Vous êtes sur la page 1sur 24


Monday Tuesday Wednesday Thursday Friday Saturday Sunday

28 29 30 31 1 2 3

G.B.M. Hopper
d. 1992
4 5 6 7 8 9 10

C. Herschel d. 1848
M.G. Agnesi d. 1799 R. Moufang b. 1905
11 12 13 14 15 16 17

S. Kovalevskaya
G.M. Cox b. 1900 b. 1850
18 19 20 21 22 23 24

25 26 27 28 29 30 31

S.A. Janovskaja
b. 1896


The Friendship Theorem In a finite graph in which any two distinct vertices share exactly one common
neighbour some vertex is adjacent to all other vertices.
Almost more interesting than the theorem
. . . in other words, if every pair of people at itself is the fact that it is false for infinite
a party shares exactly one mutual graphs. Start with G0 = C5 , the penta-
friend at the party then some gon, as shown on the immediate left.
guest is a friend of every- Now construct Gn+1 by creating
body present. More is a new vertex adjacent to each
is true: friendships pair of vertices in Gn having
must be arranged in no common neighbour. If
triangles which in- we repeat this indefin-
tersect only in the ately the result is
universally popular an infinite graph
guest. So the graphs satisfying the hy-
satisfying the hy- pothesis of the
pothesis of the theo- theorem (for any
rem are those shown pair of vertices
on the immediate we have, by defi-
right, usually nition, ensured a
known as common neigh-
‘windmill’ bour) but not the
graphs. conclusion (since
Despite it is clearly not a
the almost windmill graph).
trivial nature The graphs G0 ,
of these graphs G1 and G2 are
the Friendship The- shown on the
orem seems surpris- left; G3 is shown
ingly resistent to dir- This theorem was first proved by Paul in the centre.
ect combinatorial proof; Erdős, Alfréd Rényi and Vera Sós in 1966. A Challenge: try
and yet it emerges from a to find a formula for the
few lines of linear algebra
Although many other proofs have been
number of vertices in Gn : it is
as though it were the most discovered, theirs remains the one 100 for G3 while for G4 the number is 3695.
natural thing in the world. ‘from the book’.
Web link: the ‘book’ proof is well described here: www.princeton.edu/∼jacobfox/MAT307/lecture20.pdf; for
the fascinating full story see this MS Dissertation by Katie Leonard: www.mth.pdx.edu/∼caughman/Katie.pdf.
Further reading: Proofs from the Book, by Martin Aigner and Gunter M. Ziegler, Springer-Verlag, Berlin, 3rd Revised Edition, 2003, chapter 34.
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
1 2 3 4 5 6 7

8 9 10 11 12 13 14
S. Kovalevskaya
d. 1891
V. Malone-Mayes
b. 1932 H. Neumann b. 1914
15 16 17 18 19 20 21

R. Péter d. 1977 R. Péter b. 1905 E. Klein b. 1910

22 23 24 25 26 27 28

R. Gentry b. 1862


The Transversal Matroid Theorem Let E be a finite set and let A = {Ai | i = 1, . . . , t} be a family of
subsets of E. Let T denote the collection of partial transversals of A, i.e. those subsets X of E having
the property that each element of X may to choose a distinct member of A to which it belongs (thus if
X = {x1, . . . , xk } then there is a set Y ⊆ {1, . . . , t} also having k elements, y1, . . . , yk , say, and satisfying
xi ∈ Ayi , i = 1, . . . , k). Then T forms the collection of independent sets of a matroid.

1 2 3 4 5 6 7
 
R  r1 r3 
 
P  p1 p2 p4 p6 
Y  y2 y5
 
 
H  h1 h3 h4 
B  b1
 
 
N n2 n4 n5 n6 n7
The above incidence structure shows which
workers (1, 2, . . . , 7) may be assigned to which
jobs (R, . . . , N). The entries, r1 , p1 , etc, are
indeterminates—they allow us to express any as-
signment as a ‘valueless’ product. For instance, the
encircled assignment is expressed as r3 p6 y2 h4 b1 n7 .
This assignments gets all 6 jobs done at a cost of
e15. But can it be done cheaper?
In the illustration above, each of the six remaining Millennium Problems of the Clay Mathematics Institute is to be assigned to a different subcontractor for Edsger
Dijkstra’s fictitious Mathematics Inc. company. Obviously we want to solve all six problems as cheaply as possible! The greedy approach is to always to take the
first, cheapest option: this would give us an assignment starting r1 p2 h3 y5 . But now we are stuck because only subcontractor 1 can be assigned to problem B, and we
used her for problem R; we will have to backtrack. Thus we cannot match up subcontractors to problems greedily. To say that partial transversals give a matroid is
precisely to say that at least a cheapest maximal transversal can be selected greedily. That is, we can choose the cheapest subcontractors first and make the assignment
to problems later. This may be more a difference in mathematics than in practice: it leaves open the question of how to build transversals without finding a matching.
This theorem was proved in the mid-1960s by Jack Edmonds and Delbert Ray Fulkerson in the USA and, independently (and
also in an infinite version) by Leon Mirksy and Hazel Perfect at Sheffield University in the UK.
Web link: www.math.lsu.edu/˜oxley/dominic.pdf. Read about Dijkstra’s Mathematics Inc. at www.cs.utexas.edu/˜EWD/ (see e.g. no. 1224).
Further reading: Introduction to Graph Theory, 4th Ed., by Robin Wilson, Longman, 1996, chapters 8 and 9 (look out for 5th Ed., due in 2010!)
MARCH 2010
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
1 2 3 4 5 6 7

C. Ladd-Franklin
d. 1930
8 9 10 11 12 13 14
Women’s Day
O.C. Hazlett d. 1974 A.S. Baxter d. 1917
15 16 17 18 19 20 21

G. Chisholm Young
b. 1868 C.L. Herschel b.1750 A.S. Baxter b. 1870 S. Macintyre d. 1960
22 23 24 25 26 27 28

H. Geiringer von A.J.P. Wheeler

Mises d. 1973 E. Noether b. 1882 d. 1966
29 30 31 1 2 3 4

G. Chisholm Young
d. 1944 H.A. Merrill b. 1864


The McIver–Neumann Half-n Bound Let Ω be a set of order n, n , 3, and let G be a permutation group
acting on Ω. Then G can be generated by ⌊n/2⌋ elements.

The generation of a permutation group from a set of permutations is illuminated by the wonderful
algorithm which Charles Sims derived from a 1927 lemma of Otto Schreier. In the illustration
above, the algorithm is applied to a set of two permutations. The main work is done by the
Sift routine, which first places (1 2)(3 4) into row 1 column 2 cell in the 5 × 5 table M. The
next permutation σ = (1 2 3 5) is a candidate for the same cell since σ(1) = 2; since the cell is
occupied by (1 2)(3 4) we calculate σ := σ × M12 = (1 2 3 5) × (1 2)(3 4) = (2 4 3 5). The new
cell M24 is now indicated for σ because now σ(2) = 4. Subseqently, the queue is augmented in
the main algorithm by pre- and post-multiplying (2 4 3 5) with each non-identity entry of M. The
queue now begins with (1 2)(3 4) × (2 4 3 5) = (1 4 5 2) which Sift places directly into position M1,4 . Although the queue tends to grow rapidly it must eventually
become empty; in the current example the algorithm terminates with table shown above, bottom-right. This table describes the generated group in the following
way: the order of the group is the product of the numbers of permutations in each row, in this case 5 × 4 × 1 × 1 × 1 = 20 (the group happens to be the Frobenius
group of order 20, aka the Galois group of the polynomial x5 − 2); membership of the group is tested by running Sift on a candidate permutation σ which will be
allocated to an empty table location if and only if it does not belong to the group. Exercise: try the algorithm with Ω = {1, . . . , 8} and S = {(1 2), (3 4), (5 6), (7 8)}.
It is easy to generate large groups from a very few permutations; n/2 permutations may sometimes be necessary as the Exercise
above easily shows. But Annabelle McIver and Peter M. Neumann’s 1987 theorem seems to be deep and mysterious: it rests
on the Classification of the Finite Simple Groups and suggests no obvious way of finding a generating set meeting its bound.
Web link: www.maths.qmul.ac.uk/˜pjc/slides/pggt.pdf. Read about Frobenius 20 here: manu.amiot.free.fr/pdf/Articles/chords.pdf.
Further reading: Permutation Groups by P.J. Cameron, Cambridge University Press, 1999, sections 1.13 and 1.14.
APRIL 2010
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
29 30 31 1 2 3 4

M.L. Cartwright
S. Germain b. 1776 d. 1998
5 6 7 8 9 10 11

C.C. Krieger Dunaij

b. 1894
12 13 14 15 16 17 18

A.I. Maddison
b. 1869 E. Noether d. 1935
19 20 21 22 23 24 25

S. Macintyre b. 1910
26 27 28 29 30 1 2


Heath’s Finitely Discontinuous Function Theorem Let G and H be graphs whose edge and vertex sets have
cardinality eG and eH , and vG and vH , respectively. There is a k-to-1 finitely discontinuous function from G
onto H if and only if
eG − vG ≤ k(eH − vH ),
with equality being required to hold in the case k = 2.

For the
graph edges
with distinct endpoints are homeomorphic images of the interval [0,1] and loops are
homeomorphic images of the circle. For the graphs G and H above left we get a strict
inequality eG − vG = −1 < 0 = 2(eH − vH ) so the theorem says it is impossible for H (i.e.
the space of all points in all edges of H) to be the image of a finitely discontinuous function
with every point the image of exactly k = 2 points of G. But for k = 3 a continuous function is
possible: the line wxyw represents H by identifying the points w; the line ab is repeatedly folded
back on itself, the lengths of the folds tending to zero (a device affectionately termed a ‘wiggle’ by the mathematician Anthony Hilton). For the graphs G and H above
right, the equality condition of the theorem is met so that k = 2 is possible. This is achieved in the drawing using three discontinuities (represented by unfilled vertices ).

This surprising graph-theoretic answer to a subtle topological question, reminiscent of Kuratowski’s Theorem, is due to Jo
Heath, 1988. In the case where G is a single edge it generalises 1930s results of Orville G. Harrold and George Earl Schweigert.
Web link: arxiv.org/abs/math/9810129
Further reading: The Foundations of Analysis: a Straightforward Introduction. Book 2 Topological Ideas by K.G. Binmore, CUP, 1981.
MAY 2010
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
26 27 28 29 30 1 2

H.A. Merrill d. 1949

3 4 5 6 7 8 9

A.J.P. Wheeler
b. 1883
10 11 12 13 14 15 16

E. Kramer Lassar
b. 1902 E. Lutz b. 1914 M.G. Agnesi b. 1718
17 18 19 20 21 22 23

I. Flügge-Lotz
d. 1974
24 25 26 27 28 29 30

31 F.J. MacWilliams
d. 1990


The Classification of Archimedean 4-Polytopes The Archimedean polytopes in four dimensions are:
1. the polytopes of Boole Stott–Wythoff type;
2. the prismatic polytopes;
3. Cartesian products of regular polygons;
4. Thorold Gosset’s snub-24-cell (1900);
5. the Grand Antiprism (Conway–Guy, 1965)

Dwellers in a two-dimensional world might investigate the properties of a

Platonic solid, such as the octahedron, by examining the cross-sections pro-
duced when a plane intersects it. Above, a plane intersection parallel with one
of the faces produces a hexagonal section; a different intersection might reveal a
square or a rectangle. Archimedean four-dimensional polytopes consist of copies
of prisms and Archimedean solids meeting at vertices in identical configurations
(more precisely, the symmetry group of the polytope acts transitively on its ver-
tices). A drawing by Boole Stott is adapted, right, to show how a regular polytope,
the 24-cell, consisting of 24 octahedra with six meeting at each of its at 24 vertices,
may be ‘unfolded’ so that 3-dimensional sections through it may be visualised.
Following pioneering work by Alicia Boole Stott, Willem Wythoff and Thorold Gosset in the early 1900s,
work on polytopes was systematised by H.S.M. Coxeter in the 1940s. The classification of Archimedean
4-polyopes was completed by J.H. Conway and Michael Guy in the 1960s using computer search.
Web link: www.liga.ens.fr/∼dutour/Regular/. Fine descriptions of Boole Stott’s work are dissertations.ub.rug.nl/faculties/science/2007/i.polo.blanco/ by
Irene Polo-Blanco (ch. 5) and www.ams.org/featurecolumn/archive/boole.html by Tony Phillips (see also his book review here: www.ams.org/notices/200704/).
Further reading: The Symmetries of Things by John H. Conway, Heidi Burgiel and Chaim Goodman-Strauss, A.K. Peters, 2008, Chapter 26.
JUNE 2010
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
31 1 2 3 4 5 6

7 8 9 10 11 12 13

C.A. Scott b. 1858 V. Malone-Mayes

A.Boole Stott b. 1860 d. 1995
14 15 16 17 18 19 20

L.S. Hay b. 1935

21 22 23 24 25 26 27

S. Germain d. 1831
28 29 30 1 2 3 4


A Theorem on Rectangular Tensegrities Let L be a rectangular framework consisting of m rows and n

columns of square cells constructed from rigid rods connected to each other by pivot joints, allowing
rotation of the rods in the plane. Now suppose that in certain of the square cells ‘non-stretching’ cables
are placed between either or both pairs of diagonally opposite points, preventing these pairs from moving
further apart. Then the framework L is rigid in the plane if and only if the following bipartite directed
graph D(L) is strongly connected: D(L) has a vertex for each row and column and row i has a directed
edge to (from) column j if and only if the corresponding cell has a top-left to bottom-right (bottom-left to
top-right) cable.
Is this framework rigid? In more for-
mal terms we are asking if any contin-
uous transformation of the structure
must be isometric (distance preserv-
ing). Given that cables can bend, the
answer is NO and indeed, the trans-
formation on the far right is clearly
not isometric. How can we add extra
cables to prevent this? The theorem
gives an unexpectedly discrete answer
to this continuous problem.

On the left we construct the digraph D(L) specified by the theorem: bottom-left-top-right ca-
bles (red) are directed upwards; top-left-bottom-right cables (blue) are directed downwards;
the black double-headed arrows represent cells in which both red and blue cables are present.
D(L) is strongly connected if any vertex can reach any other along a consecutive sequence of
directed edges. And we can see that this condition fails because no edges exit from the ver-
tex set {R1 , C1 , C3 }. However, a blue edge directed from R1 to C2 is enough to bestow strong
connectivity; so a blue cable in the cell in row 1 and column 2 will make the framework rigid.
Tensegrity was the name given by Buckminster Fuller to frameworks of rods
and cables. The undirected version of this theorem, in which diagonal cables are replaced by rigid rods, was discovered by
Ethan Bolker and Henry Crapo in 1977. The generalisation stated above is due to Jenny Baglivo and Jack Graver in 1983.
Web link: www.math.cornell.edu/˜connelly/BasicsI.BasicsII.pdf (900K).
Further reading: Counting on Frameworks by Jack E. Graver, Mathematical Association of America, 2001.
JULY 2010
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
28 29 30 1 2 3 4

5 6 7 8 9 10 11

E. Kramer Lassar
d. 1984
12 13 14 15 16 17 18

I. Flügge-Lotz
b. 1903
19 20 21 22 23 24 25

26 27 28 29 30 31 1

J.B. Robinson
d. 1985 E. Lutz d. 2008


The Lecture Hall Partition Theorem A list (h1, h2, . . . , hN ) of non-negative integers is said to be a
lecture hall partition if
h1 h2 hN
0≤ ≤ ≤ ... ≤ .
1 2 N
For a fixed N (the length of the lecture hall), the number of lec-
ture hall partitions of any positive integer n equals the number
of partitions of n into odd parts smaller than 2N.
The lecture hall inequalities ensure that each row in a tiered lecture hall will have a clear
view of the lecturer. The hall shown on the left has rows at height 1, 3, 5 and 7; since
1 ≤ 3/2 ≤ 5/3 ≤ 7/4, the partition of 16 given by 1 + 3 + 5 + 7 is a lecture hall
partition of length N = 4. The partition of 16 given by 1 + 4 + 5 + 6, on the
other hand, represents a lecture hall in which the third and fourth row views are
impeded. A famous partition identity of Leonhard Euler says that the partitions
of n into distinct parts are equinumerous with those into odd parts; because
Lecture Hall Partitions permit no equal (non-zero) parts, today’s theorem is a
version of Euler’s identity with restricted maximum odd parts. If we allow
the lecture hall to become infinitely long (removing the limit on N) we
recover Euler’s identity in the limit. Here are the 21 lecture halls of
length N = 4 partitioning n = 16 (note, that zeros are allowed):
0, 0, 0, 16 0, 0, 1, 15 0, 0, 2, 14 0, 0, 3, 13 0, 1, 2, 13 0, 0, 4, 12 0, 1, 3, 12
0, 0, 5, 11 0, 1, 4, 11 0, 2, 3, 11 0, 0, 6, 10 0, 1, 5, 10 0, 2, 4, 10 1, 2, 3, 10
0, 1, 6, 9 0, 2, 5, 9 1, 2, 4, 9 0, 2, 6, 8 0, 3, 5, 8 1, 2, 5, 8 1, 3, 5, 7
The idea of a lecture hall partition arose unexpectedly in
Kimmo Eriksson’s work on Coxeter groups. He and Mireille
Bousquet-Mélou proved the above (apparently difficult) iden-
tity and several deep generalisations in two influential papers in
the late 1990s.
Web link: www-igm.univ-mlv.fr/∼fpsac/FPSAC08/fpsac08.html: the ‘List
of Presentations’ has a fine (but 3MB) talk by Carla Savage. (There is also a
talk by Bousquet-Mélou although not on partitions).
Further reading: Integer Partitions, 2nd revised ed. by George E. Andrews
and Kimmo Eriksson, Cambridge University Press, 2004, chapter 9.
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
26 27 28 29 30 31 1

E.M. Nelson d. 1987

2 3 4 5 6 7 8

M.S. Rees b. 1902

9 10 11 12 13 14 15

16 17 18 19 20 21 22
International Conference of Women
Mathematicians, Hyderabad

C.C. Krieger Dunaij

d. 1974
23 24 25 26 27 28 29

30 31
O. Taussky-Todd
b. 1906 R.A. Bari d. 2005 E. Klein d. 2005


Quadratic Nonresidue is Zero-Knowledge Provable There exists a protocol whereby an oracle, given
positive integers x < y, y coprime to x, can respond Yes or No to the question “Is y is a quadratic nonres-
idue mod x?” and impart no other information that is not polynomial-time computable from x and y.
To check if z2 = y mod x for
some z < x, that is, to check
if y is a quadratic residue or a
quadratic nonresidue, appears to
be as hard as factoring x. The
protocol described below and il-
lustrated on the right is probabilis-
tically polynomial for quadratic
nonresidue. To simplify notation
‘quadratic nonresidue’ is replaced
by ‘square root’.
The Basic Idea∗
Victor (the ‘verifier’) can com-
pute squares but not square roots.
The (mathematically) omniscient
Peggy (the ‘prover’) will reveal
whether or not y is a square num-
ber. Victor secretly chooses a ran-
dom integer r and tosses a coin. If
the coin is Heads he shows Peggy

w = r2 ; if Tails he shows her w = The Small Print
r2 y. Now he asks Peggy: “Was my What if Victor tries to exploit Peggy’s omniscience? The delicate exchange illustrated above forces Victor to prove he is adhering
coin Heads or Tails?” If y is not to the protocol. He produces a series of random pairs (r1 , r2 ) and uses them to make ‘dummy’ w pairs which he shows (the π’s) to
square then Peggy answers infalli- Peggy, with their order reversed when a tossed coin (coin2) is Tails. Peggy tosses a coin (coin3) and demands further information:
bly, since w is a square if and only Tails: she must see (r1 , r2 ) and she confirms this pair by checking it contains the square root of one of the π entries; Heads: she must
if the coin was Heads. However, see ri w/r and she confirms this by checking it is a square root of one of the entries of π × w. These checks are enough to confirm that
if y is a square then Peggy cannot w is either r2 or r2 y but not enough to reveal which one.
tell whether w incorporates y or not Zero-knowledge provability was introduced by Shafi Goldwasser, Silvio Micali and Charles Rackoff in the mid-
and will answer falsely with prob- 1980s. Their work was one of the catalysts for a decade of breakthrough results in computability culminating in
ability 1/2. Now performing the the celebrated PCP Theorem. They shared the first Gödel Prize in 1993 with Lazlo Babai and Shlomo Moran.
exchange 20 times will reduce the Web link: www.math.clemson.edu/∼rheindl/crypto/zeroknowledge.pdf
chance of a consistently false an- Further reading: The Nature of Computation by Stephan Mertens and Chris Moore, Oxford University Press, 2010 (preview at
swer to 1/220 ≈ 1 in a million. www.nature-of-computation.org).
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
30 31 1 2 3 4 5

6 7 8 9 10 11 12

E. du Châtelet
W.E. Merrill d. 1951 d. 1749
13 14 15 16 17 18 19

20 21 22 23 24 25 26

E.A. Hayes b. 1851 W.E. Merrill b. 1862

27 28 29 30 1 2 3

H. Geiringer von
Mises b. 1893


Praeger’s Theorem on Bounded Movement Let G be a permutation group acting without fixed points
on a set Ω. Denote by m the maximum size of any subset Γ ⊆ Ω whose image under some group
element is disjoint from Γ; and suppose that m is finite. Then Ω is a finite set; the number t of G
orbits is at most 2m − 1; each orbit has length at most 3m; and |Ω| ≤ 3m + t − 1 ≤ 5m − 2.

The two examples above will serve to illustrate a couple of points of interest. One the left we see that m depends not only on the group G
but also on its action; the cyclic group C5 acts transitively on the pentagon: any point may be rotated to any other. The action has bounded
movement with m = 2 since any three points will intersect with themselves under rotation. If instead C5 acts on the ten unordered pairs of points
then m = 4 (e.g., Γ = {12, 13, 34, 35} becomes disjoint from itself under rotation). This latter action has two orbits and we can check that indeed
|Ω| = 10 ≤ 3m + t − 1 = 12 + 2 − 1 = 13, as ordained by the theorem. The bound of |Ω| ≤ 5m − 2 becomes important when we do not know the
value of t. Subsequent work by Praeger with Peter M. Neumann reduced it to |Ω| ≤ (9m − 3)/2, exhibiting moreover infinitely many examples
of groups which attain this new bound. Notice, however, that our transitive C5 action fails to meet the length bound of 3m on its single orbit.
The same is true for the transitive point action of the rotational symmetry group of the tetrahedron, isomorphic to the alternating group A4 as
shown above right. We have m = 2 and |Ω| = 4 < 3m = 6. However, the transitive action of A4 on the six edges of the tetrahedron again has
m = 2 and does attain the bound; such transitive actions have been completely determined by A. Gardiner, Avinoam Mann and Praeger.
This 1991 theorem of Cheryl Praeger links back to work by BH Neumann in the 1950s. It has opened up a new line of enquiry
for permutation group theorists, some of this work being alluded to above.
Web link: www.math.helsinki.fi/EWM/meetings/ewm_praeger.pdf actually about regular permutation groups but with masses of extra goodies
Further reading: Permutation Groups by P.J. Cameron, Cambridge University Press, 1999.
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
27 28 29 30 1 2 3

4 5 6 7 8 9 10

M.C. Fasenmyer O. Taussky-Todd M.C.B. Weiss

b. 1906 d. 1995 d. 1966
11 12 13 14 15 16 17

R. Gentry d. 1917 G.M. Cox d. 1978

18 19 20 21 22 23 24

A.I. Maddison S.A. Janovskaja

d. 1950 d. 1966
25 26 27 28 29 30 31

E.A. Hayes d. 1930

M.S. Rees d. 1997 O.C. Hazlett b. 1890 L.S. Hay d. 1989


The Panarboreal Formula Let Tn denote the set of all unlabelled trees on n edges and denote by s(Tn)
the minimum number of edges which an (n + 1)-vertex graph must have in order that it contains every tree
in Tn as a subgraph. Then s(Tn) ∼ cn log n where c is a constant satisfying 1/2 ≤ c ≤ 5/ log 4.

There are 23 unlabelled trees having 7

edges; they are shown on the right,
lexicographically ordered by degree
sequence, together with an 8-vertex
graph with 13 edges in which each one
may be found as a subgraph. This appears
to be the largest n for which s(Tn ) has
been calculated exactly; but it is easy
to establish that s(Tn ) ≥ (1/2)n log n.
For, given k, 1 ≤ k ≤ n + 1 we may
always choose a tree in whose deg-
ree sequence the k-th entry ≥ n/k.
But now the same must hold for the
degree sequence of any graph G con-
taining this tree. So if G contains each
tree in Tn and has degree sequence, say,
(d1 , P
. . . , dn+1 ), then number of edges in G
1 Pn+1
= 12 n+1 k=1 kd ≥ 2
k=1 n/k > 2 n log n.

Fan Chung and Ron Graham proved the

easy lower bound on s(Tn ) and the difficult
upper bound of (5/ log 4)n log n + O(n) in 1979,
mentioning that 5/ log 4 ≈ 3.6067 could probably be
improved, possibly even down to the minimum possible value
of 1/2. This challenge has yet to be met.
Web link: math.ucsd.edu/∼fan/ (where all Chung’s papers may be found in pdf format)
Further reading: Erdős on Graphs: His Legacy of Unsolved Problems, by Fan Chung and Ronald Graham, AK Peters, 1998, section 3.5.1.
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
1 2 3 4 5 6 7

8 9 10 11 12 13 14

C.A. Scott d. 1931 H. Neumann d. 1971

15 16 17 18 19 20 21

R.A. Bari b. 1917

22 23 24 25 26 27 28

A.A.K. Byron
E.M. Nelson b. 1943 R. Moufang d. 1977 d. 1852

29 30 1 2 3 4 5

M.F.G. Sommerville
d. 1872


The Diaconis–Holmes–Montgomery Coin Tossing Theorem Suppose a coin toss is represented by:
ω, the initial angular velocity; t, the flight time; and ψ, the initial angle between the angular momentum
vector and the normal to the coin surface, with this surface initially ‘heads up’. Consider the pair (ω, t)
as a smooth, compactly supported random variable, and let the centre of its distribution tend to infinity
in the positive orthant (corresponding to large spin and long flight time). Then with ψ fixed, the limiting
probability p(ψ) of the coin landing heads is given by
 1 1 −1 π 3π
 2 + π sin (cot ψ) if 4 < ψ < 4

p(ψ) = 
 if 0 < ψ < π or 3π < ψ < π. 4 4

The key advance encapsulated in this theorem is the inclusion of precession: the normal N to the coin surface

rotates about the base of the angular momentum vector M as shown on the left. The angle ψ between the two vec-
tors is shown to remain constant. The effect on the orientation of the coin can be viewed in terms of a sphere, above
right, whose equator is the flat ‘heads up’ plane. Only if ψ is sufficiently large will the coin ever flip over into the tails
orientation. Note, in particular, that if ψ = π/2 then the coin spends an equal time heads up and tails up; if ψ < π/4 it will nev-
er be tails up and a heads outcome is a certainty. This is in accordance with the function p(ψ) whose graph is plotted above centre.
Persi Diaconis, Susan Holmes and Richard Montgomery published this analysis in 2007, together with supp-
orting data from detailed experiments. Although the heads bias seems conclusively demonstrated they remark
that, taking random external factors into consideration: “for tossed coins, the classical assumptions of indepen-
dence with probability 1/2 are pretty solid.”
Web link: www-stat.stanford.edu/∼susan/papers/headswithJ.pdf (5MB) (the above images are adapted from this wonderful paper; the images of the
US dollar coin are used courtesy of the United States Mint).
Further reading: The Feynman Lectures on Physics: Definitive and Extended Edition by Richard P. Feynman, Robert B. Leighton and Matthew Sands,
Addison Wesley, 2005, vol. 1, chapter 18.
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
29 30 1 2 3 4 5

C. Ladd-Franklin
b. 1847
6 7 8 9 10 11 12

J.B. Robinson G.B.M. Hopper A.A.K. Byron M.C.B. Weiss

b. 1919 b. 1906 b. 1815 b. 1930
13 14 15 16 17 18 19
E. du Châtelet
b. 1706
M.L. Cartwright
b. 1900
A.Boole Stott d.1940 K.S. Fenchel d. 1983
20 21 22 23 24 25 26

M.F.G. Sommerville
K.S. Fenchel b. 1905 b. 1780
27 28 29 30 31 1 2

M.C. Fasenmyer
d. 1996


The Albert-Brauer-Hasse-Noether Main Theorem Every finite-dimensional noncommutative division

algebra over a number field is a cyclic algebra.

A number field is a finite extension of the rationals Q; cyclic algebras are constructed on cyclic field extensions. For example, the equation x7 − 1 = 0 has
solutions, such as ω = e2πi/7 , not in Q. The extension Q(ω) is the smallest field containing Q and ω; it is cyclic of degree 6 because those of its symmetries
(permutations preserving the field structure) which leave unmoved all of Q form a cyclic group of order 6 (namely C6 = {1, τ, τ2 , . . . , τ5 }, τ : ω 7→ ω3 ). In
1914, L.E. Dickson, the inventor of cyclic algebras, observed that the square of τ, acting on α = ω + ω6 , also creates a cyclic group: with σ = τ2 , {1, σ, σ2 } =
C3 . Now α is an irrational number solving x3 + x2 − 2x − 1 = 0, so we get an intermediate extension Q(α) which is also cyclic. Dickson’s cyclic algebra
construction adds an indeterminate u, together with the rules: u3 = q, q ∈ Q and uα = (α2 − 2)u. This construction generalises to cyclic extensions of any
degree over any field; for appropriate choices of q the result is also a noncommutative division algebra: a vector space (necessarily having dimension a square)
with a multiplication which never gives zero for pairs of nonzero elements. Here, the result is a 9-dimensional algebra over Q, in which a typical element is
a0 + a1 α + a2 α2 + a3 u + a4 αu + a5 α2 u + a6 u2 + a7 αu2 + a8 α2 u2 , ai ∈ Q.
Richard Brauer, Helmut Hasse and Emmy Noether announced this result in 1931 with the American A. Adrian Albert close
behind. Some of the key ideas were anticipated by Emil Artin’s student Käte Hey in her 1927 PhD thesis.
Web link: www.maths.tcd.ie/pub/ims/bull57/: click on the excellent survey article by David W. Lewis.
Further reading: The Brauer-Hasse-Noether Theorem in Historical Perspective by Peter Roquette, Springer 2004. Dickson’s degree-3 cyclic algebra
is described in chapter 5 of A First Course in Noncommutative Rings by Tsit-Yuen Lam, Springer, 2009.