AN INVARIANTIVE

INVESTIGATION
OF IRREDUCIBLE
BINARYMODULAR
FORMS*
BY

LEONARD EUGENE DICKSON

1. A fundamental system of invariants of the group of all binary linear

transformations in a finite field is shown in 2-6 to consist of two invariants,
one the product of the distinct linear forms and the other the product of the
distinct irreducible quadratic forms, where in each case no two factors have a
constant ratio. The product irm of the irreducible forms of degree m can be
expressed in terms of the fundamental invariants ; this is accomplished in
7-9 by means of the remarkable three-term recursion formula (15).
Two forms shall be said to belong to the same class if and only if one of them
can be transformed into a constant multiple of the other by a linear transforma-
tion of determinant unity in the field. It is shown in 10 that there are as
many classes of irreducible binary forms of degree m as there are irreducible
factors of 7rm when expressed as a function of a certain pair of invariants.
The choice of the latter is different in the two cases, p = 2, p > 2, where p is
the modulus of the field ; this is due to the fact that, in the respective cases,
there are one or two similarity transformations of determinant unity. The
investigation is completed for the values of m less than 8. The difficulties
encountered increase as the number of factors of m increases. Certain prob-
lems arise for which the present invariantive theory affords an indirect solution,
whereas a direct solution appears to be quite difficult (cf. end of 10, and end
of 17). For m = 6, it was necessary to enumerate the irreducible cubics in
the GF\_pn~\ whose roots are squares in the GF\_pZn~\ and have a given sum.

Determination of a Fundamental System of Invariants, 2-6.

2. Let G be a group of finite order g composed of linear homogeneous

transformations on m ( m > 1 ) variables with coefficients in any given field F.
The term point will be used in the sense of homogeneous coordinates, so that
{ xx, , xm ) is identified with ( /mx1, , pxm ), while ( 0, , 0 ) is excluded.

* Presented to the Society, September 7, 1910.

Trans. Am. Math. Soc. 1 1

License or copyright restrictions may apply to redistribution; see [Link]

 2 L. E. DICKSON : INVARIANTIVE THEORY [January

A transformation which leaves every point unaltered is necessarily a similarity

transformation
(1) x'. = axt ( i = 1, , m ).

In fact, by employing the points all but one of whose coordinates are zero, we
see that the transformation is of the form

Bj a,, (=1, ,!).

Then by employing the points all but two of whose coordinates are zero, we see
that the a. are equal. Let 7 be the number of transformations (1) contained in
G, and set a>= g y. We assume that a>> 1. Any point is one of at most
<o distinct conjugates under G. A point is called special if it is one of fewer
than a) conjugates, that is, if it is invariant under at least one transformation
other than (1) of the group G. If a point P is invariant under T and if S
replaces P by P', then P' is invariant under S_1TS. To determine all
special points it therefore suffices to employ a representative of each set of con-
jugate transformations other than (1), obtain the points invariant under the
representative, select the distinct points so obtained, and find the conjugates to-
each under G.
3. Let G be the group of all binary linear homogeneous transformations of
determinant unity in the Galois field GF\_p"~\ of order p". The order of G is

(2) g = ,/(/ -1).

Within G any transformation with irreducible characteristic determinant
A ( p ) is equivalent to a transformation *

r=(i ~o)' M/0-^-./w + i.

The only points invariant under T are (p, 1 ), where A(/a) = 0. The trans-
formations of G which leave ( p, 1 ) unaltered are

7 ), V abc + c3 = 1,
\ c b caj

namely, the transformations commutative with T. The quadratic condition

may be written in the form
(b-pc)"+l = l,

where p is a root of A(p) = 0, and hence has pn + 1 sets of solutions in the

GF\_pn~\. Thus (p, 1) is one of p2n p" conjugate points under G. The
resulting system of special points is therefore (ep +f, 1), where e is any element

*These Transactions, vol. 2 (1901), p. 117.

License or copyright restrictions may apply to redistribution; see [Link]

 1911] OF IRREDUCIBLE FORMS 3

=)=0 of the GF[p"~\ and f is any element. Such a point together with
(ep +f, 1), where p is the second root of A(p) = 0, determine a quadratic
form x2 + which vanishes only for these two points. Hence a relative
invariant under G is given by the product Q of the | ( p2n pn ) irreducible
quadratic forms x2 -f . We have *

(3) Q= y = ar-<*-'V (8-P-i).

^ y j=o

Consider next a transformation S for which

where k is in the GF[pn~\. There exists a linear function I which S replaces

by kI. If k2 4= 1, S is therefore equivalent within 6? with (J j^), which
leaves only (1, 0) and (0, 1) unaltered. If k2 = 1, /S' is equivalent to

^ = ( /3 il)'
Since xSuis of type (1), we may set =(=0 ; then S leaves only (0,1) invariant.
Hence when A(/j) is reducible, the resulting system of special points is com-
posed of the p" + 1 points (1,0) and (a, 1), where a ranges over the field.
There results the relative invariant

(4) L = y JJ (x - ay) = xpny- xyp\

a

4. Lemma. If an integral function is invariant under the group G of all

m-ary linear homogeneous transformations of determinant unity in the
GF[_pn~j, it is an absolute invariant of G.
The group G is generated f by transformations of the type Brsk, which alters
only one variable xr, replacing it by xr + Xxs. Jf B multiplies a function by
fi, Bv multiplies it by \iF. Now B is of period p. Hence \tf = 1. Employ-
ing the power pn~x, we get
pP = L = 1.

5. We may now prove that any integral invariant / of the binary group G
is an integral function of the invariants Q and L. If / vanishes for a special
point it has the factor Q or L. It remains to investigate the invariants / which
vanish for no special point. The number of transformations (1) contained in
#is
(5) 7 = 1 if p mm
2, 7 = 2 if p > 2.
*DlCKSON, Linear Groups (Leipzig, 1903), page 17.
t Linear Groups, page 78.

License or copyright restrictions may apply to redistribution; see [Link]

 4 L. E. DICKSON : INVARIANTIVE THEORY [January

Any non-special point is one of = g/y conjugates under G. Now the forms

(6) q = Q(i,"+1>'Y, = jJP'(P'-i)ly

are of degree . Let Fpn be the field composed of all roots of all algebraic
equations with coefficients in the GF [/>"]. The invariant /vanishes for a set
of as conjugate points ; if ( c, d ) is one of these points, the ratio c : d belongs to
the field Fpn. Hence we can determine t in the latter field such that
q(c, d) + tI(c, d) = 0. Thus /has the factor q + tI, and / is a product of
such linear functions of q, I. HI has its coefficients in the GF[pn~\, and t
is a root of an equation of degree t irreducible in the GF[pn~\, the presence of
the factor q + rl in / implies that of the conjugate factors q -f- rpml and hence
of a function of degree t in q and I with coefficients in the GF[pn~\.
Theorem. Any integral invariant with coefficients in the GF[pn] of the
binary group G is an integral function of Q and L with coefficients in the
GF[pn].

6. The transformation which leaves y unaltered and replaces x by Dx, where

D" = 1, leaves Q unaltered and replaces L by DL. Hence Q is an absolute
invariant and L a relative invariant of weight unity under the group G' of all
binary transformations in the GF\_pn], Hence any absolute invariant of G'
is expressible in terms of Q and L", where s = p" 1.
The invariants irm (m > 2 ) investigated in 7 are absolute invariants under
G' which vanish for no special points. Hence they are expressible in terms of
invariants (6) in such a manner that, for p > 2, I occurs only to even powers,
and in view of the homogenity q occurs only to even powers. Hence, for any
p, they are expressible in terms of t/and K, where

(7) J=Q**+\ K=L>p\

The Invariants irm as Functions of the Fundamental Invariants, 79.

7. For m ^> 1, let irm denote the product of all binary forms of degree m
which are irreducible in the GF\_pn] and have unity as the coefficient of xm.
Let
(8) -?? ?;".
where ql, , q are the distinct prime factors > 1 of m. Set

(9) *.- JT (.-r-i).

Thus Fx = 1, F2 = Q. It follows readily from Linear Groups, page 18, that

FmUFmlUF,
0) 7rm=
^Fmqi^Fmqmk-

License or copyright restrictions may apply to redistribution; see [Link]

 1911] OF-IRREDUCIBLE FORMS 5

in which the first product in the numerator extends over the Jft( /x 1) combi-
nations of qt, -, q^ two at a time, and similarly for the remaining products.
By 6, "Tm(m > 2) is an integral function of Jand Koi degree

(11) d _Pnm-^Pqi
p"(p2"-l)

For example, if m is an odd prime,

7i(m1) _ i

(12) 7Tm= Fm, dm = x (m an odd prime).

8. Since d3 = 1, 7r3= F3 is a linear function of J and K. By (3),

pn

Hence by (5),
(13) e7"= x^+^-'y (t = kp+j).

The terms of (13) given by j = pn and & <_pB are

s s

But every coefficient in the expansion of (a b)" is congruent to unity modulo

p. Hence the preceding sum equals, in the field,

(14) (xpinypn - xp"y"'ny =(Lpny = K.

Next the terms of (13) with Jc =pn are

pn pl"+pn

y X'<.P"-J)y<.P-n+') _ V" a.(P2"+P"-T)yT>

j=0 T=y2"

When the last sum is combined with the terms of (13) given by j <ipn, & <pn>
and hence by t = 0, 1, , p2n 1, we get
J)2+P" V.P3"-! _ yi3"-!
ZXS(.P*+P"-t)y>t
" _ _&_8 _ . _ ff 3 '_
=o ' y

It follows that 7T3= J if.

9. We make use of the following identity :

(af** _ y'Pn){xv"'-i - yP'"-1) = (a^2"-1 - ys-i)(ajf*p - yp--^)

- ( *y ) sp" ( Xs - y> ) ( xp"-p" - y^-P" ),

License or copyright restrictions may apply to redistribution; see [Link]

 6 L. E. DICKSON : INVARIANTIVE THEORY [January

where s = pn l. We divide by (je* y")(x,p" y!p'1) and insert the factor

xapln y"p'n in the numerator and denominator of the last term. By (14),

( xv \sp" ( xsp!n_vsp2")
-r-y*- - (xy)*"(x*" - yp")p"->= (x^tT ~ *PW = K.

Hence we obtain the recursion formula

(15) Ft= QFti -KFf-i (F,=i, F, = Q).

Taking t = 3 and applying (7) and (12), we get

(16) ^ = P3 = J-K,
in agreement with 8. By (10) and (15) for t = 4,

irt= FJQ, F^QFr-KQT-

Applying (16) and (7) we get

(17) tt4 = </"- /""-1 ir - ir*-.

In a similar manner,

(18) tt5 = J"2"^ - Jp"K- Jp"-p"+i Kn - JKp2n + /r*2*^,

(19) 7T6= Jp*"+p"-*- Jp*>-p*+p-i

xp"n - jp^Kp** - T*"( ^^- 1,

(20) 7T7= Jp*"-p'+i(Jp* Kpn) {Jptn - Jp^-^Kp'" - Kptn)

- JKp"n ( J>" - ir*4") - Kirf.

The Number of Classes of Irreducible Forms of Degree m, 10-17.

10. Let (f>(x, y) xm + be an irreducible binary form in the GF\pn~\.
Let cj)l= <p, <f>, , <f>kbe the distinct forms having unity as the coefficient of
xm which are equivalent to a constant multiple of $ under the group G of deter-
minant unity. By 4, the product P = (j> $k is an absolute invariant of
G. By 5, if m > 2, Pm is an integral function with coefficients in the
GF\_pn~\ of the invariants q and I, defined by (6). The function Pm(q, I) is
obviously irreducible in the field.
Two binary forms shall be said to belong to the same class if and only if one
of them is equivalent to a constant multiple of the other under the group G of
all binary transformations in the GF\_pn~\ of determinant unity.
The preceding remarks lead to the
Theorem. There are as many classes of irreducible binary forms of degree
m as there are irreducible factors in the GF\_pn~\ of the invariant irm(q, I)-

License or copyright restrictions may apply to redistribution; see [Link]

 1911] OF IRREDUCIBLE FORMS 7

To decide whether or not two given forms <pand yjr belong to the same class,
we employ a root p = x/y of </>= 0 and a root a = xjy of ir = 0, each belong-
ing to the G F \_pnm~\,and determine the values of q/l for x/y = p and a, respec-
tively. According as these values of q/l are the roots of the same factor or
different irreducible factors of ^m(q, I), the given forms <f>and i/r belong to the
same or different classes.
Since these values of q/l belong to the GF[pnm], the irreducible factors of
Tm(q, I) are of degree m or a divisor of m.

11. We proceed to the investigation of the irreducible factors of the function

irm{J, K) when p = 2, and those of this function with t/and K replaced by
q2 and I2 when p > 2 .
From (16) we conclude that the cubic forms fall into a single class if p = 2,
and into two classes if p > 2.
For m = 4, set K= pj in (17). Thus 7r4 vanishes only if

p"=l-p.
Then p belongs to the GF[p~n] since

pTm^l-pTmrnp.

For p = 2, p does not belong to the G F[2"]. Hence tt is the product of

2-i irreducible quadratic factors

(K- pJ){K- p"nJ) = K2 - KJ+ tJ\

where
(21) t = p"n+l = p-p2.

Hence,* for p = 2, t must be a root of

(22) + +*... +*-' = 1.

The latter has 2"_1 roots in the GF[2"]. Thus, forp = 2, the quarticforms
fall into 2"~l classes differentiated by the roots of (22).
Next, let p > 2. The only value of p in the GF[pn] is /a = i. The cor-
responding factor j2 2l2 is reducible if p" = 81 1, irreducible if />"= 8 3.
For the remaining pn 1 values of p, p2 p -{- t is irreducible, so that has
the J(i>" 1 ) values in the GF\_pn] for which 1 4t is a not-square. For
Z/^ = \, we have \2 = p. Now X will belong to the G F [p"] only when /a is
a square in the latter field and hence, by (21), only when
(pn-lJ/2 _ p(p-*-l)l2 _ 1 _

*In accord with Linear Groups, p. 29, formula (23) for A = v = l, p = 2.

License or copyright restrictions may apply to redistribution; see [Link]

 8 L. E. DICKSON: INVARIANT1VE THEORY [January

The condition is therefore that t be a square in the G F [/>"]. The number of

squares t for which 1 4 is a not-square is \ (/>" 1 ) or \ ( pn 3 ) accord-
ing as 1 is a square or a not-square. Indeed, * 1 2 = fir)2, where p. is a
not-square, has p" 1 sets of solutions according as 1 is a square or a not-
square. The sets f = 1, v = 0 are here excluded. We have now shown
that, when t has any one of the \ ( p" 1 ) values for which p2 p + t is irre-
ducible in the GF[pn~\, f(\) = X4 X2+ t is irreducible when is a not-
square, but is reducible when t is a square. In the latter case, set t = t2.
Then the factors of f( X) are

(23) X2\Vl+ 2t + t,
where t is that square root of for which 1 -f 2t is a square. A partia.
summary of our results for p > 2 is given in the following table giving the
number of irreducible factors of ir^q, I)-
! linear quadratic quartic total number

8A+1 2 i(y-l) i(3y+5)

8* 1 2 *(!> +1) i(8/>" + 8)
8 + 3 0 JO"-!) i(Sp--l)
8*-S 0 i(3/-+l).
In the following examples we give all the irreducible factors :

;V = 3. q2-2P, V-Ptf-q*.
pn = 5. q2-2l\ P2lq-q2, P-Pq2+2qi.

p" = l. q SI, I2 2lq - 2q2, P-Ptf + tq' (i = 3, 6).

p"=9, 2=;+l(uiod3). q(i-r-l)l, P ilq - iq\ P(2i + l)lq-(2i+l)q2,
P-Pq2 + tqi (*=<, 2+ l).
pn = 11. q2 - 2P, I2 Uq + q\ I2db Uq + 2q2,
PPq2 + tqi {i = 6, 7,8).

For p>2, the irreducible quartic forms fall into 6k + 2, 6k, 6k + 2, or

6k 2 classes, according as p" 8k + 1, 8& 1, 8k + 3, or 8k 3,
r-especiiWy.
Classes of Irreducible Quintic Forms.

12. For quintic forms, (18) vanishes for K= pj if

(24) i,P2+i_p,>_/3,."_i, + l = 0.

* Linear Groups, page 46.

License or copyright restrictions may apply to redistribution; see [Link]

 1911] OF IRREDUCIBLE FORMS 9

As remarked at the end of 10, each root of (24) belongs to the ri^lj/"].
We may give a direct proof as follows.

/>**=1 + p3" = 1 4- +
p-V p-i^(p-i)(pP'-iy

P ppbn = 1 + PP- = p.
pp<-= 1 4

If a root of (21) belongs to the GF[pn~], it satisfies the equation

(25) p2 - 3p + 1 = 0.
For p = 2, a root of (25) satisfies the equation p3 = 1 and hence belongs to the
GiJP[2"] if and only if n is even. If p=2, the quintic forms fall into
i ( 22" + 1 ) or 2 -f ^ ( 22" 1 ) classes, according as n is odd or even. For
example, the 6 irreducible quintic forms in the GF\_2~\ are equivalent.
For p > 2, we must consider (24) for p replaced by X2. By 10, the result-
ing equation for A has all its roots in the G F [j/" ]. If one of its roots belongs
to the GF[pn~\, it must satisfy the equation

(25') X4-3X2 + 1 = (\2+ X-1)(X2-X-1) = 0.

Hence (2Xl)2 = 5,so thatp = 5 orpn = bh 1. In the respective cases,

we obtain the factors I 2q or

Z+ i(lV5)?, Z-i(lV6)?.
Hence, if p > 2, the irreducible factors are of the following types :

linear qnintic

5" 2 2.521
6ife 1 4 W-i)
5/fc2 0 1(^ + 1)
For example, if p" = 3, the factors are

X _ \< _ \3 + X2 + j ^ x5 + X4 - X3 - X2 - 1, X5 - X4 + 1, X5 + X4

^Por p > 2, the irreducible quintic forms fall into

2(52-*+ l), 4 + |(/"-l), f(/ + l)

classes, according as p = 5, ^" = 5& 1, or 2?" = 5& 2.

Classes of Irreducible Sextic Forms.

13. For sextic forms, (19) vanishes for K'= pj'ii

License or copyright restrictions may apply to redistribution; see [Link]

 10 L. E. DICKSON: INVARIANTIVE THEORY [January

(26) 1 -pp-p?8"-pW ~P" J = 0.

For p = 1 the final term is a multiple of p, so that p = 1 is not a root. For

p = 2 or ^ = 3, there is no root in the GF [ p" ] , while for ^ > 3, the only
root is p = ^.
We determine the roots of (26) which belong to the GF\_p2n] by a simple
device. Any such root satisfies the equation

<27) l-p-p'^pp'^y).

The first member equals its ^"-th power. Hence

(28) p=l-p-s,
-where s is an element of the GF\_p"~] . Hence (27) gives

_ (1~P-')(P+t)
1- p
(29) p*+ p(s_l)+ ss = 0.

Hence any root of (26) which belongs to the G F [p2n], but not to the GF[pn],
satisfies an equation
of type (29) irreducible in the GF[pn~j. Conversely, any
root p of an irreducible equation (29) is a root of (26). Indeed, the second root
is ppn and the sum of the roots is 1 s ; thus (28) and hence also (27) is satisfied.
Now, for p > 2, (29) is irreducible if and only if (l + s)(l 3s) is a not-
square in the G F [/>"] We readily verify* that this is the case for the fol-
lowing number of values of s :

(30) i(y-3) if pn =3-f 2, \{pn - 1) if p = 3 or pn = SI + 1.

Forp>2, we must investigate the reducibility of (29) when p is replaced
by X2. By the product of the roots,

pp"+1=s2, pip--w = + 1,

so that p is a square in the GF[p2"~\. Hence the quartic in X is reducible.

Its factors are
X2 + aX s, X2 aX s,

where a2 = 1 s 2s. The product of the two values for a2 is the not-square
(1 + s ) ( 1 3s ), so that one value is a square.

* By means of Linear Groups, p. 48, 66. Note that 3 is a square or not-square according
as p" = 31+ 1 or 31+ 2.

License or copyright restrictions may apply to redistribution; see [Link]

 1911] OF IRREDUCIBLE FORMS 11

Examples. If pn = 3, then s = 1 and

X4 + 1 = (X2 + X - 1)(X2 - X - 1).

If pn = 5, then s = 2, and
X4+ 2X2 - 1 = (X2 + 2X - 2 ) (X2 - 2X - 2 ).

If p = 2, we multiply (29) by (s I )~2 and set

a = p(s-l)-1, t = s2(s2 + 1)-1.

Then cr2+ tr -f = 0. This is irreducible if and only if t is one of the 2"_1

roots of (22) in the GF[2n~\. Unless t = 1, the corresponding value of s is
uniquely determined in the field. But t = 1 is a root of (22) if and only if n
is odd. Hence the number of values of s for which (29) is irreducible in the
GF[2n] is
(31) 2"-1 - 1 for n odd, 2""1 for n even.

The number is zero for n = 1 : for w=2,3,4,sisa root of

s2 + s+l = 0, s3 + s + l = 0, (s4 + s +l)(s4 + s3 + l) = 0.

It remains to determine the roots of (26) which belong to the GF[pin~\.

These must satisfy the equation
(32) p + prn + pp2n = 1.

Conversely, any root of the latter belongs to the GF^p3"1^ and, with the
exception of p = 1 when p = 2, satisfies (26). A root of (32) belongs to the
GF[pn~\ only when 3/> = 1. Hence the irreducible cubic factors of (26) are
of the form p3 p2 -f- and their number is

(33) 32"-Mf^ = 3, |(y_l) if t, 4= 3.

By 15, these factors give all the existing irreducible cubics as3 x2 +
The irreducible factors of (26) are of degree 6 or a divisor of 6. Denote by
L, Q, C the number of linear, quadratic and cubic factors, respectively. Then
the number of degree 6 is

S = l(p**+ p-l-L-2Q-SC).
By the above results we have, whether p = 2 or p > 2,

^=i(3"-//"-r-2) for " = 3Z+2,

(34)
S = i(j53"-/") for p = 3 or pn = 3Z + 1.

In view of (31), (33), (34), we have the

License or copyright restrictions may apply to redistribution; see [Link]

 12 L. E. DICKSON: INVARIANTIVE THEORY [January

Theorem. If p = 2, the number of classes of irreducible sextic forms is

i(23"+ 22" + 3.2"-6)/or n odd, |(23"+ 22" f 3-2" - 2) for n even.
For p" = 2 this number is 2. The irreducible factors of (26) are then

p3 + p2 + 1, P6 + p5 + 1

Irreducible Cubics whose Roots are Squares with a given Sum.

14. For sextic forms when p > 2 there remains the difficult problem : to
determine which of the cubics

(35) C=p3-p2 + ap-b

are irreducible and give irreducible sextics when p is replaced by X2, and which
give reducible sextics. The first or second case arises according as p is a not-
square or a square in the G F [ p3n ], namely, according as 6 is a not-square or
a square in the GF [ pn ]. In fact,

b = [Link]", Up"-l)l2 = pt^-1)'2.

In the second case, b = e2, and X6 X4 -f- a\2 e2 has the factors

(36) X3cX2+i(c2-l)Xe,
where
c<_2c2- 8ec + l-4a=0.

This quartic has the resolvent cubic

y+ 2y2-4(l-4a)y-8(l-4a)-64e2=0.

We set y = 2z and multiply the resulting equation by 8, and get

a3 + z2 + ( 4a _ i \jj _ i + 4a + 86 = 0.

If in the cubic 8C= 0 we replace p by \(z + 1), we obtain the preceding

cubic. Hence the resolvent cubic is irreducible, so that* the quartic has one
and but one root c in the GF [pn ]. The unique pair of factors of the sextic
in X are therefore of the form (36).
To determine the irreducible cubics (35) in which b = e2,

(37) C=p<-P2+ap-e2 (, + 0)r

we proceed indirectly and find those which are reducible.

(a) First, let C have a linear factor p k and an irreducible quadratic
factor. The latter must be of the form p2+(kl)p+ks2, where s2=e2/i2=j=0.

* Bulletin of the American Mathematical Society, vol. 13 (1906), p. 5, middle

of p. 7.

License or copyright restrictions may apply to redistribution; see [Link]

 1911] OF IRREDUCIBLE FORMS 13

The quadratic is irreducible if and only if ( k 1 )2 4s2 is a not-square. Let

v be a fixed not-square and let pn = 41 1, the upper or lower sign holding
according as 1 is a square or a not-square. Then

(k - I)2 - 4ks2 = vz*

has pn rp 1 sets of solutions s, z when k is any chosen not-square, p" 1 sets

when k is any chosen square 4= 1, while for k = 0 or 1 the solutions are
obvious. After excluding the sets in which either s = 0 or z = 0, and dividing
by 4, we obtain the number of values of s2 4= 0 for which ( k 1 )2 4s2 is a
not-square. If 1 is a square the number of sets s2, k, and hence the number
of irreducible quadratics, is

i(^-l).|(^-l) + i(p"-l)-J(^-3) = Hi'n-l)(i'B-2).

If 1 is a not-square, the number is

i(p" + l)-h(pn-l) + l(pn-Z)-:k(pn-3) + h(pn-l) = h(p2n-pn+2).

In the following examples we give all the irreducible quadratics : For pn = 3,
s2 = l, = l. Fory = 5, s2=l, = 2, 3, 4. For p" = 7, s2 = 1,
= 1,3,5; s2= 2, = 1,2,4,6; s2=4, =1,3,5,6. Forp" = ll,
52= 1,=1,5,7,8,9,10; s2 = 3, = l,2,6,7,8; s2 = 4, = 1, 2,
5,6, 9; s2= 5,=l,3,4,7,8,10; s2=9,=1,2,3,4,6,10. For
y = 9, i2= -l(mod3), s2=l, = i + l,i-l; s2=,= -1,
i, i, =F + 1, =F 1.
(o) Next let (37) have three roots in the GF[pn~\. Either all three are
squares or only one root is a square.
(bj) Let the three roots be squares. We seek the number of systems of three
squares si with the sum unity, two systems being identified if they are composed
of the same s in different orders.
If t is the number of systems with s{ = s2 = s3, we have

(38) 5= 0 if p = 3 or p" = 12 5, t = 1 if pn = 12 1.

Let p be the number of sets with j = s2. Now x2 + 2y2 = 1 has pn 1 or

pn + 1 sets of solutions according as 2 is a square or a not-square. Of these,
4 or 2 have xy = 0, according as 2 is a square or a not-square, namely, accord-
ing as p" = 8Z 1 or 81 3. Hence
p,= |(y_5) if pn = 8l + l, i(i>n-3) if p" = 8l-l or 8Z+ 3,
i(^"-l) if pn = 8Z+ 5.

The number of sets s{ in which exactly two are equal is therefore Z(/i t).

License or copyright restrictions may apply to redistribution; see [Link]

 14 L. E. DICKSON: INVARIANTIVE THEORY [January

Next, the total number of sets of solutions of *

x2 + y2 + z2 = 1

is p2n pn according as pn = 4Z 1. These include pn =pl sets with z = 0.

Hence the numbers of sets of squares s( whose sum is unity is

k = % { p2n pn - 3(pn =pl - ) - 6}.

Hence there are ^ {k 3 (/t t) t) systems in which sp s2, s3 are all dis-
tinct. As shown above, there are /x t systems in which exactly two of the s.
are equal, and t systems with all three equal. The total number of systems is
therefore $(k + 3/x + 2t).
(62) Let one of the roots be a square s and the other two not-squares nx, n2.
If the latter are equal, we employ the pn 1 or pn -f 1 sets of solutions of
x2 + 2vif = 1, according as 2 is a not-square or a square, namely, according
as pn = SI 1, SI S or p" = SI + 1, SI + 3. Now there are two sets of
solutions with x = 0 if 2 is a not-square, namely, if pn = 81 3 . Hence the
number of sets s, nx = n2 is

N=i(pn-1) ii pn=Sl+l, i(p_3) if y = 8Z-l or 8 + 3,

i(/>"-5) if y = Sl + 5.
The total number of sets of solutions of x2 -f vy2 + vz2 = 1 is />2n pn accord-
ing as pn = 4Z 1. These include ^)" =p 1 sets with x = 0 and />" 1 sets with
y = 0. Hence the total number of sets s, n^, n2 is

Jf = |{/*|)"-(|)q:l)-2(/l-2)-2}.
The total number of systems is therefore J(M iV) + iV= Jf-f JiV. Now

k + \M= -^{p2" p" - 3p + 3),

while N+ p. = \{pn 3) for all values of pn. Hence the number of cubic
(37) with three roots in the GF[pn] is ^(p2n pn 6) -f ii, according as
pn = 4Z j= 1.
Combining this number with the number obtained in case (a), we get

i {p2n - 2pn +t)iipn = l+l, i (p2" -p" + t)ipn = l-l.

The total number of cubics (37) is \ (pn 1 )pn. Hence, by (38), we obtain the-
Theorem. The number of irreducible cubics (37) is

* Linear Groups, p. 48.

License or copyright restrictions may apply to redistribution; see [Link]

 1911] OF IRREDUCIBLE FORMS 15

l(p"Tl)(p2)i/>" = 12l,
-lp"(pn dr 1 ) ifpn = 12 5, or ifp = 3 and 1 = ( -1 )"..

A check upon the above results is afforded by the following examples which
give all systems of three elements in the GF[pn~^ whose sum is unity and whose
product is a square :

p" = 3, none;j>"=5, 1, 1, 4 ; 1, 2, 3 ; p"= 7, 4, 2, 2 ; 2, 3, 3 ; 4, 5, 6 ;

p" = S2, 2=-l(mod3),l,l,2;i, -, 1; 1, i=kl, - tq=l;
1,1 + 1, +l;, 1, azi 1 ;.
pn=ll, 4,4,4;3,4,5;5,9,9:3,2,7;3,10,10;
4,2,6;5,8,10;9,7,7;9,6,8;
i>"= 13, 1, 1, 12; 1, 3, 10; 1, 4, 9; 3, 12, 12; 9, 9, 9; 1, 2, 11; 1,6, 8;
1, 6, 7 ; 3, 5, 6 ; 4, 2, 8 ; 4, 5, 5 ; 9, 7, 11 ; 10, 2, 2 ; 10, 6, 11 ; 12, 7, 8.
By excluding the resulting cubics and those given by the examples under case
(a), we find the only irreducible cubics (37) are those in which a, e2 is one of
the following sets :

p" = 3, 2,l;p=5, 1,-1;3,-1;0,1;4,1;3,1;

p" = 7, 0,1;3,1;5,1;1,2;5,2;2,4;5,4;
p" = 9,2 = -l(mod3),0,-l; 1,-1; i, -1; -1, 1;
t-f- 1, 1; i 1, 1; 0, d=i ; 1, i; i 1, qpi;
p"=ll, 3,1; 7,1; 0,3; 1, 3; 2, 3; 2, 4; 6, 4; 7, 4;
8, 4; 0,5; 7, 5; 10, 5; 1,9; 2, 9; 3, 9; 5, 9.
15. The investigation made in 14 enables us to determine the number N t of
irreducible cubics x3 gx2 -\- hx + , in which g is a given element and k ranges
over the squares, and the corresponding number Ngn when k ranges over the
not-squares.
Let C be the number of all the irreducible cubics x3 gx2 + in which g
is a given element. If p is a root of such a cubic,
(40) p + ppn + pp"-n = g, pp3n = p-

Since the latter equation follows from the former, we conclude that (40J has p2"
roots in the GF[p3n~\ . If such an element p is distinct from ppn it is a root of
a cubic irreducible in the GF[pn]. Now a common root of p = ppn and (lOj)

License or copyright restrictions may apply to redistribution; see [Link]

 16 L. E. DICKSON: INVARIANTIVE THEORY [January

makes 3p = g, so that the two equations have a single solution if p =|=3, no

solution if p = 3, g =(=0, and 3" solutions if p = 3 , g = 0. Hence

(41)<7,~J(p*-l)ifp+S, C0=i(32"-3")ifp=3, C9,=32-1ifp=3,y'+0.

For any p, the total number of irreducible cubics is therefore ^p" ( p2n 1 ), in
agreement with 7. Conversely, if * p =(=3, the latter result implies (41),
since x = x + ^(g + g') transforms each x3 gx2 + into a cubic
x3 g'x2 + -.
If v is a fixed not-square in the GF[pn~\, (40j gives

vp + {vp)pn + (vpy1* = vg.

First, let p be a square in the Cri^fp3"]. Since v is a not-square in the

GF[p3"~\, vp is a not-square. Hence

N 0* = N On ,' Ngs =N. y n ,'

where one of the elements g, g' is any square and the other any not-square.
Thus if <r is any square and v any not-square in the GF\_pn~\,

(42) iV0i= N0n= iC0, Nm = Nat = JV;., Nvs = Nan = C, - Nu,

where C0 and Cx are given by (41) and Nu is givenby (39).

16. We may now enumerate the irreducible factors in the GF[pn~\, p > 2,
of the function of degree d = 2(p3,t -f p" 1 ) derived from (26) by replacing
p by X2. First, if p =)=3, X2 ^ has t + 1 factors, where t is defined by (38).
Next, there are 2r further irreducible quadratic factors, where r is defined by
(30). Again, there are 2^^ irreducible cubic factors and Nln irreducible sextic
factors, all derived from the cubic functions of p ( 14). The remaining factors
are irreducible sextics ; their number is therefore

l(d e 4r 6iVIs QNU) (e = 0ifi) = 3, e = 2ifj> + 3).

Thus if = 0 when p = 3,/8=i-|-|ifp^=3. the total number of factors is

ld++*r + Nu.
Hence we may state the following
Theorem. The number of classes of irreducible sextics in the G F [ p" ],
p>2, is i(2p3"+p2" + 7pn-a), a=-2 if p"=12+l, a = 10 if
pn= l2k + 5, a = 6 if p 8, n even; \{2p3n + p2n + 5p" b), 6 = 2 */
p" = 12k - 5, b = 6 ifp" = 12 - 1 or ifp = S, n odd.
* By means of %' = gx we couclnde that, for any p, Cg = Ciit g + 0. The value of C0 was
determined otherwise in Bulletin of the American Mathematical Society, 1. c, pp.
3,4.

License or copyright restrictions may apply to redistribution; see [Link]

 1911] OF IRREDUCIBLE FORMS 17

Forp" = 3, the 12 irreducible factors* are

X2 X - 1, \3 A2 =f 1, X6 - A4 + 1, X6 - X4 + X2 + 1,

X6 db X6 - X4 + X2 - 1, X6 X5 X3 - 1, X6 X5 =f X3 + 1.

For p" = 5, the 50 irreducible factors are 34 sextics f and

X2-2, X22X-2, X3=hX22, X32X2-X2, X3=p2X2-Xl,

X3 + 2X1, X3X21, \6_X4-X2-3, \-\4 + \2_3, X6-X4-2.

For p" = 7, the 128 irreducible factors are 98 sextics f and

X2-5, X23X+1, X23X-2, X22X + 3, X3 2X2- 2X 1, X3X21,

X33X2+4X1, X3X23, X33X2+4X3, X3+ 3X2, X32X2-2XT2,
X6-X4+4X2-6, \_\4+6X2-3, X6-X4-3, \"_\4+\'-5, X6-X4+5X2-3,

\6_\*+2\2_6, X6-X4+3X2-5, \_\4+4X2-3, X6-X4+6X2-5.

Classes of Irreducible Septic Forms.

17. For septic forms we set K= pjin (20) and obtain an equation of degree
d = p4n + p2n + 1 in p, having d roots in the GF[p7n~\. For pp" = p, the
equation becomes
(43) 1 -5P + 6p*-p3 = Q.
If p = 2, a root of (43) belongs to the exponent 7 and hence belongs to the
GF [ 2" ] if and only if 7 is a divisor of 2" 1, namely, if n is a multiple of
3. In the latter case there are 3 linear and \{d 3) irreducible septic factors.
But if n is prime to 3, all the factors are septics. For example, if n = 1, 2, or
3, the number of septic factors is 3, 39, or 594, respectively. If p = 2, the
number of classes of irreducible septic forms is 3 -+-^(<7 3) or ^d, accord-
ing as n is a multiple of 3 or prime to 3.
If we set p = o-+ 2 in (43), we get J

(44) o-3- la - 7 = 0.

Hence if p = 7, p = 2 is the only root in the G F [ pn ]. Upon replacing /a

by X2 we obtain the linear factors X 3 and ^ ( 2c? 2 ) irreducible septic
factors. For p = 7, iAe number of classes of irreducible septic forms is
2.74"-1 -f 2-72"-1 + 2.
*Two factors with an ambiguous sign are obtained from the same irreduoible function of p.
t The omitted sextics are by paire factors of an irreducible sextic in p = A2.
tThis cubic is a resolvent for the seventh roots of unity. For its numerical solution Bee
Sereet, Algbre Suprieure, vol. 1, pp. 338, 343.
Trans. Am. Math. Soc.

License or copyright restrictions may apply to redistribution; see [Link]

 18 L. E. DICKSON: BINARY MODULAR FORMS

For p =(=2, p =J=7, we determine the number N of distinct roots of (43) in

the GF[pn~\ by the following indirect process. Since the equation of degree
d has d Nroots in the GF\_p7n~\, not belonging to the GF[pn~\, it has

E=\(pi"+p2n+l-N)

irreducible septic factors. Thus E must be an integer. Since the discriminant

of (43) or (44) is the square 72, we have N= 0 or 3 (Bulletin, 1. c, p. 1).
Now p2n is of the form 11 -f- s, where 8 = 1,2 or 4. Thus E is an integer
only when N= 3, 0 or 0, respectively. Hence for p =f=2, p 4= 7, the cubic (43)
has 3 roots in the GF [pn] if pn = Ik 1, no root if p" = Ik 2, 3. In
the former case, each root r is a square in the GF[pn~\; for, if not, each root
of X2 = r would belong to the GF[p2"~\ and not to the GF[pn~\, whereas all
the roots of the X-equation belonging to the GF\_p'n'\. Hence if p" = Ik 1,
p > 2, there are six linear factors. For example,

_p"=13,X2,X4, X8; p" = 29, X4, X 8, X10:

p" = 41, X 3, X 10, X 15 ; pn = 43, X 9, X 16, X 20.

For ])=(= 2,]) 4=^1 the number of<classes of irreducible septics is

6 +f(pin +p2"-2) or f(pin+p2n+ 1),

according as pn is or is not of the form Ik 1.

The University of Chicago,
July 14, 1910.

License or copyright restrictions may apply to redistribution; see [Link]