461]=[451][467](348) ==> [487][431]=[481][437](267) ==> [721][764]=[724][761](357) ==> [751][734]=[754][731](789) <== [781][794]=[784][791]
132
654
9 8 7
Same cancellation pattern as before
Pappos’s Theorem
(123) ==> [124][137]=[127][134](159) ==> [154][197]=[157][194](168) ==> [184][167]=[187][164](249) ==> [427][491]=[421][497](456) ==> [457][461]=[451][467](348) ==> [487][431]=[481][437](267) ==> [721][764]=[724][761](357) ==> [751][734]=[754][731](789) <== [781][794]=[784][791]
132
654
9 8 7
Same cancellation pattern as before
Pappos’s Theorem
(123) ==> [124][137]=[127][134](159) ==> [154][197]=[157][194](168) ==> [184][167]=[187][164](249) ==> [427][491]=[421][497](456) ==> [457][461]=[451][467](348) ==> [487][431]=[481][437](267) ==> [721][764]=[724][761](357) ==> [751][734]=[754][731](789) <== [781][794]=[784][791]
132
654
9 8 7
Same cancellation pattern as before
An automatic method
Generate all “binomial equations” that come from the hypotheses
Generate all “binomial equations” that come from the conclusion
Find a suitable combination of the conclusion by the hypotheses
A linear problem:Is the conclusion in the span of the hypotheses ?
(binomial-proofs)
Desargues’s by cancellation
==> (768) or (745)
(479) ==> [471][496] = [476][491](916) ==> [914][962] = [912][964](259) ==> [256][291] = [251][296](240) ==> [248][203] = [243][208](083) ==> [082][035] = [085][032](570) ==> [573][508] = [578][503](213) ==> [215][234] = [214][235](418) ==> [412][487] = [417][482](536) ==> [532][567] = [537][562]
54
821
09
876
[764][785] = [765][784]
A non-realizable configuration
jh
d
g
e f
c
bai
(abi) ==> [abh][agi]=+[abg][ahi](acf) ==> [adf][acj]=-[acd][afj](adh) ==> [abd][afh]=-[abh][adf](bce) ==> [bcd][bej]=-[bde][bcj](bdg) ==> [abg][bde]=-[abd][beg](cdj) ==> [acd][bcj]=+[acj][bcd](efj) ==> [afj][egj]=+[aej][fgj](egi) ==> [aeg][ghi]=-[agi][egh](fhi) ==> [ahi][fgh]=-[afh][ghi](ghj) ==> [egh][fgj]=+[egj][fgh]
[aeg][bej]=+[aej][beg]
==> (abe) or (egj)
A non-realizable configuration
jh
d
g
e f
c
bai
(abi) ==> [abh][agi]=+[abg][ahi](acf) ==> [adf][acj]=-[acd][afj](adh) ==> [abd][afh]=-[abh][adf](bce) ==> [bcd][bej]=-[bde][bcj](bdg) ==> [abg][bde]=-[abd][beg](cdj) ==> [acd][bcj]=+[acj][bcd](efj) ==> [afj][egj]=+[aej][fgj](egi) ==> [aeg][ghi]=-[agi][egh](fhi) ==> [ahi][fgh]=-[afh][ghi](ghj) ==> [egh][fgj]=+[egj][fgh]
[aeg][bej]=+[aej][beg]
==> (abe) or (egj)
A non-realizable configuration
jh
d
g
e f
c
bai
(abi) ==> [abh][agi]=+[abg][ahi](acf) ==> [adf][acj]=-[acd][afj](adh) ==> [abd][afh]=-[abh][adf](bce) ==> [bcd][bej]=-[bde][bcj](bdg) ==> [abg][bde]=-[abd][beg](cdj) ==> [acd][bcj]=+[acj][bcd](efj) ==> [afj][egj]=+[aej][fgj](egi) ==> [aeg][ghi]=-[agi][egh](fhi) ==> [ahi][fgh]=-[afh][ghi](ghj) ==> [egh][fgj]=+[egj][fgh]
[aeg][bej]=+[aej][beg]
==> (abe) or (egj)
Six points on a conic1
2
3
4
5
6
[123][156][426][453] = [456][126][254][423]
Six points 1,2,3,4,5,6 are on a conic <==>
Pascal’s Theorem
[159][257] = -[125][579] [126][368] = +[136][268] [245][279] = -[249][257] [249][268] = -[246][289] [346][358] = +[345][368] [135][589] = -[159][358][125][136][246][345] = +[126][135][245][346]
12
3
4
5
6
98 7
[289][579] = +[279][589]
Problems of this method
Usually large search space
What is the “structure” of the proof
How to cut down the search space in advance
(binomial-proofs)
Problems of this method
Usually large search space
What is the “structure” of the proof
How to cut down the search space in advance
(binomial-proofs)
A ambitious dream:Look at a theorem.... “see” its structure........ and produce a proof immediately!!
Coxeter’s proof of Pappos’s Theorem
The Theoremsof Ceva and Menelaos
Ceva and Menelaosa1a2b1b2c1c2ABCDEFG
a1a2b1b2c1c2ABCDEH
A B
Y
C
Z
X
A B
Y
C
Z
X
|AZ|·|BX|·|CY||ZB|·|XC|·|YA| = 1 |AZ|·|BX|·|CY|
|ZB|·|XC|·|YA| = -1
Ceva and Menelaosa1a2b1b2c1c2ABCDEFG
a1a2b1b2c1c2ABCDEH
A B
Y
C
Z
X
A B
Y
C
Z
X
|AZ|·|BX|·|CY||ZB|·|XC|·|YA| = 1 |AZ|·|BX|·|CY|
|ZB|·|XC|·|YA| = -1
· ·[CDA] [ADB] [BDC][CDB] [ADC] [BDA]= -1
[XYA] [XYB] [XYC][XYB] [XYC] [XYA]· · = 1
D
Glueing
|AX|·|BY|·|CZ||XB|·|YC|·|ZA| = 1
ABCDEFGHKLM
A
B
C
DX
Y
Z
S
R
|AZ|·|CR|·|DS||ZC|·|RD|·|SA| = 1
Glueing
|AX|·|BY|·|CZ||XB|·|YC|·|ZA| = 1
ABCDEFGHKLM
A
B
C
DX
Y
Z
S
R
|AZ|·|CR|·|DS||ZC|·|RD|·|SA| = 1
|AX|·|BY|·|CR|·|DS||XB|·|YC|·|RD|·|SA| = 1
Glueing
ABCDEFGHKLM
a1
b2
b3
b4
b1
a2a3
a4
= 1a1
b2 b3 b4b1
a2 a3 a4
Glueinga1a2b1b2c1c2d1d2f2e1e2f1
a1
b2
b3
b4
b5
b6
b1
a2
a3 a4
a5
a6 = 1a1
b2 b3 b4 b5b6b1
a2 a3 a4 a5a6
Glueing
ABCDEFHKMP
ABCDEFHKM
ABCDEFKMP+ =
A “factory” for geometric theorems
A theorem on a tetrahedron
ABCDEFKMP
Front: two triangles with Ceva
A theorem on a tetrahedron
ABCDEFHKM
Back: two triangles with Ceva
A theorem on a tetrahedron
ABCDEFHKM
ABCDEFKMP
After glueing: an incidence theorem
4 Times Ceva
spatial interpretations
interesting degenerate situations
A special case of Pappus’s Theorem as special case
Ceva on four sides of a tetrahedron
A census of incidence theorems
MM M M+
CC C C+
CC
MM
+
CM M C+
MM C C+
MM C C+==> harmonic points
==> many interesing degenerate cases
==> Desargues’s Thm.
Harmonic QuadruplesCC
MM
+
Front: two triangles with Ceva
Harmonic QuadruplesCC
MM
+
Back: two triangles with Menelaos
Harmonic QuadruplesCC
MM
+
Glued: Uniques of harmonic point construction
Desargues by GlueingMM M M+
Front: two triangles with Menelaos
Desargues by GlueingMM M M+
Back: two triangles with Menelaos
Desargues by GlueingMM M M+
Glued: Desargues’s Theorem
Six triangles
1
1
12
1
34
5
1
1
1 2
2
2
Double pyramid over triangle
Degenerate torus
...and other degenerate spheres
Six triangles
1
1
12
1
34
5
1
1
1 2
2
2
Double pyramid over triangle
Degenerate torus
...and other degenerate spheres
Six triangles
Six times Ceva
Six triangles
Six times Ceva
Folding the triangles
Six triangles
Six times Ceva -->after identification
Six triangles
Pappus’s Theorem
Moving points to infinity
Pappus affine Pappus
Grid theorems
+1
-1
-1
-1
+1
+1
“row sums” = “colums sums” = “diagonal sums” = 0
Another Theorem+1
-1
+1
+1
+1
-1
-1
-1
“row sums” = “colums sums” = “diagonal sums” = 0
Larger Grid Theorems
+ =
Composition of grid theorems
“space of such theorems” is a vector space
“little” Pappus configurations are a basis
Larger Grid Theorems
Composition can be interpreted topologically
„Geometrie der Waben“compare Blaschke 1937
Arrangements of pseudolines
Rombic Tilings with three Directions
Rombic Tilings with three Directions
Conversion of proofs
12
3
4
4
4
5
6
7
7
8
9
A
AB
B
C
C
7
[124][137]=[127][134][154][197]=[157][194][184][167]=[187][164][427][491]=[421][497][457][461]=[451][467][487][431]=[481][437][721][764]=[724][761][751][734]=[754][731][781][794]=[784][791]
Conversion of proofs
12
3
4
4
4
5
6
7
7
8
9
A
AB
B
C
C
7
[124][137]=[127][134][154][197]=[157][194][184][167]=[187][164][427][491]=[421][497][457][461]=[451][467][487][431]=[481][437][721][764]=[724][761][751][734]=[754][731][781][794]=[784][791]
427479
478
347
457
467
146 137154 157
178
148
134167
127
124149 197
- Works in general- Use Tutte-Groups- and homotopy
A different Story
From: Self-Dual Configurations and Regular Graphs Coxeter, 1950
Graph: vertices -> brackets, edges -> bracktes differing by one letter
Glue versus matter
[12x][13y]-[12y][13x]
Grassmann Menelaus Ceva
Glue versus matter
Glue versus matter
Glue versus matter
Glue versus matter
BFP <--> CM
And Conics ?a1
b3
b1
a3
b5
a5b2 b4
a2
a4
b6
a6
= 1a1
b2 b3 b4 b5b6b1
a2 a3 a4 a5a6
Carnot’s Theorem
A Conic Theorem
Works for any orientable triangulated manifold
A Conic Theorem
Works for any orientable triangulated manifold
A Conic Theorem
Works for any orientable triangulated manifold
Pascal’s Theorem
B
A
B
CA
C
MM
M
M
CarB
A
C
+
Loose Ends on Circles
K
L
M
Loose Ends on Circles
P7
P8
Loose Ends on Circles
P7
P8
Loose Ends on Circles
P7
P8
Loose Ends on Circles
P7
P8
Loose Ends on Circles
P7
P8