Collins Complete Woodworkers Manual

Order now at alibris.com

( 4.7/5.0 ★ | 163 downloads )
-- Click the link to download --

https://click.linksynergy.com/link?id=*C/UgjGtUZ8&offerid=1494105.26
539780007164424&type=15&murl=http%3A%2F%2Fwww.alibris.com%2Fsearch%2
Fbooks%2Fisbn%2F9780007164424
Collins Complete Woodworkers Manual

ISBN: 9780007164424
Category: Media > Books
File Fomat: PDF, EPUB, DOC...
File Details: 6.5 MB
Language: English
Website: alibris.com
Short description: Good Connecting readers with great books since
1972! Used textbooks may not include companion materials such as
access codes, etc. May have some wear or writing/highlighting. We ship
orders daily and Customer Service is our top priority!

DOWNLOAD: https://click.linksynergy.com/link?id=*C/UgjGtUZ8&
offerid=1494105.26539780007164424&type=15&murl=http%3A%2F%2F
www.alibris.com%2Fsearch%2Fbooks%2Fisbn%2F9780007164424
 Collins Complete
Woodworkers Manual

• Click the link: https://click.linksynergy.com/link?id=*C/UgjGtUZ8&offerid=1494105.2653978000716442

4&type=15&murl=http%3A%2F%2Fwww.alibris.com%2Fsearch%2Fbooks%2Fisbn%2F9780007164424 to do
latest version of Collins Complete Woodworkers Manual in multiple formats such as PDF, EPUB, and
more.
• Don’t miss the chance to explore our extensive collection of high-quality resources, books, and guides on
our website. Visit us regularly to stay updated with new titles and gain access to even more valuable
materials.
.
The text on this page is estimated to be only 24.78%
accurate

CHAP. I] BINARY QUADRATIC FORMS. 29 D=—l, ±2, ±3, 5,

after showing that there exists a g.c.d. process for algebraic
numbers a + bVD. G. Cantor84 proved, without the use of continued
fractions, the double theorem due under restrictions to Gopel.41 If D
is a prime p = 8n + 3 [or 8n+7] or its double, and if Z>=>2 + 2^2
[or
The text on this page is estimated to be only 24.81%
accurate

30 HlSTOEY OF THE THEORY OF CUMBERS. [CHAP. I

successive right neighboring forms) to one whose first root exceeds
2 and whose third coefficient is greater than or equal to the first
coefficient (apart from signs), and called such a form (and those in
the same period) reduced (although not in the sense of Gauss or
Hermite53). Using continued fractions with negative quotients, he
developed the first root of a reduced form F, the g.c.d. of whose
coefficients is 0. Denoting its factors by (|±£1)o;+ (rj + rj^y, we
have (20) e-g = a, bi-Srti = k, ^-r,l = b. Take any set of real
numbers £, £1? 77, ^ satisfying (20) and insert them in (19) ; we
obtain a positive form (A, K, B) corresponding to (a, k, b). Necessary
and sufficient conditions for corresponding forms are i. e., their
determinants are equal except as to sign, and their simultaneous
invariant vanishes. If the same substitution ( " $ ) be applied to an
indefinite form and to the corresponding positive form, the new
positive form corresponds to the new indefinite form. An indefinite
form (a, k,b) with a>0 is called reduced if the corresponding positive
form (A, E,B) is reduced whatever set of real solutions of (20) is
employed, with the restriction that K is zero for one of these sets.
The purpose of the last restriction and a>0 is to insure that only one
of the forms (a, k, b), (b, g, c), 91 Math. Annalen, 7, 1874, 382. 92
Jour, fiir Math., 77, 1874, 143-164; revision (in French) in Jour, de
Math., (3), 3, 1877, 21-42.
The text on this page is estimated to be only 24.34%
accurate

CHAP. I] BINARY QUADRATIC FORMS. 31 (cf Ti, a) shall be

reduced, where h, g, c are defined by a+h + k = Q, c + g + h = 0.
It is proved by use of Gauss' geometric interpretation that an
indefinite form is reduced if and only if its first coefficient is positive
and third coefficient is negative. Applying to a reduced form (a, k, b)
the substitutions (J J), (J J), we obtain (a, a+k, a + 2k + b), (a+2k
+ b, k + b, b), the first or second of which is reduced, according as
a +2k + 1 is negative or positive. Any reduced form determines in
this manner a finite period of reduced forms. A special reduced form
is one for which not only a and b, but also a— 2k + b and a+2k + b,
are of opposite signs; as noted by Lipschitz (Berichte Akad. Wiss.
Berlin, 1865, 184), these conditions are equivalent to those of Gauss
and Dirichlet57 for a reduced form. Selling compared his definition of
reduced forms with that of Hermite63, and considered at length the
case in which I = k2 — ab is a square. Cf. Voronoi42 of Ch. XI. E.
Hiibner93 noted that, of the conditions that ax2 + 2bxx1 + alxl
become cy for x=ay + fiyi, xi = aly + /3iyI) two determine c and Ci,
and the third is If a, "b, a^ have no common factor, all solutions of
the third condition in which a is prime to a17 and ft to ft, are found
by assigning relatively prime values to a, c^ and taking ft and — ft
as the quotients of aa + ba^ and 'ba + a1a1 by their g.c.d. If me
and me+k are representable properly by a form of determinant —Df
when m is prime to D} some power of m between me and m2e+k is
representable properly by (1, 0, D). If D=8n — 1>0, at least one
power 2k is representable properly by (1, 0, D) ; the least k is g 2H
— 1, where H is the number of classes of forms of determinant — D.
If D is not divisible by the pr.ime p, and if — D is a quadratic residue
of p, some power pk is representable properly by (1, 0, D) ; the
least k is g 2H—1. R. Gent94 proved that, if n>l is an odd integer,
the number of decompositions of 4n into x2 + 3y2, where x and y
are odd and positive, is the excess of the number of divisors of the
form 3/1 + 1 of n over the number of the form 3h + 2. He tabulated
all decompositions 4ip = x2 + 3y2 for the possible ^<500. The
difficulty with the determinant — 7 is that there are two reduced
forms ( 1, 0, 7) and (2, 1, 4). He conjectured that the number of
solutions of 8n = x2 + 7y2, #>0, y>0, is the excess of the number
of divisors of n which are quadratic residues of 7 over the number
which are quadratic non-residues of 7. H. J. S. Smith95 associated
with the form (a, b, c) of determinant N=b2 — ac>0 the semi-circle
, , , 2X A [a, b, c] : a+2bx+c(x2 + y2)=Q, #>0. Two points w and n
in the upper half H of the complex plane are called equivalent if 93
Ueber die Transformation einer homog. binaeren quad. Form in ein
Aggregat von 2 Quadraten, Progr., Memel, 1875. 94 Zur Zerlegung
der Zahlen in Quadrate, Progr., Liegnitz, 1877. 95Atti R. Accad.
Lincei, Mem. fis. mat., (3), 1, 1876-7, 136-49; abstr. in Transunti,
(3), I, 68-69; Coll. Math. Papers, II, 224-241.
The text on this page is estimated to be only 25.24%
accurate

32 HISTORY OF THE THEORY OF NUMBERS. [CHAP. I

where a. . . ., 8 are integers such that /? = y = 0 (mod 2), a = and
O equivalent numbers if they are connected by a relation (21) where
a, . . ., 8 are any integers. All complex numbers equivalent to a given
one are equivalent to each other and are said to form a class. Every
complex number o) = x+yi, y>Q, and every rational real number, is
equivalent to one (and in general to only one) number w0 = x0-\-y0i
for which x0 ^ J, x0 ^ — -J, zl + yl 5> 1, whence the point 0. It is a
sequel to the papers by Gopel41 and Smith.79 96 Post, fragment,
1827; Werke, III, 1876, 477-8 (386) ; VIII, 1900, 105 (remarks by
R. Fricke). 97 Jour, fur Math., 83, 1877, 269-273. Cf . Hurwitz,i°6
Mathews,137 Weber.145 Exposition by A. L. Baker, Amer. Math.
Monthly, 8. 1901, 163-6. 98Comptes Rendus Paris, 87, 1878, 399-
402; Assoc. frang., 7, 1878, 40-49. 99 Atti Accad. Pont. Nuovi Lincei.
32, 1878-9, 81-87. 100 Proc. London Math. Soc., 10, 1878-9, 29^1.
Partial report in Vol. II, p. 383.
The text on this page is estimated to be only 25.83%
accurate

CHAP. I] BINARY QUADRATIC FORMS. 33 A. Korkine and Gr.

Zolotareff101 stated that the precise limit of the minima of all binary
quadratic forms of positive determinant D for integral values, not all
zero, of x, y is V|A which is the minimum of f0=y%D(x2 — xy — y2)
and forms equivalent to it ; while the precise limit of the minima for
all other forms is Vl^. Proof by Humbert.183 A. Markoff102 stated
that Korkine had communicated to him the fact that VJ1> is the
minimum of forms equivalent to /x = V %D (x2 — 2xy — y2). Here
Markoff proved that VlOOZ)/221 is the precise limit of the minima of
all forms equivalent to neither /0 nor /±, and is the minimum of
forms equivalent to and obtained a continuation of this series down
to /9. By the use of continued fractions, he proved that if I is a given
number >f, there is only a finite number of classes of forms, of a
given determinant D, whose values are not numerically nf y = cm +
dn (m, n ranging over all integers). 101 Math. Annalen, 6, 1873,
369-370; Korkine's Coll. Papers, 1, 1911, 296. Report in Ch. XI." 102
/bid., 15, 1879, 381-406; 17, 1880, 379-399. 108Comptes Rendus
Paris, 89, 1879, 897-9. For details on related matters, see his
papers.104- 109> 157. 10*Jour. ecole polyt., t. 28, cah. 47, 1880,
177-245. Report of Parts IV, V is made in Ch. Ill .«
The text on this page is estimated to be only 26.28%
accurate

34 HISTORY OF THE THEORY OF NUMBERS. [CHAP. I Call

ad — be the norm of the lattice. A lattice A is called a multiple of a
lattice B if all points of A belong to B. Two lattices are equivalent if
all points of each belong to the other. If A' — AE under matrix
multiplication, lattice R is said to be the ratio of A' to A. Proof is
given of Eisenstein's theorem that every lattice with integral
elements is equivalent to one with d = Q. The system of points
common to two lattices is called their least common multiple. The
g.c.d. of A and A' is defined to be the system of points = am + bn +
a'm' + b'n', ra b a'&'~| \_c d c' d']: where ra, n, m', nf take all
integral values. If d=d' = Q, it is shown how to find the l.c.m. and
g.c.d., they being of the form [' Q]. A prime lattice is one whose
norm is a prime. There is a study of special lattices composed of the
points whose coordinates are integers satisfying a congruence
ax+by = Q (mod c). Part II (pp. 200-9) deals with the
representation of numbers a+b^/D by points. If D<0 it is usually
represented by the point m with the coordinates af &V— D in a
plane P. Take a plane Q cutting P along the z-axis and making with it
a dihedral angle equal to arc cos l/y—D. The coordinates of the
projection of m on Q are a, b. Whether D is positive or negative, the
point (x, y) is taken as the representative of a; + yV^ whose
modulus and argument are defined as yx2 — Dy2, ; arc tan If real,
the modulus is the ratio of the vector from the origin 0 to C= (x, y)
to the vector from 0 to the intersection of the former vector with the
ellipse or hyperbola I2 — Dr)2 = l. But if x2 — Dy2“<0 the present
representation differs from the classic one only in multiplying the
coordinates of points of the lattice by fixed factors. Henceforth let
D>0. By a fundamental triangle OAB is meant (p. 222) one formed
by the origin 0 and two points A and B of the lattice and containing
in its interior no other point of the lattice. It is called an ambiguous
triangle if the first "asymptote" ^/T)x — y is interior to the angle at
0 of the tri ”
The text on this page is estimated to be only 25.93%
accurate

CHAP. I] BINAEY QUADRATIC FORMS. 35 angle, while the

second asymptote — ~\/Dx=y is exterior to this angle. If we
complete the parallelogram OABC whose half is a fundamental
triangle OAB, the triangles OAC and OBC are aso fundamental and
are called derived triangles of OAB. Also, OA B is said to be the
primitive of OAC. Any lattice has ambiguous triangles. If a triangle is
ambiguous, one and only one of its two derived triangles is
ambiguous, and a single one of its two primitives is ambiguous.
Hence there exist infinitely many ambiguous triangles forming a
period such that each of them is the derived of the preceding
triangle and the primitive of the following one. Each triangle of the
period has a side in common with the following triangle. Thus in
general several consecutive triangles of the period have a common
side and are said to form a series, so that the period is divided into
series. The last triangle of a series is the first one of the next series ;
such a triangle belonging to two series corresponds to a reduced
form. The lattice (23) is designated by Am + Bn + A'm' + B'n',
where In particular, the lattice (22) is Am + Bn. Employing also the
conjugates of A, B, we see by the first sentence of the report on Part
III that*the quadratic form (in the variables m, n) (Am + Bn) (Am +
Bn) = (am + l>n)2-D(cm + dn)2 is represented by either of the
lattices Am+Bn, Am + Bn. F. Mertens105 gave an elementary proof,
without using continued fractions (as had Dirichlet57), of Gauss'
theorem (Art. 193) that two properly equivalent reduced forms of
positive determinant belong to the same period. Cf. Frobenius,175
Mertens.139' 19° A. Hurwitz106 developed the theory of
Dedekind's97 fundamental region and noted that it leads to a simple
geometrical theory of the reduction of quadratic forms. K. Kiipper107
investigated primes p = m2 + kn2. H. J. S. Smith108 proved, by
means of the functions (q» ..., qn) giving the numerators of the
continued fraction with the quotients g1? q2, ---- that every prime
12n + 7 is of the form x2 + 3y2, and every prime 12n + ll is of the
form 3x2-y2. By the same method he treated the representation of
primes by 2x2±y2. Cf. Gopel.41 H. Poincare109 obtained
arithmetical invariants of a linear or quadratic form F(xf y], i. e.,
functions of the coefficients of F which are unaltered under every
linear substitution X = ax' + Py', y = yx' + $y', where a, ft y, 8 are
integers such that a8 — Py — 1 . Consider «fr ( g) s 2 (g™ + ")"'*' _
105 Jour, fur Math., 89, 1880, 332-8. ™« Math. Annalen, 18, 1881,
528-540. 107 Casopis math, fys., Prag, 10, 1881, 10 (Bohemian).
108 Coll. Math, in memoriam D. Chelini, Milan, 1881, 117-143; Coll.
Math. Papers, II, 287-311. 109Assoc. frang. av. scf., 10 (Alger),
1881, 109-117. See Poincare.157
The text on this page is estimated to be only 24.24%
accurate

36 HISTORY OP THE THEORY OF NUMBERS. [CHAP. I

summed for all pairs of integers m, n except 0, 0. It can be
expressed as a definite integral and is holomorphic if q is not real.
Further, is an arithmetical invariant of ax+by. We may employ * 0,
let t, u be the least positive integers such that t2 — Du2 = l. Then F
has the invariant summed for all pairs of integers m, n for which
ra>0, n ^ 0, m/nl. This series may be expressed as a double
integral. The same is true of a series defining a certain arithmetical
invariant of a pair of linear forms. J. Hermes110 gave an algorithm
for computing 8 in passing from a form to a reduced right
neighboring form (Dirichlet57). Hermes111 considered a chain 19
...,>« of neighboring forms of positive determinant Df where >i_i is
transformed into i by (_J J ). Then (JJ|) replaces (/>0 by n if a=
db[/c2, . . .jKn-i], 0=.±[jc2, ..., jc»], y=:t[Ki, - .., Kn-J, 8= ± [KI? . .
., ic»], * Since the series for 0i is only semiconvergent, further
definition of 0i is needed. It is essentially log A, where A is the
modular invariant of Klein-Fricke, Theorie der Elliptischen
Modulfunctionen, 1890. The limit for k = 1 of the series for
The text on this page is estimated to be only 24.28%
accurate

CHAP. I] BINARY QUADRATIC FORMS. 37 in the notation of

continued fractions (Gauss, Art. 27), where K2ui = 82i+i, *2i=
-821Let $0, of determinant D, be transformed into the same n by
using A1? . . . , ATO instead of AC'S. If 00 is transformed into 0 by (
£ J ) , then r/A= (KI? . . ., *„.!, 0, qpAm-i, . . ., H=Ai), where the
upper or lower signs hold according as the total number of elements
of this continued fraction is odd or even. Hence A= ± |>2, . . ., TAJ,
B = ± [*,, . . ., ^Xi, fe], r= ± [KI? . . ., TAJ, A= ± [K,, . . ., TAt, fc],
Two forms which lead to different periods are not equivalent. T. J.
Stieltjes112 noted that if F(n) is the number of representations of n
by x2 + 2yz, and if dt is the number of divisors of the form Sk + i of
n, then (Genocchi56) The sum in brackets is transformed into 8 + Sl-
\(\)9 where 8 is a like sum ending with ±[n/(2A-l)], while A = [ J (
V'8n +T+ 1] , (z) = 2 sin2 L. Kronecker113 noted that every
substitution with integral coefficients of determinant unity is the
product of one of the following six T I1 °\ (0~l\ I l l\ /-1 1\ /°-A / ! °
\ HOI/' (i ;o/' U-io/' V o-i/' (i i/' (-1 ij by a substitution congruent to
the identity I modulo 2. By interchanging the two columns we get six
substitutions of determinant — 1 to one of which any substitution of
determinant —1 is congruent modulo 2. Two forms are called
completely (or incompletely) equivalent if they can (or cannot) be
transformed into each other by a substitution =1 (mod 2). Thus to
any form of negative determinant belong five properly, but
incompletely, equivalent forms and six improperly and incompletely
equivalent forms derived by the above substitutions. Corresponding
to a form of negative determinant, there may not exist one
completely equivalent to it and satisfying the conditions c2^a2^
(2&)2 for a Lagrange reduced form; however, to every form
corresponds a completely equivalent form whose coefficients a0, ...
satisfy the conditions al !> 60, cl §&<>. such a form being here
called a reduced form. Two such reduced positive forms (i. e., with
outer coefficients positive and not smaller than the absolute value of
the middle coefficient) can be completely equivalent only if they both
occur in the above set of 12 forms. All the forms of the same
negative determinant which are completely equivalent are said to
constitute a class. The number of classes equals the number of
reduced forms (a0, 60, c0) if we retain only one of two reduced
forms (a0, ±a0, C0) and only one of (a0, ±c0, c0). 112 Comptes
Rendus Paris, 97, 1883, 891. 113 Abh. Akad. Wiss. Berlin, 1883, II,
No. 2; Werke, II, 433-444.
The text on this page is estimated to be only 24.91%
accurate

38 HISTORY OF THE THEORY OF NUMBERS. [CHAP. I E.

Cesaro114 concluded that the number of positive integral solutions
of Ax2 + Bxy + Cy2 = n(A>Q,C>0) is in mean ir/(2B) -B/S2, where
82 = 4:AC-B2. According to Encyclopedic des sc. math., t. I, vol. 3,
p. 350., there was an error in the limits of integration, and the
correct value is 7T 1 B ^ — «- arc tan -s- . /CO O O Note, however,
that Gegenbauer123 agrees with Cesaro. T. Pepin116 proved that
every prime m = Sl + 3 is of the form x2 + 2y2 by use of the sum of
the divisors of m and the sum of the odd divisors of each m — n2.
Pepin116 applied general theorems of Dirichlet42 (§ VII) on the
number of representations by a quadratic form of negative
determinant D to show that the number of solutions of 2am = xz +
2y2 is double the excess of the number of divisors of the forms 8/ +
1 or 8Z + 3 of m over the number of divisors of the form 8Z + 5 or
8Z + 7. If m is odd and not divisible by 3, the number of
representations of 3^ra by x2-\-3y2 is 2cu (ra, 3), where 0) by 2x2
+ 2xy + 2y2 or by x2 + 3y2 is 6w (m, 3). The number of
representations of 2a3^m by that one of the forms (1, 0, 6), (2, 0,
3), which is suitable to it, is the double of 2(-6/i). L. Gegenbauer117
proved that the number of ways any integer r can be represented by
a system of binary quadratic forms of discriminant A is the product
of the number of linear automorphs of a form of discriminant A by
the sum of the numbers of solutions of z2 = A (mod ds), where d2
ranges over those divisors of r whose complementary divisors are
squares. The number of ways an odd integer r can be represented
by x2 + 2y2 is double the number of decompositions into two
relatively prime factors of those divisors of r which have only prime
factors 8s + 1 or 85 + 3 and whose complementary factors are
squares. Similarly for x2 + 3y2. The number of representations, by
quadratic forms of discriminant A, of those divisors of a number,
whose complementary divisor is the product of an even number of
primes, exceeds the number of representations of the remaining
divisors by the product of the number of automorphs of a form of
discriminant A by the excess of the number of those divisors d2,
with complementary square divisors, for which Jacobi-Legendre's
symbol (A/2) is +1, over the number of divisors d2 for which (A/d2)
= — 1. There are corollaries for x2 + 2y2 and x2 + 3y2 like those
for x2 + y2 quoted in this History, Vol. II, pp. 247-8. A. E. Pellet118
noted that if A= -1, ±2, ±3, 5, -7, -11, 13, so that factorization in
the field defined by V A is unique, every divisor of m2 — An2 or of
m2 + ran + nz(l — A)/4:, according as A = —1 or +1 (mod 4), can
be given the same form apart from sign. 114 Mem. Soc. R. Sc. de
Liege, (2), 10, 1883, No. 6, 197-9. 115 Atti Accad. Pont. Nuovi
Lincei, 37, 1883-4, 42. 116 Ibid., 38, 1884-5, 163-170. Cf . Pepin.isi
i" Sitzungsber. Akad. Wiss. Wien (Math.), 90, II. 1884, 437-448. 118
Comptes Rendus Paris, 98, 1884, 1482.
The text on this page is estimated to be only 25.00%
accurate

CHAP. I] BINARY QUADRATIC FORMS. 39 F. Cajori119

noted that, if 2b is divisible by neither a nor c, and if (a, b} c) is of
negative determinant -A and is transformed into (a', 0, c') by (Jj),
Ap-/xv = ±1, then aA/x + &(Ap + /xv)+cvp = 0. The quadratic in ^
obtained by eliminating A has real roots if a2 g: 4v2p2A. The
quadratic in p obtained by eliminating v has real roots if c2 ^
4/*2A2A. Each coefficient of the substitution is numerically > 0. J. C.
Fields120 considered the preceding question when Ap — /xv=-f-l,
and proved easily that there must be two numbers a and y whose
product is —A such that ap2 — yv2 = aj oA + &v=0 (mod a) for
integers v, p. If integers v, p can be found, the two forms are
equivalent. He stated similar conditions for the equivalence of (a, I,
c) to (a', &', a') or to (a', Ja', c'). L. Kronecker121 proved that the
number of representations of n by a quadratic form of negative
determinant D is in mean 7r/V— D. J. W. L. Glaisher122 proved by
means of products of infinite series that the excess of the number of
representations of 24ri + 3 by x2 + 2y2, in which xf y are both of
the form 12m ±1 or both of the form 12m ±5, over the number of
representations in which x is of one form and y of the other is equal
to double the excess of the number of representations of 871 + 1 by
x2 + 2y2, in which x is odd and y is a multiple of 4, over the number
of representations in which x is odd and y = 2 (mod 4) ; either
excess may be positive or negative. L. Gegenbauer123 gave a simple
proof of E. CesaroV14 theorem that, if a>0, & ^ 0, c>0, A = 4ac —
&2>0, the mean number of those representations of an integer by
(^ — &, c), of negative discriminant —A and with like sign for the
values of the two variables, is 7T/VA + 2&/A. He found expressions
for the mean number of divisors of x and of y in the various
representations by ax2 + by2 of an integer in a given interval or of
an integer with s digits, also when the divisors are of specified
character, such as divisible by A, or divisible by no o-th power.
Gegenbauer124 found, for the various representations of an integer
with 5 digits by ax2 + cy2, where a>0, c>0, the mean number of
divisors of x such that the divisors are =a (mod A) and have
complementary divisors =/? (mod /x), and also for other conditions
on the divisors. J. Vivanti125 proved that a form (a, b, —c) of
positive determinant D, not a square, is reduced if a and b are
positive and c = a+b, and called it a Null form. A Null form occurs in
the system of reduced forms of determinant D if and only if the
exponents of the prime factors of the form 6n+ 5 and 2 of D are all
even. If (a, b, —c) is a Null form, also (b, a, —c) is a Null form; they
are improperly equivalent by means of x = x', y — xf — y'\ they are
properly equivalent if and only if there exist integral solutions of
x2D/d2 — 3 = y2, where d is the g.c.d. of a, b. He126 gave obvious
theorems on these Null forms. The conditions for integral 119 Johns
Hopkins Univ. Circular, 4, 1885, 122 (in full). 120 Ibid., 5, 1885, 38.
121 Sitzungsber. Akad. Berlin, 1885, 775. Cf. H. Brix, Monatshefte
Math. Phya., 21, 1910, 309-325 (p. 325) ; Dedekind, Jour, fur Math.,
121, 1900, 115. Cf. Landau.148 122 Quar. Jour. Math., 20, 1885, 96.
123 Sitzungsber. Akad. Wiss. Wien (Math.), 92, II, 1885, 380-409.
124 Ibid., 93, II, 1886, 90-105. 125 Zeitschrift Math. Phys., 31,
1886, 273-282. 126 Ibid., 32, 1887, 287-300.
The text on this page is estimated to be only 26.12%
accurate

40 HISTORY OF THE THEORY OF NUMBERS. [CHAP. I

solutions of x2D — 3 = y2 are not known. In practice, it is solved by
means of Pell's equation and the theory of binary forms. For 1
obtained from the various Pi and all sets of integers a, /?, y, 8 for
which the determinant is Pi, form a group H. In fact, the product of
(26) by the analogous substitution with the parameters a', /3', y', 8',
Pj is such that the four coefficients are divisible by the product Ph of
the modules common to Pi and Pj, and the resulting substitution is
of the form (26) with Pi replaced by Pk=PiPj/Pl. Such a group H has
a finite number of generators. Consider another such group 0 with
Qi in place of Pi. In order that G be a subgroup of H it is necessary
and sufficient that each module of 0 be a product of modules of H
(each entering only once as a factor) and that the product of all the
modules of 0 be equal to that for H. Then G is an invariant subgroup
of H. For example, the group defined by the modules 2, 3, 5 has as
invariant subgroups those defined by the modules 2, 15; 6, 5; 3, 10;
1, 30. If a, b, c are integers such that a, ct bPi have no common
factor, axz + bPiXy + cLy2 is called a form attached to the group G.
This form is called equivalent to the form a'xl -f b'PiX^i + c'L$ (in
which a', b', c' are integers) obtained from it by applying the
corresponding homogeneous substitution x= of determinant unity.
All forms equivalent to a given one are said to form a class. There is
given a process to select one or more representatives of each class
and to test their equivalence. Since that process is laborious, it is
desirable to enumerate the classes without knowing their
representatives. Under certain restrictions the class number is found
by Dirichlet's classic method (Dirichlet19 of Ch. VI). There is
suggested a generalization to several sets of modules and also to
substitutions with irrational coefficients. A. Hurwitz129 called a pair
of numbers x, y reduced if the point (x, y) lies in the region E
composed of two infinite strips, the first bounded by the lines x = 2,
y — r, y — r—\^ and the second by x— —2, y— —r, y — \ — r}
where r = %(3 — V5), the first strip extending to the right of x =2,
and the second to the left of x— —2, the points 127 Elemente der
Zahlentheorie, 1887, 237-374. "« Annales Fac. Sc. Toulouse, 3, 1889,
B. 1-28. 329 Acta Math., 12, 1889, 397-401.
The text on this page is estimated to be only 26.14%
accurate

CHAP. I] BINARY QUADRATIC FORMS. 41 (3 — rf r) and (

— 3 + r, — r) of the boundary being the only ones counted as
belonging to R. A form (a, ~b, c) of positive determinant, not a
square, is called a reduced form if its roots form a reduced pair of
irrational numbers. Every form is equivalent to a reduced form. Two
reduced forms of determinant D are equivalent or not according as
they belong to the same period or not. While these theorems had
been proved by use of the ordinary development into continued
fractions, they are here proved by use of a development of any real
number x0 into a continued fraction by means of x0 = a° — 1/Xi, x1
= di — l/x-2, ..., where an is an integer chosen so that Xn — dn lies
between — 1/2 and + 1/2. *T. Pepin130 discussed the number of
representations by (1, 0, 8) and (1, 0, 16). T. Pepin131 derived
his116 former results from formulas of Liouville (this History, Vol. II,
Ch. XI). He added a theorem on the number of representations by a
form of determinant — 12. F. Klein132 called a form ax* + 2bxy +
cy2, of negative determinant D=b2—ac, reduced if its root ( — 6+tV
— D}/Q> is a reduced number in the sense of Dedekind97 (i. e.,
represents a point lying in the region R defined by him). A form of
positive determinant D is called reduced if R is crossed by the semi-
circle whose diameter joins the points of the z-axis representing the
real roots (values of x/y) of the form. For the second case D>0, this
representation of the form by a semi-circle goes back to Smith.95
but with now a simpler definition of equivalent points and a simpler
fundamental region. L. K. Lachtine133 called a form of positive
determinant reduced if its positive root is developable into a pure
continued fraction with only positive elements, and gave a method
of finding all automorphs of a reduced form without use of a Pell
equation. P. V. Prebrazenskij134 showed that Lachtine's method
differs from Dirichlet's in that Lachtine does not exclude improper
equivalence. Prebrazenskij135 treated together forms of positive and
negative determinants and divided the classes of forms of any
determinant D into two types. Those of the first type (vollkommen)
are such that if a prime is represent able by such a form, it can be
represented by the form in four ways if D“>0. For the second type a
prime can be represented in only two ways or in two infinitudes of
ways, respectively. For D= — 11, there are two classes of the first
type and one of the second. He136 later extended this method to
complex variables. For D negative, imaginary periods occur. The Pell
equation can be replaced by an addition theorem. G. B. Mathews137
developed DedekindV7 theory of reduction of complex numbers and
its application to the geometrical theory of quadratic forms. He
discussed Poincare's104 use of lattices (nets) in the geometrical
reduction of indefinite forms, employing " hyperbolic complex
quantities " x+yj, where J2 = l, 1 •;=;•! = ;'. 130 Memorie Pont.
Accad. Nuovi Lincei, 5, 1889, 131-151. "i Jour, de Math., (4), 6,
1890, 8-11. 132 Klein-Fricke, Elliptischen Modulf unction en, I. 1890,
243-260; II, 1892, 160-9. Summary in Encyclopedic des sc. math., 1.
1, vol. 3, 116-9. 138 Math. Soc. Moscow, 14, 1890, 487-526; 15,
1891, 573 (Russian). ™*Ibid., 15, 1891, 118-121. 135 Bull Imper.
Soc. Univ. Moscow, 65, 1890, 62-83 (Russian). 130 X. Vers. Russ.
Nat., 331 (Jahrbuch Fortschritte der Math., 1898, 176). 137 Theory
of Numbers, 1892, 103-131.”
The text on this page is estimated to be only 25.89%
accurate

42 HISTORY OF THE THEORY OF NUMBERS. [CHAP. I K.

Th. Vahlen138 proved that every prime Sn + I or 8n + 3 is of the
form and every prime Grz + l is of the form a2 + 3&2 by using the
fact that the number of representations of an odd integer s as a sum
of 4 squares is 8 times the sum of the divisors of s, account being
taken of permutations and signs of the roots of the squares. F.
Mertens139 gave a simple proof of the chief theorem on reduced
forms of positive determinant : two reduced forms are equivalent if
and only if they belong to the same period. Use is made of the fact
that if (a, 6, —c) and (a', &', — c') have positive first coefficients and
negative third coefficients and the first form is transformed into the
second by (" f ), then a8>0. M. Lerch140 found series which give
arithmetical invariants of vm + rn and F=am2 + 2'bmn + cnz, the
latter being a positive form of negative determinant tf-ac= -A. In
particular (cf. Hurwitz143 of Ch. VI), m, n satisfies the reciprocity
relation r(s) A. Hurwitz141 gave a preliminary account of his142
theory. Hurwitz142 gave a geometrical theory of reduction of forms,
not necessarily with integral coefficients, of positive or negative
determinants. The new principle consists in first investigating the
degenerate forms and then applying the conclusions to the general
form. A form /= (a, ~b, c) of vanishing determinant D corresponds
to a point A with the homogeneous coordinates a:b:c=l: — A: A2 on
the conic D = Q, which by choice of the coordinate system may be
taken as a circle K such that the points, 0, 1, co appear at the
vertices of an inscribed regular triangle. If r, s, u, v are integers for
which rv — su= ±1, the line joining the points r/u and s/v is called
an elementary chord of K. A triangle inscribed in K is called
elementary if its three sides are elementary chords. Consider all
rational numbers r/u whose numerators and denominators taken
positively do not exceed the positive integer n; the corresponding
points r/u of K are the vertices of a convex polygon Pn called the nth
Farey polygon, the parameters of whose successive vertices form the
nth Farey series (this History, Vol. I, pp. 155-8). Thus each side of
Pn is an elementary chord and conversely. Now each elementary
chord is a side of just two elementary triangles which lie on opposite
sides of the chord. The elementary triangles which can be formed
from the vertices of Pn cover it fully without overlapping. As n
increases indefinitely, Pn approaches K. Hence all the elementary
triangles cover the interior of K without overlapping. «8 Jour, fur
Math, 112, 1893, 32-33. 139 Sitzungsber. Akad. Wiss. Wien (Math.),
103, Ila, 1894, 995-1004. 140Rozpravy Akad. Fr. Josefa, Prag, 2,
1893, No. 4 (in Bohemian). Fortschritte Math., 189&-4, 790-1. 141
Math. Papers Chicago Congress of 1893, 1896, 125-132. French
transl. in Nouv. Ann. Math., (3), 16, 1897,491-501. "2 Math.
Annalen, 45, 1894, 85-117. Exposition by Klein"* (pp. 173-266).
The text on this page is estimated to be only 25.82%
accurate

CHAP. I] BINARY QUADRATIC FORMS. 43 Under a linear

transformation with integral coefficients S: x = ax' + py', y = 7x' +
Sy', o8-£y = l, the point A is transformed into a point A' such that
(27) *= By computing the points into which are transformed the end
points r/u and s/v of an elementary chord, we see that the latter is
transformed into an elementary chord. If p, q, r are the vertices of
any elementary triangle, then being three successive terms of a
Farey series. By (27), S replaces this triangle by the elementary
triangle T with the vertices 0, 1, oo, which has exactly three
automorphs (55). (!:!)• (i-S)Every quadratic form of negative
determinant is represented by a point inside the circle K and will be
called reduced if that point lies inside or on the boundary of T. The
representative point certainly is inside or on some one elementary
triangle; hence each of the three transformations which carry this
triangle to T will replace the form by a reduced form. The reduced
forms whose representative points lie inside (and not on the
boundary of) T therefore occur in triples of equivalent forms, viz., by
(28), (a, I, c), (c, -l-c, a + 2& + c), (a+2& + c, -a-6, a). Two such
reduced forms are equivalent only if they belong to the same triple.
The reduced forms whose representative points lie on the sides of T
fall into sets of six equivalent forms. Next, every f=(a, &, c) of
positive determinant is represented by a point outside the circle K, or
preferably by its polar az — 2by+cx=Q with respect to K. The
parameters of its intersections with K are the roots Call a form
reduced if its first root A! is positive and its second root A, is
negative (and hence if a>0, c<0). Any elementary chord which
intersects the line AXA2 can be transformed into Ooo by two
transformations, one of which evidently carries / into a reduced
form. Hence every / is equivalent to a reduced form. Let A be any
elementary triangle crossed by the line AiA2, and a, o-' the sides met
by AtA2 such that a point travelling from A2 to Aj along A^ crosses a
on entering A and o-' on leaving it. Call o-' the right neighboring
chord to
The text on this page is estimated to be only 26.43%
accurate

44 HISTORY OF THE THEORY OF NU:MJU;I;S. [CHAP. I

corresponds a unique transformation Si which replaces / by a
reduced form
The text on this page is estimated to be only 24.13%
accurate

CHAP. 1] BINARY QUADRATIC FOR:MS. 45 zero for which

f(x, y}^ n. has for largo values of n the asymptotic expression
iL.+W-l«44-.I»^L where e is of the order of magnitude of 1/Vn, C is
Euler's constant 0.57721..., a and —/3 are the roots of f(xt l) =0,
and 6i(xf0, &>0, we employ a parallelogram three of whose vertices
have the preceding coordinates with V —D replaced by VZJ. The
form now represents the hyperbolic distance V(z — o^)2 — (y —
y*Y between points (x, y) and (xly y^) of the lattice, one being the
origin. He also represented (p. 177) the form by the point150 with
the homogeneous coordinates a, I, c. H. Minkowski151 gave another
geometrical theory of indefinite forms by use of a chain of
parallelograms representing a chain of substitutions (J £). C.
Cellerier152 noted that the problem to represent p = 5n±l by x2 —
xy — y2 reduces, by the substitution t — 2x — y, n — y, to the
solution of If one solution a, ft of the latter is known, all solutions
are found by when /,, g range over all solutions of f~ — og2 = 4:.
Hence from / = 3, g— — 1, we get a second solution having u= (3(3
— a)/2. By repetitions of the process, we must reach a solution with
a>5/?. Hence there exists a solution with and only one such
solution. The work of finding it is abbreviated by noting that , and
(a2-4p)/5 is a square 148 Jour, fur Math., 125, 1903, 165, seq. Of.
M. Lorch, Archiv Math. Phys., (3), 6? 1903, 85-94. 149 Ausgewahlte
Kapitel der Zahlenthcorie, 1, 1896, 70. Summary in Math. Annalen,
48, 1897, 562-588. 150 Cf. R. Fricke and F. Klein, Automorphen
Funktionen, 1, 1897, 491; Hurwitz.14;J «i Geometrie der Zahlen, 1,
1896, 164 (196). 152 Mem. Soc. Physique Hist. Nat. de Geneve, 32,
1894-7, No. 7 (end).
The text on this page is estimated to be only 24.92%
accurate

46 HISTORY OF THE THEORY OF NUMBERS. [CHAP. I E. de

Jonquieres,153 by a slight modification of Gauss (Art. 162), showed
how to obtain solutions of t2 — Du2= — m2 if we are given a
transformation of F = (A, B, C) into (a. 1), c) and a transformation,
whose determinant has the same sign as the former, of F into (—a,
b, —c). Such a pair of transformations exist when D is a prime 4k +
1. A. Thue154 proved that every prime divisor >^of x2 + 2y2 is of
that form, and similarly for x2 + 3y2. A. Cunningham155 easily
proved that 22- 7- 13 -19 -31 is the least N for which x2 + 3y2 = N
has 24 sets of positive integral solutions. T. Pepin156 supplemented
the investigation by Gauss (Arts. 213-4) of a form f= (a, b, c) of
determinant D transformable into F= (A, B, C) of determinant De2
by a proper substitution of determinant e, by showing that every
form of determinant De2 is obtained by transformation from some
form of determinant D. H. Poincare157 called a uniform function
$(0,, b) an arithmetical invariant of ax + by if it is unaltered by every
linear substitution with integral coefficients of determinant unity. An
example is summed for all pairs of integers m, n except 0. 0. If, as in
this example, (a, b) is homogeneous of order — 2k in a, b, and if k
is an integer >1, then (a, 1) is a thetafuchsian function
corresponding to the modular group. Let H(x, y) be a rational
function, homogeneous of order — 2k in x, y. Then, if the
summations extend over all sets of integers a, /?, y, 8 for which aS
— /?y = l, is a thetafuchsian series and ^H(aa + ^b, ya + 8b) is an
arithmetical invariant, under specified conditions for convergence.
Consider a rational function H (z, z') and the series summed for all
sets of integers a, /?, y. 8 for which 08 — /ty = l, and under
specified conditions for convergence (including the fact that k is an
integer >1). Write Then (29) 3H( is obviously an arithmetical
invariant of the two forms I = ax + by, l' = a'x + b'y. It will be an
invariant of the quadratic form IV if H(z, zf) is symmetric in z, z'. As
a 153 Comptes Rendus Paris, 127, 1898, 596-601, 694-700. 154
Forh. Vid. Selsk. Kristiania, 1902, No. 7, 21 pp. (Norwegian). 155
Math. Quest. Educ. Times, 3, 1903, 28-29. 156 Jour, de Math., (6),
1, 1905, 333-346. 157 Jour, fur Math., 129, 1905, 89-150. For
corrections and additions, Annales fac. sc. Toulouse, (3), 3, 1911,
125-149. Cf. Poincare.103 His related papers in Acta Math., 1, 3, 5,
1882-4, and other journals, are reprinted in his Oeuvres, II.
The text on this page is estimated to be only 23.97%
accurate

CHAP. I] BINARY QUADRATIC FORMS. 47 generalization, let

H(a, ~b, a', &') be any rational function, homogeneous of degree
-4fc in a, I, of, V '; then the series (29), when convergent, is an
invariant of I, V. He amplified his109 earlier investigation of
invariants of an indefinite form F = am2 + 2bmn + cn2 of positive
determinant and studied not only ^F~8 but also the series 2 ±
qFxmyn. 0. Spiess158 proved that if (A, B, C), (A', B', C") have the
determinants D, D' and then for all values of alf a2 we can
determine ft, ft such that identically in t. In order that there shall
hold at the same time the similar identity with A and A', . . ., C and
Cf interchanged, it is necessary and sufficient that D — D' or D=
±D', according as A + 2B + C is or is not zero. Th. Pepin159 gave an
exposition of the classic theory of binary quadratic forms. J.
Sommer160 applied algebraic numbers to prove that every prime =
1 or 3 (mod 8) is representable by x2 + 2y2, #>0, i/>0, in one and
but one way, likewise a prime =1 (mod 3) by x2 + 3y2, while x2 —
2y2 — 8n± 1 = prime has an infinitude of solutions. A. Aubry160a
gave a summary on quadratic forms without exact references. H.
Minkowski160b proved that if f = ax2 + 2l>xy + cy2 is a positive
form with B = ac — 62>0, &>0, we can assign integral values not
both zero to x and y such that The equality sign holds only when / is
equivalent to \/S(x? + xy + y2). He gave a simple geometrical
interpretation by means of the thickest packing of circles. L. E.
Dickson161 obtained necessary and sufficient conditions that two
pairs of binary quadratic forms with coefficients in any field (or
domain of rationality) F shall be equivalent under linear
transformation with coefficients in F. D. A. Grave162 let (^) and
(*',|J) be substitutions of determinants e and e' which transform (A,
B, C) into (a, I, c) and (a±, 61? cj, and transform (A, B~C) into (a,
1), c} and (a19 b19 ct), respectively. Define a', c', W by (13) of
Gauss. Let a', ?, 26' denote the analogous functions of ~A,~Bf C
with y and ft, y' and p', interchanged. Then 158 Archiv Math. Phys.,
(3), 9, 1905, 340-4. 159Memorie Pontif. Accad. Romana Nuovi
Lincei, 24, 1906, 243-288; 25, 1907, 83-107 (reduction for negative
determinant, table of linear forms of divisors of x* -\- A for A = l, ...,
31); 27, 1909, 309-351 (positive determinant, Pell equation); 28,
1910, 307-348 (periods of reduced forms, equivalence of primitive
forms of positive determinant) ; 29, 1911, 319-339 (distribution of
classes into genera; see report under quadratic reciprocity law). 160
Vorlesungen liber Zahlentheorie, 1907, 125-7; French transl. by A.
Levy, Paris, 1911, 132-4. i6°a L'enseignement math., 9, 1907, 289-
294, 431-2, 436, 442-3. ieob Diophantische Approximationen, 1907,
55. 161 Amer. Jour. Math., 31, 1909, 103-8. 162 Comptes Rendus
Paris, 149, 1909, 770-2. His a,, . . ., 81 are here replaced by a', ....
5' to fit Gauss' notations.
The text on this page is estimated to be only 22.60%
accurate

48 HISTORY OF THE THEORY OF NUMBERS. [CHAP. I from

which we obtain Q' and Qa by replacing a, . . . , c by a', . . ., c' and
by al9 . . . , cl9 respectively. We have the identity 0" - COi = DN* +
I)N2 where D = B*-AC, D=B2-AC, while N is the same function of A,
E, C with f3 and y, /?' and y' interchanged. By means of this identity
we readily obtain all the details of Gauss (Arts. 162-6). For, if at = af
etc., we have e\ = ez. If e = e^ then JY2-Z>A = 0 and O'2 - OOj =
IW2. If e= — fij, we conclude that ambiguous forms exist. F. de
Helguero163 made an elementary study, without the theory of
quadratic forms, of the numbers represent able properly by x2 + xy
+ y2 = •{ x, y \. If a = \x^ , y , \ and b = -{x.,, y«\, then ab = \a, -
f3\ = \a, y\=\fi, -y\, where These and the three representations of
ab derived by interchanging x2 and y* are distinct in general. These
six representations are proper if a is prime to b. If a is a prime,
every proper representation of ab is obtained from one proper
representation of a and all proper representations of b by means of
these product formulas. A prime p is representable by \x, y\ if and
only if p = o// + 1 or p = 3. Every product of powers of n distinct
primes 3/& + 1, or 3 times such a product, has exactly 3 • 2""1
different proper representations by x2 + xy + y2. J. V. Uspenskij1"4
applied an algorithm closely analogous to ordinary continued
fractions to the reduction of indefinite binary quadratic forms and
obtained the periods of reduced forms more rapidly than had (Jauss.
G. Fontene165 noted that every prime factor =^=3 of a number of
the form x* + xy + y2 is of the form 3fc + l. H. C. Pocklington166
gave an elementary discussion of the form of the prime divisors of
mx*+ny* and of any divisor of ax2 + 2bxy + cy2. L. Aubry167
proved that every divisor of xz + ky* (x prime to y) whicli is not
<2VWfor fc>0 or not < Vfc [meaning V[F|] for fc<0 is of the form
/= (vz + lcuz)/d, where d ^ 2 V^/3 for Je>0 or d ^ Vfcfor fc<0. He
readily transformed / into a reduced form. Use is made of the lemma
that X2 + lcY2 = DE, Xq-uE=±pY} Xp-vE=+TcqY} E=~ r (p* + kf)
imply D=~ (w2 + fcw2), Xu-qD=+vY, Xv-pD=±kuY. As corollaries,