Miniature geometries

Author(s): BURTON W. JONES

Source: The Mathematics Teacher, Vol. 52, No. 2 (FEBRUARY 1959), pp. 66-71
Published by: National Council of Teachers of Mathematics
Stable URL: http://www.jstor.org/stable/27955823
Accessed: 13-12-2015 15:26 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/27955823?seq=1&cid=pdf-reference#references_tab_contents

You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact [email protected].

National Council of Teachers of Mathematics is collaborating with JSTOR to digitize, preserve and extend access to The
Mathematics Teacher.

http://www.jstor.org

This content downloaded from 129.16.69.49 on Sun, 13 Dec 2015 15:26:08 UTC
All use subject to JSTOR Terms and Conditions
Miniature geometries
burton w. University of Colorado, Boulder, Colorado.
jones,
Recent interest in emphasizing "patterns" in high schoolmathematics
may lead some teachers to considerminiature geometries
as suitable for a unit for gifted students.

In a previous article1 miniature number Notice that in the first three statements
systems were developed?that is, number we set up a terminology. This is especially
systems which have most of the usual true of the third statement. We could
properties of numbers with which we are merely write "L contains p" or "p is con
familiar but which contain only a finite tained in L," but ifwe are to get help from
number of numbers. Similarly, there are geometrical visualization (and this is our
geometries that are "miniature" in the chief guide in this development) it is use
sense that they contain only a finite num ful to preserve as much of the geometrical
ber of points. Here, however, we are in the terminology as possible. Property 4, on the
beginning forced to make a number of other hand, is the first really restrictive
choices: Is it to be plane geometry, solid requirement.

geometry, or a geometry of many dimen It to impose two other re

is natural
sions? Is it to be Euclidean, projective, or quirements, the first to make it a finite
any one of the other geometries we know geometry and the second to keep it from
about? being too trivial :
Our first arbitrary choice is to confine
5. The geometry contains only a finite
ourselves here to plane geometry. That
number of points.
being the case, the following requirements 6. The geometry contains four points, no
are quite natural:
three of which lie on a line.
1. The geometry consists of a set of unde
fined elements called points. Now we are faced with our second
2. It contains certain subsets (smaller sets) choice. If the geometry were to be Euclid
of points called lines. ean2 we would specify:
3. If L is a line and is a point in the set
E7. If L is a line and a point not on L,
of points which comprise L, we call
"a point of L" or write "p lies on L." there is exactly one line Lp through
The same idea is expressed by writing which has no points in common with
that "L contains p" or "L passes L. Lp is called "parallel to L."

through p."
4. Any two points "determine a line"; On the other hand, if the geometry were
that is, given any two points and q, to be projective we would specify:
there is exactly one line L passing P7. of lines has a common
Every pair
through these two points. point.
1Burton W.
Jones, "Miniature Number Systems,"
2R. H. in the Founda
The Mathematics Teacher, LI (April 1958), 226 Bruck, "Recent Advances
231. This will hereafter be referred to as "Article I," tions of Euclidean Plane Geometry," The American
and its contents will be presupposed in much of this Mathematical Monthly, 62 (August-September, II,
article. 1955), 2-17.

66 The Mathematics Teacher | February, 1959

This content downloaded from 129.16.69.49 on Sun, 13 Dec 2015 15:26:08 UTC
All use subject to JSTOR Terms and Conditions
 In this article we make the second some point not on either line. Let p0 be
choice since it is simpler in a number of the point common to L0 and L and pi,
???
respects; ifwe are to manufacture a geom P2, , Pm the other points on L0. Then
etry we might as well make it as simple as the ra lines determined by and the points
possible. Notice that Property 4 with on L0, except p0, will intersect L inra dis
Property P7 affirms that any pair of dis tinct points. Thus L has at least as many
tinct lines has exactly one common point. points as L0. By reversing the argument
we can show that L0 has at least as many
A finite geometry
points as L and hence both lines have the
with seven points
same number of points.
Before considering the general theory, Next we can show that ra+1 is the
let us look at one particular finite geome number of lines through a point. Given a
try : one with seven points and seven lines. point and some line L not through p,
(This is actually the "smallest" geometry.) it follows that L has ra+1 points and that
Number the points 0, 1, 2, 3, 4, 5, 6 and each of these with determines a line
the seven lines may be taken as the fol through p. But any line through inter
lowing seven sets of three points each: sects L in one of these points. Hence there
L0:0, 1, 3; I*: 1, 2, 4; L2: 2, 3, 5; L3:3, 4, 6; are just as many lines through as there
L4:4, 5, 0; L5:5, 6,1; L6: 6, 0, 2. Inspection are points on L, that is, there are ra+1
shows that each pair of lines has exactly lines through p.
one common point and that each pair of Third, we can count the number of
points determines a line. Also each point points in the geometry. Given a point p,
has three lines through it, and each line every point of the geometry will lie on a
contains three points. This can be seen line through p. Each such line contains ra
from the sets of points above or from the points besides and there are ra+1 such
diagram. lines. Hence the total number of points is

4 ra(ra+l) + l = ra2+ra+l.
In a similar fashion it can be shown that
there are ra2+ra+l lines in the geometry.
Notice that 7 is the least number of lines
in a finite geometry satisfying the proper
ties listed above since that is the value of
ra2+ra+l when ra = 2 and ifra=l there
would be only three points in the geometry
denying Property 6.

Figure 1 Duality

Perhaps the most important conse

Notice that one of the lines, L5, is not quence of our choice in Property P7 is the
straight, but recall that the lines are sets principle of duality: any true statement
of points and the only purpose of drawing about lines and points is also true when
the lines is to indicate the sets of points "line" and "point" are interchanged and
which comprise them. the corresponding change in connective
used. For instance, "any line has the same
Counting lines and points
number of points as any other line" implies,
More generally, suppose one line con by duality, "any point lies on the same
tains ra+1 points. First we shall show number of lines as any other point." This
that every line contains exactly ra+1 principle seems like something too general
points. Let L0 be the given line and L any to establish until we realize that by virtue
other line. From Property 6 there will be of the fact that we assume only Properties

Miniature geometries 67

This content downloaded from 129.16.69.49 on Sun, 13 Dec 2015 15:26:08 UTC
All use subject to JSTOR Terms and Conditions
1 to 6 and P7, we need merely verify dual same point. For example, (1, 1, 0) and
ity for these seven properties. This can (?1, ?1, 0) represent the same point.
easily be done. The number of triplesexcluding (0, 0, 0)
Suppose we consider "another" geome is 33?1 = 26 and the number of propor
try defined by interchanging "point" and tionality factors different from zero is
1 through 6 and ? the number of points in
"line" in the Properties 3 1=2. Hence
P7, that is, a dual geometry. We now this geometry is 26/2 = 13. Notice that
show that this will have all the properties 13= ra2+m+l form = 3. Then the follow
of the given geometry. In this dual geome ing triples will represent the points of this
try, a point would be identified with the geometry :
set of lines through it. Property 3 is its
P0: (0, 0, 1); Pi: (0, 1, 0); P2: (1, 0, 0);
own dual, and Property P7 is the dual of
P3:(0, 1, -1);P4:(1, -1, 0); :(-1, 1,
Property 4 if included in P7 is the itali -1);P6: (1, 1, 1); P7: (1, -1, -1); P8:
cized statement at the end of the first
0, -1); P9: (0, 1, 1); P10: (1, 1, 0);
section. If the geometry contains only a
Pu: (1, 1, -l);Pi2: (1, 0, -1).
finite number of lines, it must contain
Now we must determine the lines. Let
only a finite number of pairs of lines, and
the point be associated with the ordered
hence a finite number of points. =
It remains to consider the dual of triple ? (a, 6, c) and the point P' with
?'=(a', e'). Then the line determined
Property 6, namely 6d : the geometry con
by and P' will be definedto be all those
tains four lines, no three of which have a
points associated with the triples rt+r't',
common point. To show that this implies
that is
Property 6, letPa& be the point determined
(ra + rV, rb + r'6', rc + r'c')
by lines a and 6, Pbc the point determined
by lines b and c, and similarly define where r and r' range over the numbers of
da-Suppose Pab, Pbc and PCd were = = =
PCd and GF(3) excluding r 0 r'. Thus, ifP P0,
collinear. The first two of these lie on the P' = Ph r= 0, r' = l yields the point P0
line b and the second two on the line c; above; r=l, r' = 0 yields Pi; r' = l=?r
moreover they determine the lines b and c, yields P3, and r=l=r' yields P9. (Notice
respectively. But b and c are distinct that r=?1, r' = l, being proportional to
lines and hence Pab, Pbc and PCd are not 1 and ?1, yield no new point.) We call
collinear. In similar fashion any three of L0 the line composed of these four points.
the four designated points can be shown to It can be seen that the quadruples of
be noncollinear. Thus we have shown that subscripts of points associated with the
a geometry which satisfies the dual prop lines of this geometry will be :
erties, satisfies the given ones. In similar L0: (0, 1,3, 9); Li: (1, 2, 4, 10); L2: (2,3, 5,
fashion one may show that a geometry
11); L3: (3, 4, 6, 12); L4: (4, 5, 7, 0); L6:
which satisfies the given properties, satis
(5, 6, 8, 1);L6:(6, 7, 9, 2);L7:(7, 8, 10,3);
fies the dual ones. Hence the two geome
L8: (8, 9, 11, 4); L9: (9, 10, 12, 5); L10:
tries are the same, and the principle of
(10, 11, 0, 6); Lu: (11, 12, 1, 7); Li2: (12,
duality is established.
0, 2, 8).
Construction of It can be shown without much trouble
finite geometries for m = pn that this geometry satisfies the required
= 3.
First, let us illustrate the construction properties of a finite geometry with m
m =
for = 3, n = l, m = 3. Associate with each Similarly if 2, there will be seven
point an ordered triple (a, 6, c) where points determined by the triples:
a, bj c are numbers of GF(3) (see Article P0: (0, 0, 1); : (0, 1,0); P2: (1, 0, 0); P3:
I), not all are zero, and the triples (a, 6, c) (0, 1, 1); P4: (1, 1, 0); 5' (1, 1, 1); P6:
and (kdj kb} kc) are associated with the (1, 0, 1).

68 The Mathematics Teacher | February, 1959

This content downloaded from 129.16.69.49 on Sun, 13 Dec 2015 15:26:08 UTC
All use subject to JSTOR Terms and Conditions
The lines will be those given above, and ti= (di, bi, Ci) ; tj= (ay, by,cy) ; tk=(dk, bk, ck),
the numbers of the triples will be inGF(2).
where the a's, b's and c's are numbers in
In general, then form = pn we consider
GF(ra). Suppose the points represented
the triples (a, 6, c) where a, 6, c range over
not all are zero, by these triples lie on a line, that is:
the numbers in GF(ra),
two triples in which a, 6, c are pro ak= ra,i+say, = =
and b* rbt+sby, ck rd+scj,
portional are associated with the same where r and are numbers of GF(ra) and
?
point. Then there will be ra3 1 triples
not both are zero. Then x^diX^+biX+Ci
not all zero and ra?1 proportionality with similar expressions for k and j imply
factors different from zero; hence (ra3?1)
xk= rxi+sxJ' and thus, for every integer u,
/(ra?l)=ra2+ra+l points. The lines are xk+u
= Hence =
rxi+u+sx1+u. dk+u rdl+u
defined as sets of points, as in the above
+sdj+u similarly for b and c. Thus
and
example. This system can be shown to and the points Pk+U,
tk+u^rti+u+stj+u
have all the required properties of a finite
Pi+u, and Py+W are collinear.
geometry.
Thus we have shown why the Unes are
Determination of lines formed from the initial one by the process
by means of galois flelds described. There is one further question
which should be answered: Since GF(ms)
The reader may have noticed a curious
contains mz numbers, there will be m3
pattern in the two examples above for different powers of but only m2+m+l
ra = 2 and ra = 3 in the formation of the
different points; what is the explanation
successive fines. For instance, for ra = 2,
of this apparent discrepancy? The answer
the line L0 is the set of points P0, Pi, Pz
to this question is left to the interested
and the line Li isPi, P2, P4. The subscripts
reader.
of the points of the latter are those of the
In general, then, to form a co-ordinate
former increased by 1. The subscripts of we
system with co-ordinates in GF(ra),
the points in the line L2 are 2, 3, 5, those
may find a generator of GF(m3) (not
for L3 are 3, 4, 6 and for L4 are 4, 5, 0
necessarily as above) whose powers are
where 1 is replaced by 0, addition
6+
all quadratic polynomials with coefficients
being (mod 7). The same pattern holds in in GF(ra). We use the coefficients of the
the geometry with ra = 3.
polynomials corresponding to successive
Why does this pattern hold? To answer
powers of the generator, up to the power
this forra = 2, observe in a table given at
m2+m+l, as the co-ordinates of the suc
the end ofArticle I, that is a generator of
cessive points of our geometry. This ele
GF(8), that is, each power of will be of
gant device is due to James Singer.3
the form ax2+bx+c (mod 2, xz+x+l).
Furthermore, every number of GF (8) will Perfect difference sets

be a power of x. Thus xi will be expressible

Another approach to the problem of
in the form aix2+biX+Ci where a?, 6,?, ct
designating the points constituting the
are in GF(2). Hence we can associate x{
various lines of a finite geometry is by
with the point whose triple is (a,?, &,?,c?).
way of arithmetical properties of the set
Thus ?? = 0??2+1 is associated with P0:
of subscripts of the points of line L0. Con
x = 0 x+0 is associated
(0, 0, 1), x2+l sider the numbers 0, 1, 3 in the number
?? ? xA= with
with Px: (0, 1, 0), , x2+x+0
system (mod 7). The six differences are:
P4: (1, 1, 0), etc. Since is also a generator
1-0 = 1, 0-1=6, 3-1=2, 1-3 = 5, 3-0
of GF(27), a similar development holds =
3, 0-3=4 (mod 7). These are just the
for ra=3.

Now suppose that is a generator of 3James

Singer, "A Theorem in Finite Projective
Geometry and Some Applications to Number Theory,"
GF(ra3) and consider the following three American Mathematical 43
Transactions Society,
triples: (1938), 377-385.

Miniature geometries 69

This content downloaded from 129.16.69.49 on Sun, 13 Dec 2015 15:26:08 UTC
All use subject to JSTOR Terms and Conditions
numbers 1, 2, 3, 4, 5, 6. In general, a set example the finite geometry with seven
?
of numbers 0, 1, 2, ? ? ,m whose differ lines and seven points, suppose one wanted
ences (mod m2+m+l) are 1, 2, 3, ? ? ? , to test seven different varieties of seeds by
m2+m in some order is called a perfect planting. One might have seven different
difference set. Whenever such a set can plots of ground. Let these plots correspond
be found, it can be used as the subscripts to the lines of the geometry and the varie
for a set of points on a line which may be ties of seed correspond to the points. A
called L0. (For instance, form = 3, a per seed then would be planted in a given
fect difference set is 0, 1, 3, 9.) The other plot of ground if the corresponding point
lines then are gotten by increasing the sub occurred on the corresponding line. Then
scripts by 1, by 2, etc. It is not too difficult there would be a desirable symmetry of
to prove that such a difference set leads to treatment since each seed would be
a finite geometry. In fact, the first three planted in three different plots of ground,
properties of a finite geometry and the each plot of ground would contain three
fifth already hold. Hence it remains only different varieties of seed, and each seed
to show Properties 4, 6, and P7. Here we would be competing with each other seed
shall only give an indication of how to show in a plot of ground exactly once. Each
Property 4, leaving the rest to the reader. finite geometry yields a design with
Let m2+m+l be denoted by s and sup analogous properties.
pose
Desargtjes7 Theorem
? ? ?
a>0) a , , am
One of the most interesting theorems
is a perfect difference set (mod s). The of projective geometry is Desargues'
points with these subscripts will con Theorem. We shall show that this theorem
stitute the line L0, and the line Lu will holds in every finite geometry which has a
consist of the points whose subscripts co-ordinatization as described above. Re
are call that this theorem may be stated as
- ?? follows: If a, b, c and a', c' are two
ao+u, di+Uj , am+u (mod s)
sets of three points each such that the
where u takes the values 0, 1, 2, ? ? ? ,m. lines aa', bb', ccr are concurrent, then the
Let P? and Py be two points. We wish to points determined by the following pairs
show that they lie on one and only one of lines are collinear: ab, a'b'ac, a'c';
line L, that is, there is only one set of be, b'c'. To prove this let A, B, C and
three numbers: a*, ar, u such that i= ak+u A', Bf, C be triples corresponding to the
and j = ar+u. Since the a's form a perfect given points. Then, since aa', bb', cc' are
difference set, there will be exactly one concurrent, numbers rh r2, r3, si, s2, Sz can
? =
pair such that ak ar i?j (mod s). Then be chosen so that
there is exactly one u in the range indi =
nA +SlA' r2B+s2B' = nC+szC.
cated such that i ? au+u (mod s). For
example, ifm=3 and the difference set is These equations may also be written in
3, 9 with s= we find the
0, 1, 13, suppose the following form:
line determined by P5 and P9. Then ? A = '?
9-0 = 5-9 = r2B siA s2B',
(mod 13) shows that ak 9,
= 0 and 5 =
ar 9+u (mod 13) shows that riA-r3C=-SiA'+S3C',
u = 9. Hence the points P5 and P9 lie on L9.
r2B-nC=-s2B'+ssC.

Block designs
The left side of the first equation is a
Recent interest in finite
geometries triple associated with a point on the line
has arisen from their connection with de ab and the right-hand side is a triple as
signs used in experiments. Using as an sociated with a point on the line a'b'. Since

70 The Mathematics Teacher | February, 1959

This content downloaded from 129.16.69.49 on Sun, 13 Dec 2015 15:26:08 UTC
All use subject to JSTOR Terms and Conditions
these triples are equal, each represents a than a multiple of 4 and if it containsto
point which is on each of the lines and an odd power some prime factor which is
hence their point of intersection. Simi 1 less than a multiple of 4, there is no finite
larly the triple on each side of the second geometry with m+1 points on a line.
equation represents the point of inter Thus there is no finite geometry forwhich
section of ac and a'c', while the triples m = 6 since m contains to the first power
in the third equation represent the point the prime factor 3. But it is not known
of intersection of be and b'c'. But r%B whether there is a finite geometry for
? = m =10. In fact, the only finite geometries
?nA +nA r3C r2B?r3C shows the
third point lies on the line determined by known have m a power of a prime. There
the first two, that is, the points of inter is thus a wide gap between the values of
section of the following pairs of lines are m for which finite geometries are known
collinear: a&, a'bac, a'&';bc,b'&'. This is to exist and those values for which it has
what we wished to prove. The converse of been proved that no finite geometries exist.
this result may be proved in a similar The existence of this gap in our knowledge
manner or may be seen to follow from is the reason for much of the interest in
the principle of duality. this subject today.

The existence References

of finite geometries A. A. Bennett, "Modular Geometry," The
American Mathematical Monthly, 27 (1920)
Since finite fields with pn elements exist pp. 357-361.
for every prime and every natural num "Contributions to Geometry," The American
Mathematical Monthly, 62 (August-Septem
ber, one may, using the methods above, ber, 1955), Part II. This contains eight
show that a finite geometry exists having papers, the first four of which have to do
points on a line whenever m is a
with finite references are
m+1 geometries. Many
given.
power of a prime. Furthermore, if the
E. R. Ott, "Finite Projective Geometries,
geometry is constructed using the co PG(k, pn)," The American Mathematical
ordinatization above, Desargues' Theorem Monthly, 44 (1937), pp. 86-92.
C. E. Rickart, "The Pascal Configuration in
holds. There are, however, finite geome Finite The American
Projective Geometry,"
tries with m+1 points on a line and m Mathematical Monthly, 47 (1940), p. 89.
a power of a prime, forwhich Desargues' M. F. Rosskopf and R. M. Exner, "Modern
Emphases in the Teaching of Geometry,"
Theorem does not hold. The Mathematics Teacher, XL (1957),
On the other hand, ifm is 1 or 2 more pp. 272-279.

Any attempt to be precise and pay attention

to meanings would involve us with . . . prob
lems . . .which are really quite irrelevant for
mathematics. For mathematics, it is the form
that must be considered, and the meaning can
be dispensed with.?Taken from Logic for
Mathematicians by Rosser.

Miniature geometries 71

This content downloaded from 129.16.69.49 on Sun, 13 Dec 2015 15:26:08 UTC
All use subject to JSTOR Terms and Conditions