You are on page 1of 25

Why the Perceived Flaw in Kempe’s 1879

Graphical ‘Proof’ of the Four Colour Theorem


is Not Fatal When Expressed Geometrically

Bhupinder Singh Anand

bhup.anand@gmail.com
Mumbai, India

Prof. P. C. Vaidya National Conference on Mathematical


Sciences
15-16 March 2022, Department of Mathematics, Sardar Patel University, Vallabh Vidyanagar - 388120,
Gujarat, India

Bhupinder Singh Anand Why Four Colours Suffice


01: Why we need a computer-free proof of 4CT

Although the Four Colour Theorem 4CT is widely considered as ‘proven’ by


Kenneth Appel and Wolfgang Haken* in 1977; and by Neil Robertson and his
colleagues** in 1997:

“There are two reasons why the Appel-Haken proof is not com-
pletely satisfactory.

Part of the Appel-Haken proof uses a computer, and cannot


be verified by hand, and

even the part that is supposedly hand-checkable is


extraordinarily complicated and tedious, and as far as we
know, no one has verified it in its entirety."
. . . Robertson et al***, 1995, Pre-publication remarks.

*[AH77] Kenneth Appel and Wolfgang Haken. 1977. Every planar map is four colorable. Part I: Discharging. In
Illinois Journal of Mathematics, Volume 21, Issue 3 (1977), pp. 429-490, University of Illinois, Urbana-Champaign.
**[RSST] Neil Robertson, Daniel Sanders, Paul Seymour, and Robin Thomas. 1997. The four-colour theorem. In
Journal of Combinatorial Theory, Series B, Volume 70, Issue 1, May 1997, Pages 2-44.
***[RSSp] Neil Robertson, Daniel Sanders, Paul Seymour, and Robin Thomas. 1995. The four-colour theorem.
Pre-publication summary.

Bhupinder Singh Anand Why Four Colours Suffice


02: The goal

The goal of this paper is therefore to give a computer-independent argument


that transparently illustrates:

Why four colours are sufficient to colour any planar map;

and:

Why the flaw in Alfred Kempe’s graphical 1879 ‘proof’* of the Four
Colour Theorem is not fatal when the argument is expressed
geometrically .

*[Kmp79] Alfred Bray Kempe. 1879. On the Geographical Problem of the Four Colours. In American Journal of
Mathematics, 2 (3): 193-220, doi:10.2307/2369235, JSTOR 2369235

Bhupinder Singh Anand Why Four Colours Suffice


03: An elementary proof of 4CT

We begin by articulating:

The Four Colour Problem

Can every planar map on a globe be coloured with at most four colours
in such a way that neighbouring countries are coloured differently ?

Bhupinder Singh Anand Why Four Colours Suffice


04: An elementary proof of 4CT (contd.)

All illustrations should be treated as notionally representing the surface of a


globe.

✬✩✬✩✬✩✬✩
4


1 1 1
✟ ✡


1 3
❍ 3
✫✪✫✪✫✪✫✪
2 2 ❅
❅ 2 ❅

Clearly any map with only 1, 2, 3 or 4 countries requires, at most, four


colours if no two countries with the same colour are neighbours
(defined as having a common, non-zero, boundary).

Bhupinder Singh Anand Why Four Colours Suffice


05: An elementary proof of 4CT (contd.)

QUERY: Is there a minimum number n > 4 of countries in a minimal map


Mmin such that:

— a 5th colour is necessary in every coloring of Mmin ;

whilst

— any sub-minimal map, say Ssubm , with less than n countries, can always
be 4-coloured?

ANSWER: No. Here’s why:

Bhupinder Singh Anand Why Four Colours Suffice


06: An elementary proof of 4CT (contd.)

Fig.1: The minimal planar map Mmin

B ✬✩
G B
C

F
A ✫✪
D

R Y

Now, if there is a minimal map Mmin , then it must always contain some
country F which is surrounded by at least four countries
A, B, C, D—identified by, say, the colours Red, Green, Blue, Yellow.
Also, since any ‘equatorial belt’ divides a globe into two regions, we
cannot have that ‘B and D are neighbours’ and, at the same time, that
‘A and C are neighbours’.
We shall therefore assume that A and C are not neighbours in the
minimal map Mmin .
Bhupinder Singh Anand Why Four Colours Suffice
07: An elementary proof of 4CT (contd.)

Fig.2: Shrinking F to an apex PF yields a sub-minimal planar map Ssubm

G B
B C ✘✘ Apex PF in Ssubm
✾✘✘

A D
R Y

CASE 1: We first consider the case where country F neighbours only the
four, all differently coloured, countries A, B, C and D.

If we now eliminate country F by shrinking it to a point, say the apex


PF , in Fig.2;

— then the number of countries will be less than n;

— thereby reducing Mmin to a sub-minimal, 4-colourable, map Ssubm .

Bhupinder Singh Anand Why Four Colours Suffice


08: An elementary proof of 4CT (contd.)


Fig.3: Merging A and C at apex PF in Ssubm and recolouring Ssubm
G R

B C

A D
R Y

Merging the countries A and C at the apex PF into one country (as
shown in Fig.3),

— now yields another sub-minimal, hence 4-colourable, map Ssubm .

Keeping A − Red, B − Green, and D − Yellow, the sub-minimal map



Ssubm can now be recoloured,

— so that the areas A, B, C, D require only three colours (as shown in


Fig.3).

Bhupinder Singh Anand Why Four Colours Suffice


09: An elementary proof of 4CT (contd.)


Fig.4: Restoring A and C at PF in Ssubm with colours inherited from Ssubm

G R
B C ✘✘ Apex PF in Ssubm
✾✘✘

A D
R Y

If we now restore the apex PF in the sub-minimal map Ssubm :

— then A and C are again non-neighbouring countries,


— and we get a fresh 4-colouring of Ssubm ,
— in which only three colours meet at the apex PF (see Fig.4).

Bhupinder Singh Anand Why Four Colours Suffice


10: An elementary proof of 4CT (contd.)

Fig.5: Countries in the minimal planar map Mmin with F recreated in Ssubm

B ✬✩
G R
C

F
A ✫✪
D

R Y

Recreating the country F , at the Apex PF in the sub-minimal map


Ssubm , now yields a fresh colouring of the minimal map Mmin :
— which requires only a 4th colour for F ,
— thus contradicting the minimality of Mmin .

Conclusion 1: No minimal map Mmin can contain a country F that is


surrounded by only four countries if F necessarily requires a 5th colour.

Bhupinder Singh Anand Why Four Colours Suffice


11: An elementary proof of 4CT (contd.)

Fig.6: Country F shares a non-zero boundary segment with 5 others in Mmin

B ✬✩
G B
C


F d✲ E R
❍❍
A ✫✪
D

R Y

CASE 2: Hence any minimal map Mmin must contain a country F which
is surrounded by at least five countries,

— of which four, say A, B, C, D, are necessarily coloured Red,


Green, Blue, Yellow;
— whilst a fifth country, say E, which neighbours F (but not A) must
repeat some colour (say Red).

Without loss of generality, we need to only consider the case where F is surrounded by just five countries
(see Fig.6), to show why F cannot neighbour two countries in Mmin which have the same colour.

Bhupinder Singh Anand Why Four Colours Suffice


12: An elementary proof of 4CT (contd.)

Fig.7: Annexing E into F yields a sub-minimal map Ssubm

B ✬✩
G B
C


F d✲x F(E)
❍❍
A ✫✪
D

R Y

If country F now annexes country E,

— by deleting their common boundary d (as indicated in Fig.7),

Then this reduces the minimal map Mmin to a sub-minimal map Ssubm ,

— which has one less country.

Bhupinder Singh Anand Why Four Colours Suffice


13: An elementary proof of 4CT (contd.)

Fig.8: The sub-minimal map Ssubm would be 4-colourable

B ✬✩
G R or Y
C
✟ ✟
F d✲x F(E)

A ✫✪

D
R Y

Moreover, keeping A − Red, B − Green, and D − Yellow:


— the sub-minimal map Ssubm is now 4-colourable,
— where the country A and the annexed area F (E) are necessarily
differently coloured (as shown in Fig.8).

Further, since F neighbours A, the areas A and F (E) cannot have the
same colour in any 4-colouring of Ssubm .

Bhupinder Singh Anand Why Four Colours Suffice


14: An elementary proof of 4CT (contd.)

Fig.9: Countries A and E cannot share identical colours in Mmin as postulated

B ✬✩
G R or Y
C
✟ ✟
F d✲ E B

A ✫✪

D

R Y

We conclude that no matter how we 4-colour the sub-minimal map


Ssubm :

— restoring F and E to their original status,


— as independent countries in the minimal map Mmin (Fig.9),
— does not allow countries A and E to have the same colour in Mmin .

Bhupinder Singh Anand Why Four Colours Suffice


15: An elementary proof of 4CT (contd.)

However, this contradicts our premise that:

— countries A and E can be assumed to have the same colour ,

— in some colouring of the minimal map Mmin ,

— if the country F necessarily requires a 5th colour.

Conclusion 2: Any country in a minimal map Mmin that necessarily


requires a 5th colour can have at most four neighbours.

Bhupinder Singh Anand Why Four Colours Suffice


16: An elementary proof of 4CT (conclusion)

Since Conclusion 2 contradicts Conclusion 1, we further conclude


that:

There is no minimal map Mmin such that a 5th colour is necessary


in every possible coloring of Mmin .

We thus have an elementary, geometrical, proof of the Four Colour


Theorem:

Every planar map can be coloured with at most four colours in


such a way that neighbouring countries are coloured differently.

Bhupinder Singh Anand Why Four Colours Suffice


17: Why the perceived ‘flaw’ in Kempe’s 1879 ‘proof’ of 4CT is not fatal

Moreover , we have shown that the perceived ‘flaw’ in Alfred Kempe’s


1879 ‘proof’ [Kmp79] of 4CT is not fatal.

Reason: What Kempe sought to prove—by an identical


argument—follows immediately from Conclusion 2:
— Every colouring of a planar map in which a country has four, all
differently-coloured, neighbours can be 4-coloured.

The question arises:


Why did Kempe’s identical argument fail to prove the Four Colour
Theorem graphically ?

Bhupinder Singh Anand Why Four Colours Suffice


18: Kempe’s 1879 argument

In his 1879 paper [Kmp79], Kempe explicitly preferred appeal to Euler’s


formula V + E = F + 2 in a graphical, rather than geometrical,
interpretation of 4CT; where he correctly argued that:
(1) Every minimal map Mmin must contain a country F with either only
four , or only five, neighbours, such that F necessarily requires a
5th colour;
(2) Any minimal map Mmin containing a country F with only four , all
differently coloured, neighbours can be 4-coloured—thus
contradicting the minimality of Mmin .

However, Kempe then incorrectly argued that:


(3) Any minimal map Mmin containing a country F with only five
neighbours, four of which are all differently coloured, can be
4-coloured—thus contradicting the minimality of Mmin .

Bhupinder Singh Anand Why Four Colours Suffice


19: The flaw in Kempe’s argument

The flaw in Kempe’s argument (3) is the (seemingly innocent) implicit,


but fallacious, premise that:
— A minimal map Mmin can be assumed to contain a country F , with
only five neighbours, such that F necessarily requires a 5th colour.

In other words, Kempe’s graphical argument needed to, but apparently


could not*, prove Conclusion 2; which geometrically shows:
— No minimal map Mmin can be assumed to contain a country F :
— with two of its neighbours similarly coloured,
— if F necessarily requires a 5th colour.

Bhupinder Singh Anand Why Four Colours Suffice


20: Why the pictorial proof of 4CT may not be evident graphically

Fig.10: Graphical representation of Mmin Fig.11: Geometrical representation of Mmin

B ✬✩
B C G B

❅ C
❅ ✟


F E F d✲ E R
❍❍
A ✫✪
❅ ❍
D


A D R Y

Reason: It is not immediately obvious that the geometrical proof of 4CT is


expressible graphically since, as shown in Fig.10, classical graph theory
represents:

- the areas of countries of the minimal map Mmin as points (coloured


vertices); and

- a non-zero boundary between two countries of Mmin as a line (edge)


joining two points (coloured vertices).

Bhupinder Singh Anand Why Four Colours Suffice


21: Why the pictorial proof of 4CT may not be evident graphically
Fig.12: Graphical representation of Mmin Fig.13: Geometrical representation of Mmin

B ✬✩
B C G B

❅ C
❅ ✟


F+E F d✲x F(E)
❍❍
A ✫✪
❅ ❍
D


A D R Y

Hence, a graphical representation:


— cannot recognise the boundary segment d, shown in Fig.13, as a
distinct entity such that,
— after the annexation of E by F to form a 4-colourable sub-minimal map
Sminm ,
— it is evident in the graphical representation in Fig.12 also—as is
obvious in Fig.13—that A and E could not have been identically
coloured in Mmin .
Bhupinder Singh Anand Why Four Colours Suffice
22: Why the pictorial proof of 4CT may not be evident graphically

Fig.14: Pseudo-Graphical representation of Mmin Fig.15: Geometrical representation of Mmin

B ✬✩
B C G B

❅ C
❅ ✟


F E F d✲x F(E)
❍❍
A ✫✪
❅ ❍
D


A D R Y

However, it is conceivable—perhaps even likely —that:

— it was intuitively evident to Kempe that 4CT must follow since,

— in the above pseudo-graphical representation, as shown in Fig.14 just


before their merger,

— countries A and E obviously could not have been identically coloured in


Mmin .

Bhupinder Singh Anand Why Four Colours Suffice


23: Is the flaw inherited in computer-dependent proofs of 4CT?

We conclude by merely noting, without further comment, that:

The flaw in Kempe’s 1879 ‘proof’ is apparently inherited in the


computer-dependent ‘proofs’ of the Four Colour Theorem:

— by Appel/Haken in 1977 [AH77], who seemingly argue that:


— If every member of an ‘unavoidable’ set of 1,482 configurations, in a minimal map Mmin , is
proven ‘reducible’ by a computer , then any configuration in Mmin consisting of a country F ,
with five neighbours, such that F necessarily requires a 5th colour, is ‘reducible’.

— by Robertson et al in 1997 [RSST], who argue similarly , but for an


‘unavoidable’ set of only 633 configurations.

Reason: Both arguments seem to appeal implicitly to the false premise:

A minimal map Mmin can be assumed to contain a country F , with five


neighbours, such that F necessarily requires a 5th colour.

If so, then both ‘proofs’ may need to be treated as vacuously true.

Bhupinder Singh Anand Why Four Colours Suffice


End

That concludes this overview of the computer-free argument* that seeks to


transparently illustrate

Why four colours suffice to colour any planar map


and
Why the perceived flaw in Kempe’s 1879 ‘proof’ of 4CT is not fatal

Thank you
*See [An22e] Bhupinder Singh Anand. 2022. Why the Perceived Flaw in Kempe’s 1879 Graphical ‘Proof’ of the Four
Colour Theorem is Not Fatal When Expressed Geometrically. Paper presented at PCVNCMS-2022. Link to preprint.

Bhupinder Singh Anand Why Four Colours Suffice

You might also like