Chapter II.
The Definition of a Group.

In the present chapter we shall enter on our main subject and we shall begin with definitions, explanations and examples of what is meant by a group.

Definition. Let

A,B,C,
represent a set of operations, which can be performed on the same object or set of objects. Suppose this set of operations has the following characteristics.

(α) The operations of the set are all distinct, so that no two of them produce the same change in every possible application.

(β) The result of performing successively any number of operations of the set, say A, B, …, K, is another definite operation of the set, which depends only on the component operations and the sequence in which they are carried out, and not on the way in which they may be regarded as associated. Thus A followed by B and B followed by C are operations of the set, say D and E; and D followed by C is the same operation as A followed by E.

(γ) A being any operation of the set, there is always another operation A1 belonging to the set, such that A followed by A1 produces no change in any object.

The operation A1 is called the inverse of A.

The set of operations is then said to form a Group.

From the definition of the inverse of A given in (γ), it follows directly that A is the inverse of A1. For if A changes any object Ω into Ω, A1 must change Ω into Ω. Hence A1 followed by A leaves Ω, and therefore every object, unchanged.

The operation resulting from the successive performance of the operations A, B, …, K in the sequence given is denoted by the symbol ABK; and if Ω is any object on which the operations may be performed, the result of carrying out this compound operation on Ω is denoted by Ω ABK.

If the component operations are all the same, say A, and r in number, the abbreviation Ar will be used for the resultant operation, and it will be called the rth power of A.

Definition. Two operations, A and B, are said to be permutable when AB and BA are the same operation.

13. If AB and AC are the same operation, so also are A1AB and A1AC. But the operation A1A produces no change in any object and therefore A1AB and B, producing the same change in every object, are the same operation. Hence B and C are the same operation.

This is expressed symbolically by saying that, if

AB = AC,
then
B = C;
the sign of equality being used to imply that the symbols represent the same operation.

In a similar way, if

BA = CA,
it follows that
B = C.

From conditions (β) and (γ), AA1 must be a definite operation of the group. This operation, by definition, produces no change in any possible object, and it must, by condition (α), be unique. It is called the identical operation. If it is represented by A0 and if A be any other operation, then

A0A = A = AA0,
and for every integer r,
A0r = A 0.

Hence A0 may, without ambiguity, be replaced by 1, wherever it occurs.

14. The number of distinct operations contained in a group may be either finite or infinite. When the number is infinite, the group may contain operations which produce an infinitesimal change in every possible object or operand.

Thus the totality of distinct displacements of a rigid body evidently forms a group, for they satisfy conditions (α), (β) and (γ) of the definition. Moreover this group contains operations of the kind in question, namely infinitesimal twists; and each operation of the group can be constructed by the continual repetition of a suitably chosen infinitesimal twist.

Next, the set of translations, that arise by shifting a cube parallel to its edges through distances which are any multiples of an edge, forms a group containing an infinite number of operations; but this group contains no operation which effects an infinitesimal change in the position of the cube.

As a third example, consider the set of displacements by which a complete right circular cone is brought to coincidence with itself. It consists of rotations through any angle about the axis of the cone, and rotations through two right angles about any line through the vertex at right angles to the axis. Once again this set of displacements satisfies the conditions (α), (β) and (γ) of the definition and forms a group.

This last group contains infinitesimal operations, namely rotations round the axis through an infinitesimal angle; and every finite rotation round the axis can be formed by the continued repetition of an infinitesimal rotation. There is however in this case no infinitesimal displacement of the group by whose continued repetition a rotation through two right angles about a line through the vertex at right angles to the axis can be constructed. Of these three groups with an infinite number of operations, the first is said to be a continuous group, the second a discontinuous group, and the third a mixed group.

Continuous groups and mixed groups lie entirely outside the plan of the present treatise; and though, later on, some of the properties of discontinuous groups with an infinite number of operations will be considered, such groups will be approached from a point of view suggested by the treatment of groups containing a finite number of operations. It is not therefore necessary here to deal in detail with the classification of infinite groups which is indicated by the three examples given above; and we pass on at once to the case of groups which contain a finite number only of distinct operations.

15. Definition. If the number of distinct operations contained in a group be finite, the number is called the order of the group.

Let S be an operation of a group of finite order N. Then the infinite series of operations

S,S2,S3,
must all be contained in the group, and therefore a finite number of them only can be distinct. If Sm+1 is the first of the series which is the same as S, and if S1 is the operation inverse to S, then
Sm+1S 1 = SS1 = 1,
or
Sm = 1.

Exactly as in § 8, it may be shewn that, if

Sμ = 1,
μ must be a multiple of m, and that the operations S, S2, …, Sm1 are all distinct.

Since the group contains only N distinct operations, m must be equal to or less than N. It will be seen later that, if m is less than N, it must be a factor of N.

The integer m is called the order of the operation S. The order m of the operation Sx is the least integer for which

Sxm = 1,
that is, for which
xm 0(modm).

Hence, if g is the greatest common factor of x and m,

m = m g ;
and, if m is prime, all the powers of S, whose indices are less than m, are of order m.

Since

SxSmx = Sm = 1,(x < m),
and
Sx(Sx) 1 = 1,
it follows that
(Sx) 1 = Smx.

If now a meaning be attached to Sx, by assuming that the equation

Sx+y = SxSy
holds when either x or y is a negative integer, then
Smx = SmSx = Sx,
and
(Sx) 1 = Sx,
so that Sx denotes the inverse of the operation Sx.

Ex. If Sa, Sb, …, Sc, Sd are operations of a group, shew that the operation inverse to SaαSbβScγSdδ is SdδScγSbβSaα.

16. If

1,S1,S2,,SN1
are the N operations of a group of order N, the set of N operations
Sr,SrS1,SrS2,,SrSN1
are (§ 13) all distinct; and their number is equal to the order of the group. Hence every operation of the group occurs once and only once in this set.

Similarly every operation of the group occurs once and only once in the set

Sr,S1Sr,S2Sr,,SN1Sr.

Every operation of the group can therefore be represented as the product of two operations of the group, and either the first factor or the second factor can be chosen at will.

A relation of the form

Sp = SqSr
between three operations of the group will not in general involve any necessary relation between the order of Sp and the orders of Sq and Sr. If however the two latter are permutable, the relation requires that, for all values of x,
Spx = S qxS rx;
and in that case the order of Sp is the least common multiple of the orders of Sq and Sr.

Suppose now that S, an operation of the group, is of order mn, where m and n are relatively prime. Then we may shew that, of the various ways in which S may be represented as the product of two operations of the group, there is just one in which the operations are permutable and of orders m and n respectively.

Thus let

Sn = M,
and
Sm = N,
so that MN are operations of orders m and n. Since Sm and Sn are permutable, so also are M and N, and powers of M and N.

If x0y0 are integers satisfying the equation

xn + ym = 1,
every other integral solution is given by
x = x0 + tm,y = y0 tn,
where t is an integer.

Now

MxNy = Sxn+ym = S;
and since x and m are relatively prime, as also are y and n, Mx and Ny are permutable operations of orders m and n, so that S is expressed in the desired form.

Moreover, it is the only expression of this form; for let

S = M1N1,
where M1 and N1 are permutable and of orders m and n.

Then Sn = M1n, since N1n = 1.

Hence

M1n = M,
or
M1xn = Mx,
or
M11ym = Mx.
But M1m = 1, and therefore M1ym = 1; hence
M1 = Mx.
In the same way it is shewn that N1 is the same as Ny. The representation of S in the desired form is therefore unique.

17. Two given operations of a group successively performed give rise to a third operation of the group which, when the operations are of known concrete form, may be determined by actually carrying out the two given operations. Thus the set of finite rotations, which bring a regular solid to coincidence with itself, evidently form a group; and it is a purely geometrical problem to determine that particular rotation of the group which arises from the successive performance of two given rotations of the group.

When the operations are represented by symbols, the relation in question is represented by an equation of the form

AB = C;
but the equation indicates nothing of the nature of the actual operations. Now it may happen, when the operations of two groups of equal order are represented by symbols,

 (i) 1,A,B,C,  (ii) 1,A,B,C,

that, to every relation of the form

AB = C
between operations of the first group, there corresponds the relation
AB = C
between operations of the second group. In such a case, although the nature of the actual operations in the first group may be entirely different from the nature of those in the second, the laws according to which the operations of each group combine among themselves are identical. The following series of groups of operations, of order six, will at once illustrate the possibility just mentioned, and will serve as concrete examples to familiarize the reader with the conception of a group of operations.

I. Group of inversions. Let P, Q, R be three circles with a common radical axis and let each pair of them intersect at an angle 1 3π. Denote the operations of inversion with respect to PQR by CDE; and denote successive inversions at PR and at PQ by A and B. The object of operation may be any point in the plane of the circles, except the two common points in which they intersect. Then it is easy to verify, from the geometrical properties of inversion, that the operations

1,A,B,C,D,E
are all distinct, and that they form a group. For instance, DE represents successive inversions at Q and R. But successive inversions at Q and R produce the same displacement of points as successive inversions at P and Q, and therefore
DE = B.

II. Group of rotations. Let POP, QOQ, ROR be three concurrent lines in a plane such that each of the angles POQ and QOR is 1 3π, and let IOI be a perpendicular to their plane. Denote by A a rotation round II through 2 3π bringing PP to RR; and by B a rotation round II through 4 3π bringing PP to QQ. Denote also by CDE rotations through two right angles round PP, QQRR. The object of the rotations may be any point or set of points in space. Then it may again be verified, by simple geometrical considerations, that the operations

1,A,B,C,D,E
are distinct and that they form a group.

III. Group of linear transformations of a single variable. The operation of replacing x by a given function f(x) of itself is sometimes represented by the symbol (x,f(x)). With this notation, if

A = x, 1 1 x, B = x, x 1 x , C = x, 1 x, D = (x, 1 x), E = x, x x 1 , 1 = (x,x),

it may again be verified without difficulty that these six operations form a group.

IV. Group of linear transformations of two variables. With a similar notation, the six operations

A = x, y x; y, 1 x, B = x, 1 y; y, x y, C = (x,y; y,x), D = x, 1 x; y, y x, E = x, x y; y, 1 y, 1 = (x,y; x,y)

form a group.

V. Group of linear transformations to a prime modulus.

The six operations defined by

A = (x,x + 1), B = (x,x + 2), C = (x, 2x), D = (x, 2x + 2), E = (x, 2x + 1), 1 = (x,x),

where each transformation is taken to modulus 3, form a group.

VI. Group of substitutions of 3 symbols. The six substitutions

1,A = (xyz),B = (xzy),C = x(yz),D = y(zx),E = z(xy)
are the only substitutions that can be formed with three symbols; they must therefore form a group.

VII. Group of substitutions of 6 symbols. The substitutions

1,A = (xyz)(abc),B = (xzy)(acb),C = (xa)(yc)(zb), D = (xb)(ya)(zc),E = (xc)(yb)(za)

may be verified to form a group.

VIII. Group of substitutions of 6 symbols. The substitutions

1,A = (xaybzc),B = (xyz)(abc),C = (xb)(yc)(za), D = (xzy)(acb),E = (xczbya)

form a group.

The operations in the first seven of these groups, as well as the objects of operation, are quite different from one group to another; but it may be shewn that the laws according to which the operations, denoted by the same letters in the different groups, combine together are identical for all seven. There is no difficulty in verifying that in each instance

A3 = 1,C2 = 1,B = A2,D = AC = CA2,E = A2C = CA;
and from these relations the complete system, according to which the six operations in each of the seven groups combine together, may be at once constructed. This is given by the following multiplication table, where the left-hand vertical column gives the first factor and the top horizontal line the second factor in each product; thus the table is to be read A1 = A, AB = 1, AC = D, and so on.

1 A B C D E






1 1 A B C D E
A A B 1 D E C
B B 1 A E C D
C C E D 1 B A
D D C E A 1 B
E E D C B A 1

But, though the operations of the seventh and eighth groups are of the same nature and though the operands are identical, the laws according to which the six operations combine together are quite distinct for the two groups. Thus, for the last group, it may be shewn that

B = A2,C = A3,D = A4,E = A5,A6 = 1,
so that the operations of this group may, in fact, be represented by
1,A,A2,A3,A4,A5.

18. If we pay no attention to the nature of the actual operations and operands, and consider only the number of the former and the laws according to which they combine, the first seven groups of the preceding paragraph are identical with each other. From this point of view a group, abstractly considered, is completely defined by its multiplication table; and, conversely, the multiplication table must implicitly contain all properties of the group which are independent of any special mode of representation.

It is of course obvious that this table cannot be arbitrarily constructed. Thus, if

AB = P andBC = Q,
the entry in the table for PC must be the same as that for AQ. Except in the very simplest cases, the attempt to form a consistent multiplication table, merely by trial, would be most laborious.

The very existence of the table shews that the symbols denoting the different operations of the group are not all independent of each other; and since the number of symbols is finite, it follows that there must exist a set of symbols S1, S2, …, Sn no one of which can be expressed in terms of the remainder, while every operation of the group is expressible in terms of the set. Such a set is called a set of fundamental or generating operations of the group. Moreover though no one of the generating operations can be expressed in terms of the remainder, there must be relations of the general form

SmaS nbS pc = 1
among them, as otherwise the group would be of infinite order; and the number of these relations, which are independent of one another, must be finite. Among them there necessarily occur the relations
S1a1 = 1,S 2a2 = 1,,S nan = 1
giving the orders of the fundamental operations.

We thus arrive at a virtually new conception of a group; it can be regarded as arising from a finite number of fundamental operations connected by a finite number of independent relations. But it is to be noted that there is no reason for supposing that such an origin for a group is unique; indeed, in general, it is not so. Thus there is no difficulty in verifying that the group, whose multiplication table is given in § 17, is completely specified either by the system of relations

A3 = 1,C2 = 1,(AC)2 = 1,
or by the system
C2 = 1,D2 = 1,(CD)3 = 1.

In other words, it may be generated by two operations of orders 2 and 3, or by two operations of order 2. So also the last group of § 17 is specified either by

A6 = 1,
or by
B3 = 1,C2 = 1,BC = CB.

19. Definition. Let G and G be two groups of equal order. If a correspondence can be established between the operations of G and G, so that to every operation of G there corresponds a single operation of G and to every operation of G there corresponds a single operation of G, while to the product AB of any two operations of G there corresponds the product AB of the two corresponding operations of G’, the groups G and G are said to be simply isomorphic4.

Two simply isomorphic groups are, abstractly considered, identical. In discussing the properties of groups, some definite mode of representation is, in general, indispensable; and as long as we are dealing with the properties of a group per se, and not with properties which depend on the form of representation, the group may, if convenient, be replaced by any group which is simply isomorphic with it. For the discussion of such properties, it would be most natural to suppose the group given either by its multiplication table or by its fundamental operations and the relations connecting them; and as far as possible we shall follow this course. Unfortunately, however, these purely abstract modes of representing a group are by no means the easiest to deal with. It thus becomes an important question to determine as far as possible what different concrete forms of representation any particular group may be capable of; and we shall accordingly end the present chapter with a demonstration of the following general theorem bearing on this question.

20. THEOREM. Every group of finite order N is capable of representation as a group of substitutions of N symbols5.

Let

1,S1,S2,,Si,,SN1
be the N operations of the group; and form the complete multiplication table
1, S1, S2, , Si, , SN1, S1, S12, S2S1, , SiS1, , SN1S1, ....................................................... Si, S1Si, S2Si, , Si2, , SN1Si, ....................................................... SN1, S1SN1, S2SN1, , SiSN1, , SN12,
that results from multiplying the symbols in the first horizontal line by 1S1, S2, …, SN1 in order. Each horizontal line in the table so obtained contains the same N symbols as the original line, but any given symbol occupies a different place in each line; for the supposition that the two symbols SiSp and SiSq were identical would involve
Sp = Sq,
which is not true.

It thus appears that the first line in the table, taken with the (i + 1)th line, defines a substitution

1 , S1 , S2 ,, SN1 Si,S1Si,S2Si,,SN1Si
performed on the N symbols 1S1, S2, …, SN1; and by taking the first line with each of the others, including itself, a set of N substitutions performed on these N symbols is obtained. Now this set of substitutions forms a group simply isomorphic to the given group. For if
SpSi = Sq andSqSj = Sr,
the substitutions
1 , S1 , S2 ,, SN1 Si,S1Si,S2Si,,SN1Si
and
1 , S1 , S2 ,, SN1 Sj,S1Sj,S2Sj,,SN1Sj,
successively performed, change Sp into Sr.

But

SpSiSj = Sr,
and therefore the product of the two substitutions, in the order given, is the substitution
1 , S1 , S2 ,, SN1 SiSj,S1SiSj,S2SiSj,,SN1SiSj,
or the product of any two of the N substitutions is again one of the substitutions of the set. Moreover the above reasoning shews that the result of performing successively the substitutions corresponding to the operations Si and Sj gives the substitution corresponding to the operation SiSj. Hence, since the number of substitutions is equal to the number of operations, the given group and the group of substitutions are simply isomorphic.

It may also be shewn that each of the N substitutions is regular (§ 9) in the N symbols. For the substitution

1 , S1 , S2 ,, SN1 Si,S1Si,S2Si,,SN1Si
changes Sr into SrSi, SrSi into SrSi2, and so on. If then m is the order of Si, the cycle of the substitution which contains Sr will be
(Sr,SrSi,SrSi2,,S rSim1).

Now Sr may be any symbol of the set; hence all the cycles of the substitution must contain the same number, m, of symbols.

The substitution is therefore regular in the N symbols, and this can only be the case if m is a factor of N. It follows at once, as was stated in § 15, that the order of any operation of a group of order N must be equal to or a factor of N.

All the substitutions in this form of representing a group being regular, the group itself is said to be expressed as a regular substitution group.

21. The form, in which it has just been shewn that every group can be represented, is by no means the only form of representation possessing this property. Thus Hurwitz6 has shewn that every group can be expressed as a group of rational and reversible transformations which change a suitably chosen algebraic curve (or Riemann’s surface) into itself. We shall see in Chapter XIV that every group can also be expressed as a group of linear transformations of a finite set of variables to a prime modulus.