and the equation is iri'educible, there are found two expressions for p as a
quaternary quadratic; viz.,
p=iAl + Al + Al + AI- (A,Ar + AiAi + A,Ai)
= Al + Ai + Al +AI- {Ao A2 + AoAz + Ai A,)
the identical relation among the A's being
Ao {Al -A2) + Ai{A2-Az) + Az {A, - ^0) = 0.
In passing, the product of two (and therefore of any number) of primes of the
form 5??: + 1 is written at once Ln this quadratic form, for multiplying either
factor (flo, fli, 02, as, 04) of the first, p, with either factor (ba,bi,b2, bs, b^)
of the second, g , then the product pq is of the above form, where
Ao = Qo (bo — 64) + ai {bt — 63) + 02 (63 — 62) + 04 (61 — 60),
Al = ao (61 — ^4) + ai (bo— 63) + 02 (64 — 62) + 04 (62 — 60),
A2 = 00(^2 — ^4) + fli (bi — 63) 4-02(60 — 62) 4- 04 (63 — bo),
A3 - ao (63 — 64) + fli (&2 - bi) !- 03 (61 — 62) + 04 (64 — bo);
and when
fl« = Oi — 04,- 6', = 6,-64,
12 • ON THE CYCLOTOMIC QUINARY QUINTIC.
the conditions are
flo a'l + fl'j "2 + ^2 03 = a'o a\ + 0003 + a'l 03 ,
^0 ^1 + ^1 ''i + ^'2 ^3 = ^0 ^2 + ^0 ^3 + ^1 ^3 .
A^Ai + ^1^2 + M A3 = A0A2 + A0A3 + A1A3.
It is easily seen that there are four possible schemes for the ^'s, corresponding
to each of two possible choices of factors of p , g to be multiplied together at
first, giving in all, in this way, eight quadratic representations of pq; the process
can obviously be extended. Some further properties of p will be taken up in
considering the transformation of F (ao, •■•); before returning to F ( ),
one simple result on the surface may be obtained. It has just been seen that
p may be expressed in the form (ao+aip+a2P^+a3p') (ao-\-aip^-\-aip^-\-a3p-) ,
where the a's are integers; writing in the values of p", the coefficient of Vs
in the simplified result must vanish identically, and the rational part of the
product must be equal to p; whence, the values of p" being
4p= V5- l+i^lO+2^; 4:p^= - V5- 1 + WlO- 2V5;
4p4 = V5 - 1 - i^lO + 2V5; 4p' = - VS - 1 - i'^lO- 2^5;
the above product becomes
16p = (4ao —01 — 02—03)^+5 (oi — 02 — 03)^+ lOof + 10 (02 — Os)**
+ 2'V5 { oj — (02 — 03)^ + (4ao — Oi — 02 — 03) (oi — 02 — 03)
+ 4oi (02 — 03) } -
The quantity in { } must vanish identically (when for o's are put their nu-
merical values calculated from a given prime lOn + 1 ) , or { j = at once,
on writing
4ao (oi — 02 — 03) = 4 (oi 03 — oi 02 — 02 03)
and expanding. Hence, if p is a prime of the form lOri + 1 , then always
lGp = x'+ 52/" + lOu" + lOr ,
where x, y,u,v are integers. Example:
16 X 61 = 976 = 1 + 5 X 25 + 40 + 810.
Recalling that the oo, • • • , 03 are respectively equal to Aq — At, Ai — Ai,
A2 — Ai, Ai ~ Aif as above defined, also that
Ao + Ai + Ai + A3 + Ai = p - 2,
the o's satisfy
Oo + Oi + flj + 03 = P — 2 — 5a<;
ON THE CYCLOTOMIC QUINARY QUINTIC. 13
and
X = 4ao — Oi — Qi — az; y = ai — ai — a^; u = ai; v — 02 — az.
Combining this result with the formulas for the product of two primes of the
form 10?i + 1 , it is clear that integral solutions of the equation
a;" = M^ _|. 5j,2 ^ 1022 + lOw^
(for X, y, u, V, z, w) may always be found; the method of procedure to be
followed has been sufficiently indicated; e. g., for ?/= 7, the simplest method
would be to multiply together 7 primes of the form lOre + 1 , first factor of
each by first factor of each, and then multiply the result by the conjugate
expression as above, putting in the values for p" .
Returning now to the form (4), and considering the manner in which it was
written down, let {a, b, c, d, g), ( a' , b' , c' , d' , g') be the two complex 5-
numbers giving rise to the forms F (ao, • • •) , F {a'o, •• •) , then if
(a.b,c,d,g) ■ ia',b',c\d',g') ^(A,B,C,D,G),
A = aa' •{■ bg' + cd' + dc' + gb' = ao «» + P (^i «* + "2 ot'3 + as aj + «< «i ) >
B = cb'+ ba' + eg' + dd' + gc' = (ao a\ + ai Oq ) v^l + ("2 a'i + en a^ ) 4'i^i
+ ofsas lAl,
C = ac' +bb' + ca! + dg' + gd' = (aoa'i + a2 «„) v^2 + aia\ \Pl
+ ("3 ai + a4 ofj ) yffi 4'3 I
D = ad' + be' + cb' + da' + gg' — (as aj, + ao aj ) fs + (on «2 + "2 "i ) h 4'i
+ oci a\ \l'l ,
G = ag' + bd' + cc' + db' + g ' = (ao a\ + ai aj ) vt^ + (ai aj -|- ai a'j )\f'3 V^i
+ "J a'j ^ •
Hence, if
B=^,B'; C=4'iC'; D = izD'; G = ^^G' ,
the table (7) gives the following (12), which will be recognised later as con-
taining the necessary results for the duplication (and hence multiplication) of
F (ao, • • •);
G' = ao a\ + "4 0:0 + Pi ("1 a's + 0:3 a', ) + P2 «; orj
D' = ao aj + as a'o + Pi (02 a\ + ai a'^) + pt a^ a\
(12)
B' = ao a, + at Oo + P* («4 «2 + «2 «! ) + Ps aj Oj
C = ao aj + as aJ, -f Pa ("3 «'« + «< a'j ) + pi ai a.
14
ON THE CYCLOTOMIC QUIXARY QUIXTIC.
Whence it is seen that D' , B' , C are obtained from G' by doubling and re-
duction, as it is evident a priori that they should be.
The remaining sections are concerned with the more elementary properties
of the form F {ao, ai, ao, az, a^) which has been constructed, such that
3,125 (a:"- 1) / (x - 1) ^ F;
some aspects being considered where the a's have the definite meanings
assigned for
p = lOra + 1 ,
prime, and some where the a's are perfectly general quantities. AH formulte
deduced in accordance with (7), such as (12), and their consequences, are
equally true for all interpretations of p, for they are formal consequences of
the definitions of ^ , x •
In the next section some simple properties of F ( ) = 1 are considered;
this equation is of importance later.
§ 2. If Ti are the automorphics of F , Sj any transformation of F into F' ,
all such transformations are included in the group generated by Ti, Sj, say
{ Ti, Sj } . Denoting by T{u) the effect of the transformation T on u, let
Xo, Xi, Xi, Xz, Xi
yoi
T= 2o,
Uo,
Wo,
Ui ^ao + ai p + Qi p" + az p*' + 04 p*';
T(U,)=Ul,
transformations T will be sought which are automorphic for F{ao , aj , aj , as , «< )
4
U\ = Ea/ (a-, + P* yj + P-' zs + p" wy + p^'' Wj) .
By S', a linear transformation, F may be reduced to the canonical form
ao «i 0^2 0:3 «4
04 ao ai Oi az
az ai ao ci
aj az at ao ai
ai 02 az at «o
and
= Fc{ao, ■■• ),
ON THE. CYCLOTOMIC QUINARY QUINTIC.
15
and it will be assumed that this reduction has been made. Since in the ex-
pansion of Fc, the coefficients of all the highest powers of the variables are
unity, it follow that in order that T may be automorphic for Fcj and hence
TS' automorphic for F , must
r[(a-y+ p'2/; + p-'2,+ p''wy+ p''wi) = 1; 0'= 0,1,2,3,4).
These five conditions are necessary; it remains to be seen whether or not they
are sufficient. First it is clear that stated in this way the conditions are
redundant; for if
then
/ ( z , j ) = a-y + p'' yi + P^' Zj + f?' Uj + p^' Wj ,
(i=0, •••,4);
n/0".^) = ii/(i.2o = n/o\3i) = n/o\4i);
i i i i
SO that in place of five fundamental equations
i = fl/(i,i); (i = o,---,4),
there may be put the same equation
l = tlfU,i)
in five different forms, if
n/(i. 50 = n/0"' 0)
i i
be taken as the representative form f (j , ) •/(;', 1 ) ■ f {j , 2 ) • f U> 4 )
' f U > ^) ■ It is therefore not necessary to consider five distinct \(f (i, j) JT
forms; one only is sufficient; and this is the meaning to be attached to the
" redundancy " of the first found five conditions. That the so-found sub-
stitution actually transforms Fc into itself is obvious, for the result of the
transformation may be written as the direct product of F (a) and n/(),
and this product is Fc , since the last factor is unity. Hence finally F is trans-
formed into itself by T" S' (S' applied first), where S' is the multiplicative
substitution changing F into Fc , and T is of the form just found. The coef-
ficients of T may be easily deduced from the equations of transformation.
5 =
1
: T =
TT-l
-F^
-V'
n\
Wo
Ml
■Ui
Us
Ui
Ui
Wo
«1
Ui
W3
Ui
Ui
Wo
Wl
Wj
Ui
Us
Ui
Wo
Wl
Ml
U2
U}
Wl
Wo
16 ON THE CYCLOTOMIC QTJINARY QUINTIC.
The problem is reduced to finding the solutions of
Fc (Wo, Ml, W2, Ui, Ui) = 1,
The coefficients Ui are a well-known set of numbers, " Fibonacci's Series,"
according to E. Lucas.* A moment's reflection will show that these numbers
(being found), must occur in this problem; if it were a question of
p = 7n + 1
and the corresponding 7-ic, the analogous numbers would belong to a recurring
series whose scale of relation is (irreducible) of the third degree (or '"' order ").
Putting
ijo = P + P*; '71= p" + p'
(as always), then
Vo = 71 + 2 = 1 — jjo; >?? = 1 — ii; etc.,
whence
10 = 'lo; 10= — 10 + 1; lo = 2j7o — 1; r)l= — 3j?o + 2; jjo = 5?7o — 3, • - • ,
the coefficients being Fibonacci's numbers, and the general term in the series
(found as usual from the solution of a linear difference equation),! is
{ ( 1 + a'5 )" - ( 1 - V5 )"} / 2" V5, whence
Writing €„,. n= p^ vl for m constant, e^, „ is a linear function of ;;o , m with
positive integral coefficients or of 170 or jji alone with integral coefficients, and.
if n is fixed, «„, „ is of period 5 with respect to m . Putting
;„^ (-)-!{(! + V5)'- (1-V5yi/2W,
in the notation of complex numbers used previously, and denoting by e^ ^
the values of p^n\> then
<0,n=(;in-l, Mn> 0, 0, Iht ) '■
«0»n= (/in-1, 0, Hn, 1^. ; 0)
<1, n = (/in, A^-1, /In, 0, ) :
*'l,n=(0, /in-1, 0, /In, fin}
«2,n=(0, /!„, /i„_l, Mn, ):
<2,n=(Mn, 0, Mn-1, 0, ^)
•Amer. Jour., I., "Th(?orie des fonctions numi^riqucs simplement p^riodiques."
t Or more simply by Lagrange's method, " Sur les suites rdcurrentes," e u v r e s , t. 7.
ON THE CYCLOTOMIC QUINARY QUINTIC. 17
«3,n=(0, 0, ll„, /In-l, t^n ) -
«3.n=(Mn> M»> 0, fln-l, 0)
<4jn=(;tn, 0, 0, /i„, JU„_l) :
«4.n=(0, /:„, ;tn. 0, /tn-l).
The ;i's are connected by the following relations:*
a + 6=l; ah= — l, ;:„ = (o" - 6")/ (a — 6); ^.+2 = /v+i + /^;
Po = 0, ^1=1.
For
n = 0, 1, 2, 3, 4, 5, 6, •••, /i„ = 0, 1, 1, 2, 3, 5, 8, •••
that is, (In is the nth " Fibonacci's number." If two or more T's are to be
multiplied together, the formula,t
Ps - M»-i J^+i = (- 1)""^
may be used to effect reductions. Or, among many others the n's may be
used in the form
v'osm(^2^°^r::^j-
Now
whence
/(p)-/(p')-/(p')-/(p^) = C«ovi)'"=i;
and
2m» + Mn-l = A^ + (;in + ;in_l ) = l^n+i',
also
'?0 = Mt. IJO + /ln-1 .
Therefore it is easy to see that if
Mo = t^^lMf^+i', Ml = "4 = Pn/V/i„+2; M2 = M3 = 0,
^e(Wo, Ml, tt2, W3, ''< ) = 1,
and the u's are the required elements in T , where fin is the nth Fibonacci
number.
Also, it is clear that the question of automorphics is identical with that of
finding the fundamental units in the field (p) , and that the above implicitly
contains a solution of this problem; but the units are also easily proved to be
p" jjS* independently, which is done presently. The above coefficients in
•Lucas, Amer. Jour., I, p. 184 eqq.
t Ibid., p. 196, Nr. 30.
Jlbid., Nr. 5.
18 ON THE CYCLOTOMIC QUIXARY QUINTIC.
T are not in general rational, (or integral) ; when these conditions are imposed
on T the automorphics are shown later to exist by a method due to Dirichlet.
At present another aspect of F ( ) = 1 , analogous to the foregoing is con-
sidered; and it is asked, what functions serve to give solutions of f ( ) = 1
in the same manner that sinh w and cosh u furnish those of the Pellian
equation? A general case is considered.
Let
a;" — pi x"~^ 4- p2 a;""' + • • • =^ po =
be irreducible in the domain of rational numbers, its coefficients also being in
that domain, its roots ai, • • • , «„; then anj"- rational function of a of degree m ,
where a is any root, may be reduced by means of the equation, to the form
Zo + 2iq:+ ••• + Zn-iaT^,
m < n .
Putting
Zi= Zo+ Ziai+ ••• + Zn-i o^"' ,
then since only symmetric functions of the roots enter in
N^ILZi,
it follows that N is a, (homogeneous) function of the s's with rational integral
functions of the p's as coefficients; N is in fact the norm of the algebraic
integer ( zo , zi , • • • > 2n-i ) , and is of the form
N ^ (co,Ci, •••) (zo,2i, ••• 2„_i)" = F (Za, Zi, ••• Z„-i).
Let it be required to determine the 2,- so that
F(z9, ••■ z„-i) = 1;
a perfectly general solution will be given if it is possible to determine At
such that
Zo + ziai + Zi<4+ •■• + 2„_i al'^ = Ao ,
20 + Zi as + 2; ccl + • • • + 2n_i aj"' — A\, ,
2o + 2l an + 22 a^ + • • • + 2„_i 0^~^ = An-l ,
AoAi ■ • • An-\ = 1.
If the Ai may be chosen to fulfil the last condition, then c, may be found to
satisfy
F(2o, ••• z„_i) = 1;
for, the determinant cannot vanish unless at least two a.-, ay are equal, which
ON THE CYCLOTOMIC QDINARY QUINTIC. 19
contradicts the hypothesis that the fundamental equation is irreducible.
Hence if the Ai are determined as indicated, so also are the z,-. It is assumed
that all the Ai are finite, hence no one of them can be zero. The equation
AoAi ••', A„-i = 1
is indeterminate to any degree; the simplest solution in which the restriction
that the Ai shall be integers is imposed, is that in which the At are nth. roots
of unity; hence this hypothesis is chosen. Moreover the Ai are to be distinct
roots of unity, so that no condition may be imposed tacitly on the fundamental
equation. This simplest solution gives rise to nl possibilities for the z,-,
for the roots 1 , p , • • • p"~^ of p" — 1 = may be permuted in any way to
correspond to Ao, Ai, •■• , An-i. But more generally, if / (0) , / ( 1 ) , • • • ,
fin— 1) are any functions whatever whose sum is zero, then
Ai= /;^'«, (i= 0, ■■■,n- 1)
furnishes a solution of
^(zo, •••)= 1.
The choice of / ( ) is subject only to the condition
S/(i) = 0;
k is any finite non-zero constant. The function f (i) may be conveniently
chosen as a complex factor of a real integer in the domain of nth roots of
unity; one reason for such a choice being that the resulting /'s lead to a simple
parametric representation of F ( ) = 1 in n-space, analogous to the parametric
solution of the Pellian equation, with which the closest possible analogy is
being sought. Choose therefore, with no loss of generality, k ^ e, the base
of the Napierian logarithms. If w is a prime, any root p of
P"-1 = 0,
except p = 1 is selected, if n is composite, a primitive root. Choose the
" coordinates " Xy, and/ (i) so that
/ (i) = Xo + Xi P-' + X, p=' + . • . +Xn-i pf"-"'.
As i ranges from to n — I ,f (i) takes n distinct values; also
I:p*' = 0;
whence
/(0 = E(Xi- Xo)p*';
20 ON THE CTCLOTOlilC QUINARY QUINTIC.
or changing the notation, this may be written in the form
ii-i
Also
whence
or say
for
*=o
11—1 «— I / n-l \
Z/(0 = Z X;E(P')'' =0,
■^i — V; ( Xl , Xz , • • • , Xn-l ) ,
*=0
In this the Aj are considered as functions of ra -^ 1 parameters Xjt; from the
form of ;•( Xi , X2 ,•••, Xn ) ,
where the m are the periods; from the form of > in order that the /j,- should be
periods, it is necessary and sufficient that for some systems, not all zeros of
the n's, the following equations shall hold:
Ml + M2 + • • • + Atn-i = 2reo Tri
p/il + P^ AI2 + • • • + P""' /in-l = 2wi wi
- (i^V-l).
P"-' m + p'<"-" t^2+ ■■• + P^"-"' M«-i = 2n^i « J
To find /!,• multiply the equations in turn by 1 / 1 , 1 / p' , 1 / p'' , ••■,!/ p^"""' ,
and add; since p is a primitive root, all the coefficients in the result except that
of fij vanish, hence
Uftj = 2wo Tri-j- 2ni tt,/ p' + 27J2 «'/ p'' + • • •
or
ntij = 2wi ( p^"' ki+ ••■ + p''-("-»'- i-„_i ) ;
where
ki ~ rii — TJc,
etc. For reasons of convenience that are made evident in the case of ti = 5 ,
the k's are now replaced by their values deduced from
i'l + A"2 + • • • + ^'n-i — (mod n ) .
It is clear that every solution of this congruence furnishes a set of periods.
In the case n = 5 it is shown that all the sets so furnished are not distinct
from a simple set, to be determined; non-simple sets being derivable from
ON THE CTCLOTOMIC QUINART QUINTIC.
21
simple sets by multiplications. Not seeking here tie statement of the general
analogous result, the simplest solution of the above congruence is
(mod n).
whence, denoting by (oq, • • ■ , a„_i) the complex number Oo + «! P + * * •
+ a„_i p"~'; the periods are given by
n^y = 2«(±1, 0, 0, •••, =f1, 0, 0, •••)•
Regarding now the particular case of the foregoing for n — 5, let
(xo, xi, Xi, Zi, Xi) denote a point in space of the appropriate number of di-
mensions, whose coordinates are
^0 = =^ 1 ,
=^1. •
±1
il = ±l,
0, •
h = 0,
=^1, •
G
kn-l ^ 0,
0, ••
=f1
a:i;
Fixa, •■•, Z4) = 1
is the equation of a particular surface in that space, F having always the sig-
nification of § 1, the coordinates of any point on whose surface may be expressed
as periodic functions of four parameters, as follows:
If
Xi = ^ (Xo -H p' Zi + p'' Xi + p'' Xz + p*' X,), (i = 0,
.4);
then
when and only when
F{zo, ••-, Xi) = 1
Xo Xi Xi A3 Xi = 1
(1)
as is seen at once from equations (11) of § 1.
Particular solutions of (1) are evidently
1);
(Xo, ■••, Z4) = (i, p, p^ p', p^); or =(l,l,p^p^
or = (1, p, 1, 1, p'); or = (1, 1, 1, 1, 1);
of which the first ■•• fourth give respectively 120, 20, 20, 1 solutions, by
permutations; these special solutions are seen later to correspond to " planes,"
" vertices " of the " pentahedroid of reference " in S^ . But these are all
special values of c^"'"'', and a general form to which they all belong is A^^''> ,
where /(i) is so chosen that
i:/(t) = o.
There is no loss in generality in choosing A ^ e, which is done in order to
facilitate expansions, should such be wished.
22
ON THE CYCLOTOMIC QUINARY QUINTIC.
Thus
where
5iriXi = 2p"'e/f">;
Z/(.0 = o.
(2)
A general linear solution, in terms of 4 parameters, of (2) is
/(O = p'Xi+P='X2+p"X3+p*'X4; (i = 0. •■-.4) (3)
or if the X's are to be put in evidence, /,• ( Xi , Xj , X3 , X4 ) . Writing
ef^''> = Xi, or s Xi (Xi, X3, X3, X4),
5a;o = Xo + Xi + X2 + A's + Xu
5iri a;i = Xo + p* -^i + p' ^2 + p^ X3 + pX4
5T2a:2 = Xb+ p'Xi+ pX2+ p^X3+ p^X4
5t3K3 = Xo + p^'Xi + p^Xi + pXs + p=' X4
5^4x4 = Xo+ pXi + p^Xa + p'Xj + p*X4 J
Multiplying in turn by p', p^*, p'', p*'", (i = 0, — , 4), and adding,
(Xo, TTlZl, TTiXi, Tsa-J, T4.T4)
1111
(4)
(5)
So
— (Xi, Xi, X2, X4, X3).
(5^"' 5^/^' 5;rj^" 5;^^" 5^^V'
is a solution of
F {Xo, Xi, Xi, X3, Xi) = I.
The Xi are periodic in the same way that e^ is periodic, viz.,
gt _- gl+2n«r».
the Xi being functions of the Xi, it is required to express their periodicity
with respect to the Xi; viz., quantities m must be found so that
/i(Xi + w. X2 + M2, Xj + ^Js, X4 + ^'4) = /• (Xi, Xj, X3, X4)
ON THE CYCLOTOMIC QUINARY QUINTIC.
23
with the condition
Pi + m + 1^3 + /ii = 2niri .
By (3) this requires
P' Ml + P^M2 + P" /K3 + P" = 27lTi,
where n is a real integer. Hence the m.- must be functions of p such that the
left member is independent of p , and hence n,, must contain p^'"; {n = 1 , • • • ,
4) . Or, using the condition that n must be real, there are four equations to
find the periods:
W + w + /'a + Mi= 2no -ri
Pfii + p'^M2+ p' f3 + p*Mi= 2ni Trt
p* Ml + p* /i2 + PM3 + p' M4 = 2ji2 Trt \ whence
p* Ml + p' M2 + p* Ms + p;i4 = 2714 TX
p' Ml + PM2 + p'' P3 + P^P4 = 27^3 ■^ ,
It will be seen presently that if p is a real prime of the form 5re + 1 , two
(conjugate) resolutions of p into complex p-f actors furnish a set of periods;
indeed this is evident from § 1. The above forms for the periods, on elim-
inating p* , are reduced to
5mi = 2xi ( no jTii, Ks , 7i2 , ni )
5a12 = 25rt' (Ko , «2 , «4 » Wl , 7l3 , )
5m3 = 2x1 ( 7!o , 7l3 , Kl ,ni,n'i)
bin = 2Tn (no , Til , rii , Tis , Hi) ,
2rt , , , • n\
Ml) M2, M3, Mi — -c-(n4> «3» "2. ^i, U;;
2rt
("2,«4> k'i, n'3, 0);
2« , , . ■ n\
— («3, 72i, 7?4, 7l2, 0),
2xi
(n's, ??2, n-j, n'i, 0)
respectively, and since the ( ) must each furnish a multiple of 5 (otherwise
there is no period), there must be the congruence
or writing
what is the same thing.
"1 + n'2+ ti3 + n\ = (mod 5) ,
ni + Ui + ns + Hi = n ,
4no — 71 = (mod 5) ,
which has the solutions
nfr=0, 1, 2, 3, 4; n = 0, i, 3, 2, 1,
(«0, 71) = (0, I, II, III, IV),
say
which give:
(I) no^l;
and
(IV) Tio = - 1; n = l
n = — 1; no
1, ^h ^1,
1
n2^ 0, =F 1, 0,
7l3 = 0, 0, =F 1,
Ui = 0, 0, 0, =P 1.
24
(11) no = 2;
and
(III) no = - 2; n = 2
ON THE CYCLOTOMIC QtriNARY QUINTIC.
n s - 2; no = ± 2, =fc 2, ± 2, ± 2, ±2, ± 2,
± 2, ± 2, ± 2,± 2.
7ms=f2, 0, 0, 0, =F 1, =f1, =1= 1,
0, 0,
0, =f2.
"2
ns
0, 0, =p2.
0,
0, =f1, 0, 0,
=f1, =i=l,
2,
0, 0,=i=l, 0,
=^1, 0,=f1
0,
=p2, 0, 0,=p1,
0,=f1,=p1.
n^s 0, 0,
(all to modulus 5). In each set of 5 the top signs throughout go with the
first written congruence; the upper and lower signs are distinct, viz., no value
of fii deduced from an upper line can go with any m deduced from a lower line.
From these are deduced 2S sets of values for the periods
[5m/2in, 5}i2/^Tn, djis/^wi, 5fnl2iri]
= l(no, Ui, Us, n2, ni); (no, nz, n*, ni, ns);
(7io, ^3, ni, rii, ria); (no, ni, n2, riz, nt)],
but it may be easily verified at once that of these 28 only 4 are distinct modulo
5; that is, all other periods are derivable from these 4 by additions of two or
more of the 4; then by additions of the results, and the originals, etc. In
the subjoined table, each column gives one of the four-sets of primitive periods.
(2tz75).
(1112)
(1121)
(1211)
(2111)
(1121)
(2111)
(1112)
(1211)
(1211)
(1112)
(2111)
(1121)
(2111)
(1211)
(1121)
(1112),
(mod 5).
The mode of formation is apparent; the first complex number (1112) is
written down, being a simplest number satisfying all the conditions; the
remaining numbers in the same columns are derived from the first by cyclical
permutations; the three so-found numbers become the first numbers in the
remaining columns, whose remaining numbers are obtained as in the first
column. Had any other (non-primitive) period number, e. g., (2, 2, 2, 4)
been given, all the 4 periods could have been derived as indicated. Here
ON THE CTCLOTOMIC QUINARY QUINTIC.
25
(abed) means ap + bp^ -\- cp^ + dp* . The four complex numbers in a
column constitute a set of periods, all the columns the four primitive sets.
E, g., from the first column, the values of the m's are:
Mi= (1112) . 2«/5; ,i2= (1121) • 27rz75; /i, = (1211) • 27rj7 5;
/X4= (2111) • 27rt/5.
or
Ml = (P + p" + p' + 2p*) 2«7 5; fii = etc.
The above may be expressed in terms of the prime p-factors of 5, which are
also of importance in the finding of units. Throughout these factors pi, Pi, pz,
Pi of 5 are exceptional, playing the same role as 1 — i in the theory of primary-
prime factors of o + ti numbers. Writing pi , ps , pj , pi for p^ — 1 , p' — 1 ,
P* — 1 , P ~ 1 respectively, the table of periods becomes
Pi
Pi Pi Pi
(2«75)
Pi
Pi Pi Pi
(mod 5).
Pi
Pi Pi Pi
Pi
Pi PZ pi .
or finally.
Hi/2ri
fiif2n
M3/2:rt
inf 2iri'
i/ Pi Pi Pi
1 / Pi P2 Pi
1 / Pi P3 Pi
If Pi Pi Pi
1/pl pi Pi
If Pi Pi Pi
if Pi Pi Pz
if Pi Pi Pi
= [moI (mod 5).
1 / Pi P3 P4
If Pi Pi P3
If Pi Pi Pi
if Pi Pi Pi
i/p2 P3P4
if Pi Pi Pi
If Pi Pi Pi
if Pi Pi Pi
Hence
^3 ( ^1 > ^2 1 Xs >
\i) = Xj(\i-{- Sim, y^i + snii, X3 + 5/X13, Xi + ^Mu)
— Xj ( Xi + 5' A(2I ,
Xt+sVzi)
=
= Xj ( Xi + s" nn ,
X4+5"/i34)
= Zy(Xi + «"V41,
\i + s"' ^U).
or.
(;■ = 0, 1, 2, 3, 4; s, a', s", s'" any integers.)
Xj ( Xi , X2 , X3 , X4 ) = Xj ( Xi + sixii + s' p.ii + s" 1131 + s'" im ,
X2 + Sfin + s'fiii + s" P35 + s'" nn , X3 + s/iu + 5' ,^23 + s" na + s'" im ,
X4 + 5^14 + S' ll2i + S" H3i + S'" /X44 ) ,
for all integral values of the s's; which completes the setting forth of the A'j^ )
26 ON THE CYCLOTOMIC QUINARY QUINTIC.
as periodic functions, and hence the expression of the coordinates of a point
on f = 1 as was required. Not pursuing the study of these functions here
in detail, it is merely noted in passing that all may be expressed in terms of four
simpler functions; viz., if any or all of the X's = 0, there ^nll still be solutions
of F () = 1; also, since
that is
X,-(Xi,X2, Xs, X4) = X,(Xi, 0,0,0) -.Z.CO, X2,0,0) • .Y.(0,0, X3,0)
X:{0, 0, 0, X4).
From this, and allied relations, it is possible to express the Xi algebraically in
terms of the simpler incomplete functions of a single argument. Again
X ("Si, X2 , X3 , X4 ) = X ( Xi — oi , X2 , — 02 , X3 — 03 ,
X4 — 04) • X (Oi, 02, 03, Oi) .
The addition theorem may be deduced without any difficulty; abstractly it is
identical with the theorem which permits the duplication oi F {) , and the
formulae requisite for either implicitly contain the other.
Reverting now to the equation F = 1 , it is clear that the problem of iso-
morphics (as treated in this section), and that of finding the fundamental
units in [ p] , th« field of numbers constructed from complex fifth roots of unity,
are identical. It is the form of this equation F = 1, and the necessary manner
of its solution, that enables the present forms to be included strictly with the
two correspondents
4 (a;"- l)(x- 1) = Y(xy- (- l)'^'' Zix)\
and
27 (x" — 1) / {x — 1) =w^+ p2h 1^ + ppi u^ + Spici'u .
These forms of degrees 2, 3, 5 could all doubtless be considered, — as all
possible forms of a " Kreistheilungskorper " — at once, but the general method
or theori' is sometimes helped by the use of elementary methods. In the
theory of any such forms, the solution of Fe = 1 (or the corresponding equation)
is fundamental, aiid is identical abstractly with the determination of a system
(if more than a finite number exist) of fundamental units for the field consid-
ered, that is for ( x" — l)/(z— 1 ) for a given prime J) . For p = 3 the number
of fundamental units is finite; the units are ± 1 , ± u, ± u-, co an imaginary
cube root of unity. It p = 5 it is easy to show (see below) that there is but a
single " system " of fundamental units that is independent; i. e., in terms of
whkh all the other units can be expressed. But, as proved by Dirichlet,
if p > 5 there is an infinite number of such fundamental systems. For
ON THE CTCLOTOMIC QDIN.VRY QUINTIC. 27
p= 5 (see below) the single system is ± p^tiJ; ijo = p+ p*- Thus for
p ^ 5 there is an essential distinction from the classes p > 7; the case p = 5
is intermediate between two very different species, viz., p < 3 and p > 7;
in the first fall those forms which depend on a finite number of units, in the
second those whose complete expression requires the use of an infinite number
of systems of fundamental units, each system containing an infinity of units;
the " dividing case " p = 5 is a link between the two others, for the number
of systems is finite (one), and the number of units infinite. Nor does the
sequence of coefficients in the case p = v admit of a simple explicit form for the
automorphics, as was pointed out. In short, from this standpoint simplicity
ends once the case p = 5 is passed. It may be worth while to indicate another
reason for the detailed consideration of these and allied forms by strictly ele-
mentary methods. Denoting by any law (or process) of combination, it is a
definite question to ask what forms fijft, fs of the same kind, that is, on the
same numbers of variables, and of the same degree, e. g., the/'s may be three
m-ary 7i-ics, satisfy the relation
0(Ji,f,)=fz.
If is X (multiplication), and the domain of the coefficients identical with the
rational domain, the question is completely answered by Gauss' theory of
composition, if the /'s are binary quadratics; if the forms are m-ary m-ics,
Ka-GRAnge has given sufficient but not as yet proved necessary conditions,
(0 still being X ) • In addition to these well-known solutions there are a few
isolated examples of the same kind; e. g., Eisexstein's theorem on the trans-
formation of the discriminant of the binary cubic into its third power; a result
to which he attached great importance, especially in his triplication of class
problems in the theory of binarj' cubics. It is possible to regard the arithmetical
theory of forms from the standpoint of the operation (not necessarily multi-
plication); clearly such a view presents new problems and interests, especially
in the domain of quasi-algebraical number imaginaries. From this basis, the
simplest forms arising for = X are the 2-ary 2-ic3; if either the degree or the
order be independently extended, the theorj' becomes at each step more
special. Thus for quadratic forms of 3 , 6 , 7 or more than S indetcrminates the
composition-theory in any general form is non-existent, so that one point of
analogj' is lost by merely extending the order; but if the order is constant
( = 2 ) , and the degree be extended, any theorj- becomes special to the vanish-
ing point. But if both order and degree be simultaneously extended, complete
analogj' is restored, and as the special form i^ + y^ was the basis (historically)
for the composition theory of quadratics, so the 0-forms here are the guides,
and the special theory for at least algebraic numbers of the nth degree, n
being the order of the form, is approached, 'W'here = X the forms considered
28 ON THE CYCLOTOMIC QUIX.VRY QUINTIC.
distribute themselves according to degree (and order), so that it is only neces-
sary to consider the cases of prime degree. The properties that are easily
approachable of such forms are in direct analogy to those of binary quadratics,
owing to the existence of an analogue for the Pellian equation.
The question of units, etc., is briefly considered in the next section (end) ;
first the " divisors of F," and allied questions are taken up.
In § 1, F was derived in the form
F (oro, '•• , cii) = iolih^zU,
where
^0 = (ao+ s'iai+ 'rsa2+ r-s aj -}" TTiOH, 0, 0, 0, 0);
li = (ao, iriQii, TriaifTzaz, irioj); f 2 = (oro, Tsas, Tiai, xiofi, tjotj);
f J = (oro, TTiOi, ■K^a^, xiai, rjofs); It = (ao, ""404, Traaj, X2a2j nai);
where, e. g.,
to = a'+Ti(6' + c'p* + +pxi+ p= Z2 + p' a;3 + Pi 3^
is prime to p (and hence to pi and pz » for p = pi pz ) j and further let
{/(p)P = PP?P2 (modg)
which implies that ppl pn is a quintic residue of q . The congruence is equiva-
lent to the following,
that IS
{/(p)-Ti)(/(p)-pn)(/(p)-p'Ti)(/(p)-p'n)(/(p)-p*n)
= (mod q);
say
ffo (p) ■ gi(p) • giip) • gi{p) • gtip) = 0;
that is, say,
G(p)=0.
and therefore by the usual theory,
G(p)-(?(p')-G(p^)-G(p^) = 0.
This last is the product of 5 factors,
i = 0, 1, 2, 3, 4;
(/(p) - P'Vi) (/(p^) - p«tO (/(p') - (p'-Va) (/(p^) - p«T4) = [i],
say; consider now [0]. If
(/(p)-^i)(/(pM-r4)=/'(p),
then
(/(p')-T2)(/(p')-7r3)=/'(p2),
and
IO]=/'(p)./'(p^);
also
f'(p)=f(p)fip*)-Tif{p*)- rif(p) + p,
•Crelle, 30; 107 sq.
t Ibid., p. 115.
30 ON THE CYCLOrOMIC QUINARY QUINTIC.
whence [0] is of the form X + wi Xi + ^rj Xn + irj Xz +^4X4, where the
X's are to be examined more closely. It is evident that {f {p) ^' {p^))
Xo= {f(p)-fip*) + p] ' {/(p^)-/(p') + 2>}
Xi = pJip')-fip)-f(p*){fip-^)-fi(?) + p]
^i = P3f(p)S{p')-f(p'){f(p)-fip') + p]
Xi = P2fip')f{p')-f(p') {/(P) ■fip') + p]
X* = Pif (P*) ■ f (p') - f (p) {f (p')f ip') + p} ,
say
Xi = h{p),
then, as is evident a priori by doubling and reduction,
Xi = h(p); Xz=-h(pn; X2=h(pn; Xi = h{p'),
and Xo is a real integer, = {a + pno + yvi) {a + yrjo + firji) + p^ + p
+ p(2a— /3— y), (see below f or a , P, y) .
Hence IIo [i] is of the form F (Xo, ■■• , Xi), where the Xo, • • • , X4 are
functions of the same kind^the ao , • • • , ai'm F {ao, • • • , 04 ) ; for, as in i'' (a )
each of the variables a is a complex 5-number with real coefficients, and all
four contain the same coefficients in different cyclic orders, so here the
h(p)> •'• > h(p*) clearly satisfy all the similar conditions. Hence G (,p)
• G (p'^) • G (p^) G (p*) is identical in all respects of form to F (a) and it
has been so constructed that it is divisible by q. Writing for the moment
f(p')^\,
the above expressions for Xo, •••, X4 become respectively ai 02 03 at
+ p (0104 + 0203) + p^; PiaiUi — a^aiai — pui, ps ai a2 — 03 ai Oi — pat
Pi 0.3 tti — Oi 02 Oi — pui', pi 02 04 — oi 02 03 — pa\ . Now these are all func-
tions built up by multiplications by constants and additions of the results, of
the essential variables in the G-form, whence if these essential variables have
any common divisor different from unity, the above functions are also divisible
by this common divisor. From these functions, by obvious multiplications
and additions are constructed the following:
Pi az a\ + pai 03 + p^; 2>2 04 of + pa^. 04 + p';
Pz di al + pai 04 + p"^; p\ 02 al + pan oj + p^
Let now oi, 02, 03, 04 be conjugate complex factors of a real number A^ ,
that is
Oi 04 = 02 O3 = A,
which is necessitated by the assumed congruence of = 5rf , and the thereby
ON THE CYCLOTOMIC QUINARY QUINTIC. 31
determined three remaining congruences
a% = vl; 0.1 = ^1; a\ ^ tj\ {moA. q) ,
and it be assumed as is legitimate on the assumption already made that ai
is prime to -ppl pt , that A^ and p^ are relatively prime, then, by additions, etc.,
of the last found four expressions, the common divisors of the essential variables
are all common divisors of pi 02 of; p2 04 of; Pi 13 ol, Ps ai a', and by what
immediately precedes, these last are relatively prime. Therefore the essential
variables in the G-form give a proper, i. e., primitive, representation of q in
the form F . In other words, all letters having the same significance as above,
the necessary and sufficient conditions that o be a divisor of the form F are
//^7? Tz-f2^ jr^B pni^
any one of which entrains the others, and { ra / n } is the quintic character of
m with respect to n.
The actual reduction of G to the form F , and hence the primitive repre-
sentation of an appropriate number, i. e., one satisfying the congruence con-
ditions, is readily effected by means of the formulae for multiplication of
6-numbers, etc., given in § 1 or § 3; or the writing of G in the required form
may be put forth db initio; it is however less laborious to write down the re-
sults in accordance with § 3.
As shown above, [0] is of the form X,o + tti Zi -(- t2 X2 -}- ttj X3 -|- ti Xt;
and G is then F (Xo, Xi, X2, Xz, Xi);it is required to find the forms of the
X's which must not contain any tt's explicitly, and must be as proved, with
the exception of Xo which is real, complex numbers of the forms 6 -f- cp* + rfp'
+ gp^; b + cp^ + dp + gp*; b + cp'^ + dp* + gp; b + cp + dp'' + gp^ . Now
/(/')-/(p')=«+P^+p'r+p'7+pV,
fip') •/(p=') = a+PT'+p''^+p'/3+/r,
a s xl + xl + xl+ 4 + xl;
fi ~ XoXt -|- Xl X2 + X2 X3 + X3Xi + xo xi;
y = Xo X3 + Xl Xl + X2 cci + 23 Xl + Xo X2 .
Hence
{/(p)/(p') + p}{/(p')-/(p') + p}
= (a+ firio+ ym) ia+ 7'?0 + 0Vi) + p' + p (2a - /3 — 7) ,
which is a real integer; say I . Put now
To- 4+XiX2+X2Xi + Xz Xl + X4 X3; r3 = Xo .T2 + Xi X4 + X; Xi + xl + X4 Xo;
hence
where
32 ON THE CYCLOTOMIC QUINARY QUINTIC.
fi = XoXi + xl + X2 xz + ars a;o + Xt x^; u = xo xi + a:i 3:3 + X2 Xq + 2:3 a-j + yj;
r2 = a;o 2:3 + a;i aro + arf + ars 3:4 + x, a;i; Pi = So + si p + S2 p^ + S3 p' + 54 p*;
and
-X'o = ro(5o —Xo) + ri{Si — a:s) + Tiisz — Xi) + rj(j2 — a;4)+r4(5i— a;2 )— paro,
^i = ro(si — X3) + ri(so — a;i) + r2(si — X2) + tjC^s — a:4)+r4(52— a:o)— pa-i,
T2 = ro{s2 — Xi) + ri{si — Xi) + r2{so — X2) + r}{Si — xa)+ri{s3—Xz) — fX2 ,
X'i = ro(*3 — 3:4) + ri(*2 - a;2) + r^^si — Jo) + rsC^o — a;3)+r4(54— a;i)-2)a-3,
XI = ra{Si - 2:2) + ri{Si ~ Xq) -\- r2{si — Xi) + T^isi — x{)-\-ri{s^-xC)-V^\',
It may be easily verified that the coefficient of 7r4 in [ ] is ( Z; , X\ , Zj , X^ , Z; ) ,
a complex 5-number -ftith real coefficients, for the actual coefficient is
^(p) = Pt/(p^) •/(P^)-/(P) {/(P^) •/(p') + p}.
The coefficients of jra, V2, tx are F(p^), II ip^), II {p*) respectively, or
(ZqjZs, Aj, Z4, Zj); (Zq, Zj, Z4, Zj, Zj); (Zo, Z4, Zj, Z2, Z'l) ,
and the absolute term is I. Finally, putting 6 , c, J , ^ for Xp — Z; ; Z'j - Z^;
ZJ — Zi;Zi — Z4, then [0] becomes 7+ {b,0 ,g,d,c) ti+ {b,d,0,c,g)T2
+ (b, g, c, 0, d) 7r3+ (b, c, d, g, 0) Tn; or say
[0] = ao + ai Tti + a2 T2 + 0:3 xj + 0:4 T4,
then by the preceding parts G is F (ao, ai, orz, as, cii); that is the number 5 is re-
presented properly in F by a system of variables of the same nature as those
in the original F. In this connection note the distinction between F(a)
and F (fi) , where in the first case the variables a are restricted to be five very
special numbers, viz., ao a real integer and aj , • • - , a4 four complex 5-numbers
conjugate in pairs, and not independent of ao; whereas in F (/3) the variables
may be any quantities whatever. It is for this reason that b, c, d, g (and ao)
are called the essential variables. When attention is fixed on the essentia],
and not on the a-variables, the form obviously has a wider significance than
its connection with the cyclotomic function out of which it arose, although in
many respects the formal aspects of the two interpretations are identical.
When F = k is considered as a locus of points in 5-space, the two interpreta-
tions arc united; all possible systems of values satisfying the equation represent
points on the surface, whereas only those points which belong to a certain
5-fold net-work (corresponding to the nets of unit squares in the plane, and
of unit cubes in ordinary space) are relevant to F = k considered as a cyclo-
tomic function.
ON THE CYCLOTOMIC QUINARY QUINTIC. 33
Some further properties of the numbers occurring in the above are now
considered.
A " wnumber " is defined to be a linear function of vi, t^, va, T4 in which
the coefficients are rational complex numbers constructed with fifth roots of
unity, a special case being that in which all the coefficients are real rational
numbers. It will now be shown that no ir-number with non-zero coefficients
can vanish; or what is the same thing that iri, ttj, tz, -iti are linearly inde-
pendent in ( p ) . Throughout, the fifth roots are to be so chosen that
""l S"4 = T2 ffS = P .
The jr's are given by
ir\ - pp\ Pi = 0; rl- ppl Pi = 0;
■^l - ppl Pi =
0; 7rJ-pp|p3
From these
_I0 _ „« _5. _I0 _ ~.B ^S.
^1 — Pi ^2> ^i — Pi T'it
whence
< = Pi t|;
_10 _ .-.B _6
^3 — Pz^l'
»Jt _16 _ „5 _10
Ps Tl — Pi ^3
and
xi=p*p«7r|,
■Ki — Pi T2 — Pi Pi TTi ,
so that the ir-number
N = a+bri + c' ir2 + d'ir3 + g'Ti= a+ bwi + — vl-\ tt? + -r" ^i
Pi P1P2 PI Pi
s o + Stti + CffJ + dTrl + JttJ ,
say, where, a, • • •, ^ are rational. Hence / = — ppf ps is a common root of
a^ +/ = and a + hx + C3? + do? + gz'^ — .
Eliminating successively the highest powers and absolute terms from the two
equations, and repeating two relations with rational coefficients are found,
R£+Sx+ T=0; R'x'+S'x+ T' = 0;
the R, •• • , T' being rational integral functions of a , •••,/. Eliminating s?,
iSR'- S'R)x+ {TR'- T'R) = 0,
whence x being finite and determinate, neither of the () = 0,andarisa rational
function of a, ■••,/. Hence VppJ p2 is rational, that is '^plpl is rational
which is impossible, since pi , p2 belong to different resolutions of p into factors.
Therefore N =^ 0; but N was any x-number with non-zero coefficients; hence
the result is established. From this it follows that if two 7r-numbers are equal
then they are identical; for let
a+ biri+ • • • + girt = a' + b' TTi + •• • + g' ivi,
34 ON THE CrCLOTOMIC QUINARY QUINTIC.
then
(a-a')+ ■■■ + ig-9')7r, = 0,
which is impossible unless
a = a', ■•' , 9 = g'.
Now, if possible let F (a) be resolved in two distinct ways into the product
of five ;r-numbers. The coefHcients of the r's in the various factors are
independent of the it's, hence any factor in either resolution cannot be the
product of two factors in the other, for if it were such a product, this factor
could be written in a form in which the coefficients of the r's would not be
independent of the it's . Hence in some order the factors of the first resolution
are equal, each to each, to the factors in the second resolution. Therefore,
as has just been seen, the factors, in the two resolutions are identical. Hence
F {a) is uniquely resoluble into the product of five ir-numbers.
A further definition is now needed; the 7r-number a + ri/ (p) + ^2/ (p^)
+ ""s/Cp') + ■!^if (p*) in which a and the coefficients b, c, d, g in
f(p)^b + cp + dp' + gp' + ep'
are real rational numbers, is called a rational Tr-7iuviber; if in addition the
coefficients are all integers, it is called a T-infeger.
It will now be shown that F (a) is the product of five Tr-integers. For,
bp + c + dp' + gp'^ ic-b)+{d-b)p'+ig- b)p'- bp^;
bp^ + cp + d + gp' =(d-c) + (g-e)p'- cp' + (b - c)p^;
bp' + cp^+dp + g^ig-d)-dp'+{b- d)p' + (c - cZ)p=;
bp* + cp' + dp' + gp=-g+ib-g)p'+{c-g)p'-i-{d-g)p'.
But F (a) =/o(a, b, c, d, g) •■■fiia, ■•■ , g), where
/o ( ) = a + ^1 (6 + cp*+ dp' + gp") + tto (6 + cp' + dp-^ gp')
+ 7r3 (& + cp" + dp' + gp) + 7r4 (6 -f cp + dp-" + ?p') ,
and /i ( ), • • • ,/4 ( ) are obtained from this by changing ti, t^, tj, r^ into
pri , P^ TTt, p' T3 , p* Wi] p' ri , p^ TTo , pVi , p^ Ti]
p' Ti, pTTi, p^ ITS, p^ T4; p^ xi, p' a-2, p' Tz, pr^;
respectively. Hence F (a) is the product of/o(a, b, c, d, g); fo{a, —g,
b — g, c — g, g — d); fo(a, g — d, — d, b — d, c — d); fo{a,d—c,
g- c, - c, b- c); fo(.a,c-b,d-b,g-b,- b); and by definition
/o ( ) is a TT-integer.
Clearly, the product of any number of Trrintegers is a r-integer, as is also
ON THE CYCLOTOMIC QUINARY QUINTIC. 35
the sum or the difference. Hence any algebraic function of 7r-integers is a
s--number, for it is easy to see that the quotient of two 7r-integers is a tt-
number. ' In certain cases, to be examined presently, the quotient of one
TT-integer by another is again a Tr-integer.
If TTi and T2 are two ;r-Integers such that ti / ^2 is again a ir-integer, then
vi is defined to be a T-unit. These units are of fundamental importance in
the theory of these forms, and may be investigated by a method, or rather
slight modification of the same, due to Dirichlet; " Verallgemeinerung eines
Satzes aus der Lehre von den Kettenbriichen nebst einigen Anwendungen auf
die Theorie der Zahlen,"* also " Sur la Theorie des Nombres." In passing, it
is noted that the basic idea in questions of this sort is as old as Leonardo
of Pisa's identity concerning products each factor of which is the sum of two
squares. Dirichlet himself points out that his theorems arose in considering
questions of this genre given in Lagrange's theory of the multiplication of
forms. There is a well-connected chain of theorems from Leonardo, on
through Fermat, Euler's four-square result, and Lagrange's wide gener-
alization — (in one direction only, however), up to Dirichlet. Of this general
kind of theorem is Lagrange's quaternary quadratic identity, so fundamental
in the theory of these forms. But the full importance of self-reproducing
identities, of which the Lagrange-Dirichlet kind form only a special type,
was fully and clearly recognized by Eisenstein, " Uber eine merkwiirdige
identische Gleichung." f It is suggested that the general theory underlying
all these isolated phenomena, may be a natural method of approach to the
more arithmetical and less algebraical consideration of the properties of
integers. That such has proved to be the case in the existing theory is obvious
upon slight reflection. The most general theory of this sort would probably
relate itself to operators , as suggested above.
Applying now Diricpilet's methods (as cited), it will first be shown that
it is possible to assign values to the variables in a :r-number which will, render
this number less than any assigned finite non-zero quantity. The application
may be made only on the fact already proved that no 7r-number with not all
zero coefficients can vanish. By an adaptation of the method, it may be
shown that there is at least one integer which may be represented in an infinite
number of ways integrally in F; next, from any two of these representations
a solution of F = 1 is found, and it will then be shown that all solutions, as in
the theory of the Pellian equation are derivable from a certain determinate
fundamental set. The first part of the proof consists in showing that if
Pi = ao + P' cii TTi + P^' Oil T2 + p'' as Ts + p*' at tt^ ,
* Werke, I, Nr. XXXV.
t Crelle, Jour., 1844, especially the remarks at the foot of p. 105.
36 ON THE CTCLOTOMIC QUINARY QUINTIC.
then integers a.- may be assigned which will render TLUo I Pi I less than a finite
positive constant (to be determined), in an infinite number of ways. It is
essential to note, that since x,- is irrational, then for | a,- 1 or a,- selected, ao
may be assigned as an integer so that < | Po [ < 1. Consider the set N of
( 2 1 n I + 1 ) integers,
-In|, -|n-l|, ... -1, 0, +1, ■■•+\n-l\, +\n\^N,
and assign to | or,| (i = 1, 2, 3, 4) any one of the (2| k| + 1)^ possible
sets of 4 values selected from N . Now if ai is given, so are implicitly 02,03,014,
for ai, a4 are conjugate, and either one determines b , c, d, g , the essential
variables, that both a^ and as are given. By the assigning of values to | cti | is
meant the assigning of values to the essential variables in a,-, and the sub-
sequent determination of | a, 1 . There are clearly ( 2 | n | + 1 )^ — 1 possible
Don zero sets. It is immaterial whether the values of | on | , 1 0:2 1 , \az\, \ai\
together are spoken of, or whether the values of the essential variables in a
single a< ( 1 , . . • , 4 ) are spoken of, so long as the exact meaning of the assign-
ment is borne in mind. It is also to be noticed that ao is a real integer. Let
I tti 1 be such a set of four values, then ao may be assigned as an integer so that
«o + I ai 1 1 iTi i -f [a2 I k2 I + I 03 1 1 X3 1 + I ai 1 1 T4 I = Po
(0< Po< 1); that is, Po is a positive proper fraction, and the a's are all
integers. There are (2|7i|-f- 1)*— 1 such positive proper fractions, for if
|ai|^ ... =|a,| = 0,
then ao = . Now divide the interval from zero to unity into some number of
parts less in number than the total number of positive proper fractional
values of Po, so that at least some one pair of these values for Po must have
their end-points in a same interval; it is convenient to divide the — 1 in-
terval into 16 I n \^ parts. Let P'q and PJJ be a pair whose end points lie in
the same subdivision, then the absolute value of their difference is less than
l/16|7iP. Put
Fo-P'o = Po=(|a;i-ia;i) + (|a;i-ia';i).|nl + (I«2l-k';i)-U2i
+ (kal — l«3l) • U3I+ (|a'J — lai'l) • I ^r^ | = ao
+ i ai I • I Ti H" I a2 i ■ I 7r2 1 + I as [ • | tts i + | a4 1 -1^41,
then [a.|^2|n!, (i=l, ■■■, 4), and < | Fo - FJK 1/ 16| n|^
the first part of the inequality coming from the fact that the difference is
positive, and cannot be zero, for if it were the two non identical rr-numbers
would be equal, which is impossible. Again, since
\ai\^2\n\ (i=l,---,4), l/|ai|^Si/16ln|*,
ON THE CYCLOTOMIC QXnNARY QUINTIC. 37
hence
< I ao + ai Ti + aj ir2 + as 7r3 + a4 T4 I < 1 / 1 a^ I ,
where | a ] is the greatest of a,- , the a,- being chosen as above.
Consider next
|Pil = ko + aip (1 + X)^ which is represented by an infinite number of sets of values
for the a's.
Again, the congruence a< s a.' (mod m) has at most m incongruent roots;
hence there are at most m^ incongruent sets (a,), {a\) for i= 0, •••, 4.
Let two such sets (of 5 values each) be selected from the co sets which satisfy
F (a) = m, say (a), (a'); and put
/,• (a ) = ao + ai p*' TTi + ■ ■ • + at p*''
TTi.
38 ON THE CYCLOTOMIC QUIXARY QUINTIC.
Prom (a ) and (a) may be derived ( as Dirichlet, " Sur laTheorie des Nombres,"
1. c), a solution oi F {a) = 1 in the following manner.
/o(«)//o(«')^/o(a')-/l(a') •/2(a')-/3(a')-/4(a')/F(a')-.-(A)
By reductions which have been frequently used in former sections, the nu-
merator on the right is of the form Ao + ^i tti + A^ t2 + Aiir3-\- Ai -Ki,
and the denominator is m by hypothesis. It \d\\ be shown that ^4.- =
(mod m) and hence that/o {a) I fa {a!) is an integer. Dirichlet remarks that
if the variables (a) be replaced by the (a') , the resulting A\ mil be congruent
to the Ai, since (a) = (a') (mod vi) , and deduces the result in question
therefrom, but without some changes (replacing ttj, ttj, ta by their corre-
sponding TTi expressions); the conclusion cannot be drawn in the same manner
in this case; it is easily seen directly hovrever; for
m =/o («') •/!(«') -h (a') -/3 («') -U W),
Aq + ^1 :ri -f Az TTi + /I3 ""3 + Ai -Ki
^fo(cc)-Ma')-fAa') • /,(«') •/4(a'),
•whence
Ao + AiTri-{- •••-{• Ai TTi — m
= (/o(a)-/o(a'))/i(a') ■ h (a') f^ (a') fi {a') ,
and.
/o (") — /o («') = («o — a'o) + Ti (ai — a'l) + :r2 (ao — a'a) + tj (as — q's)
+ T4 (ai — ai ) = 7« (org + ^1 '^i + "^'2 '^z + as TTa + a^' tti ) ,
since
a,- s a't ( mod 7;2 ) .
Hence
Ao + -"^l TTi + Ai 172 + ^3 TS + -^4 T4 — m
= mvlo + mA'i TTi + mylj ttj + mA'i its + w^-^i ^4 ,
but if two TT-numbers are equal they are identical, hence every coefficient
Ai, Ai, Ai, Ai is a multiple of m, so also is Ao — m, and therefore Ao is
divisible by m. Therefore the right hand side of ^ is a r-integcr; so that if
(a') . (a") are any two sets of a-values as in the above, then
/o(«')//o(«")=/o('«);
similarly
/o(«")//o(«')=/o(«),
where the coefficients in /o (a) and /o (a) are integers; and clearly, by the
same process exactly,
ON THE CICLOTOMIC QUINAKY QUINTIC. 39
hence f or i = , • • • , 4 ,
n/.(a')//.(«") = n/,(a),
that is,
or, the values of a determined as in the foregoing are a solution of i'' (a) = 1 ,
and there is evidently an infinite number of them.
It may be noticed here that Xo is real; for Xi , Xi also Xj , Xz are conjugates,
either member of a pair being derived from the other by changing p into
p~' wherever it occurs, and therefore Xi X2 X3 Xi is real, that is, 1 / Xo is real,
therefore Xo is real. [This is also evident when the transformations of F
are considered, where the actual formulee show the result.] In order to
bring out the analogy betv/een F(a) = 1 and the Pellian equation, it is
necessary to consider a quantity Introduced into Arithmetic by Eisenstein,
the " regulator," or rather considerations of a similar nature. First it will
be shown that the ratios of the real logarithms of A'o and any X,(z4= 0),
are incommensurable. Throughout, unless the contrary is expressly stated,
the solution or " point "(l,0,0,0,0)oni^(a) = l,is excluded. If possible
let Xj* = Xf , then by successive changes of -ai into p-jri , etc., and from the
expressions of the A^'s as 7r-integers, it follows that
Now Xo is rational only for the excluded point, and
XS'---Xr= {F(a) }"= 1
for all values of m , and thence Xq = 1 , which is impossible (except at the
excluded point). If possible, let X" = A'?", then as before
A-i — A.2, ^2 ~ -^3) -^3 ~ ^*i -^« ~ -^0-
Raise each side of the respective equations to powers m^ , m? n , m^ n^ , mn^ , n^ ,
so that from all, A"^"' = ^t. , fn, n independent, which is impossible. Hence
the ratio of the real logarithms of any power of Xo and any power of another X
is incommensurable, for otherwise some two powers would be equal, which
has been shown to be impossible.
Considering still the equation F (a) = 1, it will next be shown that any
expression of the form A'" Xj A'3 X| where m, ■ • • , s are any integers may be
replaced by a corresponding expression in which all the exponents are positive
integers, and which expression may be written in the form Xo ( ~ Xo , " equiva-
lent to Xo" ) . For It has been shown that Xo , • • • , X^ being the ^-factors
of f (a), then A'l, Xi, Xz, Xi ~ Xo, since all are 7r-integers. Also A'5 is
again of the form A'o, and the product of any two Xi{j , i = 1 , • • • , 4) is of
40 ON THE CYCLOTOMIC QUINARY QUINTIC.
the form X; and therefore ~ Xo; all of which is seen directly from the early-
part of this section; or may be inferred from the formulse for multiplication
given § 3. Therefore in particular when vii, mj , rris, m^ are positive in-
tegers, XJ" ••• X^ " Zo. Let now Xo be the first factor of an F (a)
that is equal to unity, then Zj is the first factor of some other F (a) which
is also equal to unity, for Z'J ~ Xo . Whence
XjcoZi; •••J2"Z4, so {F(a) }" ~ {F(a) } = 1.
Let n be negative, = — m; then
zs = 1 / xs* = zr x? x^ xr ~ xs- ,
for
(XoXlX2X3X4)'"=l.
Similarly a negative power of any other X may be replaced by a positive
power of an equivalent X by multiplying by a suitable power of 1 ^ F (a) .
Hence any expression XJ" • • • Xj"* where the n's are any integers, may be
replaced by a corresponding expression in which the m's are all positive, and
which ~ Xo . Having reduced Xj" • • • X^ in this way, the resulting Z'o deter-
mines a solution of F {a) = 1, either for thea.orfor the essential variables.
mi = m2 = TTls = 7?l4 = ,
If the excluded case is refound; hence the totality of solutions is found by letting
mv(i= 1, 2, 3, 4) assume all integral values from — » to + «> ; that is,
there are oo* solutions derived in this way.
Next it will be shown that Zp Z^ A'J" A'f' may be made to approach
as nearly as desired to 1, when mi, mj, mj, mi are arbitrary national
numbers, the quantity
mi log Xi + nii log X2 + mz log X3 + mt log X4 = I,
the logarithms being real, is of the form
I = mi h + mih-h mz h + m^ li
where no lil Ij {i "¥ j) is commensurable. For any A^• ~ Xo as has been
shown, also it was proved that hlh, is incommensurable, and the same
conclusion follows in the same way if for /, any I different from 7o is put.
Since lot h is irrational it is possible to find integers mo, mi such that
I mo 7o + mi Zi I < €, where t is a previously assigned small positive quantity.
Similarly for j mo 7o + m,- /,■ | , (i + ) . Also the ratios of h / h and /i / /<
are incommensurable, hence integers mj, m\;'in^, m\ may be found for e', 5'
previously assigned, which make | m\ h + Tn\ Ii\ihl; \S'\>h\S\.
Hence log | Xf • - • Xj" | may be made to approach zero as a limit, when
integers mi, • • ■ , TO4 are properly chosen; that is,
Xmt...xr= * 1+ «, « = o,
for TOi , • ■ -, mi some integers. Hence
izr---xri ~iXoi
may be made as nearly equal to unity as desired, and since the m's are all
integers, the values of the variables in Xo will be integral; therefore integral
values of the variables may be found which make
|XoI = l,
and at the same time
F{a)=l.
Further on, F (a) is briefly considered as a surface in 5-space; it is ad-
vantageous here to examine some o*' the simplest properties of such a surface.
First, it goes to infinity in the real direction given by Xq = 0, or Xo =
is a real asymptotic " plane "; also each Xi = is an asymptotic plane, but
these last are imaginary; they intersect In a single real line, for
X.- = Zo + P-' ii + P" U + p="" U ,
where L's are real, and
X.-=0(i=^ 1, -••,4)
have in common
Lo = Li = L2 =^ Li = 0;
hence
F(a) = l
has one real asymptotic plane
Xo-1,
and one real asymptotic line; also (1,0,0, 0,0) is a point on the surface. A
point whose coordinates are real integers is called a "net-point"; and the
surface is considered as being immersed in the net work formed by all planes
parallel to
Xo, Xi, A2, X3, X4 = 0,
at unit distances apart. Considering the sheet of the surface on the positive"
side of X'o = 1 , that is, the space which includes net-points whose coordinates
42 ON THE CTCLOTOMIC QUINARY QUINTIC.
will make F (a) positive, the volume bounded by the surface, a plane parallel
to Xo = 1 , and three planes parallel to
and at finite distances from them, does not contain an infinite number of net-
points (if any), for this bounded part of the space has in it no portion of the
surface extending to w . Hence, since Xo = 1 cuts the surface in a space which
does not lie wholly (if at all) at oo , the volume
Xo=l, Xl = Cu Z2 = C2, X3=C3, 0 ^ ,
where n is arbitrary, and it does not follow as before that n may be so chosen
that/o=0. But if
-n = E{hlT,)
then
unless IqI I'o is an integer. Hence when
io = «-*Jo
k an integer, n cannot be found so that /q' = 1; that is, if (a;', y' , z' , u' , v')
is the minimum solution, viz., the net point which lies on a plane such that
there is no other plane containing a net point between this plane and A'o = 1 ,
then any other net point lying on the surface is found by taking the successive
powers, positive or negative, the latter being replacable by the former, of this
minimum Xo . This is in entire analogy to the deduction of all solutions of the
Pellian Equation from a fundamental set.
Finally, in tliis connection consider the limited volume cut out by the four
planes and the surface, and put
\X\ = \b + cp + dp'+gp^;
6, e, (i, jr being real. In all cases
!p''xp = |zp.
Whence, multiplying in succession by p^ , ^ , p^ and bringing the results to the
same form as | X\^;
\X\^ = \{c - h) + {d- h) p-^ {g -h) p^ - hp^^
= l(<^ - c) + (? - c) p - cp5 + (6 - c) p' P
= \ig - d) - dp-\- {h ~ d) p^^- {C - d) p' P
= |6 + cp+^p'+jP»P
Putting
4ai = 1 + VS; 4a2 = - 1 + VS; iPi = ^IQ-2^1>; 4^2 + <10-2^15;
and |ZP< £,
the above give
{b-Ca,+ {d + g)a2V+{ch-id-g)fi2V< e.
{(c-b)-id-b)ai+{g~2b)a2]'+{{d~b)p,-gp.}'< e.
44 ON THE CYCLOTOMIC QUINARY QUINTIC.
{(d-c)- (<7-c)ai+(6-2c)a2P+ { (g - c) Pi + b^^}^ < «.
{i9-d) + dai-{-(b + C-2d)a2}^+[-d^i- ib-c)p2P -47, in absolute value, thus if ei, tj, eg, ei are
each > €, the inequalities may be satisfied by taking 2 sets of equations of
which the first is
cPi- (d- g) Pi = «i,
(d-fc)^i- <7/32= C2,
(5'-c)/3i+&/32= a,
-dpi- {b-c)p2 = a,
the determinant of the system is
(i3!-j8l+^ift)^-0,
so that this set is not independent; in fact
€4 = — £3 ft / /3i — £2 .
It is now required to actually assign values tob, c, d, g which will satisfy the
inequalities. It may be verified such a set of values is
o< rx - ' ■ — r+ , r- , — ~ r; c<
V2(H-a2) 2V2(/Si+ft)' V2(/3i+ft)'
2V2(l + a2) MPi+M' 2^l2(l + a2)'
Numerically these make each { J'' < e/2 in the first, and therefore also in
the others, since these were obtained from the first by multiplying by quan-
tities whose modulus is unity. Put
l2V2(H-a2)i l2V2(/Ji-|-ft)l
where E [ } means the greatest integer contained in ( j ,
l(l+a2)V2J l(^i+ft)V2j
hence b < t' + e"; c< t"; d < t' + «"; g < f' . That is, values have
been assigned to a, b, c, d, g which make | A'p < e, a previously assigned
constant. From this it follows that there is only a finite number of solutions
of F (a) = 1 for which
R = \Io- Ph- P'h- P'lz\'< e.
ON THE CTCLOTOMIC QUINARY QUINTIC. 45
where < t < oo , For the planes
Xo ^^ n^o! -^1 ^^ ki', Xt = ft2j -^3 ^^ «^j
and the surface, bound a volume which for k's finite positive or negative quan-
tities contains, if any, only a finite number of net points; which is a direct
consequence of the foregoing results. Again, it has been shown that
X^---X^>^ Xo^X'o.
and that
Z' TT" V "F^ "I^" — 1
0AjA2A3A4 — 1
when
^0 ■^l X2 X3 A 4 = 1 J
increasing the suffixes by 1 (mod 5) while the indices remain in the same
places,
X[ = X^'X^XrX^,
also
(Xo---X4)~=l,
whence
Xi = Xo"" A'J"-™ X^-^ X^"-^ .
Writing r„ for log 7„ , from this and similar relations,
Ti = — OT3 /o + (wzo — ms) I1+ (mi — viz) h + (n^j — m.}) Iz,
^2 = ('"^3 ~ ^i) h — mj Zi + (mo — w^z) /a + (mi — mi) I3,
I'z = (m2 — mi) lo + {mz — mi) h — mi h + (mo — mi) h,
Ti = (mi — mo) Jo + (mi — mo) Zi + (mj — mo) h ~ m^Iz,
I'o — J^o 7o + mi 7i •{• m^Ii -f ms J3 .
From this, among others is the relation between two solutions: —
Zq 7o /i Jj 7j
z; /i /2 /3 -(/0+/1+/2+/3)
r^ U h -(/0+/1+/2+/3) /o
I'z h -{Io+h+h+Iz)Io h
r. -(7o+/i+/2+/a) 7o 7i 7,
Since Xo, ••• , Xi give a solution of F (a) = 1 , it follows that
7o+ ••• + 7, = 0;
= 0.
46 ON THE CYCLOTOMIC QUIXAEY QUINTIC.
from this, adding all rows of the preceding determinant, if /q + 0,
0,
or (/i + 73)^= ih + hy
or (7,-/3)2=- (7, -J,)2.
Finally, the statement that all the units used in this section are contained in
the formula =fe p"* tj^ is verified. A unit is a o-number whose norm is unity;
hence if = pT rj^ is a unit,
7i
h
73
h
h
7,
h
7o
h
h
7o
h
h
7o
h
h
± p"'p2'"p3'»p'^(,(,,l)n= 1;
pUi ptm pSm pirn ^ ^lOm
"no VI
- 1;
hence p" tj, is a unit. It may be verified as in the following, that all units
belong to tiiis form. Next, it is required to resolve p into factors.
prime. Put
and
hence
so
p = 5n-\- 1,
= {a' -g') + ib'-g')p + etc.,
= a + bp+ cp'' + dp^;
f(p)-fip') = A' + B'vo+C'ni;
f(p')-f(p')^A' + B'm+C'vo;
p= iA + Bvo)(A + Bni) = A' - AB - B';
hence p is to be represented by the form x^ — xy — y^, which (as usual) is
considered as half the form 2z? + 2xy — 2y^; of determinant 5; hence 5 must
be a quadratic residue of p, denoted by 5Rp . If 5 is a non-residue of p, this
is denoted by 5Np. (See for this, and the ensuing reductions Mathews,
" Theory of Numbers," pt. I, Chs. II, III.)
From the usual theory, 5Rp if
p s ± 1 (mod 5),
but 5Np if
p s ± 2 (mod 5).
Again, the only non-equivalent " reduced forms " of determinant 5 are
(2,1, — 2), (1,2, — 1); the first is that appropriate for the representation
ON THE CTCLOTOMIC QUINARY QUINTIC. 47
of
p s =*= 1 ( mod 5 ) .
Hence p has been resolved into conjugate p-factors.
In the next section further properties of F are considered, especially the
transformations.
§ 3. If the properties of the simple quadratic form 3? + i^ , or oi 3? ->= ty^
are discussed, it soon becomes evident that the complete theorj"^ of these cannot
be given unless the more general form ax' + 2bxy + cy^ is considered. In
particular, if linear transformations of either of these forms into itself or into
the other, are sought, the general form arises. As the problem of repre-
sentation of numbers by a given form can always be reduced to that of finding
the automorphics and a problem of equivalence, in other words to the theorj' of
" associated forms," the finding of these latter is a step of fundamental im-
portance. The construction of an " associated form " is arbitrary to a certain
extent; the form here adopted will be justified by showing that this kind of
form which arises first for quadratics, is also essential in the theory of forms of
n variables of the nth degree. In brief, the associated form is derived from
the given form by multiplying the latter throughout by a sufficient power of
the leading coefficient to make that coefficient an exact power; if n is the degree
of the form, the leading coefficient is to be made an exact nth power. This
having been done, the representation of the new leading coefficient in the form
is considered, this being possible (in the case of binary and quaternary quad-
ratics, as in the present and all similar cases) by elementary methods because
the transformation has reduced the problem to one of forms which reproduce
themselves in form with respect to multiplication. For this reason the treat-
ment of quadratic forms in the fifth section of the Disqmsitiones Arithmetical
appears perfectly natural and simply intelligible, whereas the more elegant
" improvements " of the theory by other writers only obscure the funda-
mental reason for the mere possibility of the theory. For a similar reason
the theory of triplication of classes can exist in binary cubics; and it was possibly
this fact with relation to the binary cubic discriminant that led Eisexstein
to construct such a theory. The more elementary parts only of the trans-
formations-theory are touched upon here. Some formulae are collected for
reference.
(I) F ^ F(,afs,ai.,a2,az,ai) =^ al+ VlAVi'A. + VVlVi(A+ VPlVif^l
'+ VPI Vi °I ~ 55p' Qfo ai Ui as ai — op (aj Vq + pi af Ui
(1)
+ Pi(4 "2 + Pi ^3 «3 + Pi l^l "4 ) + ^P"^ ("0 yl + Pi ai Vl
+ Pi or; M^ + Pi U} til + Pi cci nl)
48 ON THE CTCLOTOMIC QUINARY QUINTIC.
where
«o = «i or4 + 02 az] «i = pz ao 02 + pa^ at; Mj = poci as + P4 "o on;
M3 = pi ao ai + pai 04; U4 = pai a2 + pa ^o as •
(II) 3125(a;''-l)/(a:-l) = F; p = 5ri+l; pip4 = p = P2P3,
Pi = So + «i P + a2 p^ + ors p' + "4 P^ »
P2 = ao + 03 p + ai p'^ + a4 p' + 02 p% (2)
^3 = ao + 02 P + a4 P* + ai p' + as P%
P4 = ao + a4 p + as p^ + a2 p' + ai p^ ,
where ao , • • • , a4 are real integers among which by (2) there are the following
relations
ao ai + ai 02 -|- a2 as + aj a4 + a4 ao — — n,
ao 02 + 02 0:4 4" a4 as -f- ai 0:3 + as ao = ~ n,
(3)
5{al+5l+al+ai+al) = ip+l,
1 + ai + a2 + as + a4 + ao = ,
(III) ffi = '^ppl vz; T2 = Vppfpl; irs = ^ppIpi; 1^4= ^pplpy,
whence the following table that will be frequently used:
TTl ITJ ITS Vi
Tl pi TTi pz ITS pi W4 P
T2 Pi 7r4 p P4 Tt\ (4)
X3 Ps Tl Pj X2
X4 P4 T3
Whence, writing a, 6, c, , z, 0, u); at = {y, z, u, v, 0)
The formulae for transformation of F now follow; in any case where the work
is not given in full, the result is evident from the formula themselves.
(VL) F will be transformed by
x=kX+moY+noZ+roU+ioV
y = hX+ ■■■
z=hX-\ I- (9)
u=hX-]
v = UX+ ■•■
50 ON THE CYCLOTOMIC QriNARY QUINTIC.
the transformed ^.- will be denoted by ?',; the determinant of the transformation
(9) is
S; (9a)
and
ki = liX+ vu Y+niZ+riU+ U V, (10)
whence
fo = h+Ti (A-i + pU-2+ p'A-3+ p2A-4)+ TTj ih+ P'h+ Ph+p'h)
+ T3 ih + p" h +pU-3 + ph)i- Ti ih + phi + p-h + pH-i)- • •
= X { Zo + ( n + TT, + 7r4) /i + (:ri p^ + TTi p' + 7r3 p2 + ;r4 p) ?2 (11)
+ ( Tl P' + T2 p -|- TTJ p'' + 7r4 p^ ) /j + ( ITl p^ + TTj P^ + ITj p
+ T4 p') Z4 } + y (ml + 2 { n } + r { r } + F (0.
where { ?« } is what the coefficient of X becomes with m written in place of I;
etc. Whence, writing
^o = LoX+MoY+N,Z+RoU+ ToV; ^,= L,X+ ■•■;
i\ = LiX+ ■■■ + T,V, (12)
Lo=lh+^iUiOhhh) + ^2(hhOhk) + ^zihhkOh) + ^,{hhhkO)]
(VII) "\Mien Lo, • • • , X4 are given, Mo, • • • ; • • • , Ti are found by writing
m , • • • , < in place of I . Writing for brevity
Xo = Zo; Xi = (h Oh h k); X2 s (/i /3 o?2 /.4);
\z = {hUkOh); \i=(hhhhO),
then
io = ( Xo I- iTi Xi + :r2 X2 + ^3X3+ 5r4 X4 , , , , ) ;
2/1= ( Xo , Xl Xl , X2 X2 , TTs X3 , X4 X4 )
^2= (Xo, ^3X3, TTlXl, ^4X4, 3'2X2) (13)
L3 = ( Xo , ^2X2, Tti X4 , TTl Xl , X3 X3 )
Li = ( Xo , X4 Xl , ITS X3 , X2 X2 , Xl Xl )
with which compare (7) ; whence i^ ( Xo , Xi , X2 , X3 , X4 ) = Lq Li L2 Lz Li , (14)
or simply F (X); then in the same way are found F (fi) , F(v), F(a-),
F (t), corresponding to the M, N, R, T; and F (X) , ■■■, F(t) are real
integers. (15)
If this last is not obvious, the equivalent of it is proved later. Again,
lFi\)y-F'= {Fi\)^o/Lo] • lFi\)^yL,\ ' {F{\)^JL2}
• {Fi\)^'JLz}.{Fi\)^JLi],
ON THE CYCLOTOMIC QUINARY QUINTIC. 51
where F' is the transformed F; but the left hand side is also the product of
(XoX+ •••+ ToV) -UUUW, {UX+ ••• + TiV)UULiLu -••;
{LiX-\- ••• -\- TiV) • LaLiL^Li. Hence the product of these five is
divisible by { F (X) j^; therefore the total coefBcient of the term
X'Y^Z-^U'V, «+••; + € = 5,
is exactly divisible by { F(X) }*. In the same way there are congruence
relations for F (>:),-••, F (t) . It will be sufficient to obtain the relations
for the first product only; the others are symmetrical with them (by \Ii),
and may be written down therefrom. In order to see in detail the relations
between the structure of F and the associated F' , the following considerations
on 5-numbers are needed.
Above, (in IV 6) the coefficients of the two 5-numbers multiplied together
were all single letters, with no distinction as to real or imaginary; the form of
the result shows that they were implicitly supposed real. Such a 5-number
is called " simple," those 5 numbers in which the coefficients are simple 5-
numbers are called " complex." A complex 5-number is again a 5-number,
and a single such, but not conversely. The most general form of a complex is
M= { (aoGiaaasQi), {bobibzbzbi) , (C0C1C2C3C4), (dodidtdzdi) ,
(9091929^9*)] ^ [ao+bi+ cs + do + gi, ai+ bo+ Ci + d3 + 92,
(16)
Ci + bi+ Co-\- di + gs, a3 + b2+ Ci+ da + g^,,
cn+bi + C2 + di + ga].
Whence M is the sum of simple numbers in 2^ — 1 ways; of simple numbers
none of whose coefficients is zero in 3,125 (=5^) ways. The theory of the
linear transformation of F forms, and more generally of the forms considered
in § 2, is identical with the theory of multiplication of complex 5-numbers,
or complex n.-numbers in the general case. Similarly for the associated forms.
The immediate object is to find in simple form the products Lo Li L2 Lt , •• • ,
etc.; for this X,- Xy must be calculated. \^Tien X,- Xy is known, X2» X2j follows
from it by doubling and reducing; also if
/(p) = (abcdg),
then/ (p*) is found from/ (p) by means of the substitution (bdgc) , as may
be verified at once. From the values of the X's given in VII, by applying (6),
find
f{p)^\i\2 = {n+hh+hh, hh+hh+n, P2+hh+hh,
hh +l2h+ h h +hh, h U +l2h+ ID
g{p)^\l=ill+2hh, q + 2hh, 2hh + 2hh, ll + 2hh,ll + 2hl2).
52 ON THE CYCLOTOMIC QUINAKT QUINTIC.
A(p) = XiX«= (/J + /i+/|+/J, hh + hh + hh, hh + hli + hh.
^ , , . hh+hh+hh, kh+hh + hh)-
Or for brevity
Hp) = {yiyiViZiyz); 9{p)= {xiXiZiXiXi); A (p) = (wziZaZszO (17)
Performing the indicated substitution,
/(p) = XiX2= (yiytyiZiys) ; g{p)='Kl= (xiXiZiXoXi); hip) = \iXi=(wziZ3Z3Zi);
f(.P^) = X2X4= (yiZiyiysyi) ; ?(p^) = X^= (xiXiX^XtZi) ; /[(p2) = XjXj= (wZiZiZiZs);
(IS)
Sip*) = X4X3= iyiyzZiyiyi) ; g{pi) = Xj= (a'ia-4a:2ZiX3);
/(p') = X3Xi= (yiyii/sy^i) ; gip^) = X§= (a;iZiar4X3a;2) ;
where
Pi + 2hh = xi;P, + hh + h k = yii 2{h U-\-klz) = zi
n + 2li h = Xi;ll + h k + Uh = y2i h k + hU + h h+ U k = Z2
Jl+2hh = Xz;ll + hU + kh^y3',hh + hh + hli=Z3 ^ '
P^+2hk=x^;ll-\-hh + h h=yi; h h -\-hk + h U =[z,
II + II + n +ll=^w
among these are the linear relations,
a:i + a;2 + 3:3 + a;4 + zi = w + 2z3 + 2z4 = yi + t/2 + Jfa + 2/1 + zj
and the quadratic relations
'^1 y\ + 2yi Zi + 22/3 2/4 = a:f + (2:3 + ^4) (2-2 + \i)
yl + 2^4 Z4 + 2yi 2/3 = a-| + {xi + 2:3) (2:4 + ii)
t/§ + 2yi 24 + 2^2 yi= xl+ {Xi + 2:4) (2;i + :i)
y\ + 2y3 24 + 2yi 7/2 = a-* + (a^i + Xi) (2:3 + :i)
which come from (18) by
lfip)V = 9ip)'g{p');
and then doubling and reducing with respect to the 2:'s and t/'s which is per-
missible by (19). The L products which will be required and which are now
being constructed are:
LiLiLsLi', Lo Li L3 Li', LoLiLzLi; LoLiLiLi; L1L1L2L3 (21).
By (13).
LlLt = ( Xo , Xl TTl , X2 X2 > X3 TTS , X4 ^4 ) " ( Xe , X4 Wl , Xj Tj , X2 TTi , Xl Tl ),
ON THE CTCLOTOMIC QUIKABT QUINTIC. 53
which by (6) is
m+ >f tJ+ XItt^H- ^f ir|+ X|:rJ, Xq X4 tt* + Xo Xi TTi + Xj Xi TTl Tj
+ Xj X3 TTi TTa + Xj X3 TTi TTz , Xq X3 TTs + Xl X4 TTl Ti + X2 Xq TTi + X3 Xj ITS Tl
(22)
+ X2 X4 ^2 Jr4 , Xo X3 TTs + X2 X4 7r2 TTi + Xl X3 Tl TTs + Xfl X2 7r2+ X4 Xl TTt TTl ,
Xo Xl TTl "1- Xl X2 Xl 5r2 -}" X2 X3 T2 ^3 "T X3 X4 X3 7r4 + X4 Xq X4 )
which by (4) is, upon rearrangement of terms with respect to resulting r's:
(Xq, X2X3P, X1X4P, X1X4P, X2X323) + Ti (Xjps, XflXi, X2 X4P4, X2X4P4, XflXi)
+ T2 ( Xf pi , X3 X4 Pa ) Xo X« , Xo X2 , X3 X4 p3 )
(22a)
+ jr3 ( X| , Xl X2 ^12 , Xo X3 , Xo X3 , Xl X2 P2 )
H-X4(X2jl2. X0X4, XiXsPi, XiXsPi, X0X4)
Putting as usual
p+ p^^ 10; p'+ p^ ~ m,
the last is
(X2+ I30PX2X3+ J?lpXiX4) + TTl (X3P3+ '?oXoXi+ 5;iP4 X2 X4)
+ T2(XiPl+ J702'3X3X4+ 771X0X2)+ 5r3(XiP4+ '?02?2XiX2+ JJlXflXs) (226)
+ T4 ( Xf P2 + 5)0 Xo X4 + Vl Pi Xl X3 )
Li Li is obtained from ii Li by doubling and reducing, which is equivalent
to an interchange of rjo , 171 in (22&) whence
X2 is = ( Xg + ?7i pXs X3 + m P^i X4 ) + TTl ( \lpi + ijiXo Xl + 770 P4 X2 X4 )
+ 51-2 (Xf pi + 771 P3 X3 X4 + 770 Xo X») + Jr3 (X4P4 + 7J1P2 Xl X2
(22c)
+ 770 Xo X3) + 7r4 ( Xj P2 + 771 Xo X* + 770 pi Xl X3)
Rearranging (226) with respect to 770, 771, the coefficients of the it's in the
coefficient of 771 are found from those of 770 by doubling and reducing,
( Xj + TTl X| p3 + T2 Xf Pl + X3 X4 P4 + 7r4 X' ^2 )
+ 770 (pXj X3 + TTl Xo Xl + TTi PZ X3 X4 + TTs ^2 Xl X2 + ir4 Xfl X4 ) {22d)
+ VI (p>n X< + ^i Pi ^2 X4 + X2 Xo Xj + 773 Xo Xs + TTi Pi Xl X3 ) ;
similarly for (22c), 770, 771 interchanged,
whence Li Li Lz Lt is of the form
(A + £770 + C771) {A + Bm + Cno) = A^ - B^ - C - AB + ZBC - CA;
where (22(f) is written A -{- r]a B -\- iji C (23). The remaining 4 products
in (21) may be similarly found. They could also be obtained by division from
(14), but it is less labor to build them up by multiplication as above; or the
(24)
54 ON THE CYCLOTOMIC QUINARY QUINTIC.
product of two numbers as (16) may be used directly. It is easiest to form
first the symmetric function of the L's, which is done. The results indicated
in VIII are to be finally exhibited in the form ^q X -\- ■ • • , where the coef-
ficients fo ■ • ■ are numbers of the same form as fo • ■ • ; hence at each step in
the reductions, functions of the it's are to be replaced by their equivalents
from (4) as soon as they arise, and final results ordered according to tt's linear.
From (13) and (4)
Lo XI = pXi + TTl Xo + X2 pi Xl + T3 P2 Xa + TTi Pl Xj ,
Lo TTi = pXs + Xl Pl X4 + X2 Xo + X3 P2 Xl + X4 P2 Xl ,
Lo TTz = pX2 + TTl ps X3 + X2 Pl X4 + X3 Xfl + X4 Pl Xl ,
Lo ITi = pXl + TTl Pl X2 + X2 P3 Xs + X3 Pl Xl + Xi Xo ,
But
Lo Li = ( io Xo , Xl Xl Lo > X2 X2 Lo , X3 X3 Lo , xi Xi Z-o) »
Lo being simple, and so for Xo Lo , Lo L3, and io Li . Putting for the moment
A = Xo ^ + ( X2 X2 + X3 X3) -B + ( Xl Xl + X4 Xl) C,
B == X4X4^+ (xiXi+ X2 X2)5 4- (Xo+ X3X3) C,
C = X3 X3 ^ + ( Xo + Xl Xj) B + ( X2 X2 + X4 Xl) C, (25)
D = X2 X2 ^ + ( Xo + Xl Xl ) jB + ( Xl Xl + X3 X3 ) C,
E = Xl Xi/1 + (X3X3+ XI X4)i?+ (Xo+ X2 X2) C;
th
en
LiLiLz= (ABODE),
i3iiXi= (ADBEC),
L2LiLi= (ACEBD),
11X2X3= (AEDCB);
(26) follows from (13), (22d) by (6); one result being found the others follow
as in (18), or directly. From (26)
X0X2X3X1 = (AXo, BXo, CXo, DXo, EXo),
and so for the others; but first the values of A, • • • , E are written down from
(27), (which could be omitted, except for convenience); the columns are the
coefficients of x.- in the functions at the left. As a check, the weight of the
X's in the column x.- is constant and equal to i modulo 5. (The table is on
next page. ) Directly from (27), ^^Titing
ON THE CYCLOTOMIC QUINARY QUINTIC. 55
Ao = \l + p (pi Xf X3 + P2 Xi X| + Pi Xi X4 + Pi X2 \l)
Ai = p\i (Xi X4 + X2 X3) + P3 Xo Xf-f 2p4 Xo X2 X4
A2 = p\2 ( Xi X4 + X2 X3 ) + pi Xo Xj + 2p3 Xo X3 X4
A3 = pXs ( Xi X4 + X2 X3 ) + 2p2 Xo Xi X2 + Pi Xo Xf
A4 = p\i ( Xi X4 + X2 X3) + 2pi Xo Xi X3 + P2 Xo Xi
Bo = 2X0 Xi X4 + Xo X2 X3 + Pi Xi Xi + pz X| X4
B2 = pXi X3 + Xg X2 + 2^1 Xo Xf + pi Pi Xi X| + Pi Pi X|
B3 = 2p\i X3 X4 + Xg X3 + pi Xo Xi X2 + P2 Pi Xi X4 (27a)
B4 = p\i ( Xi X4 + X2 X3) + 2pi Xo Xi X3 + 232 Xo Xi
Co = 2X0 X2 X3 + Xo Xi X4 + pi Xf X3 + P4-X2 Xf
Ci = 223X1 X2 X3 + Xj Xi + 233 Pi X3 Xf + 234 Xo X2 X4
C4 = 23Xi Xf -I- X? X4 + 231 Vi Xf X2 + pi Pi Xf + 232 Xo Xi
Di = pXf X4 + Xi Xi + 232 Pi Xi + 233 Xo Xf + Pi Pi X3 Xf
D2 = 223X1 X2 X4 + Xi X2 + 23i Pi Xi Xf + Pi Xo X3 X4
E3 = 2^X2 Xf + Xi X3 + Pi P2 xf + p2 Pi xi X4 + P4 Xo xf
(27)
1 Vl W2 TTs TTi
Xo A Xg p3 Xo Xf pi Xo Xf p4 Xo Xf p2 Xo Xf
Xi Ti A PP2 Xi xi Xi Xi Pi p3 Xi xf Pi p2 xf p Xi xf i
X2 TTi A ppi X2 xf P2 P4 xi xi X2 p X2 X3 Pi P2 xf X2 2
X3 iTi A ppi xf X3 P3 P4 X3 xf p X| X3 xi X3 Pi ps xi 3
X4 vi A ppi xf X4 p xf X4 P3 P4 xf p2 P4 xf X4 xi X4 4
Xo -B p Xo X2 X3 xi Xi p3 Xo X3 X4 P2 Xo Xi X2 xi X4
Xi ;ri J5 p Xo Xi X4 pXi X2 X3 pi Xo Xf p Xi X3 X4 pi P2 Xf X2 1
X2 :r2 B ppi Xi Xi p4 Xo X2 X4 p Xf X3 p2 Xo Xi X2 p X2 X3 X4 2
X3 Vi B ppi Xf X4 p Xi X2 X3 p3 Xo X3 X4 p X2 Xf pi Xo Xi Xj 3
\iiriB p Xo Xi X4 P3 P4 X3 Xf p Xi X2 X4 p4 Xo Xf p X2 X3 X4 4
Xo C p Xo Xi X4 P4 Xo X2 X4 Xo X2 Xq X3 pi Xo Xi X3
Xi «•! C ppi Xf X3 2) Xf X4 p Xi X2 X4 P2 Xo Xi X2 pi Xo Xi X3 1
X2 7r2 C p Xo X2 X3 p Xi X2 X3 p Xi X2 X4 P2 P4 Xf X4 P2 Xo Xf 2
X3 T3 C p Xo X2 X3 Ps Xo Xf pi Pa Xi Xf p Xi X3 X4 p X2 X3 X4 3
X4 Tfi ppi X2 Xf p4 Xo X2 X4 P3 Xo X3 X4 p Xi X3 X4 p Xi Xf 4
56 ON THE CTCLOTOMIC QUINARY QUINTIC.
From (27o), (26), (25),
A = Ao + iri Ai + ir2 A2 + Ts A3 + Ti Ai,
B = pBo + Xl Al + TTJ B2 + T3 B3 + Tt B4 ,
C = pCo + Xl Cl + T2 A2 + T3 B3 + X4 Cl ,
D = pCo + XI Di + X2 D2 + TTj A3 4- X< B4 ,
E = pBo + Xl Cl + X2 D2 + X3 E3 + X4 A4 .
(276)
These are not the same as (25), for Ao , • ■ • , E3 are independent of x's . Notice
also that these functions are isobaric, modulo 5; e, g., Ai is thus of weight 4
in the X's. Finally, using (24) and (276), taking the product of the two
following matrices:
Ao Ai A2 A3 A-i
pBo Ai B2 B3 B4
pCo Cl A2 B3 C4
pCo Di D2 A3 B4
pBo Cl D2 E3 A4
Xo, Xi, X2, X3, X4
pXt , Xo , pi ^1 > Pi ^2 ) Pi ^3
pXj , p4 X4 , Xo , Pi Xi , pz X2
pX2, P3X3, ^3X4, Xo, pi Xi
pXi , p4 X2 , Pi X3 ,^4X4, Xo .
(multiplication by rows of first, columns of second), find
Afl Ai Aj A3 A^
Bq Bi Bj Bj B4
Co Cj Cj C3 C4
d; d; Di D3 d;
. E» E'l E|t Ej E4
(28)
whence
ZoA = AJ, + A'lXi + A2X2 + A3X3 + A'jxi; •
whence, from (26),
Lo Li L3 Li = (AJ, + • ■ • + A.\ n;
LoE = E'o+ ••• + E;x4 (28a)
..; Ei+---+E;x4)
(286)
ioiii2i3= (Ao+--- + a;x4; •••; b;+--- + b;x«)
Consider now
^oLiLiL^Li-, Ld'iLiLzLi-, LoLi^^L^L^; LoLxL,aU; L^LxLiL^^,. (29)
and (see 11) in particular the coefficients of A', • • • , T^ in the last four. In
ON THE CYCLOTOMIC QUINARY QUINTIC. 57
all (by (14)), the coefficient of Z is f ( X ) • The coefficients of Y in the last 4
are respectively Lo Mi L- L3 L^; Lo Li M2 L3 Lu La ii 1-2 M3 W,
La In. Li Li M i] each of which is a complex 5-number, e. g., LoMiL^LzLt
being the product of (moj/^i ti, 112 t^, us tz, fn iri) and the first- written
number in (286). This number will now be put in the form (12), as required,
the reductions being effected directly from (286) and (4). It is convenient
in working to use the principle of constant weight (mod 5) already noted. The
form of the result before actual reduction to form (12) is (A" B" C" D" E");
but in this it is not necessary to calculate all the letters, B" , • • • , E", in virtue
of (286) follow from A" by permutations, represented by the substitutions
[ABCDE], [ACEBD], [ADBEC], [AEDCB].
A" = {fxo A'o -f pfii E'i + PH2 Dg + p;x3 C2 + pm B\ )
+ Ti («) A'l + ;ti Eo + Pi 112 D4 + P3 m C3 + Pi Hi B2 )
+ X2 (po A'a + pi Hi E'l + /i2 Dq + p3 Hi Ci + p3 Hi Bj )
+ X3 (m A3 + p2 Hi E2 + Pi Hi D'l + H3C'o + Pi Hi K )
+ ri (ao A4 + pi Hi E3 + Pi Hi I>2 + Pi Ms C'l + Hi'B'a) .
Finally, adding all so found results (without writing them down),
LoMiLiLsLi = {//o(AoB;c;d;e;) + PHiiKKKc'iD'i) + pm2(D3e;a;b;c3)
+ pH3iC2ViElk'2B'2)+ pHi{B\C[B\E\k\} I
+ iri[Ho{ )i+ fii(.
+P3H3{.
+ 'r2{^o( h+PiHii
+ P3H3i
+ T3{M )3+P2Hii
+ H3{
+ '^i{l^o{ )i+PiHl(
+PllJ-i{
where in any column the ( )„ means the same arrangement of A, • • • , E
as at the head, with the indicated change of suffix. From (286), referring
to (26), the value of Lo Li M2 U3 Li is written down from (30) by means of
the substitutions [BCDE] and (see 13) [ hi i^z y-i Hi] , and in the same way
LaLiLiMzLi by [BCED] and [wm: A'l Ms]; io^ ^2 is Mi by [BE] • [CD]
)o+P4W(
)i
)3+P4M4(
h\
)l+ M2(
)o
)i+P3Hi{
)s| (30)
)l-\-p2Hi{
)i
)a+PiHi{
)«}
)3+P2Hi{
)2
)l+ M*(
)ol
58 ON THE CTCLOTOMIC QUINARY QUIXTIC.
and [fiiin] • [fiifi]- It now remains to evaluate J/o 11X2X3X4 in the
same way, i. e., arranged according to it's. The result will be derived in a
slightly different form to those shown above, but finally all will be reduced
to a common form, i. e., corresponding to ^0 , 6 , h, h, ?4 , (see (7a)), where
the letters corresponding to a', ••• , g' all denote real integers.
By (23),
Xi LiL3Li= A'- B^'-C^- AB + SBC - CA ,
where
^ = Xg + iri Xf j?3 + T2 \l Pl + TTS Xj Pi + TTi Xj Pt]
B = p\i X3 + Tl Xo Xl + TTi Pi X3 X4 + TTs p-l Xl X2 + Ti Xo X4;
C = 2)Xl X4 + 2-1 Pi X2 X4 + TTJ Xo X2 + TTZ Xo X3 + Vi Pl Xl X3;
or say
^ s ao+ ai+ 02 + 03+ 04; B=ho+ 1- 64; C = Co + V U;
then after reductions, collecting together terms of same weight (mod 5),
(the self-explanatory table is appended for verifications)
XiX2X3X4= {(00+20104+20203)
-(62+25164+26263)
-(c«+2ciC4+2f2C3)
+ 3(6oCo+ 61C4+ 62C3+ 63C2+ 64C1)
— (0060+0164+0263+0362+0461)
— (CoOo+Cl04+C2a3+C302+C40i) }
+ 4 similar { ) derived from this by successive augmentations of all suffixes
by 1; and by the principle of weight used throughout these 5 in the order
indicated furnish all terms which on reduction contain respectively (tto = 1) ,
"•2, T4, Xl, Ts; call, the result o' + 6' iri + c' Tr2 + P^2 ) f'lMs) 1'i(mi , fiO , Pif^i y P3t^3 , Pil^:)
+ c'(pA'3,P4M4,M0,P2Ml,P2A'2) ■'^2(1^2 > Plfil , 1^0 , Plt^i , Pi Ml) (31)
-{-d'(j)fi2 , PzfiS ,Pif^i,I^O, PlMl) ^3(^3 , PiM2 , P2,"l , Mo , Pitii)
+9'(Plil . PiJ^i > P3P3 , Pi^ii , Mo) T4(P4 , Pll^i , Pil^i , PlMl ) Mo ) ,
Oo bo
0061
Qo 62
Oo 63
Oo 64
1 4
1
1 1
1 2
1 3
2 3
2 4
2
2 1
2 2
3 2
3 3
3 4
3
3 1
4 1
4 2
4 3
4 4
4
0=
2.01
2.02
2.03
2.04
2.14
P
2.12
2.13
2.23
2.24
22
32
2.34
42
ON THE CYCLOTOMIC QUINARY QUINTIC. 59
where on left v multiplies each number in its column, similarly for a' , ••• ,g'
on right. (Compare (30).) The nature of these coefficients will now be
examined, and it will be shown jBrst that the term free from tt's is a real integer.
That it is an integer is obvious, for only multiplications have been performed
in its derivation; the reality follows with some work. On reducing by (4),
a' = XJ - p2 (Xf X| + Xi X| + 4Xi X2 X3 X4 ) - 2p\l ( Xi X4 + X2 X3 )
+ 3p (pXi X2 X3 X4 + pi Xo X? X3 + P3 Xo Xf X4 + P2 Xo Xi Xi + Pi Xo Xi Xi )
— P ( Xg X2 X3 + ps Xo Xi X4 + Pi p2 X? X2 + pz Pi X3 X| + P2 Xo Xi xi )
— P (Xg Xi X4 + P2P4 X| X4 + p4 Xo X2 X| + jh Xo X? X3 + Pi Pi Xi xf )
or
a' = XJ - p^Xf ^l + Xi X2 X3 X4 + Xi X|) - 3pXi (Xi X4 + X2 X3)
+ 2p {pi Xf X3 + Pi x§ X4 + Pi Xi Xi + Pi xi X4). "X^
— P (Pi Vi xf X2 + Pz Pi X3 Xl + Pi Pi \l X4 + Pi pt Xi xf ) .
This could be shown to be real by actually expanding, but a reference to the
values of the p's and X's shows that the labor involved is prohibitive. How-
ever, referring to the values of h{p) , h (p^) given in (IS), it is evident at
once that the coefficients of — ^ and ~ 3p\l in a' are real. The entire coef-
ficient of -j- 2p may be generated by doubling and reduction from any one of
its terms; e. g., from pi Xf X3 are successively derived ji2XiXi, Pi'S.lXi,
Pi Xf X4 . But pi Xf X3 is a function of p , say (p'')isa real integer. Similarly the entire coefficient
of — p in a' is a real integer. Hence finally, a' is a real integer. In the same
way is found
c' = ipi Vl X| + 2pi Xi Xf + 2ppi Xi X4 ) - (pi X5 Xi + 2ppi X2 Xi X^
4- 2^X0X1X2X4) - (pXi X| + 2p Xo Xi X2 X4 + 2^3 Pi Xo Xi Xi )
4- 3 (pXo Xi X2 X4 + ppi Xi X3 Xf + PPi Xf X2 X3 + Pz Xi X3 X4 + pXo xi X3 ) (32)
— {piPi Xo XiXf -f- ppi Xf X2 X3 + Pi Pi Xo X^ + PP2 Xi Xi + Pi xi Xs X4 )
— (PP3 X2 xi X4 + xg X2 + pXo xi X3 + PPi Xi X3 xf -^ ppi xf X4) ,
that is, the coefficient of X2 is of weight 2 (mod 5), (as it should be), g' is
found from c' by doubling and reducing all suffixes (including those of p),
d' similarly from g' , and finally b' from d' in the same waj', as is evident from
(4) and the manner of writing the a, b, c expressions, or it may be verified
60 ON THE CYCLOTOMIC QUINARY QUINTIC.
directly. Hence if c' be written as a function of p, it follows as above that
9' = fip'); d'=f{pn: b'=f(p'); c'=/(p),
and therefore
b'ui,-hc'ni + d'^2 + 9'Mi=f{p')-9(p*)+f(p)-g{p'')
+f(p')-9(.p')+f(p')-g(p),
where / ( ) , g {) are rational integral functions, for by (VII) the /I's have
values of the nature indicated. If
ff (p)/ (P^) = Go+Gip+G2p' + 6'3 p' +G,p\
a form to which the product of any two rational integral functions of p is
reducible, then clearly
b' m + c' ^, + d' ^2 + g' HI ^ iGo - Gi- G2 - Gz- Gi,
a real integer. Therefore, summing up, the absolute term is a real integer;
i. e., a' lie -\^ p {b' Hi + c' tiz -\- d' H2 + g' /ixi ) is a real integer. It will next be
shown that the coefBcients of the it's are imaginary, of the form (a", b", c", d",
g") . Writing for the moment
M.-=?(p'); Pi=h{p'); a'.---,g'
as above, this being permissible by (VII) and (2), the coefficients of ri , • • • 7 ^4
respectively become:
i^i) : a' 9 (p) + f (p') g il) + f ip) h {p') g {p') + f ip*) h (p') g (p')
+ f(p')h{p')g(p'),
(ttO : a' 9 ip') + h (p)f (p^) g {p) + f {p) g (1) + f (p*) h ip') g (p*)
(33) +fipnhip')gip').
(n) : a' 9 {p'^) + h {p-')/ (p') g (P') + f (P) k(p') 9 (p) +f {p') g (I)
+ f(p')h{p')g(p'),
(n) : a'gip*) + A (p)/(p') g (p') +/(p) h {p')gip')
^-/(p^)A(p)?(p)+/(p^)?(l)
which are of the form
vip) + s(p') + cc(p) + Hp') + y(p')
r}(p') + P(p') + Hp) + yip') + ^ip')
(33o)
1? (p') + « (p') + 7 (p) + Up') + /3 (p')
V (p') + y (p") + Pip) + « (p') + 5 (P-),
(336)
ON THE CTCLOTOMIC QUINARY QUINTIC. 61
)7 , • • • , 7 rational integral functions. In any line it is clear that the coefficients
of the various powers, of, p sje, ,not iAen^ca4 for general values of the letters
involved, and thereHir-i iiv gejie^aj .^ach coefficient of t's is a complex 5-number.
Writing a(p) in the fcrm a.,-\- ••• +«4P*, 7(p) = etc.; ••• the final
form of the coefficient ic ibeordjer written in (33a) is
( a" h" c" d" g"), ( a" d" h" g" c"), ( a" c" g" b" d" ) ,
{a"g"d"c"b"),
which by eliminating g" as usual become
( a'" b'" c'" d'" ) , ( a'" d'" b'" c'" ), (a'" c'" b'" d'" ) ,
{,a"'Od"'c"'b"')
all the letters being real integers.
All the foregoing considerations lead up to the construction of the associated
forms, which latter are now examined. The factor" composing an associate
are finally to be of the forms
(34)
[0] = {UX-\-Mo Y->rN^Z+R, U+ ToV). UUUU = ^'oL\ULzU,
[2] = {Li X -\- • ■ • ) . Lo Li L3 Lt = Lo Li ^2 J^3 Li ,
[ 3 ] = ( is X + • • • ) -LoLiLi Li ^ Lq Li L2 ^'3 Li ,
l4:] = (LiX+ ••• ).LoLiL2L3 ^LoLyL^L^^^.
It has been shown that the coefficient of K in [0] is expressible in the form
indicated by (336); but if 6" had been eliminated by means of
6"(1 + P+--- + P^)^0,
instead of g" , the form would have been,
a'+ ;ri (r.orjj^) + T2 {rj.orj^) + t, (rjj,or,) + ^4 (fjjj.o),
with which compare (12). In this it has been shown that all the letters Ti , • • • ,
r^ denote real integers, and it is clear that the t' are functions of the a ,b,c,d,g
in (5), and of the I's and m's in (9). Since a, • • • , g occur similarly in all the
coefficients, there Is no need to put them in evidence; the coefficient of Y being
denoted hy tpo (,1 , m) , then those oi Z ,U .V zre ifa{l,m) , a{l,T) ,ipo{l,t) ,
and the absolute term may be put = ipo {I, I) . Similarly in [1], • • • , [4],
so (34) becomes
[i]^ 'r2a2, ir4ai, TTiofi, tts as) 7 -f- • • •
[4] = kX + {ao, Viai, Tjas. T2 «: > tiqii) F -{- •••
(The as are distinct from those in (1).) [0], • • • , [4] are of the form (7).
ON THE CYCLOTOMIC QUINARY QUINTIC. 63
The last four are obtained from the first by use of (36). Among the quantities
in (38) many relations are now apparent, of which a few are written down
before continuing with the associates. The set (38) being further shortened
to (38a),
hX + Ai Y +BiZ+ Ci V-\-DiV^ [i], (t = 0, .- . 4)
then
AoAiAi AzAi = F (ao, CKi, 02) «3, cci); Bq •• • Bi= F (/So, • • • , Pi);
Do--- Di = Fi5o, --, h); Co--- Ci = F{yo, ■-, 74).
Again, from the form of (1) and from (V), (38), (3Sa), it follows, first that
if •f(?Oi fi, fs, fs, fi) is a rational integral function of a;, ••• , v, say
G {z, y, z, u, v); then also must
FiAo,---,Ai) = G(ao,---,ai),---; F(Do, --- , D^) = G{So, ---,5^).
Further if there is any relation of the form
Fi^) = Gix, --■),
then there are at the same time
FiA.---) = G{a);----.F(D) = GW.
Comparing (11), (12) and (37), and remembering the value of fc, = Oii)
ao Tpoci poci paz pai
«1 ao P4Q;4 ^3^3 PiCC2
at piai ao pzoci P3013
az pocci p^ai ao Pia.\
I CK4 p\oiz p^ai piai ao i
where the notations mean!^ that the coefficients in the matrices go by rows
with the m's and a's respectively, and (the signs all plus), corresponding rows
are to be equated; in this way the above stands for 5 equations for the m'i.
In exactly the same way there are 5 equations for each of the n,r ,i\ the system
for the Ti's is deduced from the above by writing n in place of m on the left,
and j3 in place of a on the right; similarly the t , i sets are written down by
replacing in the above vi and a by r and 7, by i and 5 respectively. Thus,
(with (50)) there are 25 equations to determine the constants of transformation
(9), there being 5 sets of 5 equations each, and the determinant of each of the
5 systems is
Hi h\ C\ di
02 62 C-2 di
az bz Cz dz
= k
= 5/,V(l-P
k a\ jSi 7i Si
ai P2 72 ^2
OC3 Pz 73 ^3
04 Pi 74 ^i S") = I. and S-S-'S" = S";
also if
SS"'=-S". S"' = S-^S",
and if S'" is an S,
(S^iS") = + l,
• Papers, I; especially
ON THE CTCLOTOIIIC QTHNAKT QUINTIC.
71
which requires that a determinant of identical form to D shall = + 1 , hence
the S'" is not distinct. Thus all the required transformations are found
uniquely. If S is any one of the transformations, and for Z) = 1 , l\ , /^ , /j , f^
any integers, SS' gives all transformations, where S' = S" S'" , and
S" =
1
m'l
?«2
^3
m\
"'i
«2
"3
n\
•■i
^2
J-s
r'i
A
U
^3
h
S"' =
\l h h \ h\
I
10
10
1
Put
\a^hxCiiz\ = k,
the cofactor of 1 in S" is 2) = 1 . Then <» solutions of this for a^, bo, Co, do
may be found; hence also further solutions from the multiplication of J^ (as
a determinant) and D , say Z> • i ao > • • • j <^3 1 ; multiphing by rows of first and
columns of second. The greatest divisor of ao,bo, Co, do is a factor of Jc, and
by the general theory ( S" ) may be reduced to the form in which the principal
diagonal of the (4th order) determinant is composed of factors of k , the numbers
below it of zeros, and those above it of positive or zero elements less than the
corresponding /c-factor in the diagonal, — in fact by the usual theory of the re-
duction of a matrix to canonical form. Now if any set aQ,bo, Co do be chosen
as above, which make
I Co &i Cidil = k,
then 00 sets are found on multiplying by the « values of D, as above, but
there is in each of the classes so determined always only a finitie number ob-
tained by use of the reduced matrix; hence in each class is a " reduced " form
from which the others in the same class may be derived, being equivalent
thereto.
As the entire idea of reduction and classes is worked as well for n-ary n-ics
arising from Lagrange's kind of equation in fact more clearly, the special
case of this paper is carried no further, but will be recurred to when the general
theory is given. One result may be stated which (if new) is of interest, the
number of classes cannot exceed the number of divisors of k" , where k is an
integer represented in the n-ary 7i-\c.
Finally in this connection it may be readily shown how from one transfor-
mation of a [ 0] • • • [ 4] into a [ 0'] • • • [ 4'] all may be found by means of
the F = 1 equation of § 2; as this generalizes at once to forms of nth order
and degree, it is given.
72
In (9)
ON THE CTCLOTOMIC QUINARY QUINTIC.
= S ^ Z / lo mo • ■ • to
let the result of priming all the letters, thus /J,, •
[4]; F'=[Q']
, f^ be denoted by S',
[4']
similarly for S"; and let
F^ [0]
be such that
S'{F') = F,
viz., let F' be transformed into F by means of S'; let also
S'S ^S",
and further let
SiF) = F
so that S is any (or all) of the automorphics of F , then all the transformations
of F' into F are uniquely determined by { S"\ . As usual, in multiplying
the two matrices S' and S, the multiplication proceeds by rows of S' and
columns of S. Let the corresponding coefficients to k, A, ••• , etc., of F
heinF', k' , A' , etc. Then from (49) it is seen that the S" are determined
uniquely by
k% + A'ji + B-ji + Co /3 + z'o /;- = I'o/o (ro,r„r„r,,r,)
k'm'o + A'oVj'i + Fomg + Com'^ + D'om'i = Xo/o( m'o , ■•■ m\)
(52)
A-Yo+--- =A'o/i(''o, ••• f<)
where the/' are the correspondents in F' of the/ in F , (see 46, 47), and (see 49)
Xo is the first factor of the equation
F {_ao, ••• . a^) = 1,
where note, F has nothing to do with F of the two forms, it is the symbol of
the equation considered /^ = 1, in § 2; there can be no confusion. For it
may be verified, on expanding, or directly from the formula; in (30), (34), (37),
that the necessary and sufficient conditions for S' (F') = F are
kk'^^^foir,, r„ •••, r, )./;(/'o, Ti --o-'-fiCo
Aok'* = fo (wi'o, m'l,
m
\)-fAro,n
5oi''=fo(«o.
Co K == Jo \ ''o I
Dck'*=fAfo,
, n\) -fiir^, Ti ■■■)
•)---fAro-
■fAro---).
■fAro---).
■fATo---),
),
),
(53)
ON THE CYCLOTOMIC QUINARY QUINTIC. 73
and these, combined with (49) give
Vo(^o. ■■■ri) + hfo){m'o, ■•■ m\)+ [-lifoifo, •••O
~ Xofo ( • • • U)
(54)
fofoil'o, ••■ ri) + tifo{m'o, ■■■ m\)+ \-tifo(.fo, ■ • ■ Q
= Xofo (4» ■ * ■ U) •
and by direct substitution, or by inspection, (52) follows from (54). Re-
writing (52) in final form, all transformations of Fi into F are deduced from
any one such transformation by means of the formulae
/o ( '4'. w'l . I'h I'z, ^4 ) = ^ofo ( Wo . w'l . ^2 ,^<'3'^''^)> (55)
where u stands for the letters 1 , m, n, r, t in turn, and as pointed out in the
automorphics, equations such as (55) implicitlj' contain 25 linear equations.
In conclusion of this section, a very brief indication of the forms in the general
theory analogous to those of this section, is given. It is seen that the present
paper treats only a very special case of those described now. The form
[0] • • • [ 4] here considered has arisen in the theory of the division of the circle,
or what is the same thing, in relation to the irreducible in /J [ 1 ] equation
a;"-i + a;"-2+ 1- 1 = 0,
where
n= (p- l)/5,
and pis a (real) prime of the form lOn + 1 . The equation F = 1 of §§ 1, 2,
has been seen to give rise to the automorphics of [0] ••• [4](= G , say ) ,
also to the theory of (Gaussian) equivalence of (r-forms, simllarlj- to the com-
position of such. But in the most general case, where all numbers considered
are to be ordinary algebraic integers, the irreducible equation ^dll be of the
type
x" + aiX"-^+ l-a„= 0,
where the a's are real integers. The most general case, where the a's are al-
gebraic integers, is treated in the same way precisely as this (apparently)
more simple case. Call the equation here corresponding to the i^ = 1 of
§§ 1, 2, the " F equation "; it is constructed as the norm of the general alge-
braic integer of the nth degree (belonging to the above equation). This is
in fact {F) & Lagrange's duplication form, and the resulting forms are of the
nth degree on n variables. However this is not strictly the generalization of
the § 3 forms G , nor is the F derived in this way so general as that of § 1.
74 ON THE CYCLOTOMIC QUINARY QUINTIC.
To find this general F, proceed as follows; let p be any real integer in the field,
whose factors are da, i>i, d^, d^, ■■• , t?^i; so that the O's involve no ir-
rationality not in the field. Then if
Ai= U+aidJ{ai) + aU2f{ao)-{ + a",-' t?„-i/ (a„_i) ,
where
/ (a,) = Xo 4- Xi a, + a-o a^ + • • • + Xn-2 oci"'^ ,
the " i^-equatlon " is
JlAi =1, (i = 0, •••, h - 1),
where the a:'s and u are the n variables; by a slight modification of Dirichlet's
method, it may be shown that this F-equation has oo"-i solutions, as in the
case of § 2, the f -equation had oo-* solutions, and in the same sense. Next,
to find the G-forms. Subject the variables u, x; to a linear transformation
» — / '00 'ol • ■ • ^On-l
, ^n— Ij In— I, 1 ■ • " 'ti— 1, n— 1 i
then if before the transformation Ai was F (u, Xg, xi, •• • , x„-2) , let it after
the transformation become A\ , so that
A', = F (looJl, 0, • •• , J-n-l, o) U' + aiF (loiylii, • •• , In-l, l) X' + ■ • •
+ \- ar' F(lo,n, ■■■, In-l, n-l) ^'n-l ,
and let Z,y denote what Ai becomes when the jth column of 5 is put for the
variables, (each for each), and let
Li ^ Ltjj Llj • • • i/n— 1, J ,
then the general factor [ /] is A\ Lj Ltj (i = 0, • ■ • , n — \); and j = 0,
•• • ,n — 1 gives n such factors as a " base." For j fixed, the product ir[i]
is constructed; and gives the most general G-form. It may be shown by the
methods of this section, that
[ i] = ku' + Ail x'l -f yi,2 a-2 + • • • + A^n-i a-^-i ,
where the A,j are all of the same form, and for thc/values of
i = , • • • , n — 1 ,
the/orm5 of the A,j run through, in some order, the forms of the At . That is,
there is a principle of permanence of form with respect to w-adic number
coefficients, as well as with respect to the variables, in the G forms considered.
The stu<3y of these G-forms seems to constitute the most natural extension
of the properties of binary quadratic forms; for the theory is not specialized
ON THE CYCLOTOMIC QflNARY QOTNTIC. 75
by extension of degree, that is, it exists for these forms in toto. The G forms
are rational integral functions of their variables; the coefficients are numbers in
the n-field; when expanded in full, real also in coefficients of
ar" + ai x"'^ • • • = .
§4. A form being given, it may be interpreted as the embodiment of
arithmetical or of geometrical facts. The two interpretations are not mutually
helpful; thus few, if any new arithmetical truths have been discovered by
geometrical reasoning. But often the statement of an arithmetical theorem
or its proof, in geometrical language tends to conciseness; e. g., Eisenstein's
proof of the Law of Quadratic Reciprocity, Poinsot's proofs of Feeii.vt's
and allied theorems, or Catlet's proof of Wilson's theorem, also the works
of PoiNCAKE and Klein in the theory of forms, and Sylvester's theories of
Residuation and Partitions. But if analogous methods are applied to higher
forms, one essential feature is at once lost, viz., space-intuition; thus in place
of the simple and obvious fact that the diagonal of a rectangle divides the
rectangle into two congruent right triangles, which observation obviates
tedious arithmetical calculations, there is a complicated question regarding
the " super-lines " of symmetry of a regular solid in ?i-space. Sight intuition
is useless here, as it was not in the plane, and all that is accomplished is a
suggestive simplification of language. On the other hand, forms which present
themselves in the study of numbers are frequently those of intrinsic geometric
interest. As an example that has possibly not been explicitly stated before,
the triplication of class property in binary cubics, due to Eisenstein, has its
analogue in the geometrical theorem that the equation of the general develop-
able of the fourth degree in tetrahedral coordinates is the binary cubic dis-
criminant = 0, the two theorems are formally equivalent. This theorem is
not isolated; its generalization to n-space is involved implicitly in Lagrange's
n-ary n-ic forms; the discussion of this is reserved; there is a difficulty in stating
the degree of multiplication. Here it is simply pointed out that the geo-
metrical interpretation of F = k is not likely to suggest any new arithmetic;
however, the properties of so simple and symmetrical a surface, it being in
many respects the 5-space analogue of the sphere, may not be without interest.
The complete discussion of this " surface " is too extensive to be undertaken
here; only those properties which result from a direct reading of its equations
are considered. Although the interpretation is in terms of Ss, all may be
carried out in ordinary space, for the 5 coordinates of the points considered
as belonging to Ss, may by the usual convention be regarded as the "coor-
dinates of a sphere." The points particularly considered are those which form
a self-congruent net work under the automorphic transformations. By reason
of che Imaginary nature of the coordinates in Si, the immediately foIlo\\'ing
76 ON THE CYCLOTOMIC QUINARY QUINTIC.
configuration of 120 points reduces in S3 to points, not spheres. For, the
coordinates of any point of the set are projectively transformed into (p°, p*,
p". p''> p") , a, ■ • • , <7 all different, and the (120) spheres (xi, • • • , ars) in S3
which correspond reduce to points, for
xi + xl + xl + zl+4=0.
But there yet remain 3,005 special points (as seen presently) whose coordinates
do not satisfy this equation, and these correspond to actual spheres, though
possibly of imaginary center or radius in S3; in general one x at least is imag-
inary. The whole surface must be interpreted with reference to these ideal
elements; if F be expanded into real form, the G of § 2, a different configuration
is obtained, containing in its equation over 3,000 terms, and obviously such
a collection of symbols is quite unmanageable. The sphere interpretation
is only pointed out; the natural method of discussion seems to be based on
considerations of Ss . One special surface is of such frequent occurrence that
it is designated by a particular name, viz., " 5-oid "; the rest of the nomen-
clature (where needed) is taken from Cayley's " On the super lines of a quadric
surface in space of five dimensions." * The 5-oid is
ul+ul + vl + vl+ul = 5,
and bears much the same relation to F = k that the director circle bears
to conies, or the director sphere to quadrics. Let
= (p (mo, • •• . W4)
be the equation of a surface in S5 which when referred to homogeneous co-
ordinates Wo, ••■ , lis becomes $ = 0, then the tangent surface at the point
( w'o , • • • , «i ) of ¥> is defined to be
(WO *„.„ + h M5 *„'.) = 0, M6 = 1 ,
where the notation means that u' is to be substituted for u in the partial
derivatives of # ^ith respect to «o, • • • . "5, and finally ws is to be put = 1 .
This definition is in accordance with usage. Let
F^^l^UoUiUiUzUi^l; Ui= wo+P*«i+p""2+p"w3+p''w4,
the general F = k having been reduced as in § 2 to this form by a linear
transformation (equivalent to an affine transformation), then
F^.= {l/Uo+P'lU\+p''/U',+ p''IV3 + p''/U\)F'; (i = 0, ...4). (1)
Hence the equation of the tangent surface to F = 1 at the point (u'o, • • • . «!)
is
'Papers, Vol. IX., p. 79.
ON THE CTCLOTOMIG QUINARY QUINTIC. 77
uo{ilu;-{'iiu[+i/u; + i!u; + iiu;,o, 0,0,0)
l/U;, 1/ U-,) + u, {1/ U'o,!/ U;,ll U;,lf U\, II IT,)
+ u, (1/ f/'o, 1/ U;, 1/ U'„ II U\, 1/ U\) -5 = 0; (2)
the ( ) being the usual notation for complex 5-numbers; F' = \, since the
point ( w'o , • • • , k'i ) is on the surface F = \. The equation may be shortened
to
Uc to + Wi t)i -f W2 1)2 + W3 fS + W4 f 4 — 5 = ,
where the V's are the quantities in ( ) of (1). In virtue of the fact that points
and surfaces of the same kind as the tangent surface are dual elements in Ss,
the c's will be referred to as the coordinates of the tangent surface. Nearly
all the properties of F = 1 that follow are from a direct reading of these two
equations. The surface whose coordinates are I'o , • • • , »4 will be denoted by
(»o, ti, V2, V), Vi) . Multiplying the c's by 1, 1, 1, 1, 1 and adding, it is
seen that the point (1,1,1,1,1) will lie on the tangent surface (2) if U'q=1,
that is, if the point of tangency lies on the surface ( 1 , 1 , 1 , 1 , 1 ) ; or, the
tangent surface to the surface F = 1 at any point of the intersection of F
and the surface (1, 1, 1, 1, 1), passes through the point (1, 1, 1, 1, 1);
or again, the surfaces which pass through the point (1,1,1,1,1) and touch
F, meet the latter in the points common to
Fo = Ui Uo h\ ir, - 1 =
and F . By analogy, Fo is called the "polaf 4-surface " with regard to the
point (1,1,1,1,1), and the surface (1,1,1,1,1) the " polar 1-surface "
of, etc. The " 4-surface," " 1-surface," are so named to distinguish two
things having similar properties. As in the foregoing, similarly for 1 , p ,
p'j p'j p^ , multiplying in turn by these and adding, it is seen that the 1-surfaces
enveloping the point (1 , p, p-, p', p^) and touching F , meet the surface (F)
in the points common to F and the polar 4-surface
Fi = C7o C/j Uz Ui-l = 0;
and so on for each of the remaining 3 special points ( 1 , p' , p* , p , p' ) ,
( 1 , p' , p , p* , p^ ) , ( 1 , p* , p' , p^ , p ) whose polar 4-surfaces are respectively
F^UeUiUiUi- 1 = 0; F3^UoUiU2Ui-l = 0, and
F4S UoUiUiUi- 1 = 0.
These 5 special points are among the 120 which lie at zero distance from the
origin on the 5-oid
nt + l4+y|+^'l+^'l-5=^ 0;
78 ON THE CYCLOTOMIC QUINARY QUINTIC.
distance being defined as the square root of the usual quadratic function, in
accordance with one usage^ not as the fifth root of a quintic function, which
in some respects seems better here. On this 5-oid there are at once 3,125
( = 5* ) special points obtained by assigning all 5 possible values of p* to each
of the w's in the general point ( j/q , Mi , ^2 , «3 , «4 ) . The 5-oid is transformed
into itself by any substitution which merely permutes the suffixes, hence it is
invariant under the symmetric group of degree 5. Also, if a^, • • • , gj be any
of the 5 numbers p', then
Sj^ fay
bj
Cj
dj
ffy
transforms the 5-oid into itself; and the Sj generate an abelian group of order
3,125; hence the special 5-oid is invariant under the abelian group of order 5^,
If Tij denotes the result of putting all the elements in Sj except that one in
the tth row = 0, then the group is generated by T.-y, (i = 1, • • • , 5); that
is by 5 operators of order 5, Ti, • ■ • , Ts, each of which generates a cyclic 5-
group, etc., so that the entire group may be represented by { Ti , ^2 , Ts , 2^4 , Ts} .
Returning to the 5 tangent 1-surfaces already considered, it is easily seen that
each passes through five special points not on the 5-oid (nor on F). Also it
appears presently that each of the 1-surfaces gives rise to four others, so that
associated with any of the 1-surfaces there are 25 special fundamental points
lying on the associated 5-oid, and these 25 points correspond to the powers of
Ti, • • • , Tb. Again, the mean-center of each of the sets of five points lies
on the corresponding l-surface, and the five mean centers lie on a 5-oid of
5~^ times the linear dimensions of the original 5-oid. Here in passing, the
remark made concerning the analogy of the 5-oid to the director sphere or
circle may be partly justified; the working out in detail is reserved. The
theorems quoted regarding circles, etc., all follow at once from the parametric
equation of a circle, and the resulting sin (a =*= 6) formula that such repre-
sentation implies. The analogy here is possibly due at bottom to the fact
that the sine is singly periodic of a single argument, the functions used here
to represent a point on F = I are by § 2, 4-fold periodic on 4 arguments; the
addition theorem is algebraic, and geometrically can be made to interpret
the properties of mean centers in the same way that the trigonometric functions
give rise to the following. The tangents at the vertices of the parabolas
touching each three of 4 lines joining 4 concyclic points and having those
ON THE CTCLOTOMIC QUINARY QUINTIC.
79
points as foci pass through the center of the rectangular hj-perbola circum-
scribing the 4-point, and if 5 points be taken, the corresponding tangents at
the vertices pass by 4's through 5 points lying on a circle of one-half the linear
dimensions of the original circle; and if a sixth point be taken, the six cor-
responding circles are concurrent; the seven such points obtained by taking a
seventh point on the circle, are concyclic, and so on indefinitely, all the circles
being equal to one half the original circle. Again, ]\Ii^uel's theorem states
that the intersections of the 5 circles circumscribing the triangles formed by
producing the sides of a 5-side are concyclic, (the intersections being the foci
of 4-line parabolas; and it may be shown that the 5-point circle is the director
of the conic touching the 5 lines; taking a sixth line, a seventh, • • •, there is a
chain of theorems as above. But the center of the rectangular hyperbola is
midway between the mean center of the 4 concylic points and the center of
the circle. In the same way there is a surface in Sj corresponding to the hyper-
bola, and so on; there is a chain of theorems. But again, as indicated in § 2,
the theorems may be extended to surfaces of lower degree in other spaces;
e. g., to quadrics in 4 space, the existence of a generalization of these circle-
properties depending on that of an algebraic theory of composition for the
degree-order considered. For ordinary space, Eisenstein's form quoted in
§ 1 could be used to find results precisely analogous to the present for Ss .
Returning; for brevity denote by 0;, 1,-, 2,, 3,-, 4,- respectively the points
all of whose coordinates except the first, . . . fifth, are zero, the non-zero
coordinate being p'. Also by {tq, • ■ • , lu; Ic} the 1-surface
lio I'o + • • • -\- UiVi = 5k ,
(?\ p'\ p«;5pM;also
1, P'
and by {0, i, 2i, 3j, 4f, k } the 1-surface
by [i,j,k, I; c] the 4-surface
Ui Uj U, Ui = c;
and hyl/s {0 ,{,2i,3i ,Ai) the point whose coordinates are 1 / S , p* / S , p^'/ 5 ,
P^'/S, .p*'/S, then:
1-Sur- Contains
face Points
(Ii) {0,4,3,3,1
(III) {0,4,3,2,1
(IIIi){0,4,3,2,l
(IVr) {0,4,3,2,1
04,lo,2i,32,43 (4, 0, 1, 2, 3)/o [0,1,2,3
Oo,l2,24,3i,43 (0, 2, 4, 1, 3)/o [0,1,2,4
(Vi) {0,4,3,2,1
ih) {0,3,1,4,2
(II2) {0,3,1,4,2
(III2) {0,3,1,4,2
(IV2) {0,3,1,4,2
(V2) {0,3,1,4,2
WTiose Mean And Polar 4-Surf. Has for
Center is Corrcsp. Pole
0} ; Oo, 1:, 2., 83, 4, (0, 1, 2, 3, 4)/5 [0, 1, 2, 3
1) Oi, 12,23, 34, 4o (1, 2, .3, 4, 0)/5 [0,1,2,3
2} 02,13, 24, 3o,4i (2,3,4,0,l)/5 [0,1,2,3
03,l4,2o,3i,42 (3,4,0,l,2)/5 [0,1,2,3
Oi, I3, 2o, 3;, 44 (1, 3, 0, 2, 4)/5 [0, 1, 2, 4
02,14, 2,, 33, 4o (2, 4, 1, 3, 0)/5 [0,1,2,4
03,lo,22,34,4i (.3,0,2,4,l)/5 [0,1,2,4
04,li,23,.3o.42 (4, 1, 3, 0, 2)/5 [0,1,2,4
0] (0,1,2,3,4)
1] (1,2,-3,4,0)
2] (2, 3, 4, 0, 1)
3] (3,4,0,1,2)
4] (4,0,1,2,3)
0] (0,2,4,1,3)
1] (1,3,0,2,4)
2] (2,4,1,3,0)
3] (3,0,2,4,1)
4] (4,1,3,0,2)
Q
80 ON TUE CTCLOTOMIC QUINARY QUINTIC.
(13) {0, 2, 4, 1, 3; 0} Oo, I3, 2i, 84, 42 (0, 3, 1, 4, 2)/5 [0, 1, 3, 4; 0] (0, 3, 1, 4, 2)
(113) {0,2,4,1,3;!} Oi, l4,22, 3o,43 (l,4,2,0,3)/o [0,1,3,4;1] (1,4,2,0,3)
(Ills)! 0,2, 4, 1,3; 2} O2, lo,23, 3i,44 (2,0, 3, 1, 4)/5 [0,1, 3, 4; 2] (2,0,3,1,4)
(IV)3 {0,2,4,1,3:3} 03,li,24,32,4o (3, 1, 4, 2, 0)/5 [0,1, 3, 4; 3] (3,1,4,2,0)
(V3) {0, 2, 4, 1, 3; 4| O4, I2, 2o, 3,, 4i (4, 2, 0, 3, l)/5 [0, 1, 3, 4; 4] (4, 2, 0, 3, 1)
(14) {0, 1, 2, 3, 4; 0} Oo, I4, 23, 32, 4i (0, 4, 3, 2, l)/5 [0, 2, 3, 4; 0] (0, 4, 3, 2, 1)
(114) {0, 1, 2, 3, 4; 1 } Oi, lo, 24, 83, 42 (1, 0, 4, 3, 2)/5 [0, 2, 3, 4; 1] (1, 0, 4, 3, 2)
(III4){0,1,2,3,4;2} O2, li,2o,34, 4, (2, 1,0,4, 3)/5 [0,2, 3, 4; 2] (2,1,0,4,3)
(IV4) {0, 1, 2, 3, 4; 3} O3, I2, 2i, 3o, 44 (3, 2, 1, 0, 4)/5 [0, 2, 3, 4; 3] (3, 2, 1, 0, 4)
(V4) {0, 1, 2, 3, 4; 4) O4, I3, 22, 3i, 4o (4, 3, 2, 1, 0)/5 [0, 2, 3, 4; 4] (4, 3, 2, 1, 0)
(lo) {0, 0, 0, 0, 0; 0) Oo, lo, 2o, 3o, 4o (0, 0, 0, 0, 0)/5 [1, 2, 3, 4; 0] (0, 0, 0, 0, 0)
(IIo) {0, 0, 0, 0, 0; 1 } Oi, li, 2i, 3i, 4, (1, 1, 1, 1, l)/5 [1, 2, 3, 4; 1] (1, 1, 1, 1, 1)
(Vo) {0,0, 0,0,0; 4} 04,14,24,34,44 (4, 4, 4, 4, 4)/5 [1,2, 3, 4; 4] (4,4,4,4,4;
The whole configuration may be expressed by the symbol | 1 2 3 4 j; the first
set of elements is written down from this, the l-surface by replacing each non-
zero symbol by its complement (mod 5), the points are clearly enough indicated,
as also is the mean center, the polar 4-surface is written down by changing the
last symbol 4 to 0, and the pole comes at once (as throughout) either from the
l-surface, the points, or the mear. center. The first row being found, the 4
following come by addition of unity (mod 5) to the variable symbols. This
gives the complete set (h) • • • (Vj). The sets (I2) • ■ • (Fo); (I4) • • • (V4);
• • • (I3) • • • (V3) come by doubling and reducing each corresponding symbol
in the first set. The last set starts with zero as a basis, and then is written
down as was the first set.
There are here 25 of the special points, and 25 of the special 1-surfaces;
each l-surface contains 5 of the special points, and through each point pass 5
of the 1-surfaces. In order to place the mean points, some further definitions
regarding a 5-oid are required. The special 5-oid was
wg+ |-w»= 5,
which by making the equation homogeneous may be written
the point (0, 0, 0, 0, 0) will be called the center, and the constant Ws the
radius of the 5-oid. The tangent l-surface at (ro, • ■ • , Ps) is
Wo rj + «l V\ + U2Tt+ «3 li +Uiii-2o = 0.
It is now seen at once that the 25 mean centers all lie on a 5-oid whose center
ON THE CTCLOTOMIC QUINARY QUINTIC.
81
is at ( , • • • , ) , and whose radius is 1 / 5^'* . Also the 25 poles lie on the 5-oid
whose center is (0, ••• . 0), and radius 1/5^'^; that is, the "mean 5-oid"
is 5~* times ths linear dimensions of the " polar 5-oid." The tangent 1-surface
at ( 7)o , • • • , ^4 ) to the mean 5-oid being
«o «4 + • • • + M4 cj — 5"-* = 0,
at the point (0, 1, 2, 3, 4)/ 5 this is
«0 + P*Ui+ p' Ui + p^U3+ pUi— 1 = 0,
which is parallel to the 1-surface corresponding to this mean center, viz., (Ij).
In the same way, it may be shown that for each of the 25 mean centers, the
tangent 1-surface to the mean 5-oid at a given mean center is parallel to the
1-surface containing the corresponding pole, or what is the same thing, is
parallel to the 1-surface containing the 5 points of which the point of contact
is the center of mean position. As usual, defining the distance between two
points u, u' of (Sg to be { (wo — y'oY + • • • + ("4 — ■w'*)" V'^, the distances
from, the common center of the two 5-oids to the special points ij are equal in
sets of 5 to 1 , p , p^ , p' , p^ . Hence these 25 points lie by 5's on the 2-surfaces,
Wo+«?+J/2 +«!+«!- P"'= (i = 0, 1, 2, 3, 4);
and by Cayley's theory (1. c.) these 2-surfaces are ruled quadrics in S^.
These special 25 points are in fact on 25 special surfaces connected rvitli the
configuration |01234|, a 1, 2, 3, 4, 5-surface respectively, an t-surface
being one of the ith degree; the equations being
«^+ W+W2+ W3 + W4- P'= 0; (i,y = l,2,---,5).
Each of these surfaces contains 5 of the special points; all the surfaces of the
same kind contain all 25 points, and through each point pass 5 surfaces, one
of each kind. The entire ( 1 2 3 4 | configuration is clearly only part of a
more general configuration connected with the surface in S^; which will now
be examined. The figure is more easily kept in hand if a special symbol
1,
2,
4,
3
2,
4,
3,
1
3,
1,
9
4
4,
3,
1,
2
is introduced; the only defining properties of S to be (at present). (1°) That
any number occurring in S is positive, < 5 , and hence only one or more of the
numbers 0,1,2,3,4; (2°) An interchange of any two rows leaves S unchanged ;
(3°) An interchange of two columns is not permissible. The last restriction
82
ON THE CYCLOTOMIC QUIN'ARY QUINTIC.
may be removed later, as will be seen, once the properties and manner of
operating with S have been established. Considering now the difTerent
possible values of 5 when the same (permissible, i. e., > 5 ) constant is added
to each member of a row, \eta,b,c,dhe the numbers 1 , 2 , 3 , 4 in any order,
but no two of the letters are to have the same value; put
S'= 1 + a. 2 + a, 4 + a, 3 + a
2 + b,
4+6,
3 + 6,
1 + 6
3 + c,
1 + c,
2 + c,
4 + c
4 + cf,
3 + d,
1 + d,
2 + d
then S' has 24 distinct values, for by definition an interchange of columns
changes the value of S' for any system of \'alues for ( a , 6 , c , cf ) , so if two values
of 5' become identical, this can happen only in virtue of any two rows becoming
identical. Say the first two rows of two values for S become identical (if
the two identical rows are not the first, the definition permits them to become
so), say
1 + a, 2 + a, 4+0, 3+a = 1 + a', 2 + a', 4 + a', 3 + a',
in this order (necessary) ; this requires that a = a' (mod 5) , and since a ,
a' < 5 , must a =^ a' . But the values oi a,h , c, d must all be different in a
given S, therefore S takes 4 • 3 ■ 2 • 1 = 24 difTerent values in all. Hence,
if the sum of two symbols S , 5' of the above kind be defined as the symbol of
the same kind resulting from the addition of corresponding numbers in cor-
responding columns of S and S' , it follows that with respect to addition the
foregoing 24 values belong to the sj-mmetric group of degree 4. Arranging the
24 values according to the number of zero values that may occur in any column,
consider first the case of 4 zeros in a column; the system (4, 3 , 2 , 1 ) of values
for {a,b rC,d) reduces the first column to all zeros, and therefore the systems
of values obtained from (4, 3, 2, 1) by doubling and reducing, reduce the
second, third and fourth columns respectively to all zeros, for these columns
in S are obtainable from the first by the same process. Or again, the com-
plementary S to a given S being for a given column that in which all the columns
are zeros except the column whose elements are complements modulo 5 of
the given column, the initial S is reduced to the form in which its first, . . .,
fourth column is all zeros by adding to S the complementary S for respectively
the first . . . fourth columns. There is no form of S in which three elements
of any cohimn are zeros, unless the fourth element in the column is also zero,
for a , 6 , c ,
3 4 10
J
4 2 1
)
13 2
J
12 3 4
13 2
2 4 3 1
3 4 2
4 10 3
2 14
14 2
2 13
3 12 4
4 2 3
3 4 1
10 3 4
2 14
3 2 1
4 3 12
4 2 3
after which it recurs. The effect is to reduce the number of zeros by one.
Since each of the X^ contains 4 zeros, and no three are in a column, and since
no 4 elements in a column are identical, no one of the X^ may be reduced by
addition of Si to a form containing less than or more than, and therefore exactly
3- zeros. Hence the 20 values of P ( = ( 1/0 > ~ X^ ) , are in two sets, the first of
which contains 4 S's such that their ( 1 / X) (or X^) parts contain 4 zeros; the
second set contains 16 S's each of whose X^ parts has only three zeros. (By X^
part is meant what remains in an J? when the first row and column are deleted).
Furthermore these 20 values are all distinct values for the Z^; for no one of these
F can contain more than 4 elements each equal to the same integer , • • • , 4;
and each of them contains exactly 4 equal respectively to 1 , 2 , 3 , or 4 , except
the first set of 5 which has 4 zeros, these numbers 1 , • • • , 4 only occurring
(in the 4-way distribution) in the places of the original zeros. Clearly then,
no two of the sixteen can be identical. Considering now a typical X', say
2
3
4
2
1
4
3
4
1
3
2
1
it is seen in the same way as for the X^, that each has exactly 3 zeros by the
addition of Si, and as each has at first not more than 3 identical elements of
any kind except zeros, no addition of S,- to any of the X' can produce a form
in which there are more than 3 zeros; also, since there are exactly 3 identical
elements at each stage, each of the P , except the initial 4, have 3 and only
3 zeros. It is seen at once that the sixteen corresponding values of the P
(each with 3 zeros), are distinct from the sixteen P, found above; for in no two
of the l^ , P initially is the distribution of zeros similar; hence in no two of the
derived P , P can the 4 identical elements be similarly arranged. By precisely
similar reasoning, it is easy to see that no two of the similarly derived P , P , P
CMi be identical among themselves, or to any of the I, P, P. Hence the
distinctness of the whole 120 l,t-,P,P,P,Pis proved, also the separation of
them into classes according to the distribution of their zeros in one case,
identical elements in others. Summing up; there are of all the Z", 4 which
86 ON THE CYCLOTOMIC QUINARY QUINTIC.
contain in their (X) part 4 zeros in the same column; 4 with 4 zeros, 2 zeros
in each of two columns; and 16 with 4 zeros, 2 zeros in one column and one in
each of two others; that is, 24 with any distribution of 4 zeros in all; and 6
sets of 16 each derived from the above by additions of multiples of Si, con-
taining each 3 zeros, each set of 16 being derived from an initial X, X^ • • • , X'
by the addition of 5i , 2Si , 3Si , 4Si . Of these last, just 8 have their 3 zeros
in the same column; and so on.
To each of the 120 symbols so derived corresponds a configuration of one
point ard a definite surface such that the tangent 1-surfaces to the original
surface F at its points of intersection with the definite surface, all pass through
the fixed point. This 120 configuration is not studied in detail here, nor is
the total configuration of 3,125 fixed points of which the 120 is only a part.
In order to investigate the figure completely, it will be necessary to consider
along vdxh F , its " reflected " surfaces; the idea of a reflected curve in a plane,
e. g., of a circle in the axis of reals by change of sign of the ?/-co6rdinate, etc.,
being carried over into 5-space (ideal), where the F-surface has " reflections "
in several planes, or about several axes. The extension is natural and con-
venient, for reflection as ordinarily defined consists in a multiplication of a
coordinate by — 1 , that is by one of the units of the natural numbers; in the
ideal 5-space here considered, " reflections " are accomplished by multipli-
cations of one (or more) coordinates by units. As there is an infinity of units,
so each point has an infinity of reflections about oo lines, super-lines, etc., but
all may be referred to the fundamental system =tp'(i=0, 1,2,3, 4).
This extension is necessary, if a complete discussion of the geometry of the
" configTiration 3,125 " is to be given; algebraically it is a discussion of certain
associated forms F {p"'ao, p" ai , p*" ao , p' aj , p' Ui) , which present themselves
naturally in considering F {ao, ■ ■ ■ . Ui) . The totality of distinct forms F
obtained in this way, constitute in the ideal Ss, all the " images " of F ob-
tained by reflections. (Two forms are not distinct if F' = eF , where e is a
complex unit). Thus, the more general configuration contains 3,125 points,
an equal number of polar surfaces, and the whole configuration is invariant
under the group which leaves F and its fundamental images invariant. The
special surfaces considered are separated into species according to the form of
the 5 (and other symbols constructed in the same way), corresponding to
them. For clearly, each 5 corresponds to one and only one form of the equa-
tion of condition that the tangent surface to F shall pass through a given point
(p° p'' p' p"* p"); the symbols S correspond to a , b , c , d, g = 1 , 2 , 3 , i , 5;
no two letters having the same value. The rest of the configuration is derivable
on the hypothesis that at least two of the a , ■•• , g have the same value.
Again, all these points lie on the special 5-oid. The 120 point-surface con-
figuration has been selected out of the total because its properties may be
ON THE CYCLOTOMIC QUINARY QUINTIC.
87
seen by direct reading, also its points are all those and only those no two of
whose coordinates are equal; also, as already noticed, in the sphere-inter-
pretation, these 120 are those and all those which reduce to points (as spheres).
It is clearly not invariant under the group of all possible simple reflections;
this group is that already found of order 5'; hence the substitutions which
leave the 120 points unchanged, that is, permute some or all of them among
themselves, must be in number 1, 5, 25, 125 or 625. First, looking at a
given pole, e. g., that represented by (0, 1,2, 3, 4), the only substitution
of the group which leaves this unchanged is the identity, but obviously, it is
transformed into another pole of the set by
r= fp
P
p.
and therefore also by T-, "P , T^. Similarly, it is seen that { T ] transforms
any pole into 5 poles of the set. Also no transformation which contains two
dissimilar powers of p can change any one of the set into another of the set.
Therefore the poles may be arranged in 24 sets of 5 each, such that all in a given
set are derived from any one of them by the transformations { T } . Five such
sets have been shown in the table, for the pole sets (Ii) , (h) , ih) , (h) , (lo)',
(III), (Ila), (Ills), (IV3), (lo), etc. The configuration of 120 poles is
thus composed of 5 sets of 5 five-point figures which have a point in common,
— ^that is, from the fundamental points or" vertices " lo, Ho, HIo, IVq, Vo,
are " drawn " sets of five-point figures; 5 from each fundamental vertex, and
any set of 5 five-point figures having a common vertex is transformed into
itself by 1 , T , 'P , P; T^; and the whole is inscribed in the special 5-oid
vl + ti\ + t/H- «! + lil
5 = 0.
Other properties were given above. The entire set of 120 poles, etc., is char-
acterized by the symbol {ah cd g) , all the letters distinct; the remaining
3,005 are given by the symbols in sets (aaedg);{a,a,a,d,g);{a,a,a,a,ff}
(^a, a , a , a , a) .
A set of values for a, • • • , ^ in a given S was denoted hy (a, b, c, d, g);
and it has been seen that from a X , • • • , X^ , other values of S are derived by
additions of multiples of Si; and it is clear that the resulting S is the same
regarded as an (Z) • • • (/^) if the sjmbols (a, b , c, d, g) are augmented by
addition of a constant to each letter, or if a distinct set of symbols (a, ■••,§)
is found for each I as it arises; that is, from the set ( 1 , 2, 3, 4) , 0?/ the 120
symbols are derived by successively increasing each number in the symbol by
88
ON THE CYCLOTOMIC QUINARY QUIXTIC.
a constant, and then reducing modulo 5 . But the /'s are 5 columned and 5-
rowed symbols; hence it is not sufficient to increase the numbers in the (12 3 4),
etc.; a means of identification must be given to show to what particular / this
4-row belongs. Clearly this is only a convenience, not a necessity, for no
(12 3 4), • • • can belong to more than a single I, and any such symbol being
given, it is possible to identify the I to which it appertains, but if a suffix
indicating the constant by which every number is to be increased, is subjoined
to the symbol, the trouble of identification is avoided. Hence the whole 120
symbols are conveniently represented by
(Xy.), (X5,), ••., (Xy.)«; (i = l,2, 3, 4, 0); (y = l, 2,3,4);
thus
Xi2= (1, 2, 3, 4)2^(3, 4, 0, 1);
n.z= (3, 2, 1, 4)3= (1, 0,4, 2);
if the j denotes the index of the column in which the given function of X occurs
in the " anharmonic table." For the sake of uniformity with preceding parts,
the order will be
X,,,= (1234),-; X,.,= (2413);-; X3.,= (3, 1, 4, 2),; X,.,= (4,3,2,1),-;
where, e. g.,
Ki^(3+j, l+j, i + j, 2+j) (mod 5).
Summing up; the symbols of the form S are written down as follows: the six
usual forms of the anharmonic ratio of 4 elements, 1,2,3,4 are written down,
and in each of the six classes so defined are put respectively the 4 equally valued
anharmonic ratio-symbols. The six classes so formed all have suffix zero,
the other classes correspond to suflSxes 1,2,3,4. In the 5-row symbol I
corresponding to any of these (abcd)i the first row and first column are
composed of the element j repeated in each 5 times, the (X)-part of the / is
the iS derived directly from the {ah cd)j. Having written down the six
anharmonic ratio signs as a first step, the three remaining sets of 6 each are
deduced from this by doubling and reducing modulo 5 every number Ln the
sj-mbols. As these symbols represent the polar surfaces, also the corresponding
poles, and as the form of a surface is determined by the number and location
of the zer^in its symbol, it is convenient to classify the ( ) on this basis. The
pole of
X.y ^W b' c' d')j
is the point denoted by (j, a' ,b' ,c' ,d') , that is, its coordinates are p'', ■■■, p"".
This may be taken as a definition, or an extension of the usage already adopted;
the X's arose from the properties of the configuration, now conversely the
X's may be assumed a priori and the resulting configuration deduced. The
ON THE CYCLOTOMIC QUINAKT QUINTIC. 89
pole whose coordinates are the above will be said to correspond to X,y a pole
is denoted by the corresponding X . If 5 poles X , X' , • • • , X" are so related that
they are cyclically permuted among themselves by the transformations { T ] ,
then it is natural to say that the 5-point X' X" • • • X" has been rotated about
that from which the 5 points are equidistant, distance having been already
defined. If {ao, bo, Cq, do, go), •••, (at, •••, gi) are 5 fixed points, and
(wo, ■ • • ,Vi) any point equidistant from them, then the 4 independent equa-
tions
(wo - aty +... + (u,-gJ-= (wo - a.+i)- + • • ■ + (w4 - gi+iY
(i=0, ••-. 4),
defining a line in Sj, determine, if it exists, the" axis of rotation " for the
5-point. The points X.-y give rise to 24 axes of rotation, one for each cyclic
5-point; e. g., the axis of the set Xi, o, Xi, i, Xi,?, Xi, 3, Xi,4, that is of
the poles (0 1 234), (1 2 3 40), (2 3 40 1), (3 40 1 2) (4 1 2 3) is
(«o - py+{u^- p^y + (t/2 - p')= + (W3 - p'y+ (w* - D-
= (wo - 1)^+ (wi - py+ (m2 - p^)-+ («3 - p')^ + (W4 - p*y,
(uo- p'y+im- p''y+ (th- p'y+(u,-iy+ (u,- py
= iuo - iy+{th - py+(t>2 - p'y+ («3 - p'y+iv^ - p^y,
(wo + p'y+ (wi - p'y + («2 -iy+(u,-py+ (u, - p=)=
= (wo-i)='+(wi-p)^+(«2-p=)-+(w3-p')-+(w4- p'y,
(Mo - p'y + («i - 1)^ + (W2 - py+ («3 - P-y + (u^ - p^y
= (vo-iy+in,- py+(m- p^y+im- p^y+in,- p^y.
By inspection, this axis passes through (0,0,0,0,0) the center of the 5-oid,
and in the same way all 24 axes pass through the center. It may be seen that
all the axes intersect in the locus constituted by all points all of whose coordi-
nates are equal for the above equations obviously pass through (arbitrary)
{a, a, a, a, a) when they are written
«0 (P - 1) + Wl (p' - p) + "2 (p' - p') + W3 (p' - p') + W4 (1 - P^) =
Mo (p'' - 1 ) + wi (p' - P) + W2 (P* - P-) + J/3 (1 - p') + W4 (P - P') =
i/O (p' - 1 ) + «1 (P' - P) + «2 (1 - P-) + «3 (p - p') + W4 (P- - P') =
Mo (p' - 1 ) + «1 (1 - P) + «5 (P - P') + M3 (P'- P') + «4 (p' - P') =
If the equations of the 1-surfaces corresponding to 5 poles Xi , 0, • • • , Xi , <
be solved for their corresponding intersection, it is found that the point lies
90 ON THE CYCLOTOMIC QUINARY QUINTIC.
wholly at 00; i. e., these 5 1-surfaces may be said to be parallel; in the same
way all 24 cyclic sets of 5 are parallel among themselves, so that in this case
the 120 poles and corresponding 1-surfaces may be said to constitute a regular
body in 5-space. The body from which this was obtained by projection at the
start is not regular, but symmetrical in 5 directions; in other words, the pro-
jected body corresponds to a regular solid (pentehedron), the unprojected to an
irregular star-shaped body; also the 5-oid is the correspondent in Ss of the
sphere, whereas the surface from which it was obtained, viz.,
«g/A-g+wf/A-f+ ••• =0
by a- projective transformation may be considered the analogue of the central
quadrics; the degree is 5 instead of 2, but the properties of the 5th degree are
in some respects closer analogues than those of the 2d; in all events it is
simply a fashion of speaking of algebraical (arithmetical) facts embodied in
the form.
Since
1'+ (P')'+ (P^')-+ (P")-+ (P^*)'=
if i is not divisible by 5, it follows that the 120 poles are at zero distance from
the origin, and do not coincide with it, in fact they lie on
vl+ hwf-5 = 0,
whose center is the origin. As in Ss or Si the geometrical language becomes
slightly paradoxical, but the. 5-oid may be regarded as a semi-minimal surface;
semi because all the points on it are not at zero-distance from the origin. The
set of 120 poles are all those of the entire 3,125 which have this property. A
result more consistent with ordinal^' 3-space could have been obtained by
defining distance not as a function of squares, but of fifth powers for the ideal
(Ssj such a definition (or its analogue) is not usually adopted. The apparent
contradiction arises from the ideal nature of the elements considered.
The distribution of 1-surfaces corresponding to the 120 poles in relation to
the 25 fundamental points Oo, ■ • • , 44 is readily determined. If (01234),
that is Xi, be the pole considered, the l-surface corresponding is found by
writing down the complementary symbol {04321;0j, similarly from
Xi, 3 (say), or (3 4012) is derived {0432 1;3); first by subtracting 3
from every term, and then taking the complement. The sets of fundamental
points corresponding to Xi,o and Xi,3 are respectively Oo, li, 2;, Sa, 44,
and O3, 14, 2o, 3i, 4; or as already pointed out, the suffi.xes are the same
numerals in the same order as those which occur in the pole; hence the pole
indicates by position of numerals what fundamental points lie on the cor-
responding l-surface; e. g., given pole (2 3 1 4 0), the l-surface will contain
O2, I3, 2i, 34, 4o. And generally, all X-symbols having a same numeral in
ON THE CYCLOTOMIC QUINARY QUINTIC. 91
the same position (e. g., in the third place) will be so related that the corre-
sponding 1-surfaces all pass through the so-determined point (or vertex).
It is easily seen that the point Oo is common to the 24 1-surfaces
"•10 > A20> ^30) ^iO't ••■> AlOJ '■'» ^40>
Oi is common to the 24,
^11 J X2i» Xsij X41; •••; XjiJ ■■■» X^iJ — j
and finally, O4 is common to
"14 J A24> A34 , A44; •■■; Aj4 , *••> A44 .
Similarly for the remaining 4 sets
li, ••• , lo; ••• 44 ••• 43;
each of which gives rise to a permutation of the X's as written above. Hence
the 120 1-surfaces pass by 24's through the 25 fundamental points. There
will likewise be arrangements with respect to 2 , 3 , or 4 of the fundamental
points, so that for (say) 2 chosen from among them all, there will be a deter-
minate number of 1-surfaces having these two points in common. Looking
at the first line of X's just written, it is clear that only 1 / 4 or 6 of them contain
a specified point, as li, other than Oo; these are Xj*, Xfo, \il, Xj*, \l, X%.
Similarly for any other point pair of any other set, and since there are
25-24/1 • 2 = 300 such pairs, it follows that the 120 1-surfaces are arranged
in 300 sets of six each such that each set of six has two points in common.
For the same 24 X's, looking at any set of six with two points common, e. g.
those with the arrangement (012**) are Xio , Xio . Remembering that
, 1 , 2 in any other order in the ( ) gives a distinct set of common points,
there are in this way given 25 ■ 24 • 23/ 6 = 2,300 sets of 2 . Hence the 120
1-surfaces are so disposed that 2,300 pairs of them have three points in common..
Similarly if 4 of the 25 points are to be common there is but a single 1-surface
satisfying the condition, if the points are the first 4 in (0 1 2 3 *) the 1-surface
is Xio. There are in all 12,650 such. Finally, there are in the configuration
of 25 fixed points (fundamental),
21 X 12,650/5 = 53,130
absolutely fixed 1-surfaces. From the equation of the 1-surface tangent at
(v'q, • • ■ , v\) to the surface F , the conditions that ( wj,' , ■ • • , v'i ) shall lie
on this tangent are written down from equation 1 (at beginning of this section);
if (wq ) • • • J W4 ) is the symbol of one of the 120 poles, the condition is
( P"°" + P""+' + P"'"+' + P"'"-*-' + P"'"-^' )/U[+{ p""" + p"'"+2 + p"^'+'' + p""+»
+ P"'"-^' ) / f/i + ( P"°" + P"'"+' + P"'"+' + P'""+- + P"-'""-' )/U\+( P-" + p-'^'
-f P--+' -f p-"-f p"+^ *"+=)/ U', = 5;
92 ON THE CYCLOTOMIC QUGfAHY QUINTIC.
or dropping accents, say
R1/U1+ ■■' + R,/U^= 5,
which intersects the surface F in all points at which the tangent 1-surfaces
pass through the fixed pole {ii'a , •••,<). This surface (R) contains the
fixed line
l/'i = 4i?i/5; ?72 = 47?2/5; C7s = 4i?3/5; U, = iRi/5.
There are 120 such 4-surfaces (surfaces of 4th degree) intersecting F in points
at all of which the tangent 1-surfaces pass through fixed points, — the 120
poles. A set of 24 will be considered in some detail first; putting
P" + p"^^ + P^" + p''+' + p"^! = rj/ 5 ^ r;
P'+ p'^+ P^' + p'+' +P'^'=rzl5^s,
P' + p'^' + P^i + P''^ + p<>^=n/5^t,
and for convenience
1/Ui, I/U2, I/C/3, 1/U,^x, y. z, w,
where a, • • • , g are the numbers , 1 , 2 , 3 , 4 in any order, the equation
px + ry -jr sz + iw =^ 1
represents the 24 1-surfaces corresponding to the poles derived from the
symbols S, if a = 0; if a ^ , the letters
b, c, d, g= 1, 2, 3, 4.
When any of the 24 systems of values for (fc, c, «Z, ^) in the corresponding
iS has been assigned, the values for p , r , s , f respectively are found by adding
the elements in the first, • • • fourth column of S, to form a new complex
number a', b' , c' , d' , g' . It is evident from the properties of S that the
coordinates so found will not all be equal; there can be no confusion if the
symbol {a' , ■ ■■ , g') is used for this number. Thus, e. g., let
(a, b, c, d) ^(2143);
then S becomes
3
4
1
3
4
2
2
1
3
2
1
4
ON THE CYCLOTOMIG QUINARY QUINTIC.
93
then
p=(-.-2,2.); r= (2, 1, •••. 1); * = (1, 2, • • ■ . 1);
<= (1, -1, 1, 1),
or WTitten in full
p = 2p^+2p'; r-2+p+p^ * = 1 + 2p + p^ t = 1 + p' + p' + p';
in brief, the numerals in S are the indices of the powers of p, if p" is zero, but
if a is (say) 2, 1 will have to be added to every numeral in the third place, so
in this case,
p,r,5,<= (.-3,2, •), (2,1,1, -1), (1.2,1, • 1), (1-2, 1, 1);
for a = 4 , 1 is added in 5th place,
(•-221), (21--2), (12.-2), (1 -112),
commas being omitted from ( ) as no number > 9 can occur. Writing down
the six sets of four numbers as values for p, r, s,t, corresponding to, and in
the same order as the X's in the anharmonic table; find for a as yet arbitrary;
p, r, 5, < = {
};
(table on next page). In the table p" =# , but is arbitrary; giving a the values
0, 1, 2, 3, 4 in succession, which is equivalent to increasing the digit in the
first • • • fifth place respectively by unity, 120 sets of values for p, r , s , t are
found, corresponding to the 120 poles whose coordinates are all distinct, i. e.,
in any pole no two coordinates are equal. Each of these sets gives one surface
of the fourth degree, defining a cone (in Ss); viz., the tangent 1-surfaces from
the corresponding pole touch the surface F along its intersection with the fourth
degree surface. Again, as has been seen in one special case, each of these
{ p,r, s ,t I has in it, or envelops, a fixed line, that is a skew curve fixed in Si;
there are clearly 120 such. There are 30 sets of 4 surfaces each; the surfaces
in any set intersect in 4 1-surfaces, or rather have these in common. E. g.,
caUing X; what X becomes when a is i, the 1-surfaces common
P
(X) {(-1111
(X^) {(-31 •
(X') { (2-11
(X^) {(112-
(X^) {(112-
(X«) {(• •3-
(1)
r s t
(•1111), (4 ), (-1111)1
(1 • -21), (2-11-), (1 -1 ^2)}
(• -22-), (21 • • 1), (•2- -2)}
(2-11-), (12- 1.), (. • • 13) !
(•1-3-), (1-1 -2), (21- - 1) J
(12-1-), (21- -1), (1- -21)}
94
ON THE CTCLOTOMIC QUIXAET QUINTIC.
(2)
(X)
(X«)
(X)
(X«)
(X)
(X')
(V)
(x"*)
(X«)
to the set Xo are g:
{(•
{(1
{(•
{(•
{(2
P
1111), (
• -21). (• .
•22.), (21
1-3-), (1 ■
•11-), (1-
r s t
1111), (-1111), (4, . .
•3-1), (12- 1 •), (21 • .
•)}
1)}
{(li-1-), (
•1), (-2.
■2), (21.
•2), (.31
13), (112
2), (2.11-)}
1), (112- .))
•), (1- -21)}
•), (2-11.)}
(3)
p r
{(4 ), (-1111), (.
{(2.11 .), (12.1.), (.
{(21. .1), (.2- .2), (2
{(1-1.2), (.31. .), (1
{(12-1.), (21 . .1), (1
s t
nil). (-1111)}
• -13), (112. .)}
• 11 ■), (• •22-)}
• -21), (2- 11-)}
• -21), ( .3-1)}
{(21. .1), (112. .), (.1.3.), (1 .1.2)}
(4)-
{(•nil), (4-
{(1-1-2), (21
•), (• nil), (. 1
1), (112- .), (. 1
{(•2
{(2 1
{(••
{(!•
ven bv
■2), (2-11 .), (.
•1), (1 . .21), (•
m, (112. .), (2
21), (2-11 •), (1
22-), (21
3-1), (12
11-), (12
1 -2), (-3:
t
n)}
3-)}
•1)}
1-)}
• )>
(lllll)a:+(lllll)2/+ (5 . . . .)2+(lllll)^«=l,
(lllll)a:-f(lllll)2/+(lllll)3+(5- . . ■)io=l,
(5 • - • •)a;+ (11111)2/+ (11111)3+ (11111)«;= 1,
(lllll)a; + (5. . . •)y+(lllll)2+(lllll)M=l.J
Since here (111 11) = 0,
z — \ 1 5i = X = y = \o .
But X, y , z or w = constant is the equation of a 1-surface, hence these 4
l-surfaces are contained in every surface of set Xo. The l-surfaces so found
(when they exist), are all of the form x = const., or
y, z, w = const.;
where a;, y, z, w = 1 are 4 l-surfaces from among the set of 120, so the 30
ON THE CYCLOTOIIIC QUINARY QUIXTIC. 95
sets of 4 surfaces are such that the members of each set intersect in four fixed
l-surfaces parallel to four of the 120 set. The above set of 24, (j), r, s, t)
is typical of all 5 such sets, for if any 2 values in one set are equal, the cor-
responding values in any other set will be also; that is the internal structure
of all sets of 24 4-oids is the same. The above may be written in the tj"pical
form for all;
a=(.31- •);
b
= (1
• -21);
c= (2-11.);
d
= (1-1 -2);
e= (• -3-1);
s-
= (1
2-1-);
ff=(2 1-
•1);
h
= (• • -13);
k=r (112. .),
l
= (•
1-3-),
m— (4 • •
• );
n
= (-1111);
r=(. •22-);
s
= (•
2- -2);
then
()x+()y+nz
+ {)w=l.
{)x+{)y+{)z
+ ()w=l
(Xf)
a
b
c d
(Xi)
h e
f
g
(M)
d
a
b c
(XI)
g b
e
f
(xi)
c
d
a b
(Xt)
f g
b
e
(XI)
b
c
d a
(Xf)
e f
g
b
(X|)
c
f
k k
(Xi)
d g
k
I
(xt)
k
c
f h
(Xt)
I d
g
k
(M)
h
k
f
(Xf)
k I
d
g
(XE)
f
h
k c
(Xi)
g k
I
d
(Xa)
m
n
n n
(Xi)
r g
s
c
(X4)
n
m
n n
(X!)
c r
g
s
(Xi)
n
n
m n
(XI)
s c
r
g
(X2)
n
n
n m
(Xi)
g «
c
T
Since the effect of multiplj
ing ( • • ) by p is simply to permute the numbers
within cyclically,
p'^a= (3 1 •••);
P'e
= (3-i--);
pH= (3
••!•)
1
ph= (3.-.1),
hence
p'a
+ p3e +
p^l-\- ph —
8;
also
a
+ e + A + f= -4,
and
m = 4, n = — 1; r + 5= — 1,
and so in any case the number of coefficients in the 24 set may be greatly
reduced, but the symmetry is thus lost. The 5 different sets are distinguished
by suffixes, e. g., Tz means that in the above value of r, p' is to be taken as p',
so that ra = (• 21 • 2). This arrangement of the 120 4-surfaces shows
96 ON THE CYCLOTOMIC QUINARY QUINTIC.
clearly the distribution of the skew-lines in 5-space through which the 4-
surfaces pass. Also each set of 4 is circumscribed to a symmetrical figure of
16 1-surfaces, which surfaces are parallel in sets of 4 (reading the table by
coIuQins), the 4-surface passing through the lines parallel to one of the funda-
mental axes arid common to the respective sets of 4 non-parallel l-surfaces
(reading the table by rows). In short, all the descriptive properties of the
120-configuration are summed up in the table, and it becomes a mere matter
of reading the table to elicit further inter-connection, in the same way that
the whole 3,125-configuratIon and its consequences arose from the reading of
the polar-tangent equation.
In order to completely exhaust the properties of this simple figure, the
properties of" ruled surfaces " in Si have to be considered, and account taken
of the other geometric entities, sub-surfaces, . . . super-lines, etc. (according
to Cayley). The 120 configuration being in a sense complete in itself, the
farther points are not considered in detail in this place.
ON THE CYCLOTOMIC QUINABY QUINTIC. 97
VITA^
The undersigned is a native of Scotland, (born 1SS3); entered Stanfo*>
University, California, as an advanced student in 1902, graduated thence 1904
(A.B.); entered the University of Washington, 1907, as Teaching Fellow, and
received there the A.LI. degree 1908; entered Columbia University, NovembiT,
1912, as a student of mathematics.
E. T. Bell.
April 2, 191$