<html><head> <link rel="stylesheet" href="/stylesheets/archive.

css" type="text/css"/> <link rel="SHORTCUT ICON" href="/images/logo-16.jpg"/> <title>Full text of "Higher Algebra"</title> </head> <body class="Texts"> <!--BEGIN HEADER 1--> <table style="background-color:white " cellspacing="0" width="100%" border="0" c ellpadding="0"> <tbody> <tr> <td style="width:100px; height:72px; vertical-align:top; background-color: #000000;"> <a href="/"><img style="border:0;" alt="(logo)" src="/images/logo.jpg" w idth="84" height="70"/></a> </td> <td valign="bottom" style="background-image:url(/images/blendbar.jpg);"> <table width="100%" border="0" cellpadding="5"> <tr> <td class="level1Header"> <a <a <a <a <a <a <a <a href="/web/web.php">Web</a> | href="/details/movies">Moving Images</a> | class="headerAt" href="/details/texts">Texts</a> | href="/details/audio">Audio</a> | href="/details/software">Software</a> | href="/details/education">Education</a> | href="/account/login.php">Patron Info</a> | href="/about/about.php">About IA</a> </td> </tr> </table> </td> <td style="width:80px; height:72px; vertical-align:top; text-align: right" > <a href="/details//stream/higheralgebra032813mbp/higheralgebra032813mbp_ djvu.txt"><img alt="(navigation image)" src="/images/books-small.jpg" style="mar gin:0; border:0; vertical-align:top;" /></a> </td> </tr> </tbody> </table> <!--BEGIN HEADER 2--> <table width="100%" class="level2Header"> <tbody> <tr> <td align="left" valign="top" class="level2HeaderLeft"> <a class="level2Header" href="/">Home</a> </td> <td style="width:100%;" class="level2Header"> <a href="/details/americana">American Libraries</a> | <a href="/details/toronto">Canadian Libraries</a> | <a href="/details/opensource">Open Source Books</a> | <a href="/details/gutenberg">Project Gutenberg</a> | <a href="/details/biodiversity">Biodiversity Heritage Library</a> | <a href="/details/iacl">Children's Library</a> | <a href="/details/additional_collections">Additional Collections</a> </td> </tr>

</tbody> </table> <!--BEGIN HEADER 3--> <table cellspacing="0" width="100%" border="0" cellpadding="0"><tbody> <tr> <td class="level3HeaderColorBar" colspan="4"></td> </tr> <tr> <td class="level3Header level3HeaderLeft"> <form style="margin:0;padding:0;" action="/searchresults.php" id="search form" method="post"><p style="margin:0;padding:0;"> <b>Search:</b> <input tabindex="1" size="25" name="search" value="" style="font-size: 8pt"/> <select tabindex="2" style="PADDING-TOP: 2px; font-size: 8pt;" name= "mediatype"> <option value="all">All Media Types</option> <option value="web">&nbsp;&nbsp;Wayback Machine</option> <option value="movies">&nbsp;&nbsp;Moving Images</option> <option value="movies.animationandcartoons" >&nbsp;&nbsp;&nbsp;&nbsp;Animation & Cartoons</option> <option value="movies.artsandmusicvideos" >&nbsp;&nbsp;&nbsp;&nbsp;Arts & Music </option> <option value="movies.computersandtechvideos" >&nbsp;&nbsp;&nbsp;&nbsp;Computer s & Technology</option> <option value="movies.culturalandacademicfilms" >&nbsp;&nbsp;&nbsp;&nbsp;Cultur al & Academic Films</option> <option value="movies.ephemera" >&nbsp;&nbsp;&nbsp;&nbsp;Ephemeral Films</optio n> <option value="movies.moviesandfilms" >&nbsp;&nbsp;&nbsp;&nbsp;Movies</option> <option value="movies.newsandpublicaffairs" >&nbsp;&nbsp;&nbsp;&nbsp;News & Pub lic Affairs</option> <option value="movies.foreignlanguagevideos" >&nbsp;&nbsp;&nbsp;&nbsp;Non-Engli sh Videos</option> <option value="movies.opensource_movies" >&nbsp;&nbsp;&nbsp;&nbsp;Open Source M ovies</option> <option value="movies.prelinger" >&nbsp;&nbsp;&nbsp;&nbsp;Prelinger Archives</o ption> <option value="movies.sports" >&nbsp;&nbsp;&nbsp;&nbsp;Sports Videos</option> <option value="movies.gamevideos" >&nbsp;&nbsp;&nbsp;&nbsp;Video Games</option> <option value="movies.vlogs" >&nbsp;&nbsp;&nbsp;&nbsp;Vlogs</option> <option value="movies.youth_media" >&nbsp;&nbsp;&nbsp;&nbsp;Youth Media</option > <option selected="selected" value="texts">&nbsp;&nbsp;Texts</option> <option value="texts.americana" >&nbsp;&nbsp;&nbsp;&nbsp;American Libraries</op tion> <option value="texts.toronto" >&nbsp;&nbsp;&nbsp;&nbsp;Canadian Libraries</opti on> <option value="texts.opensource" >&nbsp;&nbsp;&nbsp;&nbsp;Open Source Books</op tion> <option value="texts.gutenberg" >&nbsp;&nbsp;&nbsp;&nbsp;Project Gutenberg</opt ion> <option value="texts.biodiversity" >&nbsp;&nbsp;&nbsp;&nbsp;Biodiversity Herita ge Library</option> <option value="texts.iacl" >&nbsp;&nbsp;&nbsp;&nbsp;Children's Library</option> <option value="texts.additional_collections" >&nbsp;&nbsp;&nbsp;&nbsp;Additiona l Collections</option>

<option value="audio">&nbsp;&nbsp;Audio</option> <option value="audio.audio_bookspoetry" >&nbsp;&nbsp;&nbsp;&nbsp;Audio Books & Poetry</option> <option value="audio.audio_tech" >&nbsp;&nbsp;&nbsp;&nbsp;Computers & Technolog y</option> <option value="audio.GratefulDead" >&nbsp;&nbsp;&nbsp;&nbsp;Grateful Dead</opti on> <option value="audio.etree" >&nbsp;&nbsp;&nbsp;&nbsp;Live Music Archive</option > <option value="audio.audio_music" >&nbsp;&nbsp;&nbsp;&nbsp;Music & Arts</option > <option value="audio.netlabels" >&nbsp;&nbsp;&nbsp;&nbsp;Netlabels</option> <option value="audio.audio_news" >&nbsp;&nbsp;&nbsp;&nbsp;News & Public Affairs </option> <option value="audio.audio_foreign" >&nbsp;&nbsp;&nbsp;&nbsp;Non-English Audio< /option> <option value="audio.opensource_audio" >&nbsp;&nbsp;&nbsp;&nbsp;Open Source Aud io</option> <option value="audio.audio_podcast" >&nbsp;&nbsp;&nbsp;&nbsp;Podcasts</option> <option value="audio.radioprograms" >&nbsp;&nbsp;&nbsp;&nbsp;Radio Programs</op tion> <option value="audio.audio_religion" >&nbsp;&nbsp;&nbsp;&nbsp;Spirituality & Re ligion</option> <option value="software">&nbsp;&nbsp;Software</option> <option value="software.clasp" >&nbsp;&nbsp;&nbsp;&nbsp;CLASP</option> <option value="software.tucows" >&nbsp;&nbsp;&nbsp;&nbsp;Tucows Software Librar y</option> <option value="education">&nbsp;&nbsp;Education</option> <option value="forums" >Forums</option> <option value="faqs" >FAQs</option> </select> <input tabindex="3" style="vertical-align:bottom; text-align:absmiddle ; width:21px; height:21px; border:0px" name="gobutton" type="image" id="gobutton " value="Find" src="/images/go-button-gateway.gif"/> <input type="hidden" name="limit" value="100"/> <input type="hidden" name="start" value="0"/> <input type="hidden" name="searchAll" value="yes"/> <input type="hidden" name="submit" value="this was submitted"/> <a href="/advancedsearch.php" class="level3Header level3Head erSearch">Advanced Search</a> </p></form> </td> <td class="level3Header level3HeaderRight"><a class="backColor1 linkbutt on" style="margin-right:15px; color:#fff !important" href="http://www.archive.or g/create/">Upload</a><b>Anonymous User</b> <span style="font-size:7pt;">(<a clas s="level3Header" href="/account/login.php">login</a> or <a class="level3Header" href="/account/login.createaccount.php">join us</a>)</span>&nbsp;</td> </tr> </tbody></table> <div style="padding-bottom:17px;"></div> <!--BEGIN PAGE--> <div id="col1"> <div class="box" style="border-color:white !important;"> <a class="backColor1 linkbutton" style="margin-right:15px; color:#fff !important" href="/details/higheralgebra032813mbp">See other formats</a> </div>

</div> <div id="col2"> <div class="box"> <h1> Full text of "<a href="/details/higheralgebra032813mbp">Higher Algeb ra</a>" </h1> <pre>DAMAGE BOOK

DO _1 60295 > ^ DO

HIGHER ALGEBRA

BY

S. BARNARD, M.A.

FORMERLY ASSISTANT MASTER AT RUGBY SCHOOL, LATE FELLOW AND LECTURER AT EMMANUEL COLLEGE, CAMBRIDGE AND J. M. CHILD, B.A., B.Sc. FORMERLY LECTURER IN MATHEMATICS IN THE UNIVERSITY OF' MANCHESTER LATE HEAD OF MATHEMATICAL DEPARTMENT, TECHNICAL COLLEGE, DERBY FORMERLY SCHOLAR AT JESUS COLLEGE, CAMBRIDGE

LON-DON MACMILLAN fcf'CO LTD *v NEW YORK ST MARTIN *S PRESS 1959

This book is copyright in all countries which are signatories to the Berne Convention

First Edition 1936 Reprinted 1947, ^949> I952> *955, 1959

MACMILLAN AND COMPANY LIMITED London Bombay Calcutta Madras Melbourne THE MACMILLAN COMPANY OF CANADA LIMITED Toronto ST MARTIN'S PRESS INC New York

PRINTED IN GREAT BRITAIN BY LOWE AND BRYDONE (PRINTERS) LIMITED, LONDON, N.W.IO

CONTENTS ix IjHAPTER EXEKCISE XV (128). Minors, Expansion in Terms of Second Minors (132, 133). Product of Two Iteterminants (134). Rectangular Arrays (135). Reciprocal Deteyrrtlilnts, Two Methods of Expansion (136, 137). Use of Double Suffix, Symmetric and Skew-symmetric Determinants, Pfaffian (138143), ExERtad XVI (143) X. SYSTEMS OF EQUATIONS. Definitions, Equivalent Systems (149, 150). Linear Equations in Two Unknowns, Line at Infinity (150-152). Linear Equations in Three Unknowns, Equation to a Plane, Plane at Infinity (153-157). EXEKCISE XVII (158). Systems of Equations of any Degree, Methods of Solution for Special Types (160-164). EXERCISE XVIII (164). XL RECIPROCAL AND BINOMIAL EQUATIONS. Reduction of Reciprocal Equations (168-170). The Equation x n - 1=0, Special Roots (170, 171). The Equation x n -A =0 (172). The Equation a 17 - 1 ==0, Regular 17-sided Polygon (173-176). EXERCISE XIX (177).

AND BIQUADRATIC EQUATIONS. The Cubic Equation (roots a, jS, y), Equation whose Roots are ( -y) 2 , etc., Value of J, Character of Roots (179, 180). Cardan's Solution, Trigonometrical Solution, the Functions a -f eo/? -f-\>V> a-f a> 2 4- a>y (180, 181). Cubic as Sum of Two Cubes, the Hessftfh (182, 183). Tschirnhausen's Transformation (186). EXERCISE XX (184). The Biquadratic Equation (roots a, , y, 8) (186). The Functions A=y + aS, etc., the Functions /, J, J, Reducing Cubic, Character of Roots (187-189). Ferrari's Solution and Deductions (189-191). Descartes' Solution (191). Conditions for Four Real Roots (192-ty). Transformation into Reciprocal Form (194). Tschirnhausen's Transformation (195). EXERCISE XXI (197).

OP IRRATIONALS. Sections of the System of Rationals, Dedekind's Definition (200, 201). Equality and Inequality (202). Use of Sequences in defining a Real Number, Endless Decimals (203, 204). The Fundamental Operations of Arithmetic, Powers, Roots and Surds (204-209). Irrational Indices, Logarithms (209, 210). Definitions, Interval, Steadily Increasing Functions (210). Sections of the System of^Real Numbers, the Continuum (211, 212). Ratio and Proportion, Euclid's Definition (212, 213). EXERCISE XXII (214).

x CONTENTS CHAPTER XIV/INEQUALITIES. Weierstrass' Inequalities (216). Elementary Methods (210, 217)

For n Numbers a l9 a 2

a > \*JACJJ

n n n (a* -!)/* (a" -I)/*,, (219).

xa x ~ l (a-b)$a x -b x ^ xb x ~ l (a - 6), (219). (l+x) n ^l+nx, (220). Arithmetic and Geometric Means (221, 222).

- V ^

n and Extension (223). Maxima and Minima (223, 224). EXERCISE XXIII (224).

XV. SEQUENCES AND LIMITS. Definitions, Theorems, Monotone Sequences (228-232). E* ponential Inequalities and Limits, l\m / i\n / l\-m / 1 \ ~n 1) >(!+-) and (1--) <(l--) if m>n, m/ \ n/ \ mj \ nj / 1 \ n / l\" w lim (1-f- =lim(l--) =e, (232,233). n _ >00 V nj \ nj EXERCISE XXIV (233). General Principle of Convergence (235-237). Bounds of a Sequent Limits of Inde termination (237-240). Theorems : (1) Increasing Sequence (u n ), where u n -u n ^ l <k(u n _^\^ n _ z ^ 1 (2) If u n >0 and u n+l lu n -*l, then u n n -*L (3) If lim u n l, then lim (Ui+u 2 + ... +u n )jn I. n >oo n->oo (4) If lim a n ~a, and lim fe w = 6, then n-~>oo n~>oo lim (a n 6 1 + a w _ 1 6 1 + ...H-a 1 6 n )/n=o6, (240-243). n >ao

Complex Sequences, General Principle of Convergence (243, 244). EXERCISE XXV (244).

CONTENTS xi

XVI. \CONVERGENCE OP SERIES (1). Definitions, Elementary Theorems, Geometric Series (247, 248). Series of Positive Terms. Introduction and Removal of Brackets, Changing Order of Terms, Comparison Tests, 271 /w p , D'Alembert's and Cauchy's Tests (248-254). EXERCISE XXVI (254). Series with Terms alternately Positive and Negative (256). Series with Terms Positive or Negative. Absolute Convergence, General Condition for Convergence, Pringsheim's Theorem, Introduction and Removal of Brackets, Rearrangement of Terms, Approximate Sum, Rapidity of Convergence or Divergence (256-261). Series of Complex Terms. Condition of Convergence, Absolute Convergence, Geometric Series, Zr n cos nd, Sr n sin n6. If u n /u n+l = l+a n /n, where a n ->a>0, then u n -*Q. Convergence of Binomial Series (261-263). EXERCISE XXVII (264). XVII^CoNTiNuous VARIABLE. ^ Meaning of Continuous Variation, Limit, Tending to <*> , Theorems on Limits and Polynomials (266-268) . Continuous and Discontinuous Functions (269, 270). Continuity of Sums, Products, etc., Function of a Function, lirn <f>{f(x)}, Rational Functions, x n (271). Funda-

mental Theorems (272). Derivatives, Tangent to a Curve, Notation of the Calculus, Rules of Differentiation (273-277). Continuity of <f>{f(x)}, Derivatives of <f>{f(x)} and x n (278, 279). Meaning of Sign of J'(x) (279). Complex Functions, Higher Derivatives (2*9, 280). Maxima and Minima, Points of Inflexion (280-282). EXERCISE XXVIII (282). Inverse Functions, Bounds of a Function, Rolle's Theorem, MeanValue Theorem (284-288). Integration (289). Taylor's Theorem, Lagrange's Form of Remainder (290, 291). Function of a Complex Variable, Continuity (291, 292). EXEBCISE XXIX (293). XVIIL, THEORY OF EQUATIONS (2), POLYNOMIALS (2), RATIONAL FRAC-

TIONS (1). Multiple Roots, Rolle's Theorem, Position of Real Roots of/(&)=0 (296, 296). Newton's Theorem on Sums of Powers of the Roots of f(x) =0 (297). Order and Weight of Symmetric Functions (298, 299). Partial Derivatives, Taylor's Theorem for Polynomials in x and in x, y, ... . Euler's Theorem for Polynomials, x + y + ...~nu J dx dy (299-302). A Theorem on Partial Fractions (302). The Equation

lf + ...=0, (303). EXERCISE XXX (304).

xii CONTENTS CHAPTER XIX. EXPONENTIAL AND LOGARITHMIC FUNCTIONS AND SERIES. Continuity, Inequalities and Limits (306, 307). The Exponential Theorem, Series for a x (307, 308). Meaning of an Irrational Index, Derivatives of a x , log x and x n (309). Inequalities and Limits, the way in which e x and log x tend to oo , Euler's Constant y, Series for log 2 (310-312). The Exponential Function E(z), Complex Index (312, 313). Series for sinx, cos x arid Exponential Values (313). Use of Exponential Theorem in Summing Series (314). EXERCISE XXXI (315). Logarithmic Series and their Use in Summation of Series, Calculation of Logarithms (315-319). The Hyperbolic Functions (319-321). EXERCISE XXXII (321). XX. CONVERGENCE (2). Series of Positive Terms. Cauchy's Condensation Test, Test Series 2 - (325). Rummer's, Raabc's and Gauss's Tests (326-328). '

Binomial and Hyper-geometric Series (328, 329). De Morgan and Bertrand's Tests % (330). Series with Terms Positive or Negative. Theorem, Abel's Inequality, Dirichlct's and Abel's Tests (330, 331). Power Series, Interval and Radius of Convergence, Criterion for Identity of Power Series (332334). Binomial Series l-f?iz4-... when z is complex (334). Multiplication of Series, Merten's and Abel's Theorems (335-338). EXERCISE XXXIII (338).

XXI. JBlNOMIAL AND MULTINOMIAL THEOREMS. Statement, Vandermonde's Theorem (340). Binomial Theorem. Euler's Proof, Second Proof, Particular Ins tancesj 34 1-345). Numerically Greatest Term, Approximate Values of " ; 1 + x (345-348). EXERCISE XXXIV (349). Use of Binomial Theorem in Summing Series, ^Multinomia^Theorem (351-355). EXERCISE XXXV (355). XXII. RATIONAL FRACTIONS (2), PR,ECURRING)SERIES AND DIFFERENCE EQUATIONS. Expansion of a Rational Fraction (357-359). EXERCISE XXXVI (359). Expansions of cos nd and sin nOjsin 9 in Powers of cos 6 (360). Recurring Series, Scale of Relation, Convergence, Generating Function, Sum (360-363). Linear Difference Equations with Constant Coefficients (363-365). EXERCISE XXXVII (365). Difference Equations, General and Particular Solutions (367-370). EXERCISE XXXVIII (370).

CONTENTS xiii CHAPTER XXIII. THE OPERATORS 4> E, D. INTERPOLATION. The Operators J, E, Series for A r u x , u x+r ; U 1 + u 2 + u 3 + ... in Terms of u l9 Au^ A 2 u l9 ... (373-379). Interpolation, Lagrange's and / d \ n Bessel's Formulae (379-382). The Operator f) 9 Value of ( j- ) (uv) (382, 383). Vcte/ EXERCISE XXXIX (384). XXIV. CONTINUED FRACTIONS (1). Definitions, ForniationjoConvergents, Infinite Continued Fractions (388-391). Simple and ^eeuTrirSg; Continued Fractions (391-394). EXERCISE XL (394). Simple Continued Fractions, Properties of the Convergents, an Irrational as a Simple Continued Fraction (396-401). Approximations, Miscellaneous Theorems (402-406). Symmetric Continued Fractions, Application to Theory of Numbers (406-409). Simple

Recurring Con tinned Fractions (409-411). EXERCISE XLI (411). XXV. INDETERMINATE EQUATIONS OF THE FIRST DEGREE. Solutions of axby c (414-417). Two Equations in x, y, z ; the Equation axbycz...=k (417-419). EXERCISE XLII (419). XXVI. THEORY OF NUMBERS (2). Congruence, Numbers less than and prime to n, Value of q)(n) ; E<p(d r }n, where d r is any Divisor of n (421-424). ^ Fermat's Theorem, Euler's Extension, Wilson's and Lagrange's theorems (424-427). ' EXERCISE XL1II (427). Roots of a Congruence, the Linear Congruence, Simultaneous Congruences (428-431). Theorem on Fractions (431). The General Congruence, Division (mod n) 9 Congruences to a Prime Modulus (432-434). EXERCISE XLIV (434). XXVII. RESIDUES OF POWERS OF A NUMBER, RECURRING DECIMALS. Residues of a, ag 9 ag* 9 ... (mod n) 9 the Congruence gr*s 1 (mod n) 9 (436, 437). An Odd Prime Modulus, Primitive Roots (438, 439). Decimal Equivalent of m/n 9 Number of Figures in the Period, Short Methods of Reckoning, Prime Factors of 10* - 1, ( = 1,2,... 10), (439-444). EXERCISE XLV (442). The Congruence x n -sa (mod p) 9 Methods of Solution (444 45). EXERCISE XLVI (445).

xiv CONTENTS CHAPTER EXERCISE XLVII (447). Separation of the Roots, Sturm's Theorem, Fourier's Theorem (447-456). EXERCISE XLVIII (456). Newton's Method of Approximating to a Root, Fourier's Rule, Nearly Equal Roots (456-460). Horner's Method (461-465). EXERCISE XLIX (466).

FUNCTIONS, CURVE TRACING. Implicit Functions, Rule for Approximations (467-469). Series for Roots of a Cubic Equation (when all Real) (470). EXERCISE L (471). Tangents, Asymptotes, Intersection of a; Straight Line and Curve (473). Curve Tracing, Newton's Parallelogram (474-480). EXERCISE LI (480). XXX. INFINITE PRODUCTS. Convergence, Absolute Convergence, Derangement of Factors, Expansion as a Series (485-488). Convergence discussed by Use of Logarithms (490). EXERCISE LII (491). XXXI.yPERMUTATIONS, COMBINATIONS AND DISTRIBUTIONS. Combinations with Repetitions, Things not all Different (493). Distributions, Arrangement in Groups, Distribution in Parcels (494-497). Derangements, General Theorem (498, 499). Partition of Numbers, Table of p Partitions of n, Euler's Use of Series (500504). EXERCISE LI1I (504). EXERCISE TJV (506). XXXII. ^PROBABILITY. First Principles, Exclusive Events, Independent and Interdependent Events (508-513). Probability estimated by Frequency, Expectation, Successive Events (513-518). Probability of Causes, Value of Testimony, Appli cation of Geometry (518-522). EXERCISE LV (523). XXXIII. CONTINUED FRACTIONS (2). Expression of a Quadratic Surd as a Simple Continued Fraction, the Form (VNb l )lr l (527-530). The form V~AjB, Cycle of Quotients, the 6 and r Cycles, Solution in Integers of Bx 2 -Ay 2 =M (530-534). The Form ^/N 9 Integral Solutions of x z ~Ny 2 = M (534-538). The Cycle belonging to (^/N + 6)/r found by the G.C.M. Process (538). EXERCISE LVI (540). MISCELLANEOUS EXERCISES A (543), B (555).

HIGHER ALGEBRA CHAPTER I THEORY OF NUMBERS (1)

In this section we discuss properties peculiar to whole numbers. The word * number ' fs taken to mean l whole number,' and, unless otherwise stated^ ' positive whole number* 1. Division. (1) Let b be any positive whole number, and consider the sequence ...-36, -26, -6, 0, 6, 26, 36,... continued indefinitely both ways. Any whole number a (positive, negative or zero) is either a term of this sequence, or it lies between two consecutive terms. Thus two numbers q and r can be determined uniquely so that

a-=bq + r and 0^r<6 ........................... (A) To divide a by 6 is to find the numbers q and r which satisfy these conditions ; q is called the quotient and r the remainder. If r = 0, we say that a is divisible by 6 or is a multiple of 6, and that 6 is a divisor or a factor of a. Among the divisors of a number we count the number itself and 1. Whatever 6 may be, = 6.0 4-0; hence zero must be regarded as divisible by every whole number. (2) If r = 6~/, we have a = 6-fl-r' and

also if r >|6 then /<J6. Hence it is always possible to find numbers Q and R such that and JR<6.

Ex. 1. Every number is of one of the forms 5n, 5n 1, 5n 2. For if any number is divided by 5, the remainder is one of the numbers 0, 1, 2, 5-2, 5-1.

2 DIVISION Ex. 2. Every square number is of one of the forms 5n, 5wL The square of every number is of one of the forms (5w) 2 , (Sw^l) 2 , (5m 2) 2 . If these are divided by 5, the remainders are 0, 1, 4 ; and, since 4 = 5-1, the for ms are l, and 5n - 1.

Ex. 3. Tlte square of every odd number is of the form Sn + 1. For (2k + l) 2 -4k(k + l) + l, and either k or k + l must be even, so that k(k + l) is divisible by 2. 2. Theorems on Division. (1 ) // both a and b are divisible by c, so also is ma nb. (2)If r is the remainder when a is divided by 6, then cr is the remainder when ca is divided by cb. For if a = 6</ + r and then ca (cb)q + cr and (3) // both a and b are divisible by c, and r is the remainder when a is divided by b, then (i) r is divisible by c and (ii) r/c is the remainder when a/ c is divided by b/c. For if a~bq + r and 0^r<&, then since a and 6 are divisible by c, so also is r, which is equal to a - bq. Thus we have a/c = (b/c)q + r/c where O^r/c <6/c.

3. Theorems on the Greatest Common Divisor. (1) If r is the remainder when a is divided by b, the common divisors of a and b are the same as those of b and r. For, since a = bq + r, every common divisor of b and r is a divisor of a ; and, since r = a-bq, every common divisor of a and b is a divisor of r. This proves the statement in question. (2) If a and b are any two numbers, there exists a number g, and one only, such that the common divisors of a and b are the same as the divisors of g. For numbers (q v r x ), (q 2 , r 2 ), etc., can always be found in succession, and uniquely, so that a = 6y 1 -f-r 1 , 6=r 1 ? 2 + r 2 , r x = r 2 3 + r 3 , etc., ............... (A ) where 6>r 1 >r 2 >r 3 ... ^0. Since the number of positive integers less than b is limited, a zero remainder must occur ; and, supposing that r n ^ =0, the process terminates r -2 = r n-l?n + r n> r n-l =

GREATEST COMMON DIVISOR 3 Hence, by (A), the following pairs of numbers have the same common

divisors : (a, 6) (6, rj, (r v r 2 ) y ... , (r n ^ v r n ) ; and, since fn-i VZn+i* ^ e common divisors of r n _ 1 and r n are the divisors of r n . Hence the number g exists, and its value is r n . Since g is divisible by any common divisor of a and ft, it is the greatest common divisor of these numbers and is known in elementary arithmetic as the greatest common measure (G.C.M.) of a and ft.* (3) If a and b are each multiplied by any number m, or are divided by a common divisor m, then g, the greatest common divisor of a and ft, is multiplied or divided by m. For each of the r's in the proof of (2) is multiplied or divided by m. (4) If a, ft, c, ... , are several numbers, and g l is the greatest common divisor of a and ft, g 2 that of g 1 and c, g 3 that of g 2 and d, and so on, th en the following sets of numbers have the same common divisors : (a, ft, c, d, . . . ), (g v ft, c, d, . . . ), (# 2 , c, d, . . . ), etc. Hence we arrive eventually at a number g, whose divisors are the common divisors of a, ft, c, ... ; also g is uniquely determined. This number g is the greatest common divisor , or the greatest common measure of a, 6, c, ... . 4. Numbers Prime to each other. Two said to be prime to each other when so that they have no common divisor by saying that a is prime to ft, or theorems are fundamental. numbers, a and 6, are their greatest common divisor is 1, except 1. This is often expressed that ft is prime to a. The following

(1) // the product ab is divisible by a number m, and m is prime to one factor a, then m is a divisor of the other factor ft. For, if a is prime to m, the greatest common divisor of a and m is 1 ; hence the greatest common divisor of aft and mb is ft. But, by hypothesis, m is a divisor of aft, and is therefore a common divisor of aft and mb ; hence m is equal to, or is a divisor of, ft. (2) // a is prime to 6, and each of these numbers is a divisor of 2V, then ab is a divisor of N. Suppose that N = aq, then ft is a divisor of aq, and since ft is prime to one factor a, it must be a divisor of the other factor q. Let q = wft, then N = waft, and aft is a divisor of N. * The process, here stated algebraically, is that used in Elementary Arithmetic.

4 PRIME NUMBERS (3) If a is prime to b, positive integers x, y can be found such that ax-by 1.

For it follows from equations (A) of Art. 3, (2), that TI = a - bq v r 2 - - aq z + b (1 + qfa),

Continuing thus, every remainder can be expressed in the form (ax-~by), where x, y are positive integers. If a is prime to 6, the last remainder is 1, and the theorem follows. 5. Theorems on Prime Numbers. A number which has no divisors except itself and 1 is called a prime number, or simply a prime. Numbers which are not prime are said to be composite. (1) A prime number, p, is prime to every number which is not a multiple

For, if a is any such number, q and r can be found such that a~pq + r where 0<r<Cp. Now p and 1 are the only divisors of p, and as r<,p 9 the only common divisor of p and r is 1, that is to say, p is prime to r and therefore to a. (2) If a prime p is a divisor of a product abed ... hk, it is a divisor of at least one of the factors a, 6, ... k. For if the prime p is not a divisor of a, by Theorem (1) it is prime to a, hence it is a divisor of bed ... k. If, in addition, p is not a divisor of 6, it must be a divisor of cd ... k. Continuing thus, it can be shown that if p is not a divisor of any of the numbers a, b, c, ... h, it must be a divisor of k. 6. Theorems on Numbers, Prime or Composite. (1) Every composite number N has at least one prime divisor. For, since N is not a prime, it has a divisor, n, different from N and from 1, which is not greater than any other divisor. Further, this divisor must be a prime ; for, otherwise, it would have a divisor less than itself and greater than 1 , and this latter would be a divisor of N. This contradicts the hypothesis that n is not greater than any ot,her divisor. It follows that every composite number, N, can be expressed as the product of prime factors. For, since N has at least one prime factor, p, we have N~pa, where l<.p<N. If a is not a prime, it has at least one prime factor, q ; and a = qb, where l<w<.a. Thus, N =*papqb ; and so on.

COMPOSITE NUMBERS 5 But the numbers less than N are limited ; and N>a>b>... , therefore the set N y a, b, ... must finally end in a prime. Hence, N can be expressed in the form, N=pqr ... u, where p, q, r, ... , u are all primes, not necessarily all different.

and no prime factor can occur in one of the expressions for N which does not occur in the other. . any composite number.. i. For. Let A>a> and suppose that Aa + e. is equal to one of the set p. t M .. = P A Q B R C . R. and similarly. if 2V = a&. .. If a --A 9 one of them must be the greater. and thus the two expressions for N are identical.. v^V. For. the . . where b^a. and the following propositions are immediately deducible..^p A q B r c . . r . then. Hence. are primes ..*jN.r c u s . .. if p is any prime. (3) // m is prime to each of the numbers. 3. u are all different primes. *-*""*' (5) A number N is a prime.1 . Nevertheless. if it is not divisible by any prime number greater than 1 and less than... 5. 6 DIVISORS OF A NUMBER All the primes less than a given number N can be obtained in order by a process called tlie(Sieve of Eratosthenes). and is therefore equal to on e of them. it is a divisor of one of the factors p. q. (2) A composite number can be expressed as the product of prime factors in one way only. and erasing all multiples of the primes 2. that is.q b . . Q. . N. r. . a... and conversely.. .. R. r. k. . .. The process consists in writing down in order all the numbers from 1 to N . . . r c . .. .. . . g.That is to say. The' above theorem is one of the most important in the Theory of Numbers. q B .. . (6) The sequence of primes is endless.1 is greater than p and is not divisible by p or by any smaller prime. Q.. m ~ M . q. . In the same way each of the set P.. then a n is prime to b n .. a^. which are less than %/JV. it must have a prime divisor greater than p.. but this is impossible . For suppose that N ^=p a q b r c . . a must be equal to A . the number j p 4. and in either case a prime greater than p exists. If then p + 1 is not a prime.... Suppose then that N^p a qr c . 7. .r. where p. (4) // a is prime to b. . P. . .. r. since the prime P is a divisor of the product p a q b r c . . as the left-hand side of the equality is a number prime to p. k. then N^a 2 . . 6 = B. q. can be expressed as N=p a . and the right-hand side is divisible by p.. or equal to. it is prime to the product ab . where p. . then q b . . where n is any integer . t m ^p c .

which are prime to one another. the number of ways is either x or y. prove that n(n + I)(n + 2) is divisible by 6.. d n _ 2 . where p. . the different ways of expressing N as the product of two factors are : did n . Of the two consecutive numbers.. (2) The number of ways in which N can be expressed as the product of two factors. . q. pa+l _ 1 yft+l _ l r c+l _ I p-l q-l r-l For the divisors of N are the terms in the expansion of and the expressions for n and s follow immediately. . since 2 is prime to 3. a perfect square. Further. 1. 2 n . d 2 . We have 3 2r2 + 1 = 3 . d 2 d n _ v c^ n _ 2> . For. each of the numbers. Prove that 3 2n+1 + 2"+ 2 i s divisible by 7.. at least one of these numbers is odd. d 2 d n . . and d^d n . when = 2y. l9 d s d n _ 2 . d y d y+l . n(/if-l)(n + 2) is divisible by G. d x d x . either |(w + l) or \n. r c .. n + 1. PRODUCT OF CONSECUTIVE INTEGERS 7 (3) The number of ways in which N can be expressed as the product of two mctors. are primes . If n is any number. and therefore n is odd .. The Divisors of a Given Number N. n + 2. 2. and one of the three consecutive numbers. . are the divi sors of A' in ascending order. q b . Let N=p a . Ex. Hence the pro duct n(tt + l)(tt-f-2) is divisible by 2 and by 3 . and therefore n is even. 6. according as N is. rf 3 . including 1 and N. is 2 m ~ 1 ... a perfect square. and let n be the number.. .. . n. . and s the sum. including N and 1... is %(n + l) or %n. of the divisors of N. Hence. a. ' but if N is not a perfect square. is divisible by 3. . n and n + 1. (1) n = (a + l)(6 + l)(c + l) . 9 W = 3 (7 + 2) n = Ik + 3 . r.. . one is divisible by 2 . and. according as N is. Ex. if ^( = 1).problem of discovering whether a large number is prime or composite is one of great difficulty. d n _ v d n (=N). is even. c. Then. by the Binomial theor em.-i. or is not. when n=*2a?-l. or is not.e. if N is a perfect square. where m is the number of . and 7.

r + 1) is divisible by [/% aijid . of these I[n/p 2 ] are divisible by p 2 . 8 NUMBERS IN ARITHMETICAL PROGRESSION (2) Ifn is a prime. . For by the preceding n(n .. of the numbers from 1 to n inclusive. ' and m==3 > 8. . For (m + l)(w + 2) . 5 = 540. w 3 = ay 3 -f r 3 . and so on . If a is a fraction or an irrational number. n 3 . there are I[n/p] which are divisible by p . . Theorems.Different prime factors of N. . C? is divisible by n. the symbol / (a) will be used to denote the integral part of a. and s is their sum and a is any number. 5 r 2 == ^[ n 2/ a ]> > ^ e resu lt follows. are any integers.l)(n . 3 3 . (m-f n)/\ n = \ m + nj\ m w.. then Let n = aq^ + r l9 n 2 = aq% + r 2 . For example.2) . since Ji^/f^i/a]. Hence the number of pairs is 2 m ~~ 1 . 9. Thus if x = qy + r where 0<r<y. hence the result follows. /[/a] and.. The Symbol l[x/y].. n = (2 + l)(3 + l)(l + 1) = 24. (2) The highest power of a prime p which is contained in \ n is For. and the same is true if we replace ft by p 2 . Now /[(m + n)/p]>/[w/p]-f /[n/p]. (1) The product of any n consecutive integers is divisible by \-n. then I(x/y) q. p*.. . Thus we have to show that /[ (m + n)/p] -f /[ (m + n)/p*] + 1[ (m -f n)/p 3 ] + . such factors are the terms in the expansion of and their number is 2 W . and to show that the last expression is an integer it is sufficient to show that any prime p which occurs in [m n^ occurs to at least as high a power in j m -f n. . then Hence.. For.. etc. (1) // n v n 2 .. (n . . if # = 2 2 . in succession : hence the result in question.

Therefore the remainders are all different. Ex. -2V +/ -/' = (</5 + 2) w . for a is prime to n and | m-m' \<n. the remainders are the numbers 0. 1. | r is prime to .' In what follows. . they must be the numbers 0. since / and /' are positive and less than 1. are divided by n. The terms . Then. | r is a divisor of (rc-l)(n-2) . Thus t/ ft t* a prime. F$ denotes the number of permutatio ns. NOTE.n. x + 2a... If we suppose that two of the terms as x + ma. then their difference (m~m')a would be divisible by n.. For let a~ga'. (n-r + 1) and ^(ft-l) . so that a' is prime to n'.2) n =/'. 2 n ~ z .2) n =an integer . all the coefficients in the expansion of (l+x) n .. /[158/5 2 ]=/[31/5J-:6. 5*( n ~ 3 ) + . n .. and as each is less than ^. taken in a certain order. Let (V5 + 2) n = #+/ where 0</<1. n . . 2 . (2) If a and n are not prime to one another and g is their greatest common divisor.(v/5 .. For the terms x + ma. which is the required result. x + m'a leave the same remainder if m = m' + qn\. Moreover. . 2.. Again. . since n is odd. 2. taken in some order or other. except the first and last.1. then if the first n terms of the arithmetical progression x. hence. are divisible by n. and thus N =2(Cf . 2 3 . C^ is divisible by n. 10. since 0< V5 -2< 1. x + m'a leave equal remainders. the remainders recur in a cycle of n/g numbers. 1. n=gri. Find the highest power of 5 contained in | 158. . If n is an odd prime. // the progression is continued beyond the n-th term > the remainders recur in the same order. (1) Let a be prime to n. Hence..1 ) + C% . since n is a prime. of n things taken r at a time. and let (>/5 . Ex. / /'. + C . we have also 0</'< 1.. This is impossible.. the integral part of (x/5 +2) n -2 W+1 is divisible by 20n.since n is a prime and r is supposed to be less than n. . Sty. We have 7[158/5]-31.. /|158/f>J=y|<$/5J 1 : therefore the required power has an index = 31 +6 + 1 =38. 1. The reader is supposed to be acquainted with what is said in elementary text-books about ' permutations and combinations ' and the * binomial theorem fo r a positive integral index. 5 + 2 n ). (n-r + l)/[r is divisible by n.. Numbers in Arithmetical Progression. 2. and C the number of combinations. x + a. and therefore N -2 n + 1 is divisible by 20w .1.

.3w . are so divisible . they recur in order. a-f 3. since /(I) is so divisible . /(3). so is/(n + 1) . 2. a(w-w') is divisible by that is. f(n + 1) -f(n) =2 2 ( w + 1 ) .m' is divisible by n'. if 8 is true when n = m. 2. Also 2 2n .3n . . etc. Hence if /(n) is divisible by 9. the theor em holds for all values of n. and. it is true when n = a + l.3 = 3(2^ . suppose we can prove that.111 1 . this can ly happen when m . Ex. Further. it follows in succession that /(2). f(ri) = (3 + l) n . /(4). prove that 11 1 1. is true when na. Suppose that a certain statement S. using the method of (9) Ex.m') is divisible by ri.1 = 3fc. and only if. after that. Since o! is prime to w'. In some cases this is the only method available.1 is divisible by 9. when Ex. Show that 2 2n . i ___ i __ i j __ i I I i i i T ~Zi T i 7^ ^|* * """rki"** ~~ ~T~ i . then /(I) =0 and is divisible by 9. a + 2. o 11.. etc. that is. in succession.1 =(3 + l) n . relating to/(w).2 2W . Method of Induction. that is. 1.. Then since S is true when n = a.1. Again. Many theorems relating to whole numbers can be proved by a process known as mathematical induction. Let/(tt) = 2 2n -3n-l. where fc is an integer. Or more easily. If n is a positive integer. so the theorem is tru e for n = l.1). . The method may be described as follows.x ' Let/(w) be a function of an integral variable n. it is also true when n = m + 1.METHOD OF INDUCTION 9 x f m'a leave the same remainder if. if a! (m . Thus the first ri terms leave different remainders and.

he nce. . . If X=ax + by and 7 a'z-H&'t/.. and b is prime to x. in succession. then t* n+ i=t.. 10 ELEMENTARY THEORY OF NUMBERS EXERCISE I 1. 3. . 5. 4.. If q is the quotient and r the remainder when a is divided by 6. if v n~^ ' "* * t^en Vti ^ f = i>l + l .+ 23 " 2n-l 2n 2n + l 1 1 ^n+i-^w- Hence if u n -v n . 4. Now the statement holds for n = l . and n is a prime. the greatest common divisor of X and y is the same as that of x and y. If a and b are prime to each other. then ax + by is prime to ab. that is. prove that (a n + b n )l(a + b) and a + 6 have no common factor.1 111 1 1 1_ 1 1 . show th*| id q is the quotient when a is divided by b + 1> provided that r^q. If a and b are prime to each other. n+l . unless a + b is a multiple of n. it is true forn=2.6 have no common factor other than 2 . 3.. If a is prime to b and y. for any value of n. show that (i) a + b and a .ab + b 2 and a +6 have no common factor other than 3.n n + 1 n + 2 2n-l 23 4 2n . where ab' -a'b = l. (ii) a* . _ -_ -J_ - n 2n 2n + 1 Again. 2.

For the values 2. and if n is an odd prime greater than 3. .. in the expressions (i) z 2 + x+ 17. 1* . Show that 2 n + 1 or 2 n .b 2 in two ways. If 2 n -f 1 is a prime. 2.1 is divisible by 3.1) is divisible by 24 . show that the remainder is either 0. where k is any integer. and their sum. ' 11. or 3m . according as n is odd or even. according as n is. then n 2 .. 3m -f 1. 13. . according as n is odd or even.n is divisible by 30 . 2. 28. Verify for x = 15. according as n is odd or eve n. If a. then n(n 2 . 10 of x. or 8. 19. 3 . 22. 21. hence prove that 3 W . . then n 6 . Express 55 in the form a 2 .6. If 2> is a prime.9?i 2 -f 3n . Show that 7*" + 16n . Hence write down nine consecutive numbers none of which is a prime. a multiple of 3. it cannot represent primes only. 15. 17v If 3 n . 7 + x is composi te.48n . hence the fifth power of any number has the same right-hand digit as the number itself. i n (ii) x<29. 7.. 8. [Let uf(x) and v~f(x + ku). according as n is of the form 3m. If n is prime to 5. If n is prime to 5. 9. Show that 7 2n . Show that 5 2n -f 1 or 5 2n 1 is divisible by 13.1 is divisible by 64. the number 2 . and therefore n 4 is of the form 5m -f 1. Show that 2 -f 1 or 2^ . 3. Iff(x) is a polynomial. Prove that 2 2W + 1 .1). then n 2 + 1 or n z . 18. takes the values 1. and in (iii) #<40. 14. 12.. (iii) x* + a.1 is divisible by 5.1 . or is not.1 is divided by 13.1 is divisible by 24. 4. (ii) 2z 2 + 29.1 is divisible by 5.1 or 3 2n -f 3 n + 1 is divisible by 13. then n must be a power of 2. and j9=a 2 -6 2 . If n is any odd number.] 10. the resulting values of the expressions are primes. provided that in (i) #<16. 39 respectively. then a=%(p + l) 9 b \(p. 3. Prove that u is a factor of v.1 is divisible by 2304. 20. + 41. Find the number of divisors of 2000. 16.2 is divisible by 54. 5 .

Prove by induction that w(n + l)(w + 2) . 37. prove that the power of p c ontained in | N is (N . 8 the products.s)i(p . excluding N itself. \ n -f 1. 35. \ m 4. . and s is the sum of these coefficients. Prove that. then ' is called a perfect number. n + 2. then | 3nis divisible by|n.S r . and F.1.r when N is equal to 2 f . (ii)P=R s . and L and M are prime to one another. a 3 6c. Prove that [ 2n is divisible by | n . if s=N. Let s be the sum of the divisors of N. | n + 1 . and N . (n + r . 7 6. then 2 n ~ 1 (2 n . d. then g . of the divisors of N 9 prove that =lfn.. $ 25. or is not. 26. If the product of the divisors of N 9 excluding N itself. 30. Prove that the power of 2 in I 3n is greater than or equal to the power Jf 2 in [n . if 2 n .1) is divisible by \r.1) 32. with each of the coefficient s less than p.n is divisible by | m ..DIVISORS 11 23. is equal to N 9 th en N is the product of two primes or the cube of a prime. when (i)p r -l<N<p r +p. 36. 29. and *nd the three least numbers given by this formula. | n -f 1. Hence fin d the smallest number having 16 divisors. If n is the number. a 3 6 3 . 6. Show that |2tt-l/(|n|w-l)is odd or even according as n is.|n + l. 31. c.|n-t-2. a are the numbers. Show that. a 15 . < 24. Find the smallest number with 24 divisors. ifn is greater than 2. M respectively.0. Prove that.1) is a perfect number . Find the highest power of the prime p in JV. it cannot have more than 4 prime factors. (u)p r -p<N<p r . and it must be of one of the forms abed. 34. 27. If N has 16 divisors.1 is a prime. and P the product. 1 rove that (i)n=r. a. r. . of the divisors of ''J7.1 when N is a power of 2. where NLM. L. If N is expressed as a polynomial in a prime p. a power of 2. 33. If ft. if g is the greatest common divisor of m and n -f 1. 28. Show that the index of the highest power of 2 contained in | N is N . R.

The Fundamental Laws of Order are as follows : (l)Ifa=6 then 6 = a. (ii) the introduction of negative numbers and zero. then a = c. If tt n = (3 + v^5) n + (3 .1 and y -* 1 are divisible by the same power 4f 2. the number system is enlarged by (i) the introduction of fractions. 39. or if a>6 and 6^c. Rational Numbers. We thus obtain the system of rational numbers or rationale. 2. with the number zero.A. where it is assumed that zero is not used as a divisor : (5) If a = .V5) n . 40. (3) If a^b and 6>c. p 9 then is an integer divisible by p. Starting with the natural numbers. B. then a>c.. With reference to any two rationals x and y. . ' To say that x>y or y<x is to say that x follows y on the scale. are numbers whose sum is a prime. consisting of positive and negative integers and fractions. If a. This system can be arranged in a definite order so as to form the rational scale. To say that x = y is to say that x and y stand for the same number. show that w n is an integer. c. CHAPTER II RATIONALS AND IRRATIONALS 1. order that division may be always possible .. 6.C.38. and that Hence prove that the integer next greater than (3 + */5) n is divisible by 2 n . Hence we deduce the following rules for equalities and inequalities. in. (4) If a>b then -a< .6. (2) If a = 6 and 6 = c. The integer next greater than (s/7 + \^3) an is divisible by 2 2n . 41 1 If 2" -f 1 =&2/ prove that x . so that subtraction may be always possible. . the terms * greater than/ c equal to ' and ' less than ' are defined as referring to their relative positions on the rational scale.

multiplication and /jvision. the fourth the Distributive Law. in OX take a point 1 so that the segment 01 contains the unit of length. nd (5) (ab)c = a(bc). The point which represents . Theorem of Eudoxus. The fundamental laws of addition and multiplication are f (1) a + 6 = 6 + a. 4.* If a and b are any two positive rationals. the second and fifth the Associative Law. along OX. then ax>by. (8) If a>b and x>y.ac + be. a-x = b-x ) ax bx 9 a/x b/x. then a + x>b + y. To find the point a which is to represent any positive rational a. 5. Also ax^bx and a/x>b/x according as x is positive or negative. Any two rationals can j8 combined by the operations of addition. and if a and y or b and x ai! both positive. equal to m of these parts. . then (7) If a>b then and a-x>b-x. (4) (a + b)c .a is in OX'. the result in each case being a definite rational number. Fundamental Laws of Arithmetic. (6) If a = b and x = y. subtraction. excepting lat zero cannot be used as a divisor.a + x b + x. let a = m/n. The first and third constitute the Commutative Law. This is what is meant when it is iid that the system of rationals is closed for these operations. Take a straight line X'OX as axis . (2) (a + 6) + c = a + (6 + c). an integer n exists such that nb>a. (3) ab = ba. This simply amounts to saying that an integer exists which is greater than a/6. Divide the segment 01 into n equal parts and set off a length Oa. a REPRESENTATION OF NUMBERS BY POINTS 13 3. i i i I X' -a Ola X . at the same distance from as the point a. Representation of Numbers by Points on a Line.

but small in settir out a mile course. then x>a or x< -a.FIG. 2V 3 . . . 6. then any number x is said to be small if | x \ <. N 2 . The absolute or numerical value of a is -fa or . The point X is generally taken to the right of so that. to say that | x |<a is the same as saying that -a<#<a. . 14 NOMENCLATURE 7. j a i |>a. if a>6. If for purposes of comparison we choose some positive number c whuih we regard as small. which is regarded as large. If x/y is neither large nor small. We use the abbreviation ' y is 0(x) ' to indicate that y is of the same order as x. are called large numbers of the first. third. 2 . Points constructed thus represent the rational numbers in the following respects : (i) For every number there is one point and one only. . Large and Small Numbers. Thus |a-6|=|6-a|. 1. A error of 6 inches would be large in measuring a table. . .. Whether a number is regarded i large or small depends on the purpose to which numbers are applied. .6 . the point a is to the right of the point 6. orders respectively. a + 6 > a . orders respectively. x and y are said to be large or small numbers of the same order. .. e 3 . numbers which are of the same order as e. Absolute Values.. We say that x is large or small compared with y when x/y is large or small. If JV is large. according as a is positive or negative. If is small. second.. are called small numbers of the first. f x Sometimes ascribed to Archimedes. (ii) The points occur in the order in which the corresponding numbers stand on the rational scale. numbers of the same order as N. Any number x is said to be large if | x \ >N where N is some previously chosen positive number. second. and is denoted by | a \ .a. j a is positive. . third. It is obvious that + 6.

and it is unnecessary that we should be able to express u n in terms of n by an algebraical formula. .. For instance. u n . Suppose that the object of an experiment is to determine a certain number represented by A.. If x . and we write x-> a -0.y is small. as x tends to zero from above. which is generally denoted by (u n ). and . we say that x tends to a. If x tends to a and is always greater than a. however great that number may be. Such a rule defines u n as a function of the positive integral variable n. it is unlikely that the exact . u 3 . If x tends to a and is always less than a. we say that their number is infinite^ & ^ ^ have what is called an infinite set. l/x tends to oo . 9. APPROXIMATE VALUES 15 10. In this case we also say that . u n may denote the nth prime number. or from the left. or/rom the right. The rule may be quite arbitrary. is called a sequence. . we say that x tends to a from above. to some definite rule.. and we write x -> a. or the integral part of Jn. A succession of numbers u ly u 2 .. For if a<b and k is any positive rational. we say that x tends to a from below. Aggregate. An important property of the system of rationals is that between any two rationals a and b there are infinitely many rationals. This is expressed by writing x-> 0.x tends to . This is expressed by writing x -> a -f 0. the numbers themselves are the elements of the set. where a is constant. System Everywhere Dense. 8.' To say that x tends to zero is to say that x varies in such a way that its numerical value becomes and remains less than any positive number that we may choose. 12. . no matter how small.oo . Sequence. 1 1 . Approximate Values.a tends to zero. No matter what care may be taken to ensure accuracy.l/x tends to . To say that x tends to infinity (x -> oo ) is to say that x becomes and remains greater than any positive number that we may choose. formed according. Any collection of numbers is called an aggregate or set . A sequence in which each term is followed by another term is called an infinite sequence. If the number of elements exceeds any positive integer ^ choose.If x . Thus. Meaning of ' Tends.00 . we say that x and y are nearly equal. however great. it is easily seen that This fact is sometimes expressed by saying that the system of rationals is everywhere dense.

Observe that the error A . then (i) a + b<A + B<a' + b' . the first in defect and the second in excess with errors numerically less than a' . the following theorems are of fundamental importance.a )/a.). or . and we say that 3-1416 is.value of A can be found. . therefore a 1 . In theoretical and practical work.a is an upper limit to the error in taking a to represent A. 4. 9 are those which actually occur in the decimal representation of TT.. Fundamental Theorems. or to 4 places of decimals. and (a' . if a is positive. a'. we shall mean that the figures 3. 16 FUNDAMENTAL THEOREMS In writing down decimal approximations of a number.b . We estimate the comparative accuracy of different experiments by comparing the relative errors : the smaller the relative error.a)/a is an upper limit to the relative error. If a is an approximate value of A. in taking a to represent A. we have l/A<l/a and A -a<a' -a . . we adopt the followinjg scheme : (i) If we write 7r = 3-1416 (approx. and (A . If we find that a<A<a'.B<a' ..a. 1.a)/ A the Relative Error. If a<A<a'. we say that a and a! are approximate values of A. The error in representing TT by 3*14159 is po sitive and is less than 1/10 5 .a cannot exceed a' .. 13. the value of TT correct to 15 significant figures. b 9 b' are given numbers and A. hence (A-a)/A<( a '.. B are any numbers such that a<A<a' and b<B<b f . . a and a' are also called approximate values of A..a . If a. Since 3-14155<7r< 3-14165. we shall mean that TT is some number lyin ig between 3-14155 and 3-14165. Definition. 2i 10 (ii) If we write -n 3-14159 . and that TT is some nu mber lying between 3-14159 and 3-14160. All that can be done in most cases is to find two numbers a and a' between which A must lie.V<A . Again. the greater is the (degree of accuracy. we call a a Lower Limit and a' an Upper Limit to the value of A . the error in representing TT by 3-1416 is numerically less than 0-00005. (ii) a .. A -a is called the (Absolute) Error.

1234567 12 Ex. (a + B)-(B + b')<(A+b')-(B + b f ). Non-perfect 'Powers.0400709 oO lir . 2. 2. therefore aB<Ab' (A) Dividing each side of (A) by Bb'. (iii) ab<AB< a'V . ^|-^< 0-39 -0-34 =0-05. Subtracting B + V from each member of the last inequality. and ^=0-38 . Since a<A and 5<6'.. By division we find that .. and so it cannot be equal to the integer A. For. There are many purposes for which rational numbers are inadequate .b. which is positive. (ii) if the diameter of a circle is 1 inch and we . then I -2x is an approximate value of I/ (I -fa:) 2 with an er ror in defect less than 5a? 2 ..= 1 . and a. Proof of (ii). let x p/q. The truth of (i) and (iii) follows from Art. x cannot be an integer.. If possible. Hence. a~b'<A-B. 13 00 Wehave 35 < 3456789 < 34' Also ^=0-34... a + B<A + b'.2x + 5z 2 . 1. B.. by hypothesis.. 7/0< x< 1.is less than 0-05. 14. a A Bo f< Bb'' ' b f< B m Similarly it can be shown that A/B<a'/b. Since a<A and B<b'. //" . for example : (i) there is no rational whose square equals 7 . .B<a' . Show that the error in representing ^ by . whore p/q is a fraction in lowest terms. therefore the error is less than 0-05. IRRATIONAL NUMBERS 17 Ex.. 12 -.< 1 . Proof of (iv).*. then no rational x exists such that x n = A. 12 1234567 13 . 15. V are positive.and if all the letters denote positive numbers. (iv) a/6' < A/B < a'/b.2x + . oJB Ab' . Similarly it can be shown that A .~.4 is an integer which is not a perfect n-th power. . Then (p/q) n is a fraction in lowest terms. (8). Need for New Numbers. . A.--. /.

.. respectively in defect and excess. (2-70) 2 .. 2-65. 2-64. which is such that 7 lies between the two classes of numbers 2 2 . (2-64) 2 . An irrational is defined by some rule which gives it ordinal rank with the rational numbers. 2-645. Such a rule is the following. (2-645) 2 . No rational exists whose square is 7.65) 2 . However far this process is carried on. this is only an abbreviation for saying that 2/3 lies between the two classes of . 2-6. Take the following instance. for otherwise it would be possible to find a rational whose square is 7. (2-2) 2 . We extend our number-system by inventing a new class of numbers called Irrationals... . (2-1) 2 . 3. Meaning of "/A. Hence 7 lies between two consecutive terms of the sequence I 2 . Observe that ^/7 is related to the decimal 2-6457 .. . therefore 7 lies between two consecutive terms of (2-0) 2 .. 18 IRRATIONALS IN PRACTICAL RECKONING The process therefore gives rise to an endless decimal 2-6457 . 3 2 .. For if we say that 2/3 is equal to the endless decimal 0-6666 .. a' are positive ratio rials such that a n <A<a' n . This is what is meant when we write /7 = 2-6457 . . therefore 7 lies between consecutive terms of (2-60) 2 .. 4 2 . in the same way that 2/3 is related to 0-6666 . Thus if a. 3 2 .. Now 2 2 <7<3 2 . and ^. it can be continued further . (2-6) 2 . 2 2 .. it can be shown that x is not a rational number. 16.. (3-0) 2 .. less than a' . (2-61) 2 ...denote the length of the circumference by x inches.. and . a' are approximate values of ^JA with errors.J1 lies between the two classes of numbers 2. We find that (2-6) 2 <7<(2-7) 2 .. then a. . The symbol *IA (the nth root of A) is used to denote the irrational whose place among the numbers on the rational scale is determined by the following rule : \IA is to follow every positive rational whose n-th power is less than A. 2-7.(2-646)2.. and we find th at (2-64) 2 <7<(2-65) 2 .... and . We consider a positive number A which is not a perfect nth power. and to precede every rational whose n-th power is greater than A. (2 .a. so that no rational exists whose nth power is A... Representation of a Number by an Endless Decimal. (2-7) 2 .2-646. .

of course. l-229< N /5/^/8< 1-232. Now 2-236/1-818 = 1-229 . Given that ^5 =2-236 . . ... We have 2-236 / 1-818< ^5 f /6< 2-237/ 1-817. /. in practice other methods would be used. subtraction. 0-667. 0-66... XIII the rules for equalities and inequalities and the definitions of addition. - Hence we are justified in writing ( %/2 ) 8 = 2 . Ex. find a decimal approximation to s/5 / x/6. 1. Given that $2 = 1-259921 . It should further be noticed that the decimal equivalent of ^7 cannot recur.. . By a process similar to that just described we can find the decimal representation of any root of a given number . 0-666. are extended so as to apply to all real numbers.numbers 0-6. multiplication and division. Real Numbers. Find also an upper limit to the error. The rational and irrational numbers together form the system of Real Numbers. 0-67. ^/7 would be a rational number.. therefore 2 -rf < rf'J -<** = (d' n -d n )(d* + rf w rf' Now d n <d' n <l-3. Letd' n =d w + l/10 n .d* -* as n-> oc . but it is unnecessary for purposes of practical reckoning.. then d%<2<d'l. 1-229 is an approximate value of <s/5/*2/6 with an error in defect less than 0-003. ^8 = 1-817 . .. . In Ch. and 2-237/1-817 = 1-231 . 0-7. but. ...3<C.. EXPONENTIAL FUNCTION 19 Ex. 2. for if it did. /. and .. 17... prove that where d n is the decimal continued to n places. It follows that 2 . ' *l + Wn + *l< (1'3) 2 . Theoretically it is essential to make this extension. already given for rationals. where we replace any irrational which may occur by a sufficiently close rational approximation.

then a^-xx) or as .1.( 1 . let a = l/i. then a n <l. For.\a* = N . <Aen a x >a y according as agl. however great. ^en a n ^l and a n ^l according as a^l. (2) 7/a>0. For we can choose m so that m(a . The Function a*. by the Binomial theorem. for otherwise we should have a = (a n ) w <l. if a<l.1)>N .a. show that 1 -a/2 is an approximate valu e of -a) with an error in excess less than a 2 /2. For if a>l. then a n >l-f-n(a-l). and then by (1). then a x >l according as For if x=p/q. Applying the square root process to 1 . we have -a -a-f^a 2 and fa 2 . if a>l. Now |a 2 -|a 3 -ia 4 =^a 2 (3-2a-a 2 )=4a 2 (3+a)(l-a). Similarly. (1) Ifa>l and n is a positive integer. 3.1.nb. then a n >l. For a a? /a y = a aj ~ I/ >l according as a>l. if a = I -h 6. then 6>1 and Again. then a n >l . a*(3 + )(!. a n = (1 -f 6) n >l 4. If a is positive and less than 1. (5) Ifa>l and N is any positive number. a positive integer m can be found such that a n >N for n^m.|a 2 ) 2 . (4) //a>0 and >y. according as a^.Ex. /. a x = %Ja p ^ 1. and since 0<a<l.\a . (3) //a>0 and x ts a positive rational. 20 APPROXIMATE VALUES (6) // a is positive and not equal to 1.a) >0 18. If a<l.fa 3 .

a x >a m >N. let a = 1/6 and x = . a x are known numbers. m can be chosen so that ( . where a. First. assuming that 0<e<l. let #--> through positive values. a x 1->0 and a*->l. (8) If a is positive and x > 0.y. and any number c lying between a and of is chosen as an approximate value of A. (l-c) m <a.. (1 -e) n <(l -e) m <. if n>m. however small. If a>l. let a = 1/6. (7) //"a>0. If a = 1.1 >-. the numerical value of the error in representing A by c cannot exceed the greater of the numbers (c-^a). and as x -> oo . If x->0 through negative values. if J(a-f-a') is chosen as an approximate value of A. m can be chosen so that (1 -f e) m >a. and therefore A l-a n <. and so a x -+<x>. the greatest numerical value of the error is less than it would be for any other approximation. if z>w. then a 7 * . > JL / d that is. a positive integer m can be found such that | a n . If a<l. If a>l. (a'-c). then if #<l/w.1 =0 for all values of n. as in (5). Hence if and a"-l<. a x . f 1 \ m '1 If a<l. choose m as in the first part of (7) .awarding as a>\. Hence show that. where is any positive number. Then a x = 1 /6 V ->1 . choose m as in the second part of (7) . then if x<l/m. First let a>l.I<a . If a<A <a'. ZAen a x > 1 . . Thus in both cases.l<e. Then in (5).1 |< for n^m. 6* ~> oo and a 35 = 1/6* -> 0. If a<l. EXERCISE II APPROXIMATE VALUES 1. Then 6>1. a x >a nt and 1 -a x <l -a m <. Hence.

b are observed values of two numbers A and J5. show that an upper limit to the numerical value of the error in taking (i) a + b to represent A +B is a-f /? . whose numerical values cannot exceed a and j3 respective ly. and the numerical value of the error in taking a' to represent a canno t exceed a'.is . subject to errors (positive or negative). (iii) y-x 5. APPROXIMATE VALUES 21 4. If the number of steps which he takes per mile is calculate assumption that he steps exactly 30 in.1 8* 4 | < 17* 2 .B is a -f ft . show that x* x 2 + y* + z z z z z x+y+z x' 6. If a. If a is small. show that the error in the resul as much as 35.--h .] 7.e. show that 1 (i)0-3<2/-z<0-5. (ii) a .. given that a-a<A<a + (x. if 0<a<0-01. Also show that.. If 0<x<l 9 show that the error in taking .3x. when | x \ < 1. The to the d on the t may be length of a man's step is approximately 30 in. (ii) 2< <4 . (iii) a& to represent ^4J5 is afc-f (a-f a) .4x 2 . If the numerical value of the error in taking a to represent A cannot exceed a. show that !/(! + a) 3 is nearly equal to l-3a. and a'-a'<a<a'4-a'. then a -fa' is an upper limit of the error in taking a! for A. this value being correct nearest inch. 9.b to represent A .x as the value of . If f(x) = 2 + 3x . If x 9 y. [i. 8.2. \e\ = \ -4x*-5x* + 8x* \ < \ 4* 2 1 -f | 5x 3 1 4. but cannot exceed 36.5x 3 + 8x* and x is small. z are all positive and x<y<z. If 0-2<cc<0-3 and . find an upper limit to t he numerical value of the error in representing f(x) by 2 -f. (iv) ajb to represent yl/7? is i/( a ~l~)"0K where 6' is any positive number less than b ft. it is required to show that :a' + (a + a'). [If e is the error.0-6<y<0-7.] 3. then the error < 0-0007.

is nearly equal to 1 + (a . 22 APPROXIMATE VALUES 15. 10.as the value of .is less than 0-0007. show that the value of / as calculated from the observed values may be in error as much as 9-9 per cent. By the division transformation. where w l9 w 2 are the weights (in grammes) of the solid (i ) in vacuo.l/u.] 11. By means of Ex.b (a . of the true value. A ~r~ OX (ii) if a = 5. The focal length (/) of a lens is given by the formula l//=l/t> . show that I+ax . of the corresponding true values. 12. c =4 and x < 0-01. . (ii) in water.X numerically less than . of its true value. 6^3. v R (1 + where R = (b + c 13. the error in this approximation is less than 0003. 4 3 -h 4# [Divide 3# + 2 by 4# -f 3 in descending powers of x. (l + ox)(l+cx) Show also that if a = 2. the error in each case may be as much as 1 per cent. If x is small. If the true values of w lf w 2 are 13 and 10.i o y o -f" Q. show that if x is small. In determining the values of w i9 w 2 . If the true values of u and v are 20 and 13.b) x 2 . . b = 2 and #<0*1. o o o i o j. the error in taking . r~r/l "\ ^ nearly equal to 1+ (a~b-c)x. show that the value of w calculated from the observed values of w l<t ti> a may be in error by as much as 7 per cent. the error in this approximation is less than 0-0 1. of its true value. The weight (w grammes) of water displaced by a solid is given by the formula w=w l -w 2 . 14. If x> 100. show that (i) .. ' The values of u and v may be in error by as much as 2 per cent.x 2 .. .b)x . 12.

If 0<a 3 -A <e. 1 i [The error < (0-5 + 04 xO5)< . Given that the endless decimal corresponding to /123 is 2-618060 .y 3 = (a . put a = y 2 /x*. If x is small.] 21. show that an upper limit to the error in taking a/6 for A IB is Ii 10 4 .uj _ ( a _M4 value of Va6 with an error in excess less than .( 1 4.[Use Ex.y) (a 2 + ay + 1/ 2 ). Approximate values of A.x)<---. . 6 = 8377. and is nu merically less than # 2 /8. then x-y*j(2x) is an approximate value of s/(# 2 .r x . Using Ex. 17.] o/ 2tl 22. If x and y are positive and x>y.-j~ is an approximate 4(d -}. prove that (i) N/T + x is nearly equal to 1-f rr/3 .a) (a' 4 + a' 3 a -f a' 2 a 2 -f a'a 3 + a 4 ). (ii) if a. using the identity a' 6 . [Proceed as in Ex.[The greatest and least possible values of w 1 are 13-13 and 12-87. then a .] 18.] 16. prove that 123 -(2-618060) 6 < 0-0003. is positive. (ii) s/14/4/3 . Given that ^14 = 3-741.(a + 6) .. 17. . 7. If a and b are nearly equal. for each of the following obtain a decimal approximation with an error in defect. (x\* 10# 2 1 H. B are a = 3083. 3 of Art.) . show that .. [In Ex. 17.4/3.a 5 = (a' . (iii) W 14 4. 4/3-1-442. those of w z are 10-1 and 9-9.. finding also an upper limit to the error : (i) ^14 x 4/3 .i/ 2 ) with an error in excess less than t/ 4 /(2a. each correct to the nearest digit. 20... and use the last example.. o(^JA) [Put y-ZJA in the identity a 8 ...] 17.] 3 19.] . hence those of w l -to* are 3-23 and 2-77. the error in taking 1 +a:/3 to represent Z/l + xis negative. 1 of Art..V A <---.. 3 ).

. <.^. .. second. functions respectively. An expression of the form a x n + a^". a n \x. such that 2<x-f 4<1/10 3 .02..... . the polynomial is said to be homogeneous. n are positive integers. . . Now 6<2 and a + a<4. . third. 24.1 4. A polynomial in any number of variables is similarly defined.. .. A polynomial in x and y is the sum of a number of terms of "the form ax m y n . Show that the numerical value of the error in taking this as the value of A /B is less than 111 W*2 "iO"' Hence show that it is useless to carry on the division to more than four places of decimals. b so that a6-|-/3(a + a)<l/10 8 .. quartic. 7 find how many figures must be kept in A. If the sum of the indices of the variables in each term of a polynomial is a constant number. .a 2 x n ~ 2 y 2 + . a homogeneous polynomial is called a quantic. we may take a. In some parts of Algebra. a n are independent of x. . 6 = 1-4142. suppose a/b expressed as a decimal to n places. + a n y n . If ^=3-141592.a 2 z n ~ 2 + . fourth. quadratic. is a homogeneous polynomial in x. + a n y n . atfP + a^-^y + a 2 x n ~ 2 y 2 + . a 1? .1416. This is shortly denoted by (00. Polynomials of the first.*! . to the nearest digit. where a . is called a polynomial in x of the n-th degree. B = 1 -414213 . by using Ex. y of degree n is often written in the form atfc n + na^-iy 4.^. cubic. [Choose a. Definitions. /.23. A constant may be regarded as a polynomial of degree zero. degrees are known as linear. A homogeneous polynomial in x. . This is satisfied if a<^ . B to find an approximate value of AB with an error numerically less than I/ 10 s . and m. and is sometimes called a rational integral function of the variables : its degree is that of its highest term. In the last example. . .. y of degree n. We may therefore take a = 3. . where binomial coefficients are introduced for convenience.y) n . + a w . For instance..] 6 1( ^ 6 1U ^ CHAPTER III POLYNOMIALS (1) 1 . where a is independent of x } y..

cr=ph + c.4 is said to be (exactly) divisible by J? of by C. c.1 . 24 SYNTHETIC DIVISION 2.2x . Division. We divide by x . A being of higher degree than J5.1 by x + 2.e. the reckoning may be arranged as on the left : a +b + c +d x-h)ax* + fcz 2 + ex-}.x* f 2x 2 .2 . To ' divide A by B ' is to find an identity of the form where Q and R are polynomials (or in special cases constants) and R is of lower degree than B.( . I) 3 =aa? 3 -h 36x 2 4. This process is known as synthetic division.6cx 2 4. Divide 3* 4 . Such division is always possible : the result is unique and is called the division transformation. If A.idx 4.3cx + d. Ex. or of the same degree. In dividing ax 3 -f bx 2 + ex -f d by x-A.2) as follows : l-(-2) 3 -1 +2 . these must be supplied with zero coefficients. cubic and quartic functions of x may be denoted by (a. then JB and C are called factors ol divisors of 4. Let A and B be polynomials in x. ( (a. C are polynomials and A=BC. and .d(ax z +px + q 7 7 ax* .2 + cx a +p + q + r px *-phx where p = ah + b.Using this notation. c. 3.akx 2 ah + yA + g ^ . Q being the quotient and R the remainder. 5. 1. If any powers of x are missing. d$x. 6. r and the remainder = r. r = qh + d : r > 2 r > 2 qx-qn and then . I) 4 = ax 4 + 4&z 3 4. It is justified by a comparison with the reckoning on the right. Synthetic Division. It is evident that the method can be used to divide any polynomia 1 in x by x-h. d y e\x. the quotient = ax 2 -f px + q. 6.

b.I in the form .34 . 2. ll(2-3) = 22-33 . APPLICATIONS 25 Ex. 8 + 3 = 11. the required quotient = J (a.7z 2 + Ux . The first term in line (c) is 4. then Q/a and R are respectively the quotient and remainder in the division of A by ax. write the divisor in the form x 2 . 4 (2 -3) = 8 -12. remainder = 67. Ex.-0 +14 -32 +68 3 -7 +16 -34 +67 . The reckoning will be understood if it is compared with ordinary division. Quotient = 3z 3 . Express f(x)= 4x* + 3# 3 + x .'. and so on.3) and proceed thus. /. put 22 in (6) and -33 in (a). put 8 in line (6) and -12 in (a). 2 + f a.^. 7/ the divisor is of the second or higher degree. it is seen that 5 y.b/a. To divide a polynomial A by ax. + 10 and the remainder -12z-31.1 Dividing by x .J) . . as on the right.J. 4 +3 +0 +1 -1 -12-33-30 (a) 8 + 22 + 20 (b) 10-12-31 (c) The quotient is 4z 2 + llo.4 +t +4 .J and the remainder = ^.6. 2 + # .(2x . remainder = . Dividex* + 2x 2 ~3x~4:by2x-l. we may use the following rule : * Find the quotient Q and the remainder R in the division of A by x.^ the quotient = a. 3. the process is as follows : to divide 4# 4 + 3# 3 + x-l by 2 -2z + 3. l + 2 -3 .

a n _ 2 is the remainder when Q' is divided by x-h. 26 REMAINDER THEOREM Ex. Then a r is the remainder when/(#) is divided by x .. overleaf. . 6. 1.2 and the quotient 4. The remainder is -12rc-31 and the quotient 4# 2 + 1 Ix + 10. 1-2 1-2 1-2 . . The reckoning is arranged as in Ex. 1. ItQ is the quotient. Let/(x) be a polynomial in x. we can find all the coefficients.2x + 3.h. Hence -12 8 . The remainder is 19a.12* -31. If Q' is the quotient.h. a n-1 is the remainder when Q is divided by x . Suppose that f(x)=a (x-h) n +a l (x-h) n . Divide f(x) by x*-2x + 3 as above. (I) To express f(x) as a polynomial in x-h. e are independent of x.3 as a polynomial in x-2. Continuing thus. .+a n _ l (x-h) + a n .. ~ 2 4. Applications of Synthetic Division.l + .where a. The following transformations are often required. 4 + 11 + 10 Divide 4z 2 + 1 Ix + 10 by x 2 . . Express 2x z + x 2 5x .

h.3 + 10 2 + 5 + 4 + 5 + 18 + 7 2 + 9 + 4 4-23 2 +13 (2) To expand f(x + h) in powers of x. and then change x into x + h. . 1 and substitute x + 2 for x in the result. (3) Iff(n) is a polynomial in n of degree r. .5 + 10 . thus f(x + 2) = 2x* + 13* 2 + 23* + 7.2 + 1 + 4 . 2. we express f(x) as a polynomial in x . Ex. to express f(n) in the form a Q + a^n + a 2 n(n + 1) + a%n(n + 1) (n + 2) + . Proceed as in Ex. .

.4n + 10( + 1) . Express n 4 + 3n 2 + 2 in the form Divide n 4 +3n 2 + 2 by w. the quotient by n + 1.This can be done by dividing by w. . . The Remainder Theorem. 1+1 1+2 1+3 1+0+3+0 -1+1-4 1 _i +4-4 -2+6 1 -3 +10 -3 1 -6 /.. The reckoning is shown below. and the re quired coefficients are the numbers in thick type. n + 2.h. and finally the last quotient by n + 3. f or R is independent of #. 2fo. Iff(x) is a polynomial. then f(h) is the remainder when f(x) is divided by x . w-hl. in succession. This follows on substituting J^ fof xirTtfre* identity where Q and J? are respectively the quotient and remainder in the division of f(x) by x-h. n 4 + 3n 2 + 2 = 2 . 3. the quotient thus obtained by n + 2.n(n + !)( + 2) 5.

f(0-2) is the remainder in the division of f(x) by a. .IMPORTANT THEOREMS 27 Corollary. R Ex. hence if x . 2.*./md *fce w/we off (0-2). ( . ##. an + x = ( _ ! jn f ! _ _ X + X _ . Since n is odd. in the preceding. 1. Iff(h) = 0. if n is odd. x + 1 is a factor of x n + 1. Show that.) = 3** + 2** .0-2. ///(a.1 is a factor of x n -+ 1 . x 4.1. then x~h is a factor off(x).6x .l) n = .4. Performing the divi sion : 1-0-2 3 + 2 + 0-6 4-0-52 -6 + 0-104 -4 -1-1792 3 + 2-6 T?.5-896 -5-1792 . . For.fTn? + 0-52 . .1.

Hence . If for more than n values of x. c B.. and (x^-(x-^)(x where / 2 (x) is a polynomial. ax n -f bx n " 1 -t. (z-a w ) is a factor off(x). Theorem. Now /(a 2 ) = and aa-aj^O. . + hx + k = a'x n + b'x n ~ l -f .. k^=k'. & = &' .. Theorem. which may be real or imaginary.. . 6) the product of more than n expressions of the form x-cx. For by the given conditions the polynomial (a . which is impossible. a polynomial of the nih degree would have a factor of degree higher than n. we can show that (x-oc 1 )(x-oc 2 ). it follows that a^a'.a) -f (x . that is to say the polynomials are identically equal. .h')x + (k . would be a factor of the polynomial. NOTE. h = h f . hence. -f h'x + k'. 7. .. 6. Therefore /(a 2 ) = (a 2 . Later it will be shown that every equation of the nth degree has exactly n roots. not necessarily all different. . A polynomial f(x) of the n-th degree cannot vanish for more than n values of x unless all its coefficients are zero. 2 is a factor oif^x). For otherwise (by Art. then the product (x-oc 1 )(x-(x 2 )(x-oc 3 ) . then a = a'. .C.f l (x) y where / x (x) is a polynomial. . therefore /i(a 2 )=0. An equation of the n-th degree cannot have more than n roots. // f(x) is a polynomial which vanishes when x has the values a 1? a 3 .6> n ~ 1 + .a')x n -h (6 . For since /(a^^O. 6 = 6'.(x-oc n ) is a factor oif(x).. That is to say. . h = h'....k') vanishes for more than n values of x .aj)^ (a 2 ).. Proceeding in this way. a n . 28 UNDETERMINED COEFFICIENTS 8. Corollary.b) vanishes identically . .. for it is of the first degree in x. Thus (x -a)(b-c) + (x.6. If all the coefficients of f(x) are zero. by the Remainder theorem f(x) = (x~oc l ) .A. . and consequently x-oe. Theorem. f(x) is said to vanish identically.. and it vanishes for the values a. c of x.. 6 = 6'. -f (h . by the preceding article.b) (c .c) (a . no two of which are equal.

6. where some of the coefficients are unknown. NOTE. which is supposed to vanish for all values of x and y . 6 = 1. hence a = 0. . (1) ///(x. c= -1. which vanishes for all values of x . The fact that two polynomials. 6 = 0. The right-hand side ^ a (a.2/y . u -3A"5.1) This is equal to the left-hand side for all values of x.c 1. are identically equal is often expres sed by -writing u^v. if values for a. y) is a polynomial in x and y which vanishes for all values ofx and y y then all the coefficients off(x. the function can be expressed in the specified form. Show that values of a. /=0. and for any given value of y. If these equations have a solution. a~2. 2 -x -2) + b(x 2 ~ 3z + 2) +c(z 2 . y) are zero. e-0. It is often necessary to enquire if a given function can be expressed in a certain specified farm. 10. c = 0. (2) If for all values of x and y the 'polynomials f(x. then the coefficients of like terms in the polynomials are equal. c c an be found such that a ~ b-~<. This follows immediately from (1). The following method applies in all cases. QUADRATIC FUNCTION 23 Ex. the last expression is a polynomial in x. a = 0. These equations have one solution. v. rf = 0. The Method of Undetermined Coefficients. In cases of this kind we may use the method of 4 Undetermined Coefficients/ which may be described as follows. Assume that the function is expressed in the given form. y) and F(x. namely. 6. Let the polynomial be u = ax 2 + bxy -f cy 2 + dx + ey -h/. 1. y) have equal values. In order that the identity involved in this assumption may be true. Polynomials in Two or More Variables. c exist such that and find these values. -a . then u = ax 2 + y(by + d) + (cy 2 -f ey +f) . the.~'2. /. unknown coefficients must satisfy certain equations. cy 2 + e?/+/=0 (A) Also the relations (A) hold for all values of y . u.9. by + d = Q.

they are therefore rational functions of y..3/> 4 7 .--.. (C) 30 DISCRIMINANT These values of x must be the same as (qy + r) and .. i/ t has been shown that an identity of the specifie d form exists. Ex..... Assume that S = (px + qy + r) (p'x -f q'y -f r') and consider the equation S=^0 .... wo can find er. 2 of Art. Hence the expression under the root sign in (C) is a perfect square. 2.. y. giving x (hy+g) (hy+gf-a(by* + 2fy + c) .af 2 ... 10 it appears that the cases in which an expression of the form (B) can be resolved into factors are exceptional..3?/ 2 -x + 10// .. Since 2x 2 ..f (q'y -f r ') . Thus....10..2///2 -f 2(jzx + 2hxy . 11.... b. (A) When z^l this reduces to az* + 2hxy + by z + 2gx + 2fy + c. c..1.3y a + a?(^ -i-2q) +y(q -3^) +y^. (B) which is the general form of a quadratic function of x and y... (/ = !. From Ex.. and these values hajtpcii to satisfy the th ird condition..... Quadratic Functions of x and y.. From the iirst two. -1...In a question of this kind..3.5xy .. assuming that an identity of the specified form exists.ch* .2q -..3?/ 2 (2x + y) (x . Search for factors of #=3 2x~ ...0.3y) we assume that ti^(2x + y+p)(x ~3?/-f q) =2x~ . z of the second degree is of the form ax 2 -f 6// 2 + cz* 4.Itf . the values of the constants may be found by giving special valuer to the variables... Every homogeneous function of x. The condition that this may he possible is investigated in the next article.5xy . Theorem.. by putting a:!. pq :-... Equating coefficients.2 in succession. .3....0.5r// .. we have tluj three conditions : p -\.... y>-^ -3.... The necessary and sufficient condition that the quadratic faction s ^ aj? + 2hxy + by* + 2gx + 2fy + c can be expressed as the product of two linear factors u A -= abc + 2fgh .. this can be written ax 2 + 2x (hy +g) + (bif + 2fy -f c) . .. Now .. 12.

then S = 2hxy + 2gx + 2fy + c for in this case A = 2fgh-ch 2 . DIVISION AND FACTORS 31 EXERCISE III DIVISION AND FACTORS In Exx.. (iii) // both a and b are zero. Conversely. (D) (ii) //a = the reasoning fails . or y2(h 2 -ab)+2y(hg-af)+g*-ac.. The expression abc + 2fgh-af z bg*-ch 2 is generally denoted by J.. (iv) //a = 6 = A = 0.be must be po sitive. //. from the preceding. ca-g\ ab-h\ are all negative .ab and g 2 -ac are positive.a/ 2 . p'. The square root of (hy + g) 2 . p'.a (by 2 + 2fy + c). This is evident in all cases.. 1.. then S can be resolved into linear factors. but. if S can be resolved into linear factors. . r. except when 6 = 0. NOTE. A =0.. p. we shall be led to the condition (D) by solving for y and proceeding as before. Hence. if A 0.. by solving S =0 as a quadratic in y t we find that/ 2 . q' t r' are all real... then J=0. and only if h 2 .. q' 9 r' t are not necess arily all real.. S is not a quadratic function.. Observe that p. q.. Ex.and the condition that this may be a perfect square is or a (abc + 2fyh . q.... and only if.. prove that 6c-/ 2 . r.ch?) = 0. in all cases. is real if. it follows that abc + 2fgh-af-bg*-ch 2 =0 .. Similarly. and is called the discriminant of S.. 1-6 find by synthetic division the quotient and remainder when the first expression is divided by the second. in the above. and vice versa...bg 2 .. The following cases must now be considered : (i) // a is not zero. and A^O. Hence S can be resolved into linear factors if...

find the equation giving */ when (i) i/.3y 4 .*. (iii) x 4.I . 13. x*-4x* + 6x z -4x + I . Find the condition that ax 3 + bx -f c and a'x 3 + b'x -f c' may have a comm on linear factor.a. [The function is of the form (x . we obtain the required condition. [x-<x.c'a) 2 = (be' . and shall have the values 4 and 28 when x~ -1 and # = 2 respectively. [The polynomial is of the form (x . (ii) y 3#~. 3.1. Express x 3 + 2x* + x + 80 as a polynomial in (i) x + 5 . 6.2) (ax + b).] 19.] 16.1. a. Prove that the condition that ax 2 + bx + c and a'x 2 + b'x + c' may have a common linear factor is (ca' ./(-f).] 17.] 15. Find a homogeneous function of x and y of the second degree which shall vanish when x~y and also when # 4 and y~3.1. show that x m + 1 is a factor of x mn -f. 5. 9. . x* . 18. 11. x* + 2x-I. + 4. Find a polynomial in x of the third degree which shall vanish when x = 1 and when x -2. 12. find the value of a. 4. will be a common factor if a 2 -f6a + c=0 and a x a 2 -i-6'a-hc'=0. 10. Find a quadratic function of x which shall vanish when x -f and have the values 24 and 62 when x = 3 and x ~ 4 respectively. Factorise the expressions in Exx.a'b).J.3x 3 2/ . 3x*-5x*-llx* + x-I. 3 -:e 2 -2. Express # 15 + 1 as the product of four factors. 12.6z 2 */ 2 .6x + 1 is divisible by 2x . [It follows from Ex. Iff(x)---Gx* + 4x* + 3x + 2. Express n 4 in the form a + bn + cn(n + l) + dn(n + l)(n + 2) + en(n + l)(n + 2)(n + 3). 2x + l.b'c) (aV . 3z 5 + 6z 4 ~2o.] 20. If 4z 4 . [The function is of the form (7x + 3)(ax + b). 5 -!. /(J). 7. -2 . 2. 3x-2.(a . x*-2x~2. 5x* + 6x + 2. If n is odd. x + 4. x*-2x*-lx* + 8x+l2. and have the value 2 when x 2 and y = 1 . (iii) y = 3x + 4. 2z 4 . 8. find the values of /( -!).5x* -f Ix* .8# + 3 = 0. Eliminating a between these equations.1 . If x is given by the equation # 3 .l)x 3 -f ax 2 . 13. x-2.y)(ax + by). (ii) x -f 1 . 14.1) (x -f.a 2 . 17 that cc 3 + 1 and x 5 + 1 are factors.8xi/ 3 .

29. q 2 . 2x 2 + 3x-l. 6. Find this equation. ivp + 3q. z. r 2 . r. 6.] 26. c. and find the cube root. [This follows from Ex. a? + l. 27. c.1x f. and find this equation. d such that 24.IGy . 3x~i-xy~4y 2 + 8x + 13?/-3. z 2 . c. 32 DIVISION AND FACTORS 22. s 2 are connected by a linear equation. x 5 + 2x 3 + ax 2 + b . b can be found for which the first expression is divisible by the second. and find these values. and resolve the given expression into factors in each case. 21.] Resolve the expressions in Exx. (i) Find the remainder in the division of . Show that x 6 ~ 6z 5 -f 24z 4 . t/ 2 . [Find the remainder in the division of the first expression by the second. and find 6.96x + 64 is a perfect cube. show that n z can be expressed in the form rt(rc-l and find the values of a. In Exx. ax* + bx*-l2x 2 + 21x-5. y~p-q.9x -h c is the product of three factors. [Write x p-3q. two of which are i dentical.21. If x t y. Show that constants a. 3x 2 + 2xy . w 2 are connected by a linear equation with consta nt coefficients. If p. show that c is either 5 or . [Arrange in ascending powers of x before dividing. and make this identically zero. 31. If ft is any number. 25.6. 6. If x 3 -f 3x 2 . 30. 28 into factors. 25. w are four consecutive terms of an arithmetical progression. c can be found such that and find the values of a. s are four consecutive numbers. 26 show that values of a. 28.56z 3 -f 96x 2 .] 23. prove that x 2 . zp + q.27 . show that p 2 . 27.Sy 2 . q.

x z -xy + 3y-2 + \(x 2 -y 2 ).4 and -h 6 inclusive.x 3 + 3px-fg by x 2 -2ax + a 2 .Wy -f 24 = 0. [Solve for x and y in succession. what is the relation connecting p and q ? 32. or the values. show that /= *Jb c.] 40. show that the remainder in the division of /(#) by (x . 6 are unequal. of A for which the given expression can be resolved into factors. where x and y are real. where x and y are real numbers. provided that the negative sign does not precede an odd number of the radicals.12r. (ii) If atP + 3px + q has a factor of the form (x-a) 2 . and y between . (iii) If the equation x 3 +3px + q has two equal roots. g Vca.b) is [(x-a)f(b)-(x-b)f(a)]/(b-a).I2x -8?/-f 15 = 0. x z + y 2 -6y + 4 + \(x 2 -3 35. 36. prove that ax 3 bx + c and ex* i-bx 2 + a have a common .] EXPANSION OF PRODUCTS 33 37. 33-34 find the value.|. show that x cannot lie between I Jl and y cannot lie between 1 and 3. If 5x 2 -f. If x*+px* + qx + r can be expressed in the form (x -a) 3 (x -b). show that a 2 c*=ab. h= Va6. In Exx. (i) If ax 5 + bx + c has a factor of the form x 2 + px -f 1. If a. 2 -f xy-y 2 + 2x~y+ 1 0. where x and y are real.^5 and 2+^5 inclusive.a)(x -. If f(x) is a rational integral function of a. [The remainder is a linear function of x. 33.24x . 39. show that 8^p 3 -: -21q* and p 2 = . show "that y cannot lie between and . 38. show that # a + 4/> 3 = 0. and . 34. biit x can have any real value whatever.4xy -f y* . and may therefore be assumed to be A(x-a)+B(x-b). If ax 2 -f 2Lxy + by 2 + 2gx + 2fy + c is a perfect square. and for each value of A resolve the expression int o factors. (ii) In this case. show that x must lie between 2 . If 2# 2 -f 4xy + y 2 .

// p l9 p z . . f a n ) or p^x*. we then multiply each of the products so formed by each term of (x + y + z + w). . (i) If ax* + bx 2 -\-c has a factor of the form x 2 +px + 1. two.. (ii) In this case. Choosing x out of any (n .. (A) we multiply each term of (a -f b) by each term of (p + q + r) . . we obtain x n ~ l (a l + a 2 + d3 + . In general.... prove that <7# 5 4 6x 2 -f c and ca.. and add the results. 5 4. denote the sums of the products of i> #2 a s> ? a n> token one... (x 4. the product of any number of polynomials is equal to the sum of all the products which can be formed by choosing any term out of each polynomial and multiplying them together. in the identity ax* + bx + c -(x*+ px + l)(ax + c) put I/a.. Theorem.. we obtain the term x n of the expansion..... 13.. . we obtain.r) of the factors and an a from each of the r remaining factors.. Expansion of Products.. 34 BINOMIAL THEOREM Choosing x out of each factor.... three. 4.2) of the factors and an a from each of the two remaining factors... To expand the product (a + b)(p + q + r)(x + y + z + w).quadratic factor...6x 3 + a have a com mon quadratic factor. 14...1 .. for x and multiply each side by # 3 . [For (ii). p 39 .. Hence the expression (A) is equal to the sum of all the products which can be formed by choosing any terra out of each of the factors and multiplying them together. we obtain x n ~ <L (a^i -h a^ 3 -f . .c 2 ) (a 2 ~ c 2 + 6c)=a 8 6 8 . Choosing x out of any (n . at a time.a n ) i equal to the sum of all t he products which can be formed by choosing a term out of each of the factors and multiplying these terms together.a 2 a 3 -f . ) or p 2 x n ~ 2 . Choosing x out of any (n 1) of the factors and the a out of the remaining factor..jo r x n ~ r .. . respectively. then For (x + a x ) (x + a 2 ) (x + a 3 ) . prove that (a 2 . .] 41...

. Find the coefficient of x r in (1 -f x + x 2 4. Ex. Use the identity (1 -f x) 2n = (l +2x+x 2 ) n to prove that . (x + a n ) =x n + p^" 1 + p 2 x n ~ 2 -f .) n ~ 1 is CJ .l a + C%x n . 2.. .---3) _ |2? TUT UH We have The coefficient of x n in (1 + x) zn is <7^ w . the coefficient of z n in C"(x 2 + 2x) n -* is (7J . Many identities can be obtained by expanding a function of x in two different ways by the Binomial theorem and equating coefficients. C J" 1 .. can sometimes be avoided. The result follows by equating coefficients in the identity. . . (x + aj (x 4. we use synthetic division. 1. When the coefficients in f(x) are known numbers.z n -f C%x n . 4 . . Here /(x) is supposed to be a polynomial in x. if r = 2m.a 3 ) . A 22 and so on. a n .x 3 ) 11 . -f p n . .Finally. if r= MULTINOMIAL THEOREM 35 15. In particular. the coefficient of x n in (x z +2x) n is 2 n . 2 n ~ 2 . . as in Ex. Here 1 + Sa^ r = (1 +a? + x* + x*) n = (l +x) n (l +x*) n Hence. we obtain a^a^a^ . . then and. as in Art.a n .a) n . 2. The use of the Multinomial theorem. if a a 2 = a 3 . which is the Binomial Theorem for a positive integral index. . Expansion of f (x + h) in Powers of x. = a n = a. 2 W ~ 4 . . Ex. the coefficient of x n in C^(a: 2 + 2a. explained in Art. .2 a* 4. choosing an a out of each of the factors.'.. we have (x -f. 3.a 2 ) (x 4. . 16 on the opposite page.C"x n ~ r a r 4. (7^~ 2 .

k. to w factors is the sum of all the products which can be formed by choosing a term out of each of the n factors and multiplying these terms together.. For by the Binomial theorem..+A n .. .a 2 . Multinomial Theorem for a Positive Integral Index... -f k) . I {/ \JvfvJU ~T~ A/-J i*/ "T~ /i'o* / i ~ n/y "^ <^ tt where A. 2. .a 1 . 1 of this chapter. where A = a Q h + a ly A 2 ^ a h 2 + 2%/fc -f a 2 . a r \h.. -f k) n .-fic = tt (A) . A 3 = a h 3 + Sa^h 2 + 3a 2 h + a 3 . 16. Ai / r\ 1 // f(x) = a x n then f(x + h) = a^c n + nA^ 1 ' 1 + ----'.... . where each index may have any of the values 0... ^ r -(a . l) r .A^ n ^ + . . r = a C J?A r -f a!^ iC^-Ti A 7 *" 1 + agC^^Y^^ 7 ^" 2 + to r + 1 terms. The product (a + 6 + c + .. +k)(a + b + c + .. w.. The expansion is therefore the sum of a number of terms of the form a a b^c y .The following general theorem is required later : the notation is explained in Art. Now C*!>C*rI% n |r-j)|n-r Thus k r ^C^(a Q which proves the theorem. ft 'YW I ~L* r*n\ \ j/> /y7t 2 I _1 jf nffl~~T I . subject to the condition .. (1) Expansion of (a + 6 -f c + . 1.. .

K are to have all possible sets of values which can be cho sen from the numbers 0.. . . . The . Denote this value by v.aWcv . respectively. the numbe r of times which this particular term occurs). )S. To obtain the coefficient of the term a a b^c v . has | a coefficient | n I \ a | ^ | y . . and so on.c -f . Take a out of each factor of the first group. . . . then v is the least value of | j8 .. K which satisfy this condition. . . . 1.. -f k) n is where q is the quotient and r the remainder when n is divided by m. Jt. (B) a 7 and this is the coefficient of a a b^c y . \K has its greatest value when | a | j8 . . .36 MULTINOMIAL THEOREM Choose any positive integral or zero values for a. | * if a and j3 alone vary and y. . .e.. .. This result is called the Multinomial Theorem for a positive integral index. Thus in the expansion of (a ..... 2. jS. . k" in the expansion (i. +K = n. . b out of each factor of the second group. y. . c. n. the greatest coefficient in the expan sion of (a + b 4. K factors.. l/c its least value. y.. b.. Hence .. (C) where a. The number of ways in which this can be done is In .. subject to the condition a + j8-f y-f. .. jS. . . y.b . k. k" y divide the n factors into groups containing a.. K are constant. .c -f d) 6 the coefficient of a*b 2 c is (2) If there are m numbers a.

the coefficient of x r in the expanI n sion of (a + bx + ex 2 + . . which proves the theorem... y are shown in the margin . 5 such that a ^ V = 5 and The solutions of the second equation are (4. j8.r~~ &*&& .. EXAMPLES 37 We conclude that some (possibly all) of the set a.e.. /?. w subject to the conditions This follows from (1) by putting 6x. . 0).. Let r of the numbers be equal to q + 1 and m . 2). . . j_a j8 | y where a. 6. . where C is a constant. c. 6. . j^o:.. Thus a and /? must be equal or differ by 1. j8 = i(C-l) if C is odd. l a . and when a = i(C + l). where _ a. each of the rest being equal to q + 1. Hence q is the quotient and r the remainder when n is divided by m.Thus | a | j8 must have its least value subject to the condition a 4.r of them equal to j.... y. ex 2 . .j8= C. as agKC-fl). 3v. i. 1. . and the same is true for any pair of the numbers a. that is qm + r = n.. c. Hence u a has its least value when a=j8=4<7 if C is even.. A in equation (C). . .. y have any of the values 0. .. (0. .. (3) // there are m numbers a. k". k. 6. 2. 2. k. 20 . 1. are equal to a certain number q. 1). + kx m ) n is Z ro~f^ . then 0<r<w and r(q + l) + (m-r)q = n. jS. 3 2 and the possible values of a. 1. Let w a =|a|/J = |a|C-a... Find the coefficient of x* in the expansion of 5 The coefficient = 2 T -^ . &x m for 6. (2. . then ^~ = ^~ and w a = t/ a _ 1 according as a^C~a + l../<: Aave any o/^Ae values 0. .

+(ri-l)c n = 1271 6. EXERCISE IV BINOMIAL COEFFICIENTS If (1 +ic) n ~c + c 1 o.. 38 COEFFICIENTS IN EXPANSIONS 3.^ |2n 7. . r n x" 9 prove the statements in Exx.'. 2 . the coefficient = .. 112. c + 2c 1 x + 3c 2 x z + . + c w _ r c n = n-r . r~l .| 5 | 5 ' | 5 /. 3 + 1.L rr-r 2 2 .-. Find the greatest coefficient in the expansion of (a + 6+c) 7 .+(n + l)c n x n = {I + (n + 4. + c 2 x 2 +. c c r -f C!C r+1 -f c 2 c r+2 + . 2. 3 + r ^. c n-i " .. 3 + <--: .. 2 4 .+(-l 5. c ( >-2c l + 3c 2 -. Here m = 3 and 7^-2 .. . .. c a + 2c 3 + 3c 4 + .. . C . greatest coefficient = | 7/(| 2) 2 . ' . 3 |4|J) ' |2 2 1 ' |3 = 80 + 360 + 90 = 530.... | 3. so that q^2. Ex..

according as n is even or odd. show that the coefficient In of x r is .I 13..l/( | n . 3. 10. -f c^) 11. 1 1 1 1 n-l|l [n ' 14.1 ) n <5 = ... I L .. c n taken two toge ther \2n is equal to 2 2n ~ 1 . c l9 c 2 .. . r . . Using the result of Ex. Using the result of Ex. Show that the sum of the products of c . Henc e show that the sum of this series is j 2n . 12. ~ or zero.. If the terms of the series are expanded and the whole is arranged in powers of x.9. . . ..(3 w ~ r \L\^l. -f ( . 1. Show that c!5 . Prove that if n> 1. show that is equal to the coefficient of x n in the expansion of (n + 2) 2n-l Hence show that the sum of this series is I n | n. H-c n ) 2 -(c^-f c^-f Cg-f-.c? + c^ .I) 2 . [The required sum^JffCo + ^-HCjj-f . show that is equal to the coefficient of x n ~ l in the expansion of n(l +x) 2n ~ l ..

+ C 3n x 3n : (i) Find the value of c -f Cj 4..+c 2n x 2n . .. ""-.. In the expansion of . C 2 =c 8n _ 2 .j. If (l+# + a.+.. . -f c 3n .c 2w /. .. abcde in the expansion of (a + b + c+d + e) 5 . . c i C 2n-i> C r (iii) C Q -f C 2 -f c 4 4.r 3 ) n :=c -f c^ -f c 2 a: 2 4. + C 2n _2 c 2n ~ c n+l [In (i) put I/a. 19. 4. In (ii) put .15. a*b 2 cd in the expansion of (ab + ac + a f d + bc + bd + cd)*. x n(n-l) tt(n-l)(rc-2) . 11. y. x . .] IDENTITIES BY BINOMIAL THEOREM 39 In Ex. Show thatZ l .. etc. .] 18... and write down the last term of this series when n is even and when n is odd..x for a:. (v)c n = l + v ' + . the sum of the series is equal to the coefficient of x n in the e xpansion of n 2 (l^-x) 2n ~ 2 .. according as n is or is not a multiple of 3. 16-18 find the coefficients of the products stated : 16.c 2 + .. n(n~l) n(n-l)(n-2)(n-3) . . prove t hat (i) C = C 2n . Ci^Cg^. a 3 6 4 c 5 in the expansion of (bc + ca + ab)*. and observe that 20. . . for x and multiply by x m . [Write a*b'c 5 = (bc) x (ca)v(ab) z and find x. -f ( -l) n c w is equal to l-. 21.+CrX r +..or zero.= Ci + Ca + Cs + (iii) Prove v that C = c 3n . . . A . 17.P~l2* . If (l+x + x z ) n =c + c l x + c< t x 2 + .v |0 c 2 -~ r^ . (iv)c -nc 1 + ... [Using Ex. (vi) C C 2 C t C z + C 2 C 4 . z. 2 + . !In . . (ii) Prove that C + c 2 + c 4 -f c e + ...7 c 3 -f .

.n ii(l+3C.}. show that (ii)a 2 =4n 2 --2 2 . (iv) Show that the coefficients of the terms equidistant from the beginning and end are equal.n n(n -]){!. By expanding the two sides of the identity and equating the coefficients of x n . the last term in the bracket being C" r or (^ + l)^7(H-i) accorc ^ n g as r is c ven or odd. show that the sum of the | 371 coefficients of the terms of degree r is -..2x} m in two different ways. 22. prove that. (iii) at = \V. In particular. If prove that (i) a r = 2 r .(i) What is the term containing the highest power of x ? (ii) Find the sum of the coefficients. 5C?" 2 4-5 . In the expansion of (l+z) w (l+ y) n (l+ z) n .u -~ ---. 7f7> 2 l ~ 2 4.1) .. (ii) when n is odd. == . 40 HIGHEST COMMON FACTOR 24.n {C* n + CjCj n " 2 + C^Cf ~ 4 + . 3 4-3 . fi2m 9 /->m/nr2w-2 i 92 /-TfW/t2 4 __/nfW ~^ L/ L . show that Also write down the last term of the series on the left (i) when n is even..* 25..l + .l + 5C%. to (n .. By expanding {(1 4-z) 2 . |r 3n-r 23."^ . (iii) Show that the sum of the coefficients of the even terms is equal to that of the odd terms. if m>n. to n terms)..

(3) The H.. then R n _ 2 = R nnl Q n . Highest Common Factor (H. If R n is independent of x. (2) If B is not of higher degree than A. R 3 ) have a common factor. B. so that R n ^ is a .. must vanish identically. PROCESS In practice. the common factors of A and B are the same as those of B and R. If R n Q. B. every common factor of (A. B). can be found such that where the degree of any one of the set B.C. every common factor of B and R is a factor of A and. B) is a factor of J? n _j. say R n . since R~A BQ. this is a common factor of every pair . . (B. (h) if one pair have no common factor. These will be denoted by capital letters. ($ 3 . which is therefore the H. We consider polynomials in a single variable x. R 2 ). or be independent of x.. B).).. (Q 2 . RJ.terms}. B) A (Q. R l . For since A BQ 4.F. Moreover. H. Q. Moreover. the process must terminate and that in one of two ways : the last remainder.C. R are polynomials. every common factor of A and B is a factor of 7?. J? 3 ) . R n -i) have no common factor. is less than that of t he preceding. of A. [We have (1 4-2x4. and the same is true of (A. 41 BQi .R.C. the same is true for every pair. It follows from the last theorem that (a) if one of the pairs (A. as represented below.2x 2 )^ 17.common factor of (7? n _ 2 ^n-i)> an d therefore also of (A. Equations (1) of the last section are found by successive divisions.F. then (J? n _ 2 . (1) // A = BQ + R. polynomials or constants (Q }9 Rj).F. (R 2 .C. Process. (R v R 2 ). where A.F. B). R 2 . the reckoning is generally arranged as on the right.

therefore every common factor of X and Y is a common factor of A and B. A(lm'-Vm) = m'X-mY and B(lm r -l'm) = lY ~-VX. Find the H. Qi etc.C. I'.F. B A BQ. by any constant or by any polynomial which is not a factor of the preceding member of the set. Moreover. m. of X and Y is the same as that of A and B. the reckonin g is as follows : ..C. R l9 R 2 . then the H. . For if we find which.(Q 2 etc. (ii) If we arrive at a remainder R n which can be completely factorised. The process may often be shortened by noticing that : (i) We may multiply or divide any of the set A. the process need not be continued. m' are constants different from zero and such that Im'-l'm^Q. if any. of these factors is a factor of R n __ v we can write down all the common factors of (iii) We may use the following theorem : // X^lA + mB and Y ^I'A + m'B. where Z. Ex. of A = 3# 3 +# + 4 and B = 2x* Paying attention to the remark (i) and using detached coefficients. J5. For every common factor of A and B is a common factor of X and Y.F.. 1.

3+0+ 1+ 4 2-1+0+3 3 6+0+ 2+ 8 6-3+ 0+ 9 2 6-3+0+9 6 + 4-2 3+ 2.1 7 -7 -7+2+9 -7-7 21 + 14.7 21.6-27 9 9+9 20)20 + 20 9+9 1+ 1 (a) .

R 19 and consequently that of A and B. . relating to whole numbers. Hence M is equal to. But M is a factor of AB. I. 3 A -2B = 20x 5 + 15* 3 . A = 12x 5 + 5x* + S and ^8* 5 + 5* 2 + 12. which shows that E t = 3x 2 + 2x . Theorem. 18. Now x + 1 is. Hence the H. 42 PRIME AND COMPOSITE FUNCTIONS Ex. Therefore x + 1 is the H.C. then polynomials X.F. then M is a factor of B.2). and is therefore a common factor of MB and AB. of A and B where. of C and D t and also that of A and 5. Thus.3B = 10* 3 .1 is not.Hence the H. a factor of B.C. of A and B is the same as that of C and D.F.20 . Further.15z 2 . Such expressions have no H. 2(7 .F.C. For if M is prime to A. 4 and 5. Arts. =x + l. Putting C = 2x 3 -So. 2 -4 and Z> = 4# 3 + 3x .C.2. We have 2 A . and 3x . it follows that the H.4). the reader should have no difficulty in making the necessary verbal alterations.C. the theorem corresponding to that of 4 (i) is as follows : If A.1 and x + 2.C. of C -2D and 2C -D is 2x 2 + x + 2 : this is therefore the H.6z 2 + 3z + 6 = 3 (2z 2 + # + 2).Wx* = 5x*(4:X* -f 3x .C. Y. these two have no common factor.1 = (3z . 2.F. Thus x 2 + 3x -f 2 is a composite function whose prime factors are x 4. B. Polynomials which have no common factor (except constants) are said to be prime to each other. of MB and AB is B. But it is unnecessary to go beyond the step marked (a). Remembering the distinction between arithmetical and algebraical primeness. If a polynomial has no factors except itself (and constants) it is said to be prime : otherwise it is said to be composite.3* 2 . F. Hence the H.F.2D . Find the H. 3. of B. M are polynomials and M is a factor of AB and prime to A. Thus 2(x + 1) and 4(cc 2 -f 1) are prime to each other. C . If A and B are polynomials in x.C.1 ) (x + 1 ).F. We can prove theorems for polynomials analogous to those of Ch.5 (2* 3 . JB.D = 6x* + 3* 2 + 6z = 3a? (2x 2 + x -f 2). Prime and Composite Functions.F. or is a factor of.

2# 5 -5o.C. NOTE. 12z 3 + 2* 2 -2LK-4. This theorem is fundamental in the Theory of Partial Fractions. .13z 2 + x + 15. 7-9. IMPORTANT USE OF H..C. .16.prime to each other. X + -f.. are the quotients and B l9 J? 2 . 1-6. 3 -f 9* 2 . In Exx. the H. 6. 3z 4 .F.F.F. can be found such that AX + BY = l or AX + BY = G. 43 Therefore we obtain in succession the following equations : Continuing thus. of the functions in Exx.11 x 2 + I2x + 9. .C.F..-8.F. Now the last remainder is either G. 3 -fl. Y = 1 and G is a factor of A and of J3. thus in either case any common factor of X and Y would be a factor of a constant. 2z 4 . or it is a constant c. where X and Y are polynomials. which are dealt with in Ch. 2. z 5 -f5* 2 -2. AND ITS USE Find the H. we can express any remainder in the form AX + BY. 6* 3 -t-7* 2 -14a. of A and B. for the first equation may be written -^ . . For if Q 19 Q 2 . 2x*-5x 2 + 3.5x 3 -f 2. find polynomials X and Y for which the statements are identically true.C.CJF.. EXERCISE V H. .. Y such that AX+BY^G or In either case X is prime to Y. 3z 5 . In the second case. 2Lr 4 -28a. VII. the polynomials in question are X/c and Y/c. 5. etc. according as A is or is not prime to B 9 G being the H.2z 3 . of A and B in the latter case.C. the remainders in th e process of searching for the H. 3. Hence we can find X. 4.

z. y. product and quotient of two symmetric functions are also symmetric functions. and add : find A and p.z 2 x) (x 2 z -f y 2 x + z 2 y) are symm etric with regard to x. x -f y + z is represented by Hx and yz + zx + xy by Zxy. . Again. C. difference.F. For instance. Thus all the terms of x 2 y -f x 2 z -f y 2 z + y 2 x + z 2 x -h z 2 y are of th e same type. z. Terms of an expression which are such that one can be changed into the other by one or more transpositions are said to be of the same type. and thus obtain B. such that Ax + B Cx + D [Multiply the first identity by A.A. (*-l) 2 Z-(* + l) 2 F=:l. (ii) The sum. process to obtain and hence find A.C. 9. A symmetric function which is the sum of a number of terms of the same type is often written in an abbreviated form thus : Choose any one of the terms and place the letter Z (sigma) before it. and the expression is symmetric with regard to x. suc h that (7A + 10 / Li)/(llA + 28^)=2/(-3). SUBSTITUTIONS 1. then all the terms of this type occur. y.) The interchange of any two letters. 10. B. CHAPTER IV SYMMETRIC AND ALTERNATING FUNCTIONS. y. A function which is unaltered by the interchange of any two of the variables which it contains is said to be symmetric with regard to these variables. Thus yz + zx + xy and (x 2 y + y 2 z 4.7. x.C. Use the H. Symmetric Functions. (In the second expression. the second by /*. 8. It is obvious that (i) If a term of some particular type occurs in a symmetric function. (x + y -f z) 2 = x 2 + f + z 2 + 2yz + 2zx + Zxy = Zx 2 + 2Zyz. the interchange of any two letters transforms one factor into the other. is called the transposition (xy). D.

similarly. y. G = 24 and Za-b \. then -1 a.y) (x + y ~ z) = a . of the second degree. c. F actor ise (x -f y -f z) 5 . 2. Since E-0 when x . and in no other way. where a. Zx 3 + b . if a b c ~ d = 1 .y 3 . /. (ii)puts = l f y = l. Ex. The coefficient of abc is 3 .z 3 -f y-z -f yz. Expand (a + b+c+d)(ab + ac+ad + bc + bd + cd). hence. (iii) put a? = l. the required product is a 2 b + 32abc.y. y. c -2. 3 . z . are independent of x. then l=3a + 66+c.%abc = 12 + 3 .2xyz.x 5 . . z. In this assumed identity. then = 2a + 26. The product is the sum of all the products which can be obtained by multiplying any term of the first expression by any term of the second. as in the following examples. y > z.Considerations of symmetry greatly facilitate many algebraical processes. for this term is obtained as the product of a and ab. 6^1. 1 . Hence the required product is .a: 3 . The coefficient of a z b in the product is 1 ..z = 0. The remaining factor i s a homogeneous symmetric function of #.xy 2 . 2 x z y + cxyz.y) (x -f y . for this term is obtained in each of the three way s a(bc). This expression is symmetric.x) (z + x . z. y 0. b(ac) 9 c(ab). Expand (y -f z . 4 . Sab ^ 4 . ALTERNATING FUNCTIONS 45 Ex. We may therefore assume that (y + z . We therefore as sume that . Test. 6.y 5 .z). (i) put x~ 1. The number of terms of the type a-b is 12. and of the third degree in x. homogeneous.2 5 .'. Hence the terms in t he product are of one of the types a?b t abc. z 1 .24.-f z~x -f zx 1 -f x 2 y 4. it follows that x + y is a factor of E. Ea . y + z and z + x are factors. 2/-1. and the number of terms of the type abc is 4 . so that the test is satisfied. Hence. Ex. Denote the given expression by E.x) (z + x . Test the result by putting a~b=c~d = \.

k. . Similarly y -z and z -x are factors. then E is called an alternating function of x. z^l . thus E = (y-z)(z-x)(x-y).i 6=5 2. c. for the interchange of any two letters. ?/--!. where F is symmetric. of. Thus {x 2 (y-z) + y 3 (z-x)+ z* (x . transforms it into y n (x~z)+x n (z-y)+z n (y-x) = -E. . Denote the expression by E. Observe that the product and the quotient of two alternating functions are symmetric functions. h. . b. .. y. the second into the third... . This interchange of the letters is called the cyclic substitution (abc .x} (x ~ y) is symmetric with regard to x. homogeneous and of the first degree in x. and proceed thus : put x~l. for the cyclic substitution (abed) changes the first term into the second. . y. y. Equating the coefficients of x*y on each side .F.. . and the last into the first. z. Thus a 2 b + 6 2 c -h c 2 d -f d 2 a is cyclic with regard to a. Since E ~0 when x = y. when it is unaltered by changing a into b. z. . An expression is said to be cyclic with regard to the letters a. y = l. z. . c.E by the interchange of any two of the set x. y. Ex. . Such a function is x n (y-z)+ y n (z-x)+ z n (x . we find that k . b into c. then a + b^-lQ.. . y. a ^-5 put x^\. z-0 . 6. 46 IMPORTANT IDENTITIES 3.1. z. and k into a .y)}/ (y -z)(z. . h into k.. //. c into d.y) . . z. It is obvious that .. If a function of x.where a and b arc independent of x. Hence where k is independent of x. z. it follows that x ... say x and y. k). . z. thus E= -(y-z)(z-x)(x-y)(x+y + z). Alternating Functions..y is a fa ctor. /. is transformed into . Cyclic Expressions. c? (in th is order). arranged in this order. then 2a + b = 15. . Factorisex*(y-z)+y*(z-x)+ z*(x-y). 1. y.

9. . must also occur .(i) If a term of some particular type occurs in a cyclic expression. where the meaning of 27 must not be confused with that in Art. Za/J.ca(c-a) + ab(a-b) = -(b-c)(c-a)(a-b). j8. j8. and the coefficients of these terms must be equal. Ex. and quotient. Denoting the given expression by E t we have . of two cyclic expressions are also cyclic. 6. we sometimes denote x 2 (y -z) + y 2 (z-x)+z 2 (x-y) by x 2 (y-z) + .. a + & + c& + c-ac + a-ba + b-c)= -a* -b* - It will be proved later (pp. a 3 (i-c) + 6 3 (c-a) + c 3 (a-6)= -(b-c)(c~a)(a-b)(a + b + c). Thus X 2 (y-z) + y 2 (z-x)+z 2 (x-y) is often denoted by Zx 2 (y-z). a(&3. Factorise a(l -6 2 )(1 -c 2 )+ 6(1 -c 2 )(l -a 2 )+ c(l -a 2 )(l -6 2 )-4 a6c. 96) that (i) Any symmetric function of a. 1. 2... 7. 8 can be expressed in terms of 27a. + . and in proving identities. (ii) The sum. then the term which can be derived from this by the cyclic interchange. product. bc(b-c)-\. it being understood that the second term is to be obtained from the first and the third from the second by cyclic interchange. 4. Za/?y and aj SUBSTITUTIONS 47 This mode of expression is extremely useful in factorising symmetric functions. 1. In writing a cyclic expression. a 2 (b-c)+b 2 (c-a) + c 2 (a-b)= -(b-c)(c-a)(a-b). The student should be familiar with the following important identities. Again. 8. 10. it is unnecessary to write the whole. 95. y. a(b 2 -c 2 ) + b(c 2 -a 2 ) + c(a 2 -b 2 ) = (b-c)(c-a)(a-b). difference.. 5. 11. y can be expressed in terms of (ii) Any symmetric function of a.

c. it is commutative. ) or .= Sa . . in which (ab) is performed twice. is called a cyclic substitution or cycle. the interchange of two elements a. thus (abcd) = (bcda). (2) Two or more substitutions may be applied successively. This process is represented by the operator fabcd\ /abc\ . Zab . E Za . the first is changed into the second by replacing a by c. bdac of a. fabcd\ Also a substitution such as ( _ ).abc + abcZab = (1 -be -ca-a&)(a + b+c-abc).3abc . This is indicated as follows. (1) We consider processes by which one arrangement of a set of elements may be transformed into another. ) . (abcd Thus ST = (bcad)> TS = This process is called multiplication of substitutions. \cabdJ \cab/ \cabj Such a process and also the operator which effects it is called a substitution. Taking the permutations cdba. Now Za6 2 = Za . 6 by a. 6 is called the transposition (ab). As previously stated. Icaoa = bdac.Zab* + abcZab . 6. the order of operations being from right to left. Let S = (a&). . If two operators are connected by the sign = . and we write ( . and is denoted by (abed).*. Zab . Substitutions. c by 6 and leaving d unaltered. The operation indicated by (aft) (aft). Multiplication of this kind is not necessarily commutative. T = (6c). 4. then ^ Sacbd = bcad. d. the meaning is that one is equivalent to the other. but if the substitutions have no common letter. and mi om fabcd\ .4abc. and the resulting substitution is called the product.Za . (abc\ f . in which each letter is replaced by the one immediately following it and the last by the first.

' (6) Taking the elements a. e is changed to g and g to e> giving the cycle (eg). (abed) = (abc) (cd) = (ab) (be) (cd). For (abo) = (ab) (be). and is called an identical substitution. thus completing a cycle. d. (abcde) = (abed) (de) = (ab) (be) (cd) (de). We also have equalities such as (ae) (ad) (ac) (ab) = (abcde) . (5) A substitution which deranges n letters and which is the product of r cycles is equivalent to n . This expression for S is unique. This follows at once from (3) and (4) . thus completing the cycle (ac/). Also b is changed to h. 48 SUBSTITUTIONS (3) Any substitution is cyclic. c to /. As an instance. choose some arrangement.r transpositions. if a given substitution is equivalent to j transpositions. c. Thus. 6. e. and we write S~(acf)(bhd)(eg)(k) or S = (acf)(bhd)(eg). d to 6. the method applies universally. Thus if S = ( a ^ e f^ h ] \chfbgaed/ then S = (ac/) (bhd) (eg) = (ac) (c/) (bh) (hd) (eg) . The element k is unchanged. The same argument letters not contained in this cycle. or is the product of two or more cyclic substitutions which have no common element. and to prove it we introduce the notion of * inversions. S is unaltered and the number of transpositions is increased by 2. and the order of ferent. / to a. Moreover. (ab) (ac) (ad) (ae) = (edcba) .produces no change in the order of the letters. h to d. We shall prove that j = n . and so on.r + 2s where s is a positive integer or zero. making the cycle (bhd). consider the substitution s fabcdefghk\ \chfbgaedk/ Here a is changed to c. and call it the normal arrangement. If we introduce the product (a6)(a&). Next. the number j is not unique. This is a very important theorem. the factors is indiffor in effecting any letter is replaced by the applies to the set of (4) A cyclic substitution of n elements is the product ofn-1 transpositions. substitution we must arrive at a stage where some first. as abcde. .

. Keferring to (5). We say that a substitution is even or odd according as it is equivalent to an even or an odd number of transpositions. then i and j are both even or both odd. Iff. g are consecutive elements. n such interchanges. To determine the class of a substitution S we may express it as the product of cycles. namely. the margin. then j is always even or always odd. .then/ can be moved to the place occupied by g by n + 1 interchanges of consecutive elements. according as i is even or odd. and so we have the theorem stated at the end of that section... and therefore whatever value j may have.. Thus substitutions may be divided into two distinct classes. If f.. Thus bdeac contains five inversions. ec. the transposition (fg) does not alter the position of/ or of g relative to the other elements. It therefore introduces or removes a single inversion due to the interchange of /. and the theorem follows. ba.. (The steps are shown in . Therefore any transposition introduces or removes an odd number of inversions. r. Theorem 2. g are separated by n elements p... ea. and count the number of cycles with an even number of elements : then S is even or odd according as this number is even or odd. Hence the following. If i is the number of inversions which are introduced or removed by a substitution which is equivalent to j transpositions. g..gpq. da. i is a fixed number.) Thus the transposition (fg) can be effected by 2n + 1 interchanges of consecutive elements.. it must be even or odd.xgf.. #. dc. but follows it in the normal arrangement.x <? moved to the place originally occupied by / by . On this account we say that the pair ba con- INVERSIONS 49 stitutes an inversion.Consider the arrangement bdeac.pq . q. but is always even or always odd. . and then g can be fpq . If one arrangement A of a given set of elements is changed into another B by j transpositions..xf .. Here 6 precedes a.. Again. For consider the effect of a single transposition (fg).. In other words : The number of transpositions which are equivalent to a given substitution is not unique. for a given substitution. the minimum value of j is n-r. Theorem 1. Rule.

6. 4.ab) + (c 2 . (ft Y + y 2 " 2 + 2 2 ) ( + ft + y) = 6. Show that the following expressions are cyclic with regard to a. 5.ca) = -(6c-f ca-f a6)(a 2 -h6 2 4-c a -6c~ca 14. (a 2 . Prove the identities in Exx.6c) -f (a 2 .6c) (6 2 . (6 2 . 5-14. 50 EXPANSIONS AND IDENTITIES EXERCISE VI SYMMETRIC AND CYCLIC FUNCTIONS 1. d. 8. 12. (y + z-2x)(z + x-2y)(x + y-2z). 15.(6 3 c 3 + c 3 a s + a 3 6 3 ). taken in this order : -d) 2 .ab) = a6c (a 8 + 6 3 + c 3 ) . 3. (ii) (a-b)(c-d) + (b-c)(d-a). Thus we see at once that the substitution in (5) is odd. ( a . d. 14 to show that 6 8 c 3 -f c 8 a 8 -f a 3 6 8 =a6c(a 3 4. c.cy) 2 + (e x . Show that (be -ad)(ca .bd)(ab-cd) is symmetric with regard to a. 3. .6 3 -f c 8 ).ca) (c 2 . 4.j 7. c is the geometric mean between the other two. (x + y + z)* + (y + z~x)* + (z + x-y)* + (x+y-z) 3 . Expand the expressions in Exx. If one of the numbers a. Also it has 17 inversions.aft) (a 2 . 6. 6. 9. ( j3 2 y -f 0y 2 + y 2 a + ya 2 -f <x 2 -f a0 2 ) (a 4.Or we can settle the question by counting the number of inversions.+ y) =27a 3 + 227a 2 2 + 2ajSyZ'a. 10. (a 2 -f 6 2 -f c 2 ) (x* + y 2 + z 2 ) = (ax + by + cz) 2 + (bz . 2. but this generally takes longer. use Ex.6c) (6 2 .ca) (c 2 .az) 2 + (ay 13. a 3 11. c.

show that (i)2yz=o.a) + (c . r in terms of o. (ii) (y 2 -f z 2 . If a? + y+z=0.b) 4. 22. 6. Simplify the expressions in Exx. 0. z.y 2 ) (x 2 -ft/ 2 . 24-28. c. denote symmetric functions of a. 20.z 2 ) + Sx*y*z* = .b) -f (cr 1 4. and find p.c). c-a y 9 a-b z. (6.1 ) (6 + c .1 ) (a + b ./. etc. q. (a 4.ft. (in) ax 2 + by 2 + cz 2 + 2fyz + 2gzx + 2hxy can be expressed in the form px 2 + qy 2 +rz 2 .b) (b .1 + c.+ y + 8) ( 0y8 -f y 8a -f 8a + ay) =27a 2 0y + 4aj3y8. taken in some order or other. g. and notice that 18. 27aj9.(a .x 2 ) (z 2 -f x 2 . (i) Express 2(a-6)(a~c)-f2(6-c)(6-a)4-2(c-a)(c-6) as the sum of three squares.a) (a . 6.. (x-a)(x-b) (a-6)(a-c) (6-c)(6-a) . [Put b-cx. 21. (ii) Hence show that (6 . form an arithmetical progression. FACTORS OP CYCLIC EXPRESSIONS 51 Prove the identities in Exx. 19-23. 23.c) (c . except when a 6 = c.c) is neg ative for all real values of a. use Ex. c. 8 : also verify each by putting a =/3 = y = 8 = l. (aj8 + ay + a8 + jSy -f j88 + y8) 2 =27a 2 jS 2 -f 227a 2 0y -f 6aj8y8. h. If the numbers x 9 y. 2 ~i/ a -2 2 . y.a) + (c^ 1 + a" 1 ) (c + a . where 27a. 24. 3 to show that 17. 19.16. ()9y8 + y8a + 8aj3 -f ajSy) 2 =:Za 2 j8 2 y 2 + 2ajSy827aj8.

Suppose the symbol i to possess the following properties : (i) It combines with itself and with real numbers according to the laws of algebra. noting that b + c-2a=y-z. c + a -&=i/. (l- 36. c-a=y.. Consider the equation x*-2x 4-5 = 0. (6[Put 6-f c-a=. wherever this occurs. 29. (ii) We may substitute ( -I) for t 2 .} 35. (m) ( 641235 J ' CHAPTER V COMPLEX NUMBERS 1. (U) (246135J . Proceeding in the usual way. a + bz..] 32. Preliminary. or x = l2j~^l. (6-c) 2 (64-c-2a) + (c~a) 2 (c+a-26)-f(a-6) 2 (a + 6-2c). .cx 9 c+a=y. Express the following substitutions as the product of transpositions : ^ /123456X . /123456X ...] 30. we extend the number system so as to include numbers of a new class known as imaginary numbers. a?=l%/^T. [Put 64. which at present has no meaning. 29-35. a-bz. [Put 6-c=a.. /123456X < l > (654321 J . (6~ 31. In order that the operation of taking the square root of a number may be always possible. we obtain the formal solution. [Put 33.. a + 6-c=z.(a-b)(a-c) (b-c)(b-a) (c-a)(c~b) ft-c c-a -6 a(6~c) 14-ca l+a6 * l+bc "" 1+ca Factorise the expressions in Exx.. 34..

Show that if x = l+2i. under these rules will not lead to results which are mutually inconsistent. A Complex Number is represented by an expression of the form x + iy. Immediate consequences of the supposition are (i) (x + ly) + (x 1 + iy') = (x + x f ) (ii) (x + iy)-(x' + iy') = (x-x') + i(y-y f ). together with new numbers which are said to be imaginary. and they signify that the pair of real numbers (x. 1. . The complex number x + lO is to be regarded as identical with the real number x. Ex. 3.iy) (x f + iy') = xx' -yy' + i (xy f + x'y). In order that expressions of the form x + ly may be regarded as denoting numbers.) and in particular -f tQ = (B) Thus the system of complex numbers includes all the real numbers.The reader can verify for himself that (at any rate so far as addition. (iii) (x 4. where x. These things are parts of the scheme used to express numbers of a new class. The sign + does not indicate addition as hitherto understood. Definition of Equality. This may be compared with the fact that two whole numbers are required to express a fraction. and we write x + iQ = x (A.*. multiplication and division are concerned) reckoning. y are real numbers. Thus two real numbers are required to express a single complex number. THE FUNDAMENTAL OPERATIONS 53 2. 1 + itf (x f + iy') (x f . For shortness we write Q + _ /QX and in particular + tl = *. and can only be justified by an extension of the number system. (D) so that as an abbreviation for + il. y) are united to form a single complex number. subtraction. we have to say what is meant by * equality ' and to define the fundamental operations. x 2 -2z + 5= -3 + 4t-2(l+2t)+5=0. nor does the symbol i (at present) denote a number. then x* We have x a = l + 4t + 4i a = l + 4i-4 = -3+4i. We say that .iy f ) x'* + y'* x'* + y'* ' This reckoning is at present meaningless. the symbol i denotes a complex number.

.. (P) so that x + iy is the sum of the numbers denoted by x and iy. 5.. and the equations (i)-(iv) are taken as defining addition. 54 MULTIPLICATION AND DIVISION 6.... In defining these we are guided by the formal reckoning of Art..... Multiplication is defined by the equation x'y) . putting = 0...... for x-x' + i(y-y') is the number to which.... the result is x + ly. The meaning of .. we see that if k is a real number.. Putting #' = &.a .. Subtraction is defined by the equation Hence subtraction is the inverse of addition.. Addition....... (x + iy)k = xk + iyk = k(x + iy) . In applying this definition. (J) It follows that multiplication is commutative.a? ) + 4(-y) . It should be noticed that the terms ' greater than ' and ' less than ' -have no significance in connection with imaginary numbers...) + i (0-y) aB (. if x' + iy r is added.... (H) In particular. we have -y=*(-y) 7........... multiplication and division.... The Fundamental Operations. The sum of the complex numbers x-f iy and x' + iy' is defined as (x + ') + i(y + y') and we write (x + iy) + (v' + iy') = (x+x') + i(y+y') (E) It follows that x+ iy = (x + 40) -f (0+ iy).(x + ty) is defined by the equation Hence .x + ly = x' + iy' if and only if x = x' and y = y'. subtraction..... (K) .. 2/' = in the last equation. ) = (0 . 4. 1.. we are said to equate real and imaginary parts.....

..... For let? where x. multiplication and division (as defined above) are subject to the same laws as the corresponding operations for real numbers.\ z(*' 2 + 2/' 2 ) = and y(x'* + y'*)=Q...... 9.. . Thus the equation defining division is x + iy _xx' + yy' yx'-xy' x' + iy' x'* + y'* x'* + y'* ' . the process of reckoning described in Art.. Putting x'=k.. subtraction..... mi f v x Therefore *- unless x' and y' are both zero.. y' are real numbers.... ( ' excepting the case in which x' + iy' =0. the equations and x + ly mean the same thing.. we have Xx'-Yy' = x and Xy'+Yx' = y.. W e have therefore justified......)= + i ....... . (L) 8.. Hence. if we put x=x f = and y = y' = l....... For complex. . Division is the inverse of multiplication. xx'-yy'~Q and zy' + x'y Q. Zero Products.. is a real number different from zero. and attached a definite meaning to.... x' . as for real numbers.. if a product is zero. 1. that is to say. y... Then xx 1 -yy' + i (xy f + x'y) = 0......... (N) ZERO PRODUCTS 55 It is easy to see that addition...Again..... we have (0 + il) 2 =-l or i 2 =-l .. we see that if A.... then one of the factors of the product is zero... From the last equation we have Equating real and imaginary parts....... by the rule of equality.. y' =0....

+ 1 =0. and therefore x + ty = 0. Fen/y that J( . giving # = 1 or a. we have to 8 = |{ . 2 + a.If a. and therefore z' + 4j/' = 0. Denoting the given number by CD. since x'.1 V3) are ca//ed JAe ^reg cw6e roofe of unity.1 +34^3 -3(t N /3) 2 + (^3) 3 } = |( . Solve the equation x 3 = 1 . 2 + a.' 2 + y' 2 is not zero. = = (2+4) 2 ~4+4i-l"" 3+4i~~ (3+40(3~4t) -9 + 16 + 244 7 24 Ex. y' are real they must both vanish. 56 ARGAND DIAGRAMS Again. we use ' L BAG ' to denote the least angle through which AB must turn in order that it may lie along AC. Express (1 + 24) 2 /(2 + 2 in the farm X + iY. and we write AB + BA-0. Ex.1 + K/3) M a roo o/ x 3 = 1. Thus AB denotes the distance travelled by a point in moving from A to B. 11. then. 2 + a: + 1=0. Thus L CAB = .1+34^3 + 9-34^3) = 1. 1. Examples. 10. it follows that x = 0. y = 0. and EA--AB. Geometrical Representation of Complex Numbers. (1) In naming line-segments and angles the order of the letters will denote the direction of measurement.L BACy and we shall suppose that . 3. Ex. J( . we have in the usual way x = Hence the numbers 1. If x' 2 + ?/' 2 = 0. + l)=0. 2. If a. The equation can be written (#-l)(a.

then X ==sum of projections of Oz. which is essentially positive. 12. and hence r = J (x 2 + y 2 ) . The angle 6 is called the amplitude of z.-7T< (2) If z = x + iy. For if (X. then s represents z + z'. It follows that if \ z \ = then z = 0. . y = rsin0. A diagram showing points which represent complex numbers is called an Argand diagram. (1) Let z = x + ty. Oz' on OX = 05 + X'. amz will mean the principal value of the amplitude of z. zs on OX = sum of projections of Oz. and consequently z = 0. then z = x + iy = r(cos + 1 sin 5). and is denoted by am z. 2. This angle has infinitely many values differing by multiples of 2rr. Thus modz = | z\ = r = J(x 2 + y 2 ). 6) be the polar coordinates of the point z. When there is no risk of confusion. we say that the number z is represented by the point whose coordinates referred to rectangular axes are x. it will be seen that the principal value of am z is L XOz. Unless otherwise stated. where z = rcos#. is called the modulus of z and is denoted by | z \ or by mod z. y. is called the principal value of the amplitude. y = 0. tan = y/x. Addition and Subtraction. With the convention of (1). this will be called the point z. The value of 6 such that FIG. z' = x' + iy' '. Here r. Y) are the coordinates of 5. Draw the parallelogram Ozsz'. Let (r.

Now Os<0z + 0z'. it is required to find the point s n . ADDITION AND SUBTRACTION 57 Similarly Y = y + y'. by the last construction.z') = z. 3. Observe that | 2 . if or if (x or if (xy' . ~O and am (z .2' | = the length z'z.e. (3) // s n = Zt + z 2 -f . Therefore To prove this algebraically. . Y ^ en d represents z-z' . This will be the case if (x + z') 2 + (y + y')*<x* + y 2 + x' 2 + y' 2 + 2>/(x 2 + y 2 ) (x' 2 + j/ /2 ). therefore s represents z + z'.. 4.FIG. (2) Draw the parallelogram Oz'zd (Fig. FIG. i.XOz' i. and am (z + ') = L XOs. For. 5.. the sign of equality occurring if LXOz = L. It is to be observed that | z + z' | = the length Os. 4). which is the case. the number represented by d + z' = z. 4-z w .x'j/) 2 >0. Jf Od. wAere z l5 2J 2 > z n are given complex num bers. FIG. if am z = am z'.e. we have to show that where each root is to have its positive value.

+ s n _ r v Therefore | * n |<K| +| z 2 | 4-. 13. 2j + 2 2 + 23. (C) If z l *=x l + iy^ = fj (cos l -h i sin 0j)........ s 2 s 3 . then s 2 ... ... and only if. s 3 . parallel to. with similar notation for z 2 ...... (1) Let z = x + iy =r(cos^H-tsin0).... For in the last figure. z n . .. z' = x' + iy' =r'(cos 0' + 1 sin 0').. Hence also ) .. z l9 * 2 tS are in the same straight line and occur in this order. the moduli of z v z 2 > 2 s are ^ e l en g^s Oz l9 2 1$2> S 2 S 29 > an( ^ ^at f s n 1 s Os n .... Products and Quotients.... .... 2 3 .. . XOTE...... By the definition of multiplication.... Also Os n < Oz l + 2^2 + s 2 s 3 + .sin sin & + t(cos sin 0' + cos 0' sin 0)} ... we have by (1) of this article. are the points represen ting 2 1 + 2 a .. . Oz 3 ... . (A) Since division is the inverse of multiplication. This is a continued application of the first construction of this article....... and in the same sense as 02 2 . .. To prove this algebraically. +| * n |. etc. (B) For this last is the number which when multiplied by z' produces z.. | s n | ^ | s n ^ | + \z n I < I s n _ 2 1 -h I z n _i I -h \z n I .. == rr'{cos cos 0' ..Draw in succession z^... ^^rr'tcos^ + ^-ftsin^ + fl')} ..... and so on.... by continued application of equa tion ... z n have the same amplitude.... /. if z l9 z t .. The sign of equality is to be taken if. che points O...... Oz 4 . that is. 5 3 5 4> equal to...... . 58 PRODUCTS AND QUOTIENTS (4) The modulus of the sum of any number of complex numbers is less than or equal to the sum of the moduli of the numbers.. 4 = ^{cos(0-0') + 'sin(0-0')} ....

............. ... ...... nearly.|*n|.~2 w = z = r(cos0 + tsin0). by (E). 40=rr/4 approximately....... (E) where n is any positive integer...(A) we have ^ 2 ---n = ^2---M COS (^l^^+---+^n) + ^sin(0 1 4-0 2 +..(D) In (D) put Z 1 =z 2 ==... Similarly. From (D) we see that if Z is the product of the complex numbers z l9 z 2 ....... .r w = KN* 2 |.. Show that few 1 ......... |Z|=V 2 ...... (0) and axnZ --..... (P) DE MOIVRE'S THEOREM 59 Ex.. ..is nearly equal to ... We have (5 + i)=*/26(cos0 + i sin 0)....*.. z w ... then z n r n (cos n0 + 1 sin n0). and therefore (5 + i) 4 = 676(cos 40 + 1 sin 40)..... where tan = 1/5. hence we have cos 40 = 476/676.. ...+ amz n (H) but note that the last equation does not give the principal value of am Z .. (2) Modulus and Amplitude...+0 n )}. .. from (C) it follows that ) . and tan 40 = 1. sin 40 = 480/676.... But (5 + 1) 4 = (24 + 10t) 2 = 476 + 480* ..... 1..

11. a' are any numbers.+0 n < 77. Again.unless -7T<0 1 +0 2 +.. a. .(I) (J) although the last equation does not give the principal value of amz/s' unless -7r<0-0'<7T.= L z'Oz. It is important to notice that the amplitude of z/z' is the angle through which Oz' must be turned in order that it may lie along Oz. (1). z-a . ft the principal value of am . and that. Again. from (B) we have M and am = 0' am 21 am z' . with the convention of Art. if z. z .

with an extension to be given later (Art. From (E) and (F) it follows that if n is a positive or negative integer (cos + i sin 0) n = cos n9 + 1 sin n0.C. those of the point z' are r . 6. Their product is a? 2 + y 2 .the length az the length a'z : and z-a . 6 are the polar coordinates of the point z. 60 COMPLEX ROOTS OF EQUATIONS 14. z-a FIQ. am . z' are called conjugate numbers. . Conjugate Numbers.A. B.0. 7. 16). (1) // z=x + iy and z' = x-iy 9 then z. . This. is known as De Moivre's Theorem. Hence FIG. O am(x + ty) = = -am(x-iy). and if r. (3) De Moivre's Theorem. = L aza.

iy) = a + a t (x + iy) + a 2 (x + iy) 2 -f . . . Find the modulus and amplitude of s/3 + 1. where X.1.ty) = Zax m ( .ijS is also a root of f(x) = 0. Let /(a + ij8) = A + iB. where a. where A 9 B are real. Every term of f(x + iy) in which n is even or zero is real. Find 2 so that 2(3 -f i4) =2 -f 3*.iy) n . we have A = 0. We have /(# 4.(2) // /(z) is a polynomial in z with real coefficients. where every a is real and m. Y are real. The condition that (a + t6)/(a' 4. 3. j8. Let a + tjS be a root of f(x) = 0. . . Every term of f(x + iy) in which n is odd is imaginary. Express in the form JC + 1 7. . 4. V3 . and occurs with the same sign in/(x . where a . Since /(a + 1/?) == 0. and occurs with the opposite sign inf(x-ty). a x . Hence Therefore a .\/3 ~ . Hence the result follows. 2. then these occur in conjugate pairs. B = 0. .iy) = X . are real. and f(z) =f(x + iy) = X + 1 F. then f(x . are real.<s/3 -f *. (3) // an equation with real coefficients has complex roots. . We can therefore obtain expansions of the form f(x + iy) = 2ax m (iy) n ) f(x .iy). EXERCISE VII COMPLEX NUMBERS 1.4 Y. n are positive integers or zero. (i) 2~+3i f iJ (iii) COMPLEX NUMBEEb 61 . as well as the coefficients of the polynomial /(x).ib') may be real is ofe'-a'6=0.

^ 14. (i) If | 2 | = | z' | = 1 and am z = ... . (ii) if | 22.am 2'. Given that 1 + 24 is one root of the equation. 3 -h 8x 2 -7a? + 5= 0. 11. From the formula cos nO + 1 sin n^ = (cos 9 + 1 sin 0) n . Prove that | 22 cos a 4-2* |<1 if |2|<O41. then z'= -z. then # 2 4. y are variables. . 7. ___ 13. then 3 (z 2 -f y 2 ) = 4z . where ri is a pos itive integer. Ifz=x + iy and Z = Jf + 1 7. What is the principal value of the amplitude of (cos 50 4. am z' differ by TT. If 2 x 4.1 sin -C? cos n ~ 3 tf sin 3 ^ + .2 2 1 < 2 | z cos a | 4.. sin n0=(7? cos*.iy> express in the form X + iY Z 9 2~ 1 * 6. Prove that (i) if 2|z-l| = |z-2|. show that . (ii) If | 2 | = | z' | and am 2..y 2 = 1 . . 1 " ( tannf.1 z | 2 < 2r 4. find the other three roots. Prove that |24-2'| 2 4-|2-2'| 2 =2|2| 2 4-2|2 / | 2 . then | 22 cos a 4. Thus.. deduce that cos w0 = cos n -C\ cos n ~ 2 ^ sin 2 + <7 cos n ~ 4 ^ sin 4 .1 sin 50) 6 ? 8. then z' =. . Let z x + ty where x. ' 9.. then . in either case. = . the point z describes a circle.r 2 < 1 if 12. x 4 .3a.1 | = | 2-2 |. . 10.. [Let | 2 | =r.. Find the moduli of (i) (2 + 3i) 2 (3-40 3 . (ii) (2-f 3t) 2 /(3 + 4t) 3 .5.

1 ) r sin 04...cos 30 4. etc.1 sin n = cos ?i0 -Of cos (n -2)0 + 0% cos (n -4)0 . +r n ~ l cos (n .]8 4. z n . If z = cos 4. By putting 2=r (cos -f i sin 0) in the identity prove that 1 -f t coM+r 2 cos 26 + ..B) sin (0 .6 8 + c 3 = 3a6c.sin + sin y=0.. By the method of Ex.z~ n = 2i sin n0. -t-r 11 . Hence show that 2 n ~ 1 cos n 6 = cos n0 + C%coa(n -2)6 + 0% cos(?i-4)04.l Bin n 0==8mn0-C%8w(n .15.y) and 8in34-sin3)54-8in3y = 3sin(aH-)S4-y). 16... prove that z n + z -n = 2 cos nd. and if n is odd n-l (-1) 2 2 n .cos 3y = 3 cos (a 4. then and a 8 4..y) .cos 50 .1 sin and n is a positive integer.1 sin (rc ~ 1)0 ' rsin0-r n 8inn0-f r n " t " 1 sin (n~ 62 ROOTS OF COMPLEX NUMBERS 16.r 2 sin 26 + .1)2 2". and if n is even n ( .r n cos n6 -f r n+1 cos (TI . [Put a = cosa4-t sin a. similarly for (*-ar l ) n -l 17. c = cos y4-t sin y. 18. prove that 16 cos 8 sin* 0= ..] (19) From the identity (x-b)(x-c) (x-c)(x-a) t (x-a)(x-b) (a-6)(a~c)" t " (ft-cXft-a)"*" (c-a)(c-6) deduce the identities sin (0 .. b = cos j3 4.1)0 I-rco&O . If cos a + cos /? + cos y=0 and sin a 4.1 sin j5...2 cos 0. prove that cos 3 + cos 3jS 4. ..

a = cos 2a 4. By Art.2. and only if. any number whose nth power is equal to z is called afc n-th root of z.] 15..1 sin 2a. two values of u will be equal if.1 sin 0). Let ^r denote the positive arithmetical nth root of r. That this may be the case. n-1. from which it follows that u is an nth root of z. ... Consider the values of > s + L s i n I n n ) where & is any integer or zero. where r>0 and -7r<0<7r. Let z = r (cos 4. 13 we have u n = r {cos (0 -I.+ 1 sin \ . PRINCIPAL VALUES 63 Again.sin (a j5) sin (a y) [Put x = cos 20 4. (1) If n is a positive integer.Zkn)} z .i sin 20. namely those given by & = 0.2for) + 1 sin (0 4. the angles must differ by a multiple of 2ir y and the corresponding values of k must differ by a multiple of n. Therefore u has n distinct values. etc. the corresponding angles have the same sine and cosine. Thus the n-th roots ofz = r(cos 6 + i sin 6) are the values of cos- n . 1. Roots of Complex Numbers.

then since (cos 0/q + 1 sin 0/q) q = cosO + t sin 0. Q^ representing the values of /z. v^. q are integers and q is positive.. (2) To find the points representing the n values of %Jz. Let x~p/q where p. 8.. where z Here Oz=r=\/5 + 3~= and tan ^Oz =tan 8 ^= Also fy r= */2* an( * so Ci ^2 ^3 are points on the circle with centre and radius */2. then cos x0 + M*n ^ w one o/ the values of (cos 6 + isin 6) x . such that FIG. we must be able to find the length represented by $r and to divide the angle into V# n equal parts. n-1..+ i sin ~i is called the principal lvalue of ^z.. _ . . General Form of De Moivre's Theorem. 13. 1. /n 0\ . .. 1. | cos. Find the points Q lt Q 2 . 16.. Ex. . If x is rational.n J . where & = 0. 2. . A .^ v ^. The case in which 2 is a positive or negative integer has been considered in Art. therefore i cos 0/q + 1 sin 0fq is a value of (cos + t sin 0)* . v .

The remaining (n . The n-th Roots of Unity. of x"~.. let r-1 and 0=0. the n-th roots of unity are the values of 2k-rr . f 2rn 2r<jr} . given by A = 0.. . n 1. 15.. 1. 15. The n-th Roots of ( .hsm~ where # = 0.1 +a n ~ 2 +. n~l. T n 1 O 1 cos V.1) is cos . If w is even. since j? is an integer. it follows that if a is any root. 18. (2A + 1)7T . 7 1 cos -.. 1.{cos 2r7r/n + 1 sin 2rrr/n} where r = 0. cos p0/q 4.1=0.L. r = n/2. These are the values of (2* +1)77 .+<* + 1=0. 19.1 ) (x 71 . is 1.1) is given by & = -|(n . Since x n . In Art. n n If n is odd. + 1 ). other than 1.+ 1 sin .-ft sm .. (cos 0/q + 1 sin 0/q) p = cos pO/q + 1 sin j?0/# .. .. . .....also.nd x + 1 are given by r = 0.1). 1.2) factors can be grouped in pairs as follows : Since 2r7r/n + 2(n-r)7r/n = 27r. 6 = 77. n n The principal n-th root.1 + x n ~ 2 + .n n This follows from Art. 2kn . 2. . then since cos + 1 sin = 1. the root 1 is given by k \n.1 sin jo0/j is a value of (cos + 1 sin 0) . Factors of x n -1 and x n + 1. (1) The n factors ofx n -l are x . The principal n-th root of ( . This follows from Art. the root ( . then a 71 .1 ).1 = (x . n-1. (i) If n is even. therefore . 15 by putting r = l. 64 ROOTS AND FACTORS 17. the factors x-1 J. the factors corresponding to r and n .where k = Q. corresponding to k = 0. 2.r are f 2rn . . 2f7rl . 2.

The remaining n-1 factors can be grouped in pairs.. (ii) // n is oddy the factor x .. 2 2f7T 2f7T a. 2. Their actual values are 277 . /Q .2) (x*-2x C os&-^ + l) .... [ n n j I n n ] And the product of these is / 2f7T\ 2 .....cos . \ n / n n therefore (A) n / where 77 denotes the product of the factors as indicated. the other is o> 2 .. (D) 20.x -f 1 = 0.. (i) // n 15 even x + l = n^ n . 1.x..h 1 ...{ cos -.1 is given by r = 0..* sm '/ and x ...... except when r is a multiple of 3.. Moreover..(n-1).277 -.1 sm / . These are the roots of x 2 4. (A) Again..2x cos -. co r and o> 2r are the imaginary cube roots of 1. j o Denoting either of these by co.... (0) \ n / (ii) //" n is ocW. The Imaginary Cube Roots of Unity.......) + sm^ = x^ ....... . for (a> 2 ) 3 = (6t> 3 ) 2 = l. . the n factors of x n + 1 are x-{cos (2r + l)77/n + ism (2r + l)7r/n}..{ cos -..... cos i sin or ( -1 ^3)... and (B) CUBE ROOTS OF UNITY 65 (2) Similarly. l + w + co 2 = ... . where 'r = 0. and so .f.

where X=x z +2yz..o> 2 z) (x + o> 2 y 4. (E) x 3 + y 3 + z 3 .... ..y + z) (a... (B)........ (B) The following identities are important......... + coy + co 2 z) (x + co 2 y -h coz) . n + 2.-fc n . (A) ..-hc n a> n . adding (A)...... where r = n... in succession...... TT\ Now 1 +co = l + cos -^... n .........* + ........ 1-f co r + o> 2r =0. we have 1 -f o> = cos ...Sxyz = (x -f...-fc r 4-......... o> 2 for #.+c n x n and o/ the series being continued as far as possible.cos^+tsm... If (l+x) n =c Q + c 1 x + ctf(.. ...ojz) . 3 are |(2 n + 2 cos ~ ) .....)..... XT -i i 2rr. ...zx ............. 2.... show that the values of S lt # 2 ... (F) Ex......l-f-o/4.. x*-y* = (x-y)(x-o)y)(x-aj 2 y) ....=2 cos... (B) 4-. (C) to gether.......o> 2r = .... we have 3(c +c 3 +c 6 4-...... hence.+ 1 sin . since cos ~ =i........ \ * / Putting 1.....27r rt 7r/7r...... (D) x 2 + y 2 + z 2 .xy = (x + coy 4.. .. Prove *Aa* (x* +y* + z* -3xyz)* = X*+Y* + Z* -3XYZ.+c n cu 2n .......... o>.. O O O \ O O / and.+ 1 sin -5..................yz .........0=2 n + (l +o>) n + (l+o> 2 ) n .. 1..... respectively. (C) z? + y* = (x + y)(x + ajy)(x + a> 2 y) ..... (l-fl) n -c + Cl -f.2.. (C) 66 COMPLEX ROOTS If r is not a multiple of 3. Y = This follows from (F) on observing that and Ex.....

' where a.] . 3. . multiply (B) by co a . Prove that =a. O O o tJ Again.. 4. j3 are the roots of z* + z.AI i 01 27T .)=2 n +co 2 (l4-a>) n +o)(l+a> 2 ) n Al /i m / 277 .2x cos^-H l). 2-7T . . 2 (a.. nrr Also 1 +aj 2 = l 4-cos -i sin . Prove that the values of ^ . (C) by o> 2 . and 3 Finally. Ttrr .. (C) by co. v^ 2..+tsm-~ 1 (n-2)ir .^ -^~ . it will be found tha t EXERCISE VIII COMPLEX ROOTS 1. then 8(^+04+07 + .1=0. multiply (B) by o>.1 are ~^(1 t). Find the values of ^(1 + 1) and *J(l-i). Use Art. (n-2)ir ^ . ) (#.. 27T\/ TlTT . W7T\ Also a> 2 (l+eo) n =( cos -^--4 sm If cos. and add to (A) . 19 to show that 1). and (1 +o> 2 ) n cos -tsui . and add to (A) .

] 9. then (ii) if the equation has two equal roots.-f l sin . [If 01 is the unit of length. so that Oz a =2 1 O. prove that Jz=z--={'Jr + x + i*/r-x} or {Vr -f x . and if y<0 we have -7r<8<0 9 etc.COMPLEX ROOTS 67 5. [\/z */r ( cos . Give a geometrical construction to find the points z l9 z 2 corresponding to the values of *Jz. b are complex numbers.i\lr . 10. and 2 2 is on z-f) produced. Also if y>Q then 0<^<?r. then p*-p'*=:4q and pp'ty'* [(i) If a is a real root.q').] . [Put z~x + iy. (ii) In this case (p + ip')*=4:(q + i. z 2 a-*Ja*-b 2 . Prove that. by the rule of equality a 2 Eliminate a. If z ~x + ly =r (cos 6 + 1 sin 0). show that [Let 2 1 =a + */a a -6 2 .x} 9 according as y is positive or negative. y. If a.. | a -6 | -f | a -6 | 2 } = {|a + 6| + |a-6|}M 7. with regard to the quadratic z 2 + (p + ip')z (i) if the equation has one real root. . Oz l bisects the angle XOz.] 6. equate real and imaginary parts to zero and solve for x. and is a mean proporti onal to 01 and Oz . etc.J and cos 6 = . then by Exercise VII.] 8. I *i I 2 + I *$J 2 =i I *!+** ! 2 + i I *i -*2 I 2 =2 I a )t + 2 | a*-6> | .(11 + 20. Verify that the sum of the roots is -2(1 + 20 and the product . Solve the equation f + 2(l+2i)z -(11 + 20=0. Ex.

and for put n0.1^ 3 . prove that x(x + l)(x* + x + 1) is a factor of (x + l) n -x n -l. (iii) Put x =cos <f> + 1 sin ^. No w use Art.(cos . where r=0. . .2x cos (ii) * n + ar* .. n-1. 1. [Put x=Q. 15. . where a. and show that the points which represent them are collinear.2# n cos 6 + 1 ~~ {x n . co. .] 68 COMPLEX FACTORS Show that the roots of (1-f 2) n = (l -z) n are the values of i tan-^. is an imaginary cube root of unity.the roots of z n = (z + l) n .Suvw..] 14.cos nS = 2 n ..2 cos (iii) cos rz< . .1 = /Trio" 1 (* 2 . where r = 0.] 13.(cos 4 t sin 0)} {z n . . If u v = x + y + z + a(z w=x + y + z + a(x + y-2z). 1. Prove that (i) * 2W -2x n cos 194. [The roots are -J( l + i cot J. 2. but omitting n/2 if n is even. -f 2 n^ 2n > prove that a Q + a 3 + a 6 + 15.Sa^z) = w 3 -f v 3 -f.1.y 3 -f z 3 . prove thUt 21 a 2 (x 3 4.. If ( 1 + x + x 2 ) n = a -f a^ -f a 2 a: 2 + . If n is odd and not a multiple of 3.1 sin 0)}.2 cos = /T^I JJ " * (a? + ar * .\ . n 12.1 77^o ~ * cos < .cos + ~ [(i) x m . n . The corresponding points lie on the line x + i=0. 2.

Let z = r(cos0 + *sin0).2.Zfa .(d 4 x). -. L^t 1 be the point on OX which represents unity. Draw the triangle Oz'P directly similar to the triangle Olz. (1) Construction for the point representing the product zz'. r-. 6. the sign of A must be changed. Points representing the Product and Quotient of Two Given Numbers. so that 01 =1. and a. that IB.z 3 2 .2^2 = 0.a>2:2 *.1 .2ft/.OP = rr . then </( + oA+ <o 2 c)= J{> V(a 4. Z B . prove that [Consider the equation (a -f a)"" 1 -f (6 4.2 2 2 4.1 4. z' = r' (cos 0' + 1 sin 0').1 + (c + co')" 1 + (d + w'). 9. OP r . If 6<c.1 -f (6 -f co'). prove that [For x 4. z ..^^s = 0. where a> and w r are the imaginary cube roots of unity. .1 4-t\/3). For by similar triangles OP Oz .co 2 6 + we) = where D = s /( provided that 6>c. [Use Ex.1 .] PRODUCTS AND QUOTIENTS 69 21.] [IT] If 2^ 4. 2 2.3 .z t = 18. For a positive number x 9 >Jx denotes the positive square root. /. c are real.(a + w').1 = 2ar 1 .x)~ l -f (c 4. being an imaginary cube root of unity.z). Then P represents the product zz' . If o>=J( .

and z varies so that z-a i FIG. 13. and are fixed points. Therefore P represents zz'. 9. Therefore ri. az : a'z = k : 1. Then. also Lz'OP^ LlOz = 0. d' divide aa' internally and externally in the ratio k : 1. by Art. LlOP = But zz' = rr f {cos (0 + 0') -f i sin (0 + 0')}. then the point z describes a circle of which a. z-a "k. FIG. For let the bisectors of L aza' meet aa r at d. (3) // k is constant. unless k~I.i FIG. d f . (2) Construction for the point representing the quotient z/z'. 11. Draw the triangle OzQ directly similar to the triangle Oz'l . 10.z' z. by the last construction. . in which case z describes the perpendicular bisector of aa f . a! are inverse points. :. (number represented by Q) . (2). For. Then Q represents the quotient z/z'.

Z ~Q> then . a' are inverse points. > 70 DISPLACEMENTS AND VECTORS Also if c is the mid-point of dd f . K. since ca . if the point z describes a circle of which a.* (4) If z varies so that am z-a where <f> is a constant angle. FIG. az : a'z = k : 1 . say. If & = 1. k to 1. containing an angle <f>. then the point z describes an arc of a segment of a circle on aa'.ca' = cd 2 .Therefore z describes the circle on dd' as diameter. then dd' is divided internally and externally in the same ratio. hence a. \a FIG. we have ca. . then az = a'z\ and z lies on the perpendicular bisector of aa'. z-a For. Conversely. Hence. ca' =cd 2 . 12. a' are inverse points with regard to the circle. 11.

. Quantities which can be represented by lines used in this way are called vector quantities. Barnard and Child. If we draw OL equal and parallel to PQ and in the same sense.' ADDITION OP VECTORS 71 If any straight line P'Q' is drawn equal to. the actual position of the line being indifferent. The sign of <f> determines the side of aa' on which the segment lies. direction and sense. we say that L XOL is the angle which PQ makes with OX. 12 and negative in Fig. (iii) its sense. Velocities and accelerations are vector quantities. An expression (such as PQ) used to denote a line-segment with reference to its length. and in the same sense as PQ. FIG. parallel to. La'za = <f>. i. denoted by the order of the letters. To specify completely a displacement PQ we must know : (i) its magnitude. 13. p. we introduce the notions of displacement and directed length or vector. Thus <f> is positive in Fig. The change of position which a point undergoes in moving from P to Q is called the displacement PQ. This angle determines the direction and sense of (3) Vectors. and if necessary by an arrow. (2) Displacements in a given Plane. the length PQ . (1) In connection with the geometrical representation of complex numbers. 316 : ' Circle of Apollonius. (ii) its direction . Let P.FIG. 14. For by Art. Displacements and Vectors. P'Q' are said to be equal. the displacements PQ.e. * See Elements of Geometry. is called a vector. 13. 13. Q be two points in the plane OXY. 22.

a one-to-one correspondence exists between the number z and any of the following : (i) the point P .. (4) Connection with Complex Number. QR........ RS are any three displacements or vectors (Fig. If z=x + t... Any one of these three things may therefore be said to represent z. (1) Let P. (ii) the displacement OP ....... If a point moves from P to Q and then from Q to R...A force can be represented by a vector ' localised ' to lie in the line of action of the force.. the resulting change of position is the same as if it had moved directly from P to R. We therefore define the addition of Flo 15 displacements as follows : * Some writers use an underline instead of an overline........ (iii) the vector (or directed length) OP.. Q 9 R be any three points. Quantities (such as mass) which do not involve the idea of direction are called scalar. % * 23. (2) The Commutative and Associative Laws hold for the addition of dis- .. 17). 17.y and P is the point (x.... y). Q FIG.. 16. 72 ZERO AND NEGATIVE VECTORS The result of adding QR to PQ is defined to be PR\ and this is expressed by writing PQ+QR=PR . If PQ.. (A) This equation is also taken as defining the addition of vectors. Addition of Displacements and of Vectors. or to be represented by z. FIG.

k. 26. ( So that to multiply a vector by ( . DISTRIBUTIVE LAW 73 25. Definition of Multiplication by a Real Number. Further. For displacements and vectors the meaning of PQ-QR is defined by PQ-QR . The resulting displacement or vector is denoted by JcPQ or by PQ . 23. therefore and the associative law holds. To multiply a displacement or a vector PQ by a positive number k is to multiply its length by A. and 22. If after two or more displacements the moving point returns to its initial position. leaving its direction unaltered.PQ. 16. (i) In Fig. PQ + PS = PR t a fact which is expressed by saying that displacements and vectors are added by the parallelogram law. Thus we write This equation is also written in the form -PQ = QP. (1). These equations are also taken as defining the meaning of zero and negative vectors. Subtraction. NOTE.PQ + (-QR) = PQ + RQThus OP-OQ -OP + QO = QO + OP = QP. In Fig. complete the parallelogram PQRS\ then by Arts.k)PQ by the equation In particular. The Distributive Law. which defines the meaning of . (ii) In Fig.1) is to turn it through two right angles. We shall prove that if k is a real .placements and vectors. 17. Zero and Negative Displacements and Vectors. we define ( . Hence the commutative law holds. 24. and PR is called the sum of PQ and QR. 27. we say that the resulting displacement is zero. 16.

(ii) For a negative number ( -k). then (i) Let k be positive. then | z | is the length AB and am z is the angle which the directed line AB makes with the directed line Ox. It should be noticed that if a number z is represented by a vector AB. Theorem 1. If C divides AB in the ratio n : m and is any point. 18 is drawn for a value of k greater than unity : the student should see that the same result follows from a diagram in which k is les s than unity. It will now be seen that. and Q'R' = kQR\ and kPQ + kQR^PQ' Hence the theorem holds for positive numbers. NOTE. subtraction and multiplication by real numbers are concerned. Along PQ set off PQ' = kPQ. This fact is fundamental in theory and very useful in practice. For mOC^mOA+mAC. .number . complex numbers are subject to the same laws as the vectors which represent them.mOA + nOB. we have IG * and ( hence the theorem holds for negative numbers.nBC . then (m + n)OC . 74 VECTORS AND COMPLEX NUMBERS 28. Also mAC = nCB = . Thus the distributive law holds for the multiplication of displacements and vectors by real numbers. The diagram of Fig. so far as addition. Draw Q'R' parallel to QR to meet PR in R'. Complex Numbers represented by Vectors. By similar triangles.

SYMBOLS OF OPERATION Ex.O whence the result follows by addition. // a. [This is the converse of Theorem 1 of this article. We have (a + c)(W==a. hence a . OQ' in the positive direction of rotation. consecutively.* Along two straight lines at right angles set off. In particular. 6. if z is the mid-point of z^ then z = %(z l + z 2 ). FIG. Ex. OQ. B. then the corresponding numbers are connected by the relation This follows from Theorem 1. Bisect AB at C. AB^c . OP'. then OA+OB=20C and OA-OB^'BA^Z'CA. where a -f b -f c = 0. since C is mid-point of base A B. Let the symbol i applied to a vector denote the operation of turning it through a right angle in . Theorem 2. 19. 1.OA+c. OB. C are collinear. Therefore a + 6 and a -b are represented by 200"and2CJ. OC are connected by the relation 75 ? = 0. in accordance with the principle stated in this article. hence Now. prove geometrically that Let A 9 B be the points which represent a. If z is the point which divides the straight line joining z v z 2 in the ratio n:m. equal lengths OP.OC'. b are complex numbers. 2. BC. Men A. // OA. The Symbol i as an Operator.] 29.

FIG. Either of these results is written in the form ( . where i*OP is an abbreviation for i(iOP). Again. Let c be the mid-point of a6. OP. so that to multiply a vector by (. B.the positive direction of rotation.i) is to turn it through a right angle in the negativ e sense.cos . OQ' = iOP' = i(-l)OP and OQ' = -OQ = ( -l)iOP. Draw P'N perpendicular to OP. . b are complex numbers. The Operator cos0-MSin0. 22. z' so that the points z. //a. z' and a. 21 .A. 76 PRODUCT OF COMPLEX NUMBERS 30. it is obvious that FIG. Draw two equal straight lines OP. then Oc + icb .1 2 OP. i 2 OP is taken to mean t 3 (i 2 OP). find numbers z. we say that to multiply a vector by i is to turn it through a right angle in the positive sense. Along NP set off NQ equal to NP'.( . Thus i 2 and -1 denote the same operation. and in this sense we write Again.C.1)OP. Hence t 2 OP . Also ON = ^ n OP . * For the moment.1) OP^OQ'. Therefore OP' = 1(1 OP) . Thus in Fig. Ex. 2'=!. 6 may be opposite corners of a square. OP' inclined at an angle 0.6). Then OF = ON + NP' = ON + iNQ. the reader should forget his conception of I as denoting a num ber. 1 . To bring our language into conformity with that of algebra.(a + &)+* (a. if i 3 . OQ=i OP and OP' = i OQ. 21. Similarly.

Let OQ^z'OP and OR^zOQ. Here r is the stretching factor and cos -ft sin the turning factor. In accordance with Arts. These are independent of each other. 26. OP + * sin . If OQ is the vector obtained by multiplying OP by the complex number z. and the order in which they are applied is indifferent. OP. AEGAND DIAGRAMS 77 Since the stretching and turning factors may be applied in any order. 32. Hence the operations denoted by zz'.cos . if we take (cos 0-h i sin 0) n to mean that the operation . 31. z'z and rr' {cos (0 + 0') + i sin (0 + 0')} are equivalent./. which we write in the form OF-(cos0+ism0)OP.0). we write OQ^zOP arid OQ/OP^zi we also say that the ratio of OQ to OP is the number z. Division is the inverse of multiplication. so that if OQ = zOP y then OP is the result of dividing OQ by z. The result is the same as that obtained by multiplying OQ by I/z. Therefore to divide a vector OQ by z is to divide its length by r and tnr^ the resulting vector through the angle ( . Again. Multiplication and Division of a Vector by a Complex Number. we say that to multiply a vector OP by the complex number r (cos -h i sin 0) is to multiply its length by r and turn the resulting vector through the angle 0. OF . and we say that to multiply a vector by cos + i sin is to turn it through the angle 0. Product of Complex Numbers. in this order. z are to be applied in succession. OP may be transformed into OR by multiplying its length by rr' and turning the resulting vector through the angle (6 + 6'). 29 and 30 of this chapter. then we write where zz' applied to OP denotes that the operators z'.

and prove that as i varies from . If z = 3 + 2i. Z-Z'. it is required to to a given point z. 01 being numbers of which a. In this sense then (cos + i sin 0) n = cos n0 + i sin w#. Explain the constructions indicate the unit of length. z' in an Argand diagram. z by one of the following equations. 24. FIG. and fin d by geometrical construction.(cos + 1 sin 0) is to be applied n times. 4. ZZ\ Z/z'. 78 CENTROIDS 3. Let z. If a. find the point z corresponding to any given real value of t. the point Z is on the y-axis. 2. the points representing Z + Z'. the result is the same as that given by the operator cos n0 -f i sin n0. It will be seen that complex numbers used as operators on vectors conform to the laws of algebra. If Z=(l + z)/(l-z). z' 1 -f i. EXERCISE IX ARGAND DIAGRAMS : VECTORS 1. find tho point Z corresponding to a given point z t and show that if | z | = 1. mark the points z. b are constant and z varies. (iii) Z= (ii) Z=tz where t is real. a. b are given complex numbers and J^ 6^. b be complex If Z is given in terms of find the point Z corresponding d in the diagrams. (iv) Z~az+b.

prove that as t varies z describes tho line through the point a perpendicular to Oa.. Moreover. and pOA . 9.. p:q:r=& OBC : A OCA : A OAB 9 where the signs of the area# are determined by the usual rule. [Put t = tan J<^.ab. C. the point z describes once the circle with centre c and radius a. If G is the centroid of particles of mass m 19 m z .. . If A. a are given numbers. B.. AB. where t is real.-oo to + 00.it).. If c.. 6. proportional to m l9 m 2 .. then (m l + m z + m 3 + .. :.] 5. b. we may supp ose G to be the centre of parallel forces. at A 19 A 29 A & . D are any four points in a plane. . find the point z corresponding to a given value of cf> ... acting at A l9 A% . then Oz=0a+az~0a + t ..] 7. If z=a(l + it) where t is a real number. z=a + (b-a)t. . .. [Along the line ab set off the length az = t. B. . then AD . [For let p:q:r= &OBC : AOCA : &OAB. . Any three coplanar and non-parallel vectors OA. (7..)s m^ + m 2 z 2 + w 3 z 3 + . Show that this theorem is an immediate consequence of the identity 8. . then . . and O is any point. . To include cases where m l9 m 29 . are not all of the same sign. w 3 . where p 9 q 9 r are real numbers. at z l9 z 29 z a .GO to + oo . B 9 C are collinear if p + q + rQ 9 and conversely. and. a being real. Also the points A. BC^BD . r 9 then G coincides with O. AC + CD . . ab. etc.)OG = m 1 OA l + msOA 2 + maOA 3 4.. and let G be the centre of parallel forces. z describes the entire line which passes through a. acting at A. . OB 9 OC are connected by a relation of the form pOA +qOB + rOC = Q. (ii) if z=c + a(l + it) I (I . show that as t varies from . (i) if z = c+a(cos< + isin<). 10. the segment ah corresponding to the values from to 1 of t... and proportional to p 9 q.. m s . If z is the centroid of particles of mass m l9 w 2 .

then either z l9 z 2 . Prove this algebraically. we have whence the results follow immediately. then OA . if the point z l lies between z 2 and z 3 . and is any point. 10. p' 9 q'.r'y = 0.VECTORS AND COMPLEX NUMBERS 79 11. p:q:r~ A0z 2 z 3 : the signs of the areas being determined as usual. where p 9 q. 15. Theorem 1. q. Let Zj. B 9 C are collinear. the vertices occurring in the pos itive . 28. CA + OC . . or e lse O is the centre of a circle which touches the sides of the triangle Z 1 z 8 z 3 . z 3 are collinear. z a are collinear if p + q + r Q. r. p' 9 q' are real numbers. 11. Moreover. Any three complex numbers z l9 z 2 . j8 are non-parallel vectors andp. AB^O. no two of which are equal. z 2 . then (i) If the points z l9 z 2 . If A i^4 2^4 a is an equilateral triangle. z 3 are connected by a relation of the form where p. (i) //a. that is [These theorems follow at once from Ex.] 14. [If z l =x l + iy l9 z 2 =z 2 + iy l9 z s =z 3 + it/ 3 . z 3 are collinear.] 12. r' are real. Also the points z l9 z 2 .] 13. q. then these are one and the same equation. y are non-parallel coplanar vectors connected by the equations p<x + qfi + ry = Q and p'a + q'fi -fc. the ambiguous signs are both m inus. m : n : m + n=CB : AC : AB. l \ Z 2~*3\ 2|2~2l ! *8hl-2l| =Also. If the points A. and deduce Ex. (ii) If the above equation holds. [In Art. r are real numbers. (ii) If a. adeq uation of the form involves the two equations p=p' 9 q=q'. z a be complex numbers. EC + OB . and conversely.

r. . The triads of points A 9 B 9 C and X.c) + y(c . CA. (cos 30 . x c y a z b [By Ex.] AC XZ c-a z-x 18. if . prove that s ~ 4.e*. AB. 16. If ABC is a triangle and triangles BCX. the centroids of XYZ &ndABC coincide. CA Y. all outwards or all inwards.Z& = 0. 3 3 80 TRANSFORMATIONS 17.z^ . A R =AB . [Let P. R be the centroids of the triangles drawn outwards on BC.. i = . Hence show that# + v + z=a + & + c. ~ . a' 9 then the numbers represented by X 9 X' 9 Y 9 Y' are given by where 6 has the values ~. Equilateral triangles are drawn on the sides of a given triangle ABC. Also.-. [The triangles are directly similar if = . z 3 are the numbers corresponding to A l9 A t9 A B9 prove th at 3 X -f wz 2 + cj*z 3 = 0.= .= . Z are the vertices of directly simi lar triangles if the corresponding complex numbers a. y. Prove that _ _ I __ __ i QA = CA (cos 30 + L sin 30). \/O *Jd Hence show that QR \CB + ^ (CA -AB).b) =0. if z l9 z 2 . Q. ABZ are drawn on BC. b. AB. Prove that their centroids form an equilateral triangle .direction of rotation.l 1 J b-c c-a a-b y J 19. z are connected by the relation x(b .Z& .1 sin where o> is an imaginary cube root of unity. i.z a 2 + z 3 2 . and consequently z^ 4.a) + z(a . directly similar to one another. CA. and that RP has a similar value.L sin 30). 17. If A X YA'X'Y' is a regular hexagon and A 9 A' represent given complex numbers a.. Y. c and x.

a x n + c^z". O being the centr e of inversion. Show that each of the substitutions in Ex. being the centre of similitude. Under this heading we consider equations of the type f(x) =0 where /(x) is a polynomial. + a n = . and ( equation ' will mean an equation of this kind. (iii) If Z = (cos a-f tsin a) z. (v) If Z = l/z. 15. b are given complex numbers. . In this case we say that S is a magnification of s. if c passes through O. in an exceptional case. 2 and if z describes a given curve s. except that in (v). -f j0 n = 0. the curves s.] CHAPTER VI THEORY OF EQUATIONS (1) 1 . If Z. 20 converts a circle c into a circle C. 22. S can be obtained from 8 by a translation. S can be obtained from s by a rotation.2 + . 21.? -. (i) If Z = z -f a.1 + a 2 x n ~ 2 + . r? 9 therefore the transformation is equivalent to one L a a az + b ^ or more of those in Ex. (iv) If Zaz -f 6. then Z will describe a curve S. Explain the following. .Hence show that EQ = EP (cos 60 + L sin 60). The general equation of the nth degree will be written in one of the forms : x n +p l x n ~ l + pfl n . a magnification and a translation. S are similar and similarly situated . (ii) If Z=tz where t is real. into a straight line .] ' TRANSFORMATIONS ' 20. and use Ex. where a.. S is the reflection in OX of the inverse of s. then C is a straight line. Show that the substitution a'z + b' converts a circle into a circle or. S can be obtained from s by a rotation about through L a.. . [Z = ~> H --. 20. Roots of Equations. z are connected by any of the relations in Ex.

+ a n = 0. f(x) is divisible by x a.. and will be proved in another volume. 14). For let f(x) = x n +p l x n ~ l + . If f(x)^(x-a) r .a z x n ~ 2 + . we can show that where there are n linear factors on the right./J. /?. .. a is called an r-nndtiple root otf(x)=^Q.M (yt 1 \ or. then. (3) Multiple Roots. y. Hence f(x) = has n roots a. we may therefore assume that Let /J be a root of <f>(x) =0 . and no others. . we say that a is a double root. Relations connecting the Roots and Coefficients of an Equation. Thus a polynomial in x with real coefficients can be resolved into factors which are linear or quadratic functions of x with real coefficients. Therefore/(cr) is divisible by that is. . where binomial coefficients are introduced. .. A. a n $x. 82 ROOTS AND COEFFICIENTS (2) Imaginary Boots.. . a n are the roots of the equation . Theorem. The last equation is written briefly in the form (a 0> a 1? a 2 . as before. For the present.. This is the fundamental theorem of the Theory of Equations.. . so is a . (1) It follows that every equation of Ike n-th degree has exactly n roots.t/J (Ch. When r 2. and we may assume that Proceeding in this way.1 + -~ . Let the coefficients of f(x) be real . if a -Mj8 is a root. . by (x-a) 2 + 2 .+p n and let a be a root of/(x)=0.. . (f>(x) where </>(x) is not divisible by re -a. //a 1? a 2 . a Q x n + na^. <j>(x) is divisible by x . l) n = 0.. V. we assume that every equation has one root. By the Remainder theorem.

n at a time..2x .. y are the roots of x^-x-lQ. TRANSFORMATION OF EQUATIONS 83 Let y = <f>(x) and suppose that from this equation we can find a. a n satisfy the equations (B)...( .l +pzX n -*+ .. a 3 . // a...Zaj . <x.. x n ~ l + 27a.r 4. a 2 .a 2 . which we denote by <f>~ l (y).. +t) w = 0. 1..... (A) and. equating coefficients.. thre e. . 2.-|) =0 .. ^y + ya + a/J.. two. For under these circumstances the identity (A) holds... ffoiy are the roots of x n -\-p l x n ~ l 4-y 2 x n ~ a + . where <f>(x) is a given function of x. (B) Conversely. Transformation of Equations. a jSy.. .......then the sums of the products o/a t . and suppose that we require the equation whose roots are ).x 11 . +7> n =^0.(#--a. ..a n = (-l) w y n . Let a. 1..l)... )9. 3 + x* .... . which is the equation required.. Write the equation in the form x 3 -f -|# 2 4.. . .. + p M ^(ff-a. n .y. j3.. a n from equations (B) is a -f a~ x -f . j8. x n ~ 2 + . . Ex.1 =0.. l ) . be the roots of /(z) = 0.. Transforming the equation /(a?) =0 by the Substitution x=$~ l (y) 9 we obtain /{^"^(yJJ O. ..( . find the equation wJiose roots are .. y. A case in which x is not a single-valued function of y is given in Ex... a 3 . write dQwn the valu er of a + /? -f... . It follows that the result of eliminating a 2 . as a single-valued function of y. we have .)(x-a 2 ). are respectively equal to -Pi> Pv -P*> > (~ 1 )>rr For x n +piX n . . if ot^ a 2 . // a.. 3.. j^a.. y arc Mo roote o/ 2a. taken one.

. 4.A.. A*j8. . equatio n whose roots are a 3 . and the transtormation is called * diminishing the roots off(x) =0 6y A.3p 3 a ) y + p a 3 0. is/( -x)==0 .v=rc 3 . j. y . . Let a.-?/3 where . y 3 . .^+Y^-l 1-a 1-J8 1-y Ex. ft . for the result will be the same whichever cube root y* stands for.h. Let .. if # 3 .' . .Z 3 f w* + n 8 Hence the required equation is which is the same as V s + (Pi 3 . Ill. The required eq uation is therefore obtained by rationalising )-y+ p l y + p^ 4 p. -y. 2. find the. then a.. 1 a * ~p 1 y Let y~ ..1 0. then (1) the equation whose roots are -a. (3) the equation whose roots are kot. ~j8. jS 3 .. The following transformations are often required. (2) the equation whose roots are I/a. pM* ~m. If a. is f(x/k)=0. let y+p* l. j9. Hence ivrite down the value of 27(1 -f a)/(l -a).Ary. We find f(x + h) as in Ch. 1//J. 1/y. and.9 ^ . . .3p!p 2 p 3 4. To rationalis e the equation. then x . then Z-f-m + n = 0.x . y.--. = . y fere Me roofs off(x)x? 4-piX 2 +p z x+p3~Q. Hence also ' i+v + i+P. pdj* n.-.3piP 2 + 3p 3 ) y 2 + (p 2 3 .A is /(x -f A) = 0.1 /y-l\ y-l UTIJ which is equivalent to This is the required equation.. and therefore .. then y is given by l x y H../3 denotes a?^y cube root of y.' (4) The equation whose roots are a . jS. be the roots of f(x) = 0. 4.. Special Cases.. is /(1/rr) =0 . This transformation is called * multiplying the roots off(x) Q by k..

3.1 -0. + 6^0. . y are Me roots of 2x* + 3x* .4 3. .. 1.( .3 x 2 x . 2y is 1 . pro ceed as follows : Ex.3) ) 1 -h 3 . The equation whose roots are 2a.3 + + 6 . Putting x yjk and multiplying by & 3 . + 2 -8 -6 Wo diminish the roots by 6/3 = 2. j&a:. 2/S + 3.1 -f . +2 which is the one required.7=0. find the equation whos e roots are2a + 3.a: . Consider the equation of . . Write this ar> .168 = where y = 12a. The least value of & which will make every coefficient an integer is 12.^ =0. // a.f z 2 + fa: . "J ITJ 173 _^_ g The reckoning on the right shows that . 2y-i-3..6 2 + 5z + 12 = into i_2) 1 -6 4-5 +12 an equation lacking the second term.2 . 2.^_2 -4 the resulting equation is \ T2 ITy o: 3 -7a. the equation becomes f-\W +^y -!* =0.84 SPECIAL CASES OF TRANSFORMATION (5) The second term of f(x)^a& n +ai$ n "~ l + aiX n -* + . so that the required equation is y 3 .. Transform x* . For f(x -f h) = a^(x + h) n + a t (x + A) 11 .9y 2 f 90y . 1 jr. )8. +a n = can be removed by diminishing the roots by ajna^. so that the coefficient of x n ~ l iuf(x + h) is na A+a 1? and this is zero when (6) To transform /(#)=0 into an equation in which the coefficient of the first term is 1 and the other coefficients are the least possible integers . 2j3.

. the required equation is ^ J?*.. ""^"19 Find the H..... CUBIC AND BIQUADRATIC 86 Ex... (A) If x = y + h.. Therefore x . j8. and hence ....a is a common factor of f(x) and and may bo found by the H.7* .a is a common factor of ^ 2^ + 19z 2 + 53z + 36 and 2a^ + & 2 . a. a and j5 of f(x) .. a + 3..F.... Let the roots be a. therefore = -1 . 1+0-2 + 2 -3 + 9 Increasing the roots of this by 3 (as in the margin..C.. The Cubic Equation.a*d . and the equation is ^ + ^ + 03 = 0. namely a..are connected 6y i^e relation ft~ ^S(a) we can generally find them as follows: The equations /(se)=0 and /{$()} =0 have a common root. Again. of these expressions.. j8~3. (B) where H = ac-b 2 .1 and 2 are roots. The + 6 + 21 + 42 reckoning on the right shows that this equation + ? . and the third remaining root is .. If however f(x) and/{0(a:)} are identical. then JB = 0. the product of the three roots is 3 ..Ix .......6.. Find the 2 + 1 ~ 7 ~ 6 equation whose roots are a -3.. o.3abc + 2R . 8 . 4..C. G ... // two roots.. then the method fails. this becomes where B = If h= -b/a.. this proves to be # + 1.F.. process. given that the difference between two of i ke roots is 3.... + 13 +53 Hence....... 5.3 /2..6 =0. or ^ + 3^-2^-4=0. x ...sfiolve 2x* + a..~'+3--j-g-l=0.. ....2. where "j _ 3 + 7 we divide successively by x + 3). We take as the standard form wsa^ + 36z 2 + 3cz + d!:==0 .....

...Or. If x^y-bja the equation becomes .1. l/2y ..13# 2 + 9# + 9 =0... y.. + e = .... (ii) a..4a 2 6d + 6a6 2 c ... y are the roots of 2ar* 4... (A) The function u is called a quartic.4x 3 . ay + 6. y + 6/a.36 4 .1 . + i. (iii) !/(!-).. jS..i 2 .. l/(l-0).. j3.0 + 2.. given that the sum of two roots is zer . (C) If <x. 2 + 4da.. y + 2. 6 1 = a 2 d .. those of (B) are a + 6/a. The Biquadratic Equation.... y are the roots of 80^ .3a6c + 26 s (as for the cubic)...... 2y + l.. 6# 40 K (B) where T ac . (ii) 2a-f 1.. /J + 6/a.. those of (6) are a + 6/a. a"" 1 are these roots.....3x* + 4x .. y. given that the product of two roots is unity... find the equations whose roots are (i) l/2a. 86 EQUATIONS WITH CONNECTED ROOTS EXERCISE X TRANSFORMATION OF EQUATIONS 1.4# 2 -f 6# -1=0.1.. Solve 4s 4 . y + i. etc.. 8 are the roots of (A). y\ [(iii) should be deduced from (ii).1... If a. the equation is z 3 + 3ffz + <? = ..x... (a.. Solve x*-2x*-3x 2 + 4x-~l=:Q... 20 + 1. (C) If a..1=0... l/(l-y).. ]8... [If a.. )8. and those of (C) are aa+*6.. If a.. and K = a 3 e .1 and x* .a)(x . 0..] 3.. y are the roots of (A).... the equation is s 4 + 6#z 2 + 43 + #=0 .. We take as the standard form w = ax 4 + 46x 3 + 6ca..4x 3 + 3z 2 + 2x . find the equations whose roots are (i) a + i...3# 2 . (v) a 2 . a]8 + 6....2x* . 2 . (iv)a + 2. etc.or 1 ) is a common factor of x 4 ...... . a nd the roots of (C) are aa + 6.......] 4. [For (ii) use the result of (i).] 2. If z = ay = ax + 6.. if we write z = ay = ax + b. 6. 1/2)3..

[Let the roots be a . whose roots are in G.] 5. . 14.] 6. ap.3t/ 2 . [Let the roots be -38. whose roots are in A.'. a . Transform into equations with integral coefficients : (u) ^13. Solve 2# 4 . 11. .] 7.15 =0. . whose roots are in A.a are these roots.o. Solve a? 3 -7x 2 + 36=:0. .13z 2 . 10. We say that the character of the roots is known when we know how many are real and how many are imaginary. as giving tan 6 in terms of tan n$. 3a =^.lit/ + 6 = 0.9s + 9. given that the roots are in harmonical progressions [Solve 2# 3 . Solve a.15# 8 + 35o: 2 .] GENERAL THEOREM 87 7.] 9. Solve 6x 3 . 4 .3Qx + 8=0. Character and Position of the Roots of an Equation. given that the difference between two of the roots is 5. a. a-f-38. a-f 8.Ilx 2 -3x4-2=0. Transform into equations lacking the second term : 12. P.] 8. a/? 3 .8# 3 -f 14# 2 ^ 80. prove that Un n0 = (-l)Kn-i)tan0tan(V-Wn^ \n/ \n/ \ n/ V n ) [Regard the equation of Exercise VII.P. Solve 4# 3 -24o. (x ~-oc)(x + a) is a common factor of 4x*-4x*-13x* + 9x + 9 and 4* 4 + 4a?. a + 8. 2 -f 23^+18=0. Solve 4x 4 -4a? 3 -25aj 2 + a:-h6=0. given that the difference between two o f the roots is unity.P.S. given that the roots are in arithmetical progression. [If a. If n is odd.8. [Let the roots be a/>~ 3 . a/)" 1 .

a) are positive and /(O) is negative. any polynomial s its highest term. Whence it f pllows^tbat for sufficiently large values of x.. (2) Let a and b be any real numbers : then (i) ///(a) andf(b) have like signs. and is determined roughly for a non-integral root by finding two consecutive integers between which the root lies. the sign of f(x) can change only when x passes through a real root of /(x) = 0. or else there is no root between a and b.1) 2 + m 2 . {x . If n is even. .. /(a) and/( -a) have unlike signs and at least one root lies between a and . If l + im is a root of f(x) = 0. (ii) ///(a) and f(b) have unlike signs. we require Sturm's Theorem (Ch. /?.The position (jf a real root is its position on the scale of real numbers. -f p n . 88 NUMBER OF REAL ROOTS . Here we suppose that f(x) = X n +p 1 X n ~ l + p 2 X n ~~ 2 + . it has at least one real root. these being not necessarily all different. then l-im is also a root (Ch. Also this expression in factors is unique. K are the real roots of f(x) = 0. The last expression is positive for all real values of x.<l>(x).. where <f>(x) is positive for all real values of x.(x-K).. (1) If x is greater than any of the roots.and p n negative. are real numbers. a positive number a can be found so large that /(a) is positive and/( -a) has the sign of ( . if x varies.(1 4. f(x) is positive. Some general Theorems. . /(a) and /( . an even number of roots o//(z) = lie between a and b.a. where p l9 p 2 . Hence if n is odd. the equation has at least one positive root and at least one negative root. For a complete discussion. . Hence. f(x) = (x-*)(x-p). so that at least one root lies between and a and at least one between and -a. For. V. by (1). an odd number of roots off(x) = Q lie between a and b.iw)} = (x . 14) and f(x) has the quadratic factor. (ii) Iff(x)~Q is of even degree and p n is negative.(I . This is also the case when all the roots are imaginary. 8. . and we draw the following conclusions. but a good deal of information can be derived from the elementary theorems which follow...im)} {x . (3) (i) ///(#) = is an equation of odd degree. Thus if a.l) n . XXVIII).

The equation a x n + a l x n " 1 ^-a^c n ~ 2 +. v has at least as many changes of sign as u. In what follows.+ -Signs of terms of v. If a r = we say that the corresponding term is missing.-{ a n =0 is said to be complet e when no coefficient is zero. even in the most unfavourable case in which all the ambiguities are continuations : and on account of (iii) v has certainly one more change of sign than u. we shall show that if u is any polynomial and v~u(x-a). In considering the signs of the terms of a polynomial taken in order from left to right. On account of (i) and (ii).+ where indicates that the sign may be -f or . or more negative roots than f( x) has changes of sign. (ii) Unlike signs precede and follow a single ambiguity or a group of ambiguities.9.2x? . or the opposite sign. -4Z 3 . + --.. Descartes' Rule is as follows : The equation f(x) = cannot have more positive roots thanf(x) has changed of sign. and p! the number of continuations.a change of sign becomes a continuation and vice versa. we say that a continuation. To prove the first part. +-+. where n is the degree of the equation. every equation is supposed to have a term independent of x. or a change. Descartes' Rule of Signs.+ + 4---. In a complete equation : (i) Ifx is changed into -x. of sign occurs at any particular term according as that term has the same sign as the preceding. +5z 2 . so that zero roots do not occur.. and the changes at -2Z 6 . or that the corresponding term is zero.4X 3 + 5x 2 . (iii) A change of sign is introduced at the end of v. 4-7. NUMBER OF IMAGINARY ROOTS 89 . (ii) If p is the number of changes. then v. the rth sign of v is ambiguous. when expanded.+ + .' = n. Thus x 1 . observe that (i) If the rth sign of u is a continuation. then p -f p. and 4 changes of sign.6x + 7 has 2 continuations. where a is positive. of sign. has at least one more change of sign than u.3xt . First suppose that no term is missing in u and consider the following instance : Signs of terms of u.+ .. the continuations occurring at the terms -Sz 4 . In the diagram of corresponding signs. -6z.

.s =0 has one positive. m the number of positive roots of /(#)=0. Again. x-fl. m' the number of negative roots of f(x) =0 . and [JL the number of changes of sign in/(x). /(a. Hence the second part of the theorem follows from the first.. the negative roots of/(x) = are the positive roots of/(-x)=0. Show that x? -2# 2 +7 =0 has at least two imaginary roots. which is impossible. with their signs changed. r.' '. 8. j8. then ra=ju.+ Thus v has at least one more change of sign than u. // q. . (x . +//<w. therefore by (A) m + m f =/z +fi'.+ -+-0. where a. . Therefore w=/u. 1.-+00--00+ +-00 + + . Hence there cannot be more than t hree real roots.. show that the equation f(x) -x* 4. Again. are the positiv e roots of /(z)=0. and m'^p. Here /u = 2.)=0 has a positive and a negative root. namely. and m' =/// '.(p. Hence m + m'^fi +fjL'^n. m^p. Next let f(x) = <f>(x) .(^+/A') imaginary roots . Corollaries. the equation /(#) =0 has at least n . Let n be the degree of/(x). ^' = 1 . and. . 10. two positive and one negative. if all the roots are real m + m' = n. . (ii) if all the roots o//(x)=0 are real. therefore these are the only real roots. therefore /Lt-|'ft-'~ 3.j8) .. s. If </>(x) is multiplied in succession by a? -a.(m + m')^n . +/u/). if w</u. fji f the number of changes of sign in/( -x). . Ex. each multiplication introduces at least one change of sign. Now m^fji .a) (x . (A) and the number of imaginary roots = n . .. Also /* 1 and // = ! . By Art. and m'=///. For by Descartes' rule. Hence /(x) has at least as many changes of sign as/(x) = has positive roots. one negative and two imaginary roots. Ex* 2. it follows that m'^>^ . are positive..qx 2 +rx . then (i) ifp. (3).That no changes of sign are lost on account of any terms which may be missing from u appears on considering such instances as + +00.

c r . Thus the total number of changes of sign in f(x) and /( . . . /x = and // = 1 . (ii) if h. then (-x) m and (-~x) m ~ r ~ l have opposite signs.1 .x n ~ 2 -f . " J Suppose that the r terms between hx m and kx m ~ r ~ l are missing from/(x). For by the substitution x~y~ 6/a. DeGua'sRule.(/x +/z') =r ..(/x + /z') imaginary roots. If a group of r consecutive terms is missing from (i) if r is even. and (i) if h.6 2 > 0. /*' = number of changes of sign in h( ~x) m + k( -x) m "~ r ~ l . and f l (x) = t/j(x) + (f)(x).%. 90 EFFECT OF MISSING TERMS 11.. (-x) m and ( -#)-' t ~ 1 have the same..Therefore the equation has at least two imaginary roots.. 1 . and in both cases If r is odd.x) is less than the number in/^x) and/ x ( -x) by r + 1 -(/u..1 + c 2 x m ~ 2 4. the equation becomes # n + . c r x m ~ r + kx m ~ r . being zero. and -^ a 2 . Let /x = number of changes of sign in hx m + kx m ~ r ~ l .. First suppose that r is even. according as the terms which immediately precede and follow the group have like or unlike signs. k have unlike signs. and (i) if h. there are at least r + 1 or at least r .c. sign.. This completes the proof.1 imaginary roots. The number of changes of sign in ifj (x) + the number of changes of sign in 0(-z) is r + 1. /// = and r + 1 . Let f(x) = hx m + kx m ~ r .l) w -0 /wo imaginary roots.. If H = ac . the equation has at least r imaginary roots : (ii) if r is odd. <Ae equation (a. .6.. k have the same sign. -\-JJL') = r + 1. .1. where ^ (x) = hx m + c^x. none of the set. c l3 c 2 . k have unlike signs. /it' = 1 and r + 1 .l + <l>(x).(p. k have the same signs. (ii) if h. =0.-f/x'). p = 0. Hence f(x) = has at least r + 1 . /x = 1 and p =0. ft = 1.. .

Method of grouping terms. whatever is said about the roots of an equation refers to the real roots only. this method can be made to yield quite close limits. Ex. Now <(*/) =5*/ 2 (20*/ . and in Ch. For if a is the least root of /(x)=0. and if y has this value. Here Qf(x) = 6# 3 .100=0. h f > -a. (1) In this article. XXVIII. and therefore -h'<oc.60s.15) + x(x* . it is advisable to begin by finding two numbers between which the real roots lie.IQy 10. Other methods are given in the next exercise. (ii) If h" is an upper limit to the positive roots 0//(l/x)-=0. it is often advisable to multiply j(x) by some positive integer.6frr . (2) Upper Limits.by De Gua's rule this equation has at least two imaginary roots if H > 0. then -a is the greatest root of ( -a. (3) Lower Limits. 2 . To distribute the terms conveniently. 12. Thus f(x) >0 for x^l& and 15 is an upper limit to the roots. The process consists in arranging the terms of the equation /(x) = in groups.10* 2 . provided that/(z)>0 when-x^h. 1. Also. UPPER AND LOWER LIMITS 91 In searching for the roots of an equation.llx . the equation is supposed to be written with its first term positive. (i) If li is an upper limit to the roots of f( x) 0. the equation becomes <f> (y) = lOOy 3 -f lly 2 -f. Again.)=0. 20t/ 1 > 0. Limits to the Roots of an Equation. With a little ingenuity. Find an upper and a lower limit to the positive roots of /(z)=z 3 . then h is called an upper limit and I a lower limit to the roots. If k. then I/a' is the greatest positive root of /(!/#) =0. Hence. then I/A" is a lower limit to the positive roots off(x) 0. then li is a lower limit to the roots <>//(#)= 0. Any number h is an upper limit to the roots of f(x) = 0. so that we can find by inspection a number h such that the sum of the terms in each group is ^0 for x^h.66) + (x 3 ~ 600) .600 ^ 4x~(x . 1-4 The greater root of 16?/ 2 + Hty .1) + (16r + 10y . Hence A">l/a' and so l/A"<a'.1 -0 is ^^ . I are two such numbers and A>Z.1). as in the examples at the end of this article. putting x = l/y. For if a' is the least positive root of /()= 0. .

520 = 0. . To find the Rational Roots of an Equation. Now p is prime to <?. k a re integers or zero and p.3 = 0.z) = (3s 4 . h. Ex. .x(6lx* -f 127a: .(220* . q are integers prime to one another. and the only values of q/p which lie betwe en and | are 1. Search for rational roots of f(x) = 2x 3 . Thus 18 is an upper limit to the roots. Hence 4 is a lower limit to the roots of /(a. . We can test the values of q/p by synthetic division.. + k = must be included among the values of q/p. . If qjp is a root. Ex. 2 . Also /( . therefore f(x) . Again. Testing these by synthetic division. We have f(x) ~2x 2 (x --f) +5(. G B. Any such roots which may exist can be found by using the following theorem : If px~q is a factor of ax n + bx n ~ l + .r . or we may use the method of divisors given below.has no negative roots.)=0. /( .220).for x = . then p is a factor of a. .|). For. Thus 10-6 is a lower limit to the positive roots. -2.+ hqp n ~ l + kp n = 0. we find that | is a roo . Similarly k is divisible by q. Hence the rational roots of ax n + bx n ~ l + . and each group of terms is positive if . denoting the polynomial by f(x). and q is a factor of k. .C.A. .520) 4. Hence a is divisible by p. p \ or 2.^v-p ie.lo Hence <f>(y)>0 for y"^ ~~ > i-e. 6.5x 2 + 5# .. Find an upper and a lower limit to the roots of f(x) =3x*~ 61s 8 -f 127#* + 220a. The greater root of 3 2 -61o.0^4. We have f(x) =z 2 (3a.hx -f k. and therefore aq n + bq n ~ l p + . 1 .. so that aq n is divisible by p. 4. 2.Qlx + 127) -I. Hence aq n =p x an integer. q 1 or 3.4-127=0 is 17-9 . therefore p is prime to q n . for x 10-6.. . if px-q is a factor of f(x). p is a factor of a and q a factor of k ..520).x) has no positive roots. where a. i. 92 RATIONAL ROOTS 13. then 0. where p is prime to q. therefore % is an upper limit of the root s.

.3. Further.. (A) where p^ p 2 .84).. it is an integer and h is a factor of 168 =2 3 . We generally take a 1. Newton's Method of Divisors.7... and it is the only rational root. and so on.h. are the least possible integers. If x -h is a factor off(x). ...78) + (x* -672) + x (x 3 . if a is a rational root of (A).9 and 7.t ..184). Let the given equation be transformed into f(x)=x n + p l x n ~ l + p 2 x n ~ 2 + .. Since f(x) = x 2 ... Let h be any factor of p n . . it follows that all the real roots lie between .. then a is an integer and p n is divisible by a. Herein lies the advantage of Newton's method. Ex.i. the quotient is -~q n -i . (See Art. 1. To lessen the number of trials. the last number in the bottom line must be 0. thus Pn+Pn-l +Pn-2 +. this is added to p n -..p n _ 2 .) By the last theorem.. giving ~q n _ 2 h. Dividing this by h. choose any number a.... we get ~g n _ 2 > which is added to .+ft +Pl + 1 -gn-l -gn-2 -92 " gl "1 NEWTON'S METHOD OF DIVISORS 93 Here p n is divided by A.. If any q in the process is not an integer. 14.168==0..h and the reckoning need not be continued. then if x-h is a factor of/(x).+?V_ 2 +p n -l +Pn h 9l Where q^pt + h.. +p n = Q. and 4/( -x) = 2x*(x* ... Find the rational roots off(x)^x* -39z 2 + 46z .. (x* . If h is a rational root. and we divide f(x) by x .. q Newton's method consists in performing these operations in the reverse order./(a) is divisible by a -h. the usual reckoning is as follows : I-/*) 1+ft +y 2 +. f(x) is not divisible by x . .30) + 2 (23* . (6). 4..

~ 8. Again. Testing these as below by Newton's method. -I -h must be a factor of 252. 4. 3 2 . 1 .7 are the only rational roots. -7. +6. 4.-252. +3. -3. and so we can exclude 4. 7. 5 . we find that 6 and . . -7. /( .-160. Also if h is a root.-2 2 . for /(I). 1-3) -168 +46 -56 ~^To* 1-6) -168 4-46 -39 4-0 4-1 ^28 4-3 _-6 -JL Tl8 -36 -1J ~~6 1-2) -168 + 46 -39 + + 1. so that 1 are not roots.The possible values of h are therefore 1. 3. -84 -19 -29 .6. 2.2. thus we can exclude .~2 5 .1). -8 (A) A number of these can be excluded at once. 6. The remaining numbers are 2.h is a factor of 160.

^38 ~-^$ "^29 Jl+3) -168 + 46 -39. + 56 -34 ~UJ2 ^73* .

* The reckoning stops at these stages because 10 and 73 are not divisible by 3. . sc? + qx + r = Q has only one real root. Use Descartes' rule of signs to show that : (i) If q is positive.1 + 7) -168 + 46 -39 + + 1 + 24 -10 + 7 -1 70 ^19 + 7 ~~~o EXERCISE XI REAL ROOTS 1. (ii) x* .3# 4 + 2a? 3 -1=0 has at least four imaginary roots.

.2* 2 + Ix + 3 -0. [For an upper limit in Ex.IQx .Sx 2 . x 5 .x) (b . + I).l +p 2 x n ~ 2 + .350 =0. +p n =0.e.350) ..5a? 3 + 6* 2 .3z 3 . x*7. if x .10** .e. [Group thus : x(x* ..1 is an upper limit to the roots.2 + . z54-z 4 -6* 3 -8o. For a lower limit. i. x n 1 [Let sol..3z 4 .h* = are real and are separated by a and by 6.l +piX n .. 3. 6.15) + (z 4 . x n x I This holds if x n > p . if the n umerically greatest negative coefficient is equal to ... 4. if x n >p . find an upper and a lower limit to the roots of the equations in Exx. then/(z)>0 if x n >p(x n ~ l + x n -* + . l.1 > p. . group thus : x*(x* .50=0. [Group thus: x(2x*~ llx-ll) + (x-l).e. 5 .24z 3 + 95z 2 .] 11.10). then/(z)>0if x n >p(x n ~ r + X H . x* + 4x 3 .101 = 0.x) .-10:-0. 10. write x = -y and group thus : 5. For the equation f(x) =x n +p 1 x n . + l). if ..T . 2x*~llx*-I()x-l=Q..p 9 then p + 1 is an upper limit to the roots.] 9. we have By grouping terms. 4. [Denoting the left-hand side by/(#).llx . For the equation f(x) ~x n +p l x n .llx* -9z. ---..' 2 --15a.5) + (6z 2 . then Zjp 4. x* . a.] 8.11* .46x . if the nu merically greatest negative coefficient is equal to -p and the first negative coefficient is p r .I + . i.6z 2 . Show that the roots of (a .94: REAL ROOTS 2. +p n = Q. i. 3-9.

. 17. 3* 3 . 14-19.. /J.. 2 -25ar+42=0. a88 =27a 2 8 +527<xSc.5 .. ... etc .6. if (x-l)x r ~ l >p... 8. i. (2) 351. For equation (A) find the values of Za 3 yS and S**ffy. Find the rational roots of the equations in Exx. 1. .-24:=0.. The second occurs five times.6* 4. ELEMENTARY SYMMETRIC FUNCTIONS 95 15.. . c. show that the limits are (1) 4...This holds if x n >^ . 3-7.. 2. * 3 -9z* + 22o.1 . Functions of the type ZoPfPy* ...r . 27aj9yS consists of terms of the types a 2 /?y8. (A) then by Art. 4# 4 i 18... a? 3 -5* 2 ..237=0. y. j5. 6s 4 -25^4-26^ + 4s-8=0.... This consists of terms of the types a 2 2 y.2 + . 19.. Therefore Hoc . 14..18*4-72=0... (B) It will be shown that these equations can be used to express any symmetric function of the roots in terms of the coefficients. 27a= -p l9 27aj9=]9 29 27a/?y = . (3) 9. are positive integers. Symmetric Functions of the Roots of an Equation. Ex. 6. 16. 13. These can be calculated in order by a process of multiplication. (5) 7.. . namely as th e product of any one of a. c and a term from 27a/?y8... y. will be called elementary functions. c . .. -f p w = 0. find an upper limit to the roots of x 1 ..4. If the rule in Ex... ZajSy.. 11.e. and any symmetric function of the roots can be expressed in terms of them.e...51 . . 8. ... 12. 15. . Let a..4=0. (4) 6.. .j> 3 . The first of these occurs once.100# 2 .I) r >p. .H53^-95a.2* 2 .. i. 1 1 is applied to the equations in Exx. By using Ex. where a.. (i) The product Sat. 6a. which holds if (*. and (ii) Consider the product 27aj8 . a/?yf .. be the roots of x n +p 1 x n ^ 1 +p 2 x n . in the product.

y. aj8) 2 - Here the elementary functions of the second. The term a 2 /?y8 occurs three times. 1 below. (C) Using the last result and equations (B). a/tySe... the sum of the numerical c oefficients is zero. . (ay)(a/?8). 96 DIFFERENCES OF ROOT3 16. ..... 3 are important.. For reference... where h is chosen so as to remove the second term...... Ex. we find that Z^Fr^PiPi-ptPa-Sps . // a. It is important to notice that the degree n of the equation does not occur in the reckoning. A useful check is provided by the theorem in Ex. .. third and fourth degrees in a.. c.. every root of which is 1. .. namely... C? = 3(7? + 12(72 + 10(7?. indicating briefly the process of reckoning.. These are unaltered if the roots are diminished by any number h. In this way (C) becomes CJ . Let v be any symmetric function of the differences of the roots of (o c . 71 -" 1 + p 2 ..27a 2 j8y + 3Za a ]8y8 + lOZ^ySe ... Hence they may be calculated by using the equation f(y + h) = 0.a s )8y8...a 1 .. These results can be tested by putting a=]8 = y ..... The first of these occurs once. a l9 a 2 .=1. j3. .. j. y. . as each of the products (a/?)(ay8). Symmetric Functions involving only the Differences of the Roots of f(x)=0...l) n =0 f then if v is expressed in terms of a . y . The term ajSyS c occurs ten times.. + y n == 0. we give the following results. are ZAe roots of x n + fto.. (a8)(ay).. ...rn "" 2 + ..fl a .. . are calculated in order. and obviously the process can be continued so as to find any function of this kind.. 2.. ^y . and .... 8.... 1. a. The sum of the numerical coefficients is obtained by putting a^ a^ a 2 1 * n v But in this case the given equation becomes (x + l) n ... The functions in Exx. Therefore ^aft . j8....... for we can select two out of the five...a n $a:. in 10 wa y?.. (D) NOTE......(27 .

8' 3 + S' 2 (a' + j8' -f y') + 8' (j8 y + y V + '') + a'j8'y' . o are the roots of 0.. a) 2 .....fi .. (ii) The equation whose roots are .. -i .... y'.( -/3) (ii)r(a-j8)VS f ..... involves only the differences of the roots.. (B) obtained by the substitution xy-b/a..jS) (x ....d) 2 .y ) = ax 3 + 3bx.. ....+ 3cx + d. its value may be found by using the equation .... 2.. j 3. thus ^ y 8 (iii) Denote the function by v. . jS.ac).. find the values of (i) We have a (x ... + a-/3-8)(a+/3-y-8).. 8' be the roots of (B). /3... y... '. // a.... // a... we find that therefore a 2 (a 2 + 1 ) (j8 2 + 1 ) (y 2 + 1 ) .. and -0/8 = (8'ta')(S'+jB')(8' + y') . Ex.a) (x . (ii)27( j 8-y)(y-a)-^y-Za 2 -3 Ex... 3... .. SYMMETRIC FUNCTIONS OF ROOTS 97 Let a'.. in this case. 8... 6 and d. 4# K . Substituting 4....(a ...1 and .... + y+i-O.. y are the roots of ax 3 + 3bx* + 3cx + d =0. Then since Za'=0.. y..since t. (A) find the values of <!)..3c) 2 + (36 .... v=0.8Za|8 -^ (6 2 . ..is obtained by interchangi ng o and e.. a-8=jB' + /-a'-8' = -2(8' -fa 7 ). Since v is a function of the differences of a..1 for x in this identity.

!=<).....'. 5 2 -2^p 2 . j3. Substituting for a in the first equation and simplifying...Ex.. If a. The artifice employed in the next example is often useful. a 3 -za-f4r=0. we have which on expansion becomes -f 4g 3 + 27r 2 =0. 7/a + jS + y + 8 = 0.. (a-jS) 2 .. (ii) Use the result to find the equation whose roots are the six values of |~(af /?). 8 be the roots of & +p 2 %* +P& +jp 4 = 0.... Hence by addition. prove tftaf a 5 + )3 5 + y 5 + 8 5 .SjScejB .. ^J&f'l... Let 2~(j8 -y) 2 . 4... (y-a) 2 . z = (jg + y) 2 . j8.. and let s r ~ZoL r > th en by Art. 2 +pi<xQ.. (i) Find the condition that the sum of two roots a.. #*......... then since 2?a = and a/?y = -r..... . . y...827a'j8'y' = 320/a*.. Hence the required equation can be found by eliminating a from a 3 +ga + r=0. a 4 -2>i<x a + # 2 2 ~^3 +^4 =0 .......... y.. /?.. 2.....0. 15. j8 are any two roots of a^ + 46a? + 6cz 2 + 4da.... Also a 5 +p 2 <x?+p3<x.... ft of xt+p^+p^+paX+p^Q ... + e=0 ... = . Let a. Equations whose Roots are Symmetric Functions of a. find the equation whose root s are (j8-y) 2 . (A) may be zero..4j9y =a 2 f 4r/ a . 17. The following are typical examples. with three si milar equations. where a. (B) (i) Since a and -a are roots of (A). By subtraction. (z -f q)oc = 3r. a 4 + P&* + p 2 a 2 + P 3 a + Pi ... 5 3 = -3p 3 ... y are JAe roote of x? + qx + r=Q..

d... G = a*d . a^a 3 ^ -3(a' 2 d-9a&c-f9& 3 ).. If then we diminish the roots of (B) by z. I) 4 ..3a6c + 2b\ prove that : 1...9^6 -h 9c 3 ). b. c. therefore 2>1 2 P4~2>1#2JP3+2?3 2=: 0> .. a 2 {a(j3~ 7. . is ^=^0.. D-(a.. the condition (C) gives an equation of the sixth degree in 2. .. a 2 {(j8-y) 2 -f(y-a) 2 4-(a-j8) 2 }-18(6 2 -a^).. 98 ROOTS OF THE CUBIC EQUATION (ii) Let z -\ (a + j8).. (i) The condition that a.. *Explain how to derive (2) from (1) and (5) from (4). writing x~y + z. This equation is ay* + 4By* + GCy* + 4Dy + E=Q.. c. 10.* a^V 3 = 3 (^ 2 a .2* + 2bz + c..e$z. j3. is oc 3 =6 3 d..*. then we have (a . (D) where B=az + b.. (ii) The condition that they are in O. I) 3 . b.. 6... y are the roots of ax* + 36# 2 4....z) =0.. the sum of two roots of the resulting equation in y will be zero.. a 4 -fj 2 a 2 +^4=0 and a^a 2 ^ p a )=0.. (C) which is the required condition.. 0.....6a6aZ + 3ac 3 -f 36 8 rf). For equation (D). a*(a^ 2 -f ^y 2 + ya*)(ay 3 + j8a 2 -h yj8 2 ) = a*(3a 2 ]8V 8 = 9 (aW .... 0. 3. 2. a 4 {(j8-y 8.z) + (j8 .... y are in A..... 5...P. a 3 (2a-~y)(2j3-y-a)(2y-a--/3)^ -27(9. E = (a.....<%..3cx + d=0 and H = ac-b 2 ..... 4.... a 3 (^4-y 9. Now a 9^0 unless p 4 =0.P.* a f {a a (j3-y) 8 + j8 8 (y-) 8 + y 2 (a-j8)} = 18(c 8 -c). .. whose roots are the values of J(a EXERCISE XII SYMMETRIC FUNCTIONS OF ROOTS If a..

r -f a 2 ). If a..(iii) The condition that they are in H.a 5 5ax(# -f a) (# 2 -f a. Solve(x-fa-f-6) 5 -x 6 -a 5 -6 5 -:0.+ -. p. /?.3ca. Prove that a 8 -f 2ga . (ii) a 1 + b 7 + c 7 ~labc(bc + ca + ab )*. (# -fa) 7 . y may be connected by the equation is aC 3 =5 3 D. [(i) If 2a = /M-y..] 11. 0. prove that the equation whose rootsare . [Let a. 4.3dcb + 2c 3 =0.y 2 2j8' a-f j8-2y is a(aa: 2 -h 26a.] SYMMETRIC FUNCTIONS OF ROOTS 99 13. 97. and use Ex. /. + c) 3 = (aa^ + 36x 2 4. Use the last example to show that (x -f a) 5 .x 5 . 6.4. [Use Ex.] 15. 17. . If a + & + c=0 show that (i) a* + b b + c 6 =5abc(bc + ca + ab) ..J . hence the required equation is got by eliminating j8 fr om ^ 3 +g0-fr=0 and z 3 8 + qzp + r = 0. If a. + a 2 ) 2 . 11. c be the roots of a: 3 -f qx + r =0. The condition that a. ]8 are any two roots of x* + qx + r=Q.rz = 0. the equation whose roots are th e six values of a/0 is [Let z = .x* . (ii) Eliminate x between u = and ax 3 + d=0. + d) (ox -f Show that this reduces to a cubic equation.P. y are the roots of &+qx + r =0. where B= 12.is y ft a y a [Let 2 = . The equation whose roots are jgy-a 2 yq-j8 2 )8 . (iii) Put x~l/y and use condition (i).+ -.] 16.a 1 = lax(x-{-a)(x z -f aa. [Observe that x -f a -f 6 -f ( -x) + ( -) + ( -6)~0. then a= -b/a. Eliminate a between this equat ion 14. a = /fe. ~ + . is rf 2 a .

25. j8.2 . If a. b are the roots of 3 a CHAPTER VII PARTIAL FRACTIONS 1 .bd).2a r _ 1 a r+ ! -f 2a r ^ 2 a r+2 .(a . .. If P is of lower degree than Q.l) n a n =0 are the roots of V n . .%<*>ia* + 2a a 4 . are the roots of x n +p l x n . a 2 a 2 =4(4& 2 -3ac). prove that : 19. 27. 20.+p n = 0 .J 8) 2 (y 2 + y8 + S 2 ) = 144(c 2 .3abc -f 26 3 . .18. If a.2 + . (ii)^ 7 . . or a. 4. where &o = oS ft i = i 2 .~lp*(p**-pt). 2 X X' .32(7.. Ga*d.1 4.6cx 2 + 4dx + e=Q and //~ac~6 2 . a*Z(* . 0. a*i. 22. . a 3 2.a^. By means of the division transformation. If P is not of lower degree than Q y P/Q is called an improper fraction.. 23. is called a rational fraction.y)(a . prove that (1)^ = 2(^-2^).]8)(a . prove that Ev* j3 2 = .M*"" 1 + M n ~ 2 . .1 +p 2 x n .. prove that either a =6=0 or a =6 =3.ptf 26.8) = . An expression of the form P/Q. -i. y. . .a^. y. Thus ** + ! ..( .( . 8 are the roots of x*+p&*+pjc+pt=0. Show that the squares of the roots of a<F n . y. 24.a z p = I2(ad~2bc).. where P and Q are polynomials in x. If a. P/Q is called a proper fraction. an improper fraction can be expressed as the sum of an integral function and a proper fraction.1 ) n 6 w = 0. . b r =a r * .* a 2 *Note that (20) follows from (19) and (22) from (21)... .2a 2 ^2 == a a . Rational Fractions. 8 are the roots of ax* + 4bx*-\.2a r ^30 r+8 + .* a 2 21. If the equation whose roots are the squares of the roots of the cubic is identical with this cubic.

. To resolve a given fraction into partial fractions is to express it as the sum of two or more simpler fractions.= j. the polynomial A . and then C _ X Y _ _.. and therefore YX'-XY' is of lower degree than YY'. therefore C + tf-O and . Q'. Y 9 such that CX' X . X'/Y f are proper fractions. If A+^ A' + -^-. which is impossible.Theorem. Thus (BX + A Y)/A B is a proper fraction : so also is C/AB. therefore BX + AY is of lower degree than AB. X' X YX'-XY' r or A A ^7 -y. Thus Az=A' y and consequently X/Y^X'/Y'. 2.A' =0. Y r can be found so that BX'+AT = 1 9 (Ch. where the letters denote polynomials in x ix # and X/Y. Hence A .' I A and CY'/B are not proper fractions. // C/AB is a proper fraction and the factors A. Now X and 7 are respectively of lower degrees than A and B. 18. we can find polynomials Q. Partial Fractions. by division. B are prime to each other.) j^ * C CX> CT and therefore . for otherwise. CT n Y and ~ = Q +> where X/A and Y/B are proper fractions. Ill. then A^A' and X/YsX'/Y'. Fundamental Theorem. yy * Also X and X' are respectively of lower degrees than Y and Y'.A' would be identically equal to a proper fraction. FORMS OF PARTIAL FRACTIONS 101 If CX.+ -~. proper fractions X/A and Y/B can be found such that CX Y For since A is prime to B } polynomials X'. X.

. 102 TYPICAL EXAMPLES . . are independent of x.. . . It is assumed that x . B is prime to x ..b is not a factor of JB. so that JS is prime to (x . fraction of the form (iv) To a factor (x 2 +px + q) n of Q there corresponds a group of the form +- '" (x 2 +px + q) n ' Here A. We shall prove that (i) To a non-repeated factor x-a of Q there corresponds a fraction of the A form x-a (ii) To a factor (x .6) n . 4 . (ii) Let Q = (x . where the fractions on the right are proper fractions.b) n of Q there corresponds a group of the form -i -o 1 J -.a. and consequently where the fractions on the right are proper fractions. B.b) n . Proof (i).?_ J -. Let $ = (#-a) . S.3. C v (7 2 .a) +7/B. then since x-a is a non-repeated factor. Hence X is a constant.. and so P/Q = */(*... To resolve a Proper Fraction P/Q into its simplest set of Partial Fractions. . B I} B 2 . / T\O'/ 1\*> '***'.2 -.L -L. x-b (x-b) 2 (x-b)* (x-b) n * (lii) To a non-repeated quadratic factor x 2 +px + q of Q there corresponds a .

-a.as the sum of an integral function of x and three proper fractions. Express j^. X* Ex. is of the form Cx + D.a)(x ~b )(x~ c) is x + a -f b + c. and then put x 2 .. multiply each side of (A) by a. a 4 (x-a)(x-b)(x-c) (a-b)(a-c)' x-a Ex. but X is of the form Cx + D. 4. 2..2. . . Various ways of finding the constants are explained in the following examples. Ill. 1.abc. (Ch.1 at most.a) (x . 2a + xZab . (iii) The proof is similar to that of (i). = ' (a-b)(a-cY The values of B and C can be found in the same way . (iv) The proof is similar to that of (ii). B 2 ..a: 2 .. Resolve into the simplest possible partial t fractions 6 c Multiply each side by x . We may therefore assume that re 4 4 5 C (A) (x~a)(x-b)(x-c) x-a x-b x-c* To find A.+ B n .c) = x* . ' ~ * We have (x . 2 are constants.6) (x . and it can be put in the form Bi(x-b) n ~ l + B 2 (x-b) n ~*+.) where 5 1 . but each of the set J9 1 .Hence X is of degree n . which proves the statement in question. Hence it is easy to see that the quotient in the division of x 4 by (x . and put x=a.

.. a-6 + c + d=0. 6 = 1. W Multiplying by (x z + x -fl) (x* ..a? 8 -5 m .-f . equating coefficients. Note that the first and last equations in (B) can be found by (i) multiplying (A ) . and let #->oo . 6 + d=-l . J = -2.r^rr .= -.=-^ + T . 2 -5 21 43 Therefore . (B) Whence a=2.-2) x-l (*~1) 2 (a.rr* + .. put a: =0 in (A).... Find d as before.'. multiply each side of (A) by (x ..~2. we find that 3x*~2x 2 -l=(a and. /.2x 2 . 6 = 1. . C r~ ^~ 4. .... . Finally.x + 1). In this way two equations connecting the coefficients can always be written down.-!) 8 x -2 Second method.. and then put a? = l . d. 2i JL METHODS OF QUADRATIC DENOMINATORS 103 Particular attention is called to the last two steps..... --= ~a-f 6 ~c~.^ (#-l) s (a.1) 8 . a.. . T 33? .. /... l'-5 . . c = l. a= -2. Next. Next. The steps are equivalent to the following : Multiply each side of (A) by (x~-l) 3 (x-2) and equate the coefficients of sc 3 (the highest power of x) and the absolute terms on each side.1 ax + b cx + d . multiply each side of (A) by a?.

It is therefore an identity.2x 2 .a? ~ 1 and # 3 = 1.x 4. The following method is sometimes useful. The integral functions and also the fractions are identically equal." + ^ + ^l> + (^^ + ^ *" ( } We first find e and /. c = l. 1 1 " a ~(-2) 3 .x + 1=0). and so a + c = 3 9 b + 2c + d = l. From (A).l~ 8* Multiply (A) by (x -I) 3 . multiply (A) by x + 1. rf= -2. 104 PARTIAL FRACTIONS Next. This linear relation holds for two values of x (namely. and (ii) putting x=0. (m) . the roots of x 2 4. 2c + 2d= -2. whence we find that and It follows that l=3( /.by x and letting #->oo . . Multiply both sides of (A) by x 2 -f x + 1.1. and so /-2e=0 and l=-3(e+/). and then let x = . 6 = 1. Two other equations may be quickl y obtained by putting x~l and x.1 . giving a = 2.1 whence by division. giving e=-l/9 and /= -2/9.. and then let This gives D = * B+/ ' when We have & 2 = . 1 __ a 6 c c? ea. 3(/-2c)*-3(e+/) = l. -2c= -2. and then let # = 1 .+/ .1 _ ca^-f (c + d) x 2 -x -fl ~~ aX x 2 . 3s 3 .

. 2 -f-2)" * 15. . .15 ' ' 10 11 12 ' )(a? I -f2). -l=a~& + e-d+/. as the sum of a constant and three proper fractions <a-)(*-&)(*-c) ^ ^ x 3 . 3 . 3.' (a: 2 -fl) 2 (a. and let -> oo.Multiply (A) by x. r (l-ax)(l~bx) r r Hence show that if 1 -o&x 2 is divided by 1 -(a + b)x + dbx\ so as to obtain a quotient of (n 4. = a -f 6 -f e. J 2 .: r v as the sum of a constant and two proper fractions.. Express .!)(**-!) 1 1 17 1 1 1 1 1 1 i _ !** _ 1 ^ /__ 1 \9 ' / 8 s + 1'72 *-l 4 (*-l) 2r 6 (x~l) 3 9 a. In (A). /.~ . Express . giving c -i . EXERCISE PARTIAL FRACTIONS Resolve into partial fractions I. a: 3 H-a: 2 -f 1 " ' ' ' ' g --.I9x . 3( *~ 6) . 1 ' (x-l)(x>. let xQ. 16. giving 6 = ^ + ^ = ^-|.'. this quotient is and find the corresponding remainder.a.=~ .1) terms.

Hence show that _ "" IDENTITIES FROM PARTIAL FRACTIONS 105 3. . 2 34 V ... Express . 714-2 22. ..r-.. Hence show that _ _ ___ v.. as the sum of an integral function of x and r (x-a)(x-b)(x-c) 6 three proper fractions. Use Ex. " = ^ + 21. 21 and similar identities to prove that (\\ _^___fL_4__^__ -4_f_l\n .4 19. 20. show that _c c t x x+ 1 [Assume that . ..as the sum of a constant and three proper fractions (-a)(*-6)(a-c) p ^ 18.17. If ( 1 + x) n c + c v x + c 2 x 2 + . Express ~ rr-. Hence show that (a-b)(a-c)(a-d) (b-c)(b-d)(b-a) (c-d)(c-a)(c~b) _ _ (d-a)(d-b)(d-c) is equal toa-ffc-fc-fd.+ c n x n ..

_ _. 23 to prove that [In Ex.v.2. Use Ex. where p is a number which does not depend on n.5 1. and these terms cannot be added up. and the number of terms on the right (namely two) is independent of n.. Prove that 24. ( .4 " V . Let u n be a function of the positive integral variable. To sum a series to n terms or to find the sum of the first n terms of a series is to express s n in the form just described . . 23. a. The sum of the first n terms of the series . a harmonic series cannot be summed. Among the series capable of summation are arithmetic and geometric series . and a?=0. unless the value of n is specified.2 2.4 v .=2.4 4.3 2. (n + l)( + 2)(n + 3) 2(n-f3) vlvy 2.. write n -f 1 for n. 3. .4 370 T '"' 23.3. Sometimes it is possible to express s n as the sum of p terms. Meaning of Summation.3.3 3. 1. but this is not always possible. and put ! = !.] CHAPTER VIII SUMMATION OF SERIES 1 . For example. and let The function s n possesses this peculiarity it is the sum of n terms. a n =w.

to n terms. by hypothesis. J ) (n 4.v n _ 1 . beginning with the n-th.b)_ __ _ u n (a -f n + r . 2 . b) (a -f n . Series in which u n is the Product of r successive terms of an A. 4 -f 3 .1 " . 5 -f.v n _^ where v n .. Here u n ~n(n + l)(n-t-2) ' u n= v n~ v n. // we are able to express u n in the form v n .(a + w + r.where v w =-i( and since v Q. 3 + 2 . Method of Differences. />). P. 3 . b y r are constants. or simply by s n . 2. 4 .1 .. b) u n n Therefore w n = v n . Sum the series I .% . which is as follows : It is required to sum to n terms the series in which w n = (a + w&)(a + n + I .-/ and consequently 5 W = v n . where v n is some function of n. whence by addition METHOD OF DIFFERENCES 107 Ex.. we have by the preceding. where a. 3. 1. The last example is an instance of this type.u i + is often denoted by H r r ^iU r and sometimes by 2u n .. then we can sum the series to n terms. We have __u n {(a + n + r . 6)... For.2) (n + 3). and the same method can be applied to the general case.

divide by the number of factors so increased and by the common difference of the A. as in the next example . The constant is found by putting w = l. A. 5 .Here v is independent of n. p. . . ll-K.. noting that s l = I . it has been shown that a rational . H B. 3 . 2. To find the constant C. 4. .C. 7 -f 5 . e. 108 SUM OF INTEGRAL FUNCTIONS 4. and applying the rule. 7 . the ' constant ' clearly being zero. put ?i = 1 in (A). Ex. or by substituting for n in v n .34-3.integral function of n of degree r can be expressed in the form o + bn + cn(n + l)-f dn(n + l)(n + 2)-f . 3 . as in the next example. . Ex.. Sum the series 2. Sum to n term* 1 . are constants whose values may be found by synthetic division.. and add a constant. Series in which u n is a Rational Integral Function of n. In Ch.P. -f 3)(2 + 5) + 15}. 5 + 3 .6 + 4. Ill. We can therefore use the rule of the last article to sum series of this type. n + 2 in succession. * The expression in the bracket (} is formed by subtracting the term which prece des n from that which follows (n-l-2) in the A. . 9 f .l)(2w-f l)(2w + 3). 6.-f(tt + l)(n 2 -f2). I. 1. . P. .2 tt n = n( therefore 3n 8 + Wn* + 2ln + 38).. Here tt = (n + l)(n' + 2) 1 + 1 )* + 1 + 2 +2 +0 Dividing by n. Here u n ~ (2n. 3.. . the reckoning shows that 1 . hence the sum to n terms may be found by the following rule : On the right of u n introduce as a factor the next term of the A. as on the -2 right. n + 1. 1.. (3). 5 . where a.. to (r + 1) terms.

.. where 1 1 . by the rule . Here 111 n 4 (2/i-l)(2w + l) 4 1 where t> n -- The general case is as follows : It is required to sum to n terms the series in which u n is the reciprocal of (a-f n6)(a + w + 1 . . and s n = v Q -v n . C== _L. b)(a + n + 2 .5. (a + n + r. . hence the sum to n terms may be found by the following rule : Remove the left-hand factor from the denominator of u n .P. P. r are constants... A simple instance is the following : Ex. in the last example. we have _ 1 (a + n + r-1 . Thus. 6)..(a-hn-t-r~l . 6).6) Now v is independent of n. * The numerator is the difference between the first and last factors of the deno minator. Series in which u w is the Reciprocal of the Product of r successive terms of an A. and subtract the result from a constant. divide by the number of factors so diminished and by the common difference of the A.1 . beginning with the n-th. where a.. SUM OF RECIPROCALS 109 Proceeding as in the last example. or by substituting for n in v n . 6.. The constant is found either by putting n = 1.l. 6)-(a-f nb) Therefore u n ^v n _ l -v n . b) .

for to do this we should have to sum a harmonical progression.Tc J* . 2. x 11 1 1 1 ^ ""18 which is the same result as before.. by Partial Fractions. O O. put w = l in (A).3. 1 10 TWO IMPORTANT TYPES The last example is an instance of a scries whose n-th term is where P is a polynomial in n of degree m. the nth term consists of an integral and a fractional part which can be dealt with separately. If u n is the n-th term of .4* " 18* Alternative method. If w>r~l.4 3 3. If m<r .U W ( 7i "f 1 n n(n + 3) n(?i + l)(n + 2)(n 1 * * To find the constant C. c^ 11 " 1. . observing that s^ _L^ x 2 .n + 1/>) 4. + l.w6) (a 4.+ ^-^ + . . this can he summed by expressing P in the form a H. 6.1. 8 = * . If w = r . in a different form. .% (a + n&) + a 2 (a 4. .2. Another important Type. We have l n 4 5 u . the series cannot be summed.1.Ex. Sum the series 4. .

a-b+1 a~fe+l a 1 Therefore s n = v n 7 r = = r {w n (a + n) .b + I ) (n > 1 ).1 ) x n . 1.5x* + . . + n*x n ~ l . we have s ( I .. .u n x n+l . . let then s (1 . .1 ..K ( + *) .1 . Denoting the sum by s.. Sum the series I 2 + 2 2 z + 3 2 z a + . If x = l. It follows that s = P/(l ~x) mfl where P is a polynomial in x.x (m multiplications in all). we can reduce the problem of finding 8 to that of summing a geometric series.. + u n x n where u n is a Polynomial in n of Degree m. .1) 6 + b(b + 1 " then s n = . The Series U 4-u 1 x-f-u 2 x 2 4-.. AT (a-fl)-fc a Also w.T ~ = f r _. + 2a.. (a-f n . . a-6-hl a-6 + 1 ^ * ' 7. where ^i = %-^o> v 2 ^u z -u lt . this series can be summed as in Art. .a a(a + \) a (a + 1) (a + 2) a(a+ 1) . t? n = w n -w n .x) = 1 + 2x + 2x* -f ..} For where r n = u n (a + n)/ (a . thus by repeatedly multiplying by 1 . = w t .(2n ..x) = w + v& -f t'^ 2 + . + v w a: n . . If x ^1.l)x n . Ex. r Now v n is a polynomial in n of degree m .1 Let 5' = SUM OF POWERS 111 then a' (I .x) = 1 + So: H. + (2 H . v r / . w . 4.a}.

(n-r + 1) ... ... .. by Art.. it will be found that 7.-f6 r4 . and we use other methods.. when expanded.2= ' ftr .(n . such that n r ^a l n + a 2 n(n-l) + .. n r+l /(r -f.. For we can find a l9 a z . + (n-l) r .(n . that is. S r can be expressed as a polynomial in n of which the highest term is n r + l /(r + 1 ). 112 . therefore w r =/(n) -/(n . 1. 6 r ... a' (1) Theorem.. independent of n. // r is a positive integer. Show that S^ln(n + l)(2n + l) 9 S 3 ^4n 2 (n + l) 2 ..__ 8..8= and so on. may be found as follows.. (n-r + 1). . J A 1 h r h n h r(r-l)(r-2) 6r+1 ""' r = ' ftr .. 3. a r . We shall write and cr = n(n-f-l).l) r } + . the values of & r41 ..1 = ' 6r .1 =/(n). Writing n.. + 6 r Expanding and equating coefficients.. +a r n(n-l)(n~2) ..... so that n r = 6 r41 {w r+1 .. a r /(r + 1).-f n r . .l) r+1 } + b r {n r . If r is large. . n-2. 1 n r " f .... and /(n-l)-l r -f 2 r -f ..1). . b lf 6 2 .. is a polynomial in n of which the highest term is n r+1 .1 ) . (A) where obviously a r = l... We have 7i 2 -n-f W(TI~ 1) . Also (2) If /S r = 6 1 n + 6 2 n 2 -ffe 3 n 3 + . (B) The right-hand side.. cannot be found conveniently in this wa y. fe r _j.+n r ....1.. The Series 1 r + 2 r + 3 r -f .... /(tt) = l r + 2 r + . we have + -a r (w + l)w(tt-l) ... . 1 in succession for n and adding. #*.

.. (4) The values of S 2 ...1..2... according as r is even or odd. n . . /S 4 .n(n + 1).ODD AND EVEN POWERS (3) The values of 8^ S 3 . where a=n(n + l) and cr' On expansion. Putting n .. Let w n = n r (n + l) r (2n + 1) . . .. S 6 .l) r r (2n . S 5> ... we have *2r-4 .. .1.1) .2. . _ 5 + .. . and the remaining factor is a rational integral function of n(n + l). When r = l. . .. 2. 1 in succession for n and adding.. then n 2 (n + 1) 2 is a factor of/(n).. Of course. .. then W 1 + w 2 + . 1 in succession for n and adding.(n .... +w n =ri r (n + l) r (2n + l)=a r a..... We have v 1 + v 2 + . .. may be found in succession by considering the function v n = n r (n + l) r -(n-l) r n r . expanding by the Binomial theorem. Putting n .... (A) where a.+v n =n r (n + l) r . and.. this formula gives $3 =4 -87 = Thus if r is odd and >1 and S r =f(n). it does not follow that S r is arithmetically divisible by n z (n + 1) 2 ... The last term is rS r+l or S r . 3.. may be found as follows..

then n(n + l)(2n + 1) is a factor of /(ft). .. B 3 . . this formula gives r = l Jour' =3S 2 r = 2 r-3 r=5 | ' (3o* . BERNOULLI'S NUMBERS 113 Thus if r is even and $ r =/(n). . (5) Bernoulli's Numbers... for S 1 =ln(n + l) and $ 3 . 5 7 ...lOa 8 + ITo 2 . The numbers B 2> -B 4 . . (C) the last term on the left being B r or (r + 2) B r+1 .# 8 . B Q9 . according as r is even or odd.. It will presently appear that the formula (B) is immediately deducible from (A) of the last section. the zero on the right of (C) is to be replaced by |. in terms of which a large number of functions can be expanded.. B^ B ly . When r 1. are all zero.. are the famous numbers of Bernoulli. Thus ^ = 2.. according as r is even or odd. one of the best is the following. as Bernoulli's numbers.. A great many formulae can be given for calculating these numbers .. 3. and the remaining factor is a rational integral function of n(n + 1). .. We define B r as the coefficient of n in S r . and 7? 3 .. If r = 1.=0. $ 5 .. then (2r + l)B ar + iC5(2r-l)B^ 2 + iC5(2r-3)fi 2r ^ + . . . when this is expressed as a polynomial in n.(B) the last term being S r or (r + 2)S r+1 .15a + 5).. For convenience we shall refer to the whole set B v B 2 . 2. //r>l. all contain the factor n 2 (n-f I) 2 . .

(4). Hence by the formula (B) of Art. . we find that "" 691 7 p 3617 M "6* 16 ~ 510' 1 43867 _ 174611 ' 20 ~ 798 330~ * A few more of the extraordinary properties possessed by the sequence B v B 2 . in (C). 8. 3. Putting r=l..B 2 = 1 30" Continuing thus. n-2.. 1 in succession for n and adding. B 39 ... 2. the left-hand side of (C) is the coefficient of n in the expansion of %n r (n + l) r (2n + l). ... 5B 4 .For B 2m is the coefficient of n in 5 2w . . By the Binomial theorem. . Putting n-1. are given below.

. each index is to be changed into a suffix 9 i. B^ etc.. B m is to be written for B m ... They may be used for calculating 5 2 . (F). appears. (E)... we have (D) 114 SYMBOLIC NOTATION Similarly. Vol.. 11' in Chrystal's Algebra. 8. Formulae (F). in fact. Equations (D). ).. + (-ir~ 1 Cj {1 5 1 + (--ir==0 .. then (i) S r . . the Bernoullian number as defined in Art.Equating the coefficients of n on each side....rS r _i + B r <d (") S r = r [s r ^dn + nB r where B r is independent of n and is. p. . (G) are equivalent to equations 10'.. (E) From (D) and (E) by addition and subtraction. but their use requires twice the amount of labour required by (C).... (G) where each series continues till B 19 or B Q 1.+n r where r is a positive integer... from the equation it follows that n r + l =ci and consequently 1 + . after expansion. (5)...... 231. (G) can be written symbolically thus : (D) (E) (F) (G) where. (6) Theorem.e. II. // 8 f = l r + 2 r + . Symbolic notation..

l + . by successive integrations. we obtain the formula (B).H-2^ 1 + r where B r is independent of n. i(r + l)aV = C[ fl (2r4a)S 2r + C^ No constant is added. the constant B r can be found by putting n = 1 . we arrive at the equation .. Denoting this polynomial by /(?*). 8. N. 1 in succession for n. . B r is the coefficient of n in/(n)... and adding. (At any stage. (5).. and we may differentiate both sides with regard to n.. 8. for every term is divisible by n. (4). 1. Writing n-1.For S r can be expressed as a polynomial in n of degree r -f 1. Therefore rn r . we have Differentiating with regard to n. we have * r -/(*)-/(*-!) This relation involves only positive integral powers of n up to n r y and holds for all positive integral values of n.. No constant of integration is added. ft -2. BERNOULLI'S THEOREM 115 Moreover.. The second statement follows from the first. xS A = 2. /S' 2 . (7) I'sin^ the relation ^V-M^'r-i^ + //#.1 + (n-l) r . Dividing by r + 1. a == Continuing the process. from (A) of Art. Hence it is an identity.l =f(n)~f'(n-l). (3).. we can find the values of/S'j. Putting r + 1 for r in (A)..j . /'(n)f{n r . Ex.) Thus S = n. and is therefore the Bernoullian number defined in Art. 8. for S r is divisible by n. Deduce formula (B) of Art.

. starting with n 4 n 3 n z by successive integration. in the form This result is known as Bernoulli's Theorem. as above. 1. n* n 1 n 3 n = 5 ~ + ~2 + 3^~30 _w 8 n 7 In* In 4 w 2 s "8 + "2 + T2 ""24"*" 12 ^ 8 " n 8 2n 7 7?i 5 2w 3 11 _ 9 2 " 3 "15 "9 30 _n 10 w 9 3 8 __7 e w* n 2 9 "TO ^ 2 + 4" "T(J + 2 "20 Tl . we have equation (D) of Art. 8. either by substitution in (H). Verify the following. (5).2 r-1 3 |3 r-2 ' '" 7i k ! which may be written (r + I)S r = n r + l + B 1 C[' n n r + B 2 C r 2 +l n r .+B r ^ (H) and expressed symbolically. or..l + . Putting n = l. Ex.

Show that the sum of the products...3.5 + 2.2< + 3 4 ... lP + J2 2 + 13 2 + . 1. Sum to n terms the series in Exx.l) n n(n -l)(n*-n~ 1). 4. P + 3 2 + 5 2 + ..+2P.1( . 7. Show that whether n is odd or even. 3 + 3* . I 2 . 1.8 + 5..6 + 3..11 + 8. (3n -2). Find (i) the sum of the squares..6 + 7.7 + . Find the value of 10. 1-9..l)(3?i 19. (i) P-2 2 + 3 2 -4 2 + ..2 a + 2.5 + 3. 3. to (n-1) terms.. 4.2.5. of the first n natural numbers is -^n(n 2 .. 2.1).8.11. 7.1+5. .5 + 3... ~20 2 + 2 12. (ii) 1* ..* 2. 2 + 2 2 . 18. 8..5. 20. 20 3 -19 3 + 18 3 -.1) terms = |( . 11... 2. ! 16.4 + 2.) Sum to n terms the series in Exx.6 + . taken two together. .6 + . 1. 4. each side of the base containing 2n shot.. 22. Find the number of shot which can be arranged in a pyramidal pile on a triangular base.3 + 8..2.3 a + 3.. 13. to (n ..l) n n(n.4 + 2.4 4 + ..... 9. Find the number of shot in a pile of 8 courses on a triangular base.5. ... 5.4 a + .8. 14-16.+2P. 1.3 + 4. 1. 14. l 17.2 66 ' 116 SUMS OF INTEGRAL FUNCTIONS EXERCISE XIV SUMMATION OF SERIES A (Arts. 1P-12 2 + 13 2 -. lP + 12 3 + 13 3 + .4.2.4. Show that the number of shot in a pile of n courses on a square base...3. each side of the base containing 12 shot..+2 3 ~P.5 + .. 1. (ii) the sum of the products taken two together of the numbers 1. 3. each side of the base containing 10 shot..9 + . 21.3.4. 6..4+ .. is .14 + . .

A pile of shot of n courses stands on a rectangular base. 12. 3.+ n*is divisible by P-f 2 2 -h3 2 4-. 6. 6. 'Find the number of shot in a pile of a rectangular base. where m is a natural numbe r or zero. 8. Show that if n is of the form 3m -f 1.4.23. .4 3.n 3 . . . SUMS OF OTHER TYPES 11? 25. B (Arts.... 1-24. and 3.) Sum to n terms the series in Exx. the sides of the base containing 16 and 10 shot respectively... Show that if n is of the form 5m -f 1 or 5m -f 3. 2.111 1 __ i __ j ____ i 1.. Show that the number of shot in the pile is * Here the nth term is the product of the nth terms of the arithmetic series.. then l* + 2 4 + 3 4 + . 7.. t Here w f =r(w r l)=nr (r l)f. 5. 5.3 3.. 8. 5. and the top course consisting of a single row of shot. then . 5.. 24. and the top course consists of a single row of x shot. 26. .+7i 2 . . .+w 5 is divisible by F + 2 3 + 3 3 + .2 2.

1 1 1.3 3.4..4.6 6. 3 2 5 2 7 2 14. .. 1 +. 7*"" . l.. .|2-f 3.4 3.2.| 1-f 2.3. |3 + .2. 1 1 4 f I L . .+ 2i + ~ 1 (!+*)(! +2*) ~ h 16.5 1 1 1 5 __ __ __ 1.. .6 3.3.6 g + + 07II + -+ ~-f ~ + .3 2. 1* 2 3 1 1 1 * * on o 1 1 1 O i I i 1.4 2.

] (2rn-l)-l (n-fl) 2 -n 2 3. 15 "" L " [Observe that nx n ~ l = a.8.] 21. 11. 137. 9. [The value of s n (l -x) can be found by Ex.n. 11. a 1.8 6.] 118 BERNOULLI'S NUMBERS 22 n + n ~ l + n . + 2.2.4 2. _LL*L*. 14 9 F75 77 5\"9 ' * * * . 475 + M ' 273 + 2 '37I + ' -. 4^6. 4. 11 8. 6^2. 11 .1 + JLL .5.1 5 11 w a + re-l 24 -[3 + il + [5 + .* 1 " 1 (a: + n) . 115. 13.8 " 13.a: n .2 + 23 l +2 2 + 2* "' 17273 + 27374 + 3~.+ .

8. show that 64-61 4. 8. [Use Art.] 2. 26 2 .6 8 -6 4 + -.1).) 1. deduce the value of S s by differentiation. = r + 1 . 4.+ . (5).l)~n r ..56 5 + . tor terms = 2 Sum to n terms 26.. 8.+n r =... -..+ . -1 for n in the identity f(n)-f(n.. [Expand (2n + l) r+l -(2n . Given that 9 = -^a 2 (2a 3 . Prove that n(n + l) x l+n + v ft -f r^ ' + .\Y (2B . ..^0.. (5). 3..=0.5a 2 + 6or .25. . Show that formula (C) of Art. 6.../(n).46 5 -. to find the values of B 12 and J5 14 . [Put -Ti + 1..+ .. If S r ^ b^n + b 2 n 2 + b a n 3 +. . -n + 2....3) where a n (n + 1 ). Find the value of B^ by using formulae (F) and (G) of Art.36 3 + 46 4 . If F-f2 r + 3 r -f . 8. where after expansion B m is to be written for B.. 1 3* 2 1 /r* 1 a a (1 C (Art.+ b r+l n r + 1 . (2).] 3. 8. =-0.] .. show that /(-n) = (-ir+y(n-l). Continue the reckoning in Art. (5).C r b +*2 r -'B T _ 4 + . Show that C\+WB r + C$+W-*B r _ z 4. may be written symbolically in the form ' . 5.36 4 -h5. 7.!)'+>.26 3 -4.

2. is called the transposition (2 3). observe that : (i) Disregarding the signs. and is denoted by or by This determinant in its expanded form is a 1 6 2 C 3 C&ifyjCg 4. If these equations are satisfied by the same values of x and y. giving six terms. it is easy to show that The expression on the left is called a determinant of the third order. has the sign -f .^ 2 6 3 C 1 In considering the form of this. 4. The term o 1 6 2 ^3> i n which the suffixes occur in their natural order. (1). The expression a^ . If the equations a^x -f 6j = 0. c in their natural order and arranging the suffixes in all possi ble ways. 3. Many complicated expressions can be easily handled if they are expressed as ' determinants/ The theory which we are about to explain arises in connection with linear equations. IV. the interchange of two suffixes. 6.CHAPTEK IX DETERMINANTS 1 .aj) v is called a determinant of the second order. a 2 x -f b 2 y + c 2 = 0. and is denoted by or by (a^). The number of arrangements is [3. Notation. a^c -f b 2 are satisfied by the same value of x.expression. = 0. for example.a^ = 0. the terms can be obtained by writing the ' letters a. (ii) The sign of any particular term depends on the order of the suffixes. Next consider the system a i x + ^\y + <a = 0. 120 DEFINITION OF DETERMINANT . then a^ . As in Ch.

Definition of a Determinant. 6. which contains the elements a l9 6 2 . is called the leading or principal diagonal. l n .. Rule of Signs. c 3 . by an even or an odd substitution). The determinant of the nth order.l n . and a i&2 c 3 l n i fi called the leading term. The sign of any other term is -f or .. according as the arrangement of the suffixes in that term is derived from the arrangement 1 2 3 by an even or by an odd number of transpositions. 1.. Since the interchange of the two suffixes transforms a positive into a negative term. .. 2. The expanded form of the determinant has n terms : half of these have the sign + and half the sign . 1 2 3 is changed to 1 3 2 by the single transposition (2 3).. and therefore p = q... n by an even or by an odd number of transpositions (that is. The diagonal through the left-hand top corner. Ch. . n. By means of n letters a. a 2 .. c. ... we can represent n 2 numbers thus . and n suffixes 1. and one element and one only from each column. p^q* Similarly q*^p. which can be formed by writing the letters a. 4. where the suffixes occur in their natural order.. 4. c. a n b n c n . We can find the sign of any particular term by the rule at the end of Art. IV. considering the second and third terms. . Every term contains one element and one only from each row. is + . These ideas will now be generalised. Also 1 2 3 is changed to 2 3 1 by the two transpositions (1 2). 2. I in their natural order and arranging the suffixes in all possible orders. Each of the n 2 numbers a l9 6 X .. ^ s called an element of the determinant. Thus.. This rule is justified by Theorem 2 of Ch... 6 2 . each with the sign determined by the rule below.. 3. The sign of the term a 1 6 2 c 3 .. FUNDAMENTAL THEOREMS 121 Ex.The sign of any other term is + or .. (1 3) performed in this order.. I. l n . Jn the expanded form of (a 1 6 2 c 3^4 e 6)> find Me signs of (i) ajb^c^e . according as the arrangement of suffixes in that term is derived from the arrangement 1 2 3 . denoted by . or by (^i^2 c 3 ln)> i n b n c n .. For suppose that there are p terms of the first sort and q of the second. 6. l n is defined as the sum of all the products. IV.

3. an interchange of two suffixes producing a change of sign.^ (ii) ajb^c^d^. as follows. the value of A is unaltered by the interchange of these rows or columns. The remaining terms of A f are derived from a^b^ by keeping the suffixes in their natural order and arranging the letters in all possible ways. 12543. as some text-books advise. we may make single transpositions. an interchange of two letters producing a change of sign. these are 5 and 10 respecti vely the first odd and the second even. and the interchange of two columns is equivalent to the interchange of two letters. 4.) Or. Theorem. changes the sign of a determinant without altering its numerical value. Theorem. 5. But this interchange transforms A into . The same argument applies to a determinant of any order. But this takes much long er. A determinant in which two rows or two columns are identical is equal to zero. 24513. or of two columns. A determinant is unaltered by changing its rows into columns and its columns into rows. 4) in the arrangements of the suffixes in (i) and (ii). and J = 0.J. (i) f o . therefore J=~J. () (\ II I ^}=( l 5 >( 24 )' and the si s n is + Or if. For the interchange of two rows is equivalent to the interchange of the suffixes. IV. 14523.. ? t Q) = (I 2 4 X 3 5 ) = ( l 2 H 2 4 H 3 5 )> and tho si S n is ~ ( T h * e last ste P is \ <u 4 o JL o / unnecessary if we use the rule referred to. and so the sign is . The remaining terms of A are derived from a 1 6 2 c a by keeping the letters in their natural order and arranging the suffixes in all possible ways. The interchange of two rows. The results in the two cases are identical. giving the same results. Let A The leading term of each determinant is a^c^. wo count the number of * inversions ' (Ch. Hence in either case the sign of every term of the determinant is changed. Theorem. showing that three. of these are required. 12345. . If two rows or two columns of a determinant A are identical.

b%B 2 4. A can be expressed in any of the six forms : a 1 A l + b l B l -f CjCp a 1 u4 1 + a 2 A 2 4. B and C. and one element and one only from each column. Thus we have in succession *3 & 3 c. the interchange of two suffixes producing a change of sign. bring the row containing c 3 to the top. By the interchange of consecutive columns. Let A Every term in the expansion of A contains one element and one only from each row. Hence. Hence A l is equal to the determinant (6 2 c 3 ). b } B l -f b 2 B 2 c x C l + c 3 C 3 . To determine A^ every term of A which contains a 1 can be obtained from a^^s by keeping the letters in the natural order. . where A l contains no element from the row or column containing a j? a similar remark applying to every A. say C 3 . Expansion of a Determinant in Terms of the Elements of any Row or Column. obtained by erasing the row and column in A which contain a v To determine any other coefficient.122 EXPANSION BY ROW OR COLUMN 6.CgCg. bring the column containing c 3 to the extreme left. By the interchange of consecutive rows. retaining 1 as the suffix of a and arranging the other suffixes in all possible orders.

Similar reasoning applies to a determinant of any order. Every interchange of rows or of columns introduces a change of sign. The number of interchanges is equal to the number <__ of vertical and horizontal moves ' (indicated by arrows) t required to bring c 3 to the left-hand top corner. Expand the determinant A~(a 1 b z c B ) with reference to the element* (i ) in the first roiv. c 8 Hence. Prove that a h g \^abc + 2fgh-af 2 -bg 2 -ch z . 2. (ii) in the second column. 1.and now c 3 occupies the top left-hand corner. <7 3 is equal to the determinant obtained by erasing the row and column containing c 3 . h b f \ 9 f c\ . The results are (ii) J-- - -63 a a 6 2 Ex. multiplied by ( . MINORS AND COFACTORS 123 Ex. and the relative positions of the elements not belonging to the row and column containing c 3 is unaltered. which in ^ this case is 4.1) 4 .

by the rule given above Similarly O 4 = ( .l) r + s . For the element h r which is in the r-th row arid the s-th column. it will be seen that the minor of ^ 2 = (a 1 6 3 c 4 ). Find the cofactors of d 2 and c 4 in the determinant A "(a^b^d^).9-1). A=a b f-h f * h b g f = a(bc -/ 2 ) . (7 4 .h(hc -fg)+g(hf-bg) = abc 4.1) 4 ^" 3 (a 1 6 2 c? 3 ) = . Ex. Omitting the row and column which contain d 2 .e. 7.2fgh . The sign is determined as follows : Count the number of interchanges of consecutive rows and columns (i.(a^b The student who is acquainted with * partial differentiation ' should observe th at .Expanding with reference to the first row. The sign is + or . 1. The rule may be stated thus.or/ 2 . For the number of * moves ' required to bring h r to the left-hand top corner is (r-l)-f (. Denote the cofactors in question by Z) 2 . the cofactor H r = (. according as this number is even or odd.bg 2 . For the determinant the minor of any element is the determinant obtained by omitting from J the row and column containing the element. The cofactor (H r ) of an element (h r ) is the coefficient of that element in the expanded form of J. and is therefore equal to the corresponding minor with the proper sign prefixed.ch*. of vertical and horizontal ' moves ') required to bring the element to the left-hand top corner. (the minor of h r ). Minors and Cofactors.

124 MULTIPLICATION OF DETERMINANT 8. Similar identities hold for determinants of any order.. . where A 19 B ly .C..A.B.. // every element in any row. (ii) 6 1 4 1 -f b 2 A 2 + b z A^ = 0. For every term in the expansion of a determinant A contains one element and one only out of each row. therefore a 2 A^ + b 2 B l ~f c 2 C l = 0. c 2 for a 19 b l9 c 1 in A. or in any column. 6 2 . The second identity is obtained by putting b l9 6 2 . 9. b 3 for a l9 a 2 .. of a determinant is multiplied by a number k. are the cof actors of a l9 b l9 . and is the refore equal to a determinant with two rows identical. For the determinant A = (aj)&) we have a number of identities of the following types : (i) a 2 A l + b 2 B l + c 2 C l = 0. Thus '2 ^2 '3 ^3 a 2 ka . and one element and one only out of each column. (hen the determinant is multiplied by k. Theorem . Important Identities. where A 19 B 19 C^ are independent of a l9 b l9 c v Hence the expression is the result of substituting a 2 . a 3 in the e q ualit 7 A = a^! + a 2 ^ 2 -f a 3 ^ 3 .

Theorem. a 2 o 2 c \ oc 3 6 3 c. a 3 . 1 1 . a 3+ a 3 a 19 a 2 . A 3 are independent of a-. 3 where A^ A 2 . Denote the determinants on the right by A 9 A' 9 and the one on the left by A". A determinant is unaltered by adding to the elements of any row (or column) k times the corresponding elements of any other raw (or column).^ & 2 + a 2 . a 2 . a 3 . we shall prove that e. and J" = a+a-4-f a + a J + a + a^ 125 Similar reasoning applies in all cases. Now A 1 is obtained from J by substituting oq. if all the elements in any row. then k is a factor of the determinant. 3 . a 2 . And A" is obtained from J by substituting a^+v. Hence A f =oc l A l + a 2 ^ 2 + a 3 ^ 3 . where k is any given number. a 3 for a l9 & 2 . 10. ADDITION OF BOWS OR COLUMNS We have A a^A^ + a 2 A 2 -f. = a 3 + kb 3 6 3 c 3 .Conversely.a^A. or in any column. and can be taken outside. Theorem.. have a common factor k. For example. we shall prove that = a. For example. A determinant can be expressed as the sum of two determinants by expressing every element in any row or column as the sum of two numbers.

a column (or row) is replaced by one w hich contains the elements of that column (or row) all multiplied by a number h. 126 . For instance. then the determinant is multiplied by h./c x ha + I U 3 C 3 NOTE. as in Art. (ii) When more than one column (or row) is changed by a succession of changes of this kind. E x. 1. at least oni column (or row) must be left unaltered. 11. and the remainders are put in the second column. as in the last part of Art. if the second column is subtra cted from the first. ? 2 6 2 c 2 Note that by a second application of the above. (i) When.Denoting the determinant on the right by A and that on the left by A'. the sign of the determinant is changed. X -I. In simplifying determinants the following points should be carefully observed. we have /r = a c \ + Now = * ^ 6 X c x -0. 13.

a is a factor of all the elements in each of these (r . For A can be expressed as a polynomial in x. then x-a is a factor of A . and the result follows.) J For r 1 of these rows can be replaced by rows in which the elements are the differences of corresponding elements in the original rows . 1. and J=0 when x = a. Theorem.a) r ~ l (is a factor of A. ifr rows become identical when a is substituted for x. then (x . determinant on the right. 13. Ex. j 8 7 11 13 9 12 4 7 10 3 5 . Examples. Find tlie value of the. K)/ Again.\ ) substituted rows .APPLICATION OF REMAINDER THEOREM 12. If the elements of a determinant A are rational integral functions of x and two rows (or columns) become identical when x~a. so that the result follows by the Remainder theorem. The determinant in question is denoted by A. hence x .

1 1 =! .1 .14 1 6 15 We have A : 6 1 1 . 13 -3 -1 -22 ! 1 19 -2 -23 -22 7 7 -4 -9.1 31 11 -13 9 ! 14 .

23 -! 19 2 23 31 11 -13 31 11 -13 62 3 -21 i -19 5 .2 . .-5 -8 -13! ! (>2 3 -21 13 . ^i 19 .

therefore k ~ . the 3rd is subtracted from the 2nd. Hence Since J is of the sixth degree in a. // oj is a cube root of unity. 2. Prove that 1 1 1 -)(-)( -8) (0-8)(y -8). . prove that a -f&eo +c<o 2 is a factor of 127 a b c b c a cab Hence show that the determinant is equal to . are factors. J =19( . 31)^10209. If a/?. etc. the 3rd column is taken from the 2nd to form a new 2nd colum n. six t imes the 3rd column is taken from the 1st. 3.y. METHODS OF EVALUATION Ex. [In the first step.19 . Similarly ft . and consequently a-/? is a factor of J. k is independent of th^se quantit ies. In the second step. two columns in A are identical. /?. Now the term j5y 2 S 3 occurs in J.1.. and th e 4th added to the 3rd.. the 1st column is taken from the 3rd to form a new 3rd column. and on the right-hand side the coefficient of th is term is . and twice the 1st column is taken from the 4th to form a new 4th column.k. 11 +2 . 13 + 23 . 31) -f 5(19 .Expanding with reference to the bottom row. Ex.. .

and so on. similarly The rest follows as in Ex. This method applies to a determinant of any order in which the second row is derived from the first. Ex. by a cyclic substitution. 1 a c o 2 b a jab and a are factors. a 2 . the third from the second.#cu 4 6o> 2 ct 6 . 2. a4&o>4co> 2 is a factor of J.To the first column add CD times the second and w z times the third.*. therefore A | a 4 bat + co> 2 b c 6-f cco-foco 2 c a j C 4. 4. Prove that (c+a)* (6-fc) 2 (&4c) 2 a 2 -(&4c) 2 a 2 c 2 (a Subtracting the sum of the second and third rows from the first and taking 2 out side.

2abc(a 4&4c) 3 . a 4 6) 2 ^(a + 6+c) 2 (&4C) 2 6 2 c 2 a -6 -c a-6 -c c4a-6 a4fe~c be b* -6 a + b - 2(a-f6+c) 2 6c bc + ba .

if . =0. sinc e the cyclic substitution (a.b -he) 2 is a factor of J. we have whence N 2.c) 2 and in this there are three identical columns. Alternatively. If we substitute for a. which is equivalent to multiplying by be . 4 1 1 1 4 1 1 1 4 . hence a is a factor .(6 4. 12. 128 RULE OF SARRUS Hence. 6.c) is substituted for a. since A is of the sixth degree. hence. the remaining factor necessarily being of the first degree and symmetrical.ac + bc In the last step the first column is added to b times the second_ajid to c times the third. and similarly b and c are factors. Putting o = l. c~l. c) gives a determinant equal to A. (a 4. Again. by Art. b = 1. (-a) 2 a 2 a 2 6 2 (-6) a 6 2 c 2 c 2 ( .

proceeding from left to right. The positive terms in the expansion of (aj}. SIMPLE EVALUATIONS 1.according as the line which determines it is drawn downwards or upwards. Apply the rule of Ex. Hence the rule of Sarrus.^) are fliVa* btCMz. 2.= 27iV. Negative terms a 2 b.c 3 j. Cjtta&a. A more convenient way of using the rule in numerical cases is shown on the right. the sign of the produ ct being -f or . EXERCISE XV A.'3 ' C 3^ a 3 "3. which is as follows : V*/ a 3.--' '2-> ^r - D=P-N The determinant (!^ 2 C 3) is the sum of the products of the three elements lyin g on each of the six lines in the above diagram on the left. 1 to show that a h g h b f .

. 6. In the expansion of occur. If T is any term in the expansion of (a^Ca . Show that. 3. In the expansion of (a occur. In the determinant (a 1 6 a c 8 .g f c = abc -f 2fgh . according as r is odd or even.] 5.) and another term T' is derived from T by a cyclic substitution involving r letters. sign show that the terms -f aj show that the terms 4. in the expansion of (^i -f c following terms occur : and that the other positive terms are obtained from these by the substitution SIMPLE EVALUATIONS 129 4. . Z M ).a/ 2 .. [Exx. .bg* . the order of the suffixes being unchanged. 1 and 3 are instances. the sign of the term a w 6 n _ 1 c w _ 2 ^i> forming the ' second diagonal ' has the sign -f if and only if n is of the form 4k or 4& + 1.a 2 6 4 c 6 ^ 1 e 8 / . 7. then T and T' have the same sign.ch 2 . or opposite signs.

Prove that (i) I 2 2 2 3 2 = -8. (ii) I 2 2 2 3 2 4 2 =0. (iii) 7 11 13 a 6 a 2 6 2 2a6 17 19 23 6 a 6 2 2a6 a 2 29 31 37 9. . Expand the following : (i) a 6 . ii) 2a6 a 2 6 2 .8.

(iv) a b b b =-(aOla a 2 a b a a a 2 1 a a a b a a a 2 .2 2 32 42 2 2 32 42 5 2 3 2 4 2 5 2 3 2 4 2 5 2 6 2 42 52 02 72 (iii) 1 a a 2 -l+a* + a 8 .

y . m n a. 2a6 l-a 2 4-6 2 2a 26 -2a l-a 2 -6 2 10. Prove that I a. Prove that m l is equal to -^-M^-f m 2 +n z )-t(L l where L lf Jlf 2 .1 b b b a (v) 1-K & 2 -6 2 2a6 26 =(1+ a 2 -f 6 2 ) 3 . ^ 8 are the cofactors of 1 19 m^. n 3 in 11.

] 13. -e. w have any values whatever. -d to form a new top row. [From the first. m.where Z. 1 1 I i 1 1 1i i i i i a "*" a ^ a a i ' and that a similar result holds for a determinant of the nth order and of this form. Prove that + . Prove that abe -6 a -d -c d a -d -c b d = ( c b a [Multiply the rows by a. y times the fourth column.] .] 130 ADDITION OF ROWS OR COLUMNS 12. [Subtract the top row from each of the others and expand with reference to the top row. -6. second and third columns respectively take a. /?.

a 6 c 6 c d tc n bed c d e w 88 c d e o [From the fourth column subtract x 2 times the first column. and the third. If u and u ll = rove that a 6 c w u = -M. 2x times the second. prove that .14.'y 2 .] .

If u=ax* y* -xy x 2 ax + by bx + cy a'x + b'y b'x + c'y 1 y ax + by a'x \-b'y b c ' b' c' [Add x times the second column to y times the third : then add x times the first column to y times the second.(6 - APPLICATION OF REMAINDER THEOREM 131 .15.] 17.c) f .b z ) (ax 2 4 2bxy + cy 2 ). the second by y. and subtract from the third.by b c bx + cy ax + by bx + cy [Multiply the first column by x. Prove that a b ax 4.(a . Prove that the determinant a -x bed b c -x d a c d a~-x b d a b c-x is equal to (x -a -6 -c -d)(a? -a 4-6 .(ac . 16.c +d){x* .] = .

c0. &~0. prove that (.B. METHOD OF FACTORS 18. -.s-a) 2 (s -a) 2 6 2 (*~6) 2 a 3 (a . 5 6.c. and find A' by putting a 2. If 25 ~ a + b 4.] 20. Express the following in factors : (i) 1 1 y (ii) (iii) A-h/x 1 . Prove that x + a bed -x*(x + a + b+c + d).c are factors : deduce that Ns is the remaining factor. s .a.6) a (s-c) 2 [Show that s' 2 . a x+b c d a b x+c d a b c x+d 19.

If A-~y4-aS.b + c . show that a 4. prove that 22. the second by c'a'. Put a/a' A .(a + b + c 4. Prove that be bc' + b'c b'c' ca caf 4. c/c' = y. the third by a'6'.(b .] . Minors of a Determinant. Prove that x 3 a.d) { (a . 2 x I a 3 a 2 a 1 y 3 y 2 y i By equating the coefficients of x on each side. show that 1 a u a 2 1 y 3 y 2 i 132 COMPLEMENTARY MINORS 24.d) (a . '6' /x. The determinant obtained by omitting any number of rows and the same number of columns from a given determinant A is called a minor of A.baj + cw z + dw 3 is a factor of abed b c d a c d a b d a b c Hence show that the determinant is equal to . and use the last example.21.c'a c'af ab ab' + a'b a'6' [Divide the first row by b'c'. . fA = I I I A fJL V A 2 ^ v a Hence show that the following determinant has the same value : 23.c) 2 -f. 14.d) 2 }. If w is a root of # 4 = 1.

for example. the resulting determinant is called a second minor. observe that abcde occurs. Take. If from A we omit the rows and columns which contain every element of a given minor. Expansion of a Determinant by means of its Second M i n o rs . every term in A contains one element out of the first row and one out of the second. of the form where the first factor is any minor formed from elements of the first two rows (or of the first two columns). and the second is the complementary minor.a^b^c^d^) . Hence every term in the product (^^(c^d^) occurs in the expansion. Thus (a 1 6 2 ) an d (c 3 eZ 4 e 5 ) are complementary minors of By the interchange of rows and columns. and so on. we get another minor which is said to be complementary to the first. Again. the term suffixes in the same order. each with its proper sign. any minor of A can be put into the left-hand top corner. The sign of a product is determined as follows. Hence A is equal to the sum of all the products. the sum of the terms which contain a v h z is and the sum of those which contain a 2 b l is . and then the complementary minor can be seen at a glance. Consider the determinant In the expansion of A. 15.If we omit two rows and two columns. Keeping the . and these elements have different suffixes. EXPANSION BY MEANS OF MINORS 133 In the expansion of this product.

.I) r + r '+*+*'(h r k r >) .. rif' .. 7 by counting the number of moves ' required to bring h r to the place occupied by a x and k r > to the place occupied by 6 2 Thus. '. 20. In the c double-suffix ' notation of Art.Therefore.. NOTE. k r ' .. . k r > to the s'th column. h r * belong to the sth column and k r . h r . this term occurs with a negative sign. if any two rows of a determinant A are a r b r c r . the sign of (b 1 d 2 )(a 3 c 4 e^) is given b y In a similar way we can expand a determinant in terms of the rth minors which can be formed from any r rows (or columns). k r .... . The summation is to include all the minors which can be formed from the two rows. (the complementary minor of (h r k r >)).. Prove that if Pi Pz P* m l n t rc 2 . in the instance given above.. df b r r c r ' .. More generally. The sign of the product may be determined as above. in the expansion of J.. and the sign of the product is . . the rule takes the following convenient form: In (a>j> q a rs . Expand A = (a 1 6 a c 3 rf 4 ) in terms of the minors of the rows contain ing a l and a. t hen A -ZX . . where h r . 2. or as in Art. the sign is ( 1 VP+^'t" *"+*+ Ex. . 1. It is easy to verify that +(A)(to) Ex.) (complementary minor).

Product of Two Determinants of the same Order. 134 MULTIPLICATION OF DETERMINANTS 16. A is transformed into A' by the interchange of the first and fourth. therefore Again.b l m 2 b l . Whence the second result follo ws. the term a^c^^m^n^ occurs in the expansion of J. which can be formed from the elements of the first three rows. the second and fifth and the third and sixth columns. As a typical case we shall prove that 4.6 1 m 1 a^ 4. and the complement ary minor is Moreover. the only o^e not equal to zero is (a 1 6 a c a ).aiui m^ m 2 <>i c 8 c* Pa Considering the minors of J.

Denote the determinants on the left by A and A . 6. I +v . -1 -1 a l a 2 -1 Cinvn -1 where the second determinant is obtained from the first as follows : To the third column add l v times the first and m l times the second : to the fourth column add L times the first and m times the second. Hence. JJ'H-1) 2 . then we have a. as in the last example.

2 4. where 4. and therefore the sign is . two of which are and ^A.b^w a 3 Z 1 4. and the first is equal to With regard to the sign. and by the same method.b^i The determinant A is the sum of 27 determinants. observe that Iim 2 n 3 is transformed into by the same number of interchanges of letters as those required to change a 1 6 2 c 3 into b^c^.6 1 m 3 4.b^wii + c^n-^ al 2 4.bfl whence the result. This number is odd.c x 4. It is required to show that aJi 4. Second method.-1 -1 aj.b 3 m l 4.c^ a 3 2 4. The product of two determinants of (he n-th order (n>2) can be expressed in a similar form. .6 3 m 3 wliich are the same as ^i a i c \ b 2 a 2 c 2 and b 3 b 3 The second of these is zero.

. In general. 5 . a 2 "2 C 2 li m l n^ \ 7 Wl W fro I lit) tin Proceeding as in the last article. all are equal to zero except the six of the first type. Consider the arrays i a l !. A similar result holds for determinants of the nth order (n = 4. we form the determinant J = ! al + bm + cn By reasoning similar to that in Art. and the other factor is a term (with the proper sign) of (Iitn 2 n$). 17. 6 J s in another and c's in the third. for reasons similar to the above. and is sometimes called a matrix. If the number of rows is not equal to the number of columns. (a^c^) is a factor of each of the six..). there fore J = (a 6 c ) (I m n ). A number of symbols arranged in rows and columns is called a rectangular array. 16 it will be seen that This process is called multiplying the arrays. the determinant formed thus from two such arrays is equal to the sum of the products obtained by multiplying every determinant which can .Of the 27 determinants. Rectangular Arrays. RECTANGULAR ARRAYS 135 Moreover. the array is generally enclosed by double lines. in which a's are in one column. (1) Let the number of columns exceed the number of rows.

as in Art. This follows. If from the arrays a 2 b 2 I 1 2 a 3 6 3 || Z 3 we form the determinant ^2^3 -f b 2 m^ it will be seen that J =0. prove that ^2 ^3 &A . 16. (2) Let the number of rows exceed the number of columns. 1 . or from the fact that J = + bm 22 . I ^ m l 1 2 m 2 Z 3 w 3 In general.be formed from one array by the corresponding determinant formed from the other array. the determinant formed by multiplying two such arrays is zero. & x 6 2 b. 136 RECIPROCAL DETERMINANTS Ex. // * r = a r + r + y r + 3 r .

S 3 5 4 5 5 S 6 The first two results are obtained by squaring the arrays 1111, 1111 aj8y& a j3 y 8 a 2 2 y 2 S 2 The last is got by squaring the determinant in Ex. 2 of Art. 13. Ex. 2. Prove that

2(a t - a z ) (o a - a 3 ) (a 3 - ^5(6! - 6 2 ) (6 2 - 6 3 ) This result is obtained by forming the product

a, I

-2b 1 6j 2 -26 2 6 2 2

18. Reciprocal Determinants. Let ^ = (^63 ... k n ) and

where ^ij, B l9 ... are the cofactors of a l9 b v ... in A. For reasons connecte d with linear transformations, A v B 19 ... are sometimes called inverse elements, and A' is called the reciprocal determinant. (1) // A is a determinant of the n-th order and A' is the reciprocal determinant, then A' =A n ~~ l . The .following method applies in all cases. Let A==(a i b 2 c 3 ), then using identities of the type a l A l + fej JBj + c l C l == A , aA% -f 6^82 + c i@2 " 0>

it will be seen that

A J A

whence it follows that

J'= J 2 .

SPECIAL METHODS OF EXPANSION

137

(2) Any minor of A' of order r is equal to the complement of the corresponding minor of A multiplied by A r ~ l , provided that A^O. As a typical instance, let J = (a 3 6 2 c 3 d 4 ). We shall prove that, if . A and (A,B<>C) =<LJ 2 .

then We have

c 2

4 6 4 C 4

therefore

A 7? C^ 1 000

and

1

A c a d 1 A c 2 d 2 c 3 d 3 c, d.

The second identity is obtained by multiplying A by jLjL-\ JLJ-\ Oi /y. 1 11 1 juL 2 -^2 2 2 AS B 3 C 3 D 3 0001 19. Two Methods of Expansion. (1) With reference to the elements of a row and column.

Let

and ^ =

m

m n r

Let ^j, jBj, ... be the cofactors of a v b v ... in A. In the expansion of K, the sum of the terms containing r is Ar : every other term contains one of the three I, m, n and one of the three V > m', n'.

Again,

r

and

are complementary minors of A ;

hence, coefficient of IV in K= coefficient of a^r in K = - coefficient of a x in J = - A l ; and similarly, coefficient of mri m K= - coefficient of c 2 r in K = - coefficient of c 2 in J = - C 2 . Thus it will be seen that K Ar - {AJl' + B 2 mm' + C z nn f 4- C 2 mn' -f B 3 m'n + A 3 nl' + C^ril + BJm' + A z l'm}.

138

DOUBLE-SUFFIX NOTATION

(2) With reference to a diagonal.

Let

&!_ C x d l

a 2 c. 2 3 &3

and K =

a 2 x 2 c 2 d 2

Let ^ /? 2 , C 3 , D 4 be the cofactors of the elements in the leading diagonal of J, then the value of A' is Cr^efy? 4 + #3^64 -f x^ -f 7^4 -f # J B 2 -f 03 + x)

Note that the expression in brackets is the sum of x^x^a^ x } xj) 3 c 2 and terms derived from these by the cyclic transpositions (a&c), (123). The verification is left to the reader. 20. Definitions. A determinant of the nth order is often written in the form

n

lw

Denoting any element by a rs , the determinant is said to be symmetric if a sr = a rs . If a sr = -a rs , the determinant is skew-symmetric: it is implied that all the elements in the leading diagonal are zero. For example, if a h g 1 , L z y h b f m ~ z Ox g f c n -y -x I m n the determinant K is symmetric and L is skew-symmetric. We also say that the first determinant is bordered by I, m, n. 21. Symmetric Determinants. (1) // A rs , A 8r are the cofactors of the elements a rs> a sr of a symmetric determinant J, then A r8 = A sr . For A r8 is transformed into A sr by changing rows into columns. Thus if ^ = (011 a*>. OM),

a u a 12 a 31 a 32

a 21

Z 32*

SYMMETRIC DETERMINANTS (2) For the determinants J= a h g , K= a h g I h b f h b f m 9 f c 9 f c n I m n we have the following important relations : (i) EC - J 2 - Aa, GH-AF = Af, etc. ; (ii) K = - (Al* 4- Bm* + Cn*+ 2Fmn 4- 2Gnl + 2TZm), where A, B, ... are the cofactors of a, 6, ... in J. These follow from Arts. 18, 19. (3) Consider the following symmetric determinants :'

139

a h 9 , K~ a h 9 I

h b f

h b f m 9 f C

9 f c n

I m

n r

Let A, B, ... be the cofactors of a, b, ... in J. We shall prove that, if J=0, then AK==Z -(Al + Hm + Gn)*. Because A = 0, the coefficient of r in the expansion of K is zero, therefore - AK = A (Al* + J3m 2 + Cn 2 4- 2*W -h 2flW 4- 2ff Iro). Moreover, ^J5-ff 2 = Jc, C^-(? 2 -J6, GH-AF=Af, etc. Hence AB = H 2 , AC^ G 2 9 AF = G^F, etc. , and therefore - AK = ^ 2 Z 2 + # 2 m 2 + 2 n 2 + 2GHmn + 2^GwZ -f 2AHlm = (Al + Hm + Grip. In the same way we can prove that - BK = (TZ 4- Bm 4- JFn) 2 and - CX - (67 H- J?m 4- Cn) 2 . More generally, using the notation of Art. 20, let A n = (a n a 22 ... a nn ), ^ n -i==(a n a 2 2 a n-i> n-i)> the determinants being symmetric, and let A U) A 12 , ... be the cofactors of a n , a 12 , ... inA n ., r Then, ifA n _ l ) A n A n (A u a ln + A 12 ^2n + * A!, n _ 1 a w _ 1 , n ) , with similar values for A 22 A n , A^A n , .... For, expanding as in Art. 19, (1), since a sr = a r8 , A 8r ~A r8 and J w -i == 0, we have A n - - (^ n a ln 2 + ^ 2 2 a 2n 2 + + 2^ 12 m2n + ) Now, since A n _ l = 0, every minor of the reciprocal determinant (-^11^22 ^n-l> n-l) is zero. Hence, by Art. 18, (2), A n A 99 ^A^ and A^A^^A^A^. Using such identities, it follows that A n A n has the value stated. K B.C.A.

140 SKEW-SYMMETRIC DETERMINANTS (4) The following result is of great importance. Consider the symmetric determinants a-x h g h b-x f 9 f c ~ x

a-x h h b-x

a-x h g I

h b-x f m g f c-x n I m n d-x

Denote these by J 2 , J 3 , J 4 . Let A^ a-x, and suppose the sequence Jj, J 2 > ^3 t be continued indefinitely by constructing determinants of the above type. Then the roots of the equations 4=0, J 2 =o, J 8 =o,..., are all real, and the roots of any one of them are separated by those of the preceding equation. If A 1? A 2 (A 1 <A 2 ) are the roots of J 2 = 0, then A x <a<A2 (see Exercise XI, 2). Also, as in the last section, if J 2 = 0, then A^A^ is negative. Hence if x has the values - oo , A 1? A 2 , oo , the signs of zJ 3 are -f - -f - . Therefore the roots of J^O are real, and denoting them in ascending order by fa, /x 2 , /u, 3 , we have /x t < A 1 </z 2 < A 2 </n 3 . Again, as in the last section, if ^1 3 = then A^A^ is negative; hence, if x has the values oo , fa, /^ 2 , /^ 3 , oo , the signs of J 4 are +,-,+>-> + ; therefore the roots of J 4 = are real and are separated by those of Zl 3 = 0; and so on without end. 22. Skew-symmetric Determinants. (1) For the skew-symmetric determinant (^u 22 , ... a nn ), if A rs , A 8r are t he cof actors ofa rs , a sr , then A sr = ( - l) n ~ l A rs . For A r8 is transformed into A sr by changing the sign of every element and then changing rows into columns ; and since A rs is a determinant of (n-1) rows, the result follows. Thus in the determinant c b -c a -6 -a

the cofactors of c and c are -c a -b

and -

-a

and one of these is ( - 1) 2 times the other.

THE PFAFFIAN

141

(2) Every skew-symmetric determinant of odd order is equal to zero. For if A is a skew-symmetric determinant of odd order n, and A' is the determinant obtained by changing the sign of every element in A, A' = (-\} n A=-A. But A' can also be obtained from A by changing rows into columns, so thatJ'=J: whence the result follows. For example,

c -c -6 -a

-c -b

c b c

-a

-c

a b a

-b -a

(3) Every skew-symmetric determinant of even order is a perfect square. (i) Let A n = (a u a 22 . . . a nn ), J n ^ = (a n a 22 . . . a^, n _ x ), where n is even and the determinants are skew-symmetric. For n = 2,

*12

n 2 12 .

Let ^4 U , ^4 12 , ... be the cofactors of a n , a 12 , ... in J n _!. Since n i s even, by the preceding A n _ l = and A sr = (-l) n ~ 2 A rs = A rs . Remembering that a sr = -a rs , exactly as in Art. 21, (3), we can show that A ll A n = (A ll a ln + A l ^ 2n + ...+A lin _ l a n _ ltn )* ................. (A) Hence if A n is a perfect square, so also is A n ; moreover, A u is a skewsymmetric determinant of order n - 2 ; and it has been shown that such a determinant of order two is a perfect square, therefore the same is true for those of orders four, six, etc. (ii) If n is even and J n = S n 2 , the value of 8 n * is given by the following rule. Write down all the arrangements in pairs of the numbers 1, 2, 3, ... n. Denote any such arrangement by (M), (Im), ... (uv), then %n = Z( a hlfllm a uv)> where the positive or negative sign is to be taken according as the arrangement hklm ... uv is derived from 1234 ... n by an even or by an odd number of transpositions. The term a l2 a u ... a n _ lt n is to have the sign + . (a) The rule is suggested by the case when n = 4. By equation (A),

where A n , A 12 , A 13 are the cofactors of a n , a 12 , a 13 in J 3 , so that

Whence we find that

A 12 = -

which agrees with the rule. * Named the Pfafflan, by Cayley, after J. F. Pfaff.

142 VALUE OF THE PFAFFIAN (j3) Assuming that the rule holds for determinants of order n-2> we shall prove that it holds for those of order n. Then, since it holds when n = 4, it will hold for w = 6, 8, etc. By equation (A),

Now J n _ 1 = 0, therefore by Art. 18, (2), etc. Hence

where the values of *JA n , .JA^, etc., are given by the rule. We shall show that the signs on the right are alternately positive and negative. Consider the product P ^m^^n a vn\JA J>v . The values of JA n and JAgp are obtained by the rule from the sets 2, 3, 4, ... n-1 and 1,2,3, ...y-l,p + I, ... n-1. Moreover, every term in the expansion of P, with the proper sign, must occur in J n . For clearness, the element a rs will be denoted by (rs), so that (sr) = - (rs). If p is even, the product P contains a term M, which is the product of

and

By changing the order of the numbers in each of the J(n - p) 4- 1 pairs which are underlined, it will be seen that ( - !)*(*-*)+* jf j g the product of elements (rs) in which corresponding values of r y s are as below. 1, 2, 3, ...p-2, p-1, p, p + l, p + 2, 2> + 3,...w-2, n-1, n, 2, 3, 4, ...y~l, p + l, n, p, p + 3, p + 2,...n-l, n-2, 1. The numbers in the second line can be written in the same order as those in the first by %(n+p) transpositions, and since %(n-p) + l+i(n + j?) = n + l, an odd number, it follows that M is a term in J n , so that -#,> nN Mj>j> occurs in the value of 8 n . // p is odd, the corresponding process requires changes of order in i(n -p + 1) pairs, and i(n + p - 1) transpositions. Since the sum of these numbers is n, which is even, +a pn *JA p9 occurs in 8 n ; thus

It remains to be shown that the value of S n can be written down by the rule. It is obviously necessary only to consider the first term in the value of <

MISCELLANEOUS METHODS This is (pn)(l2)(34:)...(p-l,p +

143

or

according as p is even or odd. Now the arrangement p, n, 1,2,3, ..._p-l,p + l, ... n-l can be derived from 1, 2, 3, ... n by ^ + n-l transpositions. The rule asserts that the sign of the term is + or - according as p + n - 1 is even or odd, that is according as p is odd or even, which agrees with what has been proved.

EXERCISE XVI A. MISCELLANEOUS METHODS 1. If a + 6 -I- c =0, solve the equation a-x c b 0. c b -x a b a c-x 2. Prove that (&iC 2 ) (#^2) "f ( c i a z) ftA) -f (0iW (^1^/2) =0. [Use the identities tti(6 l c 2 ) + 6i(c 1 o 2 ) + c 1 (o 1 6 2 )=0, a 2 (6 1 c. 2 ) 4-62(0^2) -f 03 (^62) =0.] 3. If (a l b 2 c 3 )=0 and (^62^3) =0, prove that (& 2 c 3 )

4. Prove that

5. Prove that

c a -fa 2 a a

- a 2

ca -

6. If

l/(a + ) !/(& + *) l/(c + a)

''Q'

where Q is the product of the denominators, prove that

144

MISCELLANEOUS METHODS

7. Show that

6 a

c a b b a c a b c

8. Prove that the ' skew * determinant

= -(a + 64-c)(6-hc-a)(c+a-6)(a + 6-

is equal to

9. Prove that

c 6 6-c

c a c -a 6 a a -6 a 6 c a + 6-f-c

x c b I

c x -a m

b I am x n n x

=2(a + 6 + c){a6c - (6 - c) (c - a) (a - 6)}.

10. Prove that

is equal to (a + l)(6-

-1 6 c d

a bed a + ..a -1 c d a 6 -1 d a 6 c -1 l)(d-f -./.

13. or. Prove the following : (i) - (ii) (Hi) (iv) _ j a -1 6 -1 c 05 . etc. .1 6+1 c+ 1 d+] 1 lj [This follows from Ex. 1 +a 2 . by the same method. directly.1/6.I/a.] 11. XV. by putting 1 +a l ~ .

y 2 -1 c 6 x = C ~a y -6 a z *' y' z' w c 2 6 2 1 = (CL c 2 a 2 .

1 6 2 a 2 1 1 1 1 c 2 6 2 d 2 1 c 2 a 2 e 2 6 2 a 2 / 2 .

m. = .8. -n I m' . prove that n w /' =4 Inl'n'.8. y .j8 by /.a. y . Denoting the squares of the differences ft . -6)(a + 6-c). m' 9 n' respectively. /? . where 2*= PRODUCTS OF DETERMINANTS 145 12.ax by cz. n and those of a .(ax + by + 02) (ax' + 6t/' + cz').8 by I'. a .y.d* e 2 / 2 1 .

If J = 4 5 6 a. 5 6 7 y 2 2 3 2 4 2 5 2 y 6782 32 42 5 3 6 2 2 . ya + jSS^/x. etc. then 11' ( /u-v) 2 .-m -I On' [If jSy + oc8 = A. A' = p 2 2 3 2 4 2 a.] 13. aj3-fyS=v.

then (x . provided that the common elements are zero or are divisible by (x o) 2 . I/ 2 42 52 6 2 '7 2 w x y z w prove that 14. If in a determinant A. (ii) p rows become the same and q columns become the same.2.a) p + ff ~ ~ a is a factor. Prove that . B. Write down as a determinant the product : y b c a z x y Z X Hence show that the product of two expressions of the form a 3 + 6 3 -f c 3 . PRODUCTS 15. when x= a : (i) p rows become the same. then (x -a) p ~ l is a factor of J . 16.3a bc is of the same form.

[Consider the product 17. prove the identity on the right. c 2 6 a = (a 3 -f 6 3 4-c 3 ~3a6c) 2 . Ica6 2 a 2 a* 2a6-c 2 a b c . By forming the product on the left. .

/. .j .-a c ft 1 b c a -b a c c a b -c b a a* c. -26.2& 2 Oj 24. c. a.

n = (ab') 9 V = (ad').la' where l=(6c'). rf and 2. c. m' = (M'). 20. c'. m = (ca'). d'. By forming the product a + ib c+ld -c + id a . 5 for a. prove that the determinant A is zero. 6.3 for a'. Express (P + 2 2 4-3 2 + 5 2 ) 2 as the sum of four squares by taking 1. 5. By multiplying the arrays on the left. I V m m' n n' V I m' m n' n 2ir Im' + l'm nl' + n'l lm' -f I'm 2mm' mn' + m'n . 19.6 2 + c z + d 2 ) (a /2 -f 6' 2 + c'* + d'*) b'+la' d' + ic' -d'+lc' V . 146 PRODUCTS OF ARRAYS 18. 1.7 12 2J /31 / 2J 3 where J 1 =a 1 c 1 -6 1 1 . I^^ -^b^ etc. &'. n' = (ci'). 2.ib prove that (a 2 H. 3.

Prove that =0.nl' + n'l mn' + m'n 2nn' Hence prove that if ax*2hxy+ by* +-2gX'*r2fy + c \ the product of two linear factors.. 1 -26 t V etc. [Multiply the two arrays of four rows || a^ a l 1 etc. C. Prove that (i) x 1 1 .... 1*1.. MISCELLANEOUS THEOREMS AND EXAMPLES 22. the determinant being of the nth order. then a h g h b f g f c =0. . 21.. 1 1 X .

(ii) a ab a + 2b a + nb a a + b. -1 . (n- a + b a + 2b 23. . If APPLICATION TO GEOMETRY 147 <*1 1 .. a + (n..1)6 a + nb a a + nb a...

-1 1 .. . .... -1 a* 1 ... .. . ..2 1 .......

.. If a b a' V c c' d (TOO =0. .] 24. Such determinants are called continuants. -1 prove that u n ^a n u n [Expand with reference to the bottom row. a" b" c" (TOO a b c d ..

prove that determinant t x y z . (i) By using the fact that the on the right is zero. b' c' d' a" c" d" a? d" b" c" d" 25.a' b' c' d' a" b" c" d" prove that a c d a' c' d' 2 a b d a' V d' x b c d .

L = (ad')> M = (bd') 9 N = (cd'). y.1111 t(y-z}(z-x}(x-y)+x(y-z)(y-t)(z-t) t x y z + y(z-x)(z-t)(x-t)+z(x-y)(x-t)(y-t) = 0. d'. c./?) sin (a .a) sin (a .8) = 0 . b'. (x t y t ) and the equation to the plane through x y 1 -0. d' on lines perpendicular to d. d and a'. &'. z ~ cos 2y + 1 sin 2y. 6. . pro ve that sin(a-8)__ _ _sin(ft- 148 APPLICATION TO GEOMETRY 26./?) + sin (/? . n=(a6 x ). t 2 x 2 y 2 z 2 (ii) Hence show that. /?. respectively. d are parallel t o a 7 . c'. sin (j3 . Prove that the equations (i). 6. By projecting th e circuits a.8) sin (a . c. 6. d and a'. d' are the lengths of the sides of two quadrilaterals such that the sides a. 6. [(ii) In (i) put x = cos 2a 4.8) sin (y . b'. 2 = cos 28 + 1 sin 28. y = cos 2)9 -f t sin 2/J.8) -fsin (a . b. c. c'. m = (ca'). 8 are any angles. If a.y) sin (j3 . (iii) Use the last result to show that if a. a.1 sin 2a. d make angles a. y. c. (ii) x .8) f sin (y -a) sin (y .8) sin (ft . c'. (ii) are respectively the equation to the line joining (x^y^.y) sin (y . (iii) Let a. c. $ with a fixed line. where Z = (6c'). then LMN + Lmn + Mnl + NlmQ.

[Let (x l9 y^ (x 2 . fi.'i Vi *i 2/2 * ^2 2/2 ^2 '3 2/3 2 3 Show also that the last equation may be written in the form '>__'* 41 _ <!/ A/ X/j W Mj ^2 "~ #1 2/2 ~ 2/1 27. y 2 ). y 9 z. C. c. prove that these are connected by the equation c 2 6 2 z 2 1 a 2 y 2 1 -0. x. If ^4. BD. CA. AD. AB. b 2 a 2 z 2 I x* y* z* Q 1 11110 be the coordinates of the points. CD are denoted by a. b. Z> are four points in a plane and the lengths BC. Multiply the arrays 1 ! X 2 l/ 2 1 000 1 -2z .

. the lengths BC. d. (x 2 . CD are denoted by a... the radius of the circumscri bing sphere is E. In a tetrahedron A BCD. prove equation (ii).. (ii) -36fl 2 F 2 =A c 2 b 2 d 2 ' X 2 7/2 Z 2 1 c 2 a 2 e 2 X* 2/3 2 3 1 6 2 a 2 / 2 x t i/ 4 2 4 1 .AD. e.. y 2 . BD. where (x v y lt Zj). . 6. (i) 6F = x l y l z l 1 .AB. are the coor dinates of A y B . CA. / respectively . Assuming equation (i). z 2 ). c.28. the volume is F.

) is called a solution of the system u^Q.d* e* / 2 [Take the origin at the centre of the sphere. Equations which have no common solution are said to be inconsistent. . ..) is called a solution of u = 0. . have the common solution (0. 0.. .. etc. i )^c>. v~0 9 . 11. If the equations have a common solution (x l9 y l9 . . (iv)..). v = 0. If w = is an equation connecting the variables x. .. . . .. which is satisfied when x = x ly y = y ly . j/j.. in which each of the variables is zero : but they are not said to . .. and (x l9 y v ... .. -6RV = x i 2/i z i ~R .] * Note the value of this determinant given in Ex. be a system of equations connecting x. then (xj. y. t/. .).. x% y 2 Z 2 j\ x 2 y 2 z$ It and use the equations z. CHAPTER X SYSTEMS OF EQUATIONS 1... Let u = 0.y... . . Equations which are homogeneous in x. Systems of Equations. they are said to be consistent. Multiply the determinam 6RV= x l y^ z v E .

Moreover. So also are the equations u~ 0. the equations are said to be not independent. .. . which satisfy w = 0. V 9 m' are constants such that Zw'-Z'ra^O. which satisfy the second pair also satisfy m'(lu + mv) .. 2. any values of x.I'm) v = 0. which represent parallel line s (or planes).m (Vu + m'v) = 0.. which are the same as (lm f . y.. To solve a system of equations is to obtain all the solutions (if any exist). any solution of the second system is a solution of the first. v = 0.I'm) u . t> = 0.l'(lu + mv)+ l(l'u + m'v)~0. The equations x + y = 1. these are the same as w = 0.be consistent unless some other common solution exists. then the pair of equations w = 0. (2) // a v a 2 . are inconsistent. . v = also satisfy the second pair. (Im 1 . To eliminate a variable x from two equations is to obtain from these two an equation free from x. is equivalent to the system For any solution of the first system is a solution of the second and. which satisfy the others. x + y = 2. The equations u 0.. 2% + 3v = 4. If one equation of a system is satisfied by all the values of x... . w 2 =0. For any values of the variables z.. In such cases we may say that one system is the same as the other. w n = 0. and since Irri -Z'w^O. are constants and a l ^0 ) the system of equations MI = O. v = 0. y. . since a l is a constant (not zero). Equivalent Systems. (1) Two systems of equations are said to be equivalent when every solution of either system is a solution of the other. Vu + m'v = 0. . Any result so obtained must hold if the equations are consistent. v is equivalent to the pair lu + mv 0. lu + tnv-0 are not independent. This is done by a process of elimination. y. m.0. 150 SIMULTANEOUS LINEAR EQUATIONS (3) If I. It is left to the reader to supply proofs of the following : .. Illustrations..

(3). (i) If (a 1 6 2 )^0. w = and v = 0. W 2 = v 2 . w = 0. the equations are inconsistent unless also (b^) =0 and For if (ai& 2 ) = 0. But the second pair of equations is equivalent to the two pairs w 1 = t. For in this case uj) 2 =zuj)i and u^a^u^a v Hence any solution of one equation is a solution of the other. The latter alternative cannot occur. in which case there is no solution. 2.(4) The pair of equations uv = 0. t.u 2 b s Hence the equations cannot have a common solution unless (b^) and ( c i a %) are both zero. (ft^). 6=0. (2) Consider the equations excluding the case in which either is of the form Ox + Oy + c = 0. (iii) If (a 1 6 2 ). A formal proof is hardly required. or else an a is zero and a 6 is zero. (c^) are all zero. u% = v 2 and w 2 = 0. the chief difficulty is to include all the special cases which may arise. there is an unique solution. e^O. For by Art. and there are infinitely many solutions. 2 = 0. unless <& = 0. 1 . For example. . Systems of Linear Equations. the two equations are one and the same. we have w x 6 2 . M 2 = V 2 is a solution of 1/^2-^2. which are the same as INCONSISTENT EQUATIONS 151 (ii) // (a 1 6 2 ) = 0. suppose that a^O.w 2 &! =0. then by hypothesis fc^O and . the equations are equivalent to Wj6 2 . 3. (5) Any solution of the system ^1=^1. w==0 is equivalent to the two pairs w=0.w x a 2 + u<p t = 0. (1) The equation ax + by + c has infinitely many solutions. In considering this subject.

then (6 x c 2 ) =0 or else a x =0. Ex. (C 2 A 9 ).. B 3 =0. For we have a l C l 4 a 2 C 2 = 0. Whence the result follows. i. // (o^) =0 and (a x c 2 ) ^=0. 3. Three Linear Equations. or else (6 1 c 2 )=0.6 2 : a 2 . .a 2 6 1 = a 1 6 2 = 0. (c lf c 2 ) there is at least one in which neither quantity is zerb. the equations are satisfied by infinite values of x. if (afi^^Q these equations have solutions other than (0. then either C l =0. a^ .. therefore a 2 = and b 2 ^0. (iv) The case in which the equations are inconsistent may be regarded from another point of view. (3) Ifa^x + b l y = and a 2 x + b<$ = 0. Prove that if A =0 and C B = Q.bja l9 . c l C l + c 2 O 2 = 0.afa ~0. Conversely. Hence. a^ =0 . Suppose the coefficients of x. which contradicts the hypothesis. B^-0. then #-><x> . y which are in the ratio . &!<?! + 6 2 C 2 = 0. if a^O. Consider the equations 0. C 2 =0 or else A B =0.. they are consistent.fe x : a l or . it will be seen that (Ci-4 t ). For suppose that fl^O. (c 1 a 8 )=0. (C 2 B 3 ) are all zero. We shall say that in the limiting case when (a^) = 0. // (a 1 6 8 )=0..e. then if a a =0 we should have o^ =0. a a =0. then of the pairs (a l9 a 2 ). (b l9 6 2 ).y = Qor else (a^) = 0 . ^4 3 =0. Or thus: Using identities of the type (^^3)= Ja t . . (6 1 c s )=0 and (c l a 2 )=0. Ex. Hence the equations are one and the same. where a lf b lt q are not all zero and #2 ^2> c a are not all zero. Let A =(0^3) an d l& A\> B oe the cof actors of a lt 6j . 152 LINE AT INFINITY 4. therefore either ^=0.aj} { =0. y->oo and the fractions x/y. then Ex. Cg =0. (OjBa). This is a particular case of the preceding. and therefor e 6 t =0. then either x = 0. y to vary in such a way that (a^-^O while (6^2) and (c^) remain finite .&2/a 2 tend to equality. 1. i. C 2 =0. For a lf a 2 satisfy the equations a x 6 2 . The following results are required later. 2.e.0).

u + k = Q intersect on the line Ox-f Oy + k = 0. (iv) // every C is zero and the equations are distinct. w 2 = have an unique solution. which is the solution of the system. Two Equations in Three Unknowns. we have C^ + (7 2 w 2 Hence.u 3 C 3 = and w 2 = 0. The ideal point in which a straight line w = meets the line at infinity is called the point at infinity on that line. and let . In both of these cases. y which are in the ratio . For by the preceding. then w 3 ==0. in which case each of the equations (assumed to be distinct) is inconsistent with either of the others. then J = 0. y. C 2 7^0. represents the line at infinity. b l9 .. the equations w 1 = 0. This case is excluded. SYMMETRICAL FORM 153 6.. // u 1 =0. we must have AQ. C 2 = 0. (i) // no C is zero. then the necessary and sufficient condition that they may meet at a point (which may be at infinity) is A =0. Parallel straight lines are to be regarded as meeting on the line at infinity. . In this case we may say that the equations are satisfied by infinite values of x. Illustration. w 2 0. the equations are not independent. since C 3 ^0.Let ^^(a^Cg). then u 2 C 2 4. These conventions enable us to include various special cases under one heading. C 3 =Q. be the cofactors of a x . (ii) // CjL-0. w 3 = are the same equation. (iii) 7/C' 1 = 0. For the sake of complete generality we say that the ' equation/ Ox + 0y-hc = where c^O. each is inconsistent with either of the others. Consider the equations .6 X : a r 5. if the equations are satisfied by the same values of x. u l C l + u 2 C 2 + w 3 <7 3 = 0... u 3 =0 (as above) are the equations to three str aight lines. (2) If A = Ae equations are consistent and have an unique solution except when every C is zero . C 3 9^0. (1) //"<A6 equations are consistent. The Line at Infinity. For the lines w = 0. For on account of the identities a l C l + a 2 C 2 -f o 3 C 3 = 0.4^ B l9 .

. 6 2 ........ by Art..) Also we have t^aj-wg^ss -((*&) ... 2. c 2 ) there is at least one..... ....... (F) . . say (a l9 a 2 ).... (b^). 0)). 0..... there are infinitely many solutions....%a 2 + u 2 a l = 0. Wj&jj ~ w^i ^ which are the same as -(aA)y + (Cia 2 )z=(a l d 2 )........ y) is any solution of (A)...) Thus in every case... y 0.. Cg are all zero... (E) Thus the equations are of the form 1^ = 0..... (1) Of the three (0^). 1. c 2 = 0... (2) Suppose that (a^). provided that (fc 1 c 2 ) and (c^) are not both zero..... (c^) are all zero. They are equivalent to ^/(b^) =y/(c l a 2 )=z/(a 1 b 2 ).. say (a l b 2 ) > is not zero. (B ) excluding the case in which a l9 b l9 Cj or a 2 ........ (In fact. (See Art.. 3... (^2)5 ( c i a a)> oppose that at least one............ Then of the pairs (a v 2 ).. If (6^2) =0 and (0^2) = 0. 3... If (6^2) =0. provided that no denominator is zero... then the equations are equivalent to .. .... j3. (C) -(b l c 2 )z + (a l b 2 )x^(b l d 2 ) .. there are definite values of x and y.. Ex. they have solutions other than (0.. (4) If (a... and z may have any value. (D) Corresponding to any value of z... we have ^ = 0... y/(Cia 2 )=z/(a l b 2 )...c^ = 0........ these may be written which are equivalent to (x-a)/(V 2 ) = ( 2 /-^)/(c 1 a 2 ) = (z-y)/(aA) ) . (B).. in whi ch neither quantity is zero.... where k~(a l d 2 )/a 2 y and they are inconsistent unless k = 0...... a 2 x + 6gt/ 4. (3) In particular it follows that the equations a i x + b^j -f cz = 0......... (A) = 0.... then x and y have definite values and z may have any value... then x = 0......... (b ly 6 2 )j (Cj........ are always consistent (i.. ^ + ^ = 0... Ex... If (6 1 c 2 ) = and (c 1 a 2 ) = 0. they are equivalent to x = 0.e.^O......

. equation (D) can be der ived from (C) in this way.. (c^) are all zero. (b^).. u 2 = do not intersect. (2).. 6. In general equations to the line are of the forms described in . If (a^). Eeferring to two planes.... let ^ = 0... For example.. The equations may be written in one Art. 154 INTERSECTION OF TWO PLANES NOTE... where (a. 1. 7. y) is any point on the line The equation to any plane through the line u 1 ==0.. then the line is in the plane.. We therefore say that parallel planes intersect in a line on the plane at infinity.. the resulting equation will also hold. they are equivalent to If (6^) =0 and (0^2) = 0. where A. 6... Equations (A)... If the line meets the plane at (x... and we say that the 1^=0. y==)8... then x=oc. that is to say they are parallel. (4).. /S. c are all zero.. (xyz). (a*6 2 c 2 ). Hence the equation to any plane parallel to the plane ^==0 is of the form u l + k^0... Ex.provided that no denominator is zero. w 2 = be the equations these intersect in a line. If then we make these changes in any equation derived from (A) and (B). The condition that the line whose equations are (x-a)ll = (y-fllm = (z-y)ln . to Art. u 2 Q. (A) may be parallel to the plane is la + mb + nc == 0. the planes ^ = 0. and z may have any value. // also au + bp+cy + d=Q. Parallel Planes. 6. is a constant.. w 2 = is of the form u l + ku 2 t where k is a constant.. 6. (2) Two Planes. Much labour can be saved by the following considerations. then . (1) The equation represents a plane unless a.. (B) are unchanged by the simultaneous cyclic substitutions (Oi6 1 c 1 ). The Equation to a Plane. z) and each fraction in (A) is denoted by r. y. For the sake of complete generality we say that the * equation ' Ox + 0y + 0z + d = where d^O represents an ideal plane called the plane at infinity. If (6^2) =0. All this follows from Art.

If the lines meet at (x. we have -a / ) = 0. In the latter case equations (A) are satisfied by infinite values of x.I'm) =0. and the lines are parallel if lll / = m/m / n/n'. r' are infinite. Ex.) y A 2 ^(a 1 d 2 c 3 ) ) J 8 -( On account of the identities . no value of r exists which satisfies the last equation unless o + bp -f cy + d = 0. CRITERION FOR UNIQUE SOLUTION 155 These equations are satisfied by the same (finite) values of r.fi') (nl' . (nl') 9 (lm') are not all zero : but if these three are all zero. 8.n'l) + (y .m'ri) +(f$. the values of r..a') (mn' . in A. Also let A l ^(d 1 b 2 c. be the cofactors of a l9 b l9 .y') (l m' .(A) Let A==(a 1 b 2 c 3 ) and let A 19 B ly . P~p' m m' y-y' n ri provided that (mri). .. (x-<xf)ll'~r f . 2.where r(la + mb + nc)-faa If /a-fra6 + wc=0. Consider the equations i2/ ~*~ c i z + d-i = . r' if a -a' J I' =0. Three Equations in Three Unknowns. y y z) and (x -a)/J=r. in which case r may have any value. Hence the results foll ow. nr-nV-f (y -/)=0. z and the lines are parallel... The condition thai the lines whose equations are I ~~ m ~~ n ' I' ~ m' ~ n' ' may meet in a point if ( . mr~wV / 4-(j8~^ / )=0. y.

J 3 . If we suppose the coefficients to vary so that J->0.. Hence the equations have the unique solution.. J 3 remains finite... the equations are equivalent to that is to U2 = o. See Chrystal.. Because (A^C^) =A 2 ^0. A^ : B 3 : C 3 . Algebra. ( C) and (1) Suppose that A^O. y..... Ax . so that no one of A 19 B 2> C 3 is zero. The last identities show that if a solution exists ..... - * It is essential to show that suffixes p. u ^ = 0. 360. therefore at least one term of (A^B^C^ is not zero. then at least one of x. This step has been omitted in the text-books. w 3 = 0..... Vol. z tends to infinity. suppose that J 1 ^0. Az= ....CJU 156 THREE PLANES (2) If A = and A v A 2 . L B. A l :B 1 :C lJ A 2 : J5 2 : C 2 . and at least one of A v J 2 .. I... A 3 are not all zero.. the equations are inconsistent. because B 2 ^0 and C 3 ^ 0.. y= -'A^/A. then since UiAi + UiAz + UsA^Av (E) it follows that the equations have no common solution. Neglecting d l9 d 2 .* Because A^^Q.. ( ..A r In the same way. %==0.. all different.. Jy=~J 2 ... z satisfy the equations.we have and similarly. (D) We can show by actual substitution that these values of #. . w 2 = 0. such that none of A$ t J5 a . u l B l -i-u 2 B 2 + u^B 3 ^ Ay + A 2y . we see that in general the following ratios tend to equality : x:y:z.. 3 in comparison with the large terms. d. y.. p. so that verification is unnecessary. q t r exist. it is given by 3= -4/J...A-JA.... z=-AJA . C r is zero. in which every step is reversible.. ^ = 0.. For example. Suppose that A^gC^^O.... the equations are equivalent to each of the sets. Or we may use the following method.

similar results follow. Therefore the only remaining case is that in which every minor of A is zero. and w 2 = 0. (xyz). z are obtained by the cyclic substitutions (a6c). every A is replaced by the corresponding B or C. y. the equations are not independent. // a. (3) If A. (iii) If A L = 0. the equations M 1 = 0. We have 1 1 1 a b c a 2 6 2 c 2 1 1 1 d b c <P b* c 2 x (d-b)(d-c) (a-6)(a-c)' and the values of y. then u 2 A 2 + u B A 3 = Q. w 3 = are the same equation. For we have u^A^ + u 2 A 2 + u^A 3 = (F) (i) If no A is zero. y. we have w 3 sO. A 2 =-Q. ^ 3 7^0. (ii) // A l = 9 A^O. and then every two of the equations are inconsistent . 6. in the above. we say that (he equations have a solution (x. A^Q. c are unequal. In both cases. z is infinite and x:y: z = A l : B l : C^A 2 : 2 : C a = 4 3 : # 3 : C 8 . or they are one and the same. Ex. w 2 = have infinitely many solutions which are solutions of the system. A 3 are all zero. (iv) If. .In the limiting case when A=0. z) such that at least one of x.^Q. since A. the equations are not all independent and there are infinitely many solutions . A l9 J 2 . or else each of the equations (assumed to be distinct) is inconsistent with either of the others. 1. which case is excluded. solve the equations x+y+z~I.

or else they hare a common line of intersection........ w 2 = 0. We have J 1 .... their line of intersection does not meet the plane % = 0. Referring to 8.. FOUR PLANES 157 (iii) // A. Wg=0 be the equations to three planes.... (A) A l . at least one A is not zero. For. 8.. as in Art. two of the planes meet in a li ne which is parallel to the third plane.... which may be at infinity. Four Equations in Three Unknowns..... on account of (B).. The truth of (iii) follows from Art. Hence we conclude that. if J=0... w 3 = 0... ^Ae equations are consistent and have an unique solution except when every D is zero. w 2 = 0. let ^=0. then the planes w 2 = 0. Consider the equations u^a^x + b^j + c x z -f d x =0..9. J 2 > ^s are a M zero > the planes have a common line of inter section or they are all parallel.. Three Planes. J x .... Proof of (ii). (1).. 8... .. (3). = 0... (ii) // A =0 and A l9 A 2 . (B) On account of (A). (1) If the equations are consistent... Let A = (^i^a^) an ^ ^ ^i> BI> ^ e ^he cof actors of a l9 b .... we have u l D 1 + w 2 D 2 + w 3 J5 3 -f ^4^4 = d (2) // A =0. 10..... (1). Observe that all the planes are parallel to the line which is parallel to the line... w 3 = are not parallel and.. J 3 are not all zero. Suppose that J x ^ 0. = 0. Let A l ^ 0. the planes either meet in a point at infinity.... 4. then : (i) If A^O they meet in the point given by the equations in Art. . This follows from the identity .... then A =0.

17.3a&c == 5. 3x . show that the first can be put in the form and find I : m.1. z are not all zero and ax -f. and find I : m : n. z are not all zero. . z bx + ay where x.az ~cx -}~ ay + bz = 0. y. x + y + z6. 3y. and lead to similar results.4z) + m (x + y + z) = n.2z = . (ii) x y + z ~ w . 2x + 3y-4z~ -4.3y . show that the first can be expressed in the form l(2x 4.3y . prove that a 3 -f. Solve by means of determinants : (i) x-y + z~Q. Hence show that the equations are not independent. unless every minor of A is zero.7y . Hence show that the equations are inconsistent. 4. Having given the equations x cy + bz. (3) // every D is zero. This is explained with reference to four planes in Exercise XVII. y. If x.1. The methods already described apply to systems of linear equations connecting more than three variables. the equations are inconsistent. but they are to be regarded as having a common infinite solution.by + cz bx -f cy -f.22 4-3^ 2.== 0.4y . y az + cx.I4z = -24. 2# + 32/-4z= -4. 3. in which case there are infinitely many solutions. With regard to the equations 4x + ly.5z = 7.142 = 10. 158 SIMULTANEOUS EQUATIONS EXERCISE XVII SIMULTANEOUS LINEAR EQUATIONS 1. prove that and x*j(l -a 2 )^i/ 2 /(l -b 2 )^z 2 /(l -c 2 ).6 3 + c 3 H. 2x 4. With regard to the equations 4x 4.

y~cz + ax + du.a 2 .b cos ^4. (^.6y 4- 6~c c a a -b are consistent. A l : A^: A 3 =B l : B z : B 3 . 6 c cos . [Since d=Q. Prove that if c^O.] 10. Having given the equations a = b cos C -f. ax + by + cz=d. |). show that if t* l =0. and that the area of the triangle is i JV^A^- 12. z=ax-\-bydu.c) and a*x + b*y + c*z= d* ~(d -a)(d -b)(d -c)( 8. =6/sin J5 =c/sin 6 f . show that J 2 =0 and J 3 unless b 1 : Cj =6 2 : c 2 =6 3 : c a . [Put each fraction equal to k and eliminate x. and since d l A l +d< t therefore J 2 = d l B l + dpB 2 + d 3 B z =Q unless ^ ^4 2 . u 2 =0. If the three equations of Art.] 11. &. d is equal to . deduce the following : cos 2 A -f-cos 2 jftH-cos 2 (74-2 cos ^4 cos B cos (7 1. XVI. the coordinates of the vertices of the triangle f ormed by them are (4.= 1 . Having given that a? 4. |). 8. w 8 =0 are the e quations to three straight lines. uax + by + cz are consistent and no one of a. prove that a 3 x 4.6 3 ?/ 4. 7. If the equations xby + cz + du. [Ex.] APPLICATION TO GEOMETRY . y.c cos J5. d is equal to .or (b -c)(c -a)(a -b)=abc.y 4.bz __ az -f ex _ bx + ay _ ax 4.6. ( . c. If the equations cy 4.1. 6. ^ 3 are all zero.1. c.6 2 t/ 4. then one of 6. Using the notation of Art. c =a cos jE? -f. w 2 = 0. |). then Also if a = . the equations ^=0. 26c cos A 6 2 -he 2 . z. u^=0 of Art.1. prove that either a + b + c~Q .c*zd 3 -(d-a)(d. 10.4 + a cos C f . a/sin J.159 9.C 2 z = cZ 2 . a z x 4. 8 are equivalent to = X . 8 are distinct and if A =0 and Ji=0.b) (d .

13. 2/4. (x 2 .and if A ^ deduce the solution in the form already given. [The required equation is a (x . [We have u 1 D 1 +u 2 D 2 -\-u 3 D 3 -^-u t D^A 9 therefore the required planes are ~\=0 and 15. Using the notation of Art.ft) + c (z . show that if ^=0. (# 4 . Using the notation of Art. ~ i n i y\ "i * ^* 9 2/3 2 3 1 r 4 2/4 Z 4 1 14..(x^ 9 t/ s . 10.] 16. 10. w 4 =0 are the equations to four planes. etc. w 4 =0. It is required to find the equation to a plane which passes through either of these and is parallel to the other. is one-sixth of the determinant in Ex. t* 8 =0.] 160 EQUATIONS OTHER THAN LINEAR . the vertices of the tetrahedron bounded by them / 'A B C \ /A B C \ are the points f^r> y^ jr) > (jr* ff > 77 ) > e ^ c > an( ^ * na ^ tn volume of the tetrahedron is **-A z ID D D D [It is shown in works on Coordinate Qeom&try that the volume of the tetrahedron whose vertices are (x l9 y l9 Zj). % 8 =0.y ) = where al + bm + en = and aV + bm' + en' =0. 1*2=0 and w a =0.a) + b (y . y 29 z z ) 9 . 13. 2 4 ) may be in the same plane is . consider the lines whose equations are U!=Q. Consider the lines whose equations are S-ql/-j8S.X-aS m n m Show that the equation to the plane which contains the first line and is paralle l to the second is Deduce the condition that the lines may meet in a point or be parallel. Show that the condition that the points (x l9 y l9 2 X ).

17. Ex. 18. they have a common point at infinity. 1. Given the equations a 2 c 3 + a 3 c 2 . that is to say.. and X. (/?) If every minor of A is zero.. if A^O and we consider the planes u l9 u 2 . prove that X= P N where L lf Jf . 19. w 3 0. Y are given by N l X-N t Y=L 1 -M t . say D 19 is not zero. Here we illustrate various methods of dealing with systems of special types. . considering the planes. (1) Systems which are symmetrical with regard to x 9 y.2bJ ) 2 = 0.. are the cofactors of l lt m lt etc. 9 u^... y are given by = m v -f ^ 2 . a t c 2 + a^ . in the determinant 20. 11. then (i) If at least one Z>. and suppose that J = 0. Using the notation of Art. then (a) If all the fninors of A are not zero. For example. Hence all the planes are parallel to the li ne x\ (6^2) = yj (Cjda) . M 2 =0. u l9 u. w 4 be the equations to four planes. since /) 3 =0 the plane w 4 contains or is parallel to the line u l ~0. let t^ 0. Also.zf (a^). prove that the corresponding values of #.. Find the rational solutions of . the four planes are parallel to a cert ain line. If a. i/ 3 . the planes meet at the point (ii) Suppose that every D is zero.. two of these. The question of elimination is considered in detail in another volume. say u l9 u 29 meet in a line which is parallel to the third. the planes have a common line of intersection or they are all parallel. w 2 0.26^3 = 0. prove that [Prove that gl = ~ 2&1 = Cl ^ 2 (i c i-^i 2 ) -. 1' by the equations of Ex. Since D 4 =0.26 2 6 3 = 0. ]0. y are connected with J. etc. a 3 c l -f iC 3 . Systems of Equations of any Degree.

y 3 . Let x l/X.60 and x. z = l/Z. If x=p. Ex. If p =2. Hence if (x. 2. r ~ . (2) A solution may sometimes be discovered by considering the properties of a particular determinant. z(z-e)=xy. Ex. Consider the determinant x z z y y x The y 3xyz .z 2 . . b. ir'+^+z 3 .32^2 = 146. (y+ 2 )(z +*)(*+#) = 14. in the top line. y. ca -b 2 =Ay f ab -c 2 ~Az. Solve x(x-a)=yz. c are all zero. y 2 zx = b.I50p -f 292 =0. 4.5 taken in any order. and there is no o ther rational solution. be -a 2 Ax. Syz~q t x\jzr y the given equations are equivalent to Z> 2 -2g = 50. Thus x.23. z 2 .x". x z cof actors of x 9 y. b. pq-r = U. Therefore the only possible solutions are given by a 2 -be ~~b 2 -ca "c 2 -ab ~~ *J (a 3 + 6 s -f c 3 .x 3 .1 ^147. 3abc . zx ~ y 2 .x *+ y * +558=50.6 s .yz = a. y 1/7. (zx~y 2 )(xy-z 2 )-(yz-x z ) 2 ^Ax.20 2 -230 + 60=0.z 3 . (3) A change of the variables. z) is a solution. 3.A 2 . c. and similarly. SPECIAL TYPES OP EQUATIONS 161 whence we find that p 3 . then q ~ . giving p 2 or . /. are yz . z are the roots of 0*. Multiplying these by a.xy c. z may have the values 3. then the equations are equivalent to X 2 -YZ Y*-2 .a 3 . z.c 8 = A (ax + by+cz) .3a6c) * It is easily verified by substitution that these are solutions unless a 3 in which there is no solution unless a. y. p 3 -3^-146. y(y-b)=zx. and adding. Solve x 2 . xy . y.

cm'-c / m = and y .A EC ... c'y 2 m'y. 0). x 2 (a'+b't + c'P) -x(l' + m't). In this one value of < is infinite. 0=c 2 -ab. as in Ex. In equations involving x. and if y = tx the equations become x *( a + bt + c2) -. 4.a o c Therefore. the solutions are given by x (l + mt)/(a + bt + d 2 ). 2 = the equations become . y the substitution y = tx is sometimes useful. can be guessed. (5) 7/a solution x = x 1 . 162 SOLUTION BY SUBSTITUTION (4) In the case of two variables x. giving k (a 8 + 6 3 -f c 8 . /. If x~Q and y^O. . z. we may write y = tz. then a suitable substitu tion is "V "V x ~~~ Ju\ i -A. y. 2. either k =0.) y ~~ y\ i * j *&r. Hence if x^Q. /S'o/ve One solution is (6 -t-c.m/c. c -fa. giving the solution (0. with similar values for y. or k (BC ~A 2 )= a ABC. where (a + bt + ct 2 ) (V -f m't) = (' + 6'e + c'J 2 ) (Z + mi).. x {i + m j). Substituting i n the first equation k_(k__ a \ &_ A\A therefore. B=b 2 -ca. the equations become cy* = my.y = y l . a + 6). Solve ax* + bxy -f cy 2 = lx + my. where A a 2 -be. z t'x. and k is to be determined. y tx. 0. 0). y=c+o-l-7. z. and if we write ar = 6+c + JT. One solution is (0. 5. a'x 2 + b f xy + c'y a = I'x -f m'y. Ex.3a6c) = .

From the second and third of the given equations. ^ may have any values such that 7Z + a(7+Z)=0.. 7=0 and Z^O. 2_1_1 1 ' X~a b c' with similar equations... F. and solving for X z .. z*+2xy=c. c=Q and 7. we obtain the values stated in the question. (A) are given by where -X" 2 =2a-6~c2v/(a 2 + 6 2 -fc 2 -bc-ca-ab) and k=>J(a + b+c)... X. x + y + z-k.. If JT=0.3 =6+c-a... /. then a=0. By addition. we have 1 |__1 I !___! 1 L_^I F + Z"" a 9 Z + X~ I 9 X+Y~~ c .. then 6=0. SIMULTANEOUS QUADRATICS 163 Adding the second and third of equations (A) and subtracting the first. Z^Q.. where X~k~Zx.. by subtraction. i# zero.. Ex. (+i/ + z) 2 =a + 6+c. beside the obvious one. 6.. Special cases.. y* + 2zx=b..... The eight solutions of x* + 2yz=a... This gi ^es the only solution. 6=0 and Z may have any value. If JC=0.0.. we easil y find that A/ ^V . Another solution. 7^.Hence if none of the three.. . Equations (A) are equivalent to . .. Z..

.. Therefore the eight solutions are given by x + y + z=JA......... (ii) If =0.. Thus... y^O. then abc + 2fgh-af*-bg*-ch*=Q ..... and/=0.. *JC.. x+w z y+a)Z=*/B. Show that. by* = 2fyz. y=0 and 2^0. 7.+.. then C2 a =0... Ex.JC 9 ....... From equations (B).... %c and ( ^^ ) These results agree with the preceding if we change the sign of X. which is perm issible. x+a)y+a) z z=...... 0. where X=*JB+>JC. 164 COMMON ROOT OF TWO QUADRATICS . and therefore Also.... Substituting in (C) and dividing by x*y*z* (assumed not to be zero)... Special cases. ax *(b*y* + c 2 z*) + ..... the equations reduce to the single equation c 8 =0.. the result (B) follows. 3a?= N /^+Z. (B) From equations (A).. and thus do not constitute a system of equations. Considering the original equations it will be seen that (i) If z=0... with two similar equations. (x+uy +o> 2 z) 2 =a + oj 2 6 +o>c = (7.. and c=0. hence the equations become.+2abcx z y*z*=Sfghx*y*z* ... if the equations (A) form a system of equations with a solution other than (0. by multiplication.. then the condition (B) holds. if the equations (A) are consistent. where o> is an imaginary cube root of 1. (B) where *JA is either square root of A and so for *JB.. from the second and third of equations (B).. 3^0.-A (say). (C) Also 6V + c 2 z 4 = (4/ 2 .26c) y*z 2 .. and we have 6=0..... &y*=0.. 0)...... (x + w 2 y + ojz) 2 ~ a + o>6 + co 2 c = B... by addition..

. Zax(ax-by-cz)^X 2 (b-c)(c-a)(a-b) .~. we may write y(a-b)+z(c -a)k. y. 0. (a'. x~ . z(b -c) +x(a-b)=k. z. Hence. x y z xyz (b-c) 2 (c.cz whence also A ~ -a(b~c) 2 +b(c-a) 2 +c(a-b)* (b+c)(c-a)(a~b)' :. z. b. ft).2cazx .z). w o find that xy{y(a-b)+z(c-a)}=xy{z(b-c)+x(a-b)}.2abxy ' The given equations can be written x( -ax + by + cz) -\-y(ax -by + cz) -f z(ax + by -cz)=Q.Ex. x( k(b -c) Hence. 2a(b 2 -c 2 ) -A 2 (6-c)2(c-a) 2 (a-6 and the result in question follows. and two similar equations can be obtained. c are equal.2xy =. (ax + by .2zx .2yz . Prove that. (c a) (a 0) . then. assuming that none of the thre e. If (a.2bcyz . The consideration of special cases is left to the reader. 8. EXERCISE XVIII 1. j8') are the roots of . if no two of a..ax -by .j~ .a) 2 (a~b ) 2 " a^x 2 + b 2 y 2 -f c 2 z 2 . is zero. if no two of a.cz) respectively and subtracting. c are equal. if and x 2 + y 2 + z 2 . Multiplying these by (x + y . b.

X 2 y + xy 2 + x + y^3. y(I-x*)=:2x. a.u prove that a^^^-^^ and consequently the necessary and sufficient condition that the equations have a common root is E=0. y x . (a'. * 3 + 3^ = 8. Eliminate x. x 2 + y 2 + z 2 ^ 29.aa'bb'z* + (b z a'c f + b /2 ac . (x + * 2 y 9 25 8 ' + " 8 ' 9. 2.l) 5. SYSTEMS OF EQUATIONS 165 Solve the equations in Exx. 6. x + y + z = $. /a x . where (a. a//T. /?</. y from the equations and show that the roots of the resulting equation in z are aa'. 3-28. (y + z)(z + x)(x + y)=*.b'c). xy+(x + y)=2. jS/T . a/3'. 0). Show that this equation is a 2 a' z z* . ^--^Q-xy.(ab' .2ocaV) z 2 .. j3') are the roots of ax 2 + bx + c-0.bcb'c'z + c 2 c /2 = 0. 2(x .a f c) 2 . [Put 4. a'x 2 + b'x + c'Q. 10.1 7. 3. where R = (ac'. Prove also that the equation whose roots are a/a'.l)(y . //T is obtained from the last equation by interchanging a' and c'.a'b) (be' . x(l-y*)=2y. -=l6-xy.

X4-7/-1 21. _ B a y a a? a z --y~ a z -x. y i l^ f ----. [( + y)(a? + 2)=o a . (l~rr) 2 + (l-2/) 2 -(l-o:)(l-2/)(l-a: ? y). 23. 20. . .] 16. (y ~-z)(y -x)=by t (z-x)(z-y)=cz. xy(c-z)cz 2 . yz + zx {-xy= a 2 .-.z-f re. 112 112 ' ' 27.x 2 = 6 s . y b z c x a [Put x= 18. and prove that A 2 -X(a-b) + a-b and 14.1 --. 1_ 1 1 1 11 1 1 X d V . [Put a.xy) ~abc(x + y .2 2 . 22..c). 15. x . y2 /. ayz + by + cz= bzx + cz -f ax ~ 19.2 + 0. (x-i/)(2:-2)=aa.a 13. 2: 2 4-a.l-A. -. x+y=x' + y'=r. 25.. y l-p.O .7/ 2 c 2 .a> 4.12 ^ 4. 26. _ = -_ = . xy(ay+bx . V . i/ 2 + 2 2 ==a + 2/"f z. 2 6-i. <2 C ~~ . .--.= -. zx(b-y)=by 2 . x(bc-xy)=y(xy~ac). yz(a-x)=ax 2 . etc. 24.

If a + 6-fc=0 and x + y + z = Q. 2 2 z 2 + (z-fir) 2 + A2^ = l4-A and x^y.bzx + cxy = and x 3 + y 3 -f z 3 -f Xxyz = 0. (ii) 3a 8 -3a 2 + 3a. If yV + (y + z) z + Ayz . If .--2) and ax + by + cz-Q._ _ __ i __ _ x y-\~z a 9 y z + x b 9 z x-\-y c 9 prove that x(b + c a)=y(c-\-a-b)z(a + b c): and hence solve the equations. 33. 34. z satisfy the equations 111111111 __ i __ __ _.1 =0. prove that (i) xy = a*(a-l)l(2a-l). CL C 32. If -}-^ = ~ + =a + b. y=bY. If ax* + by* -f cz 2 = 0. and a:. 30. zndshov? (X+Y + I)(aX + bY -a-b)=Q. __ i __ _. Eliminate #. y are unequal.Aa:y = 1 -f A.(x + i/) 2 -f. y. a o x y a b [Puta^aJf. If x. 166 ELIMINATIONS 29. then will a: 2 !/ 2 4.1 + A. (iv) 3a(z 3 4-y 3 ) = 2-a. prove that .2 8 equations. If x 3 -y*~ y*-x*=a* and aM-y= -a. and solve the 5a? 3 5t/ . show that either -4-^4-1=0 or x + y=a + b.x y' y y x' x 28. y. p r0 ve that y+\=x.] 31. z from - --. 4 + t/ 4 4. (iii) 3a 2 (s 2 4-t/ 2 )= ~ 1. and (v) 3a(a. \^e + 1 + *Jy~-2 = *Jx -1 + N/jMh2 = 3. 35. prove that 36.l) = 2+4a 2 . ayz -f.

y. (y + z-x)(z+x-y)=.[Use the identities (6y + cz . etc. ?/=&. -8y-f52= 3. Eliminate a? and y from 41. 3/ 7V 2 .2bcyz . + 2y 4-32=3. 44. a: 2 + t/ a -2 2 ==6 2 . x 8 + y 8 -2 3 =c 3 .on/ f (a.ayz ~y 2 . z=c(pa 3 -gfc 8 ) is another solution. Eliminate x. 40. 3o. i/ = &(rc 3 -pa 3 ).2cazx .ax) 2 = (bz + cy) 2 .b 2 y 2 + c z z 2 . Show that the equations are consistent. z from x + y-z=a. prove that a 2 -f & 2 4-c 2 = 2a&c + l and z a /(l -a 2 )^2/ 2 /(l -6 2 )=z 2 /( l -c 2 ). Given that and a 2 o: 2 -f. 43.] 37.2a6a:y = 0. 2 + 1/ 2 + z 2 ). z from ( (x + y-z)(y+z~x)~bzx.cxy. y. + y4-2 = 3. 2 4-t/ 2 -f2 2 = l. showing that the result is abc~(a + b + c-4) 2 . and find x : y : z. Sh W 38. then xa(qb* ~rc 3 ). z are not all zero and x 2 . Eliminate x. 39.bzx = z 2 . z=c is a solution of px* + #y 3 -f rz 3 = &M/Z. xyzd*. CONSISTENT EQUATIONS 167 42. prove that a. y. If x. Show that if # a. If a.

4k 2 + 2) a:^.. v .+ -=0.. where a. y. y. prove that x 1 2 + x s 2 ~(k*. 49..c'z -f rf^ =0. (i) Eliminate x. N=(cd').(x + y-z) and J-fw + n=0. z 9 1 in Ex.-f. 6. Eliminate x.. n = (ab'). (i). . (c-a. . t from the equations ax + by + cz + eft = 0. -I. (iii). 48.-f.. (LMN). z\ t 2 for x. j8. (a -x)(b -y)=c 2 .+ . . y. u 2 + are equal. etc.] . a & c d .45. are all different and 2 2 2 2 _ . a x a. (Imn). a' 6' c x d' . prove that each is equal to %(x z + y 2 + z 2 + u 2 ). XVI. If -(y + z-x) *. c. .. d make with a fixed line. If the four expressions xz+zu).6'y -I. 25. x y z t x y z t showing that the result is L M N + Lmn + Mnl + Nlm = 0. .. y from (fc~#)(c-t/)=a 2 . I L [(i) Prove that .+ . 25. LV } (x*-t 2 )yz (y 2 -z*)xt the unwritten fractions being obtained by the cyclic substitutions (xyz). where l-(bc f ) 9 m = (ca / ).-f .(z + x-y) = . prove that the four given equations hold. y = cos -f t sin j5. M=^(bd'). 8 are the angles which the sides a. Now put x\ t/ 2 . z. L = (ad'). show that 46... Xf. 47.. If x lt x 2 .)(a-i/)=& 2 . By projecting along and perpendicular to this line. (ii) Let # = cos a-f-tsina. .T =0. (ii) Use this to prove Exercise XVI. .

J =/() \3// Let n be odd and equal to 2m + 1. .. +x x + X hence $(x) = is a reciprocal equation of the first type of degree 2m. then I/a is also a root..... Let f(x) = a x n + ajX**" 1 -f . .. is any root.. We shall say that Reciprocal Equations are of the first or second types according as k +1 or k= -1. x n f ( .. // this is of the first type. .. +a m x m (x + 1) = 0. be the roots of a x n + a^".. Hence (A) is a reciprocal equation if and only if Denoting each of these fractions by k. are the roots of a n x n + art^x*" 1 + . and if </>(x) is the quotient... .. we have a n a so that &= 1.. l/)3. Let a. TAe solution of any reciprocal equation depends on that of a reciprocal equation of the first type and of even degree.. Therefore/(z) is divisible by x -h 1. Theorem 1 . .. .. + a n _yX + a n = 0.1 4.. Reciprocal Equations.. . n... TYPES OF RECIPROCAL EQUATIONS 169 .. .. -f a n = be a reciprocal equation. + a x x + a = 0.. /or r = 0. (A) then I/a... /J. Then the equation may be written f(x) = a (z 2m + 1 + 1) 4-a 1 x(a. 2m ~ 1 + 1) + .. /&aZ i5 according as a r = a n _ r or a r = ~a n _ r .CHAPTER XI RECIPROCAL AND BINOMIAL EQUATIONS 1. 2. A reciprocal equation is an equation which possesses the following property : If a. . ... 1.

1) + . and therefore a m = Q..r ) + ( x*~* -f -i-A x/\ x 7 ^. this may be written 1 \ / _ . x/ Also x + l } (x'~ l + -^ ) = f ^ + . if <(x) is the quotient.1 4. if n = 2m. we can show that /(x) is divisible by x . Therefore /(x) is divisible by x 2 -!. 1 \ ~ . / ~ Thus in every case the solution of /(x) = depends on that of a reciprocal equation </>(x) = of the first type and of even degree. Grouping the terms as before.=t/ r . A . and if <^(x) is the quotient. Dividing by x m . . with the same result as before.1) + a 1 x(x 2 . then x x r 1\2 + -) -2 = 2 -2.1 (x 2 .2 / . Let the equation be a x 2m 4..1.2 .Iff(x) = is of the second type. x n /( . The solution of a reciprocal equation of the first type and of degree 2m depends on that of an equation of degree m. +a m _ 1 tf m . and that. Again.) = -/(#) x/ Let n = 2w + 1.. Theorem 2. so that u*=z . We say that a reciprocal equation is of the standard form when it is of the first type and of even degree. 4 ax + a = 0. Hence the equation may be written /(x) =a (x 2 .1 / \ xv V x r ".fljX 2 . since a r ~ -a 2 m-r> ^ follows that a m = ~a m . .(A) Let x f -^=z and x r + .1) = 0. . then x 2m < (-} = <f>(x).

1 =0. Thus the roots are -1. 2 = 2" or ~^sr an d x i 8 gi y ^ n hy x-\ =|. pm ~ qn = 1. i^Aere r is any integer. (3) If n is a prime number and a is any imaginary root of x n . 2. For every one of this set is a root and no two of them are equal. The Binomial Equation x n -1=0. = 2.1 1* 4 . and the equations x n . For suppose that the two a a and of (a>b) are equal. (2) // m is prime to n. Dividing by x + 1.1.".33z 2 + 1 1* + 6 = 0. the corresponding values of xare given by a? 2 . |.38x 2 + 5x + 6 =0. and so on. 1 . Then a a ~ fe = l.33** . in succession. |. then a pm = l and a fln = l. j9 and 5 can be found so that pm-qn= 1.therefore w r = 2w Putting r = 3. Therefore (x. 170 THE BINOMIAL EQUATION The reader may verify that 14z 3 .16z 2 + 2. that is x . and if z 1 is one of its roots.. 2.or - Whence we find a. Solve 6s 5 4. #*. w 8 = z 8 . so aZso is a r . (1) If a is a root of x n . . (See Ch. For if a is a common root.z^x + 1=0. a* 1 "" 1 . a 2 .1 =0 and o^~ b = l have a common imaginary root. Dividing by # 2 and grouping the terms. where p. and since m is prime to n. $e n roots are 1. 4. we have 6x 4 + 5z 3 . This is . 4.) Hence a = l. -3. (3).1 =0. . I. One root is .8z 6 + 20s 4 . a. the equations m -l=0 and x n -l=0 have no common root except 1.. -3 or ~J. this becomes 6(z 2 -2)+5z-38=0.7z. q are any positive integers.. Thus equation (A) can be expressed as an equation of degree m in z. + -. If 2 = 3..

for instance. the values of z are x o v 10 2 cos where r = l. jS o/a?-l=0.1 is divided by x . then <f>(x)Q is a reciprocal equation. .. . similar reasoning applying in all cases... /x i -i cos -. and if this is transformed by the substitution z = x + -. For (a a ) n = (a*) a * r = l. 2.or -. rt 71-1 Mcos . ..r. if aj8 6 y c =a '^9 &/ y c/ > ^ft-fty-c'^a'-a^ w hich is impossible. SPECIAL ROOTS 171 If any two terms are equal. q. thus n is prime to a . n n The imaginary roots are therefore 2r7T .1 or x* .. r . . ete. Take the case of three factors p 9 q. (4) If n=pqr . 1. where r = 0.1 =0 are . V. w-1. and since p is prime to qr. these equati ons have no common root except 1. is a root.impossible . 2. 2. (5) As explained in Ch.. . the roots of x n .. and similarly (/J 6 ) n =* 1 and (y 5 )* 1 = 1. for instance a /?}/ 5 .1 according as n is odd or even and <f>(x) is the quotient.6.. for n is a prime and l<a .4 sin .1 = are the n terms of the product where a is a root o/x p -l=0. or according as n is odd or even.1 =0 . 2riT .. for jg*-*y-' is a root of z r -l=0 and oc a '~ a is a root of x p .. where r = l. wAere p.. 17. the roots of x n .. . . yo/x r -l=0. If x n ..6<w. are primes or powers of primes. .. Any term of the product.htsin .

and therefore 2f7T .. a.h t sin -. where r<n and r n n prime to n. etc.n & & 3. I 1 or. the n roots are 1.1 =0.C. so that a is a root of x~ b = l. Definition. If r<w and prime to n then n -. The special roots of # w ./ 2r7r . 2rm7T . and since n is prime to r. fa. . 2rrr . Any root of x n . This is impossible.= 1 .hi sin .1 =0 which is not a root of an equation of the same type and of lower degree is called a special root of the equation. if cos -.1 = 0. For every one of the set is a root. m must be divisible by n. \ n n / . 2rir . TH . fj8. a 71 "" 1 .1 =0. 2r7T cos db i sin n n are special roots. then cos ---. This is impossible. a 2 . 2rm?r . n n Therefore rm is divisible by n. These can be arranged in pairs. then cos -. ^ M B. -j.h i sm ---.r is prime to n.. And if two of them are equal. for m<n. where <f>(n) is the number of numbers less than n and prime to n..is a root of v ' r n n x n . 2rn\ m . and of the same type. including unity.1=0 are cos -. (1) If r is less than n and prime to n. . x n -l=0 has <f>(n) special roots. where m<n. 172 SOLUTION OF BINOMIAL EQUATION (2) If QL is a special root of x n . ^J. for the last equation is of lower degree than x n . so that they are the roots of a reciprocal equation. . if a a =a b > then a a ~ b = l... but i is not a root of any equation of lower degree. for instance. Special Roots of x n -1 -0.h i sm . Thus.) = 1.A.

2. a 77 in (J)..j n n \ n n/ \ n o. Deduce the values of co# 30 and cos 72.-f i sm . .I =0.. (1) Let A =a(cos a + i sin a).j cos -. And since a n 1. n / it follows that the n values of ^A may he found by multiplying any value of by the n-th roots of unity.(3) //a is any special root of x n . taken in some order or other (Oh. For if a is any number less than n arid prime to it.. . 2. ( a + 2r7i . . Ex. n This is obtained by putting a^=l.).. where a. 6. c. 4.. a (n ~ 1)a .= cos . are the numbers less than n and prime t o it. 2.4. a 3a .. Dividing by x .. a 3 . . (n-l)a are divided by n are the numbers 1. a . n 1. ..i sin . I w n J For by De Moivre's theorem. (2) The roots of x n + 1 =0 are . . where r = 0.--------.. a 2 .. where a = | 4 |.. \ cos -ft sin J-. 1. Therefore a a is a special root. where r = ().. occurs somewh ere in the set a.3. Solve x 5 . r?r \ bmce cos ---. w-1. 1... . this expression is a root for every r. 3a . . .-f i sm . 1. _ ^ _ t are ^/a .h t sin . a c .i-ivsm . 2a.^4 = .. the complete set of special roots is a a . including unity. . I. a\ .-----. . it follows that every number in the set oc a . cos . n ~ 1. the remainders when a. The real root is 1. A .1 ~0. a 6 . . a n-1 . then the roots of x n . a -h 2r7rl . . and its n values are all different. ... The Equation x n -A^O. 10).1. a 2a .

thatis..=cos 72 = ^.x 1 + ar 1 . 15-2. z* .= .^0 V 5 "-! . 15-7.1 is (x 2 4.M. of x 3 . 15-4.1 and x* .x + 1 ) (x 5 .a: Let 2=3 + . 2?r . and cos -. are cos . giving 2-|(-ls/5). The L.C. Wh at are its roots? The numbers < 15 and prime to 15 are 1.then 2 2 -fz-1^0.x*+x*+x* + z + 1 ~0. This can be .( r = 1 . Now cos . 15-1. 15 5. # 2 -f . 7). 7.. 4. Find the equation of which thes e are the roots. 2f7T of which the roots are 2 cos -r-^.. The values of a.1 -0.1 and x b . or.+ai-f -.t sin . 2 ) . cos . Show that this can be reduced to an equation of the fourth degree. 2. and 2 cos . 2. 7). The Equation x 17 . + 1 =0.. 4.(r = l.4z 2 + 4z + 1 =0. If ZX + -. The special roots are cos -~ i sin (r = 1. 2.1 =0.+ 1 =0..cos -.a.x* +# 3 .z 3 . 15 ID The non -special roots are the roots of z 3 . 5 5 55 TT 0JO s/5 + l . 4. o 3 Therefore the values of z are 2 cos -.. What are the special roots of a.'.=coa 36 ^-. this becomes that is. 54 54 SOLUTION BY QUADRATICS 173 Ex. 15 -1=0. Gauss discovered that the solution of this equation depends on that of four quadratic equations.1) and Therefore the equation whose roots are the special roots is x 8 . that is. 2.

................. (H) 4a 2 =l+2 j 8 + y . ay -i to* 2/1^8+ 2/42/2 +w^^ therefore ay = ........ (J) Then by (E) A+^= ~i Also 4A/x...y8 4... which has a remarkable cyclic property...... ... then we have yn-r = y r and 2y r y s ....... (F) Consequently j88 = -^ .. will lead to a like result..... we find that .y f+J 4.. ^ rk o o cos -=y L sm -y=-..... 8.. 2f77 ...... Let then 174 EXPRESSION IN SURD FORM It follows that any relation connecting a..... y........ 2. y g .. Such relations are the following.. 3.........1...proved as follows....-^ ..y r ..... ................... using equations (A).......1 =0 are 2f7T .................... s ... Starting with the pair y^y^ we have On trial........... Hence by (H) and three equations derived from this by the cyclic substitution (a/?yS). ...... therefore Equations (A) lead to an arrangement in pairs of y l9 y 2 ........ j8...... Let y r = cos --y .............. taken with t/j..aS.. it will appear that no y except y 4 .. (E) Again...... + S = 2/x ......... (I) Finally let a + y = 2A. = <xj8 -f jSy 4.. By (B) a + j3 + y + 8= -i ...... The imaginary roots of x 17 ....... (G) Similarly we can show that 16aj8=-l+4a-4j9......... 8 continues to hold after the cyclic substitution (a/3yS).. where r = 1... (A) Also the sum of the imaginary roots is .............

2ocy + j3 = 2/3. etc.. the values of y l9 y 2 . f . cos -.= { .2/5 y 2 ~2fy + y -0 2/2>2/8 2/ 2 ~2yy + 8 -0 2/6. Ex. g^ (34 +2^/1 7).Thus y l9 y^ are the roots of y 2 . . Also and by (I). the greater root of each quadratic is that which is stated first.2/7 y 2 (K) ^/* 2/ 2 + 82/-iV = Since cos decreases as 6 increases from to 77. Express cos as a surd. REGULAR 17-GON 175 therefor. Thus. and = ~ (34 -2^17).1 J^{17 +3^17 . starting with the last equation and working upwards. y s can be found.1 =0 are given by a -2^3 + 1=0. and then the roots of x 17 . .^(34 -2^17) -2^(34 + 2^17)}. 1..

to effect the following construction.. Draw x'Ox. B'. y'Oy at right angles. .. the surd values of the sines of these angles involve one more radical sign.. -A) ( i. To inscribe a regular 17-sided Polygon in a Circle. 22.can be expressed in a similar form. A' and JB. (0. 1. .9 . t/j). 17. .NOTE. we have to draw the circles in the following tabl e. y s of the points 1. 8 are given by the equations of Art.. 5. (0. by the use of ruler and compasses only. Circle Centre Ends of chord y Ends of diamotor x Name of circle * 2 + y a + iy-A^O (0. (i) The cosines of . ~. observe that the roots of y 2 . -^=. (0. 6. 2... Let be the centre.2ay + 6 = are the ordinates of the points where the axis x = cuts the circle x 2 + y 2 .) (ii) Any length determined by a quadratic equation can be found geometrically. y 5 in Fig. It is therefore possible. M ) 0) . XIX . A). 0) (0. y 2 ). 27).. the point coinciding with B. cutting the circle in A. (Ex.. Let the vertices of a regular inscribed 17-gon be marked 0. j/ 8 ) (two of which are marked t/ 3 ... Take the radius of the circle as unit of length. 2. To solve these geometrically. The ordinates y l9 y 2 .2ay + 6 = Thus to find the points (0. .

y t ).) )8 . a).0) (0.) a a. (0.0) (0. S) M x* + y*-2oty+ 0=0 (0. A) (t.j8) (^-y. y) A x* + y 2 -2fjiy-&=() (0. (0. 2 + t/ 2 -2$/+ y = (0.). (0./i) (t.<x) (</-/J. (0.O) (0. y. y. y.x* + y*-2Xy-& = (0.0) (0. |8).

y. 27. .8) (V. S are equal to the tangents from B to the circles j8.r) (^8.0) (0. respectively.) 8 The centres of the circles are marked o>. /z. in Fig. A. J58. For. .a .). . we can draw the circles j8 und y. (0.y=0. . 176 REGULAR 17-GON (ii) The diameters of the circles a. j9.) y x.f$. since the equation to the circle on J5j8 as diameter is _ 2 / ~ 2 which is satisfied by x= *J .x* + y 2 -2yy+ 8=0 (0.+ y*. J5a as diameters. A. y.. The circles co.2oy+ =0 (0. 8. a. giving the points a. etc. To draw the last four circles notice that : (i) The circles a.0) (0. y. y. y. &). S meet y=0 at the same points as the circles on J5jS. /3. respectively. a. By. j3. (0. 8. p are easily drawn.

[Write this x* + x-* + (x + x~ 1 + 2)*~2(x + x~ l + 1) 4 . 4 . 2 *j Hence show that . These lines meet the given circle in the vertices 3. 4. 5. + 2=0. Hence we can draw the circles a. 3x 5 . 1-9 : 1.2/2 -f y. y. 14. + 2=0. y. 2x* + x* -I7x* + x + 2^0. x* + Qx*-5x* + 6x + l=Q.10* 4 . 2# 4 + 3z 3 + 5* 2 + 3a. 7. 2z 7 -a. Through y 3 . x'-Sx* + Ilx*-Sx+ 1=0. 12 of the required 17-gon.10* + 3=0.5 n-. Or we may draw the circles a.3* 3 . RECIPROCAL AND BINOMIAL EQUATIONS 177 EXERCISE XIX RECIPROCAL AND BINOMIAL EQUATIONS Solve the equations in Exx. 3.n and z=x + x~ l . 2x* + 5x*-5x-2=0. prove tha t u n+l -U n j *n=-T-9-. 2. and the square of tangent from JB to Q j8 = 1 .+u n where u n x n +x. 8. With centre jS and radius $6 draw a circle cutting yOy' in t/ 3 .3^-0. and show that this re duces to 10.For (diameter of Q a) 2 = 4 (a 2 . p. Along Ox take OJ^^OA. With centre ft and radius /i7 draw a circle cutting yOy' in jS.. 14 and 5.4j8.j3) = (2jS + y + 1) . j/ 5 .-32. Draw the circle on By as diameter cutting Ox in G. + 2=0.and *n= w n-i. y 5 draw parallels to x'Ox.3s a . 8 as explained above. With centre A and radius A J draw a circle cutting yOy f in a. which can be completed by bisecting the arc 35 at 4 and setting off arcs equal to 34. 9. * 8 -f l+(s + l) 8 =2(* 2 + :r + I) 4 .. 2z 6 -z 5 -2z 3 ~a. 8. 8. If s n ~l+u 1 + u 2 + . 6. Along Oy r take With centre co and radius oo7 draw a circle cutting yOy f in A. Construction.

2 + a 5 .1 =0. \ b by [To find the equation.i sin .C. Show that there is a value of k for which x* .( cos . where r ~ 1.1. a s + a 4 is *.] A C JL C 11.is a ^ y adz* + (ac + bd)z* + (ab + bc + cd.(a-fa.M. of x* . that is.kx -f 1. the special roots of x 2n -1^0 are the roots of x n -f 1 =0 . 13. x* 1. and write (x* + l)fx=z. For the equation x*+px* + qx* + j34-l=0.d) 2 = 0. 4.1 ) + l} = {(aa. a.] 16.c) 2 + (6 . of x 2 . /?. If a. prove that the equatio n 1 * l l whose roots are a-f-. i. g-(N/3 t).1. divide a. 5 + -. and their values are d. the equation whose roots are a -fa 6 . If n is a prime. 2. The special roots of x 12 -1=0 are the roots of # 4 -a. [Show that x*-15x*-Sx* + 2=0 has two roots a. and their 2T7T 2f7T values are cos i sin ~ . y are the roots of o# 3 -f bx 2 + cx + d=0. x 9 . 12.=3 8 + z a -22J .1 by the L.n = 15 the equation on which the solution of x* n + l -10 depends. and find this value. p such that a/? = l.M. 14. 12 .C. The special roots of x 9 -1=0 are the roots of x* + x* + 1~0. Then 178 CONSTRUCTION FOR REGULAR PENTAGON 17. prove that . If a is an imaginary root of # 7 . 3 + bx* + cx + d) (dx* + ex 2 + bx + a)}.: + l and (1 -ar)(l -ar')=2-z. by the L.) . 6 .Bad) z + (a .L] 15. [Prove that adIJ{x 2 -a. 1 x n 1 x~ n [1 +s n = .1 =0.e. a -f 1 =0.1. y-f . # 3 .1.I5x 3 -&r 2 + 2 is divisible by x* 4.

(ii) the roots are all imaginary if. the only real root is 1. show that o and deduce the construction indicated in the figure for inscribing a regular pentagon in a circle. 20. the roots of are the values of tan 2fTT (4r+l)7r 21. (iii) two roots are real and two imaginary if (fg-f I)*<p 2 . the given equation has three real and two imaginary roots. Let z l9 z 2 be the values of z. and so p*>4(q-2).l)*=a(x* .] 19. The conditions in question can now be obtained by considering the equation whose roots are z^ . th en &=i(i>/z7^4) or i(z 2 N/z a 2 -4). // in the last example all the roots of the given equation are imaginary and the values of z are real. prove that \ p \ <4 and q<6.4. z a 2 . one and therefore both values of z must be real. and only if. If any value of a: is real. but if a <0 or a> 16.(i) the roots are all real if. prove that x~ 2(1 -a)' ' Also if 0<a> 16. and only if.2 = 0.pz + q . [If z x + x~ l t then z* -f. In the figure OH=\QA. and . [For then 4(q . p*<($q + I) 2 and at least one of the two inequalities p*<4(q -2). If (x . p*<2q + 4 holds . if r is any integer.1) and x^ 1.2) <p* <2q + 4.] 18. 4(g-2)<p 2 <(itf+l) 2 and 2g + 4<p 2 .4. Show that. etc. If y r = cos-~ .

.. (A) If x y bja this equation becomes where // = ac-6 2 ....... The standard form of the cubic equation is w = ra 3 + 3ta 2 + 3cz + rf-0 .....27( 2 4-4# 3 ) ........... Equation whose Roots are the Squares of the Differences of the Roots.. those of (C) are aa + 6... The resulting equation is aV + 18a 4 ^ 2 -h81a 2 fl r2 2.... fl + b/a.36 2 c 2 ... Ex.... (F) where A = a?d 2 . y + ft/a. is obtained by putting q = 3H{a 2 . y are the roots of (A)... Hence the equation whose roots are (-y) 2 > (y-) 2 ... (<*-) 2 ........................ G=^a*d If 2 = ay = ax + 6..-f27(G ?2 -h4Jy 3 )==0........ Prove that the values of sin and cos-^ are respectively -^^ and A (I -v'17-f N /(34-2 N /17)}--i N /{174-3 N /17 + N /(34-2 x /17) + 2 [Referring to Art.... 2.. ... sin=V(i-~iy2) and {! + A - CHAPTER XII CUBIC AND BIQUADRATIC EQUATIONS 1. (D) whence the important equality a 6 (]3-y) 2 (y-a) 2 (a-j8) 2 = .....22......... ...... r = G/a* in Ch........... VI..... a/3 + 6.... The differences of the roots are the same for equations (A) and (B)..... ......... (C) If a.......6a6crf + 4ac 3 + 46 3 rf .... 5... j8. ay + b.. (0) . those of (B) are a -ft/a.... TheCubic Equation...... 17. (E) It is easily verified that 2 + 4# 3 = a 2 J... the equation is z 3 + 3//z + 6-' = . 1.

.. ifA>0...... all the roots of the cubic are real.. then z 3 . <*) 2 Q. Therefore m and n are the roots of t 2 + <#-# 3 =0.. the possibilities are (i) All the roots may be real. 180 CARDAN'S SOLUTION 3.. and its vanishing is the necessary and sufficient condition that the equation t/ = should have two equal roots.......A) 2 + /it 2 } 2 ... denoting the roots by a. Also.. Character of the Roots.. and remembering that imaginary roots occur in pairs.......... because m 3 w* = -/?. ajQ... (B) Let z = w*~ + w*..... A IIJL. O>Q- NOTE......*{ (a .. we find that the product of the squares of the differences of the roots is equal to ............ (D) and we may take m = J ( ... but Cardan's soluti on gives them in an imaginary form.... and. m *n* = -H... (A) becomes z 3 + 3#z + #=0 ....4p...... where CD is an imaginary cube root of unity..... ... (ii) One root may be real and two imaginary .so that a 4 (^-~y) 2 (y~a) 2 (a-j3) 2 =->27J ....(m + n) = ...... Hence if J<0.. m + n= -G.. one root is real and two are imaginary.. all the roots are real and unequal .. (C) Comparing (B) and (C). 6r=0 are the necessary and sufficient conditions for three equal roots. which is very unsuitable for practical purposes ... the equation az 3 + 3&z 2 -h3cz-hd = .. (H) The function A is called the discriminant of the cubic u.. If C? 2 +4H 8 <0..6. Cardan's Solution.. are Q-H/Q....G + >/(r 2 + 4/73).. If z=az + 6.. Also it follows from equation (D) that J7 = 0...... and then by (H).. Hence the values of z.... so that A >0.. If Q denotes any one of the three values of the three values of m* are Q.......3rwn*z .. -<*HIQ... J<0. Excluding the case in which two roots are equal. that is of ax 4. the corresponding values of ri* are -H/Q.. 4..

. = GI*Jh\ and the ....* In this case a solution can be easily obtained as follows.. Trigonometrical Solution when G 2 +4H 3 <0. If z = ax + 6. cos is determined by cos 3 0-f cos 0-i cos 30 = (B) Let z = q cos 0..7b + Q = 0.. a solution can be obtained by using the equation This is sometimes called the irreducible cote. // all the roots of ax 3 + 3&# 2 + 3cx + d =0 are reai.t. If a is any such value... (D) 2 3 2>/-/P Now G ?2 < . #<0 and h>0.. where 27/x 2 <4.. 5. where p and q are real. ... Taking the equation z 3 + 3Hz 4-0=0.... and so a real value of can be found to satisfy the last equation.___.... 2 +4# 3 <0..ifif 3 .. by a substitution of the form xp + qt... the equation becomes z 3 . the roots of (A) are ^27T jcosl -aj ^x.... FUNCTIONS OF THE ROOTS 181 If cos 30 is known. /.... a/iow tJiat the e quation can be reduced to 3-j-f^=0. 1. the equation becomes t*~t+p. Writing z=*Jh... where A= -3#=3(6 2 -oc).. so that 3ff G cos 3 + ^-cos0+ -^ = (C) q 2 q* v ' Equations (B) and (C) are identical if 4 $ ff = 2V-# and cos30-~-~= . (A) when 6r 2 + 4Jff 3 <0. Since aU the roots are real..~ where jj..

XXIX.. 1 + CO + C0 2 = 0... (A) i 3 -M 3 =-3(<o-a. a> + cu 2 M = 2y .... .........0..a) ..j8) . /J.. iJf = 2k 2 ..t....) 6..... j8 + h.cu W .. (2) L and M are functions of the differences o/a> j8.oj 2 ) (a .( co . Also.. 3. a> 2 L ... a a NOTE. where a> is an imaginary cube root of unity.. 182 SUM OF TWO CUBES (3) We have the following equalities : orL + wM = 2j8 ~ y .. and observe that CU 3 =1.L=a-hco/J-f co 2 y. (-JaM) 3 are Ae roofs of which is the auxiliary quadratic in Cardan's solution. The roots of the cubic can be expressed as convergent series by using this reduction.. y transforms either of the functions L 3 .wM = ( a> ... .. 9... Thus the transposition (a. 2 )( j 8... y + h f or a. Let . they are unaltered by writing a + A...substitution is 6 >Jh t x~ + ~..a> 2 ) (y ...... (1) TAe interchange of any two of the letters a. CO -^=1^3... (C) (4) From equations (A) and (C) it follows that (-|a) 3 .y )( y -a)(a-jS) ... M =a-f a> 2 /J-f coy..... j3.. by Ex.... Two Important Functions of the Roots.. y.. (B) Moreover. (See Ch. y... XII..a .2?j8y = -9#/a 2 .. It is the existence of functions possessing this property which makes the solution of a cubic equation depend on that of a quadratic.... M 3 into the other. For since 1-f co-f a> 2 =0. j8) changes i 3 into (0 + coa + a> 2 y) 3 == (cua + co 3 ^8 + co 2 y) 3 = coW 3 = M 3 . ' a>L .. Therefore Z 3 + M 3 = -27/a 3 ...a.

by symmetry..s .6c) 2 . then aA/Lt + o(A+jLt)4-c = 0. we find that (ad .6 2 ) x 2 + (acZ .B^- Eliminating A.. TJ A i c ^ o<Z-&c &i-c 2 = and //....6 2 ) (bd . (C) Also A. Therefore A and.... ... the reasoning fails... a( A 2 jit . /x are the roots of H s= (ac ..(A~/t).. B are given by A + B =a. and only if.. B from the first three and from the last three of these equations.. ^Aew constants A . A-4 +fJ3 = . A. p... ^ 6 v c oc-o 2 so that also in this case A and p are the roots of H = 0.A/u 2 ) + 6( A 2 -p?) + c(A -/x) -0. if w = has two equal roots. w=0 has two equal roots....... 2. Assume that neither A nor /u. is zero and that A^/x. and Eliminating /^.(bx + c) 2 = 0. // A jit..... THE HESSIAN 183 // A = 0. ^ are the roots of (ax + b) (ex + d) . . For using equation (G) of Art.. The Cubic as the Sum of Two Cubes.= = ----...4 (ao .... (A) provided that u has no square factor . and divide the first of these equations by A-/x and the second by Aju.7.6.be) x + (M ..A 2 ) = 0. B....c 2 ) = A Hence. Also A.= -. then by equations (B).. // u == ax 3 -h 36x 2 -f 3cx + rf. so also has H = and vice versa...c 2 ) = 0.. By expanding and equating coefficients it will be seen that the identity (A)holdsif A+B = a \A + ^. can be found such that u^A(x-\Y + B(x-tf\ . This case will arise if.. which is the same as H(ac~6 2 )x 2 + (od~6c)x + (6d!-c 2 )==0 ..

...I ) = k .. J R = (x~y)(a-j8)..a>aj8. M =a + eo 2 j8 + coy. 7?^ -3. The roots of H = will be denoted by A. (1) The function H which occurs in the last article is called the Hessian.1 . Let L =<x -H co/J + o> 2 y...1) 3 3(#-2) 3 . UG the preceding to solve the equation 2x* + 3# 2 . Hence the given equation may be written 5(x.. (A) (2) Using the identities of Exercise XII. _ 5"^ -2A. 5. 4. and we shall write where h Q = H = ac..4=5. Q = (a?-j8)(y-a)./75 -4/45)... so that A 1 2 ~4A A 2 = A . if P = (z-a)(j8~y)... (A) Then A.. 1. A+2B=../i=2.-l. and the roots are given by -^/5 (x . M' = /Jy -f co 2 ya 4. Let 2x* + 3x*-2lx + l$ = A(x-X)* + B(x-iJL)* .. Then.. are the roots of (2* + 1)( -lx + 19) -(#-7) 2 =0. /J. y as follows. h^ad. 6.a> 2 aj8...... A+B = 2. The Hessian..--a) 2 ()8~y) 2 4-(a:--j3) 2 (y-a) 2 -f(x~ y ) 2 (a^j9) 2 } ... where k 1... giving A = l..6 2 .Ex.. /3 (a: ..... Z/ = j8y + coya 4..... in terms of a. so that .w)L' =0. co.. //. A(> 2 .be. .. ft. (B) (3) We can find the values of A. co 2 . Since ^ 3 = 1 .. 8..3^ _ (5Jfc3*) ( Hence the roots are |( .227Q/J and 184 AUXILIARY QUADRATIC It follows from (B) that one root A of H=0 is given by that is.... ..c 2 .2lx + 19 =0.2)... therefore Z"P 2 = ........ and is of great importance in the theory of the cubic. we find that ~18H = a 2 {(o. equating the coefficients of x 3 and # 2 in (A).. we have P + <? + = 0....<*>)L + (co 2 ... and.. p. h 2 = bd.

If a.. Find to five places of decimals the real roots of (i) x* -3a.a. show tl t (i) (0-. a . L'. (vi) If the roots are all real and a> j8> y. (ii) 3.j3 is (iv) a a + 2 y + y a a=|(r + Jfc).so that A = -L'/l and similarly ^ = -M'/Jtf .She) x -f 9 (a 2 d 2 . (C) (4) It is easy to verify that the substitution reduces the equation h^x? + hfl + h 2 = to which is the auxiliary quadratic in Cardan's solution. If a. y . the difference between any two of the roots cannot exceed 2 N /(3g).3a 2 (acZ .... 0..3qx +r =0 and k = >/3(4g 3 -r 2 ). y are the roots of a: 3 . Find to five places of decimals the real root of (i) z s + 293-97=0.. (ii) z 8 + 6a.. a 2 y-f ^ 2 a + y 2 ^ = f (r-t). 0. M' 9 h ... M. y .. (iii) The equation whose roots are j9 .y.. (ii) 08 -.. y are the roots (i) the equation whose roots are j8 . (v) a a j8 + ^ 8 y-f-y 8 a = a 3 y-hj5 3 a + y 3 )8= -9g 2 .....y.fabcd + 3ac 3 + 36 8 d ) = 0. (iii) 2.-Hl=0.. (ii) the equation whose roots are a 2 j8 + ]8 2 y -4r y 2 a. Thus the roots of the auxiliary quadratic are EXERCISE XX THE CUBIC 1. a ... a 2 y + 2 a -f y 2 )9 is a 4 * 2 . In &e following examples ike letters L. h l9 h 2 have the meanin gs ..-26^0.jS is ) = . 2 H-27a.a.. and the difference between the greatest and least must exceed 4.

h n=h 29 where & iV( -JJ). o> 2 ). yoc + Jy + ma-fn-0.L'/L.A) + J3*(j8 .A) 3 -f B(x . y are the roots of . If u A (x . o>. and these are ( where 3& 2 =4A ^ 2 ~^i 2 and ^ > ^i> ^2 are the coefficients of the Hessian.3# 2 + 3=0 into itself. 5. Find the values of I. Show that L t + "Jf=3. m.JUT '/ Jtf . 1. a/? + Za + w/? + tt=0 . [Multiply equations (A) by (1. add and use Ex. there are two substitutions of the form which will transform the cubic equation u=Q into itself. (A) Hence prove that A/B= . [This is merely another way of stating the last part of Ex.assigned in Art.[ = ^(*i + i )> and a 2 y + j3 2 a + y 2 + 3^ = ~ (/H-fc)^ 8. o> 2 .JJ). 1). proving that 2^0^-^ + ^. a a TYPES OF SUBSTITUTIONS 185 6. Explain why this fails if A ~0.] 7.5. j9. If k= s/( . and unless otherwise stated a.. Find substitutions of the form x~(ly + m}l(l'y + m f ) which will transform the equation x 3 .x)=a) 2 ^*(y .A) +*(a . (1. (1. assume th at .A) -f JS*(y . 2h mh l -k.L 9 /M\ A = . y are the roots of u =0. ^ = .ft)=0. j8) o/ i^e cubic equation u =0 are connected b y the ' homographic ' relation ^ ^ 9. 8. prove that l jB + /8V4V + 8 . jS. That is to say.M'.] 10. any iwo roots (a.M )=a>4*(|3 . showing that these are * = (2y-3)/(y-l) and * = (y- ./x) 3 and a. 8. n such that y + J/? + ray -fr&=0. Except when J =0. w). and M* + L=3L'.

If a. Prove that the equation tt=0 is transformed into y*=k? by the substitution * y x* + px + q. Further. (1. If . 6 prove that p= -fA. (In other words. m. n the substitution transforms the equation x 3 . 2c .) 15.3qx + r=Q hi to itself. y are real and a> > y. 1S6 BIQUADRATIC EQUATION 14. . o>). + 35=0. r are the same numbers whatever pair is chosen. and add. -f 1 ^0 into itself. /j9 2 + mj5 + n = y. (1.where p 9 q. n such that Za 2 + ma 4.-. then 3i=Jf(A-. show that a 2 + 2a. o> 2 ). showing that these are 12. 36 N 2c N 6 .ft = /?.2 = are connected by a relation of the form pa. 1).*) = 27/2 [Multiply by (1. j3 of the equation x 3 -3ic 2 4. j8.60. find I.3# 2 . to 8 . y are the roots of a?-21a. then where jfc Hence show that with these values of I. 1. Show that any pair of roots a. jS. y are the roots of the cubic w=0 and <**+p<z + q = k. g= hAa a a where A is a root of the Hessian H =0. p 2 +pp + q = a>k. if a. if A = -L'/A ^= -Jf'/Jf.3s.] 16. Z proving that. y are the roots of x* -30#-f r=0. Find substitutions of the form x~(ly + m)l(l'y + m') which will transform the equation z 3 4. m.14 is equal to either ft or y. the substitution y=x* + 2x~ 14 transforms the equation into itself. j3. proving that the relation is 13. y ^ * 36 . -\-qfi + r=. <o. If a.11. /?.

will reduce tfc=0 to J^^S 3 . .6/a. giving the value of k in each case. ay t + _^ + _ !f+? .O f (B) where iTac~6 2 .if p= +A. The Standard Form of the Biquadratic Equation is u=*ax* + 46x 3 -f 6c# 2 + 4dz-f e=0 (A) If x = y . y + 6/a. a a a where A is a root of H=0 and k is determined as in Ex. if y-x 2 + 4x + 11.6. 17. the equation becomes z* + &Hz* + Gz + K=Q (C) If > j5. we must take L = .-f 11 and y=# 2 -f-7. 9. g= + A-. aS + 6. of t* =4 3 . Let A = /Jy-fa8. =a 3 e-4a 2 6d + 6a6 2 c-36 4 . 17. Some Important Functions of the Roots.) REDUCING CUBIC 187 10. a/J 4. 8 -f 6/a. S are the roots of (A). giving ic a -f 4x + 3=0.3 =0 and ( L) 3 . 17. M = 6. (JJif ) 3 are the roots of i. If we take A = 1. p. fi = . are the roote of x* + 2x .3. If 2= ay = ax + 6. 15. ay -f 6. [A.' (See Art. Explain why it is not to be expected that both roots of this equation satisfy u =0. y. In Ex. /A = ya-f-/JS.] 18. we may have i/ = 8. those of (B) are a + 6/a.6. G s =a 2 d~3a6o + 26 s . /? + &/#. y~x 2 + 4a. * This is a case of * Techirnhausen't Transformation. this becomes .6H 2 4G iC A m . Find substitutions of the form y ~x 2 + px -f q which will transform =0 into the form y 3 = ^ 3 . It is thus found that each of the substitutions. and those of (C) are ooc + 6.e. i>=a/J-fyS.

A. (2) The cyclic substitution (a/Jy) changes A to /it. v are equal. /i.. + aj3y 827a 2 ~ (4d 2 . j8....A = 2:aj8 = 6c/a...... /x.. (3) The functions A.. We have (j8-y)(a-S)=i>-/z (y-a)(j3-SHA-v . (H) Expressed as a determinant a & c bed c d e It will be shown that the solution of (A) depends on equation (G). v into itself or into one of the other two.(1) Any rearrangement of the letters a. K B.... and vice versa.. then A=/i = *>.Sac) a 2 ' a 3 The second term of this equation can be removed by the substitution and the equation becomes 4^3_/^ + j = o. 7a . v are Ae roote o/* a*y*-&a 2 cf + 4a(bd-ae)y-8(2ad* + 2eb*-3ace) = Q . 8 are equal....... Also if three of a. which is called the reducing cubic. (G) where / = ae ...... 188 THE DISCRIMINANT . (D) Hence if two of a. The functions 7 and J are of great importance in the theory of the biquadratic.C.ae)/a 2 .. and vice versa.. 8 transforms any one of the functions A. //. j8.4aj8y8 = 4 (4W ... y. ZajSy .4bd + Sc 2 . 8 are equal.. (E) For 2. y... y.. /z to v. v to A..See) + ^ (4fc 2 . It will be found that the existence of functions having this property makes the solution of (A) depend on that of a cubic equation. /J. then two of A.

.. or if a.... 8 are all real...... and therefore also t v Z 2 . For A... or if they are all imaginary... By equations (D).. jS... /x... then v is real and A...... (M) Another important equality is ( ) 8-y)2(a-8)2 + (y-a) 2 ( j 8-8) 2 + (a-j3)2(y-3)2 = 24//a2 . Then by (F) ^ = A-^-A-i(A+ / i + v)=i{A.. Also if a. . y. ..... then only one of the three t v t% y t% is real........ (5) The Functions I. V im' .. (0) where J = / 3 -27... Moreover.. t 3 ar e all real. v respectively.. (P) .. /?... j8 are real and y = / -f ^m.. /z. y..... 8 are of the forms lim... and therefore in equations (H) and (I) we may put 6 = 0. so also are /. t s be the roots of equation (Q) corresponding to A.. y....... giving a z I = K + 3H* and a*J^HK-G*-H*> whence the important identities = a 2 /-3ff 2 .. (K) a a whence.. v.. t^^a{(y . one of these cases must arise. y..... 8 are real and two are imaginary ..... 3 are functions of the differences of a... t z .. (L) with similar values for 2 2 .(4) Boots of the Reducing Cubic. y..^ft ...... J.... It follows that / and J are the same for equations (A) and (B)... c-H/a... by (A). (N) For and (6) The Discriminant. a(j8. then t v t 2 .y )(y -)( -/3)(-8)( ..... 3 .. 8 == / m..... and so the converse statements are true.. J.a )(j8-8)-(a-jB)(y-8)}.. e = #/a 3 .. jS.. .. 8 are a ll real.< 3 ) 2 (< 3 .... .7 2 .... /x are imaginary.. Let t^t 2 .... 8..< 2 ) 2 by equation (K) = 256J by Art.......... G* + H* = a*(HI-aJ) . and conversely. 2./t -(v-A)}... are all real if a.... d = /a 2 . If a. If two of the roots a..... /?. and conversely. Since t l9 t%.....

we find that the product of the squares of the differences of the roots is 16w 2 m' 2 {(Z . n 2 = s 2 -ae (C) Eliminating m. and so J<0.1) + w 8 }*. (5) If J>0. 15. If two roots are equal. Excluding this case and remembering that if there are any imaginary roots. and then J>0. 12.Z') 2 + (w + m')*}*{(l + Z') 2 + . Denoting them by lim. Writing the equation u=ax* + bx* + &cx 2 + 4dx + e^0. (3) All the roots may be imaginary.' (See Art. two roots are real and two imaginary. Ferrari's Solution of the Biquadratic. the roots are all real or all imaginary. /'iw'.ae) (as + 2V* -Sac) = 2 (bs . (F) which is the c reducing cubic.) Equations (C) become i = 2bt + bc-ad I (G) . (2) Two roots may be real and two imaginary. /?. we have 2w 2 = as + 26 2 ~3ac.m') 2 } 2 .ad) 2 . and so A>0. see Art. then these occur in pairs.REDUCING CUBIC 189 The function A is called the discriminant of the quartic w. 11. (s 2 . it is easily shown that the product of the squares of the differences of the roots is -4w 2 (a-0) 2 {(a . For the criterion distinguishing the two cases of (5). I im. Hence (4) // J<0. wn = &s-ad. Character of the Roots. which reduces to 5 3 -3c5 2 + (46d-ae)5 + (3ace-2ad 2 -266 2 ) = (D) The second term can be removed by the substitution s = 2t + c (E) and equation (D) becomes & 3 ~/* + J = 0. (A) we assume that au = (ax* + 2bx + s)*~(2mx + n)* (B) Expanding and equating coefficients. then J = 0. and its vanishing is the necessary and sufficient condition that the equation w = may have two equal roots. Denoting the roots by a.1)* + !*}*{ (0 . the possible cases are : (1) All the roots may be real. n. 10.

. (a.. 10.. Now a(3y = c + 2t l + n v therefore a (/?y 4.. Also a(j8y-a8) = 2w 1 . let (y. wo have The last equation is satisfied if s -~ . y) and (1. a).. + 5 -f n} {a. (H) then a.. Expanding and equating coefficients.. Hence (i) the values of m are . Solve n =x* + 3x 3 + x* .aS) = 2c + 4^. It should be noticed tha^ the three roots of (F) correspond to the three ways of expressing u as the product of two quadratic factors. 2 + (p ..77^) a.0.... (I) Further... n t = (26^ + 6c .ad)/m v the equation u^=0 can be put into the form (ax 2 + 26z + c + 2^) 2 . + n x ) = .. 13. if fj is a root of (F) and m l =\/ (at v + b 2 .. Therefore the roots are -1 >/3.?/i) a. inn.. j8. =i.4.. 3) of the suffixes.... + s . Ex. 1.. a(]8-f y-a-8)== -4m 1} ..(2^0. . Thus w e ma y take a =--j.190 FERRARI'S SOLUTION Thus. and then ?yr -1+4. which agrees with equation (F) of Art. y be the roots of ax 2 + 2bx + c + 2^ + (2m 1 x + 7i 1 )=0.2 = 0. S are the roots of ax 2 + 26x + c + 2/ x . and its roots are the roots of the quadratics ax 2 + 2bx + c + 2^ = (2m 1 x + 14)..(2m 1 a? + n x ) 2 .. and u [x* + (p -f..... 2. Let ]8... (J) with similar equations given by the cyclic substitutions (a.ac). =-|..n} ^(x*+2x-2)(x*+x + l). /?) be the roots of the equations obtained from (H) by changing the suffix 1 into 2 and 3 respectively. \vhero w is an imaginary cube root of 1... w.(mx + n) 2 .J-.. a> 2 ... Let u = (x 2 +px + s) 2 . Deductions from Ferrari's Solution...

(G) it follows that these functions are the values of z found by eliminating tfrom __ _ 1 2bt + bc~ad Z ~ ~2' 7 and ... o y + oc ~ p S ot + p y o and from (F).. (M) 14... ~ are 2m l 2m 9 r( 2 ya-)8S ___ J^ZY^ 3 + y Of. Ex. Biquadratic as the Product of Quadratic Factors. and assume that .4foe 3 -f 6cx 2 -!. (ii) The values of n are \a(fiy -aS). VI.... obtained in another way in Art. . Descartes 9 method.. |a(ya .yS). 96. (iii) The values of -~. (K) DESCARTES' SOLUTION Also. 16.j8S). 3. -f e.In connection with these functions it should be noted that by Ch. Let u ax* 4.2 .4c?o. p. (iii).> -~. 15... la(oc^ .. the equation whose roots are 191 This is an important equation. by (F) and (G).

a a m + m' -0. ... etc.. m + m'. we have a b 3c-2all f b all' d 3c-2all' d e \ =0.u = a (x 2 + 2Zx -f ?/i ) (x 2 + Expanding and equating coefficients. a a Now 1 1 I l f m m' 1 1 |l f I m' m k d . and multiplying each row by a/2. we have l + l = 2 -. 2 Z -f V 211' lm' + I'm m + m' lm +l'm 2mm' Substituting the above values for 1 + 1'. -m?w =. m + m = 6 ~ 4a . e : .

Four Real Roots. we proceed as in Art. by one of the methods described in Ch. NOTE. which is obtained as above for the biquadratic.3# 2 x -f 2 = 0. m. The complete set of conditions may be found by using Sturm's theorem. J>0. the conditions can be found much more easily by the use of either Ferrari's or Descartes' solution of the biquadratic.= 0. x* . we can now find values of I. it has been shown that J>0 is a necessary condition that the four roots should be real . it will be found that one r oot can be found approximately with very little trouble. involves rather tedious reckoning . / .2p/3 for I 2 . to obtain the cubic in the form P(l 2 +p) 2 irl 2 y 2 = 0. This. VI. 11. In numerical work. at least two of the roots must be imaginary. XXVII. since it is only necessary to have one way of breaking up u into quadratic factors. given later. the usual course adopted in text-books.15 =0. Ex. by De Gua's rule (Ch. 2 8 . If we write the equation becomes which when expanded is Corresponding to any root ^ of this equation. 11. if the equation is given in the form u^-ofl + px 2 + qx + r = Q. a b c + 2t b c-t d c + 2t d e 192 POUR REAL ROOTS Again.* Later.1 \z . the student can convin ce himself of this by verifying the steps of Cardan's method for solving the equati on. 15. using I instead of 2Z. the solution beco mes very laborious by the methods described in this chapter . 15. In Art. V. also. Hence. unless this cubic has rational roots. but these two alone are not sufficient. This is all that is require d. which reduces to an equation lacking the second term on substituting z . 77 <0 are necessary conditions that all the roots should be real . if #>0. 1). m' which will satisfy the conditions.

. 10. (A) which. (M).. JJL.e. . v = has the same number of real roots as w = 0. and supposing that v 55 (z 2 + 2lz + m) (z* . ut?ll/3... we may denote them by Ait/i.. and 12# 2 > 2 /. the values of I 2 are the roots of 2 =0.Theorem. the three values of I 2 must be all real and positive. with the substitution. the roots of (A).G 2 = 1 2 K..e. -i+t>.. where A-Fju-h2A' = . A'i/*'. J(A -f X' t///) 2 . and conversely. y + H = at. or their equivalents (ii) J>(). z is given by v = z* + 6Hz* + 4:Gz + K = O t where K = a 2 I-3H 2 .. and then the values of Z 2 . 9. of which one is real and t he other two imaginary . (2Z 3 -f 3HI) 2 . and thus the values of I 2 are A 2 . since K=a*I -3H 2 . For v = 0. Now. from which ttS(*H2-42042s+l-63580)(s-2-42042a:+l-22265)=0. then. by equating coefficients. and Art. since their sum is zero.. t*-t?= 5-27397. Hence. and conversely.. if. -A-ti/i/. and 2Hl>3uJ. It follows that the roots of v = 0.3-85845 . Hence... the roots of (A) are all real and positive.. using the method of Descartes. The necessary and sufficient conditions that the roots of u == ax* -f 46x 3 -f 6cx 2 + kdx -f e = aw all real i* (i) J>0. are A' 2 .. Eliminating m and m' from these equations. i. Let z = ax + b .. but only one is positive . (i) If the roots of v = are all real.. and therefore those of u = 0..e. FOUR IMAGINARY ROOTS 193 Now the three values of I 2 correspond to the three ways of expressing v as the product of two quadratic factors. the three roots of (A) are all positive.. i. (ii) If the roots of v = are all imaginary. t>=l-0943. or W + l2Hl* + (9H*-K)l*-G*=Q. ti=-2-16423. H<(). tt'+v^lS. and only if. i. we may denote them by A. (iii) If two of the roots of v = are real and two imaginary. //<0. are all real. (C). then by Art. the roots of (A) are all real.. reduces to 4J 3 - *Thus. -4(/4+/z') 2 > -i(/A-/i') 2 ...2lz + m') 9 we get.

aJ) . | 2HI\>\ 3aJ\ . also. is equation (A) on p. where the last factor is the product of pairs of complex conjugate factors. using Ferrari's method. by addition. all real and positive. . positive or negative. First observe that. the roots of (A) are real if those of 4 3 . which is negative. but / is positive. * In general. G 2 + I2H*<a 2 HI . if n is odd. suppose that J>0. hence p n ^ q n according &spr^q. Then.It + J = are real . it follows that p* : q 3 according as p g q* Then.. hence. G 2 + 4# 3 . 12H 2 >a 2 I. suppose first that //<0> and i2H 2 >a 2 I. since 2HI<3aJ. But. 10). p 2 + pq + q 2 = (p + %q) 2 + f <? 2 >0 . To prove the equivalence of the second set of conditions. and only if. A = / 3 -27e/ 2 >0 . we have />0 . since p 3 -q 3 = (p.e. +g n ~ l ). which proves the first part of the theorem.Now. therefore 2HI<3aJ. pnqn =*(p-~q)(pn. and J>0. since J=/ 3 -27/ 2 >0. hence pn^^n as |p|<|ff|. we have 8/P/ 3 <27a 3 J 3 <a 3 /V. if. and not as p :g q. Thus. and 2HI<3aJ. when expanded. and therefore 8H*<a*J.. so that 4# 2 / 2 >9a 2 </ 2 . if n is even. hence.1 +pn-*q+ . if p and q are any real numbers. 9. and this is the case if.q) (p 2 + pq + q 2 ). and dividing by H. since HI is negative. VI. i. and is therefore positive . by Descartes' rule (Ch. therefore l2H 3 <. H is negative and 12H 2 -a 2 I is positive .a 2 (HI . it follows that HI<0 Also. ff<0. the roots of (A) are.a 2 HI. 192. since //<0. pnqn^ty q)(p+q) (a positive factor). the equation whose roots are s which. There are three possibilities. the two sets of conditions are equivalent. Next. Now. I2H 2 P>a 2 P>27a 2 J 2 . hence. and. 194 EECIPEOCAL TRANSFORMATION Alternatively. and only if.

...(i) AH the roots of u = may be real .... The substitution x = py + q transforms the equation w = into apY + *BrPy* + Wp*y* + 4:Dpy + E = Q.. is given in Exercise XXI... which are all three real.. y. 13.. 17. 8 = A' I/A' . 1)3... in that case we may take a = A + i/z.. (C) and therefore Hence by Art. due to Euler. efa I) 4 .. E-(a... and that the roots of (A) are !/i> 2/2> J/3> 2/4 where J/22/3 == J/iJ/4 :== l* Let a. 6.. 6... C = aq 2 + 2bq + c... so that It follows that ? 2 H~?)(y-?) = (a~?)(S-?)...... -iW) 2 > 1 2 (A. that is... This will be a reciprocal equation if ap* = E. (A) where B = ag + b... j8 to be the real roots and y = A + t/i. p* = D/B . in that case the roots of (A) are all three real and positive. if q. of which one only is real.. y = A' 4..an d 2 . (iii)..... . d. <%. S be the corr esponding roots of w = 0.. but owe owfa/ is positive.... Bp*==Dp. (ii) All the roots of u = may be imaginary . j8.... 16... c... c... Transformation into the Reciprocal Form. /? = A-<fc..t/i .. 8 = A ..... the values of q are the roots of the equation in z foun d by eliminating tfrom S' i LO ..A') 2 ... (iii) Two of the roots of w = may be real arid two imaginary . (B ) Suppose that the conditions (B) are satisfied. . Another method of solving the biquadratic. and then are the roots of (A).. we may take a.. D = (a.. p are given by aZ> 2 = 2^. The proof proceeds as before.I/JL'. and the roots of (A) are -iaV-fO 2 ..

.TSCHIRNHAUSEN'S TRANSFORMATION 195 NOTE... the corresponding value of p 2 is -i + |~|_ * Taking p = 1. d= -^..*.. That is to say. This process is called Tschirrihausen's Transformation : it has been applied to the cubic in Exercise XX. one substitution of the kind required is * =y .. *=i(.. . Use this to solve the equation. /. Dividing by y 2 and putting z=y-f-y"" 1 .C8=CJ 2 ==JF / C a .. Tschirnhausen's Transformation. . 2/=i(2V2 a -4).. .1 + a 2 x n ~ 2 + . The equation 4* 3 -/* + Jr=0 is 4e 3 -f^-^=0...... y) and (a. in which (j8.1..1+^5^-2 + 6^5). Here o = l. Substituting in (A).3* + 1 = 0. and aj=J(z~2db^ a ~4) . j8... in general p v p 2 .3t/ -f 1 =0.Cy = Ca... n + a^. 8. the equation becomes y 4 ~ %!? + 3y 2 . One solution is $=-J.. 17. we find that 2 a . Thus theoretically. the corresponding value of q is and by (B) of the text.. y. 6 = J. ... c=-l. that is (4*) 8 +2 (40 -3=0. Ex... 1. If we eliminate x between /(x) = a o.. giving z = | (3 /5).. are respectively the centre C and the foci F y F' of the involution determined by the "p/ COtfl F Y fi" points a..... the line segments are such that C f j8.. Also y 1 -#s + l=0. The points corresponding to q and qp 9 as given by (C).. 16.. p r can be chosen so that r of the coefficients b l9 6 2 > ^n are z ^ro. c = 0. 8) FIG 29 are corresponding pairs of points... + a n = and y=x r + p^" 1 + ^2^ r "" 2 + + Pr we shall obtain an equation of the form for a single value of y corresponds to each of the n values of x. Find a substitution of the form xpy+q which will change the equation x* + x*-2x-l=Q . or i( -1 -^^^2 -6^6)... . (A) into the reciprocal form. and by (E) of the text...

......... the resulting equation is of the form 2/ 4 +/</ 2 + <7 = .. j8......... Let a.. where 2 = * . y-a-8) = 0.. (D) and s 2 +^5 1 + 4y = 0.. (A) Further.. We shall prove that in general three sets of values of p and q can be found such that if x is eliminated between u = and y = x 2 + pz + q....-Zz...... ...... Hence we have. .... q be chosen so that the coefficients of y and y 3 are zero.... (B) where z is one of the three j8y~<xS ya-/3S aff-yS J3Ty-a-S ' y+a-jS^ ' a + jS-y-S * One value of y corresponds to each of the four values of x ................ from (D)........ the resulting equation in y is therefore of the fourth degree............. y. 8 be the roots of u = ax 4 -f 46x 3 -f 6cz 2 + 4dz + e = 0. aq = 3c + 2bz.... ....196 TSCHIRNHAUSEN'S TRANSFORMATION Case of the biquadratic....... and let y l9 y 2 be the values of y.. s z = Z<x?....... Let p... Now (j3H-y) 2 -( j8y-a8 therefore -p^s.... (E) where s l = 2oL........ the values of p and q are given by ap=-Ab + 2az.. Then we may take 4-j>y + ?= -~2/2 whence by addition and subtraction.

1 . . e = . and using Art. and use the result to sol ve the equation.. which reduces to .. g.. we have * y(x* . 13. frcttn (A) .... Find the values off..200-0.. One value of t is 1. it is obvious that the equation in y is of the form (A)...9x 2 . d = 3...5..9x 2 . we have z^ where 4* 3 +5*-9=0. q=0..I2x -H 5.. J=-9. But x z + 43# y ... c =0. we find that a/ a a \ a 3c 2 giving ( } Moreover... SUBSTITUTIONS 197 As will be seen in the next example.. so that 7=-6. q have these values... b =0.. One such substitution is therefore 2 O flH\ y x H~ oiC.. each value of y gives one and only one value of x which* satisfies the equation w = 0.. giving z~f and then p=3... 4 ..3x) =a. Here a = 1. (F) Substituting in (E). ( iii).46 Hence p = + 2z .... Find a substitution of the form y=x z +px + q which will reduce the equa tion to the form y 4 +/?/ 2 + g ~0.. if p...... Ex. \D j To eliminate x from (A) and (B).

[Dimmish the roots of u~Q by 2 and use Ex. I) 4 . I) 3 . y.The values of y are \/2. 3. tt=0. c. 4. If ( j8 -H y)8 = (-*. [Deduce from Ex. If + y = a + 8. and the corresponding values of x are given by Hence we find that the roots of (A) are -lN/2. 10*./?S a/? . If L show that the interchange of any two of the four a.4d(a: + e =0. y. show that the eq uation .3edc = 0. e$z. b. 4 -fg# 2 -f ra-f s ~0. 6. th e method for obtaining the second quadratic (G) is y [ao? +(46 ~ap)x] * ( and thus. show that e 2 6 -f 2d 3 . EXERCISE XXI THE BIQUADRATIC Unless otherwise stated a. j8. d$z. 8 changes L 3 into J/ 8 and M 3 into L 3 . Show that the equation whose roots are j8y . D-(a. 2. y. show that od 2 =6 2 e.] 3. If 0y=a8. show that a z d + 26 3 . 1. If a. ft. 2 4. #-(a. d. y**&+px+q. and the substitution. c.] 6.yS /94-y-a-o' -/ is where B-az + b. j8. 8 are the roots of u^ax* + 4bx* + 6co. * In the case of the general equation.8) j8y. 1. 198 FERRARI'S METHOD 5.a8 ya . 3 are the roots of w==o.3ofo = 0.

(y 4-a . 10. If a. * 4 -f 12x 3 4-54^ 2 -f 96x4.6a: -h 3 as the product of quadratic factors in three different ways. 13.ft .225 = 0. Hence show that the equation has two and only two real roots unless a . x* .S) 2 . (Hi). Show that the equation whose roots are s 3 2 3 [Use equations (D). 14.3z 2 . jS. = 1 or 0.3x 3 4-# 2 -20.4x ~ 3 ~0.5. S are the roots of x* -\.19z 2 -f 243z . (G) of Art. Solve x* .r 2 = 0. 7.2x* -\-a(2x .1 ^0. y. prove that the equati on whose roots are (j34-y-a-S) 2 .4sz -f 4g# . 11.<jz 2 . and that if 2 X is a root of this equation. 10.] 8.4x 3 + lx z . . 13.1) ^0 by putting it in the form and choosing s so that the right-hand side is a perfect square. a 4 4 x*~2x.] 16. (a + j8-y-S) 2 is z 3 . Express # 4 .u 4 2 = 0. * 4 + 12:r-5=0.4* 3 4. 3.40 = 0. 15. 12. x 4 .whose roots are is z 3 . and Exercise XX. then where A ls A 2 are the roots of A^Az^-s^O. Show that the equation whose roots are and two similar expressions is 4(*4-L) 3 -e 2 /(z4-L)4-eV=0 where L=ce-d*. [Use equation (L) of Art. Solve by Ferrari's method : 9.

] EULER'S SOLUTION 199 17.z + l~. 19. (Cf. m. *Jn denote either of the square roots of /. 13. etc.>/n.. Find a substitution of the form x=py + q which will transform the equation into the reciprocal form.] 20. s/n has the same sign as G.) [The left-hand side is 5H(z 4 + 12* . 13 by changing a. (ii) If this equation is identical with z 4 4.Jl-~ *Jm .*Jm + -s/n. *Jm.2x . (iii) The complete solution of the biquadratic u = is given by aa -f b = . s/w .22!tom =0. 13. Use thi s to solve the equation.[This follows from equation (L) of Art. show that the other four values of x are the roots of x* f 12. Observe that the values of I. 0. 18. n are i\a 2 (j3 + y-a-8) 2 .6//z 2 + 4Gz + A" = 0. we get eight values of x. see equation s (L) and(K) of Art. then Z.(9// 2 ~-K)s. the square roots being chosen so that v/Z . 8 into their reciprocals. four of which sati sfy Without solving any equations. Euler's Solution. 1 of Art.*Jn. m.G 2 0. 17. a8 -h 6 = . af3 + b~ *Jl. n are the values of a given by 4s 3 -f 127/s 2 4. In Ex.1.5).r 3 + 54x 2 + 96x f 40 = 0.*Jl + sjm + >Jn. n and z = *Jl + \/m -f \/n. which is the same as 4(s-f//) 3 -a 2 /(s + //)+a 3 </-0. .200^0 is substituted for y in y x 2 + 3x. Find a substitution of the form y=x 2 +px + q which will reduce the equation x* + x 3 . Show that one such substitution is x = 2y . m.223 . z 2 . Ex. if each value of y which satisfies the equation y * + 9% 2 . ay + 6 = ^Z + Jin . prove that (i) z 4 .&*Jl*Jm*/n . If Jl. y.1 to the form y 4 -f fy 2 + g = 0.

(Cf. Use this to solve the equation. Suppose P to represent some rational. Ex. The number 3 may be assigned to either dass. If 3 belongs to A. which will be called the lower class A and the upper dass A'. showing that one such substitution is this reducing the equation to 4(1 CHAPTER XIII </ THEORY OF IRRATIONALS 1. and which therefore belong to the class A . Sections of the System of Rational Numbers. according as the point P does. Lower class A Upper class A O P FIG. any number in A is less tlutn any number in A'. The system of rationals is thus divided into two parts. ) 21. The class A contains every rational less than 3. Two distinct cases arise. Similarly if 3 is assigned to -4'. and there is no least number in A'. 30.Show that one such substitution is y =x 2 . it is the greatest number of this class. 3 for example. represent a rational number. For if a' is supposed to be the least number in A '. it is the least number of this class. First case. or does not. this reducing the equation to (2y) 4 + 26 (2y) 2 -11^0. and the class A has no greatest number. Find a substitution of the form y=x 2 +px + q which will reduce the equation to the form y*+fy 2 + g=Q. Suppose the rational numbers to be represented by points in a straight line. and let the line be cut at any point P.x . yationals exist which are less than a' and greater than 3. and the class A' contains every rational greater than 3. 10. Thus.1. and the class A' contains all those to the right of P. . The class A contains all the rationals to the left of P.

This classification is said to define the irrational number which we denote by -s/7. according as a 2 <7 or iJJKDedekind's Definition. For no rational exists whose square is equal to 7. This will be the case if 7 -a 2 7 -a 2 6-flK-r . so that any number in A is less than any number in A' . and a is not the greatest number in A. Suppose that a certain rule enables us to divide the whole system of rationals into two classes. and so a<a'. For if a is any positive number in A and a' is any number in A'. For. b -fa 2a Thus. 6 belongs to the class A. Two cases arise : (i) If A has a greatest number or if A ' has a least number.J1 is to follow or to precede any positive rational a. we have to find b (greater than a). (ii) Any number in A is less than any number in A. we have a 2 <7<a' 2 . then a 2 <7.Thus corresponding to any rational. (iii) The class A has no greatest number and the class A has no least number. namely the greatest number in A or the least number in A . The upper class A' is to contain every positive number whose square is greater than 7. DEDEKIND'S DEFINITION 201 ticcond caw. and we can find a rational b such that 6>a and 6 2 <7. a lower class A and an upper class A'. the classification defines an irrational number which is to follow all the numbers in A and to precede all those in A. . the classification is such that either the dass A has a greatest number or the dass A 1 has a least number. let OP be equal to the side of a square whose area is 7 square units. A'. Similarly it can be shown that A' has no least number. so that & 2 -a 2 <7 a 2 . Suppose that the point P does not represent a rational number. Classify the rationale according to the following rule : The lower class A is to contain all the negative numbers.<-s . suppose that a is the greatest number of A. The classification is such that (i) Every rational is included in one or other of two classes A. the classification defines a rational number. zero and every positive number whose square is less than 7. and we assign it a place on the scale with rationals as follows : . To do this. For example. (ii) If there is no greatest number in A and no least number in A'.

10. (B. then r is the least number in A and the greatest in B. each of the classifications (A. the section (A. Equality and Inequality. 202 EQUALITY AND INEQUALITY The classification just described will be called the classification or the section (A. If A contains some positive rationals. A). (i). these definitions must be in agreement with those already given for the system of rationals. Thus at least two. For let a<j9. If r is the only rational between A' ex B Fio. Moreover. where a and j8 are defined by the classifications (A. (A. A') defines a positive real number. infinitely many rationals lie between them. (J5. which is impossible. by Art. 2. Then r follows a and precedes jS. a and . and the real number defined by it may be known as the number (A. We are not justified in regarding a ' real number ' as a ' number ' until we have given fresh definitions of equality. A'). rationals lie between a andjS. Two real numbers are said to be equal when they are defined by the same classification of rationals. 31. II.) . A'). we say that a is less than j3 and that j8 is greater than a. If A' contains some negative rationals. and we should have <x = r = /?. (See Ch. A') defines a negative real number. If a and /J are real numbers. A). at least one rational r lies between a and jS. B') defines the same number r. and therefore infinitely many. The real number zero is defined by the section (A. therefore r belongs to each of the classes A' and B. inequality and the fundamental operations of arithmetic. Since these are different classifications. B f ) respectively. for a</J. A'). Hence. // a and j8 are unequal real numbers. and some of the rationals which follow a precede /J. where A contains all the negative rationals and A contains all the positive rationals.Any rational or irrational number is called a real number^ and the aggregate of rationals and irrationals is known as the system of real numbers.

5. for we can choose a and a! so that a' -a<V -I. Divide the system of rationals into two classes according to the following rule : The lower class is to contain every rational less than any number a' in A f . Let A and A' be two classes of rationals such that (i) Each class contains at least one number. and only one. exists such that # n ^a<a w ' for every n. however small. This theorem may also be stated as follows : If (a n ) and (a n ') are sequences of rationals such that (i) a 1 <a 2 < 3 . and (ii) it is possible to find n such that <*n -<**<> where e is any positive number we may choose. so that a' -a<. If there is one rational which escapes classification. Endless Decimals..Theorem. <a 3 '<a 2 '<<.... Proof. where is any positive number we may choose. <a n . however small) then one real number a.. so that either a or a' lies between I and I'. then for every a and a' This contradicts the hypothesis. let a -a 1 o 2 . * ** (iii) Two numbers a and a' can be chosen from A and A' respectively. .a n ... Then there is one and only one real number a such that a<a<a' where a is any number in A and a' is any number in A'. (ii) Any number in A is less than any number in A 1 ./ . ENDLESS DECIMALS 203 . then it is to be assigned to either the upper or the lower class. For if two rationals Z and V (l<l f ) so escape.. the classification defines a real number a which satisfies the Conditions stated above. where the successive figures are formed according to some definite rule. Thus. The upper class is to contain every rational greater than any number a in A.. It will be proved that not more than one rational can escape classification.. <a. be an endless decimal. by Dedekind's definition. Using the ordinary notation.

6) = (a f . and therefore a real number 8 is defined by We say that the decimal represents or is equal to this real number S. . any real number 8 can be represented by a decimal. . then for any suffix n. j3 are real numbers defined by a<a<a' and &<<&'. if 8 is known. it follows that a non-recurring endless decimal represents an irrational number. after a certain stage. a'. a small Roman letter represents a rational. however small. Of course. This classification defines a real number. A! of rationals which satisfy the conditions of the theorem. 4. and then (a f -f b') . a' are any numbers in classes A. For. o B.Let d n = a a^ . 204 FUNDAMENTAL OPERATIONS Conversely. b' so that a' a<^e and b' 6<i . Show It is obvious that a -f ft and j3 -Ha are defined by the same classification of . 6. we can find d n so that d n <8<d n '. it is presumed that a. using the same notation as in the preceding. Real numbers will be defined by using the theorem of Art. (iii) We can choose a. we can calculate to any number of figures a decimal which represents S. where e is any positive number.A. and when we say that a real number a is defined by a^a^a'. (ii) Any a -f b < any a' + b'. That is to say. then a -f j8 is defined by a -f 6<a -f /2<a' + b'. a w and d n ' = d n + 1/10 W . and conversely.(a -f. a terminating decimal may be regarded as an endless decimal in which. We define a -f /? + y as meaning (a -f jS) + y.b)<. and it is easy to show that the commutative and associative laws hold good. The fundamental Operations of Arithmetic. The sequences (d n ) and (d n f ) obviously satisfy the conditions of the last theorem. for (i) There is at least one a -f 6 and one a' + b'.O. If a. (1) Addition. Since any rational can be represented by a decimal which terminates or recurs. In what follows. . all the figures are zeros. a Greek letter denotes a real number. 1.a) + (V .

a'. a + 6< a< a' -f 6'. Prove that a + ( . x. If a. a a a' '// > /} 6' k FIG.a'< a 4. Let a be defined by a< a< a'. 2.a'< 0< a' . These conditions define a real number.rationals.a)< a' . then if a'<A. then a + ( . (ii) Every a6<any a'b' . If a is defined by a< a< a'. Now every a< any a'. therefore a . We define <x~j3 as a-f (-/?).a. Hence a + ( -a) and are defined by the same classification. Choose rationals h. k. then a + is defined by a -f 6< + 0< a' -f 6'. and are equal. MULTIPLICATION OF IRRATIONALS 205 . &' so that a'b' -ab<. . and since every 6 is negative and every 6 X positive. Prove that a + = a. (2) Subtraction. 6. (i\\) We can choose a. for (3) Multiplication. Thus a + and a are defined by the same classification of rationals and are equal .a.( . Zero is defined by b< 0< &'. for (i) There is at least one ab and one a'b' . For a'b' -ab = a'(b' -b) + b(a' -a).a) =0. where is any assigned positive number. where 6 is any negative rational and b' is any positive rational.a) is defined by a . /? are positive real numbers defined by a^oc^a' and 6</3<6'. so that A>a and &>/J. 3. Whence it follows that (a-j8)+j8=a. then aj3 is defined by afe<a/3<a'6'. 32.

^7 is defined by a 2 < 7< a' 2 . First choose h so thai. If a is a positive real number defined by a^a^a'. jB) =a. for (i) There is at least one I/a' and one I/a.a). <Aen a/j9 is defined by a/6'< a/< a' ft. we have to show that a. I//?) . Further. 206 POWERS AND ROOTS (4) Division. Further definitions are l/( -a) = .^1 = 7. 4. .b) + k(a' . a'. jB =a(l/jB . a so that a>/i and a' a</re.ab<h(b' . a' are any positive rationals such that a 2 < 7< a' 2 . Consequently rio meaning is assigned to 1/0. a' "i . whence it follows that (a/j3) . Next choose a. NOTE. JSfo. a(-j8H-aj8 = (~a) and (-)( -J8)=aj8. Since h<a<a'. Show that ^1. a. and this classification also define s the number 7. a/Sy is taken to mean (<xj8)y. we have h 2 <aa' and a'-a<aa'e. for (a . 5. It is easy now to show that the commutative. This follows from the definition of l/j3 and multiplication. ^ Ex. /? =a . = = 0. where rt. so that 6' . distributive and associative laws hold good. These conditions define a real number. (ii) Every l/a'< any I/a.b</2h and a' -a<e/2k.I/a and a/j8=a .ab<. j9 are positive real numbers defined by a< a< a' arw2 b< j8< &'.a'6' . For zero and negative factors.0<A<a. can be chosen so that p IG ^ a -a<aa f . We can now choose a. ^/7 is defined by a< N /7< a'. To prove this. 6. (iii) We can choose a. // a. then I/a is defined by l/a'<l/a<l/a. where e is any assigned positive number. the definitions are a. a' so that l/a-l/a'<. l/j8 . Tliis reasoning depends on the existence of positive rationals less than a . 6'.7* ' "~~l . and then a'6' . Hence *y7 .

whose n-th power is equal to a. then positive rationals x l9 #/ exist such that For if x<x<x'. . If a = z n where # is a positive rational. Then (i) no rational escapes classification . for no rational x exists such that x n =a. there is one positive real number . Theorem 1 . This follows from the last theorem. and every positive number x such that x n <oc. Again. for sc n <Ca<J?' n . however small. to n factors. zero. for the nth powers of unequal positive rationals are unequal. a .7. divide the system of rationals into a lower class A and an upper class A' '. The existence of a number x such that a<x 1 ' n <x /n can be proved in a similar way. nx^. If n is a positive integer and a a real number.x n . then x/ n ~x 1 n <. x/ n ~x 1 w <(x 1 / -ajj) . for every x and every x' '. a . (iii) There is no greatest x and no least x f .x n <a . If m. PRINCIPAL ROOTS 207 Theorem 2. the nth power of a (written a n ) is defined by a n ==a . n a positive integer and x. The class A' is to contain every positive number x such that x' n >a. Therefore the classification defines a real number . x' positive rationals such that n <a<x' n . . (ii) Every #<every x'. //a is a positive real number. If no such x exists. and one only. That these formulae may hold for zero and negative values of m and n. Powers and Roots. according to the following rule : The class A is to contain every negative number. . provided For as in the preceding. (ii) // is any positive number. n are positive integers. then We can choose ^ 1 >x.1 . there is only one such number x. so that x x ~#<(a -x n )fnx' n ~ l : and then x x w . so that x^ <a. it follows that a m . we must have a = 1 and a~ n = l/a n .a n =a m + n and (a w ) w =a mn . (i) // a is a positive real number.

as in Theorem 1. Thus (. y. Two surds of the same order are like surds when their quotient is rational : otherwise they are unlike surds. and therefore no real number exists which is the nth root of a negative number. . This definition assigns no meaning to a x when x is irrational. 8./a) n =a. Definitions.^a . If a is not a perfect nth power. where e is any positive number. Therefore . therefore these may be chosen so that x' x is as small as we like. ^Ja is called a surd of the nth order.or else y and b are the squares ofrationals. The following theorems and examples depend on the fact that a rational number cannot be equal to an irrational.n== (. For suppose that x-^a and let x = a 4.z. 4 and therefore n =a. we have (. and by squaring. a. of -a as . or simply the n-th root. Surds. however small. 9. the principal n-th root of a is defined as the positive real number whose nth power is equal to a. This is written J^a. If n is odd.y]8-y(aj8) and WH^a^ y(. Also every rational is an x or an x'. ? a q is defined as *]a* or (!ja) p . 208 SURDS Indices which are Rational Fractions. When n is an odd integer. If a is a positive real number.x n <oL<x' n and x n <g n <x' n . q are positive integers..Jy is rational. This case is considered in Art.1 ) n (^ a ) w== -a. a classification which satisfies the conditions of the theorem of Art. If p. then x = a and y = b.. Hence. then z + Jy Jb. they may be chosen so that x' n -x n <. thus If n is even. If x+Jy^a + jb where x. b are rationals. Theorem 1. and is generally called the n-th root of a. the 7^th power of every real number is positive. and from the given equation Jb is rational. It follows that ya. we therefore define the principal n-th root. Thus both a and n are defined by the same classification of rationals.

and consequently a or 6~0. we obtain an identity of the form . Ex. which is not the case. and. so that */p and *Jq would be like surds . therefore c*p 2 = a 2 6 2 = a 3 c.b~p. which is not the case. \ c / Hence c=0. p-*Jq is also a root off(x) = 0. and therefore a ~ and c = 0. c. q are rationals and that q is not a perfect square. 2. 7/6 = then a -{. then b*Jp + CsJq~Q. 6=0. # are rationals and p is not a perfe ct cube. Multiplying the given equation by %Jp we have cp +a?Jp + 6Ap 2 =0. if x = a. 1. Therefore 6=0 and c = 0. c are all zero.On the other hand. 6. 6. then y 6. Therefore a6=0. it follows that b 2 = ac and c 2 p ab. and therefore also a=0 and 6=0. then Vp/\A?~ -c/6. Ex. and if f(x) is divided by (x-p)*-q. By transposing and squaring we can show that 2ab^/p c*q ~a 2 . IRRATIONAL INDICES 209 Theorem 2.) . // a + 6^/p + c^ 2 =0 wAere a. c=0. // a=0. the left-hand side would be irrational and the right-hand side would be rational . 6. If ab^Q. so that ^/p 2 would be rational. c are rationals and Jp t ^/q are unlike su rds. which is impossible. and eliminating the terms containing fy Since ^/p is irrational. then if p + *Jq is a root 0//(#)=0 where f(x) is a polynomial with rational coefficients. if 6^0. Suppose that p. If c^Q we should have p 2 =( .c^q ~ 0. // a + b^/p + c*jq where a. For we have {x-(p + Jq)}{x-(p-Jq)} = (x-p) 2 -q. then a. thena = Q.

Proof. To prove this. 7 hold when a is a positive real number. by Art. choose x. 9. defined in this way. choose a number &>. Let a be a real number greater than unity. . It is easy to show that if a>0. x f belong to classes of rationals satisfying the conditions stated in Art. and lead to the following : Definition.l/n. Then. 7/0<a<l. there exists a single real number such that a* = N. is called its principal value. At present of will stand for its principal value. 7.l)<a h (a x '~ x . (iii) we dan choose x and x' so that a x ' . x' exist such that a x <N<a x '. and then a x ' -a x <a h (a n . however small. and let be an irrational defined by x<g<x' where x. for (i) there is at least one a x and one a x> ' . the index laws a x . we have Since Jq is irrational it follows that Rp -f S = and R = 0. II. At present. Irrational Indices. rationals x. Next. The theorems of Art. Then a* is defined by These conditions define a real number. Consequently = &ndf(x) = Q{(x-p) 2 -q} 9 therefore p-Jq is a root of/(z) = 0. The real positive value of a*. Suppose that no rational x exists such that a x = N and first let n>l. (7). we can find a positive integer n such that l a n -l</a h . Theorem. If a and N are positive real numbers and a^l. We define 1* as 1. Finally. Putting x=p + Jq.a x < where is any assigned positive number. a y = a*+* and (a x ) y = a xy hold for all real values of x and y. Logarithms. where 6 = I/a. 4. ~ NOTE. S are rationals. (ii) every a*<every a x '. to distinguish it from other values which will be found later.where R.x . x' so that x' x<. 18. we are not able to assign a meaning to a* when a is negative. by Ch. the indices being rationals. 210 OPEN AND CLOSED INTERVALS 10. then a x <a h and a x ' -a x = a x (a xl . a* is defined by the classification a x '<af<a x * Or it may be defined as (I/b)*.1).

a x <N<a*' and a*<af<a*' . (iv) If x l9 x 2 are any numbers in the interval (a. so that a*<jy. or the range (a <#<&). for a x <a x '. for if x is supposed to be the greatest. 1 1 . If any positive real number a is chosen as ' base ' and N is any positive real number. The classification therefore defines a real number . we can find x 2 >x ly so that a^ aPK^N -a^ i. 6) open at a and closed at 6. If a<l. 6) or the range a<x<b. 6) and/(x) is a function of x such that f(x l )^f(x 2 ) when x 1 <x 2 . Then (i) no rational escapes classification . Definition. 4. Hence. we say that/(x) increases steadil y through the interval. We say that (i) the set of real numbers (x) which lie between a and 6 forms the (open) interval (a. 6) is often called a point in the interval.Divide the system of rationals into a lower class A and an upper class A' by the following rule : The class A is to contain every rational x such that a x <N. The class A' is to contain every rational x' such that a x '>N. Again. and then a* = N. in which case/(x) is constant in the interval. where e is any positive number. however small. 6) or the range a<x<6. and so a^ = 2V. for every x and every #'. 9. (iii) If a<x<6. DEDEKIND'S THEOREM 211 ) w hen x^x^ then /(a?) is said to increase steadily in the . (ii) The real values of x such that a^x^b form the closed interval (a. for it is assumed that no rational x exists such that a x = N.e. and we write f = log a N. these may be chosen so that x f -x is as small as we like. unless the sign is always that of equality. as in Art. (iii) There is no greatest x . as in the last article. the number given by the relation a*~N is called the logarithm of N to the base a. they may be chosen so that a x ' -a x <. the real numbers x form an interval (a. (ii) Every #<every x' . we find f so that (l/a)* = l/N. Thus both N and a* are defined by the same classification of rationals. and since every rational is an x or an x'. Similarly it can be shown that there is no least x'. Definitions. ' Interval ' generally means ' open interval/ Any number x in the interval (a. a classification which satisfies the conditions of Art.

f(x) decreases steadily. which is not the case. the aggregate of rational numbers is not perfect. it is the greatest number in A l or the least in A^. and just as in the preceding. . it is greater than all the numbers in A l and less than all those in A^. namely. All this is sometimes expressed by saying that the system of real numbers is closed. 1 that if we take any point in a straight line as origin and any length we may choose as unit of length. where e is as small as we like. we obtain a section of the system of real numbers. (1) If a is rational. It follows from Art. 12. These form two classes A l and AI. Between a and /? there are rationals greater than a which therefore belong to A and also to A'.f(x) decreases steadily in the stricter sense. If f(x l )^f(x 2 ) when # T <# 2 . Two cases arise. (v) If a is a real number. (iii) any number in A is less than any number in A' 9 then there is a single real number a. the rational numbers do not form a closed system. Thus the consideration of sections of this kind leads to no further generalisation of our idea of number. the ' neighbourhood of a 9 is the set of real numbers in the interval (a-e. Also. vx(i) Dedekind's Theorem. and if f(x 1 )>f(x^) when x l <. On the other hand. Proof. (ii) If a is irrational. a must belong to A or to A. Similarly if a is the least number in A^. Consider the rationals in A and A.x 2 . For if we take with A all the real numbers x such that x^a and with A' all the real numbers x 1 such that x'^-b. we can show that it is the greatest number in A or the least in A' The theorem just proved is of great importance. (2) // the numbers in the interval (a. or in other words. 212 ARITHMETIC CONTINUUM 13. the section determines a real number a. such that all numbers less than a belong to' A and all numbers greater than a belong to A' . For suppose that there is a number ft in A greater than a. Sections of the System of Real Numbers. a + e). then to any point P in the line there corresponds a real number p. The Continuum. b) are divided into two classes A and A' 9 as in (1). By choosing suitable axioms regarding the characteristics of a straight line we can ensure the truth of the converse of this statement. or the aggregate of real numbers is perfect. Hence j8 must belong to A'. If a is the greatest number in A v it is also the greatest number in A. for a section of the system of rationals does riot always define a rational. which form a section of the system of rationals and define a real number a. it is also the least in A.stricter sense. It shows that any section of the system of real numbers defines a real number. The number a may be regarded as belonging to either of the classes. (ii) every real number belongs to one or other of the classes. // the system of real numbers is divided into two classes A and A' such that (i) each class contains at least one number.

The exact meaning of these statements depends on the particular kind of quantity we are considering. If p is the real number corresponding to the point P of the line. We assume that a magnitude B can be divided into any number (n) of equal parts. we say A<B and B>A. If A and B are incommensurable magnitudes. Hence the system of real numbers is adequate for the measurement of the length of a straight line. the magnitudes A and B are said to be incommensurable. the ratio of A to B (written A : B) is defined as the irrational number determined by the following classification of the entire system of rationals : the Lower Class is to contain every rational a/6 for which bA>aB.' 'A is greater than or less than J5/ where A and B are quantities of the kind. If. if A can be made to coincide with J5. RATIO AND PROPORTION 213 If no integers a and 6 exist such that A = ^B. If A and B are magnitudes of the same kind. and the aggregate of the points of a straight line is called the linear continuum. If A and B are magnitudes of the same kind. B is said to contain C (exactly) n times. we say that p is the measure of the length of OP. A contains the nth part of B exactly m times. This is expressed shortly by writing B^nC or C = ^B. equal velocities. In considering how a magnitude or a quantity of any kind is to be measured. The student should reflect on the meaning of equal angles. RATIO AND PROPORTION NOTE. and C is called the nth part of B. equal forces. 15. etc. we say A = B . we denote this by writing A B. if A~j BJ. Equality and Inequality of Concrete Magnitudes. small letters will denote positive integers. let A and B denote segments of straight lines . Ratio of Two Incommensurables. Definition. the magnitudes A and B are said to be commensurable. . no matter how small A may be compared with 5. Denoting any one of these parts by C. In Arts. The aggregate of real numbers is called the arithmetic continuum. 14-16 capital letters will denote concrete magnitudes and not num bers . This is an instance of what is known as a continuous magnitude. and if integers a and b exist such that A contains the fcth part of B exactly a times (that is. 14. at the outset we must define what is meant by saying that 'A is equal to J5. For instance. 16. equal quantities of heat. an integer m can always be found such that mA>B or ^B<A. Ratio.that to every real number p there corresponds a point P in the line. if A can be made to coincide with a part of B. Thus the correspondence between the points of a straight line and the real numbers is complete. and the ratio of A to B (written A : B) is defined as the number a/b. Axiom of Archimedes.

or *Jx. and *Jxy = */ab. there is no rational a/b for which bA = aB : (2) every a/6 is less than any a'/6' . for all positive integral values of m and n. B. Let r6 1= =6 2 and (r-fl)a 1 -a 2 . and therefore rb l A>(r + 1)a 1 B. Va6 are rational. y=b or x b. then b^A>o 2 B and a^-a^r-f l)lrb{> ajb l9 so that ajbi is not the greatest a/6. Similarly it can be shown that there is no least a'/6' Hence the classification defines an irrational number. C. unless *fxy. Since y is prime to x. for a is prime to 6. Theorem. Now by the axiom of Archimedes. D are said to be in proportion. Then akx n .] 2. where a/6 is a positive fraction in its lowest terms unless a and 6 are perfect nth powers . y=a. x + y=a + b. then by hypothesis mC^nD and A/B = C/D = n/m. then by hypothesis b l A>a l B.j = . Let b = ky n . B. C. */b are all rational or all like surds. n exist such that mA = nB. Therefore a~x n and by n . D are magnitudes of the same kind and A:B = C:D. then A. If no such rationals m. then A. Ja. the ratio of A to B is equal to that of C to D if mA> =oi<nB according as mC>=of<nD. y n must be a divisor of 6. n exist.the Upper Class is to contain every rational a'/b' for which b'A<a'B. * This is Euclid's definition of the equality of two ratios. If Jx + *Jy = */a + >Jb 9 then either xa. D are four magnitudes of the same kind. CjD are irrationals defined by the same classification and are therefore equal. by hypothesis. B and also C. [Squaring. for suppose ajb^ to be the greatest.] . where k is a whole number. because (1) every rational falls into one or other of the two classes. If A. This classification defines an irrational number . C. Therefore an integer r can be found such that r(6 1 ^-a 1 JS)>a 1 J5. Then bx n =ay n . B. If A. for.* Proof. and the ratios A/B. a and b are perfect nth powers. /x \** a [Suppose that ( .where x/y is a positive fraction in its lowest terms. however great. Show that no rational exists whose nth power is equal to ajb. D are incommensurables. 214 SQUARE AND OTHER ROOTS EXERCISE XXII IRRATIONALS 1. Hence Jk = l. a multiple of a magnitude (however small the magnitude may be) can be found which exceeds any magnitude of the same kind. for ^B<A<^B : (3) there is no greatest a/6 . *Jy. that is to say. If rationals m.

ac=bdp.*. If x + $y=a + $/b where x. [We have Hence by Art. (iii) 4/2 + 4/3. 2. where a. and eliminate \^3 and V2 in turn : for the second part. c.V6 + 3) (V2 -f N/3 + a:). [Put z= a. c are all zero. i/ 8 = 110 + 18(4/2 + (iii) if z=4/2 + 4/3. since pjq is not a cube. [Proceed as in Ex. a. and (2 8 -5) 8 = RATIONAL COEFFICIENTS 215 9. If all the solutions of . b are not perfect cubes. then x=a and y 6.] 5. y. Show that *J2 and v/3 are cubic functions of J2 -f *J3 with rational coefficients. 1. -a. Find the equation of lowest degree with rational coefficients of which one root is (ii) &2 + 3^/4. find t/ 8 . show that a. c*=dp. 6. bc=ad. and show that this is equal to y + 5 when x = 1. 6. etc. Find the condition that rationals x. $y=mg/b 2 where m is rational.] 8. if c ^0 and a ^0. Now use Art. and show that x \(a dbVa a -6).3. .2. 2. 8. then p=c*. Show that 2 8 H-y~6 + 32\/6y = . If a$lp* + b$]pq + c$q*Q and neither of j? 2 ^. (ii) if y=#2 + 3#4. [(i) If * = N^ + %/S L a: 2 -l=2\a. j?^ 2 is a perfect cube. Ex. If a + b*Jp + c*Jq + d*JpqQ where *Jp 9 *Jq are unlike surds. _ _ [Show that C*q*-a*p 2 -b 3 pq = 3db(ap\/pq z + bq\/p 2 q). b are given rationals and 6 is not a perfect square.] 6. find the value of the product ( \/2 . b are rational and y. and that the expressi on in brackets is not zero. and consequently. let y = s/2 + \^3 . and that ^2-^/6 + 3 is the ratio of two linear functions of *J2 -f \/3 [For the first part. Also find x and y in terms of a and b.. y may exist such that ^fa + ^/b = *Jx + *Jy. 8. Ex. then a. 8 =5 + 34/6. .] 4.] 7. d are all zero.

yb. Hence show that aiaa . it is supposed that s n <l. Prov e that ~ . various methods of dealing with inequalities are explained. or I -ab and 1 -cry are the squares of rationals. show that a^. and continuing thus. {a 2 . x. b. (o6 / ) i -4(aA / )(A6 / ) f whence the second result follows. X If X and 7 are not zero. a a j3 8 are the roots of z 2 { (oft') f . j8 2 ). A.2> . we can prove that . then +s n ). . Next observe that the given equation is unaltered by the interchange of x and a.. are the roots of y . a'x 2 + Zh'xy + b'y* = 1 are rational and a. Weierstrass' Inequalities.as 2 + 2hxy + 6t/ 2 = 1.6') (aA')} + (a a') (6 . where (ab') =ab' .(a l + a 2 ) + a x a 2 > \-(a l + a 2 ).y-&==r.a'A) (W . in the last inequality.a') ( A6') .a'6) 2 . / Also. a'. If a.ft) ( . h' 9 b' are rational. where.a r &. 6.] CHAPTEK XIV INEQUALITIES IN this chapter.i.6') = 0.(6 . y and 6.a x ) (1 . Many of the results are of fundamental importance. ( . a n are positive numbers less than 1 whose sum is denoted by s n .4 (aA' .. 1.a 2 ) = 1 . .] 10. . and show that 1 -ab is a perfect square.h'b) are the squares of rationals. y are rationals such that prove that eiJAer a. (aY -bX)* + 4XY=Q. Ij a v a 2 .4 (a*') ( A6')} + 2z { (a .^2) be the solutions.'. [Let (a lf j8i). solve for ~ ..(a -a') (6 -&') and (oft' . then (h -A') 2 . -s n ). For (1 . a. [Put*-a = -S:.

that is. Ex. (1 -)<!/(! -*) 2. (1 H-o n )<l/(l -00(1 -o. 1 -a 1 <l/(l +%). we have l/(c + a-6) + l/(a + 6-c)>2/a. Similarly. For (a + 6 + c) (be + ca 4. l/(6 + c . Therefore j&e.. For since 6 (6 -f x) is positive. 2. x are positive.. 3(n-2) .a) + l/ (c -f a . c are positive and not all equal..) . we have (1 -a. for 0<a 1 <l.. Show that [n>w 2 . We have (|^) 8==1 n . 6. // a. then (a + b + c) (be + ca + ab) >9abc. and the result is obtained by addition. 4.. c are positive numbers any two of which are together greater th an the third. 1. 1.c) + 1/(6 -f c . 6. if a(a + 6 -c-f c-f a-6)>2(c~f-a ~6)(a + 6-c) or o 2 >a 2 -(6-c) 2 . (1 +a n )<l/(l +* n ).(l In the same way Again. and r(n~r + l)>n if r 2 -r(n + l) + n<0 .. and if s n <l.. 6. according as ax^bx or a $6. which is the case.6). 2/c respectively . that is if (r-l)(r ~n)<0 or if l<r<n. n Ex.. n . Using these and similar inequalities.ab) . since a(c + a-b)(a + b -c)>0. l/(a + 6 . Ex. A special arrangement of terms or factors is sometimes useful. according as a$b. are greater than 2/6. 7/a>0 and 6>0.) .... (1 + 0l )(l +a z ) . 1. 2(n-l) . 2.9a6c TYPICAL METHODS 217 3. Many inequalities depend on the fact that the square of a real number is positive. then (a + x)/(b + x)^a/b. then (a + b)>*Jab. .)(l -o. r(n-r + l) . and 1 +a x <l/(l -a^. (a -f x)/( b + x)^ a/b according as ab + bx^ab + ax. If a.. (1 -a n )<l/(l + 0l )(l +a 2 ) .. then Now.(l-o 1 )(l-o 8 ).a).. // a.

then it is easily seen that Ea n (a .A + 2 A + A 2 C.. For let A = a r 2 .a n (a-6)(a-c)=Z(a-6)(a-c)=27a 2 -.c) = -r. . also c n (c-a)(ctherefore a n (a-6)(a-c)+c n (c-a)(c-6)^6 n (a~6)(6-c). 6.. Therefore ^<r and *^ . 1 * 8 (2*-D AI /o ix 4 6 Also (2* + !)^-... there is no loss of generality in assuming that Ex. values of A. . -f #tA) 2: ^( a i 2 + a 2 2 + . a n ^b n ^c n and a n (a-6)(a-c)>6 n (a-6)(6 -c). we may assume that a^6>c. f 6 2 2 -f ... . First let w>0. a 2 . so that m^O. c are all positive and n^O or n^ . then for all Z(a r + A6 r ) 2 . = a n /b n . . 218 INEQUALITIES INVOLVING POWERS 6.-276c -l. . < 2:6c and 27a 2 . .. b n are two sets of ((!&! -f a 2 b 2 + . let n= -w-1. 2. a n and b l9 6 2 . Prove that ^ . real numbers. .6) (a . c. B = 2a r b r . 6 = l/jS.. . >. then a n (a-b)<a-c)+b n (b-a)(b-c)+c n (c-a)(c-b)^0. A. -f 6 n 2 ).j8) (a . C = 2b r z . 1. then -f a n 2 ) (6 X 2 .-. the sign of equality occurring only when a^Jb^ = a^jb^ = . Hence.y) ^ by the prec eding. . //!..1. On account of symmetry. ' then r.A . k. whence the result in question. " // a. 1. Let a = l/a. If both sides of an inequality are symmetric functions of a y 6. c = l/y.-. .' 5. // n=0.2?a m (a . .

. First suppose that x~p. If x and y have opposite signs. then (a r x . then a r x -a s x and a r v -a/ are both positive. t hen according as x and y have the same or opposite signs.. Hence (a r x -a 8 x )(a r y -a 8 v )^0. not all equal to one another. Therefore nEa x + v > Sa x +v + Safa* = Za r x . then 27a r +y++.. not all equalities. 2.a/)<0.a 2 ^i) 2 7. If a 19 a 2 . not all equal to one another.a.1) squares of the form (a^g . Let r. The theorem also follows from the identity where S is the sum of the \n(n . The inequality in Art. are all positive or all negative. we must have a r -h \b r = for every r. y are positive rationals.Therefore ^4-2AJ5+ A 2 C^O. s be any two of the numbers 1. If a is any positive number except 1. and consequently a x + v -h a*+ v ^*a r *a 8 v + a s x a r y ... and < must be substituted for > in the preceding. a n are n positive numbers.* then **. y = q where p. both negative or zero. 7 can be written n n n Hence it follows that if a^ a 2 . There are 2^(^-1) relations of this sort.y. Ea ^ Zgv ^ Zaf n n n n FUNDAMENTAL INEQUALITIES 219 9.. 2a r v . and so Hence the roots of A + 2XB + A 2 C = are equal and B 2 = Otherwise A -f 2A$ + A 2 C>0 for all values of A. n.1 of them and a r x a s v in only one. and x. a n are n positive numbers. . Hence by addition (n.*) (a r y . 8. q are positive integers. First suppose that x and y have the same sign... . and x.z. and therefore B 2 <AC. Also a r x + v occurs in n .\)Za r x + v >Z:a r x a s y where s^r.if x>y... . If we have (a . If the sign of equality occurs.. _ l)/p _ (aP _i _ V)/(p _ 1} .

if for all positive and * It can be shown that this is true when x. Further. I. 37-39.TT~ > -. is > .TT~~ 9 that. putting i i Ifx for x. y = q/d where p. ~ 1 Ex. we may take x=p/d.1) =a n . Hence. the truth of one of the inequalities involves that of the other.1 )>0.l)/(p . v are positive real numbers.1 .= (a .(a p ~ l 4-a*~ 2 + .l)}/p(p .1).1) > a-1.b x <x6 x ~ 1 (a ~ 6) . i i __ji i If w>l.We have to show that x .then n(a n -l)>l .1 )} and each term of this is positive or negative according as a. as a. n(a n .a~ 1 )>l . Ifa>l and n^l. y are fractions.) .. (a-l) 2 >0. the expression in the large brackets can be written { (a*. It has been shown that if a>0 and ^1. If one or both of cc. 4.3 ) + .^\.(a** 1 .x ~l )/x increases. d are positive integers and p>q. 10. If a and b are positive and unequal and x is any rational number except 1. (a p l)/p decreases as j> decreases. A . q.1 . in either of the cases. for 6>0 and NOTE.e... Proof. and therefore (a-l)/p>(a<*-l)/q if p>q. . we conclude that x(a x . (a p -l)/p. then ^-i (a 6)>a * _ 6 x >x6 x-i (a _ 6)f unless <x < 1 . i. XXIII.a*~ 2 ) + (a*. + (a~ l . that is.n(l-a n)> a ^(l . Now. P/ d ?/ d P 9 i where b = a d . Thus. passes through positive ration al values. First observe that. it follows that (a. Putting I/a for a.1) decreases and x(l ~ a x ) increases as x increases. (a x ~l)/x increases with x. (See Ex . in which case xa x ~ l (a-b) <a x . in either case.a. ---.l){pa~ l . If n = \ 9 the inequality becomes a -1>1 --. This follows from the preceding.1 . and consequently (l-a-*)/* decreases as x increases. or a (a.a*. .

n are positive and m<n. 1 1 . . the second inequalities in Art.I $x(k . Thus it is only necessary to prove one inequality in each case.a)>b x .b x ^xb x ^ l (a . Or again.b) ^a x . that is according as -y(k~l)^k-k y + l .1 . Writing n for x and 1 -f x for .. (l-x) n <l~nx if 0<n<l. changing the sign of x. Let a/b = k. 2. and x> -1 and 7^0.l (a-b)>a x -bf*.b) according as k x . or t/^0. Ex. .1 5 k (y + 1 ) . hence the inequality holds if 1 -mx>l/(l +nx). or according as (k x . it follows that.y y then xa x ~ l (a . (1 -x) n >l -nx if n<0 or >1.l (k-l)>k-v-I.l)/xg(A . a x -l<x(a-l) if 0<z<l. 9.a x '. then (i) if x is positive a x . if n is any rational except 1. or if (1 +7ix)(l ~mx)>l.1)/1. or if 0<x<(n -m)/mn.CJU 220 FUNDAMENTAL INEQUALITIES unequal values of a and 6. or according as k v + l . + a n where a v a 2 .1). and so xa*.b x accord ing as -yk~ v . or if x{(n-m) -mnx}>0.. 1. JAen (l+x) n >l+nx if n<0 or >1. 10 become a x -I>x(a~I) if x<0 or >1. that is according as xgl . This completes the proof of the theorem. // s n = a 1 + a 2 + .) (ii) if x is negative and equal to . it is possible to find positive values of x such that (l+x)< For this inequality holds if x>0 and Now (1 + x). 9) according as y + l^l. that is (by Art. so that xb*~ l (b . (l+x) n <l-fnx if 0<n<l. Ex. // m. xa x ~ l (a-~b)>a x -b x . When 6 = 1. (See Art. and therefore a x -b x >xb x ~ l (a -b).m >l -mx. we may interchange a and 6. if x<l and ^0. or according as (*+'-l)/(y + 1)2 (*-!)/!.y . Now y>0. a n are all positive t then . .P B.

then na>a + b + c + . Again..5 *_ HH-Un-^^-j* ^(l+aja+o. l ..... r=n4..... c. 10). therefore therefore u n . +k>nb and a>A>b... ... then A is neither the greatest nor the least of the set.... By repeating the process we can therefore obtain a set of n positive numbers containing two A's.... &..... 6. A* . ... but a greater geometric mean. 6 in (A).. c. 12. Now M! 0.. Let &' =a + 6 ....). 13. 6. We have Thus A + b' =a -f b and Ab'>ab...... is greater than their geometric mean.... For let A he the arithmetic mean and G the geometric mean of the n positive numbers a.. Arithmetic and Geometric Means. (A) Suppose that a^any other number of the set and that fc^any other number.....k. T/?e arithmetic mean of n positive numbers... etc..-h) and G= n J (abc . d.b'.... &)...-4 so that ?/">0...d..... Theorem.c... 4.. ARITHMETIC AND GEOMETRIC MEANS 221 Now s n r -^ n Li> ra n ^ n r ^T l 1 > when r>l (Art... then if n>l. therefore w n >0 for n=2. (B) obtained by substituting A.... but a greater geometric mean. 6. this becomes an equality.u n _ l >a n u n ^ and n >(l +a n )u n _ 1 >u n .... with the same arithmetic mean as before. (1) If a... if the numbers in (B) are not all equal. k are n numbers and A^ -(a + 6 + c+..Denote the left-hand side by u n . Hence the numbers of the set (B) have the same arithmetic mean as those of (A). n then ^4 is called the arithmetic mean and G the geometric mean of a. b' for a. and consider the set A.... ... which are not all equal to one another.. and when r = l. c. 3.. ..... .

i abc ... a n are positive and (n .3.. a 3 . Thus for the numbers 10. By the preceding..'.1)} -r/ 2 . // a x . 5.5. -f a . n + r The right-hand side is equal to A n+r . 5 . .. k are equal. 1. 7. Taking the set (A). 5. the geometric mean increasing from left to right. 7..3 .. and therefore G n A r ^A n + r .. J&r.. k y A 9 A 9 A 9 . 222 FUNDAMENTAL THEOREM Alternative proof..... 5..1). +2n . 2. (2. so that G^A. 5... .1)* = ^ + a 2 + . we have Proceeding in this way.. and since A is the geometric mean of these equal numbers. 4. The arithmetic mean of the numbers in each set is 5. (2n . 2. AV 2.!)}/> l . 6. 5 .. 5 . 1. U V '/ so that If n is not a power of 2. Shoiv that n n >l . k) > . We have (1 + 3 -f 5 + .. (2-l) and n n >l . 3 . we shall finally arrive at a set of n numbers each equal to A of which the geometric mean is greater than (?. 5.. consider the set a 9 b 9 c 9 ..1 = &{! + (2 .Continuing thus. 6. we can show that if n is a power of 2. k^ { (a -f 6+ .. 1. + (2w . the successive sets are 10. then (1 }* abc . Now 1 +3 +5 -f ... 9 where A occurs r times and n-f r is a power of 2. a 2 .5 .1). it follows that A>G.. (2*-l)..3. 2.. 7.. the sign of equality only occurring when all the numbers a. >\ / 1. 5. 6.

. k are n positive numbers which are not all equal to one another. For (* . MAXIMA AND MINIMA 223 .+ . those in the second set being rational..a 2 ) (5 . .8 ) . . . . .a 2 ) + (0 . k . 6..n . b occurs y' times. . a.1) ^ {(5 . k and x. w = w'/g.*M/ k '' ' '" Now the left-hand side is the arithmetic mean and the right-hand side is the geometric mean of a. . and m is any rational except or 1. //a. and then it is sufficient to prove that ax' + fy' + . . .. + & For let x y z w n then 1 -(a m + & m + c m + . 6. so that the result follows by the last theorem.. then n V n / according as m does not or does Iw between and 1.. where x' 9 y r . . .. ... c. // a.....a n )} l /( n ~ l \ There are n inequalities (or equalities) of this kind. where a occurs x' times. and k occurs w' times. 6.. .. (a . y. 14. 13 we may take x^x'lg..... + w m $ n. + (* .. y = y'lg.. Theorem..a 8 ) + . Theorem. then + ku For in Art. a b c k a + 6-f c-f .. and the result follo^a by multiplication..a w ) = a lf therefore a^ (n . . according as x m + y m -f z m + ..:. . w' and g are positive integers. 6.. w are two sets of n positive numbers.15..+ m ^ l-a + + c + . k.

. .. 224 APPLICATIONS Again. +wk m . 6.k and x. ... y. . by Art..... -f w is nc n . the value of x + y + . + w m -n>m(x4-y-h .. c. J according as m does not or does lie between and 1. Hence we draw the following conclusions : Suppose that x.. 11. . 1 6..b. This completes the proof of the theorem. fl and -a m + & m + c m + . In the theorems of Arts. . Application to Questions of Maxima and Minima Val ues.. if ra is any rational except or 1.. w is (c/n) n ..w are two sets ofn rationals and those of the first set are not all equal to one another. by Art.. +w m occurs when x y... from Aft... w is greatest when so that the greatest value of xyz . (2) If xyz.. 13. .If m does not lie between and 1. Theorem .. m being any ratio nal. so that x m + i/ m + .. * This depends on Art. =w?. -f w-n) = Q..+w = c then. i so that the least value of x + y+ . w is least when x = y=.. then: (1) If x + y + z + ... 17. according as m does not or does lie between and 1.. the least or the greatest value of x m +y m + . Ifa.... 3 we conclude that : (3) If x + y + z + . k are equal. the value of xyz . 15 in the same way that Art. 11. . and therefore x m -f y m -f .w = c. the sign > must be replaced by < in all these inequalities. +w n <n.w. -f t^ rw > n. then. .. . If 0<m<l.+ k m )^ l-(a Hence x m + y m + ... 14 depends on Art. 13-15 the inequalities become equalities when all the num'bers a. the value in question being n l ~ m . z. ..y... w are n positive variables and that c is a constant. +w = c. c m ..

since m does not lie between and 1.. that is when x~ y~z-2. . In Art.6) -I.c) is greatest when s -a~s -b~s -c. the value in question being n 1 " 1 /. prove that the area is gre atest when the triangle is equilateral. 2. putting w= .a) (s . Ex. . Ex.c). we have ( \x) (Xx) (^y) (py) (/z//) = 6A 2 /x 3 . if A. b. and let a -f 6 + c ~2s. Ex. Therefore the value of A is greatest when a b = c. Therefore A# + Xx+^y -f/xy+^y is least when A#=/u. the greatest or the least value of x + y + . z which satisfy the condition x + y + z = c. Hence the greatest value of (yz) (zx) (xy) is 2 8 . . y. n~3. it follows that the least value of 3z + 4t/ is Ex. Find the least value of x~* + y~ l + z~ l for positive values of x.. // the sum of the sides of a triangle w given.(4) If x m + y m + . (3). 4.+w occurs when x = y=. subject t o the condition x 2 y 3 6. then zJ 2 s(s . y.a) -f (s . -f w m = c. /x are any constants.1. c 1 /. Since o% 3 = 6. z.?/ = (6A 2 /z 3 ) y . Now (. then according as m does not or does lie between and 1.. 3. the value of (yz)(zx)(xy) is greatest when yzzx^xy.b) (s . Hence the least value of 2Az -f 3pty is 5(6A 2 /i 3 )^. Let a. Find the least value of 3x + 4y for positive values of x and y. subject to the condition Since yz + zx + xy~l2.b) (s . c be the sides of the triangle..a) (s . the le ast value of x~ l + y~ l + z' 1 is SV" 1 or 9/c. Putting 2A = 3 and 3/x 4. 17. If A is the area. and the greatest value of xy z is 8.? . Hence the value of (s . all the numbers concerned are supposed to be positive. EXERCISE XXIII In the inequalities marked with an *.(s . Find the greatest value of xyz for positive valuer of x.=w.c) s a constant. 1.

If x>0 or <-l.c .m'ny + (W + mm' -hnn') 2 = El* .. TYPICAL EXAMPLES 225 3. If a. Z7' 2 . [Let p 3g. . d are in harmonical progression. show that (b 4. cr 2 . If a. . If a>0 and x>y. then (ax n + by n )l(ax n ~ l + by n ~ l ) increases as n increases through integral values.* (i) a(a-b)(a-c)+b(b- .. 6.c) 2 ^ 6c -f ca -f ab.* -r + --. then a + d>b + c. . dareall >1... . then [E(mri . 6. etc..* If x^ty. showthat 6. 2>-l-#> ^>4-3^ be the reciprocals of a. . If J 2 + w 2 -fn 2 = l and J' 2 + m' 2 -f w' 2 ^l.6) 2 -f (a + 6 . 7. jp f. c are any real numbers. (& 2 + c 2 )/(& + c)^(& + c)...+ . then (a l -f a a + . If any two of a.] 4. 9.] 8. 10. -f a n )/(6 1 -f 6 2 + . then [Show that x r + l + x zn ' r ^ I +x 12. show that a x + a~ x >a y + a~ y . If . b. Show that a(x + y)* + bz 2 + cz(x ~-y) cannot be positive for all real values of the quantities involved. etc.] 11. c.. &4-c c+a [For. b n are two sets of numbers and all t hose in the second set are positive.1. . c.* If a. + b n ) lies between the greatest and least of a^/b^ 2 /6 2 . 2. . 6. c?. 6.] m* . then [2(ab -4. a n lb n . c are together greater than the third and x + y + 2=0.. . 5.1)> (a + !)(& + 1). a n and b 19 b 2 .a) 2 -f (c -f a . c.^ a + b + c.

n in succession.. then (p 4..1. Finally let p = l. l)f+i .6 2 (c + a) -f c 2 (a 4. If p>l and r is any rational except zero. + r )/w r4 ' 1 lies between l/(r-fl) and [In Art. [The second inequality is merely another form of the first. then [327a 8 > 27a . 14. lie between and . 6=p-l.. If a 9^6 and #.b).(ii) a 3 -f 6 3 + c 3 + 3a5c ^ a 2 (6 + c) -I. If any two of a. Hence show that (l r + 2 a + 3 r 4. b 9 c are positive and not all equal. If n>0 anda?>l.* a*b + b*c + c*a^3abc. c . .] x n ~x~. fc are n positive numbers. 20. and add. If a. Ella^n* . then (i) Za . If x> I and n is a positive integer. y are any two positive numbers such that a? + y = l.=r4-l. then [If a>6.] 13. [Use Art. 6.. 6. &=/>.] 15. 9. then .p r+i^( r + l)pr >-p r + l where the upper or lower signs are to be taken according as r does not. 10. First put 0=73 + !..1)1 (x 22. or does. c are together greater than the third.a'&nZa*. and next a=|9.] 16. then (^ \* -l<x(?-l . If a.] 226 HARMONIC SERIES 19. JEb" and Za(6-c) 2 >0. (ii) Za 3 . 3. 21. (# n . then A A ni > ~ . 2.n nx 17.. let a. El'..

r/d where p. a n <(? n . provided that all the factors are positive. . .. If a is any positive number except 1. and a. 30.. r.. Consider the A. If all the factors are positive.. !tl . (ii) If a x .+a n <nA and H n <a l a 2 . geometric and harmonic means of a^ and a n . 6.Q.x n =a n where a is a constant. !L^ and ?. z are positive variables such t hat x + y + z is constant. (ii) The least value of (x l -f ifc) (a? a + k) .] 31.z 1. y. whe re a x occurs (y -z) times and a? occurs (x y) times. z be integers and let x> y> z (Art. a n < A n . c are positive rationals and x. y..y)>0. y.. 29. then (i) abc^TI(b+c-a) . y. y. of the set of numbers a x . a 2 .. . . 6. 2 are positive and x + y-f.. denote them by p/d. a 2 . \. 25.23. a 2 .z) +a*(z . * .. a n be a sequence of positive numbers and let A.. and the O. If *=!+. Deduce the theorem of Art.. . ?.... and x. a n is a harmonical progression. then 8xt/z <( 1 -z)(l -i/)(l -z)<2 V [Observe that 27(1 -a:) = 2. The greatest value of xyz(d -ax -by -cz) is <Z 4 /4 4 o&c. q/d. then (i) If ttj. then (i) The sum of the products of every r of the x's > CJ?a r .. H resp ectively be the arithmetic..M. [First let x. . (ii) abcd^ TI(b + c+d -2a). a x .M. 26. If of. . d are given positive numbers.. 30.. q. This case can then be at once deduced from the preceding. .] ^04: 71 1ZO 71 28. a n is an arithmetical progression. a x .. show that x a y b sf has its greatest value when x/a y/b zlc.x) +a*(x .k) n . G n <a 1 a 2 .. IfzjX&t . If x. 9 from Ex. a 2 . ?. Let a l9 a 2 . ... c.] 27... z are fractions. d are inte gers and d is positive. z are rationals no two of wh ich are equal and which may be positive or negative.. * 24.(x n + k) is (a 4. nH<a 1 + a 2 + .+ -+. then a x (y . a 2 . show that A o n [Consider the sequences \. 5).+_ and n>2. If x.

then (a l + a 2 + .b n ) ^n(a 1 b 1 +a 2 b 2 + .1)1 (p ~ 1).l)lp -(a*. 9. Use the relation (a r ~a n )(b r ~b n )^0 to s how that w n >w. 34. a -I) 2 when p>q. q are positive integers... -f 1)}. b r . $(b+c+d) fora.] -f- IMPORTANT GENERALISATIONS 227 333 16 a-f b + c + d' [In Art..1 ... 15 put w = ?. 16 to show that a" 1 *) + b~ l c + cr l a a + b + c 35.. . put w= -1.. . .. n4. 15.] 37. a-a^. . a 2 .+a n *)< [In Art. OK and b l9 b 2 ... . .* [Use Ex. then by Art. If !.. etc..+a n )(b l + b 2 + .. show tfiat a* -I a a ~l p-q.* Use Art. p q 2 [Let u^(a? . If a> 1 and p. If a l9 a 2 . a n are positive and p>q then (a 1 * + aJ + .n _!..2a r . b n are two sets of positive numbers a rranged in descending order. etc. +a n b n ) .. 34. .32. [Let u n ~na r b r ..] 36. .

we find that i and by Art. 7. y such that x'>x. when *>!/.y 7 > x .1)> 1 -a. 37. / so that | -c' | <l(x-t/)(l -a" 1 ) 2 . using Ex. we can make 1 -f as nearly equal to 1 as we like. When n is large.p ~. by making n large n l enough. See Ch. [Let x' 9 y' be rational approximations to x. 11 on p. 9. // a> 1 ami x. then (a*-l)lx>(av-l)ly if x>y.] 39. then. p(p-L) a p -l a q ~l p q . Ex.. Let /( # ) (a x . If a>l and x. 1. d are positive integers. XIII. y' <y. y are positive real numbers. 15) 1 Limit of a Function of the Positive Integral Variable n. and so /(a?) -f(y)>0. further.] CHAPTER XV SEQUENCES AND LIMITS (Continued from Art. . d(a d . and there fore x' ." 38. q.1/.is nearly equal to I . y are positive rationale . Let a lld b . and suppose that then /(*) We can choose x'. 1 4.and . then x y 2 [Put x pfd.1 .1 )/. yqjd where p.

oo . (1) If corresponding to any positive number that we may choose.This is roughly what is meant when we say that 1 is the limit of 1 + as n tends to infinity.u n . no matter how great.oo . \u n \<M. If w n ->oo and v n = . In this explanation. the meaning of ' large enough/ ' as nearly equal to 1 as we like. no matter how small. and that it converges to the limit L If w n ->oo or . nor to + oo . the sequence (u n ) is said to bo divergent * THEOREMS ON LIMITS 229 If u n does not tend to a limit. then I is called the limit ofu n as n tends to infini ty. If u n is a function of the positive integral variable n> the i limit of u n as n tends to infinity ' is usually defined as follows : Definitions. nor to . or simply lim u n = L n->oc We also say that u n tends to I as n tends to infinity . u n >M. If u n -->l we say that the sequence (u n ) is convergent. and express this by u n -~>l as n-+vo . there is a positive integer m y such that for every integer n greater than or equal to m. (2) If.' is by no means clear. then u n is said to tend to infinity as n tends to infinity) and we write w n > oo or lim u n = oo . we say that v n tends to minus infinity (-<#). for every positive integer n. there is a positive integer m such that for every integer n greater than or equal to m where I is a fixed number. corresponding to any positive number M that we may choose. This is expressed by writing lim u n = /. it is said to oscillate and the sequence (u n ) is called oscillatory. . If u n oscillates and a positive number M can be found such that.

. // u n = ( . Thus u n oscillates finitely.1. (2.j . II. Ex. and so a n oscillates i nfinitely.. u zn -> 1 and u 2n .oo according as n is even or odd. It is usual to join points representing consecutive terms. 1.1. a n = 1 for every n. (3..then u n is said to oscillate finitely. 2. if|a|<l. 18. represent graphically the first Jew terms of the sequence (u n ). if a=-l. if a . u 2 ).l) n I 1 + . if a< . u n oscillates infinitely. How does the function a n behave as n-> oo ? Referring to Ch. This is done by plotting points whose coordinates are (1. u } ). and examine the diameter of the sequence as n-> oo . Otherwise. it will be seen that if a> 1. . a n -> oo or .!-> . (5). Ex. It is often useful to represent the first few terms of a sequence (u n ) graphically. a^-^O. As ?i~> oo . a n ~> oo . a n =r 1 and a n oscillates finitely. u 3 ). 2 I 7/2 7/4 1 / \ / .1 .

34. (1) Ifu n -*l and v n -*l\ . 2. Fundamental Theorems on Limits.\ o 1 / 2 \ y 4 \ 5 -1 \J * f ?t 3 ^5 \il l FIG.

then each of these inequalities will hold for n^m. Hence also | u n v n . (ii) u n -v n ->l-l'. by the definition of a limit. By the definition of a limit we can choose m so that provided only that n^m. (i) and (ii). Therefore u n v n -+ll'. (iv) l/u n ->l/l. Proof. and then n -| < |ifX' Now by the definition of a limit we can choose m so that provided only that n>m. unless I' = 0. no matter how small. Then Choose any positive number L greater than | V + j8 1. and for such values of n we have and | Wn _ Vn _ (J _ J') | sg | Mn Hence. Choose a positive number .L.then (i) u n + v n ->l + l'. then \u n v n -ll'\<\fll\+\\. unless 1 = 0. (iii) w n ^ n ->Zr. If we denote the greater of m v m 2 by m. then A 1 . By the definition of a limit we can choose MJ and w 2 so that \u n -l\<. Hence also .l for n^m l and \v n -l'\<\t for n^m%. (v) u n \v n ->l\l' .a // Z is ftoJ zero we can choose a positive number L less than | 1 + a |. (iv) Let tt n = Z+a. u + v->l + l' and 230 MONOTONE SEQUENCES (iii) Let u n = I + a and v n = T + j9.IV \ <.

or less than. We shall prove that the sequence converges to A as a limit. u n is said to decrease steadily. If w n +i<w w for all values of n. therefore | f> w -J|<| v n -u n \ +\ u n -l \.i_ * 11 < c. // every term of a monotone increasing sequence (u n ) is less than a fixed number k. <Aen v n ->l. EXISTENCE OF A LIMIT 231 Theorem 1. 3. Hence the classification defines a real number A. Now we can choose m so that for n^m both |w n -Z|<ic and |^ n -w n |<i. for some value of n (and therefore for all greater values). For Vn -Z = ( Vn . and (w n ) is called a monotone decreasing sequence. then the sequence converges to a limit equal to. and (u n ) is called a monotone increasing sequence. k. w n >a. and therefore > T .1 1 < c. u n is said to increase steadily. w Z (v) This follows from (iii) and (iv). for when r (2) Ifu n ->l and v n ~w n ->0. and therefore v n -> Z. | v n . Monotone Sequences. Divide the system of rational numbers into two classes as follows : The lower class is to contain every rational a such that. Proof.Wn ) + (w n -Z). . Hence for such values of n. If w n +i^ M for a11 values of n. No rational escapes classification. and every a is less than any a'. The upper class is to contain every rational a! such that u n <a' for all values of n.

1) where #>0. Again k^A. then y>\ and 4. This can be proved by an argument similar to the preceding. therefore 0<A-u n < if Hence limu = A. Theorems. u n ~0 for every n. and is exceeded by some term u m of the sequence and by all succeeding terms . show that u n tends to a limit as n-^& . choose a positive number e. then (v n ) is an increasing sequence of which every term is less than . For if u n does not tend to + oo . 232 EXPONENTIAL INEQUALITIES i Ex. and let a be a rational such that A-<a<A. This is impossible. Then a belongs to the lower class. If 0<#<1. then the sequence converges to a limit equal to or greater than k. let x~l/y. If x>I. no matter how small. or thus : Let v n = -u n . and by XIV.k. for such a rational would belong to the upper class. u n would exceed every rational which lies between u n and A. and so lim u n ~0. . // u n --ii(x n . then u n tends to a limit or to -oo . Hence u n tends to a limit. Hence (u n ) converges to a limit greater than k. 9. u n >0. Next. Theorem 2. For if u n >A. thus It follows that if u n increases steadily as n->oo .First observe that u n cannot exceed A. u n decreases as n increases. a positive number k exists such that u n <k for all values of n. If every term of a monotone decreasing sequence (u n ) is greater than a fixed number k. Hence also if u n decreases steadily as n->oo . which we shall denote by /(x).k. Therefore (v n ) converges to a limit less than . for otherwise k would be exceeded by some term of the sequence. 1. then ?/ n ->oo . If x 1. then u n tends to a limit or to -f oo . (1) ///or all values of n* u n is positive and u n where k is a constant greater than unity. as ?i~> x .

f 1 +. it' n ^ m where M is a fixed number. (1) If n is a positive integer. positive or negative. Exponential Inequalities and Limits. w n ->oo . after a certain stape. (2) If for all values of n. . Ex. therefore u n -+co . by the Binomial theorem. 11. The proof is similar to that of Theorem (1).. by XIV.For u n >ku n _{>k*u n _2 . then u n -0. sJww that (i) * p ->oo if x>l . then 1 \ m / 1 \ n / 1+ ) > 1 + . and then hy the definition of a limit we can find m such that u n+1 /u n ^>k for n^-m. and the results follow from Theorem (3).=( -_ ) 'X>x as rt-> oo . w n ->0. i. to n -f 1 terms 12 '(3 * Or. by Theorem (2). and u n +Q For if />! we can choose k so that >&>!. x n u / n \v If u n = -^ then . >k n ~ l u 1 . If l<] 9 we can choose k so that l<k<I. u n is positive and u n i<Jcu n where k is a positiv e constant less than unity.and 1w/ \ n/ \ m.e.. (ii) ~->0 if |*|<1. and then we can find m so that u n+l ju n <k.) <3. 1. EXPONENTIAL LIMITS 233 (2) // m and n are positive integers and m>n. For.. If p is a given integer. (3) // u n is positive and lim u n . n /u n = l t then w n ->o> ?//>!. Hence. for n>w. Hence by Theorem (1). For since m/n>\. Now &>1. 5. ..

fu ' \ (f\ \~ n 1 (ii) Again..m I \n ml f 1 \ m / l\ n 1 + . where e is a fixed number less than 3.(by XIV. m/ n m \ m/ \ n/ n f 1\* n 1 Moreover. V n 2 / n* therefore 0<w ' . and is one of the most important numbers in mathematics.u n = u n ( -*. n n n n therefore u n '-u n -+Q and u n '-+e. Wfl ' . and the limit is the same for both.->0. 11). ' 7 ' J \ n/ m n Hence l-<l..1 ) = u n \ f 1 . 1 -.. therefore by XIV. It should be noticed that u n <e<u n f . Now . It can be shown that 6-2-7182818284 . Therefore u n -e as n->co .. \u n / {\ n / j Nowifn>l. 0<w/m<l.u n < <-. Ifu n = (l+-) andu n ' = (l--) .) <1 .1 \ . u 71 tends to a finite limit. For (i) it has been shown that u n increases with n and that u n <3 for every n. .and n/ (3) The number e.thenasn-+oo \ n/ \ n ' through positive integral values. This limit is denoted by e. 11.) > 1 + . so also does u n ' . l<l--<l+n. and therefore ( 1 + 1 >l-f-j.~ 2 ) .

..-f . .+ . . and that u n tends to a limit as w-*oc . [The limit is 1.. (v) n 2 -f( n x n 2. -f vr<n.>< as ?i->oo.. .-4.fl I A* 5.. If ||<l. If u n ~ -H ---.+ . (ii) n + (-l) n . J J n 1 /I 1\ /I 1 1 1\ [Group the terms thus: -+ ^+. (iii) n (iv) 2 + (~l) n -.n * 2 o 4.. Prove that -> as n-> QO . Show that -n< . and . n 234 EXAMPLES ON LIMITS 3.4 ^ o 4 /i Hence show that 1 H. 7. . if | x \ > 1.] 1 .1 otherwise...) + ^-4. prove that f<t* n <l.. and say how each behaves as tt~>oo . then ^. H --. State how the following functions behave as n-> (i) 2 + (-l) n . --* . Represent graphically the first few terms of the following sequences.i._H ------. (Os.+ 4.o+. Find lim n .EXERCISE XXIV 1.] n->oo % ~~ W 6.) + .-4.-fr-.

Now llli 1 1 x n y n . and lim y n ^ 1 .^ . prove that if z>0 and 1 1 [See Art.5 ..1 =y n (x n .1) -f (y n .. according as x 5 y.. For the last part we have f(xy)-\im n(x n y n . ^ [Show that w n ^<w n . 1. v n ->oo and i< w n i' n < ^v /v I 1 [If />0.] /2" = A/ . hence 2 ^ 2n 1 ^^'s-srn:' " v< 2-^n and 4 6 2 4 2w .''* .8. If _____ n _^_^ tt n ->0.1) =/(a..1). w n *=u n v n and <w n <l. Wallis " has shown that this limit = A/ . 3. 46 2n + 2 1 10. If lim 71(0. 1 l Q T _ 9 . then by Ex.). Ex. 9. then .1) .] ' 7T .N/2/Hrl.. prove that w n tends to a limit which lies between l/\/2 and 1 as n->oo . If w n .

. 4.3 r -f . u 2 . the difference between u m and any subsequent term must be numerically less than e. Now u m+p . .u m \ < | u m + P .u m ). . Represent the terms u v w 2 . lim u n ^\l(r f 1).u m .GENERAL PRINCIPLE 235 11. 1).-. n in succession for ?. The necessity for the condition r~hl>0 in (ii) should be explained. Hence | u m + p . .-u m \< (A) That is to say. . (i) The condition is necessary.. we can find m so that for every positive integer p. .(a + tt) r j/tt f + 1 = l/ (r + l). and | u m ^ -u m \< e. by the points Wj..I \ + \ u m -l\ . 71 \ nj n r + l (ii) If r + l>0.] 13. If Z>1. Corresponding to any positive number that we may choose. on the x-axis. 4. n.. therefore "?i->Z where Z^l. Ex.. called the general principle of convergence. 3.(u m ]1t -l) + (l.. (ii) The condition is sufficient. / 1 r-f-i | show that (i) (r + 1)^ 7J lies between 1 and ( 1 -f . Show that lim {(a + 1 ) r 4. [In Exercise XXIII. If n ^ 3. and may be proved as follows. For if u n >l. \u m + 1( .. all the terms beginning with u m lie in a strip of width 2e. .(a -f 2) r -f . and add. / ] n [For the first part use the inequality ( 1 4 ---I <3. This statement.. for every positive integer p. put 1. General Principle of Convergence. -f n r )/ti r l . prove that n+ \/(n + 1) <s 'n.. But J n /n->oo when J>1 (XV.. If r is any rational except zero and w >4 ( l r -f 2 r -I. 2. .) . Thus s'w decreases steadily.. The necessary and sufficient condition for the convergence of the sequence (w w ) is as follows. for ?*>3 we should have ^n>L and l n /n<l. The condition asserts that m can be found so that all the points. J7. it must be possible to find m so that. so that if the terms of the sequence are represented graphically.-.^-00 6. is of the greatest importance. no matter how small. also C'^>1.] 12. Hence show that */n -> 1 as n -> oo .

for every positive e. Now u mi is within (c^. 36. . This is of length -e. it is required to prove that the sequence (u n ) is convergent. The figure illustrates the case in which the intervals overlap. for every integer w 3 >w 2 .u m | < e for every integer n greater than m.Je. u mi> is within the interval ('U mi -. and this is the limit of the sequence (u n ). lies entirely within (c/ 1? /^).|e.. Proof. such that a n <J<b n for every n. Again. Denote this interval by (a v b^. U mv +). 23G NECESSARY AND SUFFICIENT CONDITION Given that. we can make this segment as short as we like. there is one real number I and one only. (ii) for sufficiently large values of n. NOTE.C. 1. (a 2 . u niz f ie). t --> ^ ns H ~~* ^ ' Hence by Ch. By choosing e small enough. 6 T ). 6 2 )... By repeating the process. and contains u n tor // ^Hi. 6 2 ).. The necessary and sufficient condition for the convergence of . we can find a succession of intervals (a^ b^). Using the given condition. u u is within each of the intervals .>. Thus the whole or part of (w ?Wo . FIG. Q B. there exists a positive integer m such that u n . we can determine m 2 so that. Denote the common part of these intervals by (a 2 . the point u n tends to a limiting position somewhere near to u m .. for every integer w 2 >Wj. (iii) b n -ff tl * . we can determine a positive integer m l so that. XIII. he may omit the proof on the next page. no matter how small. u m ^ is within the interval (u m<i ... Hence it seems reasonable to conclude that as n->oo .lie on a segment of length 2e FlG 35 with its middle point at u m . (a. for W 3 >m 2 . 2 . If the student is satisfied with this reasoning. A. such that (i) each contains all that follow . u m ^ + \t) is within (a v b^) . b r ) .

. BOUNDS OF A SEQUENCE By the condition (A)... is equivalent to it.... 237 and then all the terms w wfl . wf . (1) If there exists a number M such that. and also the function u w are said to be bounded above (or on the right). which is bounded above and below.... no matter how small.../w every positive integer p. or a function w n ... we can find m so that I U^^-Un I <ic. 2 m+z u m+z> He in the interval Therefore the difference between any two terms which follow u m is less than .. for all values of n... Bounds of a Sequence... 7. then (w n ). If there exists a number JV such that.. Corresponding to any positive number e that we may choose... then (B) is possible when n]>m." <e . for all values of n.the sequence (u n ) is usually stated (less simply) as follows.. are said to be bounded below (or on the left). Hence we can find m so that |w n + p -tt n |< if H>w.. Therefore the conditions (A) and (B) are one and the same... the sequence (u n ).. A sequence (u n ). and also w n ... it must be possible to find m so that \ ". To prove this. (B) if >w. The condition (B). we have only to show that if (A) is possible.. is ... .. though apparently more general than (A).

. r / <Aa . or of u n (FiG. 37).said to be bounded. ? /4 1 \ / \ O V . Examples. Here (u n ) has a greatest term. namely w 2 ="2 : ^ has also a least term. namel y % These are called Ae upper and lower bounds of (M W ).2.

FIG.2 \ y 4 V 5 -1 V 7/ 5 -2 -1 2. 37. .

This is impossible. Proof. Also if e is any positive number. This number h is called the upper bound of the sequence (u n ) or of the function u n . 3 8). or of w n . in -+ 1. no term of (u n ) exceeds h. Moreover. Now a belongs to the lower class and is exceeded by some term of (u n ). for u. Hence the classification defines a real number h.1 FIG. we say that 1 is the upper bound of (u n ). // u n is bounded above there exists a number h such that (i) no term of (u n ) exceeds h .1 is the lower bound of (u n ) of u n (FiG. and every a is less than any a'. for in that case it would exceed a rational a' greater than h. The lower class is to contain every rational a such that for some value of n u n ^a. . Divide the system of rationals into two classes. for a' belongs to the upper class. (ii) Infinitely many terms exceed any number less than 1. Hence the number h satisfies the conditions stated above. For similar reasons. we say that . Although there is no greatest term. Hence at least one term of (u n ) exceeds h . both classes exist. (ii) at least one term of (u n ) exceeds any number less than h. The upper class is to contain every rational a' such that for all values of n u n <a. no matter how small. No rational escapes classification. we can find a rational a such that h-<a<h.e. Since u n is bounded above. 38. for the following reasons : (i) No term of (u n ) exceeds 1. 238 UPPER AND LOWER BOUNDS Theorem.

Similarly. infinitely many terms of the sequence lie between h and h-. u% . infinitely many terms are greater than u m . Explain how to select a subsequence which will converge to the upper boun d h If Ui^h. infinitely many terms of the sequence li e beticc.. let a 3 be the lirst term of (u n ) greater than a.e. and ther efore also greater than h . In a similar way we can show that : // u n is bounded below there exists a number I such that (i) no term of (u n ) is less than I . Upper and Lower Limits (of Indetermination) of a Sequence. t/ 3 . u 5 . The number I is called the lower bound of the sequence (u n ) or of the function u n . and at least one term u m is greater than h . For every term is less than h. 1. 2 be the first term of (u n ) greater than v t .. let a.. Since there ia no greatest term. Ex.. . ^ > * >e the upper and lower bounds of the sequences u 19 u 2 .. .. Continuing in this way we form the required s ubsequence ^i a 2 . u 2 . <x 3 . (ii) at least one term of (u n ) is less tJian any number greater than I. according to some definite rule are said to form a subsequence. if (u n ) has no least term.. respectively... (1) Let (u n ) be a hounded sequence and let h l9 h 29 A 3 . . . NOTE. If a 2 ^h. . then it has an upper bound.. The formal proof is loft to the student. u 3 . w 4 . etc..cn I and LIMITS OF INDETERMINATION 239 Terms selected from u l9 u 2 . 2 .This theorem is often stated as follows : // u n is founded above. and l v J 2 . u 39 w 4 .e. .. .. . // (u n ) JLOS no greatest term. 8.

\ / .... . h l ^h 2 .. is an increasing sequence and every term is less than h v Therefore this sequence tends to a limit L less than or equal to h r The numbers H and L are called the upper and lower limits (of indetermination) of the sequence (u n ).If fi l=sUl9 then u v is not less than any term of t/ 2 . ^ 4 .. therefo re h^h^.. Thus. Here h 1 = h 2 =h 3 = .. Hence h l9 h 2 . Similarly A 2 ^/fc 3 . Z 3 . . %. . . in any case. (2) Examples.=H. (Fig... and so h l = h 2 . A 3 . If h^u ly then Aj is the upper bound of u 2 .. A 3 >A 4 . It is usual to write lim u n = // and lim u n ~L. u 3 . is a steadily decreasing sequence and every term is greater than l v Therefore this sequence tends to a limit H greater than or equal to l v Similarly I l9 1 2 . 39. . w 4 .) 1 L //2 / M/4 7/. and so on..

\ o 1 2 \ y *\ 5 . /n s \ .U S 2 ) S2 fl* .

1 / \ / \ l / 2 \ y 4 \ 5 -1 / > 'Uz u s .

e. all the A's are equal. . Here ^ = ^3- 1\ .. Hence in all cases. Therefore every h is a term of (u n ). h n -\ are all terms of u lt w 2 . no matter how great m may be. greater than m. 40. 39. (Fig. . 240 GENERAL PRINCIPLE OF CONVERGENCE (3) It is easy to see that. FIG. FIG. after a certain stage all the terms of (u n ) are less than H + } in other . h^h^u^. (u n ) has no greatest term. that is to say. None of the terms from u m onwards can exceed H. etc.." In this case. after a certain stage. we can find w.'. .-2 U. infinitely many terms of (u n ) are greater than H e . then h n = u n and h v A 2 . That is. etc. L~tim w 8n+1 = -1. but infinitely many terms exceed H . m exists so that 7> /> /> H n m ~ Mrn+l ~ "'m+2 ~ . every term of (u n ) is less than H + . //=lim u 2n ^l. if A n >A w+1 . also. so that h n >h n+l .) =l z ~Uq-. 40. and in all cases // is such that. As w->oo two cases arise : (i) After a certain stage.. (ii) There is no such stage as that described in (i). /.. w n _ 1 ..

u m+l -u m <k(u m -u m ^) . .u n \ < Je. u u where k is a positive constant less than unity. only a finite number of terms exceed II + e. <te n .L H . It follows that. and since 'U mirn >u m .u n ^..9 (u M -u 8 ). We can therefore choose m and n (n> m) so that | H . it is required to prove that (u n ) is convergent. therefore u m+l . . if H = L = l.u n + u m -L + u n . We can at once deduce a proof that the condition in Art. 9. Therefore (u n ) has upper and lower limits // and L.u s }jk s . Theorems. For. after a certain stage. all the terms of (u n ) lie between L-e and // -f . we can find m so that & m <e(l -k)/c. Therefore H = L and (u n ) is convergent. that is between I . Also Hence by addition Choose any positive number . however small. it follows that I tWn . 2 .u m | <c.words. | u n . then the sequence is convergent. only a finite number of terms are less than L-c. the difference between the fixed numbers H and L is less than any positive number c. no matter how small . the sequence is divergent.u m . In the first case. First observe that u n cannot tend to -f co or to oo . (4) General Principle of Convergence. suppose that the conditions hold for IMPORTANT THEOREMS 241 Then if w>s. then the sequence converges to I as a limit. (1) // (u n ) is an increasing sequence and if. Given that m can be found so that \u n -u m \<e for nl>m. 6 is sufficient for convergence.u m <k m c where c = (u s+l . In a similar way we can show that infinitely many terms of u n are less than L + e and. But if u n -u n ^ } . after a certain stage. after a certain stage.and l + e. then. in other words. since &<1. That is to say. \H-L\^\H-u n \+\u n -L\+\u n ..u m \ < 3-6 . \u m .L \ < Je. and since H . all the terms are greater than L~-e.>w n .u m | <c for every n.

In the second case. . then u n n ~+L Because Z>0. i NOTE. 2. n . therefore (u ni /l m ) n >] and ?/ n n >i!. 242 IMPORTANT THEOREMS (3) // .. .9 o/ n awrf w n+1 /?/. -f u n and Urn u n = I. Then we can find m so that and by multiplication. (I -)<u n n <\ j m (t ' V 1 1 Now u m fl m is a fixed number.s n = /. Here w n n -> 1. therefore (Z-) n < J m <(Z + e) n i / \ n 1 / and ' i m ) . The converse of this theorem is not true. i t does not follow that w n4l /w n ->/. n ~>/>0 where I is a fixed i number.Therefore the sequence (u n ) converges. if n>s. 1. it follows that ^u n ^u n . . then Urn . 2. but n n fl /^ n does n ot tend to a limit. in i which the terms are alternately 1 and 2.* x n > oc ^ We can find w so that for any positive c. Since Z-e>0. For consider as a special case the sequence 1. u n+l .. we can choose a positive number e< and as small as we like.<? w = u l + u z + . That is to say.w n >w s ^ t ~ w 6 Now w s+1 u s is a fixed positive quantity. (2) // M M >0 /or a^ m^. . therefore obviously w n ->oo . if ^ n n -> /.

+ a A Let nfy^^a-fay. then where Q = ~(^i+n-ift + 4-a^). Since u n -*l we can find a fixed number k such that | u r \ <k for every r... and then I 8m I < I % | + h/ 2 | 4. 1 1 I I m 7 1 / / so that .m)(l + e) for and Now let m and n lend to oo in such a way that w/m->oo .v < A: and s n n->(>. Now by (3). +j8 n )-^0.. We can also find m so that | a s | < /k and | f3 8 \ < c/A for COMPLEX SEQUENCES 243 . l(a 1 +a 2 + .s m <(n .e)<s n .. n r =i and since a n ->0 and jS n ->0 we can find a fixed number k such that for every s |a f |<* and |ft|<t.. and b n = l> + f3 ni so that a n -~>0. (n . 7i n Thus in the inequality (A). 4. we may suppose that m~ >cc and n>m 2 .m)(l . (4) // P n = a n & 1 +a n _ 1 6 2 + a n _ 2 ' .. (For example.->0.. 3 + --. ?Z Teand we shall prove that $.+a n )->0 and i(j8 1+ j8 a + . therefore s n jn->l. We have | Q n |< ~ T| A-m I .) We shall show that under these circumstances s m /n~>0.Hence by addition. j8 n ->0..I m I < WA'. m/n and 5 m /?i tend to zero and e can betaken as small as we like.

For let z = r(cos 6+ i sin 0). difference. If corresponding to any positive e. however small. (ii) if z = l then z n = l. i. then z n ->0.Hence |a r ]9 n _ r+] |<*| j8 n-r+1 |< for n-r + l>m. therefore z n -> z. IV 10. All this is evident geometrically. 2. | oc r f3 n _ r+l \<e for l^r^n. and (i) if |z|=r<l. XV. continue to hold when the functions are complex. (A) and |r0n-r+l|<*| a r|< * r r>m ( B ) If n>2w. if x n ->x and y n ->y. so that. Again. product and quotient of two functions of n. (3) Limit of z n as n->oo when z is complex. with this^proviso. By definition. where z is a fixed number. then z is called the limit ofz n as n->oo . 7/|z[<l. Otherwise z n does not tend to a limit except when z = l.e. and if | z |> 1 then r n ->oo and z n does not tend to a limit. then r n ->0 and z n ->0. (2) The theorems of Ch. regarding the limits of the sum. for r^m.P n -~>ab. the inequality (A) will hold f^>r r<m + I. First suppose that z n -*z. so that x n +x and //>?/. (1) If z n = x n + iy n where x n and y n are functions of a positive integral variable n. For K-H=-y{(z-z) 2 + (2/--?/) 2 } (A) Therefore | z n z \ ^ | x n x \ and | z n z \ ^ | y n ~ y \ . there is a positive integer m such that \z n -z\<t for n^m. the sequence (z n ) is said to be complex. and then limz n = l. Complex Sequences. l^n" 2 !"^^? therefore |# n -#| and \y n ~y\ tend to zero. n->oo Under these circumstances the sequence (z n ) is said to converge to the limit z. from equation (A) it follows that \z n ~z\-> 0. . for \z n -z\ is the distance between the points z n and z. then z n = r ri (cos n0+ i sin nd). and we write lim z n = z or limz n = z or z n ->z. Therefore |Q n |< e: consequently Q n ->0 and . In order that lim z n z = x + ty it is necessary and sufficient that limx n = x and limy n ~y.

u n is positive for all values of T?. then. 6).k for n^m.z m \<t. Corresponding to any positive e. General Principle of Convergence.] . for u n tends to / from above. 2 e N W htn + p-^ml therefore Hence the condition is necessary.QO . NtfTE. and by Art. then (x n ) and (y n ) are convergent sequences . As in Art. however small. Next suppose that \ z m ^ . it must be possible to find m so that \ z m + p -z m \<e for every positive integer p. First let u n -> I. The necessary and sufficient condition for the convergence of a complex sequence (z n ) is as follows : Corresponding to any positive e. If u n ->--<*). \ X __ r |<-1<: MTU] lay __. all the te rms of u n have the same sign.7^40 Mm \ < ->. and similarly it can be shown that (y n ) is convergent. we can therefore choose m so that for every positive integer p. Hence (z n ) is convergent. If I <0.. 6. we can choose m so that l<u n <0 for n^m. we can find m so that u n < . however small. If (z n ) is convergent. EXERCISE XXV 1. and the condition is sufficien t./ \^l f \ ^m+y J 'm \ ^ 2 fc *iu-i | . the condition may be stated thus. If / ^ 0.244 SUBSEQUENCES OF PROPER FRACTIONS 11. after a certain stage. 6. f Kithor u n tends to a limit or to . for every positive integer p. If u n is a decreasing function of n. it must be possible to find m so that if . Proof.x m | < | z m f p Therefore (x m ) is convergent (Art. Therefore the terms of (u n ) beginning with u m are negative. then | x w +p .

[The sequence. 2. (2).) -0. 1 is true. omitting such fractions as |. [Art.] n V 2 3 n/ 6. ! arranged in a definite order. Represent the first few terms graphically . Prove that lim . + .] FIQ.] 5. . 3. [See Ex. 11. . 41.(\+ \ + 1 + . GEOMETRICAL CONVERGENCE 245 4.. Show that the positive proper fractions can be | I ' . If u n >0 for all values of n and u n+l /u n ->l } then (nu n ) n -+l. 9. If u n ^ 7. (3). indicate terms which form subsequences converging to the upper and lower bounds as a limit. If u n is an increasing function. 9. XXIV. is 1 1 2' 3' 1 3 1 I 9 4' 5' 2 3 5' 5' 5' 6' 6 The required subsequences arc indicated by the -Q dotted lines. . the statement in Ex. and Art. so as to form a sequence.

.. z~ 5.n .^ .. 1 1 1 ~ w -7i + l 8.. If 1 prove that u n -* oo . . -zr*+ + -^T. Hence show that . > where for r = l. 3.. and consequently (n 4- "n-l e. [Show that r(7i 1 1 . + i . n "~l. prove that 3(71-2) 71 .. n. show that -> e. 2.+ .

[(i) Here t* n 2 = ^ + w n -i ^ n . then ^ rt ->a. . Thus w n i^ n _! -> and it follows that The case in which u l <a can be treated in the same way. then u l 2 >u 1 +. then v w ->j8. Illustrate geometrically. (ii) if v n = */ (A: v n __ l ) and v 1 <fc. If Wn=/( w n-i) where f(x) is such that (i) f(x)>0 if *>0. Hence ^ n -> Z ^ a. -/? are the positive and negative roots of x 2 -x-k prove that (j) if u n = */ (k + u^) and %>0. 4. and (iii) the equation x=f(x) has a single positive root a. 2 <t6 1 and (u n ) is a decreasing sequence.1 ~ u n~i -Un -zi hence u n '^u n _ l9 according as u n _^'u n _ z . for n 3. and shows that (u n ) is a decreasing sequence.'.. we have and thus P r Pr ~ u r ~ u r+\* The figure illustrates the case in which Ui = ON l = N 1 p l > a.] 11.1 ~ & + w f2-2 ' w n 2 ~ w n .] 246 CONVERGENCE TO ROOTS OF EQUATIONS 10. u 2 2 = u l + k<u l 2 . J^ir5< suppose that u t > a. therefore (u n ) is monotone. Again.9. i/. /. 42. the expressions . Show that as the number of roots tends to infinity. /. FIG. u 2 >oc. In Fig. [It is advisable to draw a figure.. Similarly u n ><y. and so N r is always to the right of N. (ii) f(x) increases with x. Thus u n -> I ^ a ^ and NN r N 2 N. u 2 2 ~u l + k>u 2 + k f . Hence N r tends to N as a limiting position and u n -> a. (ii) Draw graphs of y x and y~ meeting at A. Thus J\^ r P r > N r p r for every r. 42. prove that w n ->a. If k>0 and a.k ( u t ~ a) (u l + /?).k for 'i^! 2 .u t . .

. which is the same for both. Let u n be a function of n which has a definite value for all positive integral values of n. Also lim I ( u n~i + v n-i) < Wn-i and V n = s/(^ n -i^n-i)> ^n-i.. <^ w . so that **n. 2_7)2_7 % Draw a graph of the last equation... and so does v n . to oo ^3.. is called an infinite series. prove that 2/ ^( a. If u l9 v 1 are given unequal numbers and n^^ for prove that (i) u n decreases. <u 2 <u l . and v n increases as n increases. u Z9 u^ u$ . and the sum of its first n terms by s n .7 =0 as limits. K ~ v n = i ( \^w_i . An expression of the form in which every term is followed by another term.Mn^) = lim t* n ....x . w 3 .j CHAPTEK XVI CONVERGENCE OF SERIES (1) 1 .V 7 WT"^". The sum of the p terms immediately following the nth term is denoted by R Ui v .. and that both sequences converge to 2 as a limit. is an increasing sequence. and prove that if u n ^ \7-V7-f \/7-N/7T. v ~ w n-f 1 + u n+2 ~^~ ~^~ u n+j> ~ s n+v~~ s n- . If 3 = v7-v7+y. v n = lim t. 12. to 2n roots. n _ 1 = lim (2u n . This limit i s called the arithmetico. then %.. u 5 .. is a decreasing sequence.. Definitions. 13.' U n Thus v^^ Hence u.tend to the roots of x 2 ... <v n . This series will be denoted by En^\u n .geometric mean of u and i\ (Gauss). or by 27w n . n tends to a limit.N/ v -i ) 2 > > * u n > v n for ever y ^ i . Prove also that \7-fv7. (ii) Each of the functions u n and v n tends to a limit.

neither will t n . the series is said to be divergent. If x< .. to infinity or to minus infinity. In fact. +x n ..As n tends to infinity. is convergent and its sum is zero. s n = l or 0. or it may do none of these things.1) + (1 . If x^l.. according as n is odd or even. whether finite or infinite. according as s n oscillates between finite limits. If s n tends to no limit. or between + oo and . s n ->oo or . It diverges if oscillates finitely if x== . If x= -1.. For s n = (l-z n )/(l-z) if x^l. lim s n = s. so also will t n ..+w n . The Geometric Series. We say that the series oscillates finitely or infinitely. Thus the series (1 . 248 ADDITION OF SERIES 2.. and is not a sum in the sense described in the definition of addition. hence (i) if s m+n tends to a finite limit or to oo . * n n -Sfn an d 5 m is a fixed number. or briefly. For let 5 n = w Then n = s m 4.oo according as n is odd or even.. .1.... or 2u n = s. and u m+l both diverge or both oscillate. . oscillates. We are therefore not justified in assuming that the sum of an infinite series is unaltered by changing the order of the terms or by the introduction or removal of brackets. Series which diverge or oscillate are often said to be non-convergent. (1) If m is a given number. converges if l^l^l an d its sum is l/(l-x). the series is said to oscillate.1.. or they may transform a convergent series into one which diverges or oscillates.. there are four distinct possibilities : s n may tend to a finite limit. and oscillates infinitely if x< . 3. changes of this kind may alter the sum. n->oo This is also expressed by writing w t + w 2 + w 3 + .1 -t-. Theorems. whether finite. to oo =s or 27fw n = s.1.1) + (1 .oo . 5 n ~>oo .. If |z|<l. 1 + w 2 -f. If s n tends to a finite limit s. but the series 1 1 + 1-1 + . (ii) if s m + n tends to no limit. The series 1 +x + x* + . the series Uj + Wg + tia-f 6<tfA converge.or infinite.oo . Thus s is defined by lim s n = s. If s n tends to oo or to .1) + . z n ->0 and * n ->l/(l -x). the series is said to be convergent and s is called its sum to infinity. The sum (to infinity) of a convergent series is essentially a limit. .

s o does the second.(u 2 v 2 ) -f . then ks n ~+ks. . NOTE. so that where m tends to infinity with n. neither does ks n . . so does 2v n . and v l + v 2 + . finite or infinite. . Changing the Order of Terms. . therefore lim s n = s..1. then (i) // Su n converges to a sum s. // the terms of a convergent series of positive terms are rearranged. then Eku n is convergent and its sum is ks. . 4. therefore lim t n = s. If $ n tends to no fixed limit. (ii) The proof is similar to the preceding. . -f w 2 -f .. If k 9^0 and s n -oo . and let the terms be arranged in groups without altering SERIES OF POSITIVE TERMS 249 their prder. For let s = Then for every n we can find m and p. . + v n ) = s 4. converges. so does Zu n .Vj) + (u 2 -f v 2 ) + . does ks n . the series remains convergent and its sum is unaltered. converges. . Also if the first series diverges or oscillates. Now lim s m = lim s m + 9 = s. For corresponding to every n we can find m y so that But lim t m = lim t m+l = s. and its sum is s + 1 . . unless & = 0. so.(2) If k is a fixed number and Eu n converges to a sum s. . Let Zu n be a series of positive terms. . 6. + w n ) + Iim(v 1 -f v 2 4. Denote the sum of the terms in the nth group by v n . If the series w x -f w 2 + . all the series in brackets are finite. Addition and Subtraction of Series. Under this heading we include series in which all the terms are positive beginning with some particular term. . Introduction and Removal of Brackets. (ii) If Zv n converges to a sum s. then (i) (Wj 4. converge and their sums are s and t respectively. (ii) (MJ Vj) -f. For if s n ->s. In the proof. and its sum is s t. 5. SERIES OF POSITIVE TERMS We first consider series in which all the terms are positive.

to oo = . and therefore tends to a finite limit or to <x> (Ch.. Comparison Tests. . . Theorem. Let Eu n -be convergent. therefore lim m = s. Derangements of the last sort are considered in another volume. among the first m tsrms of Zv n . The argument fails for such a derangement as MI + w a 4. For s n increases with n. Now lim 5 n ==lim 5^ +J> ==5. then Zu n will be convergent if (i) * u n ^a n for all values of n . . Let Suppose that the first n terms of Zu n are. XV. where s is a fixed number less than k (Ch.. (i) If a + a 2 + . Hence the series cannot oscillate. . we have u l + u z -r . Hence Zv n is convergent and has the same sum as Zu n . + u z + w 4 + w e + . so that every u is a v and every v is a u.Proof. and if s n is always less than some fixed number k. Denote the new series by v n . 3). but Su n is broken up into infinitely many infinite series. and since t is a fixed number. Proof. if s n <k for every n. the series converges and its sum is less than k. + u n ^. Further... a n <t . 8. (ii) * or if u n /a n is always less than some fixed positive number k . The rule for the addition of aeries applies in cases of this kind.a 1 +a 2 -f . 3). (1) // 2u n and Sa n are two series of positive terms and Ua n is known to be convergent. 250 COMPARISON TESTS 7.u 6 -f . Further. then s n ~>s.. Thus n^m^n+p. Thus u 2 does not occur till infinitely many of ti have been placed. suppose that the first m terms of 2v n are among the first (n +p) terms of Zu n . A series of positive terms cannot oscillate. where Zu n is broken up into two (or any finite number of) infix For here we cannot find m so that the first n terms of Zu n m terms of Zv n . it follows that u n is convergent. XV. NOTE. (iii) or if u n /a n tends to a finite limit. Also s n < f >m<s n+1) .. . and let the terms be rearranged in any manner.. and m tends to infinity with n.

therefore u n is convergent. and th e brackets contain 2. Group the terms as follows : 1 1 1 where the first terms in the brackets are l/2 p . +w n >JV provided that n>m.+ . The geometric series and the series of the . 8. .. at once from (ii). First let p>l.+. . its convergence or divergence is not affected by grouping the terms in brackets in any way we please. . flcient that these conditions should hold for all values of n greater than some METHOD OF COMPARISON 251 9. l/2 2p .y these tests. (2) // Eu n and Za n are two series of positive terms and Za n is known to be divergent. no matter how great. etc. (iii) This follows at once from (ii).hose which are most generally useful as test series. 4. Proof. t^-f w 2 + . such that u n /a n <k for all values of n. therefore % therefore Eu n is divergent. The series TJ + H^ + O + . terms.. we require certain series which are known or divergent. (ii) * or if u n /a n is always greater than some fixed positive number k . : s divergent. m can be found such that a l + a 2 + . Since Ea n is divergent.. values of n. Now ka n is convergent. -f a n > N provided that n> m . Now Ea n is divergent . . for if lira u n /a n is finite. we have u n <ka n . ^ s convergent if p>l and divergent if Denote the given series by S. u n >ka n . Each term in (A) after the first is less than the corresponding term in I 1 \ / 1 1 1 1 .. l/2 3p . (iii) or if u n /a n tends to a limit greater than zero.. /. for if lim u n /a n is finite. Theorem. Since S is a series of positive terms. then Uu n will be divergent if (i) * u n ^a n for all values of n . (i) Let TV be a positive number. respectively. a positive ound such that u n /a n >k for all values of n. . we can fin d a positive constant k.(ii) For all values of n.

... (B) Now (B) is a geometric series whose common ratio 1/2 P-1 is less than unity.C. we take !/ for e* n . 10. etc.... o let a n = * Then we have = 2 . we obtain 2n/n 8 or 2/n*. for p>l. Test for ro nvcrgence or divergence the series 1 x x 2 x n ~ l . Group the terms thus: i + i +( i + i) + ( i + i + i + i) + .. Is the series = ~ ..-f + o~~l r + ~* converge^ or divergent ? i . 4. 1/2 3 .. B B. the series S becomes 1+J-f^ + J-H. . each term of/S after the first is greater than the corresponding term of l+^ + J + i+. 1/2 4 .. respectively.. Next.... terms...4: o . 252 D'ALEMBERT'S TEST Ex. .... ... if p = 1... or !+_+ + . 2...... and the brackets contain 2. Now Za n is convergent .. but the last series is divergent. . Z . Lastly..*>. . and therefore S is divergent.. (C) where the last terms in the brackets are 1/2 2 .. 3 j. therefore u n is convergent...which is the same as !+_+_ + . Examples on the Comparison Tests. Disregardi ng the numerical factor 2.. 4 ..... ifp<l.. (D) Now (D) is divergent. Hence (B) is convergent. therefore S is divergent.. and therefore S is convergent. 1.A.. 1 Q K Ex.. or l+ + i + J + . 8..... * The nth tenn a w of a suitable test series is found thus : Keeping only the hi ghest powers of n in the numerator and denominator of u n . Each term of (C) is greater than the corresponding term of which is the same as 1 +i + f + f + .

Again. Now Zx n ~ l is convergent. 71 a: 1 + x n x x~ n + 1 x n If = 1. choose k so that l<k<l. (i) It is sufficient that the conditions should hold for all values of n g reate than some fixed value. then Thus w w <Wj . A. n ->0 and v n lx n ~ L -> 1.71 " 1 is convergent. a. D'Alembert's Test. Hence Su n is divergent.. . and we say that the test fails. then 2!u n is convergent whei l<\ and divergent when />!. Then by the definition of a limit u n+\l u n<k f r ^>^ ? where m is a fixed number.>l.is divergent. and therefore u n is divergent. (ii) Nothing is said as to the ease in which Urn u n+l lu n ~l. if u n+l /'u n ^l for all values of n.. a:~ n -> and w^ = = -. consider the series . TESTS DERIVED FROM THE GEOMETRIC SERIES 11...-> . NOTE. if />!. then therefore w t -f w 2 + +u n ^nu 1 -*cc . For example. Again. if u n ^/u n tends to a limit Z. In particular. Hence ZV-. Henc< u m+2 + .. 1+z l+# 2 l+o: 3 l+z n If <1. therefore convergent. Hence Wi + tt m+2 +. is convergent. k n ~ l and 2^7/j . A series Zu n of positive terms is eon vergent if u n+l ju n <Jc<\ 9 where k is a fixed number and n may have an\ value. 1 # n 1 1 1 If a. For if u n+l /u n <k<l for every n. _ . where x>0. For if Z<1. + . is divergent. In this case the seriei may be convergent or it may be divergent. and therefore Zu n is convergent. +.^ 4.+ . The series is divergent if u n+} fu n ^l for all values of n. u n = ^ } and the series is divergent.. therefore 27w n ii convergent. then m can be found so that u n+ ^>u n for n>m.

Comparing these tests. Cauchy's test is more general than D'Alembert's. but if the limit is approached from above. XV. 9. That this is the case will be seen by noting that (i) Cauchy's condition for divergence is wider than D'Alembert's. that if u n+1 /u n ->l then u n n ->l. so that always u n+l >u n the series is divergent. however. Cauchy's Test. The series 1+2+3 + 4 + . where m is a i fixed number. i The series is divergent if u n n ^l for infinitely many values of n. then u n <k n . and consequently Zu n is convergent. i (ii) It has been shown in Ch. if u n n tends to a limit L then Zu n is convergent when l<ll and divergent when l>l. is an instance of this . and consequently w n >l. A series Eu n of positive terms is convergent if i w n n <&<! where k is a fixed number and n can have any value. . for infinitely many values of n. that of D'Alembert is more generally useful than Cauchy's. and in both we have lim u n +J u n = 1 If. i If w n w >l. On the other hand. The proof is similar to that in the last article. 13. then Zu n is obviously divergent... It is sufficient that the conditions should hold for n>m. i For if u n n <k<l. i In particular. and since k<\ the series Zk n is convergent. u n+l ju n is a simpler function than u n . For in D'Alembert's test the condition has to be satisfied by all values of n greater than a fixed value.1111 _ 1 1 1 1 i + 2 + 3 + 4 + "' and r + 2i + i + 4* + -The first is divergent. u n ^^u n tends to 1 as a limit from above. the second is convergent. Zu n is divergent. For in most of the series with which we are concerned. and this is not the case in Cauchy's. Also nothing is said as to the case in which lim u n n \. (2). CAUCHY'S TEST 253 12.

lim =lim T xx. Here -^ ~ i i and D'Alerabert's test is inapplicable. Ex. = ' lim -= lim = i - Hence the series is convergent. D'Alembert's test fails... If x 1.*.. is convergent for a ll positive values of x.i But if u n n ->l. it does not follow that u n+l /u n tends to any limit. series 1 +x + x 2 / 12 + x* f (3 + . is convergent if 0<#<1 and divergent if x n x n+l w n41 n Here u n = . and if x>l. it is divergent. but in this case the series becomes and is divergent. n n n n + 1 u n n-fl Hence. 3. So we may expect that Cauchy's test will sometimes succeed when D'Alembert's fails. "~~ />. Test for convergence the series a + 6 + a 2 4. Show that the. . Using Cauchy's test.. u n+1 = ~.n-i ~n ..6 2 + a 3 H. if 0<:r<l.. 3 below. 1. as in Ex. . . since w n n ->a a o r 6 2 . * TT ! 7? II 1' iv Horo = . 254 EXAMPLES OF SLOW CONVERGENCE Ex. 2. Show that the series ic + + + + .6 3 -f . .. the series is convergent. 2 #3 yA Ex.

. J + l-+-J + . m-fl m+1 (ra + 1) 2 (iii) Hence show that. is convergent..1. 6. and divergent if a^l.) + (* + ! + i + ) .. and so on. 2..j -f .according as n is odd or even.. at least 1000 terms o 4 must be taken. with an error less than 001. the series is convergent if 0<a<l and 0<6<1. State the character of the series showing that for the second series lim$o M ~-l and lim 5 2n _ | . [Consider the sum to n terms of the second series. Criticise the following : 2(-J + i... if ci l -\.. . EXERCISE XXVI 1. . -f.rri.a 2 + ft 3 -\..] 4.. =(i + 4 +4 + .rr + : ~^r + v .. 3. Show that the scries x + x ^ + x 2 -} x~ 2 + x 3 + x~* + . J (m-fl) 2 (w-f2) 2 (m-l-3) 2 lies between and . which is absurd. show that 11 1 1 1 .-~ + /.11111 1 (iii) Given that 2 20 > 10 6 4. + a t ) 4. .. If n is a positive integer..+ i + J-f-.-. Prove also that the converse is true if a n >0. = n-i+4 + . for n > some fixed number m... in order to find the sum to infinity of the series 4. is never converg ent.J (a 2 + a 3 ) + I ( 3 4-^4) 4. .. Show that. so also is -?> (a. 5. show that the sum of the first million terms of the series I -f i f 1 4. or if b^\. 1 2. 1 <..-) = i + ^-i-*-l-i -.. Show that v ' (m + : (ii) The sum to infinity of the series .. for J <1. is less than 20. .^ + ~r + ..

1 For what positive values of x are the series in Exx.. 2 . 1 1 1 o. Explain why D' Alembert's test cannot be stated thus : A series of positive terms is convergent if the ratio of each term to the preceding is less than unit y.. .~ Test for convergence or divergence the series whose nth terms are 15. 24.2 3. __ _____ _ 19.. 23-31 convergent. 2 29.4 5. 1 + */2 2 4. .8. .lll 9. .. . . i 26 ' 1. is.CONVERGENCE OF SERIES 255 Determine whether the series in Exx. -JL_. - 32.6 1 3 _5_ 7 13. and for what values are they divergent ? 23.. . n .9 .. 17.3 4. Exx. 20. 2 n + a | n . 22. 10 a+6 a + 26 a + 36 1 . -A. .. 30. /t /_4..x 2 / 25 . N/rc 2 + l-n.+ + + .2.5. 7 J_ + __ + _?_ + . H. V^-fl-s/n 8 ... 33-37 refer to series of positive terms. -^_.6 7. 1 2 a 3 a x a. is. 31. 21.. l+2x + 3x* + 4x* + . -.. Z 1 -. 7-14 are convergent or divergent.

for From (A) it therefore follows that s 2n is positive and increases with n. so is u n z . and from (B) it is seen that $.5 3. 7 3 . u n 2 <u n .7 1 . in which the terms are alternately positive and negative.7 5. If 27w n is convergent.1 for n ^ m.w 3)-( w 4.. n ^ + . _i i 41 -i i -. 3). 37. .4.33. /. We have and this can be written in the form *2n = w l-( w 2.. so is Su n ln. The series u^-u^ + u^..7.1 ^ ^=^ .5. S 2ni . . 38-41. 39. [For w w / Find the sum to infinity of the series in Exx.3 .12 3 41) i . for w w ->0 .5. v n u n 34. 11 256 ABSOLUTELY CONVERGENT SERIES TERMS ALTERNATELY POSITIVE AND NEGATIVE 14.6 3. l ~s 2 n + u 2n+i an( l l* m U 2n+i \ therefore s 2n+l tends to t he . is convergent if each term is numerically les s than the preceding and lim u n =^0. .5 2..9 1..6. then Zt' is divergent. If Hu n is divergent and ^t. 2n tends to a limit which is less than MJ (XV. s.. 9 5 .3.u 1.6 .4 2. w 2 -i/ 3? 3 -w 4 ..A 2 3 4 .-f :r~ 1.5 3.. If 2u n z is convergent.w 2n Now Wj-i/a. are all positive. 38. 2n is always less than the fixed number u v Hence. [After a certain stage u n <l. . If Zu n is convergent and -2! ^ -^t. Again. so is Zu n f(\ +u n ). 7 . .. Theorem.] 36.6"" . then Zv n is convergent. Proof. v n u n 35. If Eu n is convergent.w 5)----.+ ~~^-~ -f .

same limit as s 2n . The series is therefore convergent, its sum is positive and less than u v Ex. 1. Show that the series 1 - ^ 4- -3"-- 4 + ... is convergent. Here the terms arc alternately positive and negative, each term is numerically l ess than the preceding and lim l/--0 ; hence the series is convergent. SERIES WITH TERMS EITHER POSITIVE OR NEGATIVE 15. Absolutely Convergent Series. The series 27w n , containing positive and negative terms, is said to be absolutely convergent if the series 2 1 u n | is convergent. If Hu n is convergent and E | u n | divergent, then Su n is said to be conditionally convergent or semi-convergent. Thus the series l-| + i-i + --- is absolutely convergent, because the series 1 -f f + \ ; + \ + . . . is convergent. But, 1-^4- 3- -$+... is condition ally convergent, for it is convergent and 1 4- 1 + 3 4- 1 4- . . . is divergent. Theorem. An absolutely convergent series is convergent. Let Zu n be absolutely convergent ; then by definition Z \ u n | is convergent. Now i/ n 4- j u n \ =2u n or 0, according as u n is positive or negative. Therefore every term of the series (u n + \ u n \) is >0 and is < the corresponding term of the convergent series 272 1 u n |. Hence Z(u n + \ u n \) is convergent and consequently, by Art. 4, Zu n is convergent.

PRINGSHEIM'S THEOREM 257 NOTE. If the reader fails to see the point in the statement of this theorem, he should note that, if we say that Zu n is absolutely convergent, we assert the convergen ce of another series, \ u n \ , and not that of 2u n itself. From Arts. 11 and 12, it follows that the series Zu n is absolutely convergent i f, after a certain stage, , i_ ?p <k or if |uj<*, where k is a fixed positive number less than unity. 16. General Condition for Convergence. Since s n is a function of the positive integral variable n, by Oh. XV, 6, (B), the necessary and

sufficient condition for the convergence of Zu n is as follows : Corresponding to any positive number e that we may choose, however small, it must be possible to find m so that y if u^in^for every positive integer p, //j/7/ -jo 7? '}/ -4- ?/ -4- -U II I <^~ f HUM t-o, j it n> j, I I Mft-fi ^ '*n~f 2 ' * * * ' n-f j?| ~Thus if Zu n is convergent and p is any integer, and in particular u n ^0. But the condition (A) is 'not sufficient to ensure the convergence of Hu n unless p is allowed to tend to oo in any way whatever. Ex. 1. Apply the general condition to the series 11 1 + 2 4 "3 + '"~ h /i' f "' ' 1 1 lp Here 7f w p ~ 4- +...+ > ' n + 1 n + 2 n+p n~\-p For any fixed value of p, p/(n H />) -> as ?i-> oo . But if p ri 2 , then p/(n f p) ~> 1 ; hence, Rn t p cannot tend to zero, and so the series is not convergent. 17. Pringsheim's Theorem. // Su n is a convergent series of positive decreasing terms, then nu r ~>0. For if p is any positive integer P u m+v< u m+l+ u m+2+ + u m+jT > as W->OQ Let p = m, therefore mu 2m - >0 and 2mu 2m ->0 ; i.e. nu n -*Q, when n = 2m. Let p = m 4- 1 , therefore ( m + 1 ) u 2m +l ~- > ; and (2m + 1) u 2m+l = 2(m + 1) u 2m+l - u 2m+l ~~>0 ; i.e. nu n >Q, when n = 2m -f 1 . This proves the theorem, whether n is odd or even. Ex. 1. Use this theorem to show that the series 27 - is divergent. n Hero nu n n- - =1, and does not tend to zero. Hence E- cannot be convergent ; n n n and, since it is a series of positive terms, it must be divergent.

258 REARRANGEMENT OF TERMS 18. Introduction and Removal of Brackets. Let the terms

of Zu n be arranged in groups without altering their order. Denote the sum of the terms of the nth group by v n , then (1) If 2u n is convergent, so also is Sv n , and these series have the same sum. For let $ = !*! + 2 + ...+M n , t n = v l +v% + ...+v n . Then for any value of n we can find m so that either s n = t m or s n = t m + (w w4l + w m+2 + . ..u m + 9 ), where the terms in the bracket are included in the group v m+1 . If Zu n is convergent, fi n,p :=s w n+1 + t w+2 +...tt w4 . J> ->0 as n~oo, also m~> QO as n ~> oo , therefore s n - Z m -> 0. Hence Sv n converges to the same sum as Su n . (2) If Sv n is convergent, Eu n is not necessarily so. For, given that Zv n is convergent, we cannot conclude that R njtf ->Q for all values of p. (3) Denote | u n \ by u n \ and suppose that the terms of Zu n f are grouped as in (1). Let v n ' be the sum of the terms of the nth group. Then, if Ev^ is convergent, so also is u n ' (Art. 5) ; and, as n -> oo ,

Therefore, as in (1), Zv n converges to the same sum as Zu n . 19. Rearrangement of Terms. The following example shows that it rearrangement of the terms of a semi-convergent series may alter its sum. and a rearrangement of the signs may change the series into one which is divergent. Ex. 1. Lft's be Ihe sum oj the tern i -convergent series

With regard to the, series t I I I I I /ri , , i 1 1 1 i 1 1 -5-4*3-5 1 + 5- ........... (B) ^r^rr^r- ......... (C) obtained by rearranging the terms and the tigns of (A) respectively, prove thai (B) is convergent with a sum {s and (C) w divergent. (i)Let w - 1 -5 -r- ---,-... to M terms. or w = l -~ --H-- -- -~ + ... to n term s : then ^

SUM ALTERED BY *.

1 1 1

--lim9 2n =-*. Also Hence, the second series converges and (ii)^t < n = i + l_l + l + l_l then *>( -+- - }+( -+- -O D I/, 1

Now the series 1 + - + - -f . . . is divergent ; 2i 3 t sn +% &U tend to oo . Hence the series (C) is Theorem. // the terms of an absolu the series remains convergent and its su? Denote the new series by v n) so tha We have u n -f | u n \ = 2u n or 0, accor And since E\ u n \ is a convergent se (u n -f | u n |), for its terms are < the c Let Z\u n \=s and Z(u n + \u n \)^ E\u n \ and (u n + \u n \) -are series o unaltered by any rearrangement of tern 2\v n \=s> 27(^ + 1^1) = ^ Ex. 1. Find a range of values of x far whicJ l+(x-* 2 ) + (x-x*) z can be arranged as a convergent series of ascendi Expanding the terms of (A) and removing h

Now, if |a?-a^|<l, the series (A) is coi justify either the removal of brackets or a derai However, by Arts. 16, 18 (1), 19, each of these convergent ; and, by Art. 5, this will be the case

where a? / =| x |, is convergent ; that is, if x' + x'<\(*J$-l) t or -tON/5which is the required range. NOTE. What has been said does not preclude can be made over a wider range.

7 ERROR ^arching for the sum of an infinite content with an approximate value.

and its sum, denoted by R n , is called >,ries (A). Thus we have

i approximate value of 5. It should be s. )er limit to the numerical value of the o find a number which | R n \ cannot t series 4-...+(-l) w W w +... , ive terms. Here

n+ 3 Brackets is positive and less than u n ^ Hence the error in taking s n as an lij less than w n _ i _ 1 . ror in taking x - x 2 /2 as the sum of the 3ss than x 3 /3. .n which all the terms are positive or in erms are all of the same sign. This is n often proceed as follows. 3 x n --K"H +... where 0<^<1. The series is 3 n

( le of s is numerically less than x n + 1 / (n + 1 ) ( 1 - x). ;ence or Divergence, For a conlue of R n for a given v&lue of n, the more ge.

ilue of r y the more rapid is the convergence

/4 2 + . . . is convergent, but in order to find mals, at least 1000 terms must be taken fore say that the series converges slowly.

-v" SERIES OF Next, consider the series 1 -f |- f _ of a million terms is less than 20 (st therefore said to diverge slowly. In prat, the reckoning so that we have to deal w ; For instance, the series, 5 = 1 - J J By grouping the terms, we find ' 1

This series converges more rapidly find that *-2[t + i(i) 8 + i(i) 5 +. rapidly than a geometrical series T 22. Series of Complex ' that the series Zx n and Sy n c s n , cr n , r n be the sums to n tei. Ch.XV,10, lim Sn = lu and we say that z n converges ti Thus Ez n is convergent if Zx n a

(2)ByCh. XV, U^t of Zz n is as follows : Corrcs 1 it must be possible to find \s n+v -s n \<, tha (3) Definition of P terms and 2\ z n \ is As for real sei convergent.

then 2\ z n \ is c Consequently (4) It is impo. 2x n and Zy n9 an The first part c

Hence if Z \ x n \ (5) The theorems c of brackets and the c complex series. This

' FOR SERIES , is complex, the series ..+z n ~ l + ... and when convergent, its sum is 1/(1 -z).

1 if | z |> 1, z n does not converge to 0, (3), *->0 and s n -*l/(l-z), 3 sum is 1/(1 -0). re 0<r<l, we have >s2tf + tsin2fi) + ... to oo - r cos + ir sin

1 - r cos 6

that these results are also true

1 1-5 FIG. 43.

then w n ->0. then we can find

ULTIMATE SIGNS OF TERMS 263 Now the series + - H + . . . is divergent, therefore m m+I m+2 uju n -><x> and w n ->0. 25. The Binomial Series. For any given value of n, the series - n(n-l) 9 n(n-l)(n-2) ... (n-r + 1) 1 + nx + -Vs ' ^ + + -~ ~ 7 * ' x + is called the Binomial Series. When n is a positive integer, the series ter-

minates and its sum is (1 + x) n . In all other cases, it is an infinite series, which we denote by + ti 1 + tt 2 + ... + t< r + ... . (1) Ultimate Signs of the Terms. From the beginning, or after a certain stage, the terms are alternately positive and negative, or they all have the same sign according as x^O. For w r +i/ M r = (n-r) xj(r + 1). Hence if m is the least of the numbers 0, 1, 2, ... which is greater than n, the terms u m , u m+l , u m+2 , ... are alternately positive and negative, or al l have the same sign according as x^Q. (2) Convergence, etc. We shall prove that the series is absolutely convergent if \ x \ < 1 ; and when x = l, the series converges if n> - 1, oscillate s finitely if n = -1 and oscillates infinitely if w< - 1. Proof. Denote the series by U Q -h u + u 2 4- . . . , then

n-r

-.x, as r->oo ; therefore lim

*r+l

u r r+1 and the series is absolutely convergent if | x \ < 1. Next let x = l. If r>n, u r ^ and u r have opposite signs, so that, after a certain stage, the terms are alternately positive and negative. Also -si +-T where a r = r-n r (i) If n>-l, as r~>oo, a,,.->n + l>0. Hence, by Art. 24, w r -0. Also | u r+ i | < | u r |. Hence, by Art. 14, the series is convergent. (ii) // n = - 1 the series is 1-1 + 1-1 + .. . and oscillates finitely, (iii) // n<-l, let n + l= -m so that w>0, then

Hence | u r \ -> oo and the series oscillates infinitely. The case in which x~ - 1 is considered in Ch. XX, 6.

264 CONVERGENCE OF SERIES EXERCISE XXVII RANGE OF CONVERGENCE Show that the series in Exx. 1-5 are convergent.

3 _J ___ L , L__JL , 4 J V' 3 1 - 3 ' 5 , a + 1 a+2 a+3 a + 4 " 2 2.4 2.4.6 . , 1 1 1 1 1 K | I __ ____ I . __ I _____ L 2 2 3 2 4 2 5 2 6 2 ' X # 2 X^ 6. Prove that the series - -- , + - - - ... is convergent if <x< 1. l+x * * 6

7. Show that the series 1 4- x + x 2 j | 2 4- ^ 3 / 1 3 4- . . . is convergent f or all values of x t real or complex. 8. Show that - + - ~ -- - -f - + ; -- - + ... is a divergent series. a a-f-1 a + 2 a + 3 a + 4 a-f5 9. Show that the series 3

is convergent if /n 2 <4/27. [Consider separately the sum of the odd and the sum of the even terms.] 10. (i) Find a range of values of x for which the series

is convergent. (ii) Find a range for which the series can be arranged as a convergent series of ascending powers of x. 11. If 0<z ^ 0-4, show that the series 1 4- (2x cos 6 - x 2 ) + (2x cos 6 - a: 2 ) 2 -f . . . can be arranged as a convergent series of ascending powers of x. [It is enough to show that with the given condition | 2x cos \ 4- x 2 < L]

12. Let the terms of Su n be arranged in groups, without altering their order, and denote the sum of the terms of the nth group by v n . Given that 2v n is convergent, prove that Eu n converges to the same sum as Ev n in the following cases : (i) If the number of terms in every bracket is finite and w n ->0. (ii) If all the terms in each bracket have the same sign. [See Art. 18. In case (i), J? n , j,->0. In case (ii), v n -+Q ; hence, every te rm in v n tends to zero, and case (i) applies.]

SUMMATION OF SERIES 265 13. Show that the series l+z + z~ + ... is absolutely convergent when 2 l/(l+i), and find its sum. 14. If z n = +-~i L > show that Ez n is convergent, but not absolutely convergent. n n Prove the results in Exx. 15-17, where R n is the remainder after n terms in the given series. 15. For the series 1 +ry- ^ 16. For the series 1 1 r + 7-5 - - + . . . , \R n \<r provided that 1 I o I ft X X^ X^ X^ tl/ + 1 17. For the series 1 +r + r^ + |- -+..., R n < i r- provided that Q<x<n + l. IJ !_ l_ 1^: n + l ~ x Prove the results stated in Exx. 18-23 : 18 27 L-- 1 19 ""' JLU< +-J + An -. /- ii/

2 i |n 20. If |*|>1, -i- + -l- + _i- + ...to*=- L

? - T . I-x* x-l 2L li + l-i4 + ri~;8 + -- tOQ0 ^ ] !^ or TI~> ^cording as | x \$l. 22. If 6-l>a>0, then

a a(a + l) a(a-f l)(a + 2) ^ 6-1 + + " f " ^ ^ '

23. If 6-2>a>0, then

o ___ __ 6 6(6 + 1) 6(6 + l)(6 + 2) ^(6"-a-l)(6-a-2)' 24. Show that the sum to n terms of the series cos + cos 2^ + cos 30 + ... is I {sin (n + 0) cosec \Q - 1}. Hence show that the series oscillates finitely except when 6 is an even multiple of TT, in which case the series diverges. n [2 cos r6 sin - -sin (r + 16) - sin (r - 10).] 25. Show that the series Z'sin nO oscillates finitely except when 6 is a multipl e of TT, in which case it converges to zero.

26. Show that the series z + |2 2 + z 3 + ... is convergent, but not absolutely onvergent, when z = cos - + 1 sin - . Illustrate the slownc 4 4 by plotting the points 8 19 s 29 ... $ 7 on an Argand diagram.

convergent, when z = cos - + 1 sin - . Illustrate the slowness of the convergenc e 4 4

CHAPTER XVII CONTINUOUS VARIABLE 1 . Definitions. We say that x varies continuously from a to b when it passes once through all real values between a and b. Under these circumstances x is called a continuous real variable. Limit of f(x). Consider the function f(x) = {(1 -f- x) 3 - l}/x. When x = Q,f(x) has no definite value, for it then takes the form 0/0. In other words the function f(x) is undefined for the value zero of x. For every value of x except zero, we have f(x) = (3x + 3x 2 + x 3 )/^ = 3 4- 3x -f x

2 . Thus, if x is nearly equal to zero, f(x) is nearly equal to 3 : further, by making x small enough, we can make the difference between f(x) and 3 as small as we please. This is roughly what is meant by saying that 3 is the limit off(x) as x tends to zero. The matter may be put precisely as follows : Choose any positive number e, no matter how small. If | x \ < 1, then

therefore | f(x) - 3 1 < e provided that | x \ < \<L. Thus, as x tends to zero, |/(z)-3| becomes and remains less than any positive number that we may choose, no matter how small. This is precisely what is meant by saying that 3 is the limit off(x) as x tends to zero. It should be observed that f(x) never attains the value 3. The usual definitions are as follows : (1) //, corresponding to any positive number c which we may choose, no matter how small, there exists a number rj such that I f( x ) ~ 1 1 < provided only that | x - a \ < TJ, where I and a are fixed numbers, then I is called the limit off(x) as x tends to a. This is expressed by writing lim f(x) = I. z-+a Or we may say that /(#)-> I as x-^a, where it is understood that x may approach a from either side, i.e. x may tend toa+Oora-0. Observe that nothing is said as to the value of f(x) when # = a. Such a value may or may not exist. In any case, we are not concerned with it.

THEOREMS ON POLYNOMIALS 267 (2) If, corresponding to any positive number c, no matter how small, there exists a positive number N such that | f(x) - 1 1 < c, provided only that x> N, where I is a fixed number, then I is called the limit off(x) as x tends to infin ity. This is expressed by writing lim /(#)== Z or f(x)->l as x->oo . 'x-+<x> ,(3) If to any positive number M, no matter how great, there corresponds a positive number N such that f(x)>M, provided only that x>N, we say that f(x) tends to infinity with x, and we write /(#)-> oo as x->oo

or (less correctly) lim f(x) oo . 3->oo The reader should be able to frame for himself corresponding definitions for the meaning of ' f(x) tends to minus infinity as X-+CG .' 2. Fundamental Theorems. Let u, v be functions of x which tend to the limits I, V as x tends to a or as x tends to infinity, then (i) lim (u -f- v) = I + V, (ii) lim (u - v) = I - 1', (iii) lim uv = IV, (iv) liml/w = l/Z unless Z--=0, (v) limufv = l/l' unless Z' = 0. Apart from some verbal changes which the reader can easily make for himself, the proofs are the same as those in Ch. XV, 2. 3. Two Theorems on Polynomials. What is said in this article holds good, no matter whether the variable x and the coefficients are real or complex. Let f(x) = a Q + ax -f a 2 x 2 4- . . . -f a n x n , then (1) If is any positive number, no matter how small, we can find a positive number 17 so that | f(x) a 1 < c provided only that \ x \ < 77. Proof. For shortness, let | x \ = x' and let k be the greatest of the numbers (a^, |a 2 |, ... , \a n \ ; then | f(x) - a Q \ < | a l \ x' -f | a 2 1 x' 2

If '<!, then kx'f(I -x')<, provided that kx'<(l-x f ), that is if Therefore | f(x) - a 1 < c, if | x \ < /(k + e) ; and so /(k+*) is a suitable value of 77.

B.C. A.

268

QUOTIENT OF POLYNOMIALS

(2) // M is a positive number, no matter how great, we can find a positive number m so that

\f( x ) \>M, provided only that \x\ The letters having the same meaning as in (1), we have

|/(*) !=*">.

n -l - +n-2 ^2

i X

1

1

Choose any number e between and | a n \ . Then by the last theorem 1 1 <

if

< - - that is if #'> . X K+ For such values of x we have | f(x) \ > x' n {|e& n \ - e}. In particular, put = i|a w |, then |/(x)|>t|a B K if x': Hence if m is the greater of the two numbers and

then f(x)>M if \x\>m. From Theorem (1) it follows that lim f(x) = a Q *

a;->0 Again, if x = y -fa, then y->0 as x-^a and /(a) is the term independent of y in the polynomial f(y + a). Hence lim /(x) - lim f(y -f a) =f(a).

. 1. // f(x) =

_ 3

Km /(a;) ; (i) -3 3 ^5' (ii) Let a; = 1/y, then y -> as x -+ oo and lim *- lim /(X) ~

0, (ii)

, (iii) cw a?-

(iii) Let x = 1 -f y, then t/ -> as a? -> 1 and

* The reader must understand that this is not a roundabout way of saying that f( x}^a Q when

CONTINUOUS FUNCTIONS

269

4. Continuous and Discontinuous Functions. We confine ourselves to one- valued functions of the real variable x. We may say that/(x) is a continuous function of or in the interval (a, 6) if the curve whose equation is yf(x) is continuous between the points

where xa and x = 6. This supposes that we know what is meant by a ' continuous curve/ A representation of the curve in Fig. 44 can be drawn without allowing the pencil to leave the paper, and we say that it is continuous. In Fig. 45, the curve cannot be so drawn near the points where x = x and x~x 2> and we say that the curve is discontinuous at these points.

FIG. 44.

FIG. 45.

We choose the following as the distinctive feature of a continuous curve : As the point P (x, y) passes along the curve, any small change in the value of x is accompanied by a small change (or by no change at all) in the value of y. (This is true everywhere in Fig. 44, but is not true near the points where x x l and x=*x% in Fig. 45.) More precisely thus : let P Q (x Q , y Q ) be any point on the curve and P(x, y) any other point on it on either side ofP . Choose a positive number c, no matter how small. Then if \y-y Q \< for all sufficiently small values of | X~X Q |, the curve is said to be continuous at P . Further, we say that a curve is continuous if it is continuous at every point on it. Thus the curve AB is continuous in Fig. 44, but in Fig. 45 it is discontinuous at the points where x=x t and x=x 2 . Definitions. (1) The single-valued function f(x) is continuous at x = x (or at the point X Q , or for # ), if for any positive number e that we may choose, however small, a number 77 exists such that \ f(x) -f(x Q ) \ <<=, provided that \ x - X Q \ <TJ. This is equivalent to the following : f(x) is continuous at x iff(x)~*f(x Q ) as X-+XQ, where x may tend to X Q either from the left or from the right.

270

TYPES OF DISCONTINUITY

It is important that the student should fully realise all that is implied in the definition just given. If /(#) is continuous at z , then (i)f(x) has a single definite value when x~x ;

so that the end points are included. Taking (as usual) Jx to denote the positive square root of the number x. 45 let y v y/ be the ordinates of C. then y-^y^ as x-^x l from the left. f(x) = 1 -f x 2 . (iii) and we must have i x =/(# ) = 2 . for all values of x except zero. as x-^a^-hO.e. but are not all equa l. 45.A). when #=0. it is supposed that y is infinite.... .x is discontinuous at x . (i) In Fig. so that f(x) is discontinuous at # = 0. Thus y is a discontinuous function of x at x = x 1 . A simple case. then /(x + A)-> a finite number L 19 and/(x . b X FIG. 45.ar) T * zero. L 19 L% all exist.^i + . x 2 x 2 If f(x) = x 2 + ^-f . but.(ii) if h tends to zero through positive values. C 7 . i. and so f(x)0 . /(O) =0. to oo .. 1/x and 1/x 2 are discontinuous at x = 0. For example. (iii) It may happen that f(x) has real values only on one side of the point x = x . is given by Goursat. Types of Discontinuities. In such a case the curve y=f(x) stops abruptly and/(x) is discontinuous at # . in which f(x Q ). a finite number L 2 . we see that v x . 1. This is expressed by writing lim f ^ '^^ and lim v = Vo." and at # = # 2 . as x-^x l -0 and y->y% as x->x l from the right. every term of f( x) is 1 ~}-X" (1 4. and lira /(*) = 1. If the interval (a. b) is closed (a<b). + 7 : .e. then. 7 7 JL / 7 a JC ^ fl/j ~ X ^ 3/1 ~H (ii) At the point where x = x 2 in Fig. Thus. *->0 (2) TAe function f(x) is continuous ih any interval if it is continuous at every point of the interval. i. f(x) is continuous at a if f(x)-~>f(a) as x->a + Q and at b if f(x)->f(b) as x-+b-Q."#.

If n is negative. so that x n = (-l) n y n where it is supposed that x n has a real value.1 and x n ~ 1 lie between a n ~ 1 and b n ~ l . hence if k is the greater of the last two numbers. // f(x) tends to a finite limit I as x->a. If then c is any assigned positive number. Now y n is continuous for positive values of y> therefore x w is continuous for negative values of x. then /(z) = Z + 7?. x >a Therefore the polynomial is continuous for all values of x. b so that 0<a<x <6.x 1 < =-. . by Art. a polynomial or the quotient of two polynomials) is continuous for all values of x except for such as make the denominator vanish. also <f>(x) is continuous at x = I. If y = (f>(u) where w=/(x). limx n = 0. If x is any positive number. XIV. 3. and is denoted by </>{f(x)}. function of a function of x. | x n . we can choose a. (1) It follows from the definition that iff(x) and <f>(x) are continuous at x = x . provided that | x . The same is true for the quotient f(x)/<f)(x). If n is positive. r . x n -x n lies between nx n ~ 1 (x~x ) and wx n ~ 1 (x~x ). therefore lim 6. and <f>(x) is continuous at x^lythen lim <f>{f(x)} = cf>{lim f(x)}.x n | <. except when <f>(x Q ) = 0. Theorems on Continuous Functions. Continuity of Rational Functions. 10. Let x be any number other than x in the interval (a. Hence also any rational function of x (i. | x n x w |<&. then by Ch. 7.FUNCTION OF A FUNCTION 271 5. k I n I Hence x n is continuous at x = x . The Function x n .|x~rX |. Now both x". however small. // n is rational. (ii) 7/x<0. x n is continuous for all values of x for which x n has definite real values. If/(x) is a polynomial. For if lim/(x) = Z. (2) Function of a Function. 6) . then y is called a.e. then. where 77 -->0 as x->a. let x= -y. where a is any real number. so also are the sum. (i) Suppose that x>0.|w|. lim/(x) =/(&). Theorem. (iii) At the point x = 0. difference and product of these functions.

we can find 77 so that f(a) . f(x) assumes at least once every value between f(a) and f(b). there are. but not for negative values of n. Hence as x varies from a to b. (1) If f(x) is continuous at x=*a and f(a)^Q. provided that | x . We assume that any straight line which passes between the points A. f(x) has the same sign as f(a).<f(x) <f(a) -f e. y-k a XQ x l x f FIG. if a-r]<x<a + r] where 77 is arbitrarily small. Suppose that a<b and f(a)<f(b). Because f(x) is continuous at x = a and f(a)<k. 6) for which f(x)>k. and let k be any number between /(a) and f(b).f(x) takes the value k at least once. It follows that the line y = k cuts the curve at least once. that is to say.s~>0 x w ->-oo or to -oo. Fundamental Theorems. 272 CONTINUITY OF A FUNCTION 8. Let A. 46. Proof. we can choose e so that f(a)-e and f(a)+e have the same sign as /(a) : and since f(x) is continuous at x a. For such values of x. . 6) for which f(x)<Jc: and because f(x) is continuous at x = b and /(&)>&.a \ < 77. according as x->4-0 or to -0. Hence x w is continuous at x = for positive. For since /(a)^0. (2) // f(x) is continuous for the range a^x^b. B cuts the curve at least once. Let k be any number between /(a) and f(b). there are values of x in the interval (a. then as x passes from the value a to the value 6. From the graphical point of view the truth of this statement is obvious. then f(x) has the same sign as f(a) for all values of x in the neighbourhood of a. B be the points on the curve y=f(x) whose abscissae are a and 6.values of x in (a.

for it is less than x . . Hence it follows that/(# ) = . which is impossible. If f(x) is a function of x such that tends to a limit as h tends to zero. b) is included.) It has been shown that both classes exist.0 if # -<<# -i-e. This is certainly the case whe n f(x) is discontinuous at X = X Q . by reason of the continuity of f(x) at x . Thus f()<k when < -f. 46. The upper class A' is to contain every number x' hot included in A. and every x is less than any x'. yet f()>k for some values of in the range a <<#'. Derivatives. If f(x )~k>0. Therefore the classification defines a real number. 9. b). f(x) has no derivative at X = X Q .DERIVATIVES 273 Divide the real numbers in the interval (a. It is possible that for some values of x. we can find so that f(x)-k<. which is either the greatest number in A or the least number in A '. Thus/'(x) is defined by the equation The function f'(x) is also called the differential coefficient of f(x).f(x + h) must tend to/(x ). Hence x + belongs to class A. (Thus in Fig. then o: . x . we can find e so that/(x)-fc>0 if X Q -^X^X Q + c. and is denoted byf'(x). b) into two classes as follows : The lower class A is to contain every number x such that f()<k where | is any number in the range a<f <#. It is easy to modify the proof to suit the case when /(#) NOTE. as it is greater than x . Thus XQ e belongs to class A'. this limit is called the derivative off(x). is the least of these values. class A consists of the range (a<x<x ). which is impossible. determined as above. If f(x) =k for more than one valuo of x in tho interval (a. Also every number in the interval (a. for although f(x')<k. If f(x )-k<0. If x is such a value. this limit does not exist. The range (#!<#' <# 2 ) belongs to class A'. It remains to prove that /(# ) = k. For in order that {f(xQ + h)-f(x Q )}/h may tend to a limit as h-+Q.

if we choose. If a straight line T'PT exists such that. just after. Notice that (i) Q is not supposed to reach P. strikes the ground at B and proceeds to C. 48. We shall suppose that the curve has a definite tangent PT at P. it is moving in the direction TB and. Let (x. It is supposed that the curve is continuous near P. 1 1 . as Q approaches P from either side. and that this tangent is not parallel . This is sometimes expressed by saying that the tangent to a curve at P is the limiting position of the chord PQ as Q tends to coincidence with P. (ii) The curve in Fig. the angle which PQ makes with T'PT becomes and (as Q continues to approach P) remains less than any angle we may choose. Strictly speaking. 10. We may. This is an instance of a continuous curve which has no definite tangent at one point. so as to approach P from either side. // f(x) =x n . namely BT and BT '. however small. Tangent to a Curve. Just as the ball hits the ground. B FIG. say that the curve has two tangents at B. so that Q may be as near to P as we like. y) be the coordinates of a point P on the curve whose equation is y=f(x). Let P be any point on a curve and let Q be a point which is supposed to move along the curve. PQ would cease to be a chord. 1. then T'PT is called the tangent to the curve at P.Ex. Gradient of Tangent. for if Q were to coincide with P. where n is a positive integer. prove that f'(x) Wo have f'(x)= lim 274 GRADIENT OF TANGENT 47. 48 is supposed to represent the path of a ball which is thrown from A. in the direction BT' . according to the definition. the curve has no tangent at B.

'. Then and since P and ($ are on the curve. Notation of the Differential Calculus. If PT is nearly parallel to OY. /. Let Q be the point on the curve whose coordinates are (x + A. y-f k). a 3 ) is 3a 2 . we have /'(#) 3# 2 . as in Fig. gradient of Pr = lim^4^W'(*)^-~>. 12. and if P tends to coincide with a point where the tangent is parallel to the y-axis. Ex. If f(jc)~x^. at the point whose abscissa is a. Find the equation of the tangent to the curve. whose equation is y = x?. k =/(* + h) -f(x) .h) . In the differential calculus the derivative of f(x) is given a different name and a different notation is used. . the gradient of the tangent at the point ( a. DIFFERENTIAL COEFFICIENT 275 Draw the ordinates PiV.2a 3 . we have y =f(x) and y + k =/(x 4. 1. as Q tends to coincidence with P. and the equation to the tangent is -a) or y = 3a 2 z . and draw PR parallel to OX to meet QM in R. we may expect that/' (x) will tend to infinity. f'(x) will be large. 49. gradient of chord PQ = \ J^^M .*. . Any number nearly equal to x is denoted .to the y-axis. PQ tends to the limiting position PT and h tends to zero . n h Now. QM. /.o fl Hence /' (x) is the gradient of the tangent to the curve y f(x) at the point fey).

We often write . and. v are functions of x. then d . (1) If u.. if <?/=/(#). and is called 2Ae differential coefficient of y ctx j (with regard to x). Rules of Differentiation. du dv . the value of y corresponding to x + Sx is denoted by y + 8y . and the process is called differentiation.we are said to differentiate y with regard to x.by z-f 8x. . ax 276 DIFFERENTIATION 13. In finding -=. thus we have ^ 5y dx X 2/ This limit is denoted by -=^. -~ in the form ~ 7 -/*(o:). % d .. (l) (. where ^~ denotes the operation dx <r v ' dx * of differentiating f(x) : it is also written shortly as -~...-J. T.T-"" T"""T"> dx v ' dx dx ^ ' dx v ' dx dx .. v du dv . .

dx\v/ dx N v / v dx dx N v/ d /u\ 1 du u dv 1 / du dvN d /^N 1 du u dv_l / du dv\ dx\v/ v dx v 2 dx~^v*\ dx dx/ We have (i) fo J3 + l)2-(2-l)8_ 5 ( 'da: (Sar+l)" ~(3a: + l)*' . d . dv du .. . d /1\ I du (Ill) T .(UV) = U -7. ^ ..and -=. .have definite values. Because -5. ~ dx ax and in (v) v^Q. be the increments in w. (iii) 8 (uv) = (u + Sw) ( v + Sv) d ^ + lim(fF + fc)lim s dv dw dx dx ' l^___Sw_^ du dx 1 du u(w-fSu) wlim(u-fSw) w 2 A^V ^ v 1 ^ rf ^ 1N . ^~ are not infinite. Sv> 8(uv).> dx dx dx ax-u/ u ax d /wN 1 / du dv dx\v/ iP\ dx dx It is assumed that -=. (iV) -y. wv.i 5 -T.+ V -7. etc... Sw and Sv tend to zero as 8x->0. corresponding to an increment 8x in x.i .. Let Su. ax ax (i) and (ii). also that in (iv) u j . Proof.. v.. etc. These are left to the student.... ..

7 rr j: *j^ ... u n 55" = !^+! du * ! dx u 3 u^ . ... . 1 dq 1 du 1 dv /!)___+_ an< i / u \_^ = --p dx u dx v ax qdx u ax v ax where p = uv and q = ujv... then \. ' z dx U dx V dx and by repeated application of the rule for a product... ax dx dx result follows. . ...._ POWER OF A FUNCTION 277 Ex. . t> 2 .. -l .. (ii) This follows in a similar manner. then ~f. if each side is divided by p(=uv) t th e . V v^ .. wtare u u 29 . t^. i* n dx rs=1 \u r dx ' SimUarly ^^ = Z 8 ~ m (-~}.. 2. ^ n . 1 dU _ 1 d% 1 rf(w 2 w 3 u n) U dx ~u dx u 2 u B .. v m ).Z zdx (i) If p uv. J V dx i ssc ^\v s dx J . 1 dp 1 du 1 dv . then __. v m . Slww that the rules for the differentiation of a product and a quotient c an be written in the forms .. are functions of x. 1 dz I dU I dV Z_ / / / [/ onH _ _ ~ _ _ _ \J I r cHlU..-V-T. (iii) Let 7 UjU 2 . and the result follows... (iii) Deduce that if 2 = (u 1 w 2 . and.... -_*. u n )l(v t v 2 ..+ u-=..

cr 'u /-\ /--v g rft^ 1 d 2 ^ c?y 1 dw ndy Hence. J w v ' wdx zdx ydx w dx y dx That is. we have *f * ( ^ )=ny -i. Prove that. .. continuous for the corresponding range of values of w. y . w dx y dx y dx y dx y dx (ii) Let n -m. unless y =s.~ = or ~ -r.--. 00. (1) If w=/(x) is a single-valued function of x. y (n fa ctors) .~ . == _^__^. then yisa continuous function of xfor a^x^b. w dx y m dx .-=. Let w y n \ and (i) let n be a positive integer . if y is a function of x..+ -~ -|~ =0. then wy* and w Q =y t) =z 9 say. 278 FUNCTION OF A FUNCTION 14. It should be carefully observed that there is always a factor y-.=.-r . _ .Ex..-=.=.+. 3.. . wy = 1 and . 1 rfw Hence. NOTB.. in all cases . then w=y~~ m . then for all rational values of n. by (i). where m is a positive integer . y ^ . and -~ exists.T .y. Idw 1 dy 1 dy 1 dy ndy f t _.* oo? oo. by (i) or (11). Idw 1 dy m /. then w~y . Function of a Function. continuous for and y~<f>(u) is a single-valued function of u. < __j____ / n terms) = . 1 ^w m dy . ^ w dx y dx w dx y dx P (iii) Let n=p/q where p and q are integers . and multiplying each side by w(=y n ). . i-^^-fl-^ and .

we can find r\ so that \u~UQ\<r). Find ^ . then y=u n .r (B) ax ow ox du dx It may happen that Su = for some value of Sx. --->. Sw->0 oo. As o#->0.= 0. a . and w . ou du Consequently dy . we can find a number 77 such that \y-~y<>\< > provided that \u-u Q \<r)./<>. y Q the corresponding values of u and y. and because /(x) is continuous at x . 8y^8y Su Sx ~8u' Sx w . fi 8u du du . Sy be the increments u and y corresponding to an increment 8x in x. and if F(x) = <f>(u) = <f>{f(x)}.^ v / = lim/-lira^--^-. for otherwise ~ would not be finite : 7 7 U/U hence = 0.and since -r~ is finite. 8y _. ax dx and therefore -+--. (2) If the derivatives f (x) and <f>' (u) exist for a<x<6 and for corresponding values of u. for 8w = 0. (i) when y = (ax + fc) n . _ In this case we must have 8y = 0. 1 . =nu n ~ l .. and then equation (A) does not hold. and =-~ .Let x be any value of x in the given range. Also -^. Thus the equation dy ^dy du dx du dx is true also in this case. then if Sw^0. and consequently \y-yo\<e if | x-x ]<?}' This proves the theorem. 8w dv dw .g. ttati. Let 8w. |=|.. (ii) when y = (i) Let u=ax + b. for any assigned positive number e. then fW-f <) .a=na(ax+b} n ~ l . Because y = <f>(u) is continuous at w . Ex. .

26a. //+.!.*. (x + h) n .hm v .. Now ~T-y q = qy 9 " 1 ~/..1 'and nhx n ~ l .u. then <fy dx Proof.* having proved that the derivative of x n is nx n ~ l when n is a positive integer. // y = x n wfore w is any rational. and by Ch.l) n w n and r- Or. Therefore dy . and differentiate both sides of the given equati on 2k 2y dy with regard to x 9 thus ~ + 7? -5? =0. let x = . ..x n lies between nh(x + A)* 1 . then y = ( . If x is positive. (x + h) n -x n -. x + h is positive for all sufficiently small values of h.. 26x = . and .=( ) . we can prove that this is also true when n=p/q as follows.' - -. XIV. Also a. 71 " 1 is a continuous function of x. a j cr eta 15. u dx \ u*J (a+bx*) 2 SIGN OF FIRST DERIVATIVE 279 &.1 1 i <^V / 1 \ ~. p j 7 Let y = a^.ax h-+Q h If x is negative.ox% then y =.dx du dx v ' /Y * 70. 10. (11) Let ti a -f. ' a 2 6 2 We regard y as a function of x. Derivative of x n . therefore y* = x p .

it does not follow that f(x) increases steadily as x increases through X Q . For {f(xQ-^h)-f(x Q )}/h tends to a positive limit as A->0. Ex.1 /. Higher Derivatives. where x l9 x 2 are in the neighbourhood of a? .= px*>. 17. For if x t <a. f . /(x + A) ~/(x ) has the same sign as A.dy . 280 MAXIMA AND MINIMA 18.. 13 obviously hold for such functions.x ^ ^ dx ^ dx qy. Hence for sufficiently small values of A. Complex Functions of a Real Variable x. If the derivative oif'(x) exists. 2 <a. Similarly t// / (x )<0. The rules of Art.. 3. it has pnly been proved that f(Xi)<f(x ) and f(x 2 )<f(x ) t and we cannot conclude that f(xi)<f(x 2 ). Iff(x )>0 9 then for all values of x in the neighbourhood of x . 13... _ dy px p ~ l p x p ~ l p --* therefore qtf. /or every x in JAe neighbourhood of x .1 q Z^v q x q 16. If y =/(#)> this is written in any of the forms So by n differentiations (when possible) we obtain the nth derivative or the nth differential coefficient of f(x) 9 which is written in any of the forms .)>0 only for the single value # of #. See also Art. it is called the second derivative or the second differential coefficient of f(x). /(x) </(x ) according as x ^ x . Meaning of the Sign of f'(x). = . then y is called a complex function of x and its derivative is f'(x) + i<f>'(x). NOTE.= .. z S a5 5 x . If /'(a.1 and -/ = . If y=/(x) + i<(x).

if a number exists such that f(x Q )>f(x) when then /(x ) is a maximum value of /(x).1 . . a n $x. Maxima and minima values of a function are sometimes called stationary or turning values. In other words.. A necessary condition that f(x Q ) may be a maximum or a minimum value of f(x) is that f'(x Q ) =0. + a n Or. If in the preceding.1 + . a n _Jx. .1) . as a increases through x . 1. l) w . then f'(x) = n(a . using the notation of Ch. and by successive differentiation Hnfn. then f(x Q ) is a minimum value of (fx). Suppose now that /"(# ) exists and is. (a^-f a x ).2 + (n " 1) ^" 2) a 2 ^~ 3 + .. 16. Hence by Art. .l)a^. Ill. We say that/(z) has a maximum value at X XQ when f(x Q )>f(x) for all values of x in the neighbourhood of z .3.. a l9 a 2 .a n) L? then /' (a?) = n f a^.. I)*. f"(x Q ) must be negative. Maxima and Minima.1 . This follows from Art.( ~^~1 a 2 x n ~* + . 19. following are important instances. 4.2.. TEST FOR MAXIMUM OR MINIMUM 281 If f(x ) is a maximum value. not zero.1 + (n . . ). the sign of f'(x) changes from -f to -. if f(x) = (a Qy a v a 2 . 16 . /(z + A)</(a. . (ii) If f(x) = a^x n + na^*.The. (i) If t/ = z n . the sign > is replaced by < .. . . Then since f'(x) decreases as x increases through rc . but the condition is not sufficient. then for sufficiently small values of A. -t = nx n .

as x increases through z . Here -=^ = 3z 2 . If the sign of / .(x) does not change as x passes through x . a maximum nor a minimum when x=0. f(x) is neither a maximum nor a minimum. FIG. Search for maxima and minima values of Here /'(*)=6* a -6*-12=6 Thus /'()=0 if *=-l or 2. but as x passes through 0.Similarly if f(x Q ) is a minimum value. a minimum or neither of these. In this case /(# ) ma 7 be a maximum. the sign of -~ does not change. the sign of f'(x) changes from + to . 282 CALCULATION OF LIMITS .6 =6(2* . Or thus. the tangents at A. leading to the same results as before. The case in which /" (x ) = will be considered later. and as x increases through 2. The sign of f'(x) for different parts of the curve is as indicated.1). Ex. B 9 C to the curve y ~f(x) are parallel to OX. Thus /( . the sign of f'(x) must change from .to 4. and f(x) has maximum values at A.. it must be positive. so that -~=0 dx ax dy when =0. 51. in Fig 51. C and a minimum value at B. f"(x) = 12s .. Fig. As x increases through -1.1)<0 and /"(2)>0.to + . the sign of /'(#) changes from . 52 represents the graph of t/ = x 3 -f 1. To illustrate this geometrically. ax Thus y is neither.1) is a maximum and /(2) a minimum value. and if f"(x Q ) is not zero. 1. so that f"( .

20. the secant QP cuts the curve at another point Q' near P. . ax . dx* If Q is a point on the curve near P. and we may regard the tangent TPT' as the limiting position of the line QPQ\ when the three points Q. (ii) as x tends to oo . In Fig. and Q' are very close together. y) on a curve at which j. lir ' 3. Thus ~ has a maximum value at P. dx* EXERCISE XXVIII Find the limits of the functions in Exx. A point (x. 53 APE is supposed to be part of a curve represented by y =/(#). Points of Inflexion. 2 and therefore -~= at P. -~ 0. Q FIG. Find the values of 4. 1-3 (i) as x tends to zero . If a point moves along the curve from A to B. P. 1. and the curve crosses the tangent. the gradient of the tangent increases along the arc AP and then decreases. 53. Definition.is a maximum or a minimum is called a point of inflexion. dty At such a point.

(l- 5. n are positive. if m.. r Bm(x + h) -sin a. . Iim 8.sin x y -=. [sin (x -4.tan x = sec 2 x. hm sin a: 9. v L = _ . Iim . Prove that sin x and cos x are continuous functions of x. If f(x)=z(x-a) m (x-b) n . 14. prove that d \ r Hence.cos x .. 10. (ii) Iim . 2 sin %h . f'(x) vanishes for a value of x between a and b. For what values of x is tan x discontinuous ? 12. 1.h) -sin a. 6.sin x = cos x 9 -=. Prove that (i) Iim cos a. CO8 (x -f \h) ~> cos x as DERIVATIVES AND TANGENTS 283 13.= 1. Prove the following : (i) . Prove that -7. =2 sin^A cos 11.(1 +*).

If y=A cos |Lia.l a/a:. If y = cos x -f sin or. + J5sin IJLX. Find the equation to the tangent at the origin to the curve whose equation is y 22. 16. . If P.] 21. Prove that the function 2x 3 .5 always increases with x. rfy 2 ^ aa: 2 / \^/ r7 2 y 18._ . Find the derivatives of (iii) (iv) sin n z. a . (ii) if x 3 + y* = (Z-iC . dx 20. then T-O " M 2 2/- 19. Q. then -~ = iv. a + k respectively. (vi) tan n tf. prove that the tangent at P is parallel to the chord QQ'. Find -% if (i) ax 2 + 2hxy + by* + 2gx + 2fy + c=Q. (v) cos n #.. cZ 2 o.3x 2 + 6x . Prove that if = tan.+ ~ l) ' 15. 23. d*y //dy\* 17.10) I / (x + ia). then [Consider the values of 1 / (x .-^ / ( / . Q' are points on the parabola y ~px 2 + qx + r whose abscissae are a.h.dxj dx (d \ n 1 _1 / _ jl 1 } (m) (dx) l-x*-2\"((l. Prove that 1 = .

C. there is a single value X Q of x such that yg=f(x Q ). Prove that x~ -b/2a gives a maximum or a minimum value of aa. We shall prove that (1) The equation y=f(x) determines x as a single-valued continuous function of y. Prove that the right circular cylinder of greatest volume which can be sent is 2 feet long and 4 feet in girth. and show that one of the values of x found in (i) gives a minimum value of f(x) and that the other gives neither a maximum nor a minimum. 31. and find the greatest value of x(x. Let f(x) be a single-valued function of x. continuous for the range a<x<6. The regulations for Parcel Post require that the sum of the length and girth of a parcel must not exceed 6 feet. for what values of x does f'(x) vanish ? (ii) Sketch roughly the graph of y=f(x) from # = to x = 2. Show that the height of the right circular cylinder of greatest volume which can be cut out from a spherical ball of radius r is 2r/\/3. Thus. Further. 2 -f bx -f c according as a ^ 0.2). A. Since f(x) is continuous. 2 . Show that the turning values of (x-l)(x -2) (2 -3) are approximately 0-38. suppose that throughout the range. corresponding to any value y Q of y between /(a) and/(6). 284 INVERSE FUNCTIONS 27. (i) When does the function 3x* . 29. between and 1. f~ l (y) is continuous.36a. (i) If f(x) =(x-l)*(x. Sketch roughly the graph of y = x(x~ I) 2 . Also f(x) takes every value once only. f(x) takes every value from /(a) to f(b) at least once. This function is denoted by/~ 1 (j/).I) 2 for values of a. either f(x) increases with x. and state the character of each. Find the volume of the box of greatest capacity which can be formed in this way. The rest of the sheet is bent so as to for m an open box on a rectangular base.4# 3 . T B. 30.24. 21. 8 inches long and 5 inches wide. or decreases as x increases. Four equal squares are cut off from the corners of a rectangular sheet of tin. so that the statements y=f(x) and x=/~ 1 (y) mean the same thing. 26.1 increase with x ? (ii) Find the turning values of the function. for f(x 1 ) =f(x 2 ) only when x 1 = x 2 . 25. Also the function/" 1 is called the inverse of/. taking . For. as x varies continuously from a to 6. Inverse Functions. Again. Proof. 28.

provided that | y and therefore x =/ -1 (y) is continuous at j/ . To show that/~ 1 (y) is continuous at y=zf(a) and y =/(&).. Joi. 6). we take y Q + k and y . Hence | x ~ x \ < ..the case in which y increases with x. we consider the variation of y for the ranges (2) // -~. increases with y. and vice versa. 8y ax .and + are within the interval (a. Therefore dx . and let 77 be the smaller of the two k.T ^dy by . then.. so that For sufficiently small values of c. let y Q be any value of y between /(a) and /(&) and XQ the corresponding value of #. k'. x . 54. PRINCIPAL VALUES 285 If y decreases as x increases. /. As y varies from y -7? to y + 7 ?> x ^ es between # -e and XQ + C. ox . Sv .is finite. Idy ~. for a.exists and is never zero for the range a<x<b. ^ ' J dx J y dy I dx For if Sx and 8y are corresponding increments of x and y.. since as Sz->0. . then ~.k' as the values of y corresponding to x ~ and X Q + . FIG.= 1 I ~. Let y Q -k I X Q ~6 X Q X^ X and y + k f be the corresponding values of y.

5^>. for these. A similar definition applies to tan^ 1 x.Si" 1 lim^ = l -. the derivatives of sin*" 1 a. dx 22. Thus the principal values of sin" 1 x and tan" 1 x lie between .tan~ l x = . dx l+x* dx Let y = tan~ 1 a.. The Inverse Circular Functions. Unless otherwise stated. 55. . The smallest positive value of y such that cos y = xi$ called the principal value of cos~ l x. and that of cos" 1 x lies between and n. The numerically smallest value of y which has the same sign as x and is such that sin y = x is called the principal value of sin~ l x. Show that -=. sin" 1 ^. Y Y Y jy=s//i" l jr FIG.. Ex. as can be seen from Fig. then x=t*ny and -=-=8 dy dy 1 ' te = r+?' ..and .. cos" 1 ^ and tan"" 1 ^ will be used to represent the principal values of the functions . and cos" 1 ^ are positive and negative respectively. I .

x) and (if our supposition is possible) not bounded in (x. (ii) at least one value of f(x) exceeds any number less than h. A function which is bounded above and below is said to be bounded. exceed any number we may choose. b). If a number N exists such that/(x)>2V for every a. Now f(x) is continuous at x=a. a number I exists such that (i) no value of f(x) is less than I . b). f(x) lies between f(a) .f(x) is said to be bounded below. If 3=0. According to the supposition. Divide the real numbers in (a. 6). where is any small positive number. Theorem 1. I. But f(x) is not bounded in any interval including zero. x). we can show that a number h exists such that (i) no value of f(x) exceeds h . Ex. b). Because f(x) is continuous at x = a. we can mak e I/a. then/(x) is said to be bounded above. if f(x) is bounded below. Let S denote the aggregate of the values of /(x) as x varies continuously from a to 6.286 BOUNDED FUNCTIONS 23. (ii) at least one value of f(x) is less than any number greater than I. This number h is called the upper bound of f(x). for by making x small enough. If f(x) is continuous in the closed interval (a. b). XIII.c and /(a) + e. b). 6). therefore for sufficiently small values . f(x)=0. If x^O. n-^-oo 1 +nx 2 f(x)~ljx. Hence there are values of x in (a. Thus f(x) has a definite value for every value of x. which is the greatest number in A or the least in A'. in (a. The classification therefore defines a real number a. Consider the function f(x) = lim . This number I is called the lower bound of f(x). Let f(x) be a function which has a definite value for every value of x in the interval (a. for sufficiently small values of x . Suppose that f(x) is bounded above in (a. 12) to the set S. it is bounded in (a. By applying Dedekind's theorem (Ch. If a number M exists such that/(z)<lf for every x in (a. both classes exist and every number in A is less than every number in A'.a. Suppose that f(x) is not bounded in (a. Bounds of a Function. b) into a lower class A and an upper class A' as follows : The number x is to be placed in A or in A' according as f(x) is or is not bounded in (a. Similarly. 6) such that f(x) is bounded in (a. 6).

B. 24. which is impossible. Therefore h=f(x) for some value of x in (a. Hence l/{h-f(x)} is not bounded. 6) and h. b) for which f(x)>h -. Similarly there is a value of x for which f(x) = I. 6). I are its upper and lower bounds 9 f(x) takes the values h and I at least once as x wries continuous ly from atob. Fig. But if h-f(x) does not vanish for some value of x in the interval. Theorem 2. A strict proof (as given on the next page) depends on the rather difficult considerations of the last article. Here f'(x)= -GO at . where is any small ROLLE'S THEOREM 287 positive number. but from the graphical point of view its truth is obvious. Hence f(x) is bounded in (a. b) and has a derivative f (x) for every x such that a<x<b. l/{h-f(x)} is continuous in (a. Rolle's Theorem. This proves the theorem. for a + >a. FIG. then f'(x)=Q for at least one value of x between a and 6. For if c is an assigned positive number. 57 represents the curve j/ = l -x$. Suppose that f(x) is continuous in the closed interval (a. 6). Let the curve y=f(x) cut the x-axis at A. however small. and that at every point it has a definite and finite gradient. The theorem asserts that there is at least one point on the curve between A and B where the tangent is parallel to the x-axis. 57. a-fe). f(x) lies between /(a-e) and /(a-f e). there is at least one value of x in (a. so that h-f(x)< and 1/{A -/(*)}> 1/c.of x-a. Suppose that it is continuous from A to B. If /(a) = and f(b) = 0. and so belongs to A. 56. FIG.* // f(x) is continuous in (a.** This theorem is of the greatest importance.

for the case in which f(x) is a polynomial./t) must be continuous in the open interval (a.. which is impossible... Iff(x) is continuous for a<x<6 and f'(x) exists for a<#<6.1 and + 1. b). Moreover. &) . for otherwise there would be values of x in the neighbourhood of h for which f(x)>h... Either (i)/(z)=0 for all values of x in (a..... is given in the next chapter. .. or (ii) some values of f(x) are positive or negative.. 6).. then m -f(a) = (b-a)f ().. ..-- -/(a)}.. unequal to /(a) or /(6) and /'(Z)=0.. in which case /'(z) = for every x .f'(x) is not zero for any value of x between . A simpler proof.. This proves the theorem. The Mean-Value Theorem. Let <f> (x) =/(&) -f(x) .. If there are negative values of f(x). b) and attains at least once its upper bound h and its lower bound /. it is bounded in (a.. then I is negative. (A) where is some number between a and b. there would be values of x near I for which f(x)<l.<&.... For if f'(l)^Q. f'(h)=*Q. Proof.. 25.. and * This proof is taken from ** If f'(x) exists for a<a.. Because f(x) is continuous in (a. If positive values of f(x) exist.. 288 MEAN-VALUE THEOREM Proof. which is impossible. h is positive and unequal to /(a) or /(&). but not necessarily in the closed interval.

then <f>( x) and t/j(x) differ by a constant. <}>(x) -i/f(z) = a constant C. (i) f'(x) is always positive. AREA UNDER A GRAPH 289 26. b are any two values of x in the interval. The theorem asserts that there is a point C on the curve between A and B where the tangent is parallel to AB. f(x) decreases as x increases . throughout any given interval. where it is supposed that a. Such a function is called an integral of f(x) (with regard to or). Corollary. then f(x) is constant throughout the interval.b respectively. (ii) if f'(x) is always negative. by Art. (iii). it is easily shown that Fia. . then f(x) increases with x . Geometrically. and is denoted by \f(x) dx. Also ^(a) = and <f>(b)=Q. B be the points on the curve y~f(x) where x = a. 25. Suppose that f(x) is a given function of x. NOTE. which proves the result in question. (ill) if f f (x) is always zero. Let A.'(x) = -f(x) {/(&) -/(a)}. I ntegration . and the ordinate NP (or NP produced) meets AB in Q. 58. y) is any point on the curve.then <f. //. hence by Rollers theorem <f>'(x) = Q for some value f of x between a and 6. All this follows from equation (A). If <f>(x) and iff (x) are any two functions such that <f>'(x) = $'(x). therefore. If P(x. It has not been assumed that /' (x) is continuous. and that we try to find a function ^>(x) such that </>'(x)~f(x). Cor.

Now S=Q when a. also known as the General Mean Value .. NP and the arc OP is equal to \l(n + 1) of the rectangle contained by ON. S corresponding to the increment ox in x. and <f> (x) = f(x) dx + C. 59. dx v ' J n + 1 NOTE.*. !f /b: 11 " 1 " 1 y rfa: = I A:a. Ex. just as -j. 4 Integration ' can be applied to find the area of a curve as in the next exampl e. PM<oS<Tect.+ C. then rect.-j. prove that the area ONP bounded by ON. FIG. Denote the area by S and let #. 8/9 be the in crements in y. Taylor's Theorem. The symbol . 2.(7. Let Sy. are two equations which mean the same thing. y = .. These are inverse operations. = ^0 and 5= 290 TAYLOR'S SERIES 27. where C is a constant. . the general value r /i r is f(x) dx + C. Since ^ x n ^ ^ ( n + 1 ) x n . 1 . // PN is the ordinate of any point P(x 9 y) on the curve represented by y ~kx n where n>0. O Of JO therefore y < . dx is to bo regarded as denoting the operation of integraj J tion.Hence if any value of (f>(x) is denoted by /(x)rfx. QN. i.denotes that of differentiation. y be the coordinates of P. Thus j-</>(x) =/(#). Ex. If Q is the point (x -f ox t y + Sy) and QM its ordinate.< y + Sy . n <fo: = -. therefore (x n dx .. NP.*.. and since Sy -> 0.e.

/"(#).a)f () + -(&. therefore by Rolle's theorem </>'(#) = for some value ^ of x between a and 6. If f(x) is continuous in the closed interval (a. then equation (A) may be written . f"(x) . Taylor's Series.../ (n) (x) exist for all values of x between a and b..h"f w (a -f 9h)->Q as n-> oo . then ... Suppose that f(x) is continuous in the closed interval (a-c. . . .Theorem. and the n derivatives f'(x).x)"- and f (x)= - Now <f>(a) = Q and ^(6)^=0. Then equation (B) holds..) 2 /" (a) + . In general.. Let b = a + h.+ . then f(b) =/() + (6 . The proof is similar to that of Art. a + c).(B) /^ i. the value of depends on n. FUNCTIONS OF A COMPLEX VARIABLE 291 Further suppose that R n = -. hence which is equivalent to equation (A). where J2 B = pA n /< w >(a + 0A) and 0<0<1. Let and then F' (x) = . b) and the first n derivatives f'(x)./ (n) (^) a/wi /Aa< | h |<c. 1 1 |>*-1 f/ " i^ where is some number between a and b. 25.p 1 (b .. For any number between a and a-h/* may be written in the form a + Oh where 0<0<1.

e..vL " . but 2x -t.. without the explicit use of x or y.. It follows that if one series of operations determines X and Y in terms of x and y. The series in (C) is called Taylor's Series and R n is Lagrange's form of :he remainder after n terms. By this we mean that the point z is to move along a continuous curve from the point z to the point z v Thus the variation of z is continuous if the variations of x and y are continuous.... . where X and Y are functions of x and y.to*>.. Consequently X.. (D) 11 This expansion is known as Maclaurin's Series. We use the equation Z=f(z) to indicate that Z is the result of definite operations performed on z. so is 2x + 3iy . we say that Z is a complex function of the real variables x and y.. independent of one another and each capable of assuming any values from oo to +00.... Thus 2x -f 3iy is a function of z in the first sense. Hence it would be perfectly reasonable to call Z a function of z. As z passes from a value z to a value z v we suppose the variation to be continuous.. z is known.(C) For if s n is the sum to n terms of the series. Here x and y are real variables.to . so also are x and y.. 29. i.A) . For if. Function of a Complex Variable z. but not in the restricted sense. and we say that Z is a function of z. The Quantity x-f ty.. THE COMPLEX VARIABLE 28... the variables x and y not occurring explicitly. Y and Z are known. But it is usual to restrict the meaning of the term ' function of z ' as follows. which we denote by z. . is called the complex variable.. If Z = X + iY. When a 0.3iy cannot be expressed in terms of z alone. If z is known.. equation (A) becomes V(0) + 1 AT / (0) + . /(a + h) ~ s n = 7? n -> so that s n ->f(a -4...

where a is a fixed number. Continuity of Rational Functions of z. (3) The function /(z) is continuous for z = z or at (the point) z if for any positive e that we may choose. is to say that 2-a->0. the point z (on an Argand diagram) is within a circle with centre and radius . Hence if z describes a continuous curve in the region of continuity of the function /(z). 77 can be found so that | /(z) . however small. All that has been said in Arts. holds for the complex variable. In particular : (i) lim (o + az + a^ 2 -f . for any positive that we may choose. . . the point z is within a circle with centre a and radius c. with regard to the real variable. which may be as small as we like. provided only that | z . (1) To say that z tends to zero (z->0) is to say that z varies in such a way that | z \ becomes and remains less than any positive c we may choose. we can find 77 so that I f( z ) ~f( z o) I < > provided only that | z . the point /(z) will also describe a continuous curve. 2 and 3. 31. After a certain stage.292 CONTINUITY OF RATIONAL FUNCTIONS 30. however small. It is assumed that /(z) has a definite value at every point in the neighbourhood of the point z . EXPANSIONS IN SERIES 293 EXERCISE X3TTX . This is the same as saying that as z tends to z along any path whatever. Thus.z 1 <^. however small. + a w z n ) = a . after a certain stage. as z approaches a by any path whatever the distance of the point f(z) from the fixed point I becomes and remains less than any length e that we may choose.a \ < rj. which may be as small as we like. (2) We say that /(z) tends to a limit I as z tends to a if. (4) The function /(z) is continuous in a region A if it is continuous at every point of the region. however small. In other words. (iii) A rational function of z is continuous for all values of z except for those which make the denominator zero.1 1 < e. Definitions relating to Limits and Continuity. To say that z tends to a (z->a). z-o (ii) A polynomial in z is continuous for all values of z. /(z) tends to /(z ) as a limit.

dui Now -7 = ~v >0.. X 3 X* (ii) sina. (ii) tan x-x increases with x in the intervals (-lU. ax + b. ax 3 + 3bx 2 + 3cx + d and ax 1 + 4bx* + bcx* + 4<fo 4. rfv and ^-=w 1 >0. and so on. Prove that for all values of x X 2 y* (l) COSX = l-+r-. Therefore s r -* cos x as r -> oo . ^!>0. /. /.+ r^ . . . ITT). ax 2 + 2bx + c. (w a . 2 ). U B respectively. for every x. |TT). [(i) Let t r be the rth term of the series arid s r the sum to r terms. If cos.. -land xi_ x du n dv n prove that ^=.. (71-. s zn < cos x < s 2n+l and s 2 +i . t. %>0.. x 2.s zn tzn+i ~> as ?i ->. / re 2 3.. if C7 = w 2 ^s -3u u l u 2 + 2u l 9 9 F=^o w 4~4ii 1 w 3 + 3% 2 .*.1<0. .*.. then -7. v = sin x .] 4. If a. = x.decreases as x increases from to TT.v . show that.. % is an increasing function. then w n =0 and v n = 0. [If #=0. also ^ = cos x. v x is an increasing function. in succession for (u 2 ..=0 and =0..1. Then by Ex.o o .to oo .^ = M ' Hence show that u n and v n are positive or negative according as n is odd or even. tO QO.] . (iii) . v 3 ). u l9 u 2 . 3. so that U and F are indepenax ax dent of a. .e are denoted by w . Prove that (i) sin x-x always decreases with x.x < 0.

. -f. then o o JLf 1 /(0)=0.cos x + sin a.] x 2 x 3 = x-j + ---. o 5 7 {Let S r =x-~ + ~-. then x* x 5 x' tan" 1 . If f(x) is a function whose first derivative is I/a. 4 from Taylor's theorem. 22. Use Taylor's theorem to prove that /.1 *-^.4-(-l) r .tooo.. that./ rv .T W & (i) sin (x + h) =sm x + h cos x .] 8.. and/ / (x) = l/(l-fa: 2 )-(l-x 2 4-a:*~. [Proceed as in Ex. ^.. .. tan~*aj lies between S r and S r+1 . show that. Deduce the expansions of sin 2 and cos x in Ex.4.*. and /(s^tan.h) = cos a. and | 8 r . Hence. .. as r -> oo .r^ cos x + . 7. .x . 2i 3 [(i) If f(x) = sin x. 294 CASES OF BINOMIAL THEOREM 7. so long as x is positive and not greater than unity... |^ n |<-|^ n |->0as n-*oo. ..1 x 2r ~ 2 . if 0<a?< 1. if 0<a:^l. Ex. = # .] 6.+(-l) r ~i .. by means of Art.S r+l 1 -> 0. 2i o h 2 h 3 (ii) cos (# -f.. 1....A sin a: .5. to QO.sin x .-f -. Prove. so tha t which is positive or negative according as r is even or odd.

j8 be consecutive real roots of f(x) = 0./(0). /'(O).9. ^ ^ z 2 + . so that .-.... show that the expansion of (a+h) n by the Binomial theorem is given by Taylor's series. prove that tan 6 lies between + i0 3 and + f0 8 ..] The Binomial Theorem 10..isanr-multiplerootoff(x) = Q.. show that /(O). are all positive. another form (named after Cauchy) of the remainder after n terms of Taylor's series must be used.+ .1 <#<().. let /(0) = + a0 3 -tan .... to oo .. * For /(#) = (#. See any treatise on the 4 Calculus. at least one real root of f (x) = lies between any two real roots of f(x) = 0. RATIONAL FRACTIONS (1) 1. / // (0).. <f>(x) where <f>(x) is a polynomial and 2.. (A) [For r>n. /"(O). Proof. in succession. are all zero... If n is any rational number and .-. so that in this case . Multiple Roots. R r may not tend t o zero.'] CHAPTER XVIII THEORY OF EQUATIONS (2).. POLYNOMIALS (2). To deduce the expansion (A) in this case. and that /'"(0)^0 throughout the interval 0<0^7r/6 if 9a>8 .a) r . explain why we are unable to say that ft r -+Q.. prove that J r ->0 as r -> QO . 11. then (x <x. hence.' x T -* IT 1 where .. If 0<0<7r/6.- (ii) If 0<a:< 1...1 <x< 1. [Here ( 1 + Ox) n ~ r > 1 for r>n. Let a.. n(n-l) .. If n is a positive integer.r <l and "' x r ->0 as r~>oo. If<x. and so far as we can tell.] (iii) If ./ / (0).. ..) r ~ l is a common factor off(x) andf'(x). // f(x) is a polynomial.. .(n-r + 2) ^ .. [For the second inequality. where f(x) is a polynomial.. (l+0x) n . Rolle's Theorem for Polynomials. (i) show that n(n-l) .

a root of f'(x) lies between o^ and a 2 . For let j8 x . not more than one root of f(x) can lie in any one of the open intervals . . Hence (i) if j3 1 <a 1 <a 2 <jS 2 . Then <f>(x) is of invariable sign for oc^a^jS and /'() = (x .. or (iii) be greater than the greatest of them. ]8 r be the real roots of /'(o. . then j8 x and /J 2 cannot be consecutive roots . so also are those of /'(x)=0. .. so also are those of f (x) = 0. Let o^. f'" (x) = 0. . Hence \fj(x) and consequently /' (x) is equal to zero for some value of x between a and /?. then ^ cannot be the least root of f'(x)=Q .f(x)=(x-*r(x-p). . then a x is one of the set j8 1? j3 2 .) If 0^:56 a 2 . and <f>(x) is not divisible by x-at or #-/?. (ii) if a 1 <a 2 <j3 1 . and the roots of any one of these equations sep arate those of the preceding equation. /' (x) is of degree n-1.<j3 r . (2) // all the roots of f(x) = are real. by Rollers theorem.a) where Now 0(a) = m(a-j8)<(a) and Therefore 0(a) and 0(jS) have unlike signs . Deductions from Rolle's Theorem. and t he roots of the latter equation separate those of the former. (Art. and suppose that jS 1 <j8 2 <. then j8 r cannot be the greatest. 1. ift(x) is everywhere continuous. (iii) if j8 r <a 1 <a 2 . moreover. 3. ( 1 ) // all the roots of f(x) = are real.1 intervals between the n roots of f(x) = 0. This follows from (1). 296 DEDUCTIONS FROM ROLLE'S THEOREM (3) Not more than one root of f(x) = can (i) lie between two consecutive roots off (x) = 0. any of which may be multiple roots. n are positive integers. = j8 r . For if f(x) is of degree n. a 2 be real roots of f(x) = 0.. f" (x) = 0.. Thus.) = 0.<i>(x) where m. and a root of /'(x) = exists in each of the n . or (ii) be less than the least of these. /J 2 . If a x =a 2 .

one between -1 and 1. .. j8 2 ..a) (x . of f'(x)=Q are known.. then (i) s r +p 1 s r _ 1 +p 2 s r _ 2 + ... 1. or the same sign.. . 2 -f-26a...)..p n =0...(a 3i -f26 + c) and the equation oa..oc< 0... . ...2 + . 1. 1. none of the roots of f'(x) = is a root of /(x)=0.3. + c =0 has two imaginary roots. (5) If f (r \x) is any derivative of f(x) and the equation / (r) (#)=0 has imaginary roots. c. and the theorem follows from (3). 2.(4) // /'(x) = Aas r razZ roote. k $ #... . // 6 a . Sums of Powers of the Roots of an Equation Newton's Theorem.. we can find the number of real roots of f(x)=Q by considering the signs of /(&).... See Ch.6* 2 + 24x + 7 = 0. If f(x) = has an in-multiple root...(a? . according as f(Pi) and /(/J 2 ) have opposite signs.. For /(->(*) = ?i(tt-l) . (B) . so that $ = n.j8) . (i) We have identically f(x) . NEWTON'S THEOREM 297 4. Ex. Here f'(x) = 12(x*-l)(x-2). and two imaginary roots. 1 ) w ~ faw a4 tecw* (uw imaginary roots. (A) and let s r ==a r + j8 r -h. j3.. /(/J 2 ). ... . <Aen f(x)=0 cannot have more than r + 1 If /(#) =0 has no multiple roots..+s l p r _ l +rp (fi) * r +Pi*r-i+Px*r-*+. 4 . * -00 -1 1 2 00 /(*) + .. (6) If all the real roots j8j.. + K T (r = 0.1. lies between j3 x and /? 2 .. 2. or no root. 6. 2..&E 3 ..+ + + and /(#)=0 has a real root < . Find the character of the roots'of f(x) = 3a.. It follows from (4) that /(#)=0 has at least as many imaginary roots as/'(z)=0... we regard this as the limiting case in which m roots tend to equality. Thus the theorem is true in all cases. Let a. Ex. AC be the roots of f(x)^x n +p l x n ~ l +p z x n ../c)... then f(x) = has at least as many.. A single root of /(x)=0...+Pn*r-n=<> Proof. . The roots of /'(*)-= are . (x .. 11. VI... then the equation f(x) = (a.1. . 1. +.. Ex.

2)^ 2 . . ^ .+p v A2=a?+p l <x+p 2 .. 2 .... ^ 3 =a 3 -hjo 1 a 2 -hjo 2 a-f j 3 . Q 3 = s 3 + p^ 2 + p^ -f np 3 . in succession . . .1 ^^ Q 2 .....V Therefore 2 + --+^-. by addition. . substitute a.. In this equation. . we have . in the expansion of (1 4-x) r ... and so on ..a) = x n ~ l + Atf".3 4-..1Now f'(x)nx n ~ l + (w-l)p 1 x n *~ 2 -h(n-2)^ 2 x n .2 -f where the coefficients are given by A l =<x.. (y~ z) r > et c. 2. K in succession for x and add. Hence the second result follows by addition.. are the coefficients of x. . theref ore. 298 ORDER AND WEIGHT Ex. 1.. -f. Q n =p n . +y n -v If the other quotients in (C) are expressed in a similar way.s l + np v Q< z . in terms of S Q . s l9 s. . p.. find the value of 27(a-/?) r . -4 n _ 1 =a n ~* 1 +j9 1 a n ~" 2 -f ..(n . . assuming the rule for differentiating a product (Ex.. hence. and Q n ^ . and similar equations hold for each root. . Similar equations can be deduced from the expansions of (/3-#) r ..v-2*i + w ^n.. and finally. wher e r is an even number..1 )Pn-l = (E) (ii) Let r^zn. . f(x) / (x . then (a .qxa 7 "' 1 + C 2 z V~ 2 ~.x) r = a r . y. For equation (A) of the last article.* n-1 -f jv n _ 2 -f . + .. ... If q..^ rt a r ...+x r . and so on.w = 0.2 -f .. 277). and comparing this with (D)..and.1 + p 2 a 7 *. then multiplying each side of /(a) = by a r ~ n we have a r -h fta 7 ". it will be seen that ' -* "-* -* + . c 2 . (D) where Q l .^ 2 -f p^ L -f np 2 . . /?. x-oc x~ By synthetic division..+Q n . +|> w _! ... x 2 .(w .

. p n . The equation becomes y n -f A^y"" 1 + A 2 j9 2 y n ~ 2 4. and in equation (A) write x = yl A. and that of ZajSy is 3.. the sum of the suffixes of every term. /J..* is equal to the weight (w) of u. let Let ^ be expressed in terms of the roots by means of the equalities Consider Jfo highest power of any root a which occurs in this expression for u.. Proof.. |8. ...A n p n . /i/s alowc : also the index s is the degree of this term. and denote it by a*.. .....l + ..... A/3. and the roots of this are Aa.. of degree s. therefore a 5 arises from the highest term Pi 2 p% and /row. in which the highest term is of degree s and Pi = i/a > ?2 ==a 2/ a o e ^ c The weight of a symmetric function (u) of the roots a. For if the equation is written in the form (A)...... involves a to the first degree only.... ... Ay.... for example... ... . namely 3. PARTIAL DIFFERENTIATION 299 Theorem...l +p& n -*+.... .... be the roots of x l n +p 1 x n . The order of a symmetric function of the roots of an equation is defined as the index of the highest power of any root which occurs in the function. Order and Weight of Symmetric Functions. . u is a function ofp l9 j> 2 .. when written at full length. a l9 a 2 . . Each of p 19 p 2 .5.... From what has been said it will be clear that the order of a symmetric function of the roots is the degree of the function ivhen expressed in terms of the coefficients p l9 p^ .. Thus the weight of Za 3 /? is 4.. 4..... . Thus the order of 27a 3 /8 is 3...... Leta.. // a symmetric function (u) of the roots of equation (A) is expressed in terms of the coefficients p v p z . y .. (B) then a Q s u is a homogeneous function of # . j8. 0... .+p n = Q . If u is a symmetric function of order s of the roots of atfv n + a l x n .. is its degree in a. ..... Denote any term in u by Pi h p 2 k p^ 1 .+a n = Q... and that of Za/3y is 1. (A) Consider any function of the coefficients . .. .....

y. Ex.y.. by / which proves the theorem. y. 3 fdu\ d*u . p 3 . and ^~ is called the first partial derivative of u with OX CfX -. p^ by A. . A 3 . 6. the effect of this is (i) to multiply u by \ w . j.. respectively. oc 2 p 2 is of order 2 and weight 4. where a. For example. z. the context implying that the differentiation is partial.. 1. I _ 1 - ___ _ .z. Write down the literal part of Za 2 /? 2 . We also write d 2 u .bp 2 2 -f cp^. c are independent of p lt p 2 . A 2 .nereiore ^ ^ (xx) ox oy oy ox It follows that equation (A) holds when u is a polynomial in x. respect to x.A. if u = x m y n . .C. The operation of differentiating u with respect to x.) is a function of the independent variables x. * Thus the term 2>i 2 j> 2 3 is written PiPiPzPzPzU B. (ii) to multiply piSpz. on the supposition that x alone varies. Partial Derivatives. _ dxdy dy\dx/ with similar notation for higher derivatives. is denoted by the symbol ^~ . (iii) to multiply pfp^pz 1 .If then we multiply every root by A. Suppose that u=f(x. _ .. .. The only terms of weight 4 are such as cont ain Pi*> Pi 2 Pz> PiPv Pz z > Pv The first two of these are of too high an order... expressed as a function of the coefficients. Sometimes the ordinary symbol -p is used in the same sense. . 300 TAYLOR'S THEOREM .. Thus we have 27a 2 j8 2 = ap\p$ 4. 6.

. a n [L \^ where a .as an operator.a + hdi + -r^r + ' + TT. h 2 h". Again. and if X = x + h we h&ve Hence. ox oy \ ox oyj ^^=^^^--5-^ dx dyJ \ dx 2 dxdyj \ dxdy if h y k are constants .1 . (1) // f(x) is a poly nomial of degree n in x. a result which we may express as follows : u.+~fM(x) & n Proof.. by successive differentiation with regard to h. XVII. A strict proof is difficult.It is proved in books on the ' Calculus ' that the commutative property represented by (A) holds when the functions in question are continuous. a v . or thus : we may assume that (A) //IN 7 2 f(x + h) . then f(x + h)=f(x) + hf'(x) + ~ f"(x) + .. a n are independent of A. dxdy The value of D n u can be written down in the same way by the Binomial theorem.= (h-r-+k-^-}u = Du. we often regard h^ + k-^. This follows from Ch. 7.. and we write 7 du da ( d 7 9 \ ^ h-~-+k-=. Taylor's Theorem for Polynomials. and here we are only concerned with polynomials.

y + k) u + Di For we may assume that where a . + -^ . . 6 . ZAen 1 1 (E) n where D denotes the operator h -. Euler's Theorem. and differentiating m-r times with regard to h and r times with regard to k. we find that ^T) /(# + A. .. ox oy oz The proof is similar to the above. A.. using the Multinomial theorem. if X = x + A..r ( d \ r ^3T/ \SV/ /(^ ^) = /tn-r. y. k.) is a polynomial of degree n in x. then . k (D) Then. y.+ l~.. ^ v ^ (3) // u?Ef(x.fm~r and putting A = 0. k we have a = u'..f (x + h) = aj + ha 2 + :-.. .+ . for we find that EULER'S THEOREM 301 1 _+*-Dn u ( B )* (2) If u z=f(x.. . Putting A = 0.. 2. we have ^ r =/ m _ r . y + k)=p r + terms involving A. (1) Let u be a polynomial homogeneous in x.a n . y and of degree n. are independent of A. The result follows by putting h = in these equations . y) = ^ which proves the theorem. y. . then f(x -f A. = 0. Y = y -f yfc. z. the left-hand side of (D) is equal to f d \ m . r (x. y) is a polynomial of degree n in x.+ k^.r(^ ^)> sa 7 ^ e .a 3 + . . 8.

the proof is similar to the preceding. in general...a is a non-repeated factor of <f>(x).. Partial Fractions.2 -5.+-^nu> ox y dy oz D r u = n(n-l)(n-2) . 6. (n-r + l)w. y. 9. ^.y ^~ 4....... ) in its expanded form. in equation (E) of Art. /(x + As. (x^ + y-^-) w = n(n-l)(n~2). z... k = Xy. J \ dx J dy dz / r J Putting h = Xx. y. whence the results follow.. 302 PARTIAL FRACTIONS (2) // u=f(x. .9 ^^dx J dy dx 2 y dxdy and.. z.4....(n-r=f \ ox oy/ Proof.. f(x + \K.. and of degree n.y+Ay) = (l + A)t .r ( } and since u is homogeneous and of degree n in x.... .) is a polynomial homogeneous in x. * See end of Art. then by Taylor's theorem...^~ =..-~.... // x . /-Az...y + Xy) du /AX > + --.... Let f(x)l(f>(x) be a rational proper fraction which is to be expressed in partial fractions. . then dn du du x a" +y~*-+ z *. the fraction corresponding to x-a is .. y. .. . y)..4. wAere D r tfawdx for ( x ^.. (f)'(a) 'x-a . (B) The coefficients of the powers of A in the expressions on the right of (A) and (B) must be equal.. 7. Let u=f(x.

we have f(a)=Aif>(a) and ^(a)=0(a).. x m . 2 .J .1. n are positive i ntegers and m^n. If x-a is any factor of x n . n u n n and denote the sum of the fractions corresponding to x . + 1--K-V n x-l r-*i r ' ' n x-l n a.a" 1 by u r . or even. 2?r /.For let <f>(x) = (x-a)^(x) and ^ = 4rv /r ^>(x) x-a From the second equation f(x) = At(x) + (x and (rom* the first.. Let a = cos + * sin . therefore A=f(a)/<l>'(a).a and x ..a. then a" 1 = cos i sin . 4-1 r--i T . Putting x = a and observing that 0(a)^0. by differentiation. w r = . i(w-l) or ^(n-2).. according as n is odd.(zl) 1 !.2a: cos ra + 1 SYMMETRIC FUNCTIONS OF ROOTS 303 Hence.l l(x n .1. then a n = 1 and x n ~ I na n ~ l x-a n x-a' The imaginary roots of x n 1 = are cos i sin^-^ where r = l..1) is equal to 1 1 +r =J(n. . ... Express x m . 2.1 /(x n ~l) in partial fractions.where a = * r 2 - r tr . 1. n n * according as n is odd or even.r tr . then by the preceding. where m. 2 # cos rmoL cos r(m l)a . Ex.

. then v satisfies the equation dv dv dv ' For the substitution x = X + h transforms w = into where A l = a l + a^i 9 J 2 = a 2 + 2a 1 A-f a A 2 . r are independent of a . 7/a.. . . ..a n ) is a symmetric function 0/a. . . 1.. a lf a 2 . Theorem..Q^~ + 2a l ^~ + 3a 2 ^ +.. we have l5 + 3a 2 ^ + 4a 3 ^ = ..... l 8 4 Q 2 = . ..=0. and since the roots o f [7 = are a -A. are f/te roote of w==(a . etc. . and by the extension of Taylor's theorem the coefficient of h is the same as that in dv dv dv a Q h ^. involving only differences of the roots.. j8. 1) 4 ~0...+ (2a x A -f a Q h 2 ) ^~ + (3aji + 3a x A 2 + ajh?) ^ + . ]8.. a 3 . (B) If the right-hand side of (B) is expanded in powers of A. . We may therefore assume that where p 9 q. And since v is a function of th e differences of a. find in terms of t he coefficients the value of This function is of order 3 and weight 3. The following theorem is often useful.. l) n =0 awd t..A... .. dv dv dv . a$x. ^or ^ biquadratic (a .. a n $#. every coefficient must vanish.. a p . ^a x aa 2 ca 3 J&x.. j3.... j8. a...10.... . ^[ n ) .. a n ) = (f>(a^ A^ ..a 1 . .'.. a lt .s^(a . we have v = (f)(a^ a l9 ... .. ...

Show that the equation x 3 . are real. 14.F.] 6. Find the values of a for which ax 3 -9x 2 -f I2x -5 = has equal roots. The equation f(x)==x* + 3qx + r=:0 has two equal roots if Prove this (i) by applying the II.This is true for all values of a . 8.3 arid . 3. . !_? Ii [Use Ex. (ii) by elimi nating x between /(#) and f'(x)=0. we find that ^=32. 3 of VI. The equation x n -qx n ~ m + r=Q has two equal roots if 5. . process to f(x) and f'(x) .. a 1$ a 2 .. which agrees with Ex.4. The value of 3? may be found by giving special values to a. All the roots of . Thus by taki ng <x=0. 8=4. y=-l.Ix + 7 = has two roots between 1 and 2 and one between . r-2p. If the equation x n +p 1 x n ~ l +p 2 x n ~ 2 + . 16. +p n ~0 has three roots. giving q-~3p. y. The condition that the equation ax* + 3bx* + 3cx + d = may have two equal roots is Show that this may be obtained by eliminating x from ax 2 + 2bx + c=Q and bx 2 + 2cx + d~Q. 2. and solve the equation in one case. = 1. 7. therefore = 0. 4. VII.C. 304 APPLICATIONS OF ROLLE'S THEOREM EXERCISE XXX 1.

c and A lf A 2 .] 14. 1. z are the roots of an equation of the form 0*-a8* + b 2 + #=0 and if B=<f> + a/3. 13. A 3 are the values of A given by the equation jc*^ y 2 ^ z 2 ^A + & 2 ^A + ^-A~ ' prove that these are all real . if . The equation x 5 + 5o 3 + b has one and only one real root in the following cases : (i) if a>0 .. 11.6) 3 + (x .. 12 by putting I/a? for a?.24x 4. b have the same sign . The equation f(x)=(x .each equal to a. this reduces to one of the form < 3 +<(6 2 . one negative and two imaginary roots.a) 3 + (x . [The roots of f'(x)=Q are -2.1.c) 3 -0 has one real and two imaginary roots. also. two real roots if . and no real root if r> 19. show that a is a root of n t x n-i + ( n _ l)*p lX n~2 + ( n _ 2) 2 p 2 X n ~* -f . If Aj. .. (ii) if a. according as 12. Show that if 36 2 <a 2 there are real values of #.] 10.-5-0 has one positive. [x 9 y.Ja 2 ) + fc'=0. A 2 . The equation x 6 + 5ax 2 + b has one and only one real root in the following cases : (i) if a. 9. The equation x 6 -5a# + 4&=:0 has three real roots or only one. 8. The equation 3x 4 + Sx 3 . y. b have opposite signs and b(b* + 108a 5 )>0.6# 2 .] FUNCTIONS OF ROOTS OF EQUATIONS 305 15. (ii) if a<0 and 6 2 -f 108a 5 >0. Show that the equation x 1 -4a^-hlOx 2 + 7a. 6.r = has four real roots if 7 13<r< -8. z which satisfy and that these values are all positive if a>0 and a 2 <46 2 .8<r< 19. then a 2 > A x > 6 2 > A 2 > c 2 > A 3 . [Deduce from Ex. +P n -i =0. A 3 are in descen ding order.

where ^^a^x-^a^. etc. and a is the harmonic mean of the roots of /'(a?) ~0. py-~x y etc. . show that ax 2 + by 2 + C2 2 + 2/yz + 2gzx + 2hxy . l) n ^0.c -fg/h. U. 0. 4. . h is zero. show that H7r s 2 tt-i^0.. If f(x)~0 is a cubic equation whose roots are a.. . is (a . n ^/. where #cos(2r-f l)wa-cos (2 // _ _ __ j __ _ __ ^^ ___ __ \_ " r ~ rc* -to cos (2r + 22. If s n is the sum of the nth powers of the roots of x. l) w ---0. u l9 u 2 ..hf/g .... /?. If a. prove that a 2 /?y. y. 19.2 ! 'vn j ______ I _________ i_ _i _____ + |j. If one of these (g) is zero. 2 ~~a x 2 + 2a 1 xy + a 2 y*.. Let k be given by the equation J'^ a-fc A ^ 6 f7 / If this equation has a rejx?ated root ^. * 2n -=-4cos 3 -. 21. u n $t. a l9 a 29 . n. The sum of the ninth powers of the roots of # 3 -|~3#4 9 is zero. and then &! = a and (a -b)(a. + |2 + ". so also is another (h). Under these circumstances. 17..+ h*~~ U ' show that ZV^O for r = 2. are the roots of the equation a-n-i x n. y. If m and n are positive integers and m^n.16. are the roots of (a . prove that according as n is odd or even. prove that k^a-gh/f^b.c) / 2 . 18. 20. . 3.. show that the equation whose roots are ay-x. provided that none of /.. If a.k t (x 2 + y* + z 2 ) .4 -x 2 4 1 -~0. /?.. g.

then 1 11 !+<! + ~<lf-. G'. m + l x m therefore (l + ~ } <( 1 -f . XVII. C' and consequently F'. 9. If a>0. where m is a positive integer.} <(l + xj \ m If z->oo . XV. (Ch. (3).] CHAPTER XIX EXPONENTIAL AND LOGARITHMIC FUNCTIONS AND SERIFS 1. 1\ m / 1 " in -f-1 / 1 \ / t * \ / i A 1+ 7 =fl+ 7 -7-I1+- . and the roots of any one of them separate those of J' 0. Continuity of a* and log a x. then m->oo and by Ch. . Hence the inverse function log a x is continuous for positive values of x. B' 0. 5. etc. XIII. the equations A' =0. m/ v ' . 21. This follows from Ch.(l+j^ e. IX.is a perfect square. (4). the function a x is continuous for all real values of x. 21. XV. Exponential Inequalities and Limits. C" = are qiiadratics in ^.. (l+-)%c and (l--}~*-*e. 2. H' are all zero.* (1) As x->oc through real values. Hence if k~kn then A\ B'. _ m + lj \ m + l (\m 1+1) . 5. [Denoting the cofactors of the elements of J' by A' 9 F'. (Continued from Ch.) The reader is reminded that log x stands for log e x. See Ch. \ xJ \ xj For let m<x<m + 1.

x .log a e = r . EXPONENTIAL THEOREM 307 (2) If z is any real number. according as z. XVII. i and consequently lim ( 1 -f x) x 6. 5.therefore Next let x = y + l.. y~>o i/->o Aoga a* 1 therefore lim . a^oo \ X/ This is obviously true if z = 0. Now by Ch.fi+h V y/ \ yj It follows that if x->x or . then as x->x . a *_i hm -----. In either case..log a {lim (1 + y) y } . *Aew . x-*Q x This is obviously true jf a = L If a^l. then as or . lim(l+^)* = . / z \x lim (14--) = e z .50. = 1/2. by the preceding. 1 1 1 lim log a (1 + y)* . . If z^O.= log a. ?/->co and =(i+ii v . (2).= loga. let a. y/ \ y (3)// a>0.oo . let a x = l+y. then as x->0.

so that each of the functions a x 1 1 .~ -. XIV. for the values 2. 3.. . The Exponential Theorem. . Hence .y f-) +. It has been shown in Ch. XIV.(n-l)/n are positive numbers less than unity. If n is a positive integer... A . and since 1/n..a~ x -. ton \ n/ n I 2 \n/ 3 \n/ + 1 terms where 308 IRRATIONALITY OF E(x) Now jo r <l.. 9. 1. then --.<log < .and x x tends to log a as a limit as #->0. For all real values of x. a*-l if x>0. x 2 x 3 x n I ' J Proof.. 2/n. . x n(n~l)/x\ 2 w(n-l)(n-2)/a\ 8 (1+-) =l+n. and its sum is a function of x which we shall denote by E(x). by Ch.1) of r we have. n .. /.- 3. The series (A) is Convergent for all values of x. (n .. that the first of these functions increases and the second decreases as x increases from zero.ic->0 X Here x may tend to zero through positive or negative values.x.~ + --Vo ^(-) + . l-o-* .

. Ex. <voj 2.1+1+ [2 + |3 + . and let | x \ =^x 1 . // a>0 and x is any real number. >-D g ^.n . Let s n be the sum to n terms of (A). we should have R | y=an integer. which is impossible on account of (B). 4.. Suppose that e=pjq t where p and q are integers .\ ^1(2 + ^" Xl + TT a!l+ . when x = 1 we have ii 11 1 6 ..3 .. then I I i.p r <r(r + l)/2n. .. Hence e cannot be a rational number. 3 -4 . 1 = lim( 1-f 71 >QO n '00^ but W-^OO This result is called the Exponential Theorem.+ n Xl j Now ^(x!) is finite for all values of x ly therefore / x\ n Iim^ n4 . $A#tt> JAaJ e i$ irrational. 0<l-. then where ? + 2 If the equality (A) holds. 1.+ 5 + from which it can be shown that e = 2-7182818284 . In particular.

Derivatives of a. When a>0 and x is irrational.... This agrees with what has been said in Ch... A Tx a h -l lim a x r = a x log a......... 9.. e x . then by Art..... 2. dx . It should be observed that.. a x has not been defined for all real values of x. if a<0.... 6...a* = e* lo 8 a = jE(zloga)... (3).. r -j-a x = lim ax In particular.. (A) that is to say 2 1 ) 3 + ..... . NOTE. (B) GRAPH OF EXPONENTIAL FUNCTION 309 6.... An Irrational Index.SXIII. (1) Suppose that a>0..... for it will be shown later that the sum of the series denoted by E(x) is a continuous function.. logx and xf. we take the equation a x ~E(x log a) as defining the meaning of a*....

On this take OA. Without' removing the pencil-point . ax prove that this is true for all real values of n. where TN l. as the unit. it follows that. lake Oa OA.n e n log a. we proceed thus : To d d n . This fact gives a rapid construction for the graph of y=a x . to the left of O. dx x dx G x (3) It has been shown that when n is rational. For instance.. the subtangent is co nstant. and mark one of the rulings as the axis of y. n _ y. for y = e x . then x = e v and -j~ = e v = x 9 ay ^ x fy 1 xi. that is./' da O FIG. for the curve y=a x 9 the gradient dyjdx is y log a. x n = nx n ~' 1 . equal to. take a sheet of ruled exercise paper and draw a line perpendicular to the ruling as the axis of x . NOTE.log x = . 4 intervals.(2) If y = logx. and on OX. with considerable accuracy. Since it has -been shown in (1) that. then TN = l/log a : i. Draw A a until it meets the first ruling to the left of OY at B. -=..e. say. _ __ e n log a? _ _ x n dx dx x x --nx n A. 60. and PN is the ordinate of P. if the tangent at a point P on the graph meet s OX in T. * d i ! therefore -.= -.

and since /(0) = 0. in Fig. the curve as seen through the back of the paper is the graph of y = log x. the foot of the ruling next to the left of or. the curve is then the graph Also. then /(0) = and /'(z) = e*-l. Hence. OX and OY are interchanged. For let f(x) = ^-1 -x. then -|x 2 <x-log (1 +x)<s x 2 /(l + &). it follows that f(x) $ according as x $ 0. Again. then 1 T 2 f'(r\ i i v**'/ -* J v ' _ ^ I+x Hence f(x) decreases as x increases from -1. This follows from (1) by taking logarithms. and draw bB until it meets the ru ling next to the left of A . and so on.log (1 + x) . with the axes in the usual position.fr6m B 9 rotate the ruler until it passes through 6. then ix 2 /(l + x)<x-log (1 + x)<-|x 2 . Find . and then turned over. if 0(#) = a.1 and ^0. then x>log (1 + x). (2) // x> . if x increases or decreases from zero. For let f(x) = x . if the paper is turned through 90 in its own plane. 310 EXPONENTIAL AND LOGARITHMIC LIMITS 7. (1) For all real values of x other than zero e*>l +x. in either case f(x) increases from zero.points on the curve to the right of Y i n the same way. so that /' (x) : according as x 5: 0. and t/ -l<x<0. (3) // x>0. 60. and replace the broken line by a ' fair curve.' If. Inequalities and Limits (continued).|-x 2 . and is therefore positive.~log(l + x)-%- .

. then as x~>0. (5) For any positive value of n* (i) lim -~. x n . and since <(0) = 0. This is obvious when r<0. (3). ______ x n x n m mx n ~ m * Choose m so that 0<w<w. (ii) Put x = l/y.=0. h are respectively the smaller and the greater of the numbers 1 and 1 -f x. lim x r /e x = 0. (4) // x> -'1 and ^0 and I. putting .w ->ao and (log x)/x n ~>0. then x 2 x 2 This is merely another way of stating the inequalities in (3). then. This proves all the inequalities in question.->oo X x >o (i) For all positive values of m we have by Art.1 x 2 . J/->QO and x w log x = ~ (log y)/y n ->0. EULER'S CONSTANT 311 (6) For all values of r.=n. 2. (ii) lim (x n logx) = 0. ~r X then Hence <f>(x) increases as x increases from -1. then a^ x~>cx) . If r>0. it follows that <f>(x) "2 according as x ^ 0. a._ __ _ __ ^ . logx 1 x m -l 1 _ 2__ <. let e x = y.

n + 1 1 . 1 n I ( I n n-l n 6 n _l n & \ n that w n decreases as n increases. since w = ->0. in succession for n.H ---. then as n->oo .. (5). (6).<log V w/ n & \ n/ w i Substituting n .. Let n >e any positive number. log 1+. log <. 9. have Also log f < 1 . and by addition. It can be shown tnat / 0-577215 . Theorem. (2). and.logn. hence .e.2... (2)... /.. 7. Again. from above. w n ->y Tt' y is a fixed number lying between and 1. by Art. lence it follows that x r /e x -+0. Proof.. NOTE.. (5). n . i. u n Thus u n is a decreasing function which is always >0. no matter how small n may be. by Art. 1 \ . It is often convenient to state the theorem as follows : .-f-<l + logn . 7. By Art. 2. therefore e* tends to infinity faster than . First suppose that #-^oo .. The Manner in which e and log x tend to Infinity. that is.+ J + . hence 0<w n <l .log x tends to oo more slowly than l/x n . therefore Zo<7# tends to infinity nore slowly than x n .. n i /i 1\ 1 . y->oo. by Art. . x n /e x -+Q. Consequently V>y where y is a fixed number >0. however great n may be. log (w But log(w + l)-logw>0. (log x)/x n ~>0. Again. 7. 7. no matter how small n may be. and. y<l. we. then -logx->ao. // M n = l+-g. . by (5). and by Art.e* y \ ' Now as x-^oo.. 8. . The number y is known as Euler's Constant. If #->0. it follows that log(n-f l)<l+!-h! + . n . (log x)/x~ n ->0.) < <-log (1-. (logy)/y w ->0. since w w <l. and. then e x ->oo and logx->oo. Now.1. 7.

Also s 2n+1 ~s 2n --0. and is denoted by E(z). since P>0.log 2 + e 2n .+ . The Exponential Function. and exactly the same proof holds when z is complex.(y + log w + ) .. where ->0 as n-^oo .. (1) For all values of z. // z = x + ly where x and y are real.. real z 2 z 3 or complex.. Thus. .. n and />cos< = l+-.= 1 +. Therefore 2n --log2. where p>0 and ~7r<^<7r. the series l+z + r. we have cos ^>0. It has been shown in Art. Let s r be the sum to r terms of the series..+-.1 -+. is convergent. 1+->0.11 1 N /I 1 1 then = (y + log 2n + 2B ) . and ^/tan ^ -> 1 . Proof. therefore <. . log2 = l-| + |-. ( z n jE r (z)=lim (1+-) =e x (cosy + 1 siny).. 3 that the first of these equalities holds fv T hen z is real. + (-l)". hence lim 7^^ = lim (n tan <^) = lim {ny/(n + x)} == y.0. (2) Theorem. and -\-n n Also tan ^ = y/(n + )-> 0. to oo. or by exp z. hence.e n .-M . Series for log 2. Its sum is I ! called the Exponential Function. 71 11. 312 EXPONENTIAL FUNCTION 10.=p(cos <f> + i sin <f>). l_-i i_ (-n-i! to oo . n n n cr For sufficiently large values of n... hence s 2n+1 ->log 2. Z 3T II Let 1 +.

we define the principal value of a z as E(z log a) and. 12. When x = 0. if a is real and positive and z is real. (3) It follows that E(z) cannot vanish for any value of z. if a>0. Therefore E(z) is unaltered by adding 2ik7r to z. For all real values of x and y. Thus if z = x 4 iy. it has been shown that E(x + ty) = e x (cos y + i sin y). this gives cos y + i sin yE(iy) Equating real and imaginary parts. therefore p w = (l-f-j sec"^. since lim see" <f> . hence. we take a z as representing this value. for neither e* nor cos y + 1 sin y can vanish.lim ( 1 + tan 2 <f>)* n = lim f 1 + ^ ) = 1 (^ 2 ) 2n = 1 > we have lim p n = e* . When z is real. SINE AND COSINE SERIES 313 (4) The function E(z) is periodic. positive or negative. Series for Sine and Cosine. 13. A Complex Index. For if k is any integer. we have. n/ n z \ n/ T r \ n/ / w2\in JL and. e x ' (cos y + i sin y) (cos y' + i sin y) = e x + x> {cos (y + y') + i sin (y 4. except when z is an integer. For the present.y')} = e* +2 '. and e z ' = e x ' (cos y' + tsint/') . This implies that for imaginary (as for real) values of z. and %' = x' 4. we take e z as denoting its principal value defined by E(z). In agreement with this.iy f . at present. and it has been shown that E(z)= the principal value of e z . thus. e z is to be regarded as a multiple-valued function. . e z . for all real values of . Again.( x \2 y2 / x \2 / x \n 1+-) +^i=(l+-) sec 2 <.The real positive value of e z is called its principal value. JE(z) = lim p n (cos n<f> + i sin n^) and jE (z) = e?(cos y -ft sin y). the values of cos y and siny are unaltered by adding 2&7r to y. thus E(z + < 2 t ikn)^E (z). We take this equation as defining the principal value ofe z when z is complex. e z ' = e x . and its period is 2i7r. e z is a multiple-valued function. we write e z = E(z) = e x (cosy-f i sin y). Hence the same index laws hold for imaginary as for real indices. a z = e zlo ^ a = E(z log a).

v 2 t/ 4 . Ex. Ex. Find the value of (n* -f 5)/ i n....l)+n-f5.. we have (2) TAe 5en*65 2/M n x n /(w + a)(nH-6).-* - 14.. a v . hence. A are unequal positive integers. sin ?/= (^ rJL~^li ""*" JWK *" 314 SUMMATION BY EXPONENTIAL THEOREM 15. a r . 2... . 1. 6. therefore cos j/ = (c ty 4.. (n-r-f 1). 2 = 3.iy)~e~~ iy . .. Find the value of Denote the sum by $. so that u n = a -f a 1 n + a 2 n(n -l)-f . a 8 = l. and then s^n /v>tt spf* 1 T** 2 r n=r n-r = (a + a^x + a 2 x 2 4..t sin y~E( .) |n w. +a r n(n~ 1) .Aere t/ n is a polynomial in n and a. Here w s -f 6=n(n-l)(n-2)-f 3n(n... -f a r x r ) e?. Series which can be summed by the Exponential Theorem. putting a 5. Series of this kind can be summed as in the following example.. where u n is a polynomial in n.e"~ l *0.. y 3 ?/ 5 .. Using e* to mean exactly the same as E(z)> by the preceding. and * = 1 in the result above. independent of n. then . 'We can find a . Exponential Values of Sine and Cosine. cos y + i sin y = E(iy) = &u 9 cos ?/.(rn-A. x n (1) The series Zu n .

Now where a.234 3579 3579 1+2 1 + 2 + 3 n + ji" f [6 4 "L?" f "" ~ii" + ni~ 1*1 1 1 2 3 3 8 4 8 + '. d=l. their values being a= -5. 2-9 : 0. c= -2. 6 = 5. Show that lim fl . Thus XT Now =->-'-*)> _ -"' 6 LOGARITHMIC THEOREM 315 EXERCISE XXXI 1. c. 2** 3a: 8 1 + " + "" ""^" Prove that : . a. rf can be found by division. Find the sum to infinity of the series in Exx.} X = 1.' 7 * A ^ 4 " " 1 " 4 "" 2 3 3 3 4 s A . 6.

Let s n be the sum to n terms of the series.. cfx 1 x 1x 1 x therefore y n >0 for every n.. then y n = when x = 0. X B. If 6 ox cosfea. 1-hx* Therefore y n ^ 0.) (ii) The case in which x = 1 has been dealt with in Art.=w -f ^^ + .+ + .!<#<!. a 16. 1+a..4. . 316 LOGARITHMIC SERIES (iii) When x<0. Proo/.C.wa: n + . Hence log(l-fx) lies between s n and s^^ for every n. But n+l-n = (~l) n ^ n " l " 1 /( w . 12... Ex. therefore s n ~>log (1 +x) as n->oo ..s n .. Theorem. then i n where = tan" 1 . which proves this case of the theorem. 10. according as n is even or odd. (i) Suppose that 0<x<l.11... it has to be shown that X 2 X 3 X n if 0<x<l. and ^ __ 9 dx 1 -f a.. XXIX. ( -=t(4S-16e).. and ~r n = ^ __ = -.f i)"^ as W-^QO. then y n = log ( 1 + x) . 8.A. (Of. cos rz^ 13. changing the sign of x. // . . then y n = when x = 0. then -log (1 -x) = x + ~ + -+. Let y n = -log (1 -x)-s n9 where s n is the sum to n terms of this series.

and.. n H....(A) Changing the sign of x..or in taking o ...7....= -./ x 3 x 5 \ /nv 3 + 5 + -.= log (1 + x) ..(D) o: 3 a. 1+x n-f 1 .+... . since 0<x<l... we nave & 1-x 7^ 2n-fl . 1 . thus y n ->0 and s n ~>-log (1 -x)....i** / -. 17. which proves the theorem in this case.. Hence 0<y w < ------... we have X X /-tv~...log (1 . and ofe n _ x n | x n 1 x n+1 )_ 1 x n+1 ~dx = i-x ~ (i -x * r^Ti ' (T 7 *) 2 / ~ ~ ^i ' ("r-lo^ .. Logarithmic Series...+ (-I) 71 . (C) ..1 4. Writing . so that x = ^ . it follows that x x 1+x .x).n+i x n+ Let Zn^yn-"* ' i ~5 then 2 n = when x = 0. If -l<x<l. 1 +# and since log . 1 x" +1 therefore z n <0.. . show that the err..4.1 1 x the last expression ->0 as n->oc . 5 L // 0<x<l and s n ~x+ .......... to n terms..

-.log. find the value of *-* ~~~ *" The series is convergent when \x\ < 1. 2 can 6e applied to sum any series of which the general term is where P is a polynomial in n and a. then ^n=oo' x n 1 f x 2 x k Ex.. we have . then since 1 1 3 1 wehave TAe method of Ex. Ex. rt Let =2. Calculate Iog e 2 to seven places of decimals. 19. CALCULATION OF LOGARITHMS 317 18. 17 to the calculation of Napierian logarithms is exhibited in the following example. 2s n as the. If k is a positive integer and \ x \ < 1.2. If \x\<l. 1.1-fa. log^ means log ]0 jV. The number e is chosen as the base of the system of logarithms used in theoretical work.. Series which can be summed by the Logarithmic Series.. just as. 6. in practical reckoning.=-. 2 -2 Carrying the reckoning to nine places.. k are unequal positive integers.. Ex. T x i+* i . Calculation of Napierian Logarithms. 1 ~~ x 2ii\> + 1 1 x If R n is the remainder after n terms of the series. Let 8 be the sum. log^V means log e N..-. value of log is less than " . 1. In theoretical work. the inventor of logarithms. .-. Such logarithms are called Napierian } after Napier. The method of applying the equations of Art. . .

1/3 =0-333 333 333 1/3=0-333 333 333 1/3 8 =0-037 037 037 l/(3 . 3 8 ) =0-012 345 679 1/3 5 =0-004 115 226 l/(5 . 3 5 ) =0-000 823 045 1/3 7 =0-000 457 247 I/ (7 . 3') =0-000 065 321 1/3 9 =0-000 050 805 I/ (9 . 3') =0-000 005 645 1/3 11 =0-000 005 645 1/(11 . 3 11 ) =0-000 000 513 1/3 18 =0-000 000 627 I/ (13 . 3 13 ) =0-000 000 048 1/3 15 =0-000 000 070 1/(15.3 15 ) =0-000 OOP 005 =0-346 573 589

0-693 147 178 /. log e 2 =0-693147178 nearly.

318 BORDA'S METHOD The error in the value obtained for log e 2 is due to two causes : (i) We have taken 2S 8 for log e 2. By Ex. 1 of Art. 17, the error on this account is less than

. JL >000 000 002. 17 1 9 (ii) The error in the calculated value of 2S Q is numerically less than 2x8x0-000 000 000 5 = 0-000 000 008. Adding the results of (i) and (ii), we see that 0-693 147 178 is an approximate value of log,, 2 with an error numerically less than 0-000 000 01. V. 0-693 147 168<log e 2<0-693 147 188. Hence, to seven figures, log e 2 = 0-6931472. Having found log e 2, the Napierian logarithms of the natural numbers can be found in succession by using formula of (D) Art. 17. The logarithms of successive prime numbers may be calculated by the following method, due to Borda, in which the series employed converge very rapidly. Using the identities x 3 -3x + 2 = (x-l) 2 (x-f2) and z 3 -3z-2 it will be seen that

(x + l) 2 (z - 2) " x 3 - 3x - 2 ~ 1 - 2/(z 3 - 3z) '

-f ... .

Putting z = 5, 6, 7, 8, it will be found that log 2-f log 3 + log 7 =0-0181838220 ... , -| log 2 + log 5 - log 7 = 0-0101013536 -2 log 2 + 2 log 3 -i log 5 =0-0062112599 ... , ~ I log 3 + i log 5 + log 7 = 0-0040983836 .... Hence it can be shown that log 2 = 0-693147 180 ... log 3 = 1-098612288 log 5 = 1-609437912 ... log 7 = 1-945910149 .... The logarithms of successive primes 11, 13, etc., can now be obtained by putting z + 2 = ll, 13, etc., in the formula. The method of the example which follows, in which logarithms to a base 10 are calculated without the use of the logarithmic series, is of historical interest, being one of the methods given by Briggs.

HYPERBOLIC FUNCTIONS 319 Ex. 2. Obtain the logarithm of 7 to the base 10, by the use of square-roots only , to 5 decimal places. In the reckoning below, when a square-root has been obtained which is greater th an 7, the square-root of the product of it and the last square-root obtained that w as less than 7 is found next ; and vice versa. Thus we have log log Jog log log log and VT6-log 3-162277 ... = log ,4 =0-5, */T6I=log 5-623413 ... = log J5=:0-75, ^105 = log 7498942 ... = log =0-875, >/]&;= log 6-493816 ... -log D = 0-8 125, s/'OD^log 6-978305 ... = log ^0-84375, */(7JE=log 7-233941 ... =log ^ = 0-859375, so on ; until we arrive at

log 7-000032 = 0-845100 log 6-399968 = 0-845096 ... j thus, Iog 10 7=0-84510 to five places, and probably is equal to 0-845098 to six places. 20. Calculation of Common Logarithms. We have logjo.V-log.tf/log.10.

The expression l/(log e 10) is called the modulus of the common system of logarithms and is denoted by /*. It can be shown that, to fifteen places of decimals, /* 0434294481903251. Thus Iog 10 tf =/*log tf tf, where ju, has the above value.

21 . The Hyperbolic Functions. (1) The functions known as the hyperbolic cosine, sine and tangent of x, and denoted by coshx, sinhx and tanh x, are defined by --* A , *~ x tannx

, , . The reciprocals of cosh x, sinh x, tanh x are called the hyperbolic secant, cosecant and cotangent of x, and we write sech x = = . cosech x = -r-i , coth x --

- . , WQViVAU. */ . i , VVVJLX rf/ 1 i . cosh x sinh x tanh x (2) Hence the following equations : cosh 2 x - sinh 2 x == 1 , 1 - tanh 2 x = sech 2 x, T- cosh x = sinh x, -7- sinh x cosh x, -7- tanh x = sech 2 x. ax ax ax

320

ADDITION FORMULAE

Also, cosh ( - x) = cosh x, sinh ( - x) = - sinh x, tanh ( - x) = - tanh x, so that cosh x is an even function, and sinh x, tanh x are odd functions of x. Graphs of these functions are shown below : rough sketches of the first two can be constructed quickly from the curves y = e x , y^e~ x (see p. 309), by observing that, in Fig. 61, C is the middle point of AB, and ED = BC. The ordinate of y = tanh x is equal to the ratio ED/EC ; this can be found numerically or by a geometrical construction.

61.

(3) Addition Formulae. We have 2 cosh x cosh y = \(e x + e~ x ) (e y + e~ y )

-f- e

Similarly,

Therefore

= cosh (x -h y) -f cosh (x-y). 2 sinh x sinh y = cosh (x + y) - cosh (x-y), 2 sinh x cosh y = sinh (x + y) + sinh (x - ^ 2 cosh x sinh y = sinh (x + y) - sinh (x -

cosh (xy) = cosh x cosh y sinh x sinh y, sinh (x y) = sinh x cosh y cosh x sinh y, and consequently A , , . tanhxtanht/ tanh (x dby) = ^ T r T r^- , v y/ ltanhxtanhy where in each equation the upper or lower sign is to be taken throughout.

INVERSE HYPERBOLIC FUNCTIONS 321 (4) Using the exponential values of sin x and cos x (Art. 14), we have

cos x =

2 sin x = = ~ sinh ix, tan x= tanh ix, i and cosh x = cos tx, sinh x = i sin *x, tanh x i tan ix.

22. Inverse Hyperbolic Functions. Let i/ = coshz, then e x _j_ e -x _ 2^ and e 2x - 2e*z/ -f 1 ==-- 0, therefore ^ = ?/v/(2/ 2 -l) and x~log{y,J(y 2 -l)}. Thus if #>1, there are two real values of x, equal in magnitude and of opposite signs, which correspond to any given value of y. The positive value is denoted by cosh" 1 y, so that cosh" 1 y = log {y + J (y 2 ~ 1 ) } . Next, let ?/ = sinhx, then 6 2a; -2e a? /-l=0, and since e x >0 for real values of #, we have eF~y + ,J(y 2 + l), so that # = log {y + J(y 2 + 1)}, and we write sinh- 1 y = log {y + Jy 2 + 1}. A * . e^-e-* Again, it / = tanhx = then e 2 * = - ~ and Hence we write ,, 1

Thus cosh" 1 / is a single- valued and continuous function of y (see Ch. XVII, 21) for all values of j/>l ; sinh"" 1 y is a single-valued and continuous function of y for all values of y ; tanh"" 1 y is a single-valued and continuous function of y over the range -l<y<l. EXERCISE XXXII The notation of Art. 9 is used in Exx. 2 and 3. 1. Show that -~H --- -+ -- - -f . . . -f ur- -> log 2 as n->oo. 2n

2. Prove that 1 -f t + + ... -J-- ~"=log 2 + Jy + J log n + 2n ~ n , so that ^71 1

322 LOGARITHMIC SERIES 3. Let the series 1-J-fJ-J-f... be deranged as follows. Write down the first p positive terms followed by the first q negative terms, then the next p positive terms followed by the next q negative terms, and so on. Prove that the series so deranged is convergent, and that its sum is log 2 + | log plq.

[Let a r be the sum of the first r groups of p positive followed by q negative terms, then 1 1 \ I/ 1 1 1

= (log 2 + Jy + \ log pr + 2pr - pr ) - i (y + log qr + c ar ) -+ log 2 + \ log plq. Also, if s n is the sum to n terms of the deranged series, r can be found so tha t all these terms are contained in oy, but not in a r i. Thus a r -s n is the sum of a finite number of terms each of which ->0. Also r->-oo as w->oo, and consequently s n -> log 2 4- i log p/q.] 4. Use the series for log ( 1 -f x) to show that : (i) if 0<x<I, then 0<#-log(l + x)<\x* ; (ii) if -l<x<0, then 0<#-log(l +x)<\x*l(l + x). [For (ii) let x= -y, then 0<t/<l and ...), etc.]

5. (i)Showthat (ii) Hence calculate Iog 10 7 to five places of decimals, given that Iog 10 2 =0-301030, p, =0-434294. 6. (i) Showthat lo glo 1-1=2,1 |l + i - i+i - 21-.+ ...} (ii) Given ^=0-434294, find Iog 10 11 to five places of decimals. 7. (i) Show that log, g=2 M {^ + 1 . ^. + L ^ + ...} . (ii) Given that Iog 10 2 =0-301030, Iog 10 3=0-477121, ^=0-434294, find log lo 13 to five places of decimals. 8. (i) Show that log e ( 1 -f x + z 2 ) = log e ( 1 - x 9 ) - log e ( 1 - x). (ii) Expand log c (l-f x + x 2 ) to six terms, and show that the coefficient of x n in the expansion is -2/n or Ijn according as n is or is not a multiple of 3. 9. Find the coefficients of x* m and x* m + l in the expansion of

Sum to infinity the series in Exx. 10-14. 1 1 1

' IT2 + 37* + 6~6

APPROXIMATE FORMULAE 323

15. In the expansion of (I +bx + cx*) log e (1 -fa;), (i) Find the coefficients of x* and a*, and determine the values of b and c so that each of these coefficients may vanish, (ii) Giving to 6 and c the values so determined, expand the given expression as far as the term containing x*.

(iii) Hence find an expression of the form - r- - - which will give the best possible approximate values of log e (1 + x) when x is small. 16. (i) Prove that (6 + 6x + x*) log e ( 1 -f x) - (6x + 3x*) . ( 1.2 2.3 3.4 t } x 6 J --- x H -- x 2 - ... V , [3.4.5 4.5.6 5.6. 7 /' (ii) Denoting the series in brackets by u v - w a -fw 3 - ... , prove that u n = (in- 1) 2 *n-i (n-l)( + 4) (iii) If 0<a?<, show that u n <u n _ l . O A<| i Q'*'2 (iv) Hence show that if 0<x<~, then - - -- is an approximate value 3 b -f bx -f x* of log e (1 +x), with an error in defect less than x 5 / 180. 759 17. Use Ex. 16 (iv) to show that Q0 ^ AO is an approximate value of log e 1-024,

with an error in defect less than 5/10 11 . Hence find Iog 10 1-024 and Iog 10 2 , each to ten places of decimals. 18. If a-b=h, where h is small, show that a-6/1 l\h h 2 h

a-b\ 2/a-6\ 3 _^ h* A 8 + " + i + * + "'"

(iii)* Hence show that, if a is^nearly equal to 6, , a a-bfl l

the error being approximately equal to (a -6) 3 /6a 3 . Sum the series in Exx. 19, 20': in v"* 83 * 00 n ~*~* ~n on v WamQO ( n + *-J x n

21. If 8 n is the sum to n terms of the series 1 1 1 1 1. 2. 3*5. 6. 7 + 9. 10. 11 13.14.15 ' prove that 8^= a , a i 1 .., | 1 , Hence show that the sum to infinity of the series is J log 2. * This is Napier's formula.

324 LOGARITHMIC SERIES of the series 1 1

22. If s n is the sum to n terms of the series

2. 3. 45. 6. 78. 9. 10 (3w - l)3n(3n + 1) prove that 111 1 2^^+^ + ^3+".+3^T- L

Hence show that the sum to infinity of the series is (log 3-1). 23. Prove that

+++ "- to -=

24. By expanding log(l-f-o;-fa; 2 ) in various ways show that (i , i- + ^Ll)-(-f)(- g ) + ... =( -i r . 2 or (-ir- according

|J8 [3 as n is of the form 3r or 3rL ,.., , w 2 n a (?i 2 -l 2 ) n 2 (n 2 -l 2 )(n 2 -2 2 ) . 1XM . lm . (u) 1-^+-^ ; - i - .-ii - ; + ... =(-!)" or (-|) n - 1 ac1^ Li l_z cording as n is of the form 3r or 3rl. ,..., , n(n + l) (n-

or

2n + 1 according as n is of the form 3r or 3rl.

-=0, -, - or - according as n is of 7Z- 71 71

the form 6r, 6rl, 6r2, 6r~3. [These are obtained from the identities :

(ii)and (ffi (iv) log(l -^)-21og(l -x 2 ) + log(l -s)=log 25. Use the identities ^-^ to show that

/ 72 1 / 72 3 U 4 -

2/ + "" J " This formula is useful in finding the logarithms of large primes.

CHAPTER XX CONVERGENCE (2) SERIES OF POSITIVE TERMS (Continued)

1. Cauchy's Condensation Test. // f(n) is a (positive*) decreasing function of n, and a is any positive integer >1, then the two series, and 2a n f(a n ), are both convergent or both divergent. Group the terms of Zf(ri) as follows : {/(I)

where, if v n denotes the sum of the terms of the nth group, v n =f(<* n - 1 + 1) +f(a n ~ l + 2) + ... +/(a). The number of these terms is a n - a n ~ l , and since f(n) is a decreasing function, (a n - a w ~ 1 )/(a n ) < v n < (a n - a n ~ l )f(a n - 1 ) ; /. -(a-lJaYKXvn^^- 1 )^" 1 /^' 1 )a Now, if a n f(a n ) is convergent, so also is v n i and therefore Sf(n) converges (Ch. XVI, 5). If a n f(a n ) is divergent, so also is v n ; and therefore 2f(n) . diverges. But a series of positive terms must converge or diverge, therefore f(n) and a n f(a n ) are both convergent or both divergent. 2. An important Test Series. The series 2r } - ^ is convergent if p>l and divergent if

If a is any positive integer, this series is convergent or divergent according as the series Zv n is convergent or divergent, where

n ~ a n (log a n )* ~ (log ay n p " Since (loga) p is constant, the given series is convergent or divergent accordin g as 21/n p is convergent or divergent, that is according as p>l or

* In any case, the terms of the decreasing function /(n) are ultimately of the s ame sign (Exercise A-A.V, 1}.

326 RAABE'S TEST 3. Kummer's Test. Let Su n and 21/d n be two series of positive terms and suppose that 2l/d n is divergent; also suppose that Vn = dnUn/Un+i-d n+l ; then (i) if a fixed number k can be found so that after a certain stage, say

for n^m, v n >k>0, the series Zu n is convergent ; (ii) if v n <0 for n^m, Zu n is divergent. Proof, (i) Putting w, m + 1, ... , n-1 in succession for n in the given condition, we have

whence u m+l + w w + 2 + . . . + u n < 7 (d m u m - d n u n ) < 7 d m u m .

Thus if s r is the sum to r terms of

m + d m u m , a fixed number ;

therefore Su n is convergent (Ch. XVI, 7).

(ii) Since d n - d n4 . 1 <0 for n^m, we have

"n+i d therefore u n >d m u m -5-. Now d m w m is a fixed number and l/d n is a n divergent, therefore 2u n is divergent. NOTE. If t> n ->-0 we cannot find k so that t; n >&>0 for n^m, and the test fail s. In part (i) of the test, S\jd n need not diverge ; in fact, d n may be any posit ive function of n. 4. Raabe's Test. Let Zu n be a series of positive terms, then (i) if after a certain stage, say for

where "k is a fixed number, then Zu n is convergent. (ii) If n( ~ 1)<1, then Zu n is divergent. Xu n+l ' This iollows at once from Kummer's test by taking d n = n, giving

GAUSS'S TEST 327 In particular, if Urn n ( - 1 J = Z, then Zu n converges when l> 1 and ^ u n+l ' diverges when l<l : but if Z = l, the test fails. This test may be tried when D'Alembert's test fails. .,,-,.., ,. . l * l .3 x 5 Ex. 1. Consider the convergence or divergence of x + - -- -f 5 -r =- + 4 o ^2 . 4 o

TT *, .. n Here -T5L -~-^ -' . -~ ; .-. hm 2- - 9 . u n+l (2-l) * a ' w n+1 x 2 By D'Alembert's test, 27w n converges if | x \ < 1 and diverges if | x \ >1. If | x \ = 1, the test fails ; with Raabe's test, however, we have 6n 2 -n 3 .

therefore 2u n is convergent. NOTE. This series is the expansion of sin" 1 x, and was discovered by Newton. From it, he deduced the series for sin x. 5. Gauss's Test. Let Zu n be a series of positive terms and suppose that u n /u n+l can be expressed in the form

^n-fi where p>l and \ b n \< a fixed number k or (in particular) b n tends to a finite limit as n-><x> , then Zu n converges if a>l and diverges if Proof. First suppose that a^l, then ( U n l\ , b n n ( i j=a+ ^ r >< \u n ,, / n p -*

as

n+l -

for | b n | <k and p>l. Therefore, by Raabe's test, Su n converges if a>l and diverges if a<l. If a = 1, Raabe's test fails and we apply Kummer's. Taking d n = n log n in Art. 3 we have

- (n + l) log (n + 1)

^n+l

= n ( 1 4- - + -j J log n - (n + 1 ) log (n + 1 )

Now (n-f-l)log ~ = (w-f l)logf 1 r)->-l as and i^.6 n ->0, for jp>l (Ch. XIX, 7, (5)), and | Therefore v n -> l and, after a certain stage, t> n <0. Hence 27w n is divergent. ,^

328 BINOMIAL SERIES In many cases u n /u n+l can be expanded in a series of powers of 1/n, and we have the following rule : r* u n ^ a b T f n .,_ _ I j __ j ___ i // i -r -r o ~r , J U M n n 2 the series being convergent for sufficiently large valws of n, then Zu n is convergent if a>l and divergent if a^l. For in this case b n is the sum of a convergent series. This is the most generally useful test for series of positive terms, and in cases where it applies, it is better to use it at once instead of beginning with D'Alembert's and Raabe's tests. P l% 32 \2 32.52 Ex. 1 . Discuss the convergence or divergence of - 2 + W~T* * 02 42 6 2 + ' " '

_ _. . . . A , x n ^ By division we can show that =[ - -r 1 + - n n

I/, i\ //, i v i where b n - =^1 + - )/(l 4-^) - - j aa n->oo. Therefore by Gauss's test the series is divergent. Here the tests of D'Alembert and Raabe both fail. CONTINUED FROM CH. XVI, 25 6. Binomial Series. When x=-l, the series (n-l) 2 n(n-l)(n-2) v 2 - /

L L converges if n>0 and diverges if n<0. When x== -1, after a certain stage all the terms have the same sign. Denoting the series by u + Wj -f w 2 + . , . , we can show by division that M- r + 1 ., n-fl b~ * _ i t * i ~j ) w r +i n - r r r z where & r = n(n + 1) l( I --)->n(n + l) as r~>oo. Hence, by Gauss's test, the series converges or diverges according as n + 1 5 1, that is according as n ^ 0. Apparently the test says that the series diverges when n 0, but here the reasoning fails because u r / u r+i * s f the form 0/0. 7. The Hyper-geometric Series. The series 1+ rT^ X+a r.2.y(y+V xi + "is known as the hyper-geometric series, and is of great importance.

HYPER-GEOMETRIC SERIES 329 If none of the constants a, j3, y is a negative integer, the series is endless,

and we shall prove that (i) the series is absolutely convergent if |x|<l, and is divergent if \x\>\; (ii) when $ 1, the series converges if y>a-f-/8, and not otherwise; (\\\) when x -1, the series converges if y-f l>a + /3, and not otherwise. Proof. Denoting the series by U Q + ^ + . . . -f u n + . . . where

, = - ... 1.2...n.y(y + l)...(y + n-l)

. we have

(i) Hence the series converges absolutely if |x|<l, and diverges if |x|>l. (ii) When x = l, w n /w n+1 ~>l. Hence, after a certain stage, the terms have the same sign. Applying Gauss's test, we can show by division that

n+i n

where \ n- \ w/ w and 4, 5 are fixed numbers. Thus b n -+A and Z"w n converges or diverges according as y-f 1 -a->l or <1, that is according as y>a-f/J or y^a-f/?. (iii) If Aen x= - 1, after a certain stage, the terms are alternately positive and negative. Also |w n +il<! w n| ^ (a + n)(]8 + n)<(l+n)(y + n) ; that is, if (a + j8)<y-fn(l+y),

or if l+y~a-j8>(ajS~y)~. This is true for sufficiently large values of n if y 4- l>a -f /?.

Again, as in the preceding, -- !*_-.! +-5 where a n ->y-f 1 ~a~)3 as w n+1 n n->oo. Hence by Ch. XVI, 24, u n ->0 if y-f I>a4-j8, but not otherwise. Therefore by Ch. XVI, 14, the series converges if y-f l>a+/}, but not otherwise.

330 ABEL'S INEQUALITY 8. De Morgan's and Bertrand's Test. Let Zu n be a series of positive terms, and suppose that u n /u n+l can be expressed in the form u n+l ~~ n nlogn' then (i) if after a certain stage a n >l + k where k is a fixed positive number, Zu n is convergent. (ii) // a n <l, Zu n is divergent. Using Rummer's test, we take d n = n log w. Then, as in Art. 5, u v n ~n log n . - (n + 1 ) log (n + 1 ) u n+l

Therefore v n ->a n -l as n->oo . Hence Zu n converges if a n >l + k and diverges if a n <l. If a n = 1, v n <0, for log ( 1 -- r )< -- = , therefore Zu n is divergent. \ n -f- 1/ n + 1 SERIES WITH TERMS POSITIVE OR NEGATIVE 9. Theorem. // Eu n is a convergent series of positive terms and a v a 2> 39 t5 a sequence of real numbers, positive or negative, and all numerically less than some fixed number k, then 2a n u n is absolutely convergent. For | a m ^u m ^ \ + 1 a m+2 w

Since 27w n is convergent, we can find m such that for all values of p,

where is any positive number, as small as we choose. Therefore | a m+l u m+l | -f | a m+2 u m+2 1 -f ... -f | a m+ ,u m+ , | <c. Hence 27a n w n is absolutely convergent. ^a:. 1. // Eu n is a convergent series of positive terms, 2u n cosnO and Su n ai nnB

Then u l = s v u^s z -s v W 3 = s 3 -.s n _ 1 + a n * n . 2. l<s m+ v-Sm<h> that is to say. Therefore 11.s 2 . we can find numbers A..^ + a ^ + m . + a n w w . ... . // Zu n converges or oscillates between finite limits and a l9 a 2 . . and their sum is a v Also s v 5 2 ..are absolutely convergent for all values of 0.of positive terms whic h tends to zero as a limit.. 10. . Now the factors a l -a 2> 2 ~ a 3> a n-i~~ a ni a n are positive or zero.a 3 Wj 4. a 3 . Let s n = u l 4. . and if a l> a a . >+ (a n ^ l -o n ).. therefore * = 1 5 1 + 2 ( 5 2 ~ S l) + ' ' ' + a n (*n ~ S n~l) = (a l . a 3 . . Dirichlet's Test.a 2 n 2 + .w 2 + + u nSince 27w n converges or oscillates finitely.. . nofr. l< u m+I + w m+ 2 + + Therefore. then a j < Q. // u v t^.a 2 }s 1 + (a 2 . for all values of m and p.a 3 )* 2 -t. 5 n are all greater than I and less than h. by Abel's inequality Now m+1 ->0 as w->oo .. ... etc. . therefore for any e we can find m so that hence ^ a n u n * s convergent.. t* a sequence of real numbers swh that for the values 1. I so that. then Ea n u n is convergent. .. is a decreasing se quence of positive terms. is a decreasing sequence.. . -f u n . + aj ^ n < a ^ DIRICHLETS TEST 331 For let tf n ~ u l + u^ 4. . 3. . and t n . ti 8 . Abel's Inequality.

j&ar. a n a: 1 n ->0.. both the series a n cos nQ and a n sinn& are convergent unless 0=0 or 2fo r. 25). Abel's Test. a n tends to some limit I. Z(a n -l)u n is convergent.A. therefore Hence by Dirichlet's test. If n n converges and a 1? a 2 . .C. For since a n x^ is convergent. that is to aay.lLu n is convergent. For if is not a multiple of 2?r. 832 CONVERGENCE OF POWER SERIES POWER SERIES A series of the type Za n x n is called a Power Series. y B. a 3 . Hence if | x \ < | x l |. For as n tends to infinity. X 7 i X k + k H-An T T X l I ^l Therefore Z \ a n x n \ is convergent and Ea n x n is absolutely convergent. r In 4 Therefore every term of 27 1 a n x n \ is less than the corresponding term of the convergent series 7 . therefore a n u n is convergent. is a decreasing sequence of positive terms. // Za n x n converges (absolutely or conditionally) for x = x ly it converges absolutely for all values of x such that \ x \ < | x 1 \ . Hence the result follows by Dirichlet's test. In Arts. 12. . in which case Ea n cos n6 diverges. 24.. then Za n u n is convergent. 13-15. 13. // (a n ) w a decreasing sequence of positive terms which tends to zero as a limit. tho series ScoBnO and 27 sin n^ oscillate betwe en finite limits (Exercise XXVII. But Zu n is convergent. a n u n . x may be real or complex. Therefore a positive number k exists such that | a n xf \<k for every n. 1.

The upper class is to contain every real number r'. The circle (0. R = <x> in the exceptional cases (i) and (ii) respectively. Fdr a complex series Za n z n . or (iii) a positive number R exists such that a n x n converges absolutely when | x \ < R and is non-convergent when \x\>R. 14. Choose a positive number a greater than | x' \ . The series converges absolutely or is non-convergent according as x is within or outside the interval. Also. The lower class is to contain every real number r such that Sa n x n converges if | x\ =r. and suppose that lim \a n+l /a n \=l Then lim = l\z\. 17. the character of the series is not determined. the interval (~R. R) with centre at the origin and radius R is called the circle of convergence. then. Nothing is said as to what happens when | x \ = R. such that Za n x n does not converge if |x|=r'. 13 and 14 that every r is less than any r'. suppose that there is a value x' of x for which it is non-convergent. With regard to the series Za n x n . by Art. these may be included in (iii).R) is called the interval of convergence. or (ii) it converges absolutely for all values of x. either (i) it converges for x = and for no other value of x. All that has been said applies to any power series. This number R is such that a n x n converges absolutely if | x \ <R. a) into two classes. Divide the real numbers in the interval (0.14. real or complex. . Let Sa n z n be any series. Suppose that there is a value of x other than zero for which 2a n x n converges. // Za n x n is non-convergent for x^x v it is non-convergent for every x such that \x\>\x\. RADIUS OF CONVERGENCE 333 By writing R = Q. and is non-convergent if | x \>R. If z is on the circumference. For a real series Za n x n . and it follows from Arts. 15. The series converges absolutely or is non-convergent according as the point z is inside or outside the circle. the series is non-convergent for x = a. 16. Both classes exist. R is called the radius of convergence. Therefore the classification defines a real positive number R which separates the classes and may be assigned to either class. Proof. For if the series converge for x x 2 where I^l-H^il* by the last theorem it converges for x^x^ which is contrary to the hypothesis. At either end point Za n x n may converge. diverge or oscillate. real or complex.

// (a n ) is a decreasing sequence of positive numbers such that a n+l /a n ->l and a n ->0. where it diverges. r z 1 r | <k . We have T? z n = a n+l z n + l + a n +2* n+2 + > and | T?Z |<| a n+l z+ l \ + 1 a n+2 z+ 2 1 + . (iii) This follows from (i) and (ii) on substituting -z for z. therefore Za n z n is absolutely convergent when |z|=l. except that it diverges at the point z = . As in Art 13. (i) if Za n converges. 1). in which case a n cosn0 diverges (Art. the following theorem is often useful. except at the point 2 = 1.+ (a n + f n)z n > where |^| >0 as |z|->0.and it follows from D'Alembert's test that R = l/l. (iii) the same statements hold for the series (-I) n a n z n . then. In considering the behaviour of a complex series at points on its circle of convergence. (ii) If Za n diverges. we can find a fixed number k.1. Moreover. we shall prove that /(z) = a + a 1 z + a 2 2 2 + --. if lim | a n \ n = l. If Za n z n converges when z=*z lt by Art. Za n z n converges (though not absolutely) at every point on the circle. then lim | a n z n \ n ^l \ z |. i i Again. (ii) if Za n diverges. such that for every r. it is absolutely convergent when | z | <| z x |. 1). 13.. both Za n cos nO and Za n sin nO are absolutely convergent (Art.l) n a n is divergent. For 1/72 = lim a n+l /a n = l. (i) if Za n converges. and by Cauchy's te st 18. Ex. Denoting its sum by /(z).. both a n cosnO and 2a n sinnd converge unless = or 2&7r. . 11. and at a point on the circle of convergence a n z n = a n (cos + i sin d) n = a n (cos nO + 1 sin nO). when Z( . Ex. 9. and therefore . 334 COMPLEX VARIABLE 19. Hence the statement (ii) follows. 2a n z n converges absolutely at every point on its circle of convergence.

// Za n z n =Zb n z n for every value of z whose modulus is less than some number /z. i: z n+l U .. 19. For by Art.. Therefore 77 1 Z l " [ 2 1 1 TJ ->0 as 1 z *r 20. . for every n. then a n = b n for every value of n. r z|<| *i I z i T T . Criterion for the Identity of Power Series. z i z n + 2 1 i _i V and thus.f z "+ 1 a r z r \<k ~ . f *i J it |2 z l 1 Jdlence it z ^z 0.

17 that (i) The series converges absolutely if | z | <1. (n r + 1) . the series converges absolutely at every point on its circle of convergence.6 1 <any positive number. the series converges (though not absolutely) at every point on the circle..1. |=1. the point z is on the circle of convergence. where it diverges. so that u r ' = | u r \. we can divide by z. therefore a l + a 2? + + ( a n + 'V)^ n ~ 1 = ^i + ^ + ' + (^n + ^)2 . we have = 1 i where w is real and z complex. (lii) If l<n<cO. it follows from Art. By making | z |-^0. according as r n^r + l. however small. it follows that a == b Q .n "" 1 * Making z->0 as before. we have a l = b v and by continuing the process. i r n u/ r + 1 " Therefore u' r+l = u r '. that is according as . / u r 4. it will be seen that | . Let accents indicate absolute values. except at the point z= . Denoting the series by n r Therefore.. 21 . Since the above equation holds for values of z other than zero. MULTIPLICATION OF SERIES 335 When 1 2. We consider the convergence of the series (n-l) 2 n(n 1) . Binomial Series. and we shall prove that (ii) // n>0. a n = 6 n for every n.. 6 are constants. .where 77 and 17' tend to zero as | z |-->0. then if r>n.. and since a . and is non-convergent if z>l.

their sums being s and t respectively. v 3 . Multiply the terms of %-f w 2 -fw 3 + . that is according as n^O. After a certain stage... we apply the theorem of Art. u r ' is a decreasing sequence. After a certain stage. 6.1.. are convergent series of po sitive terms. then the series U l v l + ( U I V 2 + U 2 V l) + ( U 1 V 3 + U 2 V 2 + U 3 V l) + is absolutely convergent and its sum is st. 18.. or if they are absolutely convergent real or complex series. the terms of Zu r are alternately positive and negative.. to apply the theorem of this article. (See Art. <->0. If ?i> . or is constant and the series cannot be convergent. and V 1 ^v 2 + v 3 + . according as Z( . 18. // t/j + Wg-f w 3 + . This array extends to infinity on the right and below.>l and. and we shall consider two ways of arranging the terms so as to form simply infinite series..) Hence the statements (ii) and (iii) follow immediately from Art. 336 PRODUCTS OF SERIES Let s n . we have * 1 _u . 18. t n be the sums to n terms of 27w n . MULTIPLICATION OF SERIES 22. XVI. and arrange t he products as below.l) r u r is convergent or divergent.If n ^ ~ 1? u r increases with r. These are the conditions required in Art. and we have to consider the convergence of u r ' '. Ev n respectively . so that Zu r r is convergent or divergent. . 25. u' r /u r '... then if a n is the sum of the first n terms of the first n rows of (A). as shown in Ch. by v l9 v 2 ..

. Fia. (B) where the nth term is the sum of the terms of (A) between the nth and the (n .* * f FIG.. .. 62. . ... (C) . . The sum to n terms of this series is cr n . 62.. Again.. and its sum to infinity is st. 63.... Then the terms in (A) can be arranged to form the infinite series u i v i + ( u i 2 + W 2 v 2 + uflj + (ttjVa + U 2 v 3 + u 3 Vz + U 3 v 2 + u^) + .. where the terms are represented by dots and the nth square just contains the terms in cr n . ..... . as in Fig... Draw a set of squares..... the terms of (A) can be arranged to form the infinite series U 1 V l + (M l Vt + U 2 V 1 ) + (tijVs + Wg^ + tVi)*.l)th squares.

+ v n . // Su n and Ev n are convergent with sums s and t respectively and one of these series.t. we have two series which differ only in the arrangement of terms.. w 3 = v n _ 2 + t. Considering the array (A) in Art. the sum of the suffixes is n + 1 . . is absolutely convergent. The first series is absolutely convergent and its sum is st . 22.. it should be noted that. The series equivalent to E(x) and E(y) are absolutely convergent.... the series (C) will be denoted by Zd n . MERTENS' THEOREM 337 23..v n _ m+1 + v n _ m+2 + .+ . 5. . prove that. w 2 . where d n = uv n -f i/ 2 tv_i -^ w s l -2 + + w w Vi .. . in every term of d n ..where the nth term is the sum of the terms between the nth and the (n-l)th diagonals.. 6 and 19)... 23 and 24. (B) where w l = v n ..) Let s n . for all value* of x and y..+u n w n _ l .+ ~n ~n~i \!L l^Lil II ! n ii 2 l?_ ! In Arte. say Zu n . drawn as in Fig. w-1 + v n .. t n .. . therefore where <C = r. (Mertens. and generally w m . If the brackets are removed in (B) and (C).^ + v n .. XVI. 2v n .. . D n be the sums to n terms of u n . Ed n respectively.. it will be seen that s n t n -D n = u 2 w l +u 3 w 2 +u 4 w 3 + . // E(x) = l +S + + . this is therefore also true for the second series (Chap.st. 63. 1. then Zd n is convergen t and its sum is . Ex.

9. |^|->0. . then \P\<s'c.) Let s n . -f | n n | <. D n he the sums to n terms of Zu ny ZV n . then Hence D i + D 2 + D 3 + . Ed n respectively. Hence. XV. // Eu n and Zv n are convergent with sums s and t respectively. . provided that n-w-hl>/x. provided that m 4. \Q\->0 and s n t n -D n ->0. where m is to be properly chosen. as ?/?. . (Abel. tend to infinity. s + + v nThe series Zv n is convergent. and hence (i) we can find a positive number p such that | v r +v r+l +v r+2 + . and consequently w. and consequently We therefore consider separately the sum of the first m terms of the series in (B). .. .. (4). -f v n |< for r>/x..2. Q ^ W then with the condition (C).1 u m +% | + . and the sum of the remaining terms. . . . .Vi + n-Va ^ *n Now s n ~->s and /->/. then its sum is st. p 1 and /u.. and if Sd n it convergent.* Let p = u ^ + u#o 2 + . if $' is the sum of the convergent series 2?|t/ n |. which proves the theorem. 338 ABKIAS THEOREM 24.' . therefore by Ch. The conditions (C) and (D) are satisfied if n>2m and m> either of the two.(C) (ii) the sequence (t n ) is bounded . + u m+l w m . + D n . hence we can find a positive number k such that | t n \<k for every /?.. and therefore l^il* l^l' l w 'm|j are all <c. Also Q and since 27 1 u n \ is convergent. where c is an arbitrarily small positive number . Hence.(D) and then | Q \ <A'e.lastly w n -i = ^2 + ?. * This artifice is often useful. .2>//. we can find a positive number pf such that | v m f 2 1 H . t n . .

M . which is convergent. the series 1 . EXERCISE XXXIII Prove that the series : t .3 a(o + l) 1./ Under suitable conditions.5 .s n t. .. + .a' > 1 and divergent if a . is convergent if 6-l>a>0. 1 Therefore D = .. . . and the statement is true in this case also.+ ^ r . + k ..r 3 -f c? 3 r 4 . .3 1. 9.=-> > show that 2/w n is convergent if ' '-* ' & a . 22.2 + "a ~ 4 + is convergent . that (log ( 1 + x) J 2 ^ rf r r 2 . XV. // x ~ 1. +$ .L -f . if D is the sum of ZW W . ! 1 H. . -prove. is divergent. .. If = . 9. .r^l. the series for log ( 1 + a. (3). .^. for it is easily shown that d n <d n ^ and d n ->0 by Ch. 1.. 24 applies. Hence Art. .a' < 1.. . -. by Ch..3. ..+ -. // -l<. ( 1 + : -f .4.) is absolutely convergent and the result in question holds by Art. the rule for the multiplication of series gives // -1<#<1. XV. .1 V . l/^.).. d n = "-. /r n 1 /<^1 2 Also. 4-.^ W.5 a(oH-l)(o+2) 2 6 + 2T4 ' MULTIPLICATION OF SERIES 339 A rr u n nP + <M?~ l + bn v ~ z + .s/. '/* 4..d% + d 3 . 1 1.. its sum la 1.. (3J. and so is d l .3. ..T -^-f.

. show that *>i9 v z> ^ 3 . Show that.d^ -f d 2 . The series 2 a cos nO and 27 ~ sin nO are absolutely convergent.J e x . 8. u 3 . [Use the following : | cos n 6 | > cos 2 nB&%(l+ cos 2n0) and | sin nB \ ^ sin 2 nO ^ -J r ( 1 . The series 27 .Using Ch.cos 2n0). . the series is convergent. when is a multiple of TT. If u l9 u.. [Deaote the series by v l .1 is absolutely convergent for all values of x. 4 6. is a decreasing sequence of positive terms tending to ze ro as a limit.v% -f v 3 . The series 27 . n 2 n 2 7.. The series 27# n cos n# and Ex n sin n0 are absolutely convergent if | x \ < 1.] 10. the rule for multiplication fails to obtain a convergent series for /i__L_j__ L_j_ [Show that the series is 1 .cos nS and 27 . show that the series whose nth term is a n .sin n6 are convergent excepting that the 71 71 first series diverges when 6 is an even multiple of IT.5. 9.cos nO and 27 .. XV. if Q<p<l. is a decreasing sequence tending to zero as a limit.sin n6 are not absolutely convergent 71 71 excepting the last series. .. If a n = ( 1 -f . .] / ^\ -~ 12.] 9.. . 1 1 1 11. where 1 n r=I r p( n _ r+ i)v and r(n-r+l)^%(n + l ) 2 . Discuss the convergence of Zu n where If. (3).

Vapdermonde's theorem asserts that if m and n are any numbers whatever. therefore n 2 .j8-f 1. Vandermonde's Theorem. then Proof.L\ . . Under what circumstances is this true (i) when z = l.] 13.1 ? CHAPTER XXI BINOMIAL AND MULTINOMIAL THEOREMS 1. is one of the values of (1 -f x) n . 2. Also 27l/n 2 is convergent.j8 > y. prove that L *Y 1 -^-y(V + 1 ) yJP(a. If FfrlKX^l+x + **^ and \x\<l.-t-r~-. and only if. *)=a(l -*)F(a-f 1 9 p + l. which enables us to deal with the case in which n is any rational number. where x is any number whatever.. (Of course. and (ii) when a= .[a n -l=-. y + 1. This leads to the following theorem. The following notation is often used . we write x r = x(x-l)(x-2) .*)-(y-a)JP(.. A General Statement. then n 1 *+ro*+. n is a positive integer .. First suppose that m and n are positive integers . (w-r + 1) ^ _J_^_J x 2+_ +__V ---. in which case the sum is (1 +x) n .. The Binomial theorem asserts that the sum of the series n(n-l) n(n-l) .) The series terminates if.. (x-r + 1). this value must be specified.y + l 9 x).+ r : 1 Z r The coefficient of x r in the product of these series is therefore equal to the coefficient of x r in (1 -f x) m + n . that is to say.J x r + mtm when convergent. in a complete statement..| 2 as n->oo..|a n -l|-*-|2.

/"(/A) = so that /(m)... This equation holds for all positive integral values of m and n . -f.. .. n. The series is absolutely convergent if | x \ < 1 ...A^x r -f . .f(p) =/(m + n +p).. the sum of the series n(n-l) nln-l) . . We first give what is essentially Euler's proof.^r ' Multiplying by r. demote its sum by f(n).. where d r = ^+ -^l + . f(m) .+^- Hence..-^A + .d v x -I. Case of a Rational Index./(n) . BINOMIAL THEOREM 341 3.. and since each side is a polynomial in m.. . f(n) are absolutely convergent... Eiderjs proof.. \n r + i j.. O\ f M A* I 1 \ n r ~n(n i)\n &) ... although in the last step we have to assume a property of infinite series which is not proved till later. f(n) = 1-1. ./(n)=/(m + n) . therefore by the rule for the multiplication of series. and let Art M f M 1 \ ///. we obtain equation (A). n it is true for all values of m.... . // n is any rational number and !.d z x 2 -f .... by Vandermonde's theorem.. (A) Hence /(m) . the series corresponding to f(m). so that n 2 2 w r r For all values of m and n.n) .r m r . and /(/).. (n-r + 1) 2 ' r is the real positive value of (1 -f x) n ......._ ^r-1^1 ^r-2^2 .../(p) =/(m 4.

. (ii) // n and x are real and \ x |< 1. With regard to sign... that is f(n) . If n=p/q where jo. the reader will find it a very useful exercise to work through the details. = (1 + NOTE. provided that | x |< 1.. We shall prove that if n and x are real and \ x |<1. Putting m~ -n in equation (B). it will be shown in another volume that f(n) is a continuous function of x in the interval - 342 BINOMIAL THEOREM Now f(n) does not vanish for any value of x in the interval.. . if n.... Second Proof for the Case of a Real Index... it can be shown that f(n) is continuous f or all values of n. /( -n)~ (1 +x)~ n ... then .. f(x) is one of the values of (l+a:) n . we conclude that /( " ) = (1 +x)* . . (i) Equations (A)-(C) hold when x is complex.... In particular....* Hence /(~j is a q-th root of (I + x) . is a positive integer... Negative index.... (B) Thus.. when n is irrational .... Hence if (l+x) n is defined as in Ch.... XIII. although it is unsuitable for reproduction in toto.. (0) therefore /( . we have (1 -Hx) n f(n).. by the preceding... The following proof depends on very elementary considerations and.. x being real.n) = . and consequently its sign is the same throughout the interval..... Positive index.... l.. \'x\<l and n rational.a similar result holding for any number of factors.(1 -f x) n . 4. </ are positive integers./(n) = l. But when cc = 0. therefore the sign is positive and/(n) is equal to the real positive value of (1 -f x) n . if the real positive gth root of (l+x) p is denoted ? /p\ ? by (1 -fx) . 7..... Let n be a positive rational. Henc e if x is complex.....

...x _ n(n-l) .. then 1/n. r has the same sign as t nt r+v After a certain stage. (C) . Therefore (l+x) n lies between s Wjt . then /i \* i n(n-l) 9 ...n so that y njl = (l -cc) n -1^0 -according as wgO...... . . and n(w-l) .. (B) If for any assigned values of n and r it can be shown that y n -\. Changing the sign of x... (A) Also ~j^ = ny n ^ r ^ and y n . (w-r + 1) v ' 2 ' lo ..... . and * Wff+1 for r>&.. r = when x = . s n> y^ s n> r ~^0 as r->oo .... T* where (1 + x) n has its real positive value. (n-r-f 1) (l~x) n = l~nx + -~2~~ x 2 ---+ (.. n-2.. it has to be shown that if 0<z<l. Also let First suppose that Q<x<l... which proves this case of the theorem... as r increases. y nt r must have the same sign as ny n _ lt r-1 ..... Here ^^ = (1 ~^) n ~n.1 ) -^ .. r .n(n-l) n(n-l) .. it follows that the three expressions y n . r-i i Q of invariable sign for 0<z<l... r+i> an( i consequently those of y nt r are alternately 4. Next let x<0. are substituted for n and observing that. by (A).-x r + . Proof. say for r > some fixed number k.... the signs of ALTERNATIVE PROOF 343 ^n.. Therefore y n ...and .. Let t nt r be the rth term and s nt r the sum to r terms of the series. n(n-l). Repeating this reasoning when n-\.^T-* .. .. .. then by (B).. ! = (!+ z) n -150 according as n^O .. But since the series is convergent....(n-r-f-2)2/ n ^ r4 .....r +i t i ^ as ^ e same sign as n-r-f 1. 1>1 . Therefore s W| r -> ( 1 + x) n as r -> oo .... ... (n-r + 1) have the same sign.... y n .

r~*Q SL8 r -^ since t ntr ->Q it follows from (F) that y n> r ->0... and we proceed thus : n . r -x ntr and (1 -x)y n ...r Therefore (l-x)* n>r = * n+1 ... r+1 ->0 as r->oo ...... therefore y n . . r ~ ^n. as before... r+1 .. and z nt r have opposite signs. r+2 so that by (E) z n>T has the same sign as (w.+ l)y nif ..r = when ^ = ....n . If yn+i.. r-fl Hence by (E). r + x ^n. it follows that the three expressions and (~l) r n(w-l) .1 it follows that y nt .... r = y n + lff + 2* n>r . (ii) //)/<-!. y n ^ r has the same sign as t nt r+1 .. r n . (n~r-f 1) have the same sign.t n> r+1 ^ 2/n+l.-i and y w .. r +i .a?)y n ..(1 -x)"" 1 ... then -M-l<X-w where //* is u positive integer.. r . 2 Wt r has the same sign as t n+lj r+2 . But ^ n .. we have +^. (E) After a certain stage.. (P) Hence ' (1 . y n _ r+l> t and n -r ~h 1 have opposite signs.also -^..... Now . the sign of t n> r+l is invariable..... say for r>k...... For every r.. r ->0 and s Wjr ->(l -x) n .. Hence.*..x) z w ..... (D) Reasoning as in the previous case and observing that by (C).. (H) 344 PARTICULAR INSTANCES (i) If n> ...... r = (1 .r = -^-i. Therefore y n>r lies between and * w .. r .

y n + m . Now when n= -1. Hence s nt r ->(l -x) n . r ->0.4. in each case it is assumed that | x \ < 1 : and in a similar way (l-z). y n +m-2.5. So that nir ->(l+z)n when w is a negative integer.. and therefore. then y n> r ~>0. Also (1 +z)y nj r = \ (compare equation (F)). 5. 1. and the theorem is proved in all cases. 21) and J n>r ->0 as r->o> . The case in which n is a fraction is considered later. XX. The reader should be familiar with the following particular instances . if y n+l} r ->0. by case (i).r a ll ^nd to zero. r . Hence as before.4.r> 2/n. . 6. Hence in succession y n . . r ~>0 for n== -2.3 .4 3. n(n-l) 2 2 Using the same notation as before. -4. z complex and \ z \ <1. We shall prove that if n is a negative integer. (iii) say for what values of x each expansion is valid. The theorem is obviously true when n is a positive integer. (ii) in descending powers of x . Expand l/(3 + 2a:) 2 to three terms (i) in ascending powers of x.5 NUMERICALLY GREATEST TERM 345 Ex. -3.3. Hence in succession it follows that y n + m _ 1 . 1. we see that the series is convergent (Chap.5 3.6 n(n-f-l) 1.2 = l+2z 3 . ..Now n-h m> -1.

o o 3 /5. w r r Let ifc be the positive integer such that i<n H.8.3. that is if | x |<f . O ^ J .12. 2. = -|.1 <k + 1. we have 7. Numerically Greatest Term in the Expansion of (1 + x) n where |x|<1.2. treating every term as above. where w~| and a. 5.4 which is equal to 2n(n + l)(n + 2)(n + 3)x*/\ 4.3.4 ' 1. Let u r be the rth term.^s 1 !-! * .16 2. Sumtheterie* 1 + + ~ To see if this can be done by the Binomial theorem.. J _______ ".f. and proceed thus : 5._.27 (iii) The first expansion is valid if | -|ar | <1.11 f. take some particular term. and let accents indicate numerical values so that x' = | x | and u r f = | t/ r | .\3 __. s ay the fourth. Let 5 be the sum of the series . The second is valid if 3 1. t/ n is positive. . \4 ' 8. that is if Ex.

the first term is the greatest.. the I ~r~ X tth term is equal to the (i + l)th. where i is a positive integer or zero and / is a positive proper fraction. then l ' therefore w r+1 ~w/ according as (m + r-l)x' = r. then w r _}_iw/ according as (n-f 1 r)x' . and each is greater than any other term. ..m. and the numerically greatest term is among the first k 4. (ii) If = i +/. .is a decreasing sequence.& -* 1 346 NUMERICALLY GREATEST TERM (i) If ~ j . numerically. and each of these is greater than any other.therefore ^4-1. (m-l)x' (iii) If ^ f = i.r~ ~i+f. .. the terms in the expansion of (l+x) n either decrease numerically from the beginning of the series or else they increase numerically .= i.. where i is a positive integer. ^ that is according as r dr.(m-l)x' that is according as r ~ -: -. 8 ' It follows that : (i) // m^l. . We have thus proved the following important property of a binomial series: // |x|<l. let n be negative and equal to . If r^. where i is a positive integer and / is a positive 1 ~\~ x proper fraction.1 terms. the (i + l)ih term is the greatest. (ii) // . Next.. <.r. then. where t is a positive integer.k. . % +2j ... the (i + l)th term is numerically the greatest term.. the iih term is equal JL "*~~ X to the (i-f l)th.7 . . .

1. 1. as follows : Ex. We have 3* 5 s L 1 1.. and each of these is greater than any other term.2 .)" 2 "" when x%? Let u r denote the rth term . CALCULATION OF ROOTS 347 8. 125. Approximate Values. The perfect cubes are 1. the sixth term is equal to the seventh. 8. 64. The Binomial series can be used to obtain approximate values. M r+i u r Hence the greatest term or terms must be included among the first 1 4 terms of t he expansion. u r+i %^ -r + l 4 54 . that is according as r ~ 6. For such values of r.4r ^ 5r. Hence. Ex. Which is the numerically greatest term (or terms) in the expansion of (I + a. 27. If r ^ 13. Search for two of these such that one is roughly twice the other. and from (A) it follows that | u r+1 1 rr \u r | according as 54 .up to a greatest term (or to two equal terms) and then decrease numerically. Calculate */2 to six places of decimals.. u r+l /u r is positive. .. This will be the case if 4r>54 or if r ^ 14.4r u r r 5 5r (A) First consider the terms for which u ril ju r is negative.

... Now....... (B) To estimate the possible error in this result..= Jj =..= 0-000 001 07 . in the seri es (A)... The error arises from two causes : (i) The error caused by taking s 5 as the value of '^2 is R 5 . let s n be the sum to n terms of the series (A) and R n the remainder after n terms....0-000 08 8 = ~ 4 = 0-000 OOP 02 ' 0-000 080 02* Therefore..2. we have =1*25....3 ......u 3 + u t ..u 5 + .. ^ -0-01. also each . . '2 = 1-260 001 07-0-000 080 02 nearly = 1-259 921 05 nearly ......L 1...... the terms beginning with the second are alternately positive and negative... .-- 1-260 001 07 *. ... + * 5 9 ""/ ( ' Denoting this series by u t + u 2 .

~-u 5 = ~~u 5 < 0-000 000 000 4. (ii) The error in the calculated value of S B arises from the errors in the calc ulated values of w 4 and u 6 .C. u= -^ . where p. p' 9 q' are independent of x. 2. we see that the error in the result (B) is n umerically less than 0-000 000 010 4. hence the error in th e calculated value of s 5 cannot exceed 0-000 000 01. If x is small and n>l. 348 NEWTON'S APPROXIMATION Choose a and b so that a-b-l/n and (a -&)&-=(&. Substituting these values for a and 6 in (B).A.is numerically less than the preceding . These cannot exceed 0-000 000 005. b-^(n-l)l2n and a = (n + l)/2/i. q. /2 = 1-259921.l)/2w 2 . i Ex. obtain an approximation to (l+x) n in the form p + qx -.'. therefore J? 5 is positive and less than w e . Thus. /. we have . . We have and (B) B. the error is less than 0-000 000 000 4. Adding the results of (i) and (ii). 5 5 3 5* Now . to six places of decimals.

on J Denoting the series in brackets by MI . ^ ~ 1 n 2 .. .-x-x 2 X < 1 -.(C) -l)x n 2 \nj \ 2 J \n J v ' If x is small..1 proximate value of (1 -f x) n . it can be shown that _ i __ .--. .. // n>l and 0<x<l.nearly.rf. . and since 0<x<l. .x) .. _ OX . it follows that .. - 9.= 7 .r ~ (r . . Theorem. for ( m _ 1 ) r _ ( r + 2) (r . .^ 2 + ^3 .LX -h .. on" 5 ^ 4n 4w .2) n<0 .l) r '^r + .} . w r _! (r-f-2)n r-1 Now (rn-l)r<(r-f2)(r-l)?t provided that r>2.+ (.1) n . i. .. w/A an error in defect less than -ys~3 ^ 3 By multiplying each side of equation (A) in the last example by 2n + (n.x. it follows that u r <u r _ l . (1 + x ) n ^ _ ---.l)x. u r rn 1 r we have . the series (A) and (C) are both convergent. and since they are id entical as far as the terms involving x 2 . The sum of the series is therefore positive and less than unity .

5/504 is an approximation to \2 with an error in defect less than 0-0000005. and find which is the first of this series of positive terms.27 liM 3 10 8 10 7 KXAMPLES OF APPROXIMATION 349 EXEECISE XXXIV 1.. the third term may be the greatest ? . which is the first negative term ih the expansion of ( 1 f x) *~ ? I 9 2. an approximation is 5750-4-12 635 . 1. Show that 63. whence the result follows. Prove this.and. (1+x)*' when ar^J ? 4. (l+x)~* when *---{?? 7. the terms in the expansion of (1 -x)~* . Here v2 * ( 1 + j J henco. in the expansion of _ 3 (I -x) 2. Between what numbers must x lie in order that. since n> 1 and x is positive. (l-x) 2 when *-:&? 5.. are all positive. beginning with a certain term. If x is positive. Which is the numerically greatest term or terms in the expansions of 35 _ 3 3. Ex. If a: is positive. (l~x)~ when x = ]l 6. i r * with an error in defect 4 750 i 6 ' r 504' -. the last expression is less than (n 2 -l)rryi2n 3 .4 12.

15.nearly. If 0<a. then *Jx = . // we have tables giving the square roots and cube roots of numbers to f sig nicant figures. (ii) 1 -3# f fix* is an approximate value of 1/(1 +x)* with an error in excess less than 000001. 3 ^ 2 9 x 5 the error being in defect arid less than 5/81 a. 13. which is the first negative 1 term in the expansion of 9. If 0<#<1. 2 is an ap])roximate value of 1/(1 ~x}* with an error in defect less than 0-000005. show that s/a.8. 1 /x \ If *Jx~a to f significant figures.r. show that (i) Vl 4 -x 1 +lx-{x.~ nearly.<0-01. show that (i) l-f2a.+ a ) to 2f significant figu res.8 . 2 \a / . show that 14. show that the roots can be found by division only. IfO<<l and /? n is the remainder after n terms in the expansion of 1/(1 -ir) 2 . If 0<#<0-01. If 0<#<1. (ii) 1 ~2x + 3x'' is an approximate value of l/(l+x*) 2 with an error in excess less than 0-000004. If x> 0. neglecting cubes and higher powers of e. (ii) \/l ~x~ \ ~\x . 11. If x is large and positive.. + 3a. and R n is the remainder after n terms in the expansion of 1/(1 -x) 3 .nearly * the error being in excess and numerically less than - 10. If is small. + 6x 2 is an approximate value of 1/(1 -x) 3 with an error in defect less than 0-000011. show that (i) l+3o. the error being in defect and less than ^x'^.( . to 2f significa nt figures as follows. 350 APPROXIMATION TO ROOTS 16. show that. show that 12. 3 -t \~x-\.

25.).] 17.3*) .|z) --. In each of the examples 21-26. Show that (i) if x is small. *J(x* + 4) . (iii) ( 1 + 4a. Given v/2 1-41421 (approx. then $x = -( + 2a ) to 2f significant figures. (i) prove the given equality . V2 =-J(l -0-02) "i 23.).154434690 (approx.*J(x* + 1) ~ 1 . 4/4 -f(l+ 0-024) "i 24. 19. that >/2 = 141 421356237 (approx. 15. that $ 10 = 2. deduce an approximate value of the given surd . show.-iVr) . (iii) find an upper limit to the numerical value of the error. the error is in defect and is less than tt 2 -! /a-6\ 3 "12^ V~T/ ' . # (8 + 3*) -r (9 . 3 / 5 21 \ (ii) if x is large. [Put a *Jx-. show that ~ ^ and that if a>6.(1 . and use Ex. #5 =(1+0-08)*. 26. show. 21. If a is nearly equal to b. (ii) by expanding the right-hand side to four terms. 4 7 .1 /x \ If $x-a to f significant figures.). v/(z 2 +4) -V(x s + 1)= 2 ~ ( 1 ~ ^ + ^) nearl y20.i* 2 + ~ 6 7 4 ~z 4 nearly . */3=i(l-0-01696)*. Given 4/10 2-1544 (approx. ^1001 = 10(1+0-001)3.). 22. -f 10x 2 ) * = 1 -f 2x + 3z 2 (iv) (v) ( I -fa: + * + *)* = l+ix-h A* 8 + ^5**. If x is small. show that approximately (i) V(4 . #4=f(l 27. by division. 18. o \a~ / Show also that in each case the error is in excess. by division.

] /543 2709 ' 28. (1 =coefft. of x n ~* in (1 -a:)-< . If a -b=^h. 9. ^ 54 2706 3/10*.. whore h is small. 10. -f w n _!a. show that a I h I h' 1 I 5 h* ft" + 2 ft~8 6~ a + T6 ft~ 3 ~T28 b** '* 1 1 7i 1 A 2 1 & 3 1 &* 2 + 4 ft" 8 ft" 8 " 4 " 16 ft" 3 ""32 6* + "" " (iii) Hence show that if a is nearly equal to 6. n< " 2 + . Show that J 5 ^^ = ' C ^TT O nearly.[Assume the result of Art.. then Va a I a + b b = a + b + 4~'b'-' nearly ' the error being approximately equal to (a-ft) 4 /128ft 4 .2 in (1 -*)-(-). [g=coefft. Ex. = |g (1 and %--f v Hence if 8 is the sum of the series in question. the error being in defect and less than 10*. therefore w l +i/ 2 o. + w 3 a.b)/b..* If prove that It will be easily seen that w 2 -|?.. of n . 1. SUMMATION OF SERIES 351 29. and put x = (a . (1) Many series may be summed by expanding some function of x as a convergent power series in two different ways and equating coefficients. Application of the Binomial Theorem to the Summation of Series. 2 / -f-.(g + l).

* This equality is required in the Theory of Probabilities. /i .r) n . .. so that all the series which occur are convergent. r-1.})x 2 (l +x) n ' 2 Ex. dividing u r and successive quotients by r. n . 2c 2 + 2 . . Find Z*^ (r + 1 ) 3 c r x r .. .(2) Theorem. and.~. A L . A 2 .a -f a^ + a 2 x 2 -f a 3 x 2 + . . in this order we can find constants A Q . fix) ~ z = For x x and the series are absolutely convergent for sufficiently small values of x. 352 SUMMATION OF SERIES (3) In what follows. + a n x n + . -f c r x r -f .* 2 = l . // f(x) . . we find that .c 2 x 2 4. as far as required. +a n is equal to the coefficient of x n in the expansion of f(x)/(l -x). . we suppose that | x \ <1 except when n is a positive integer. . .x 2 + . w ( 1 H. Z^^u r c r x r ^At(l+jc) n + A v nx(l +x) n ~ l + A 2 n(n... fo r \ x \ <k.. . .c 3 x 3 -h . . . . Hence we can sum the series *_ Q u r c r x r where u r is a polynomial in r. 3r 3 :r + ...r 2 -f .. ! l r = c x 4. We shall write (1 + x) n = c 4. . then a + aj-f a 2 -f . .* = w -f -^r ^ x 4. .~~ ----.. n(n-l) n(n-l)(w~2) a then. . Denoting the sum by *V and proceeding as above.-.2c 2 x -f 3c 3 . such that then. 2. . -f r(r - and so on. where k is any positive number.. A 3 . + rc^ r ' 1 -f . .c^x 4. For. by a succession of similar steps. r-2..

(4) Again. we can show that (Ijfx^^^ 1 GI 2 _c. " .3 j __ _ ?r _ . if u r is a polynomial in r. we can sum the series r . * Tlirse results can also bo obtained by differentiating with regard to x.r+3 . 1. . altho ugh so far the process has not been justified in ease of an infinite series. . .2 i i " -.8 }l -f (7 4 3) a: -f (6/& 2 + 8w + 3) x 2 4 (w + 1)V}.i" -i~X "r~ n + 1 ?i + l 2 r-fl and. by similar steps. S=-(l+x) n -f 1nx(l +ar) n . Jv \ i (n + l)(n + 2) (M + l)( + 2)'n + l 1. These results can also be obtained by integration.) n .l)x~(l +x) n ~ 2 + n(n-l)(n = (1 + a.v..(r + l) 3 -a +7/' + 6r(r .l)(r -2) . SPECIAL TYPES OF SERIES 353 Hence.1) +r(r .3 and so on.2 (r + l)(r + 2)' (l+x) n + 3 1 1 X + ' 2(n + l) ^. by expanding (1 +x) w+1 .2.'. ~~~^. l "V" " " ^ i ^n^ ' r *^ i. " ^.1 + 6n(w.

. we can find an equation of the form u r A l A% A k "*.. i ~~~ ":. \ / ^rr i~ . r r + 1 (r + l)(r + 2) '" (r + l)(r + 2) .r+2 rr x r + The value of Zv r c r x r is found by (3). . *"-** (MMT. Further. r + k . . r + 2. -4fc_!. n -Hi z 2 (i+x) n +*-i v. and the other sums are given by the above equations.. r -f 1 in this order. . x i i 2 r i * .. Find S = 2^1 -s <V f We have . A l are the remainders in the successive divisions and v r is the final quotient. b. For this can be reduced to the preceding by multiplying numerator and denominator of the fraction by such factors as will convert the denominator into (r -f 1 ) (r + 2) (r + 3) . (7TIj(iFT2MF?S. . (r 4. r / ' . we can sum the series where a. 2. Ex. . and then the sum of the series is zvx + z ~~ r+1 .r ' For..1.k) .. . k are unequal positive integers of which k is the greatest. dividing u r and successive quotients by r -f k.r + 3 hence.. where ^4^...

. If 6.fcc m ) n ... . . 4. then (i) /(x) can be expanded in the form 1 4 ux + u 2 x* 4. 2. then by the Binomial theorem * . Multinomial Theorem for any Real Index. .. n is real.3 H r 4.3 r + 3 and 27(r f. c. .. 11. .-f/c = For let y = 6x + ex 2 -f .1) ( H--2) .&# 4 ex 2 4. i are any m numbers. . and f(x) = (1 4... .u r x r 4... K: Aave any positive integral or zero values such that and the corresponding values of a are given by a-f/3-fy-f-. and in other cases is absolutely convergent for sufficiently small values of x.. 4.^(a + 1) . (ii) The coefficient u r of x r is the sum of all terms of the form n(n . where is the same as in Ex.a. .. /B y ^ ~ B'Ty/'-E c "" ' where /?. n-fl d+*) n+l 354 MULTINOMIAL THEOREM yiOO ^r^O 7+3 tr "" ' Wo have ~ = r . y .fcr w .. where the series terminates if n is a positive integer. 4.

... | c |.. and the corresponding values of a are given by a+/?+y + 8=~-|-.. .. In order that the terms of this series may be arranged in ascending powers of x.. y. y. (a + 1) v _JLI *_ I . .. it is sufficient that the series is convergent when all the terms are made positive. s being any positive integer.. 2#( .... y. (1).. + k'x' m <l.. |fej. (C) It follows from (A) that the coefficient of x r in the expansion of f(x) is the sum of all terms of the form n(n-l)... b'....2) 6 . . the coefficient of x r in (bx + ex 2 + . ... stand for \x\..+ K S ...... we can find a positive number p such that |y|<l when |x|<p..... 4 kx m ) s is the sum of all terms of the form EXPANSIONS 355 where j8... . for y in (A) and expanding the terms. 8 have positive integral or zero values such that /J + 2y + 33 -4... ...... 3... and by Ch. The statement (ii) follows on writing a = n-s. The term corresponding to the first set of values is a ft y S . Again.. (B) and /?+ y-K... . ic have any positive integral or zero values such that jS + 2y + .. .(n-* + l) fl T-7f i j c. r ^4. .. c'... Ex...v -_ _z i 1 . JPt'nrf f/ie coefficient of X* in the expansion of (\ + 2x .... and we can find a positi ve number a such that this is the case if | x \ <a.... and the coefficient is the sum of all terms of the form ( ~i)( -i--l)( -i ~2) .+w/c = r. . where x'. where j8. . .4# 2 . c/ .. where the values of j8.3 ~~ 8 Here n~ --J.... /S. . By substituting bx + cx 2 + .. The possible values of a... 1...provided that |y|<l.. by Ch... are given by (B) and the corresponding values of s by (C). 16.4) v ( . we obtain an expansion of /(x) in powers of x which is convergent if | x\<p.. This concludes the proof of the statement (i). This will be the case if b'x' + c'x 2 4...... y. Ill. 8 are shown in the margin. XVII. A...

and | 2 | <1.( 1 . show that the coefficient of n+r-i ^ t he expansion of (1 + x) n /(l -a. 6. show that 3. 2. 4. + a n _! = $n (n 4. If n is a positive integer.x)} n l ( 1 . 2. 0. 3.) 2 is (n + 2r)2 f| . 5 n(n .j 2> > ' -3i. If n. -2j. -4. so that the required coefficient is -3 + 15+4f + 6=22-|. 356 COMPARISON OF COEFFICIENTS 5. ^. 1. EXERCISE XXXV 1. [Expand {2 .a.: r L 2 2.2) (n + 7) 2 n . Similarly it will be found that the other terms are 15. if z = ~^.l)(n -2) (n + l)(n + 2) . Show that. 0.1 . If w is a positive integer and (1 +x) n show that + c*i + a.4 . r are positive integers (including zero).4n(n-l) 4 .] 4.) 2 . + . By considering the expansion of (1 -a:)~ n . then x + y n-f2) . show that n(n + 1) n(n + l A ~r H H r^.

where n may hav e any value.fc+1 >. 13.. if (1 -fx) n = 004-^0:4. +0^+ . show that [From the expansion of {1 -# (2 -x)}~ 1 = (I -x)~ 2 . .(r-h + l) and <f> (x)-A Q x h 4. Prove that CfCfJlJI + Cf^Cfilf H-(7J +1 C. show that p n [Divide by p r . 12..X )~^. for all values of m and n 9 = the coefficient of x r in the expansion of (1 +x) n ~ m . Prove that if n is a positive integer. put 1 .. 2 x 2 4. If n is a positive integer.] In Exx. !?L n ( n "" 1 ) JL??_ n(tt-l)(n~2) " f "~~ ' ' + " * ~ ' 8.= coefficient of z n in {2 ..A l nx h ~ 1 + A 2 n(n.. 10-13. If p 4-^ = 1 and n.. find the sum of the series.Si. . 14.r . If u r =Ao + A l r + A s r(r-I) + . r are positive integers (n>r). + .) n x (1 +a?)-< n . [Consider the product (1 -x)~( T +^ x (1 . prove that 15.. 6.q for p.T \] 16. 2 r==Q mc r x r 9 when m is IL^J. Show that..x)} n l(l ..] 7. and show that the coefficient of q k on the le fthand side when expanded is ^coefficient of z fc in (1 +3..+A h r(r-l).c.-.(1 .. + A h n(n ~I) .. to 40 terms -6' JJ.8)2 W ~ 3 .l)x h ~ 2 -f . Show that 1 3 2n + a -3 .!-=.. .x)*=(n z + In 4.( n-h + l). 9..

then by Ch.2714-3* the summation being taken for all positive integral values..^>) 2n+3 . and only if. including zero. of the equation Q Q. . | x \ < | A |. may be written in the form P a + a 1 a. The rational fraction P/Q can be expanded in a convergent series of the form U + u l x + u 2 x 2 + . If it can be shown that the series obtained by any process is convergent for all values of x in any interval including zero._ Q l+piX+p^x 2 + .. +u n x n + .(6 -f c -a) 2n + 3 . All of these terms can be expanded in a convergent series of ascending powers of x if.] * Check the answer by putting nl. 20. +p r x r Various ways of expanding such a fraction in a series of the form Zu n x n are given below.. real or imaginary. if and only if \ x \ < | A |.-. RECURRING SERIES AND DIFFERENCE EQUATIONS 1. where A is a constant. .(c 4-a .r .(a 4-6 -c) 2n + 3 . For. Expansion of a Rational Fraction. Theorem .. Any rational proper fraction in x..)~.. where A is the root of $ = with the least modulus. any other method of expansion yields identically the same series. r a positive integer and a is any root. P/Q can be expressed as the sum of several terms of the form A(x-a.. 2. [Consider the expansion of (a 4-6 +c) 2n+3 .. where A is the root of $ = with the least modulus. + a 2 z 2 + .. CHAPTER XXII RATIONAL FRACTIONS (2). +a m x r _ . whose denominator does not contain # as a factor. by the method of partial fractions. of p y q 9 r such that p + q + r n. XX.

Ill. A series having this property is called a Recurring Series. 3. ^5. after a certain stage. Show that.) Ex. the roots are -iV 2 and t(i and their moduli are -s/2il.4) 1+1+1+1 + 4 + 12 + 16 + 16 + 32 (a) -3-9-12-12-24 (b) + 2+6 + 8+8 + 16 (c) . Ex.When this is the case. For what values of x can the fraction be expanded in a series of ascending powers of x? Equating the denominator to zero. the expansion of P/Q can be obtained by adding the various series. . 1. Hence the expansion is possible if and only if | x \<*J~ .1. Methods of Expansion. any r successive coefficients are connected by a linear relation. 1. Q are as described in Art. (See Ch.(2 . (1) Suppose that where the series is assumed to be convergent and P. 1. Multiplying by Q. (2) The process just described is equivalent to synthetic division carried out as in the next example. 1 . tjie values of u Qi u v u 2 .3 4.. Expanding the right-hand side and equating coefficients. are determined by the equations Hence.. for sufficiently small values of x 9 1 + X + X 2 + X 3 The reckoning is as follows. 358 RECURRING SERIES 3. >/5.

+ 12 diagonally in (c). where 0=tan". (3) The method of Partial Fractions. the next term in (d ). Show that.2tsin(n + l)0. 2 + 1-3. +4 diagonally in (c). he will see tha t the reckoning is essentially the same as that just described. where 0=tan. 3(2-3+4)^6-9 + 12.5*( n . so that we may take a = 1 + 2*. 6 .! 2.9. (1). + .l 2. -3. Ex. 6 . 0=V5(cos0 -tsin0). find the value of u n . put 2. Using this method and that -of Partial Fractions and equating coefficients.px + qx 2 ) can be expanded in a convergent series of ascending powers of a.2x + 5 =0. j8 are the roots of x* . a n + 1 -/3 n ^ 1 =5*( n + 1 )... put 6. /. is as follows : . Find also the coefficient of x n in each e xpansion. We have 1 1 = _1 /__?_ _ _J )~a-8 \1 -a* 1 -j _ -j8*)~a-j8 where a. 1(2-3 + 4)^2-3+4. (6).) 2 (l-2*)' v ' 1 as far as the terms containing a.2.2i. If 1/(1 -2ar+5x 2 )=w +w 1 a.3 + 1 =4.fl )sin (n + l)0. (a). . ft = 1 . as in the following exercise.1+3 + 4 + 4+ 8 + 20 (d) The first term in the quotient (d) is 1 . if p and q are real numbers. hence w n =i. (a) .+M w ar n + . Use synthetic division to expand w (l-a. HOMOGENEOUS PRODUCTS 359 (4) A rational fraction can also be expanded by the Multinomial Theorem. 2. If the reader will do this example by the method in Art. the next term in (d) . the condition that 1/(1 4. Hence i i Now a=v/5(cos0 + isin0). EXERCISE XXXVI 1. (6). we can obtain many important results. 3...

1 1 -px + qx*~~ l 2 "" n where a n =p n -(n.b 2 x~* + b v x~ l + a + a^x + a&* + . . prove that and 4. . show that d n +*(b-c) + b n +*(c-a)+c n +*(a-b)=-H n (b--c)(c-a)(a--b). [H n is the coefficient of x n in the product i.] 360 SCALES OF RELATION 7. then | x \ < \ (p' ..l) T C^. 5. where p'-\p\\ (ii)ifp*<4q..(i) if p 2 >4#.e. 3.0) can be expanded in the form . If | x | lies between | a | and | |. The result is obtained by the method of partial fractions. c.I)f ^~>-l)^-^-2) . If |a|<||<|j8|. If a + p=p. show that the fraction l/(x . in the expansion of 1/(1 -ax)(l -bx)(l -ex). Prove that.l)p n ~*q + ~ -p n ~*q* . I the (r+ l)th term being ( . The fraction = (a. 6.. aj9=5 f .pr j3 n ~ 2 g + . use the identity to prove that ...fix 1 -px + qx* / __ *>\ to prove that a n + p" =p n .a) -!/(*.Vp 2 . If a-f/?=jp. .] 5. .x 1 . . Li !j_ the (r + l)th term being ( . 4=9. [Let |a|<l)B|. use the identity f- 1 . then \x\<lf*Jq. 4. . equate coefficients and use Ex.]8){l/(a.<x)(x .*q)/2q |. for sufficiently small values of x. If H n is the sum of the homogeneous products of n dimensions which can be formed with a.(n-2r + l) yn _ ifgr [Expand both sides. . . p w ~ 4 g 2 .<x. 6.zr q r .J8)}.r p n . where the series extends to oo on the right and left. .

Hence if the first 2r coefficients are given... Some authors call equation (A) the Scale of Relation of the series .a p the (r + l)th term being (-1? ( *~ r " ^ n ~^ "' (n -*V-...-.. the subsequent coefficients can be found in succession by equation (A)..] 8... (1 -x)(\ -x 2 )(l . and in one way only..^ 2 (2 cos ^ ( "... Recurring Series. We shall take it to mean either of these things.~ (2 cos B) n ~ 2 + r ^r~ (2 cos 0) n ~ 4 .4) (2 cos *) ..(2 CO8 )n -sr... u l9 ..... .. be a series in which any r + 1 successive coefficients are connected by the equation u n +PiU n -i+P2U n _ 2 +.. Li _ the (r + 1) th term being ( . p = z~ l where z~ cos 0-fi sin 0.... Show that (i) the coefficient of x m in the expansion of -~ -. others use this term to denote the polynomial 1 +p 1 x+p 2 x 2 + .. If % is a positive integer... 6.] 1 +# 9. 5. Thus a recurring series of the rth order depends on 2r constants. prove that (i) 2 cos nB = (2 cos 0) n .3 y. p r are constants... .+p r u n _ r = ) . sin I 1 12 the (r+ l)th term being ( . Also it can be continued as a recurring series of the (r + l)th order in a doubly . (1) Let u Q + u 1 x + u 2 x 2 + . (ii) 552* (2 cos 0r....is (n f 1 ) 2 . 7 put a = 2..1)' (n ~ T ~ l)(n ~*~ 2) "' ( "~ 2r) (2 cos 0)-->.. 7" [In Exx. is '' (ii) The coefficient of x* n in the expansion of .... (A) where r is a fixed number and p l9 p 2 . .x 4 ) 4. + p r x r . in general the series can be continued as a recurring series of the rth order. . u r _^ are known.l) r n ^ n ~ r ~ *) "' (*>. [Expand both sides and use Ex.. Such a series is called a Recurring Series of the r-ih order.y ... If the coefficients u .

r u 0.^ Wj = 0. q. % + p W 2 4.infinite number of ways. u & +pu^ 4.pu 9 4. so that the four conditions u l 2 -u^u 2 = Q 9 u 2 2 -u 1 u 3 =0 9 w 3 2 -w 2 w 4 0. Discuss the question of continuing 2u n x n as a recurring series when th e first six terms are known. For to do this we can give u r and u r+l any values. 1. This will be the case if "3 and (A) In case either determinant is zero.g?/ 3 4 . q..p^ + qu = 0. . If it is one of the first order.pw 4 4. w e 4. Now Su n x n can be continued as a recurring series of the third order if p. Then p. W 4 4-^ 3 4.l)th order.pu 2 4.r^j 0.gw 3 =0. That these equations may be consistent.#1*2 4. CONVERGENCE OF RECURRING SERIES 361 Ex. u 2 +pu 1 = 0.. It may happen that the first 2r terms belong to a recurring series of order r-1.qu 4. so that two conditions are necessary that Eu n x n may be a recurring se ries of the second order. u B +pu 4 0. Let the scale of relation be u n +pu n _ l + qu n _ 2 + ru n _ 3 0. suppose that the scale is u n ^-pu n _ l then u 2 4. and then the series can be continued as a recurring series of the (r . r have definite values and rr^O. the scale being u n +pu n ^ 1 =0. r are given by u 3 4.<?t*2 =0. u 4 z -u 3 u 5 =Q must be satisfied. Two conditions are necessary that this may be so.ru 2 0. each of the determinants in (A) must vanish. (2) Any recurring series u n x n is convergent for sufficiently small values . then Ut+pUQ Q. w 4 4.

.. of which the denominator is It follows from Art. For let the scale be u n +p l u n _ l +p 2 u n __ 2 +.. | u |. (C) where a = w .+p r U n __ r = Q. which is called the generating function of the series..... Continuing thus.p 2 x 2 4 ... is a rational proper fraction. ...... 362 GENERATING FUNCTION (3) The sum to infinity of any recurring series Zu n x n .... In the same way where w^ is the greatest of u^_ s _ l) u' n _ 8 _ 2 ... p%'. u$.. to QO . a l = w r 4. by multiplication we can show that | u n \ where A is the greatest of Wn-i> Wn-2> . | p 2 |........ . p.. . .....1 . .. 4 p r x?) = a 4. p 2 ^> p r xr an d adding.^WG.. and 2u n x n is convergent if | x |<l/y> for then | w n x n | is less than the (w4-l)th term of a convergent series of positive terms.. .. be denoted by U Q '.. | p l |.. Let the scale be U n +PlU>n-l+P2U n _ 2 +.... ... where # is the greater of the numbers Pi +p% + +p r ' an d 1.... .of x... .+p r u n _ r = 0.. u \ Thus | u n x n \^\gx\ n . we find that s ( 1 + p& 4.... .. a 2 = t/ 2 4 jo^ the coefficients of powers of x higher than x r ~ l vanishing on account of equation (A).. (B) Multiplying (B) by p^. (A) and let | U Q |....djX 4 a 2 x 2 4 . Then by (A). ... ....... + a r _ 1 x r .... Thus s is equal to a rational fraction.... when convergent...... . and u n ^ is the greatest of the set u^ l9 u' n __ 2 ....A. 2 that 2Ae recurring series is convergent if\x\<\ A |... UQ.. u. ... (A) and let S = w + w 1 x + w 2 x 2 -f .

when |*|<1. and the scale is (1 -a.x) = w 4 v^ 4 1 1 2 ^ 2 4 . where v n = u n -u n _ l . = -3 . Ex. . is a root a of the equation 1 .3 . + pu n _ 2 x n . Denoting the sum by 8. and S-- FINITE DIFFERENCE EQUATIONS 363 (5) The sum to n terms of a recurring series can be found by a process similar to that in (3).. Find the sum to n terms of w + u^x 4. Ex. we find that where & is independent of x. Find the sum to infinity of l 2 42 2 *43 2 a: 2 44V4. by addition. 3.) 8 . To prove this. where u n +pu n _ t + qu n . VIII. . 4 u n x n . By r steps of this kind. we can proceed as in Ch. Let 6 n =u + u l x+ u 2 x 2 + .1 + pu n _^x n t Since w r +7>w r _ 1 + ^w r _ 2 :: =0 for r>l. then s (1 .ti. . 2. 2 2 * 2 . . . = 3.. .. . so that v n is a polynomial in n of degree r-1. and therefore s(I-x) r + l ~u +(k-u Q )x. (4) // w n i a positive integral function of n of degree r.. . + u n _ l x n ~ l then pas n = 2?tt # -f >Wi# 2 + .Aer6 A i iA6 root of Q = wjAicA Aas $Ae Zeas^ modulus. 4 v n x n 4 . Let s = u 4 ux 4 u%x 2 -f . . . z = for n>l. *n (1 + JP^ + ^ 2 ) ~ u o + ( w i +P ?/ o) x ~ u n xn which determines the value of s n . unless a. which proves the theorem. . .. l*x -3 .u 2 x 2 + . 7. then Zu n x n is a recurring series of which the scale is (1 -x) r+l . .. This is a convergent recurring series. or in the last example..

(ii) expressing this as the sum of partial fractions of the form A(l-ocx)" m } where a is a root of x r +p 1 x r ~ l + p 2 x r ~ 2 + . j8 are the roots of x 2 +px + q = Q and A.. The general value of u n is called the general solution... that is to say they are arbitrary constants.. C are arbitrary constants. and shows that this value involves r arbitrary constants.... and the values of w . We give the proof of (ii).C .... The process leads to the following results.. Thus the process described in the last section enables us to find the general value of u n which satisfies equation (A). j3..In this case the sum may be deduced from the general result by supposing that #- (6) Given the scale w n -i-^ 1 w n _ 1 + .. If a = P 7^ y. where a... u v . (i) The general solution of u n + pu n _ l + qu n _ 2 ^=Q is w n = ^4a n -f J3/? " or (A -f wJ5)a n . 4-j? r w n _ r = 0. B are arbitr ary constants. B. From this point of view. +. 2 A B . then u n = (A+nB + n*C)oL n .. .. (ii) The general solution of u n +pu n ^ l + qu n _ 2 + ru n _ 3 = Q is where a.. //a^^y. (A) where p l9 p 2 . .A - 364 GENERAL SOLUTIONS (iii) For any other value of r the general solution of equation (A) has a similar form.. and the process of finding it is called solving the equation. the first or second form being taken according as a^/? or a = j8.p r w n _ r = 0..p r are constants. the solution is u n = (A+ nB) a n -f Cy n . and (iii) expanding the partial fractions and collecting the coefficients of x n . Examples are given in the next article... equation (A) is called a Linear Finite Difference Equation with constant coefficients. 5. y are the roots of x 3 +px 2 + qx + r = Q and A. provided that no two of the roots are equal. . Suppose that any r + 1 consecutive terms of the sequence (u n ) are connected by the equation u n +PiU n ~i+P2U n _ 2 + .. -f 7? r = 0.. the first r terms of the sequence may have any values whatever.. Examples of Finite Difference Equations. So far as equation (A) is concerned.. u r _ v we can find the value of u n by (i) finding the generating function .

l=. u v w 2 . (i) If no two of a. ..4a + ^. and u n = ^4o where JB'. then P/4 JD f^ ^L JJ w and u n = (4 -f nfi)a n + Cy w .r = 0.If w .ocx + (l^axp + (1-a.u n~i-u n -2^Q and ^ = 1.. where A. u 2 have any values. c are functions of ^ . l * 2 3 J 7: Q where a. = _ . 2-Ao^ + B^. Hence Un^AoP+Bp 1 .== 7: (say). %. and every other term is the sum of the two preceding tertrw. UQ + u^x + u 2 x 2 + . where a. (iii) If a = j3 = y. The first term of a sequence is 1. 1. . C are arbitrary constants. Denoting the sequence by u l9 w 2 . j8 are the roots of . we have u n . j8. y are the roots of # 3 + px* 4. j&r. 6.~ . C' are arbitrary constants. then * a + bx + cx 2 P . (ii) If a = /? 9^ y. B. Expanding the fractions on the right and equating coefficients. then P__A__ B (^ Q ~ T . . . + u n x n + . y are equal. . the second is 2. U 2 ^2. Also where a. j3. Find the n-th term.qx 4.

Denote the series by w -f u^x + u 2 x 2 + ...1 -i-[3 n + ( .4 -^2.J 8=v'5.*. Now a+0 = l.)"-'. giving p= -2. The roots of x*-2x-3^Q are 3. Ex. 3. a.w n-1 + u n _ 2 =0 ^w a rea/ /orm. 2. the sum to n terms of the series 1 -2 + 7.-. B from these equations.. q= -3. hence 2-y8 = l+a=a 2 (for a 2 -a -1=0).20 + .3 n + B(-l) n .l) n ~ l ]x n ~ l . 3. /. 4 l-3# 4 1+s 1 / v\n 1 ~iin Now lim iZL-^^ limLJL == . this may be written . .. The solution is u n Ca n + />^3 n where a. sum to n terms = .0K = (2 . .-. and let the scale be Then 7+2p+# = and 20 + 7^4-2^=0.. a/3=-l and if a>/8.h _ -.".(2 . are arbitrary constants.Eliminating A. and therefore u n =A. . Find the n-th term of a recurring series of which the first four terms ar e Find also the sum of the first n terms when x ~ 1 . j8 are the roots of x*-x + \ and (7. Therefore I -d - GENERAL TERMS 365 Ex. . we find that (a . 3 l-3 w z n 1 !-(-*)" . J'iwrf the general solution of u n .1. and similarly 2-a=/? 2 . where A+B^\.)3)a" -' . Changing the constants. and the scale is u n -2u n ^ l -3u n _ z -Q. Thus the nth term-w^a. Hence A-% B=^\.

i x 4Utfe n^ < xC/ O*< ! [The moduli of the roots of x 3 -2# 2 + 4x-3=0 are 1. 6. j3 n are cos t sin -^~. the sum to infinity is ** ( ~\ _1_ 9 1*2^ //I _ f?l* I 4. Show that. and that the sum to infinity is 7J-.77 Hence the values of a n . ft are 1 ^3 7T . 1/^/3..!. If u n is the nth term of the recurring series 14-24-34-84-134.] 366 EXAMPLES OF RECURRING SERIES 4. .. show that u n ={(3n-4)( -l) n -f 2 n + 2 }.. . Find the wth term and the sum to n terms of the recurring series : (i) 1 + 2 + 5 + 14 + . (ii) 1+2 + 5 + 12 + . . Show that the recurring series is convergent. . If 3u n -lu n _ l -{-5u n _ 2 -~u n _ 3 =0 and % = !. u 2 = 8. and so o o . and find the sum to infinity of the 8eries u ..30 4-.. ... l/>/3. i ..^f*^ i ^ ~^~ fjjL. A r and the solution is u n A cos + B am ~ u u EXERCISE XXXVII 1. . 3. Find the nth term of the recurring series l-f2-f34-54-74-94-.'/*^ . and if the first three terras are 1 + 2x + 3# 2 ..The values of a. u 3 = 17. if u n -2u n _ 1 + 4u n _ 2 -3u n _ 3 = Q 9 the series 2u n x n is convergent if | x | < . 2. find the values of u n and u l 5.

If 2u n x n .a) 4. = ^r u n x n . then Uu n 2 x n is a recurring series whose scale is [1 . If Eu n x n is a recurring series whose scale is 1+px + qx 2 . where ^4.] 10. Prove that each of the series COS a 4.affix) ( I ~ OL'fi'x). J 8.2^0^-2 -pp'OU'lln-z + <f<l' 2 Un__i =0.1)0 -u^ 2 sin (n.6V n " + Dai' n p /n . [If a. C.A: 2 x 2 . C are constants.OLOL'X) ( 1 .^?a n /3 /n 4.. and -^7r<6<^rr 9 show that n ~ 2 ?L:J \/(4</-7> 2 ) u n ain6~u 2 q 2 sin (n .2>'z 4.] 12. If ZW n . where A.q'u n __ 2 = 0..2)0. Hence the scale of. sin oc + x sin (04-a)+# 2 sin (204-a)4-.X COS (0 4.qu n _i ^ and v n + p'u n ^ 4. 11 to show that l+px + qx 2 n==0 n 13. then Eu n r n x n is a recurring series whose scale is Hence show that Unu n x n is a recurring series whose scale is (1 11. [If a.q'x*)9. then Un r n ^Aoc n ' n 4.a) 4. then 2(u n + v n )x n is a recurring series whose scale is ( 1 -\-px 4.g 2 # 2 ] ( 1 . ^8 and a'. . B. Zu n v n x n is ( 1 .. then Su n v n is a recurring series whose scal e is u n ~PP' u n-i + (PV +P' 2 <1 .2kx 4.. If Zu n x n ..(p 2 ~ 2q)x 4. where ^ 2 <4g. use Ex.. I) are constants.7. Hence the scale of Eu n 2 x n is (l-a 2 ^)(l-^)(l-^). then u n * = A(ot*) n + B(fi*) n + Cq n . B. .. j8 7 are the roots of # 2 4-^r + </~0 and # 2 -h^>'^4-g r/ ~0 respectively. where ^> 2 ^ 4<^ and >' 2 7^ 4^'. fi are the roots of z 2 4-pz4-?=-0.X 2 COS (20 4. Er n x n are recurring series of which the scales are 1+px + qx 2 and 1 +p'x + q'x 2 . where tan#= .*P'X) ( 1 .qx).flo: 2 ) ( 1 4. If w n -^ w _ 1 + gw n _ 2 =:0. ZV n are recurring series of which the scales are u n + pu n _^ 4. Ev n x n are recurring series whose scales are l+px + qx 2 and 1 . . If .

Id) _____.x n cos (a -f nO) + x n + l cos (a + n . the scale being 1 . FORMATION OF DIFFERENCE EQUATIONS 367 6. Here we give methods which apply to various cases which occur in ordinary algebra. 5 it is clear that this solution must involve one or more arbitrary constants. 2. (1) // u n = au n _ l9 then u n = Ca n where C is an arbitrary constant. The required equation is found by eliminating C' from u n ~Cn-"l and ^n-\~ V(n . Solution of some Elementary Types.2x cos 4. Finite Difference Equations.1) -2. 1. To solve a difference equation is to express u n as a function of n in the most general form. B from 7. Ex. A particular solution is obtained by giving these constants special values. prove that u n .. and from what has been said in Art. .. obtain the corresponding difference equation. u n __ v u n _%. An important class of these equations has been considered in Art._____ 9 and that the sum to n terms of the second series may be obtained from this expression by changing cos to sin in the numerator. Show also that the sum to n terms of the first series is cos a . giving (n . The general theory belongs to the Calculus of Finite Differences. The result is found by eliminating A . // u n ~AoL n +Bf$ n where A. a difference equation may be obtained as in the following examples.l)w n -nw n.x cos (a .is a recurring series.. ft are giv en unequal numbers neither of which is zero. Let (u n ) be a sequence in which some or all of the terms u n . where r is a fixed number. Ex. The result is called the general solution.x 2 in each case. Such an equation is called a Finite Difference Equation. . 1 = 2. 5. B are arbitrary constants and a. If u n is a given function of n containing one or more arbitrary constants. // u n ~-Gn-2 where C is an arbitrary constant. u n ^ r are connected by an equation which holds for all values of n^r.0) .(* + $)u n _i +aj8t* fl _ 8 -=0.

(2) // u n = a n u n-i where a n is a given function of n> then w n = Ca x a 2 . Ex. n .a n . and u v may have any value. and divide each side of the given equation by v n ..2 2. it is better to apply the method just described rather than to use the actual result.3 (n-l)n) . 368 SOLUTION OF ELEMENTARY TYPES (3) // u n .1) =2/ (n ... Dividing by n(n-l).. 2 in succession for w. The result is Un/Vn~Un-lK-I = b n/ v n' In this. .1).For w n /w n _ 1 ==M n _. 1 /w w .1.nu n _ 1 ~2. 1. n ~ 2.. Let v n = a 1 a 2 . the general solution is where C is an arbitrary constant. --a n a n -i <*2i u n-\ u n-Z u l therefore u n = a^ .. Writing n ..2. a ny a l and Wi/ a i mav have any value we choose. wK . For Un * Un ~ l ^-aa a *or ~ .u n ^l(n . we have Whence by addition._ 2 ==.. . a n where C is an arbitrary constant. In examples of this type.a n u n _i = b n where a n and b n are given functions of n.^1/^1 = &iM + 6 2 /v 2 + + & nK . Solve the equation (n -l)u n . .1. -n I U.. =w 2 /w 1 = a.fe iKThis gives the revsult in question. 2 in succession for n and adding. .. for the value of u l lv l -b l jv l is arbitrary.. .. writing n . we have u n \n .

nu n _ { = I M. so als o is C. we have Writing n-1. If a = ]8.\ [a a 2 a n j where B( = u l /ot-cft/oi) is an arbitrary constant. if a . 1. a LINEAR DIFFERENCE EQUATIONS 369 (B B 2 # n l Therefore u n = coc n { . gn _ a n r Hence. and proceeding as before. the equation becomes u n j\n--u n _ l l\n-\ l. 2.. For dividing by a n . 2 in succession for n and adding. Find the general solution of u n . 6. n n /3-a which.) Ex. (4) The general solution of ?/ n ~aw n _ 1 = cjS n is ^a n + c n +Y03--a) or according as a and j3 are unequal or equal. then i/ n = cna n - (5) The general solution of w n ~(a-f j3)w n _ 1 -faj3w n _ 2 = i u n ^AoL n + Bft n or u n ^(A + nB)oc. n . . Ex. n-2. (See Art. . we find that w n ~[n(n + (7) where C is an arbitrary constant.Hence u n ~ Cn-2 where C~ 2 + u v and since u t is an arbitrary constant.. .may be written where A(^B-cft/(ft~oc)) is an arbitrary constant. + r .. Dividing by | n.+ -^ + . .

.. (iii) The method of dealing with such equations as au n -f bu n ^ -f ... then v n + w n is the genera l solution of (A). 6. For it is obviously a solution.O ^ A / Next search for a particular solution of Assume as a trial solution w n =a + fru + en 2 . (ii) // v n is (he general solution of au n -f 6w n _ x -f cw n _ 2 + . k. fc. (See also Art.. by (4). (1) An equation of the form au n + bu n _ l + cu n __<>+ . then for all values of n. we find that . for u r+l9 w r4 .. .ku n . a + bn+cn z -5{a + b(n-l)+c(n-l) 2 } + 6{a Equating coefficients of powers of n. Linear Difference Equations. k... u n ~ocu n ^^C^ t where C is an arbitrary constant. 4. w n ==^a w + Cj8 n+1 /()8-a) or (4+Cw)a n . .. r = 0. where a.) 8. Here r is a fixed number.according as a and /? are unequal or equal.. and w n is any particular solution of equation (A).. . 370 GENERAL TERM OF A SEQUENCE Ex.. it should be noticed that (i) The general solution involves r arbitrary constants. r are constants and l n is a given function of n. r = l . 2 >-can be found in succession in terms of u l9 u^ .. (A) is called a linear difference equation. and a. -f ku n _ r = /. 5..... This establishes the result in question. B are arbitrary constants. for C ft/ (ft -of) is an arbitrary constant if a^/3. 1. The general solution of u n -5%^-f 6w n _ 2 =0 is *^tl A +/J. Solve the equation u n -Su^j-f 6w n _ 2 =n 2 -f 5 n + 2 n . + ku n . also it is the general solution. With regard to equation (A). All the equations in Art.. For the equation may be written ^-atV-i^K^-a^o).. and therefore. .. .. 7 are of this type. because v n involves r arbitrary constants...... where A.. is illustrated in the next example. according as a 7^)8 or a = /J. I are given functions of n or constants.. u r) which may have any values.. Hence by (1).

4.... Thus a solution is tS. Trying u n an......... take the equation Since 2 is a root of x 2 -5x + 6=0... .....i) n ~ 2 }.... 2... The first term of a sequence is 1..2c^l..2.. .....2 n + l .. 26-14c=0. then a(5 n -5. If 2tt n -tt n _ 1 =na.. we have a{n . show that 3....... it is clear that a .... Show that the nth term is a x where x ^i :{2 + ( ....l)2 n ... 2 W is not a solution. 2 n .. giving a = . Finally. (D) The general solution of the given equation is obtained by adding the right-hand sides of equations (A)-(D).. a -^ ~ n d so a solution is 15) ....5 n " 1 +6.. and every other term is the arithmetic mean of the two preceding terms. Show that the nth terra is thus proving that the nth term tends to J (a + 26) as a limit as n -> oo . the second is 6. and every other term is the sum of all the preceding terms. 6 J..... Thus a solution is u n = -n. take the equation and assume as a trial solution u n a . (B) Again. 2 n -5(w .. Find the nth term.. 2a giving c -J-.. and is EXERCISE XXXVIII 1... 5 n .. The first term of a sequence is 1....... giving a = ? 6 ^. and every other term is the geometric mean of the two preceding terms.......x + 6(n -2)2 n ~ 2 } =2 n .. the second is a.5 n .a )=5 n .. The first term of a sequence is a....

(a 4 ft + y) U n _i 4 (fty 4. Obtain a particular solution of by assuming u n =an + b.[u^. Deduce the general solution of the equation as given in Art.] 12.gw n _ a and u n tends to a limit other than zero as w-> oo . (ii). K/|n_= Wn _ 1 /ln_-J-K-D n /[n.l)(^ n ~i + ^n-2) an( i l = 0.l n as a trial solution.(n 10. and show that [Write the equation t* n . 11. and the roots of x z -px-q are 1 and -~q. show that |g|<l. . Find the general solution of u n -5u n _ l -i-Qu n _ 2 l n . where J5.nu n ^ = .(^n-i ~ nu n-2)-] 9.Ui + qui that hm u n .1) u n _ l ~ . . If u n = ?i(u n _ l + u n _ 2 ) y find u n in terms of w and ^. If w w = />_! -f.. 7. Find the general solution of w n -w W -i-Wn_2=ft 2 . C are arbitrary constants. 5.DIFFERENCE EQUATIONS 371 5. M 2 = l prove that V 1 .(w 4. n 1+q 6. [Assume u n ~a . ( " 1)n 1 j 3 li " L? /' [Write the equation in the form u n .! . Hence find the general solution of the equation. also show ^i x i. show that u n j\n-*e~* as ^->oo. If u n nu n _ l + (-l) n and w ~l.] 8. Show that the equation u n .ya 4 aj8) tt n _2 ~ can be written in the form Hence prove that Un-oMn-i^Bp + Cy" OT according as fi^y or j9 y. If w n (^ .

.5it w 4. Show that the general solution of the equation Vn-l + au n + bu n-l + C = can be written in the form where A is an arbitrary constant and a. /3 are the roots of a: 2 . no ^ e that there are two forms.l)?/n_ 2 . Hence show that the general solution is ~~ 2(1 -f cos ) 14. The equation then reduces to v n-l v n 4.] 16.9 0..(c .a)v n _ 1 4.} 372 GENERAL SOLUTIONS 15. Put v so that u n 4 a . If u n u n _i =.c . prove that >l.ab)v n _ 2 0..2 n 4-5 n ^ _.WTJ_I 4. in which case [Write the equation as / ^(^ n _i4-a)4-&w n _ 1 4-c^0. < /I 2 w "" 1 4-5 n ~ 1 17. then 2 .ab =0 .(6 . = fi. Find a particular solution of u n + u n-i sinna by assuming u n a cos no. unless a. find u n in terms of u^ [nu n = (n .[Take as a trial solution u n ~ an 2 + bn + c. If Vn-i + 3w w . If < w n w w _ 1 4.(a .4 w _ 1 -2=0.b)x 4. -f b sin noc.] 13. unless v n _j = 0.

(u n + w n ) = ( W n-fl ~ *n) + (V w+ i ..+Au n unless w 1 = 0. For A (u n + v n ) = (w w+1 + v w+1 ) . . If tt n+1 =-~ and w l = a. E. In this connection it should be noted that if then As n is not equal to Au v +Au% + . The Operator A.+Vn^Wn^-ti!. -' 18. and A is regarded as denoting the operation which when applied to u n produces the difference u n+l . u 3 . etc. Let u n be a function of the positive integral variable n. thus 374 TABLE OF DIFFERENCES This process may be continued to any extent and. the symbol J 2 denotes that the operation A is to be repeated. and so on . and consider the sequence u v u 2 .. from the way in which . so that Jw 1 = w 2 -^^ 1 . INTERPOLATION 1. then d(u n + v n )=Au n + dv n . For v n = w w+1 . D. *?_! = w n .-. Au 2 = u 3 -u 2 . for and (iv) Following the index notation of Algebra.V n ) =A<U n + Av n . Here n may have any value. (ii)// Au n = v n .!*__!. (i) The meaning of the symbol A is defined by ^n^tVrt-f^. prove that 2n-w n a4-w(l-a) and [Put w n ^nv w .... .... (iii) // u w v n are functions of n.tij. . u n .ti w . *Aen Vj + Va + Va + .] CHAPTER XXIII THE OPERATORS J... whence the result follows by addition. Vj = w a . .u n . and we write .

for in finding the successive differences of u x we are merely concerned with the sequence II I/ 07 u x> x+V u x+2> Illustrations.. Au 1 3 5 7 A*u 222 A*u (v) If 'u x is a function of any variable x.... a table of differences is conveniently written thus: UQ Ul u 2 W3 . 2 2 .... All that has been said about the symbol A still holds..... ... Special Cases.. J 3 w n . (B) .............^ . Hence we conclude that 4 r u n = u n+r -Clu n+r _ 1 + C' 2 u n+r _.. 3 2 ....... it is obvious that the numerical coefficients in the expressions for J 2 w n .. 2.... are the same as thos e in the expansions of (1 -x) 2 ...... 2 -.. If x is variable and J applies to x.ra( f +D..... . we write x r =x(x-l)(x-2). (i) If u n = n(n~ l)(n-2). since w 3 = n(n~l)(w-2)-f 3w(w-l)-hw.. by the preceding.....+(-I)nu n . For the sequence u Q . (1 -x) 3 ....... then and J r w n -0 for r>3.. for the sequence O 2 ...... .. Thus..(a.r x(x + l)(x-f 2) .. (A) a result which is proved more concisely in Art.. .. (A) THE OPERATOR & 375 Similarly we can show that Ax^ r = .. we write u 1 4 9 16... (ii) If w n = n 3 . 4....the successive differences are formed. (i) The following notation is often used .... then that is Jx r = rx r _ 1 .... .. ..-r-f 1) and 3_ r = . where u n n 2 . u l9 w 2 . I 2 . the meaning of the'operation A applied to u x is defined by Au x ^u x ^-u x .^ ...

1).. we find that (7 = 0.it being assumed that no denominator vanisJies. 1. 3.. S=^ (n + 1) 3 + vhere C is a constant. a? 1.. a m . If a is a constant. Shaw that n(-l) + (n-l)(n-2) + . (ii) A Polynomial. is to increase n by unity. a + E. regarded as operators. the following equations define the meaning of E + a. a^ . we write and so on . might have been applied. For A m u x \ m. Putting n =2.. VIII. Ea and a$. and then J w^ = a l + 2a 2 x x + 3a 3 # 2 -f . | r and J r ^w x = 0. Thus for all values of n. (ii). The Operator E.1 =(/t + l)?i(n . thus. Of course. Eu n ^u n+l . + a r x r . a? -2. Further. Hence the result. 2. and so on. w(7i-l)(/i-2) + (n-l)(n-2)(n-3) + .. . if r is a positive integer. Similarly for the second se ries. . .a m + terms containing x as a factor. If u x is a polynomial in x of degree r. If 8 is the sum of the first series. .. dividing by x. and since Jn 3 = 3w a > by Art.x~ r = . in succession. . a r .. l=(n + l)n(w and so on. 3.x r = rx r ~ l and -_. The effect of the operation denoted by E y when applied to w n . (E + a) u n = w n ^! + a?/ n = (a + J0) w n . Finally A r u x = a r . +3 . 2 . // w<r. . independent of #.+2.. such that u x = a o + a i^i + a 2 x 2 + . 1. ^en J^WQ ~ \m . ?x. These formulae may be compared with -r-. the rule in Ch.rx~ (r + 1 \ ax ax Ex. we can find a .

Again. the operator E combines with itself and with constants according to the laws of Algebra.-..(ii)). if p.. then (i)^u x ^u x+r -C[u x+r ^^C r 2 u x+r ^-.... the operation E is defined by 4..376 LAWS OF OPERATIONS Hence.. 6 are constants. then For the left-hand side =4^ = (Art. ...CJ^. Fundamental Theorems. More generally. if u x is a function of any variable x. so far as addition (subtraction) and multiplication are concerned..... then n(n-l) A n(n-l)(n-2) . ...+(..-(-l)X = {E r ... 1. on writing x for n. // u x is a function of the variable x 9 and r is a positive integer. (A) (ii) u x+r = u x + C[Au x + ClA 2 u x +.. The first theorem follows from the reasoning in Art.1 + Cj'.. q are positive integers...+A r u x . . and therefore (iii) If u x is a polynomial in x of degree r l... (E + a)(E + b)u n = (E + a) (u n+l + bu n ) + abu n Thus. (B) Proofs. The same is true for A. (iv) // s n is the sum to n terms of the series u -f u 2 -f u 3 -h ... (ii) u x+r =E r u x = (l +A) r u x .... E*E*u n = E* (Eu n ) = Eu n ^ = so that E*E*u n = E*>+<Ju n = E*E*u n . (iv).l) r }u x Again.. 2....+(-lYu X) .. More easily thus : (i) J'. for 4* n = ti w+1 -w n = (^-l)ti n ... if a.

Ex. . Denoting the function by u n . the result follows at once. Find a cubic function of n which has the values 1. 4 respectively.1. Ex.' For w = Similarly METHOD OF DIFFERENCES l+^) n ~X> therefore 377 Adding and using the results of Art. 13 when n = \ . 1. 1. we have n I 2 3 4 u n 1 -3 . 3. -3. 2. 2.

.3+3.+( Here u n ~ (n + l)(^ 2 +2).6 + 4.r>3. that ^=6. Writing down the first four terms. ll + . . since u n is a cubic function._ +6 j&a. 2. u n as on the right.. J 3 u 1 = 6.-1 13 Au n 4 2 14 A*u n 6 12 6 and A r u n = for. so that A r u n for n>3. Sum the series 2.. Also tt = ^ w -X = . we find. 4^ = 1 A^u^U. This is a cubic function of ?i.

. 4. Au n A 2 u 6 18 44 90 i9 12 - 1A 14 12 26 46 14 20 6 .n(n-l)(n-2)(n-3) . Ex. // 5 = * n -C7J(a: + l positive integers. then 8 if r>n and S = (-l) n \n if r = n. (Cf. Ch. 3. are Ex.Hence by (iv). VIII. 1.)* r.

. S=^(l . . VIII. (*~r + l *'~\ [2_ [3 ' LT+JL Now... Operators in a Fractional Form. B r coefficient of n in S r . {AZ A3 Ar^ J "T + ~. VIII.. 4 to show that B^ ^.... (iv) : E n -l n(n-l) . hence Again.... Hence equation (A) holds for r>0. + (-ir i }r . 5.. we have f(n + l)n (n + l)n(*-l) (n-f l)n(n. (A) For n r =tf n O r =(l+J) n O r and <d0 r =0 if m>r.. 2. (5). Apply equation (A) of Ex..E) r x n ^(-l) r A r x n f and the result follows by Art..* (Bernoulli's Numbers.. 4..... by definition.For if A applies to x... Many writers give the following ' proof ' of the theorem in Art.+( ~ 1)f " r} r= " Hence if r>l. Li II. 4... equating the coefficients of w in (B)..) If B r is defined as in Ch. (5)). +n r .. ^ = i and (-l) r J5 r =J5 r for r>l (Ch. then -l+$-. (ii). 8.. n(n-l)(n-2) . 5. Hence if # r = l f + 2 r + ... J0 r =l. 8.. Ex. 378 BERNOULLI'S NUMBERS Ex.1) .. Moreover...

.. Lagrange's Formula. when obtained. y$ for x=a. IX.. If only two values of y are known. are to be determined by making the curve go through the n points.. c. 6. 7. where f (A) and <f>(d) are polynomials in J. provided that f(A) is divisible by <(J). Thus the equation E n -l Sn ~~E^T Ul does not imply that (E . this fact justifies the process. If y=y l9 y 2 . It is natural to take for f(x) the simplest function of x which satisfies the conditions. 259-262 of his collected works). In fact the last statement is not true. can be verified by multiplication and addition . c. in such transformations. This result was obtained by Cayley by a much more difficult process (Vol. for x = a. Interpolation. and draw a * smooth ' curve through the points. . or otherwise. We conclude that.i Here the operator is written in a fractional form in order to transform a polynomial in E into a polynomial in J. andf(A). 6. Suppose that y is a function of x whose value has been determined by experiment. we assume that where the n constants p . . LAGRANGE'S INTERPOLATION FORMULA 379 The result. pp . . <f>(d) are not regardedas separate operators . the curve is the straight line joining the corresponding points.1) s n = (E n . the operator can be put in the form /(J)/^(J). and we have the rule of proportional parts as used in working with tables of logarithms and trigonometric functions. Let us suppose that the equation to the curve so drawn is y=f(x).1) %. The graphical method is to plot the points (x. for (E-l)s n = s n+1 -s n = u n+l and (E n -l)u l ^u n ^ l -u l . But at every stage the operator must be regarded as a whole.. c respect- . 6.. We assume that this curve represents as nearly as possible the graph of the function y.. The problem of interpolation is to find approximate values of y for intermediate values of x. 6. If y is known for n values of x. y) for x = a. p^ .

y%. For example.. instead of applying Lagrange's formulae. and if n is a positive integer. Construct a table of differences as below.O. sin 50 =0-7660. y 2 > ^3 ^4 when x = a. y y *(a-b)(a-c)(a-~d) with similar formulae when more than four values of y are given.ively.y \ ._L^ . y n ^f! n y = (l+A) n y 0) therefore Hence the equation represents a cubic curve which passes through the four points (# . . u Au A 2 u A*u 589 Mj-7660 -57 532 7 " -64 468 .. 1. in 55 =0-8192. ?/ ) (x l9 T/J). suppose that y y& y\. n62. 2 B B. c.' . and gives an approximate value of y corresponding to any value of x within the specified range. Let 1*3 = 10 4 sin (45+5x). 380 INTERPOLATION 8 . j/ 3 for X X Q . Similarly if y = j/ 1 . b. .A. c. Ex. then J r y = for r^4.. x v x 2 . 6. we assume that (x b)(x c) (x -a)(x c) (x a)(x b) y=syi (a-b)(a-c) +y * (b-a)(b-c)* y3 (c-a)(c-6) * For y is a quadratic function of x of which the values are y l9 y& y 3 when x = a. d we take (x-b)(x-c)(x-d) . If the values of y are known for values of x at equal intervals. y s.' + three similar terms . x$. Assume that y is a cubic function of x . Given tJiat sin 45 =0-7071. . we use the operators E and A. A-rn 60 =0-8660.

0-7 x 0-4 x 57 = . w 5 = 7-1154. // w = 4-3315. J 3 w -= . Four points on the graph of u x are known. If UQ and several other terms of the sequence U Q . = 0. w 3 = 5-6713.7. This is the same as assuming that J 4 w a . u v u%. . sin 52 =: 0-7880 to four places. 2. Ex. .57. ^ = 4-7046. The correct seven-place value is 0-7880108. the rule of proportional parts w ould give sin 52 =0-7660+f x 0-0532 =0-7873. we may proceed as follows. J 2 w = ./..15-96 | 2 ILl-JpZJ A 3 u Q ^ 1 -4 x 0-4 x 0-7 = 0-392 I 7895-992 15-96 u x = 7880-032 . Hence we have approximately 0. BESSEL'S INTERPOLATION FORMULA 381 If wo only use the values of sin 50 and sin 55. Assuming that J r w for r>3. find approximate values of u and w 4 . J^ = 589. so that x = 1-4 and u =7071 D = 1-4x589 = 824-6 MA = . approximate values of w^are given by x(x-l) A9 x(x-l)(x-2) -L_J 2 M .._j 52=45 + 5z. are known and approximate values of the missing terms are required.'. We therefore assume that it can be represented by a cubic function of x.

. the effect of second differences may be easily and accurately allowed for by using a formula due to Bessel. c can be taken from the tables. It should be noticed that the approximations just found are much closer than tho se given by ) =5-1879... if we neglect differences of the fourth and higher orders. (x + l)x(x-l). we have Now w -w_. Let u_ l9 u . . In using Mathematical Tables.. w 3 tan 80... as shown b f on the right. u lt u be four consecutive values of a function u XJ as given in the tables. 9... b-a=^d. w 2 ^5-1420.....+ ^i) ^11-4619. w 4 = \ (u 3 + u 5 )= 6-3933.. 382 THE DIFFERENTIAL OPERATOR To show that this gives a good approximation. c and d 9 e... 6.. and suppose that values of u x are required for values of x between and 1. (B) The values of u Q9 a. and therefore x(x-l).).. e-d^f. Bessel's Formula. NOTE.. W 5 ~tan82. w 1 -tan78.... so that very little labour is required in using the formula. The correct values are 5-1446 and 6-3138. In this example w ~tan77.i=a. u -i Denoting first and second differ.a ... so we have found that tan 79 ^5-1420. Bessel's formula is u * e \ c x(x-l) d + e which is the same as a) .. and to estimate the error. tan 81 ^6-3199 (approx.. ences by a 9 6. w 4 = 6-3199 (approx..).. /.

. (D + a) (D + 6) u = (Z) -I. 6.... we have z^-4..) For >/12344.. we take 045020= a -: 11 1-085553 045001=6 M 1 = N /12350 = . ^2-^/12360 = . 10....' V x(x~l) fd + e e-d\ (x + l)x(x.....a) ... When we write du d d n u/ d \ n the symbol -r.is regarded as an operator.1) so that if ?' x is the approximation given by (A).a) (Dw -f 6w).l)(2x~ I)/.. a + A Da and aD. b are constants. the following equations define the meaning of D + a. Ifo... c-a-^ .-000037. c are given in the tables. -J*(3?-l)=--06. 1. and (/) 4. 044983 -c........ C/5^ to6/65 of square roots to find Using Barlow's tables... The Operator D.-00000224 N /12344=!ll-1035o7> ~ The result is correct to the last figure.a) (D -f 6) u = Z) 2 w -f 6Z)w -f aDw -f afrw = {D 2 + (a + b)D + ab} u. (The differences a..1 ) (c . LEIBNIZ' THEOREM 383 Hence if a. Hence the error is approximately izx(x. which is sometimes denoted by 118 If a is a constant. . regarded as operators.. w = 111-085553 6= -0180004 Ja?(a? .

Differentiating w times by Leibniz' theorem. ujm* . ax ax ax As D t and D 2 are independent of each other.^) . -j (uv) = ^ This result is known as Leibniz' Theorem. the operator D combines with itself and with constants according to the laws of Algebra.k) ". = ' /^ W = 0. so far as addition. . . 1. D l (uv) = v. Differentiating with regard to x.arM! . d n n . 11. we have ww tn-i Wi _ i + x ( x z _ jt)"^ = tt w (a. .w/m 2 0. D 2 (uv) = u j.k) y 2 -f ar^ .Thus. 2 . The n-th Derivative of a Product. whence by addition.k) w a -f. Similarly (# 2 . and is often written in the form = . and that D 2 operates only on v and its differential coefficients. (x 2 . -f uv n . . If ij~u + v where u m =x + (x z . where the suffixes indicate differentiation with regard to x. therefore / ^ \ n .'. and hence ~= (uv) = (D l -f D 2 ) uv.kf* . . (x 2 . Ex. 1.. If w. so that n / \ & u r\ i \ dv .. ...( 2 . subtraction and multiplication are concerned.\dz/^ WV '"~' 1+ *' ( uv >~\ 1+11 2 21 2 2)( u )> that is. 2. and suppose that D only operates on u and its differential coefficients. v m = x . hence (# a . . v are functions of x d du dv dx dx dx' Let the symbols D l9 D 2 denote differentiation with regard to x. they obey the laws of addition and multiplication. Differentiating again.~* ! = (x* . * *.v/m? 0.k) v^-^-xv l .Jt)"" .k)^u l =u/m .y/m 2 =0.

Use the method of Art. according as n I. 2 to find the sum to n terms of the series (i) P + 3 3 + 5 3 + . 2 2 + 2 2 . S-=l-{(2 + A) n -A n }u 9 .. .) 2. 2 or 0. 6=A-(r-l)(r-2)(3r-5)... . 71. then S = ( . then S = i-{(^+ l) w ~(E . /.^2 n ~ 3 n(n + l). 2.k) y n +* + n. if n>2 [In each case let S be the sum of the series. 3. If x r = x r -f o^^! -}.1 ) n J w . 4 respectively. Also J r? ?/ r ( ~y) n | n. Show that (i) x n -C'i(x-y) n /-l. show that (x -h n) r = a.. If u n = x n and A apph'es to .(* 2 .. 3. If x r ^x(x-l)(x~2). . show that 4.] 7. 4 2 + . (ii) I 2 . XXIX . (ii)Lct w r = (a:-r)(y-r)..l)x r _ n if r>n and C^jr if Verify that. prove that a^l-r(r-l). Also Aut = (x-l)(y-\)-xy=-x-y+\ 9 J 8 M = 2. .I) n )w . given in Ch. 3 2 + 3 a . the last term being r(r.(xr + l). 4. . ^ 2 w = 2. J w w r = if n>2 (iii) Let w r = r 2 .. t/ w+1 + ^-i 7 - This example leads to a remarkable theorem on cubic equations.. (iii) (w-l) 2 C?4-(n-3) 2 r^ + (w-5) 2 Ct4-. Find the polynomial of the third degree in n which has the values 2.2x. [Equate to the coefficients of x r ~ 1 and x r " 2 on the right.E) n ii = ( . when w = l. if n 3 and r 2. the identity becomes 6.. and JX = 0if 5. (r ~n-\. . 384 USK OF OPERATOR A EXERCISE XXXIX 1.6a: r _2 -f . then S = ( 1 . 11.1 )a: r _ 2 -h . (Use Art. Jw = l. etc. n~2 or ?z->2. 4. ..n/) r '. . Show that (i) JV+^fc+JrJ'rj-JL .t. .1) . Ex. Also w = 0.r + ( ?rar r _ l + CJr (r .l) n A n u^. 32. (i) Let u r = (* .

[Use Ex. Prove that x n -Ci(x+l) n + Cl(x + 2) n -. and = ( .fC? ^-"'+ ( .C? ^.1 EXAMPLES OF IDENTITIES 385 9. prove that 1 = __ (-ir\n x Deduce that 4.. | rjh 2.- 2 (x+l)(x + 2) n(n-l)(n-2) I 3 10.(ii) J r x r + 2 = J [x 2 -f 7-a: 4.] 8. 6.1)n 1 1 1 + - 1 n(n-l) n .T m-f-1 1.2. [Use Ex..2 m + 2 . 7. Prove that 1 2 yt(n-l) 2 2 2 W * ~r 71 r. If none of the denominators vanish.^r (3r -f 1)] .1 Yl {& + r-r + iV(3r -f 1)} | r+_2 when n = r 4.+(-\Y(x\-r) n + \ when n~r-\ 1.

u.l) r r (an + b)(a . / i\n. [These results follow from Ex. If a:_ r = l/a.. 386 USE OF THE OPERATOR J 14._(r +8 )-. + 2). If n is a positive integer and y is any number such that no denominator vanishes.x (in) each series being continued to n 4. to n + 1 terms.... x= -(t/4 1) . m(w-f-l) m(m + l)(m + 2) ' m(m-f 1) .^3 n n. If u n ~\l(an + b). 12 by putting successively r = l.3 n ~ l nt/^x/..( r+1) +(7?r(r+l)a. ^n(n-l) n(n-l)(n-2) /.(a: + l)(a.^-.. prove that n n ( n ~l) n(n-l)(n-2) _ y + l l + n..1 terms.n 111 1 n n(n-l) a n(n-l)(n~2) a 2 ' ' " ' "3 12. show that ( + n)_ T = a?_ r -6 I ?ra . 13.2. Show tfcat ln + 1 . show that (i) A r u ~( .(a: + r-l). (m-fn) 11.

l)(n-2) a(a.1) ... prove that b 6(6 1) _ a-6 a-6-fl a-6-f2 a-6-fr-l ^ r Wn= ^ T 6^1 6^2~ 6-r-fl * [Putting n = l in Ex...M i o\ * // i IW/-M i o\/ M i o\ ' : + 2). find Au 0t A*u . 15. _ --. If ^^ ___ > prove that __a~b a(a... etc.to n + 1 terms= ' -.. (a-n-f 1) U== ~~ ' a-6 a-6 + 1 a(a.1) .. 1 ln + 2 -..1) .V / .... 15. as in Ex. Prove that f I _? ) M a )(1 a j . i I\/... (a-n-f 1) a-b a~6-fl a-6-fr-l a(a. If w 1 tti=7 ^2=7-77 T...l)(a-2) "* n 6 + JT~ 6(6-1) [3 6(6-l)(6-2)" to n-fl terms. .*//*.. 15.. (1 ?L \ 6/\ 6 1/\ 6-2/"\ 6 Ti__ a n(n-l)a(a-l) n(n . (a-n-f 1) ~ w 66-16 16. .. in succession from the equation s 17.> e tc.

16. show that ^12516 = 111-874929. . Given that ^12500 = 1 1 1-803399.[Using Ex.~+ ".] 1 r INTERPOLATIONS 387 19.to n ~ 1 term8 [2 [Equate the coefficients of a* in Ex.] 18. 4. ^12510 = 111-848111. sec 89 7' = 64-8657. Show that the sum of the products two together of 1 JL JL i x 9 x + l' x + 2 is equal to n(n-l) ( n-2 I (n-2)(n-3) 1 "-. In Art. sec 89 5' = 62-5072. sec 89 6' =63-6646. 21. the left-hand side = ( . Given that sec 89 4' = 61-3911. 4 can be written and ( 20. 4.l) n J n w =(l -E) n u . 17 and put #= -6. it has been shown that Prove also that and use each of these formulae to show that B^ = [Equation (D) of Ex. ^12530= 111-937483. Ex. ^12520 = 111-892806.

] on Qn 23. 6. 7.to <. 9. = 5. [(iii) and (iv)... r n . 5...l) n S .e=0.3 r WJ^^ +--+_ + .-+ I + to oo .... 15. 2. If r n = l n -"r? + r5-. prove that L_ i (u) . we have S W =(J + l) n and S n ^ = (E . 203.. A 2 fl .. 877.. 2.+ ('" I ) m i +... 3... ^ n /e=2. Any expression of the form . prove that 1 ^ 971 .! (iii) Using (ii).. 52. -1. If S n = l n + * L i +rz+... i-: .. = F-i + + + ... S 19 showing that for n = l. IJL I d or~i or-i (ii) J"5. 4140. show that for n=l. 5.. Definitions. 2.. ..T B = Tn.2 r 4 n ..show that Bessel's formula gives sec 89 5' 40" = 63-2741. -1. on on / w i i \n 22.... . CHAPTER XXIV CONTINUED FRACTIONS (1) 1 . _.. 4.....l f 3 n ... 4... 5..... LL L_ (iu) S n = -S n _ 1 (iv) S B = 5 J ^ 1 (v) Use (iii) or (iv) to find S 19 S Z9 . 3.

.. . ... are all positive. a : H ---...... are called the elements of jF.=r 1 1 J a 2 + a 3 + a n + a 2 a 2 We shall write w=^) 1 / 1 . J a. JLJLA * * 1 . a v + " .. 6 2 60 b n . \ V I 1 where the a's and 6's may be positive or negative numbers. _.. 1 ... . if each fraction contains n elements.. ......-.x ~ ^ ^ 2+4+4+44... w=^22 where .Ui ~ " ~ ".. flu 6<> 6<>6q second.. b n+l (u = 2. ... 1 #2 2^~ ^3 It is obvious that the fraction is unaltered if 6 n .... a n .. . ) are all multiplied by any number k or if the signs of 6 n . .... 3. and so any continued fraction may be expressed in the form F-a + &2 &3 bn (C) JU . .. The quantities a i ^2/ a 2 b n /a n .2+4+2+4 + 2. and the fraction obtaine d by stopping at any particular stage is called a convergent. . and is written Oj a . .. (B) The signs may be supposed to be attached to the 6's.. Thus any continued fraction may be written in the form (C) where a 2 . . a 3 ....is called a continued fraction. ... Prove that.. . l 4-f 4+ 4+ 2+ 4+ 2-h 4+ "" 1122 1111 The fraction on the left is equal to s ---.. .. . so that fi. Thus the first... (1) Let u n denote the nth convergent of the continued fraction. .. . Formation of Convergents... . ..3 ~ . . Ex. third. =x ... a n and 6 n fl are all changed. convergents are .....

.... and denote the nth convergent by u n ... therefore ( br+l \ Ur+1 ... 3... = .1 ' II I J. ..r +-' Thus the theorem holds for n = r-fl... Now p r _ v gv_ j. ..a 3 j> 2 4. if p n and q n are defined for successive values of n by the equations ? = ?-! + &nPn-* Vn = anq n -i + b n q n _ 2t . It will be found convenient to write y o== i an d ?0 = 0... q r _ z are independent of a r and b r .. but it holds for n = 2...........63/03. ... Assume that this holds for the values 2.... (A) then p n /q n is the nth convergent of jP.. (B) and it will be seen that Pz^^Pi^^Po an d ?2 ==a 29 r i + ^25 r o> so equations (A) hold for n^2... For the continued fraction #Ae quantities p n and q n are defined by the equations Pn = ttfiPn-l + ^n-2 ?n = a n?n-l + 6 w y n _ 2 . also u r m ay be transformed into u r+1 by writing a r + b r+1 / a r+i for a r . and ... I JL I f....63^ and ? 3 == Proceeding thus we can show that........ Proof.. .. (2) It is important to state all this rather more precisely : Definition. then p n /q n is the n-th convergent of F..THE RECURRENCE FORMULAE 389 Observing that u 3 may be obtained from w 2 by changing a 2 into a 2 4. If p n and q n are defined in this way.. r + ^ P '~ l+ TPr ~ Z _ 5^i (r^i Hence w^ ... we have u = ( 3 and so u$ = psjq& where y 3 .. ....... (A) with the initial values p Q = l 9 qo = 9 Pi = ^i> q\ = 1 Theorem. Equations (A) are called the recurrence formulae.. . p r _ z . then . r of n.

3. 4..1.x)-^l+~ |j 3^ y Calculating the convergents of the continued fraction. Prove that (I. Infinite Continued Fractions. F is called a terminating continued fraction.1 6-8o. With regard to the fraction ._ 69 6q 6 n If /I I 4 __ ?_ n JL' t*i T" " ) - if the number of elements is finite.. 390 INFINITE CONTINUED FRACTIONS (3) For the fraction b b b n p n and q n arc defined by equations (A) with the initial values Again. in succession.~ which provea the statement in question. so that the recurrence formulae are Pn = fln^n-l ~ &nPn-2> ?n = a n?n-l .therefore for w = 3. . Notice that no fractions are reduced to lower terms.6 n?n-2' In the following articles we shall call the reader's attention to certain types of continued fractions which will be considered in detail later. until the final stage.1-3&. . Ex. 5. we have 2. the fraction b b ft 2 3 is obtained from the fraction in (2) by changing the signs of the 6's.

diverges or oscillates. /.-. An expression of the form 1 1 1 a i ___ _ _ _ i T^ ... the value of the fraction F is defined as lim Pn/q n . however. 4. to oo . Theorem. Simple Continued Fractions. which is to be regarded as defining the sequence of convergents Pl/<ll> ?2/?2> > Pn/<ln> > and we say that the continued fraction is convergent." > a 2 + a 3 + a n + where every a is a positive integer.. its value must be unity. we have # = 3/(2-x) leading to x=lJ2i.We may. except that a l may be zero. z 2 + 2z-3 = 0. is called a simple continued fraction. if x stands for 333 2T2T2T"* t000 ' and we assume that x is a definite number.. z=*l or -3. suppose the b's and a's to be determined by a rule of such a kind that there is no last element.. Any rational number can be expressed as a simple terminating continued fraction which can be arranged so as to have either an odd or an . If the sequence converges. 1 a 2 +a 3 + SIMPLE CONTINUED FRACTIONS 391 333 Illustrations. where every element 2 2 2 after the first is -3/2. but we have not shown that a value exists.'. Let x stand for ^ ^ -= . we have x = 3/(2+se). divergent or oscillatory according as this sequence converges. In this case F stands for an infinite continued fraction.. . Again. We have thus proved that if the fraction has a value. a result which is clearly meaningless. In neither of these examples can the first step be justified without knowing that the continued fraction is convergent. and we write jF = a ! + . // we assume that x may be treated as a number..

etc. and if a n is the last quotient.. In the process of finding the G...^ -^ Again. and therefore A IB 1 r.C. let a v a 2t . as required. if >!.. be the quotients and r l9 r 2 . or an even.. Ex. B l Br r . r 1 = r 2 a 3 + r 3 . =flo"f 7. =O+ 7-. etC.. so that A^a^B + r^ JS==r 1 a 2 + r 2 . . A 111 B g. Let A/B be the number. we can write 1 1 1 a n ~a n -l+ andifa n 392 RECURRING CONTINUED FRACTIONS Thus matters can always be arranged so that the continued fraction has either an odd. the remainders.as a simple continued fraction with (i) an even and (ii) an add number of quotients.even number of quotients. . 1 -^ = #1 + ^prr.M.rr r 3 r This process terminates. of A and 5. A and B being positive integers. 1. Express ^g$.. number of quotients.

process to express (1 -x)~ z as a continued fraction. 3.C.a l . For instance. 13) are the least solutions in positive integers of 68* . -gg. The H. Use the H.F. 9 and Ch.= 2 + a 4 =5 68 63 157 136 5 5 21 20 1 JL I i i i NOTE. the penultimate convergents to 157/68 in the two forms above are 30/13 and 127/55. 2.1 respectively.157y = . (See Art.C. 55) and (30. 157 i. process is giving H2x .157y = + 1 and 68* . and it will be found that (127. XXV.e.F.) Ex. The fact that the number of quotients can be made odd or even as required is of importance.

2+ -9/2* + l/3 .x) 2 * ^ l/2a? + .L "" 1+ "^4/3+" ^9/2*+ 1/3 2s 3* -3/2 1 + 1.??.-9/2* 1 l-x/2 1 1 -4/3 1/3 ~x:l'il"~ x * 2x 2x -x* -X* . 2x .4/3 X 2 1 X 2 1 .4/3 + -9/2* + 1/3 .1/3 1 (1 .

thus _____ _____ ._____ . Therefore p 1 =Aoc + B^a=a+^ p =A +B =1.. if such exist. * . ^ = 1. (i) // a. 5. then by Ch. the statement at the end of Art..\3JLlv Uv7vA MY _. (i). where A. _ . prove that the n-th converge nt of is equal to (a n+1 -^ n+l )l(<x n .Afl r4"\7 .(P 1 ) or (l + l/n)a according as ot^fi or a =j5. form the * acyclic part ' of the fraction. 9 and the non-recurring elements. Recurring Continued Fractions. AD VA.4n = aq<n-i . First suppose that a^jS. and that its value is the numerically greater root of x 2 . -. + 6=0.. (ii) Hence show that the continued fraction is convergent if a 2 ^ 46. The cycle is usually denoted by putting asterisks under the first and last of the recurring elements. q Q 0. ft are the roots of x* -00._ _ ~~ ~ A _ ~. gvng ^l/= thus y w = ( n+1 -]8 w+1 )/(-]8) and fc = ( n -)8 n )/(-^ .s ^ e acyclic part.ax + 6 = 0... (i) Here p n and q n are given by Pn = <*<Pn-i ~ l>Pn. _ ._.1Q /l ATir'I'f'.^n-2 together with p = l. the elements recur in the same order. we have a ' recurring ' continued fraction. XXII. p^^a.. B.. after a certain stage. If . 1+2+3+4+3+4+ J 1+2+3+4 * * Here . 3-f 4+ J 1+2 J * SPECIAL TYPES 393 Ex. 1 has been used.2x 3x x 2x ~ 1+ ri2 + 3^T* In the last three steps. 5. 1.. E f are constants. .is the cycle and . The recurring elements form the * recurring period ' or the * cycle. A'. .

Let |a|>|/?|. j3 are real. Hence. ^ n /g n = (l + l/n)a. 2. And so for the g's.> n -(n + l)a n . the continued fraction is convergent and its value is a. // p n lq n ta the n4h convergent of I 1 3 a 5 2 (2w-3) 1+ 2+ 2+ 2 + " 2+ " prove that P n l3n = l ~ 3+5-. we have . 5. hence . If n is odd. leading to the same result. then by Ch. Proceeding as before.(ab +2)p n -2 +Pn-t =0. If v n denotes either p n or q n . in both cases. y n = na n . Ex. (i).leading to the first result.+ < ~ 1 ) n ~ 1 2 ^Ti ' Hence show that the fraction is convergent and that its value is w/4. then a. // p n lq n is the n-th convergent of the recurring fraction 11 1 1 aT 6TaT 6-f '" ' prove that p n .(ab + 2)p n _ 2 and q n -(ab + 2) q n If n is even. then a~/? and p n /q n (l+I/n)a-+oc. 3.1 . we find that A~B~\ > A' = Q. so that a = 2oc and 6=^2. B' = l/a. XXII. (ii) If a 2 >46. we have the same equations as before. then j8 n /a n ->0 as n->oo and q n ^ l-W ' If a 2 46.. the recurrence formulae give Pn =t>Pn Pn-2 = pn whence eliminating p n -i and p n -3> we have p n . 394 BROUNCKER'S FRACTION Ex. except that a and b are inter changed. Next let a=)S.

. "" ""'"" 1 "* " 1. . 3 .2. 4.. is convergent and its sum is Tr/4 . whence we find that q n = 1 . The reckoning on the right shows that ./. Find the first four convergent (unreduced) of r\ JLJLJL* /-\ JLAJL! w 2+3+4+5 f W l-2-3-4' 2. ther efore the fraction converges and its value is Tr/4.1). ..1/3 -f.3.. EXERCISE XL 1. the series 1 . we have ^n-i 2n-l' .. . we find that A =0.and this holds if we write n .. 2 in succession for n.. ft .. Again. Without calculating the convergent^.5. 4 . (2n . therefore n -(2.1 (2-3)(2n-5) .l) n .. Replacing the t..'s by p's and observing that jt) = and ^ = 1.lK_i=(. Putting n~ 1. Express fff as a simple continued fraction having (i) an even and (ii) an odd number of quotients. prove that 2 3 4 5_ 1 1 23 2+ 3T T 5^1+ 2T 3T 4 ' 3.1.(2-l)"" 1 3 where ^4 is independent of n. leading to the first result. .l)7n-i-0.3.). !(.-. replacing the v'a by g's and observing that (fo l an( ^ 5 F i == l we ? n -(2n. Next.1/5 .

2. /i + /a + INFINITE CONTINUED FRACTIONS 395 7.85 1_J_1 5 52"" 3. 5.. .l)/(2a*-3a) in the form 1+^ .'" 2^" a?"" nx-n + T * .. Show that the nth convergent of 2 JL_1__! 1 2-2-2. a * ~ a * """ 6.. where the a's are integers greater than unity. There being (n + 1) elements in all.4.5* 52 57 85 104 19 20 I Explain this.1 l_(n + l)a?-n ~2. show that 2.. 8. and that the value of the infinite fraction is unity.is (n + l)/n. and that the infinite fraction is divergent. and verify by calculating the convergents. . 9. Show that the nth convergent of 1 1 1 1 rr^r^r^r is n. Express (2a 2 -a.. Express ^ in the form a l .

1. 15. the numerically smaller root of x 2 -ax + 6=0 is equal to A JL J A a a # a [This follows from Art.3+ 2. Ex.1+ 1. 396 SIMPLE CONTINUED FRACTIONS 16. For the fraction b b b b a + a + a + " * a + * " ' prove that p n = . 1.10. 2 o B.] 11. g n -( n+1 .A.^ 1^4^1-4^' \2~2^2-"V" Hence show the nth convergent of the first fraction reduces to 2n/(n + 1). q n respectively are denoted by u n .1 ( 14. 5.0).a ( a n -)/(. Prove that . each fraction containing n elements. 12. v n respe ctively.C. If a z >46. 9fy . ^ a re the roots of # 2 -a# -6-~-0. fi\ l x _ _z_ z = _ v . If p n jq n IL the (unreduced) nth convergent of and the number of terms in > n . 6 are positive. then u n \v n is the nth convergent of Hence show that 13.j3 n+1 )/(a . Prove that 1 I 2 2 2 J3_ (^-1] 2 _ 1 11 1 showing that the infinite continued fraction is divergent. the fraction is convergent and equal to the positive root of x~bj(a + x). Prove that. J_ J_ JL JL _o/JLJL-i. Prove that . Hence show that if a.]3) where a.

. If n is a positive integer.-. .t> + .. 18. Simple Continued Fractions are of the form 1 l l x l a t + ~~"~" r~T * * ~j ' where the a's are positive integers.~^. + ( . > 3/3 Gt 3 + ...1 w"+ w . For the fraction 1_ I 2 2 2 3 a (?i-l) 2 r+ 1 "*. . prove by induction that n n [ n . 6. 19. are called the (partial) quotients and if 1 1 _ 11 X 2 = <X 2 + ~ 7 ~ .l)<fa_ 2 }. ' showing incidentally that. 15 is divergent.. show that m jw-l) a (w--2) 2 I 2 J-itJli i. iftl O ~~ ju'tTl ~~ <L) ~~ x Also prove this result by induction._ ..w .I) 71 " 1 Hence show that the fraction is convergent and equal to log 2...~ . (a + n) 9 then Use this to show that the fraction in Ex. Putting a~ -m.... 16. . .n ^ 1 .' ' ' 2a + 2n -I. Here a l9 a 2 . . except that a x may be zero. if k n = a (a + 1) (a + 2) ...] 17. where m is a positive integer. For the fraction .-2 2 11 n -t. .~ 2a + 5 .1+1+ "* r+ "" ' 7) 11 1 prove that ... prove that 1 + 2 + 3 + n + r J>n_ ^j ?n ^1_1_ in -hi \n-i I 12 13 Hence show that the fraction is convergent and equal to l/(e ~ 1). [Here g n -(a + H)g n _ 1 ^=(a + n. . in Ex.!){?! -(a 4.1*+ ?T-~ "2+ "" 2"+ 1 + 2 ~ w+~2 ' 20.

. are called the complete quotients. FUNDAMENTAL PROPERTIES 397 PROPERTIES OF THE CONVERGENTS 7. 8.. the continued fraction x l is obtained from its n-th convergent by substituting the complete quotient x n for the quotient a n .........3 \ . also therefore Pnfr-1-. ....... since q^q^...1 it will be seen that q n is the same function of a 2 . ..2 and so *'or the y's. they tend to infinity with n. mi . Xl "~ ai " f a 2 + *"a n ^+ x n l that is to say. and U/ 1 a 2 #o#3 -4. a n _ x ..1. /1\ 0i a^o-fl a^a. As in Art... n . The convergents are -. and in the case of an infinite continued fraction.Pn -i7n = (-1) w . p n and q n are defined by the equations with the initial values jo = 1 . . 9.1 ..... -f p n .. .. p n ~ p n ^ = (a n . (B) Again... From the recurrence formulae (A) we have Pn f ln-l -Pn-dn^fanPn-l +#n -2)^-1 ~ Pn-l (n?n-l + <ln 3) = -(Pn-i9'n-2-? > n-27n-l)This holds if we write n .a 3 + a 4 + a 4 + a 5 + then x 2) ^ 3 ... thus j) w a/?d g n .. ~*-. increase with n.... a n as p n ^ is of !...! ..1 )p n .) ... a 3 .. a 2 . 2. Thus we have JL _L_JL... ~f-o -ha... we have therefore P1n-z-Pn-&n = ( ^} n " l ^n . We have p 2 >p v and if n>3.. (0) From (B) and (C) we obtain the formulae .. .......2.. .2 in succession for n. g^ = 1 .. etc. p l = a 1? </ = 0.1 ...2 . -.. ..

. ' 2 -r i* n n-l l where in the second the quotients occur in the reverse order.. we have and n . It has been shown that if p'q'.. p r '/([r are respectively the r-th convergent of 11. ^uer?/ convergent pjq n is in its lowest terms.-.= n ---. q n ).. p/q are consecutive convergents . + a.2. . (E) (F) It follows that if p r /q r . The second of these holds if we write n . 398 ODD AND EVEN CONVERGENTS Also PI/PQ^^I and JaAh^^J therefore /n 1111 -. also p n is prime to Pn-\ and q n to q n _ v For if the numbers belonging to any of the pairs (p n . p n '/q n ' ~p n lp n -i and p n ^ 1 /? n-1 = ? n /?n-i an d by Art.10. 3 for n. From the recurrence formulae (A). 12.] . . g would divide 1. 2 for n. n .1. then on account of equation (B).. then For by the preceding. (p n . 1 1 . Pn-l Pn-l/Pn-2 ?n-l <In-J<ln-2 The first of these holds if we write n . 11 and a n + a* + a n n a-. 10 these fractions are in their lowest terms... .. n ~ 2. p n ^) 9 (q 7n-i) have a common factor g.

. and consequently q' = q" and p'=p". let pq-p'q=l where q'^q. then pq' ~-p'q = 1. Conversely. every convergent lies between the two preceding convergents. and suppose that p/q is expressed as a simple continued fraction with an even or an odd number of quotients according as the arbitrary sign is -f or . therefore Thus the odd convergents form an increasing sequence. pq f -p f q = pq" -p"q. Sequences of Odd and Even Convergents. it follows that Wn-w-i and u n -u n _ 2 are of opposite sign. We can prove that p'jq' is the convergent immediately preceding p/q. 1 4. therefore q cannot divide q'-q". W 2 n 7 U*n-l == l/?2n?2n-l-> 0. Each convergent is a closer approximation to the value of the continued fraction than the preceding. Let w n then.of a simple continued fraction. CLOSENESS OF APPROXIMATION 399 Consequently. (i) every infinite simple continued fraction is convergent. 9. In other words. Therefore u n lies between w n _ rl and w n _ 2 > that is to say. for by equation (D) of Art. the even convergents form a decreasing sequence. and these limits are the same. . Now W!<w 2 and consequently W 1 and so on. 13. which proves the theorem. as n tends to infinity. both w 2n _! and u 2n tend to finite limits. from equations (D) of Art. For if p" jq" is the last convergent but one. Now p is prime to q. 9. Hence u n tends to a limit x such that for every n. and every odd convergent is less than any even convergent. and therefore p(q r -q")q(p' -p")> . therefore q divides q' -q" or else q' = q"Also 5'<<7 and q"^q. and (ii) its value is greater than any odd convergent and (iii) less than any even convergent. by hypothesis and equation (B) of Art. 9. for pq -p'q= 1.

For let z a l -\ -.<7 n _ 1 .For let z = %-{- 1 1 1 . 14.' (zg^ .'^!. is nearer to p n /<In ^ an to p n -\l<ln-r 1 5.. . By Art. and let p'jq' be the convergent immediately preceding p/j. it is also nearer to z than p'/q'. then Pn+l/<ln+l = ( a n+lPn + ^n-l)/( a n+l?n + ?n-i) ' and therefore z = (z'p n +p n -. therefore z-- that is to say.i)/(z f q n + ?n-]) Consequently 2.. z lies between p/q and p'/gr' . therefore r/s lies between y/g and p'jq'.j> n ) = ..(2. ulwy convergent p/q of a simple continued fraction is a closer approximation to the value of the fraction than any rational fraction r/s with a denominator less than q. and 2'=a so that z' is the (n + l)th complete quotient. Moreover.. if r/s is nearer to z than ^/j. and P'i 1 .y n _!) and 2 " = Now g r n >? n _i and 2..

17. I W n .sp' is an integer> it follows that 400 LIMITS OF ERROR 16. Any positive irrational z can be expressed as a simple continued fraction. by p n /q n . therefore I t*n ~ Un+2 I <\ W n ~ * I <Ki .e. in representing the value z of the continued fraction.W n + 2 I =nW?n7nf2 and I w n ~ w n-h whence we have the inequalities (H).where a l is the integral part of z (which may be zero). g n + a = a n +29Wi + ?n> therefore ?n +2/ a n+2 ^ Also q n+l >q nJ and therefore l/9 n ?n Hence the inequalities (I) follow from (H). then by Art.l# n+ Also by equations (D). Again. w n> ^n+2> ^> U n+l9 is an increasing or a decreasing sequence. Therefore ? | ry' . If is the numerical value of the error. since ry' . Let z a l + -. 13.^jp' | <^ . then and within wider limits For if u n =p n /q n . Z 2 . and.r_ \~8 q' 9 9' i. and that in only one way.

.a - 3 3 . then Now as n tends to infinity. let Z 2 = a 2 4-l/0 3 . the fraction F converges to z as a limit.then l/z 2 <l and z 2 '>l...<z. it can be shown that if u r is the rth convergent of F. it follows in succession that a 2 ... we have a 2 <2 2 . That is to say. a 3 . a r <2 r <a r + l.. # 3 <z 3 . a 3 . Z n = <*n+l/Zn+l> where a 2 . Suppose that _JLJ_ _A 1 1 ..~ <2o. etc. . since z a >l. 0-2 + a n + Again. u 2n _ l and u 2n tend to the same limit... ----.4 EXPRESSION FOR A SURD 401 Continuing thus. therefore 1 1 ! + >z... which must therefore be equal to z. a + a * a H. a 2 + .. and we write JL JL 1 a 2 -h '" a n + It remains to show that this mode of expression is unique... a 2 a 3 j 11 1 1 and a l -f -------. .. . are positive integer s and we have 1 11 The process can be continued indefinitely. and leads to the infinite continued fraction ] j F S =O I + -. etc. Z 3 -a 3 -fl/Z 4 . for every r. etc.. a 2 +-">2 2 . are integers such that. Continuing the process.

1 a.1 ~\ 7m . and 11 Hence. therefore i j a.d H -. and N/37+8 ^37-l_.-----. = 6. The value of any infinite simple continued fraction is an irrational number. 18.o -H -7=- -7= . Express (/v/37+8)/9 a* a simple continued fraction. . and a 2 H -----. For any rational number can be expressed as a finite simple continued fraction.: ' 1 "f* ~ 9 iT "9" ~ s/37 + r 4 A " 4 V37+3' s/37-4 3 \/37 + 4 ___ N/37~5_ 4 -. The integral part of this surd is 1.. = 6 2 + T 11 2 1 t 1 a 2 -6 2 and a 3 + .. any convergent of the continued fraction will be called a convergent to z.'l. 4 N/37 + 1 \/37~3 7 1 -f "~ -.. =6 3 + r v: > C&4 -r C/4 -r and so on indefinitely. i j 1. . Thus the continued fractions are identical. by similar reasoning. -.> u -i tj -r > .^ . . Ex.: 1 H ~. When z is expressed in this way. = 6.The integral parts of these continued fractions are equal.

and the error is numerically less than I/q n q n +v an( i a fortior i less than l/q n 2 . then p n /q n is an approximate value of x in defect or excess.=l+eto. It is required to find good approximations to the value of TF in the form of vulgar fractions. ^2+ . L JL 77 7+ 15+ 1+ 292+ 1+ 1+"" The convergents are 3 22 333 356 103993 I* T* Toe* m' 33ioT'"" The denominators 106. 4 4 v/37+3 7 The complete quotient (s/37+3)/7 has occurred before.V37+4 3 3 V37+5 V37+5 . and they are both in excess.-tithe continued fraction being a recurring fraction. 402 INTERMEDIATE CONVERGENTS 19. 5 : is/37+8 1 J_ JL_ 1 9 IH. This is the required approximation. Approximations. and the subsequent work consists of repetitions of the last three steps. we calculate the convergents until we find one p n /q n such that q n q n+ i ^a. Or again. 1. Ex. according as n is odd or even. with the notation. 33102 are large compared with the preceding ones. ^37-3 . Now 2n+i = 0n+i7n + ?n-i> hence if a n+1 is large. For 355/113. thus. the conditions are satisfied if q n ^*Ja. (1) If a given number x is expressed as a simple continued fraction of which p r /q r is the rth convergent. . of Art . if we require an approximation with an error numerically less than a given fraction I/a. Of course. JL _JL _L _JL. 7 _ . therefor e 22/7 and 355/113 are specially good approximations. given that _q . the error is less than 113x33102 3740000 .1+ 3+ 2' x. p n /q n is a specially good approximation.=2+-.

The most important fractions of this kind are the so-called intermediate convergents. p'/q. (ii) Every fraction P T jQ r is in its lowest terms.p n +a/?n+a. 21. 2). one an odd and the other an even convergent. We may then proceed as follows : Choose two convergents p/q.P r Q r+l =p n+l q n -p n q n+ i . Also x lies between these convergents. (iii) Any fraction which lies between P r /Q r and P r+1 /$ r+1 has a denominator greater than either Q r or Qr +v (iv) Limits to the numerical value of the error in taking P r /Q r as the value of x are given by For we have P r+l Q r . the fraction p m p + m'p' Q mq + m'q' lies between p/q and p'/q'. m( are positive integers. Ex.( ~ l) w+1 . therefore P/Q is a closer approximation than p/q. and the denominators form an increasing sequence. whence the statements (i)-(iii) follow : and because P r IQr> Pn+2/q n +2> x > 2W]/?n+1> is either an increasing or a decreasing sequence.odd or even. Let p n /q n be the nth convergent of 1 Consider the sequence Pn Pn+Pn+l the (r-hl)th term being P r jQ r where P f APPROXIMATIONS 403 The terms of this sequence (excluding the first and last) are called the intermediate convergents between p n /q n and . The problem may be of such a kind that no suitable approximation can be found among the convergents (as in Art 22. and it may be possible to choose m.The sequence possesses the following properties : (i) It is an increasing or a decreasing sequence according as n is.20. If m. therefore . and perhaps also than p'/q' . m' to suit the conditions of the problem in question.

the one which is nearer to x is the number required. 404 APPLICATION TO CALENDAR To do this we express x as a simple continued fraction.h terms. then P/Q is the closest approximation in defect of all fractions with denominators less than a. It is required to find the fraction which. we first consider approximations in defect. and calculate the convergents p^q^. by considering the sequence (B) formed by the even convergents and their intermediates. except that it is a decreasing sequence in which each term is a closer approximation in excess to x than the preceding. For if h/k is nearer to x than P/Q.Qr Qr Since r<a n ^ 2 . 22. In the same way. or else it must lie between two suc. Now consider the sequence formed by writing down the odd cqnvergents and inserting all the intermediate convergents. P'/Q' are successive terms. then (i) This is an increasing sequence of fractions in their lowest terms. If q n <. Taking the sequence (A) formed by the odd convergents and their intermediates. and in either case k^Q'>a. any fraction which lies between them has a denominator greater than Q or Q f . and each fraction is a closer approximation in defect to x than the preceding. has the same properties as the preceding. this proves the statement. . Of these two fractions. then h/k must coincide with a term to the right of P/Q in the sequence. Again. Problem .a. (iii) If P/Q. we find successive terms P/Q. pdq& until we find p n /q n such that If q n ~a. we can find the fraction with denominator less than a which is the closest approximation in excess. (ii) The denominators form an increasing sequence. P'jQ' such that Q^a<Q' . then p n /q n is the required fraction. is the closest approximation (in defect or in excess) to a given number x. the sequence formed by the even convergents and their intermediates. of all those with denominators less than a given number a.

.. and is an over -correction. q 7 89 ' q 8 120 q g 809 For the approximation in defect. ?>io/<? io. The fraction 7/29 is inconvenient . Ex.. the year being taken as 365 days. . We therefore p roceed . This is also an over-correction. 1.e... amounts to adding 1 day every 4 years. t .. but 8/33 suggests an addition of 24 days in every 99 years. and 2\IQ 9 -. <. Ex. except at the end of a centur y. so that the required approximation is 1680/449.. . we insert intermediate convergents between p 1 lq 1 and p 9 /q 9 .. P! 333 p B 449 p 9 3027 and we find that ~ .= . calendar. 1. The mean tropical year is 365-2422642 days. CHARACTER OF APPROXIMATION ' 405 None of the other convergents give convenient approximations . As in Art 18. Finally. we can show that 14 . These are of the form P r /Q r -= (333 + 449r)/ (89 + ] 20r).(1680/449) 2 <(449/120) 2 . The conditions $ r ^500<(> rfl give r---3. i..1680/449. For the approximation in excess. we consider the convergents > 8 /<? 8 . and the same is true for the intermediate convergents in the first few intervals. It is required to find a prac tically convenient method of correcting the.Ex. 1+24-1+6 * * . make every fourth year a leap-year. . fraction with denominator less than 500 which is nearest to s/1 4.14.-. there are no intermediate (jonvorgents and the approximation in excess is ?/& =449/120.Since a 10 = l. We have < * 1 ' ' and the convergents are 1 7 j* 39 j47 321 1* 4' 29* 33' 161" f 194' 1325"" Take 1/4 as a first approximation. we can show that N /14 = 3 + . This. 2. Find the.

i t makes every fourth year a leap-year. The fact that this approximation. ! P I 1 i P 1 If z ~ \ < "7 -----. 161 161 ' The values ??i~19.39 97 19. except those at the end of a century. Tel "400 * This gives one more day to be added in 400 years than the fraction 8/33 .. can be obtained from the convergents 1/4 and 47/194. if m.= a l + --. . We choose the convergents 1/4 and 39/161.^ 9 or a fortiori if J J \ ^ 9 q\ q(q + q) \ q V then p/q is a convergent to z. 20. then.1+2.. shows that this also is an over-correc tion. 97/400.as in Art. The answer to this is contained in the following : Let p/q be expressed as a simple continued fraction with an even or an odd number of quotients according as p/q ~g z. and let p f /q' be the last convergen t but one. m f are positive integers. m'^2 give the approximation 19. For suppose that p 11 .4 + m'. . l4-m x .39 39 4 ra.e. which is an approximate value of some quantity z. It may be important to know under what circumstances p/q is a convergent to z. Let p/q be a fraction in its lowest terms.. which are both in excess. i. 1 m. MISCELLANEOUS THEOREMS 23 . 4 + 2. The error is less than a day in 4000 years. the first an oven and t he second an odd convergent . with the fu rther correction that every fourth century is to be counted as a leap-year.

If p r /<lr i s th 6 r ~th convergent of the simple continued fraction of which the value is z. for Pn>p n -i. // p/q is an even convergent. 25. of a simple continued fraction whose value is z. for then z' can be expressed as a simple continued fraction in which the first quotient is not zero.q #2+ a n where n is even or odd according as p/q g z.q') is positive and ____ q(qz'+q') \q Hence qz' + q'>q + q' and z'>l. which proves the theorem. and X/Y an odd convergent. For if k is the complete (n-f l)th quotient. according as p/q L z .to prove that z'>l. and q n >q n -i. then . 406 SYMMETRIC FRACTIONS 24. therefore in both cases q(qz f 4. and let z' be determined by 1 1 1 ^a-i a It is sufficient . then z 2 5 PnPn-il^n-v according as n is even or odd. We have *=& andthcrefore '-*-&-:** qz +q q q(qz +q) Now pq f -p'q^I or 1. which proves the theorem. The expression on the left is equal to and this is positive or negative according as n is even or odd. Therefore z^^n-i/g^n-i* according as Jrfn-l ( k Pn + Pn-l? ~PnPn-l (*? + ?-l) 2 ^ 0.

and it occurs later than p/q or else coincides with it. in which the quotients equidistant from the beginning and end are equal.^: <i _L__ 77 4+l-h4+3' 132 4+1+IT4T3' are symmetric. let X'/Y' be the convergent which immediately precedes X/Y. the fractions. therefore X X' p ~ and since X/Y is an odd convergent immediately preceded by X'/Y'.according as p/q precedes or follows X/Y in the sequence of convergents. XX' . and the latter an even number. Symmetric Continued Fractions. z*> and qq qY 26. (i) // p/q precedes X/Y. For instance. ^ . . let p'/q' be the convergent which immediately precedes p/q. therefore x _/ p 7 |<*<f: and since p/q is an even convergent immediately preceded by p'/q'. A simple finite continued fraction.9 pX (ii) // p/q follows X/Y. Then p'/q' is an odd convergent. SUM OF TWO SQUARES 407 The following is an important property of a fraction of this kind with an even number of quotients. is said to be symmetric. Then X'/Y' is an even convergent. the former having an odd number of quotients. and it occurs later than X/Y or else coincides with it.

therefore P ~Q. P'lQ' be the convergents immediately preceding p/q and P/Q. Q' = q* + q'* and Q* + l= For by Art. P/Q being in their lowest terms. 11. also therefore p_ a 7 tt r 1 + l and p a r _ x n h % 5 .. and let p'/q f . and ^ = 1 a r + a r + a r _ + a x the fractions p/q.= #! H q j .. P (1) Let . then P=p*+p'2. P/P'=P/Q./n r .

PQ / -P / Q = (~l) 2r and P /= =Q..P 1 * i 1 Q~ c F }. 1. we have therefore any common factor of X and Y divides p and p'. therefore X/Y is in its lowest terms. a 2 + '" a r + P' 1 L. // p r /q r is the r-th convergent of the continued fraction . and consequently Moreover. therefore ##. therefore P = X and Q = Y . So also is P/Q. -4I p pq+p'z" / r = pq+pq a r + q <f + q' 2 If X=p* + p' 2 and Y^pq+p'q'. which are prime to one another.

26.M. 1901 =^ a 2 +z> 4 2 = 26 a + 35 2 .C. the last being p 2 =a j p l + l.C.M. Hence the G._ Q~ " at+'" a r + a r +'" a^ prove that the remainders in the process of finding the G. The reckoning shows that 1901 _LJL_L J_ JL J_I 218 + 1+ 2+ 1+ 1+2+ 1+ 8* Hence. of Q.P 1111 ff i _____ . 1 and Art. 1901. (See Ex. process is that shown on the right. if p r /q r is the rth convergent. Given that 218 2 -f 1 =25 . &r. Q=p* r -i with sinaikr equations. express 1901 ow fe *ww of two squares.) P2 <hP* r-i Par a 1 Pz r-3 Pzr-2 i Pi Pz . 2. P are For P=p 2r .

1 1 2 21{ 6 2< $ 1901 L 157 3 35 8 2 1 8 I ? 9 1 1 408 SUM OF TWO SQUARES .i 1 sum o ftwt ? squares.

2_ h2/ 2 = p an( j ^=^2^.+ ---. . .. . we have = l4-r i ^ ^ ....3 .1 = Pq^r-v Ex. . 28... by Art.39 2 .. o9 54-14-14.a l a 2r _ 1 = a 2 .= = a 2r 4 -. if p m l^ m is the mth convergent to the last continued fraction.. and the continued fraction is symmetric. (A) Q l then if P'fQ' is the convergent immediately preceding P/Q..... then an integer Q can be found so that Q* + 1 is divisible by P..* Consequently 7. it is a divisor of Q-P' or else Q = P'. The first alternative is impossible.. y a r-1 -h a x Q a 2 + a r 4. . n . Application to the Theory of Numbers.. 4 -----. . 1.. .. ---. since P is prime to Q. any factor of Q 2 + 1 can be expressed as the sum of two squares..a r-1 4. therefore Q==P' and R = Q'.1 p P 1 11 1*1.. we have. find a solution in positive integers of the equation Using Art. Given ^Aa^ 3637 = 46 2 4.. for P>Q and P>P'..a r 4... .. Denoting the mth convergent of (A) by p m /q m . etc.= a r 4 -. . then obviously P>Q and P is prime to Q. . For let Q 2 + l=PR where P>R. Z=y r 2 4-^ i==a.. X 1 111 1 Let .a x then.a 2 4... Also ^L^r-i ~ Qpzr~\ == 1 ? therefore Q 2 4...= 0.. T z 1 1 . and writing down the continued . // Q is any positive integer.. which proves the theorem. and -...= a. Therefore P(R-Q') = Q(Q-P') and. 26. --...27. 28.. Let p l l /A* ... Conversely 9 if P x 2 + y~ where x is prime to y and x>y. and therefore a 2r = a l9 ^ J Q P 2r a 2r _ t 4.. ..

These may be grouped as follows : (i) Fractions which have no acyclic part.. If x is prime to y.. 30. (ii) Those with an acyclic part consisting of a single quotient.. then where the order of the quotients is reversed..F . 7 7 7 .3 f 1027 3637 of which the last two convergents are .1)=: RECURRING CONTINUED FRACTIONS 409 29... The conclusions are of fundamental importance.. (iii) Those with an acyclic part consisting of more than one quotient. Simple Recurring Continued Fractions.. 290). any factor of x* + y* is the sum of two squares... Or thus : the last two convergents of 3 + 1. then a is a root of a quadratic equation with integral coefficients. and we proceed to consider these types in order. 290. The reasons for this classification will appear later. . Moreover. For we can find an integer Q such that x 2 -f y 2 is a factor of Q 2 + 1 and every factor of Q 2 -f..1 4-" 5+ 1 4.. <<J90 10/27 and the required solution therefore is (1027. 31.. we have 3637 x 200 . (**) &2+ b 3+ b n+ &1+ &2 + the fraction having no acyclic part.-.1 + 5T 1 -f 1 4.. . 1111 1 i T (X> U-t T~ 7 7~~~ .1027 2 = 1. No Acyclic Part.fraction JL _ X -1 JL _L 1 l LA 1 4. if j8 is the second root of the equation ..are TT ancl i o Therefore 290 and x = >/(3637 ....1 is the sum of two squares. whence it follows that -1<JB<0.

we have OC l/i T" 7 '" ~" .. and if p r /gv is the rth convergent of the second fraction... and the corresponding roots of (C) by a'....... 'Also..l/j3 is the positive root of (D). (B) by p r /q r and Pr/y. we have a'^ + l/a.. ^-(Pn-ffn-Jy-Pn-l-O.. /? is negative. 11 is the same equation as Now equation (D) can be transformed into equation (C) by writing ..Denoting the rth convergents of the -fractions (A). j3 where a is positive.. /?'. .1. ' = 04 + 1/0. (B) and therefore p n -i(x-<*i) 2 + (Pn-q n -i)(x-<*i) -?n = . and by Art. and is consequently equal to the fraction (B). as in Art. 31.. and -^>1 .. The root a' is the value of the fraction (A).....r respectively. A single Acyclic Quotient.. 31 we have . ' T 6 2 + b n +oc therefore a is the positive root of and in the same way the fraction (B) is the positive root of Jn'zMK-CJ*-^--!^ ' which by Art.......... from (B). then a is a root of a quadratic equation whose second root cannot lie between and .Ifx for x.... Therefore . ( c ) Denoting the roots of (B) by a..... 11 11 1 1 then x-a^^I/y... from which it follows that 410 STANDARD TYPES 32. l l ] <? siw/Ze quotient a l does not recur.

. ft respectively for y in (B). whiqh is contrary to the hypothesis. which is contrary to the hypothesis...... r .......IJL/I 1 " v ' SUMMARY OF RESULTS 411 Also if we eliminate y from equations (B). ..7 . Thus a' is equal to the fraction (A) and From equation (D) it follows that a m + l/p=a m -b n -f. . It follows that a m + l/j8-/-/ or -/-/.. for if a m = b n then a w would belong to the cycle.. Consequently /?' = /--/ or -I -f where 7 is a positive integer....vl ^ t/ n ~ 7 . and therefore /J' cannot lie between Oand -1. a being positive. ... 1111111 // _ n l XI OC .. 33.. . ! 11 j 1 i 1 11 _ A i ___ __ _____ O Tlfi -. .. Now a m is not equal to b n . /. (C) we obtain a quadratic in x with rational coefficients... A) i _ _ _ _ I II l t'lii i <*l-l. for if a l = 6 n then a a would belong to the cycle.where /is a positive proper fraction. then and ?n!/ 2 -(F-?-i)!/-yn-i = .. More than one Quotient Acyclic. /?' are found by substituting a. .. = .... and t/ = 0i +.. a 2 + a m +6 1 +6 2 + y 6 2 + b n + and let P r /Q r and p r /q r be the rth convergents of the first and second fractions respectively.. then by Art.. . 1 111 . /J are the roots of (C). ..11 Let x = a t +. 31.= > ... Now a l and b n are unequal.. where / is a positive proper fraction..Uf-l "1 in which at least two quotients do not recur > then a is a root of a quadratic equation with rational coefficients whose second root is positive. (C) If a. whose roots a'...

and there fore also /?'. then j3>0.. is positive in all cases. Ex. . prove that ... 412 COMPARISON OF PHYSICAL UNITS 2. For the fraction --.C. .. where ^ = 1+ __.1 or j8>0.) 2 D B. . JL ~f~ JL ~f" o ~}~ * * Denoting the value of the fraction by x y we have i 1 1 ^ 1 1 1 ^l + _.. 2. and y = giving y = (s/37 + 3)/7 and * = EXERCISE XLI 1.. prove that p n q n -i an d provided that | x \ < | ( . For the fraction ---. Also the second root (3 of the equation is restricted as follows : (i) If there is no acyclic part..~ . Thus 1 1 1 . (iii) If the acyclic part contains at least two quotients.a + </a 2 + 4).1 and since a /m __ l is a positive integer. It has been proved that any simple recurring continued fraction is a root of a quadratic equation with rational coefficients..i+ m . then /?< .A. NOTE.. The argument depends on the existence of a m _ l9 and fails if there is no t more than one quotient in the acyclic part. 1. Find the value of 1 +.i+ or . the expression on the left. Summary. 34. 1 _ __ ft I ________ fYf ft Um . then 1 </?<0. -----.. (ii) If the acyclic part consists of a single quotient.where I is a positive integer. (See Ch.. XXII..

A straight ruler is graduated in inches and centimetres. Prove that / n . .000.1 1 1 1 /-T.111 4-1)^ CL 4 . .(ii) l>n?n-4 -S f /iJPn-4 = ( . 22. the ratio of an inch to a centimetre being N.. Prove that Deduce the following : -JT. If N is expressed as a simple continued fraction and p/q is any convergent.. \<2(<Z' 2+ 2a + 2+ ~ ' k 1 1 1 1 4. show that the distance between the graduations 4 p centimetres ' and ' q inches ' is less than the distance between any two pre ceding graduations corresponding to a convergent of the fraction. Ex..l) w ~Mnn-ian.-2 3. (See Art. 1.. Prove that (i) 1111 // 1\/ c \ (n) a -f f r~ *T .A/ a + 7 ( a + r~" 7i &4-c-f6+2a V\ 6/\ 6c-{-2/ 6.~ A 1 va 2 ~-2 a- 2a -f 2a -f 2a -f 1 1 1 5. Show that 449/120 differs from ^14 by less than 1/90.) 7.

11. are the closest approximations (i) in defect and (ii) in excess to 0-2422642. by Art./_*". (iv) ^ 2 n4i = _/2n :r:: (^n-|. For the fraction F = -. then l+frr-o: 2 _ 1 . then (iii) ap 2n = bq 2n __ l ~abd n /d l .1. 10.2)x -f 1 ~ and d n = a n . Given that 1 metre 3 -2809 ft. the error being less than half a grain. giving ^=(l--i)/^. Obtain also the usual approximation. If u n =-p 2n . giving -4~ -B=6/c/ lf /.\Bfi~p 2 ~b. Zj rBfP~p z -=c.. 2n or for > 2 n-i> then.) 9. 5.(ab 4. . If # n /g n is the ?ith convergent to -----. Given that 1 kilogramme = 2 '2046 pounds. B~-(l-fi-*)ld l9 :.CX 2 -f iC 4 1 -f fix .4 n. Ex. 2.8. Aa. obtain the two approximations. (See Art.^ 2 [If w n stands for y. and find an upper limit of error in each case.. and cc 2 < |( -c-f \ 7 c 2 h4). If p/q. prove that F lies between *Jpp'/qq' and *Jp'p"Iq'q"- SPECIAL TYPES 413 12. p"lq" are consecutive convergent^ to a simple continued fraction F.B~ p Q ~ 0. Ex. Ay. show that 44 kilogrammes is slightly greater than 97 pounds. prove that a+ b+ a+ b + r Also if a. 8 kilometres 5 miles. 7 grammes 108 grains. 2.---> then .. p are the roots of x 2 .] 13.d fi-i)/ d iFinally. then Ja \~lJf3=pi 1. (v) if e^a& + 2. 22. it is easy to show that and then the results in (v) can be obtained by equating coefficients. -----.-i .i -d n )/d 1 . j) 2w -i-=(^n. Find the fractions which. 103 kilometres 64 miles . we have . p'lq'. of all those with denominators less than 500. p^n^d^d^ If w~^? 2 n-i> since c^a-f-^.

stand for positive integers. CHAPTER XXV INDETERMINATE EQUATIONS OF THE FIRST DEGREE In this chapter a. .Prfr-i +Pr-i<lT.. It is obvious that we need only consider the forms ax by = c and ax -f by = c. then ^ =Pr (Pr +Pr-i)> G =.+ '*' a r _ t + a r + a r _ t -f '" cu-f _.. If either of these equations is satisfied by integral values of #.3# .. and we shall assume this to be the case.1 and is repeated any number of times. 6. in which a is equal a... and more particularly positive integral solutions.. no solution in integers exists. this can be removed by division. then any common divisor of a. 15. __________ _ . .. If n L H --. b must divide c. R are positive integers such that Q~-l=FR and JP>JR.i -f 2 + rt n Pw-i^s-i-gn-iPs-i is the numerator of the (n~5)th convergent to . prove that $ 2 .1 is divisible by P. must have one of three values. Q. t/.a. Hence if a. Hence there is no loss of generality in supposing that a is prime to b. = ~ and P/Q is in its lowest a 2 -f #r-i + a r + a r-i + 2 + a i V terms. If a.. The Equation ax~by = 1 (a prime to b). .' and if p m jq m is the rath convergent. . then can be expressed in the form P 1 111 11 _ __ i ^.. Prove that the fraction a ----. 2. ---.. b have a common divisor which divides c.. c. Q=23. ---. n + a n-l + n_2 + 14. . b have a common divisor which does not divide c. 3.x to .. 16.R = V2r = 3r-JL ( Verify by taking P = 33. A Single Equation of the Form axby=c.2 ~ 0. 1. and that if x satisfies the equation 2# 3 -f 3# 2 .. . the fraction sat isfies this equation.. of a single linear equation in two or more variables or of a system of m linear equations in n variables (m<n). Underthis heading we consider integral solutions. If P. _______ _ Q x a..a.

GENERAL SOLUTIONS 415 (iii) Since p/q is a convergent which precedes a/6. Therefore (x qc)/b = (y-pc)/a t. 2.. The last convergent is a/6. (ii) To find the general solution in integers. ft =pc . y = pc + at ( = 0. 2. Hence the general solution is x = qc + bt. express a/b as a simple continued fraction with an even number of quotients.. y~p + at where = 0. Let p/q be the convergent immediately preceding a/6.am. We shall call this the ' least ' solution in positive integers..1.(i) To find a solution in positive integers. and therefore (q. Ex. (iii) Let m be the integral part of the smaller of pc/a. (iv) Since a(b-q)-b(a~-p)= . qc/b 9 then the least solution in positive integers is (a. Hence a(x-q) is divisible by 6. y) is any solution in integers. where t is an integer or zero. aq-bp~l (Ch. it follows that p<a and q<b. Then aq-bp = l and aqc~bpc = c. .). The equation may be written a(x-qc) = b(y-pc) 9 and since a is prime to 6. pc) is a solution. and the general solution in positive integers is x=oc + bt. express a/6 as a simple continued fraction with an even number of quotients (Ch. the ' least ' solution in positive integers of ax -by = -1 is (b-q. then ax . y~p + at (J = 0. XXIV. etc. 1. Consequently where t is an integer or zero. 4). 4. (i) To find a solution in positive integers. Hence (qc.by = 1 =aq . 1.) Thus the equation has infinitely many positive integral solutions. 2. XXIV. and from the form of the general solution we conclude that (q. 1. The Equation ax-by = c (a prime to b).bp. p) is a solution. and since a is prime to 6. 9). This is called the general solution in integers. .6m. Find the general solution in positive integers of . Therefore a(x~~q) = b(y-p). 1. x-qc must be divisible by b. suppose that (x.. Then since a/6 is an even convergent. x . and let p/q be the convergent immediately preceding a/b.q must be divisible by 6. Hence x = q + bt. /8) where a qc.p) is the only solution in positive integers such that p<a and q<b.a-p). (ii) To find the general solution in integers.

Then aq-bp = \ 9 and the equation may be written ax + by = c(aq -bp) or a(cq-x) = b(y + cp). and let pfq be the last convergent but one.r~l7y = 5: (ii) 13z-17y= -5 .(i) 13. 7 7 . \ b. 76 + 8 = 13 .77) ~\ly -5.3 = 1.. .10=-!. . 5.77 = 14. 13. 1. 2.1) is (3. We find that l^o+J--^^*-!---^-!-! 17 1+3+4 1+3+3+1* The convergents of the continued fractions are 1 3 13 1 3 10 13 f i' I' n . Since a is prime to 6.). 15).. Hence the equation may be written 13 (# .). . (iii) 13*-172/=996. and by the preceding the least positive integral solution is given by x . P P 4' Is' IT and 13. . Similarly for the second equation a solution is given by # = 13. . 2). we find that 996 = 13 . 5 = 65.by c can be made to depend on that of an equation of the same form in which c is less than or equal to the smaller of $a. y 11.5. the reckoning may be shortened thus : Dividing 99 6 by 13 (the smaller of the coefficients). 1. 1. ).2. The Equation ax + by = c (a prime to b). and the general solution in positive integers is 416 NUMBER OF SOLUTIONS In case of the third equation. y = 10. In this way the solution of ax . cq-x must be divisible by 6. 2. (i) General solution in integers. general solution in in tegers is * = 20 + 17*.4-17. Hence for the first equation a solution is (20. The least solution in positive integers (given by t . and the gene ral solution in positive integers is x = 3 -f 17$... y = 15 + 13* (*=0. The. 5=50. and therefore . NOTE. y = 2 + 13* (t 0. Let a/b be expressed as a simple continued fraction with an even number of quotients. Therefore the general solution in positive integers is s = 91+17* f y = ll + 13* (=0.13-17.

. 165).1 </ . and the equation may be writte n 13* + 17^3000(13. ao ab Hence.. . 4 -17. b a ab Now . As in Art. 694.. (219. 1. Use the rule just given. Let m.. Find the positive integral solutions of 13# + 17/ 3000. . therefore /=0. Hence the general solution in integers is x = qc-bt 9 y^at-pc (2 = 0. 9). m + 2. if c is not divisible by ab. HT _..<14.JLV LO Li . _ =706 ^.lit. 4.17 . etc. 705.g < 1 . 3000 4 15 Now. . Ex.-L ao THREE UNKNOWNS 417 Ex. 13 <~ '-. where 0</<1 and then x.. 1. and therefore -. 3) or 13(12000 -*) = 17 (y + 9000) : and the integral solutions are given by x~ 12000 .. y~l3t. = 1 and N--. qc/b~n+g. to show that there are 13 such solutions. 2. 152). 3 = 1. c 15 9000 _ 4 Moreover. we have 13 . Hence if N is the number of solutions in positive integers.). The values of t (if such exist) for which x and y are positive are given by the conditions. also /=Yo ^"i^ so ^ nat /<# hence the number of XO. Thes e solutions are (15.g. ^692 ^ therefore the positive integral solutions are given by = 093.9000. (ii) Positive integral solutions. pc/a<t<qc/b. 4 .. n have positive integral or zero values such that pc/a = m+f.where t is an integer or zero. (32. n. 1. y will be positive integers if t^m + l. If c is divisible by ab 9 then pc/a and qc/b are integers . the number (N) of solutions in positive integers is one of the integers next to c/ab 9 and is greater than or less than c/ab according as f^.

and the general solution in integers of this equation i s M = 4-3t.. may be found as in the next example. Ex.Zij + 3z = 7. 3).. Hence the general solution in integers is x 10 . say in the firs t.. Find all the positive integral solutions of the pair of equations.. v . (2) // (a.. 2. z are positive are and 1. .3y = 4 .. . hence all integral solutions are given by (be') (ca') where g is the G... we find that 2m + 35 = 11 . Two Equations with Three Unknowns.. y.^=3/-2 = 10-9.C.solutions is 13. y = 8 . by Oh. (4)..C.. 1. (ca'). Substituting these values of x. Since 705 -693 + 1 13....... 1) and (1.. this result agrees with the preceding. 418 SINGLE EQUATIONS Now every common factor of these numbers is a factor of their G... then Thus x. 2x . (ca') y (ab f ) is zero. z = l+2.. 6.... 8. j3.9J. z will be integers for all integral values of t if and only if / is a common factor of (bc') t (ca') 9 (ab r )..... Eliminating one of the unknowns (z). we have 2x . y in one of the given equations.b'y + c'z = d' '. 1.. y. if any exist. 6.. (ab 1 ) and ( = 0.6^.M.). Denote each member of (A) by t/f where t may be any integer or zero and / is an integer to be chosen later. and therefore a.... y) is a solution in integers of the two equation ax -f by + cz = d.. (1) The general solution in integers. 2 1+2^. y--2m. (i) // none of the three (be'). therefore the onl y positive integral solutions are (10.M. the equations are equivalent to -= (te'P(ca') (aV) . The only values of t for which x. . it is required to find the general solution in integers. and tho general solution in integers of this equation is #~3w-2. (in~0. and all the positive integral solutions. of (be').. 2.4# + ly + 3z = 1 9.. 1. X. a'x 4. y~2w = 8-6*. 2.

c are prime to each other. z-y are denoted by X. namely x^ 2 t 1. lx + \\y~ 97.y = /3 + cu-aw 9 z where u. of (ca'). ^ y) is an integral solution of ax + by + cz = d (where a. . 2. 5).C. 1) and (7. w have any integral or zero values. one (7. 6. hence z<4. The other possibilities. 5. Find all the positive integral solutions of 3^ + 4x 2 -f 7# 3 + 1 1. no Thus there are two solutions in positive integers. The greatest coefficient is 2(>. the equations are equivalent to and it is easily seen that the general solution in integers is where g l is the G. 3arj. 6. Y. namely (6. (ab').r 4 0 7. GENERAL SOLUTIONS 419 This is obviously true. the . may be treated in the same way. There are 28 solutions in all. x 3 = 4. no solution. we can show that if any two of the three a. 2 = 3. are all positive. For if rr-a. 7. 2-2..2). and we may have ~ 2.M. other solutions are given by xa + bw-cv. then every integral solution can be so expresse d. then all the positive integral solutions (if any exist) may be found as follows : Ex. y~fi. Z respectively. 2 ).(ii) If (6c')=0. 7z + ll?/ = 45. 2. 1.26z = 123. and 7 + 11 +2(>2< 123. c are positive or negative integers). no solution. Further. no 2 = 4. + 4# 2 = 9.r + llr/=:71. 7# + !!?/:= 19. Find the positive integral solutions of lx + 1 1?/ 4. c.. Therefore the only possibilities are 2 1. (2) If (a. Ex. therefore a: 4 <5. A Single Equation of the Form ax + b (1) If a. ^x l + 4 2 = 6.. of which there is one solution (0. 7. v. Here Il:r 4 ^67 ~(3-f 4 + 7). 6.

.. each one foot in length.. 5.aw.. av-bu Z. 7* + 8// = 50. 29* . X. with the scales in contact... and since a is prime to 6.... 2-7.... Y . find the general solution in integers and the least positive integr al values of x.... whatever integral value Z may have.... 12. 2... 13.. AB is scaled off into 13 equal parts. show th at the distance between one scale division of AB and one of CD can never be less than I/ (13 x 17) of a foot. If.130 = 1. If the two rods are placed side by side.equation becomes a-Y + ftY + cZ-O .-41y^l. 17a. and the ends of the rods in alignment. 9. Express 68/77 as the sum of two proper fractions whose denominators are 7 and 11. and express the given fraction in these ways. v can be found so that. and CD into 17 equal parts.cu = . Show that I/ (13x17) can be expressed in two ways as the difference between two proper fractions whose denominators are 13 and 17 . 11. Find the least solution in positive integers of =l . EXERCISE XLII 1... 27a:-8y = 125. 3.. (A) Let a be prime to 6.. lLr + 7y = 151.. 10. The sum of two positive integers is 100. AB and CD are two rods.. 12a?-7y = 211. This proves the statement in question. we must have X + cv == biv. 7.. If one is divided by 7 the remainde .. 4. where w is some integer or zero. 8. 7. y.. Find all the positive integral solutions of equations 8-10. and that there are two cases in which the distance has this value. 6. Which are the pairs of divisions in these cases ? 14. 8z--27?/:--125.. then. (ii) 68s In Exx. Z are integers which satisfy (A). then integers u..

Find the numbers. 5. he paid a bill of 9s. and if the other is divided by 9 the remainder is 7. A man has in his pocket three half-crowns.S. A certain set of stamps is made up of four different sorts : a collector has one of each sort and some duplicates. respectively.. leaves the remainders 1. 23. 9x+I6y + 25z = IQ9. and the lot are worth 3s. The values of the stamps are 2d.. If c is increased by kab. Ix + lly + 133 = 201.] 17.. 19. Find all the different wa ys in which the batch could have been made up. The number of solutions in positive integers of the equation x + by~c is the integral part of c/b. seven other silver coins which are either shillings or sixpences. 15. x + 3*/-4z. 20. Is. 2681. 18. The least value of c for which the equation ax + by c has N solutions is given by c^a + b + (N -l)ab. 24. 7d. 9d. Find the possible number of different ways of doing so. 11. 22. 9d. 15.. Find the least number which when divided by 7.r is 1. . Find the least value of c in order that the equation l%x + Yty c may have 13 solutions in positive integers. 9. coupled with the fact that a + b is the least value of c for which the equation has one solution. Find also the general form of all such numbers .] 420 SOLUTIONS IN POSITIVE INTEGERS 16.\ 21. showing that the required value is 2682. 20. respectively . and eightpence in coppers . 25. a. then the number of positive integral solutions of ax + by=c is increased by k. g are unaltered. for which the equation 3x + y a has six positive integral solutions. Verify by finding the number of solutions for c-2682. [For c/ab is increased by k and/. 3d. Find all the positive integral solutions of equations 20-23. Find the least positive integer. 26. 7z = 209. [This follows from Ex.

where the symbol == is an abbreviation for ' is congruent with. x 2 =p 2 + cu . then a and b are said to be congruent with respect to the modulus n. we have a = r (mod n) and 0^r<w . If a.cw' 9 where u 9 v. 29. // a^b(modn) and a f ^b'(modn).aw. unless both b and b' are divisible by ab' . where q may be positive or negative. For example. b are any numbers.dv' . [For a(bc') + b(ca') + c(ab') = Q and a'( CHAPTER XXVI THEORY OF NUMBERS (2) 1 . Any statement of this kind is called a congruence. . positive or negative. and the absolute least residues are . Congruence.^n. This is expressed by writing a ^b (mod n) or a 6 = (mod n). (iii) aa'^bb'. such that a ~~ b is divisible by n. w. Every residue of a to the modulus n is of the form a -f qn. Hence we can always find r so that a = r'(modw) and |/|<^n. a'x + b'y + c' Q and x is an integer. How many solutions are there to this problem ? 28. thus r is the least positive residue of a. and either r or n .r) (mod n). (1) Let n be any positive whole number. and each is said to be a residue of the other. This number r' is called the absolute least residue of a.(b + b') and (a .b' are divisible by n. XIPI + an' + bv' . (2) Fundamental Theorems. 34 == 4 s .(n . a number prime to n. Thus the least positive residues of 34 and -34 are 4 and 1. If (p l9 2h> P& P*) i g an integral solution of ax l + bx z ^-cx^ + dx^ e 9 show that other solutions are given by x i ~P\ + bw -cv du'. Again. If ax + by + c~Q.27.a') -(b-b f ) are divisible by n. and in particular uc^bc. prove that y is an integer.r^.34 = -4 = 1. then a/d = b/d.'' When no doubt exists as to the modulus in question. if 5 is the modulus. u' 9 v' 9 w' have any integral or zero values. a == r = .1 and 1 respectively.b and a' . #3 ~p3 + dw' + av ~ bu.1 and . shillings and pence. which we shall call the modulus. If r is the remainder when a is divided by n. For by hypothesis a . Find the sum of money which is expressed in farthings by the same digits written in the same order as those which are required to express it in pounds. where r' may be positive or negative. (iv) // a and b are divisible by d. hence (a + a) . then (i) a -fa' ^6 + 6'.a'b. we simply write a E=&. (ii) a -&==#' -6'.

. I..z..y.. . i t 2 + .. . and if a number is prime to n. taken in a certain order (Ch. 11 or 13. a . 11. .)=f(x. z. where q is a whole number. therefore aa' -66' is divisible by n. Thus a close analogy exists between ' congruences to a prime modulus ' and * equalities/ 422 TEST OF DIVISIBILITY (3) The following are immediate consequences of the theorems in (2) : (i) If to any modulus as=a'. &==a'6' .W = (a . and gp(12) = 4. . k~k f . so also is its remainder. therefore 12345671 is divisible by 13 but not by 7 or 11. then we have J + l =1001 =7 ..6 = qn. 1. Let s-a -a 1 + rt 2 -.. then afdz=b/d(modp). 10).) is a polynomial in x. Now d is prime to n. 2. the remainders are 0. + (-l) n a n .. 11 or 13). Consequently. y. 2.. . with integral coefficien ts and x = x'. 11 . we define the value of <>(1) as 1. 5. 2... Consequently N~s (mod 7. x-i-a.). 7. 1.Again. 11 and 13.. Thus if 2V = 12345671. This proves the theorems (i)-(iii). &'.. then 5 = 671-345 + 12 = 338.y'.. In particular. . so also is qn. The Numbers less than a Given Number and prime to it.. h ence t~ -1 and t r =(-l) r . . . to any modulus. if a == 6 (mod p) wAere p is a prime and if d is any common divisor of a and b. Now 338 is divisible by 13 but not by 7 or 11 . For if these numbers are divided by n. and in particular a m z=a' m . y^-y'. 13. Here. Any number N may be written in the form N=a +a l t + a. Ex. (1) If a is prime to n. Thus the numbers less than 12 and prime to it are 1. n> it follows that a/d = b/d (mod ri).. w-1. as many terms in the progression are prime to n as there are numbers less than n and prime to it. +a n t n where f^lOOO and 0^a r <t. If n is any number greater than 1. except when d^Q(modp).6') 6. therefore q/d is a whole number. then to the same modulus f(x. x + (n-l)a which are prime to n is <p(ri). with regard to each of the moduli 7. then to the same modulus ab .. Give a test as to the divisibility of a number by 7. x + 2a. the number of terms of the arithmetical progression x.. Lastly.6)a' 4. it should be noted that 1 is to be regarded as prime to every other number.. then since a and 6 are divisible by d.(a' .. 6 = 6'. z^z...z... (ii) // f(x\ y. and since a/d-b/d = q/d . the number of positive integers less than n and prime to it is denoted by <p(n) : moreover. aa' ..

then q>(p r ) p r (l-l/p)... m + k . and their number is p r -p r ~ l y which proves the theorem. For by (2). and their number is p r ~ l . every column contains cp(n) numbers which are also prime to ft.... <p(m r ). according as k is or is not prime to m.. and so cp(mn) = <p(m) . To enumerate the numbers less than mn and prime to it.. . 2p. of the numbers 1. (n-l)m+k. It follows that there are cp (m) . VALUES OF p(n) 423 Thus the numbers in question.. . If r>l..(2) If m is prime to n.. where p. Hence and so on.. (3) // m v m 2 . w r ) = 9?(%) .. for the numbers in each column form an arithmetical progression of n terms with a common difference m which is prime to n. . <p(n). 3m (n-l)w-t-l (n-l)w-f2. (4) If p is a prime. then cp(mri) (p(m) . w 3 .. <p(m z ) . nm. (5) If n=p a . since p is a prime. are the prime factors of n. 2m + 2 .. m r are prime to one another.. r .. .. <p(m 2 ) . (f>(n). mn in n rows and m columns as below : 1 . 2 . 2m 2m-fl . r. therefore it is prime to m 1 m 2 . 3. (p(m l m 2 ) = (p(m l ) . p r . The proposition is true when r = l.. Now every member of the &th column is or is not prime to m... m-f2 .. then p/\ q/ .. Moreover. then cp^m^m^ .... <f .. 3. the numbers in question are those which are prime to both m and n. for any number of steps.... are members of the 99 (w) columns headed by numbers prime to m. Since m is prime to n.. cp(ri) numbers less than mn and prime to both m and n.. Also m 3 is prime to both m^ and w 2 ... p r . All the rest are prime to p r .. k . for cp(p)=p-I.. we arrange the numbers 1. 2m + k . those which are not prime to p r are p. 2. 2.. m w-hl . 3p. q... <p(m 2 ).. that is to mn. being prime to w.

For instance. Q^y^b.. If n>2 and x is any number less than n and prime to it. the sum of the integers less than n and prime to it is \nq>(n). 1026 = 2. when w = 1024. then every divisor of n is of the form p x qVr z ... Hence the numbers less than n and prime to it can be arranged in pairs such that each pair has a sum n. Fermat's Theorem. Let n^p^r . the number of pairs is fyp(ri) and the sum of the numbers in question is 424 FERMATS THEOREM (6) Theorem..p fl . and the result follows from (3) and (4). 19 and <p( 1026) = 1026(1 --|)(1 .) = 27{y(p*) . ..=n. Ex.q*> . 3. . (f...}. // p is a prime and a is any number prime . so also is n . where Q^x^a. d 3 . 0<z<c. Hence S is equal to the product Now 1 + (jp(p) + cp(p 2 ) -f ... are primes and let S = <p(d l ) + y(d 2 ) + ...r . where x. We have 1024=2 10 . and 3.x. + (p(d r ) = n...J)(l -j^)=324. 41 and 9? (1025) = 1025(1 -J)(l -^-)=800. then <p(dj + y(d 2 ) + 93(^3) + . hence (p (1024) = 1024 (1 -|) =512. 1026. 2. 24 ... q y r. z. 12. y.For p a .1 )= and similarly for the other series in the brackets.. 2. etc.. have all values subject to the above conditions. 8. Find the number of integers less than n and prime to it. 1 . This illustrates the irregularity in the variation of the function q>(ri). Ex.1025.. . 99(7") . d r (=-n) are the divisors of any number n.. Therefore S-=p a . 6.. 4. This is obviously true when n 2.. where p.(jp(p a ) = l+(p-l) + (p a -ji) + . . 3 3 . .. // n^2. the divisors of 24 are 1.. Thus <p(n) must be even.+(^.. r . 1025 = 5*. rf 2 . // ^( = 1).. <p(r*) .. so that S = Zcp(p*qyr* . <p(d r ). 4. are prime to one another.

2a. The fifth power of any number N has the same right-hand digit as N... For a p .a is divisible by p. less than n and prime to it. which is prime to n. therefore p is a divisor of a^^~ 1 ^ .. .. 3a..... the remainders are 1. Corollary 1. For since a is prime to p. a 2 . . Hence the remainders are the numbers a v a 2 . Now the product of the numbers in the set (A) is congruent (mod n) with the product of the remainders. (p . which proves the theorem. . 4.1 is divisible by p y which is a prime number... and consider the products aa l9 aa 2 . 2. Also the product of the numbers a. for the factors of any product are both prime to n.l)a are divided by p... a<p (w) ^a^ . For a p -a = a(a p "" 1 -l). then a p -a is divisible by p. a x a 2 . therefore a p ~ l \p-I = \p-I. For if we suppose that two products as aa r . then a p ~ l 1 is divisible by p. 2 we have N 9 ^l(mod 19). then a(a r -a s ) would be divisible by n. If p is a prime and a is any number whatever. in ascending order. 19m 1. if a. Otherwise we must have N = 19m. and since \p 1 is prime to p. Ex.to p. is congruent (mod p) with the product of the remainders. If n is any number and a is prime to n. Euler's Extension of Fermat's Theorem. therefore a^^ ^mod^). by Cor.. for a is prime to n and a r a s <in.. Also the remainders are all prime to n. 3. aa 9(n} (A) If these are divided by n. p . a 3 .. the remainders are all different.1 -! is divisible by p. and if a is prime to p. we have the result in question.1 taken in a certain order. a<p( n ). Let ^( = 1). a<p( n ) be the numbers. a 9 ( n ) taken in some order or other. Corollary 2. a p ~~ l .1 )-l)(a*^ I > + l). then afW ^ 1 (morf n). WILSON'S THEOREM 425 Ex. and dividing by a x a 2 .1 or of a^/>-i) + 1 ? since p is a prime number. a 9(n) (mod n). and a". therefore a*( n > . 1. since \ r(19 -1)=9. . then a i (11-1)52 l(modp). If p is an odd prime and a is prime to p. 3a. Hence in all cases a p . If N is prime to 19. aa^ . 2.1 -l = (a*<i.. . This is impossible.. aa s (r>s) give the same remainder. For A 75 -N is divisible by 5 and by 2. The ninth power of any number N is of one of the forms 19w. . 2a..

| 28 4.. 3....1 ) . a' are called associated residues.. Hence. Wilson's Theorem. For otherwise p would have a factor q which would divide [p-1 and [p-1 + 1. | 28 + 1-0 (mod 29). the sum of the products of tJie numbers 1. The numbers a. If p is a prime number. if \ p . .. 2. 2.1 4. Hence 128 4-233 is divisible by 29 and 31. therefore |j|8 + 233=0 (mod 29).. 2a.. .. .29. Again.. then a 2 -! must be divisible by p. and theref ore by 899. p 1. 3 . We have 899 = 29. . (x 4.. and therefore [28 4-233=0.. a. Lagrange's Theorem. it follows that [28 416=0. 2. such that the product of each pair is congruent with 1.. Let f(x) = (x 4~ 1 ) (x 4... If then a is any one of the p ..1 + 1 ^0 (mod p).1 are the only values of a for which a' can be equal to a.2) .p . it follows that either a + \p or a -1=0. there aa' =zl(modp).. 3.1 is divisible by p 9 then p is t3 prime number. to the modulus 31. >~1. such that 3. f(x) = tf>. . for every a.. which is the result in question. NOTE. p-1 taken r together is divisible by p.. p . For if a is any one of the numbers 1.. and if the products p. .2..2. . each being the associate of the other.1 taken r together.2) | 28 4-32=0.+a P _ l .. If a'=a. 1.3 numbers 2. 3a. 3..31.. Prove that | 28 4-233 is divisible by 899. and so ( ... Consequently.2 + a&*-* + . Thus 1 and p . .l +a l x*.1)( . Ex. then a' is not equal to a.. (p-l)a are divided by 1. p . therefore 30.. these numbers can be arranged in i(p-3) pairs..1 ==j?.. 3. Hence \p .1 = 0. then if a r denotes the sum of the products of 1. By Wilson's theorem. that is. and since p is a prime and a<p. we have 1 30 + 1 = 0.. Therefore 2 . If p is a prime and r is any number less than p-I.1 (mod^). (p~2)~l (mod p) and \p. then \p-l +1 is divisible by p. ... (Euler.5. also 233-1 (mod 29) and 233= 10 (mod 31). taken in some order or is one number a' and one only. the remainders are the numbers other. (A) Also we have the identity (x+p)f(x) = (x + I)f(x + l)..) 426 LAGRANGE'S THEOREM Conversely. .. 6. Since 2 is prime to 31. 1 .

(B) Equating the coefficients of x p ~ 2 . N6w. and CJ?" 1 .1 -1=0 (mod p) 9 which is Format's theorem. . aj. If p is awrime. hence f(x)=0 (mod p). a 3) _ 2 is divisible by p*\ and.1 must be divisible by p.. z p ~ 3 ... EXAMPLES OF DIVISIBILITY 427 Ex... . (2p -l)(2p -2) .. and since a v a 2 ._ 2 ar e all divisible by p.. We can immediately deduce the theorems of Wilson and Fermat. it is divisible by p. we put x equal to p and . x+p . x 4. in the identity (A). 7> 2 -f . (ii) If x is any number prime to p 9 one of the numbers x 4-1. greater than 3. . since p is a prime greater than 3. (6).. For another proof. that is \p -1 4-1 is divisible by p t which is Wilson's theorem. for r<p .I m .. thus : (i) Equating the terms independent of re in the identity (B)..2. for all integral values of m . ( I p\m is divisible by p m ~\ 3 . .1) j)~i-f ap^ 2 P+^-s . . greater than 3. since p is a prime. 2. and. see Art. a/p-s exists . prove that.. and since the left-hand sides are identical. (p + l)~2>_i ~%_2 2. 11.1 (mod p) 9 therefore x 3 *. Hence.. we have (p + l)(p + 2) (2p . we find that Now. which is the theorem in question. Also /(#)= x p ~ l 4-ap^ x p ~ l . 1 mp . 3_p .. ~ . Cf is divisible by p. a p _ 2 are divisible by p 9 so also is a^^ + 1. prove that '1 1 1 ^2 + '" p~-} If. It follows in succession that a l9 a 2 .2p in succession. x. If p is a prime. NOTE. by Lagrange's theore m. 1. Cf ~ 2 > etc. it follows that a^_ 2 * s divisible by p 2 . Ex. we have by subtraction. . are all prime to p. since 3 is prime to p.

1 is divisible by 504.2.. Consequently. and.1) is a solution of the congruence. a' 2 =s .. If p and q are different primes. (2m -f l)(mod^). 10. (4m -2) .1 = ( . and S^aQ-^^a^ .) ~ r + 1 (mod jt? 3 ) . show that and in particular (ax) p ~ax p (mod p). 2..1. Apply it to 30414 and 8110 3. X Q . 1. 2.I) n t n 9 where Z~ 10000 and Q^a r <t.1 = (mod p). 11. . 2. If p is a prime of the form 4m + 1. | (r -f. by giving r the values 1.l)fc ^ p ~^ . 2 \p -3 + 1 is divisible by p. as in Ex.1 (modp).. If N ao + a l t + a 2 t* + . Show that 118 + 1 is divisible by 437. a p ^= a (mod p) 9 and if a is not divisible by p. I /. and p is prime and greater than 3.Substituting rp for x in the identity (A). has no solution. then p q ~ l -f q p ~ l .j 2 E B. then n 12 . [Show. show that | \(p . If p is a prime. x 2 . 4. 428 ROOTS OF A CONGRUENCE 7. then n 6 . if n is a prime greater than 7. m .. .1) is a solution of the congruence. and (ii) that p is of the form 4m ..+ T h=0(inod7i). If p is a prime and x l9 x 2 . If n is any odd number.A.j ra(| p\ m =0 (mod p m + s ). show that I %(p . Hence.. where r ~l. EXERCISE XLIII 1. .1.1 is divisible by pq. 9.C. we have.+a n t n . . x 2 + 1 = (mod p). If p is a prime of the form 4m . it follows that hence. show that \n-\\ I+S + QT--. 3. Prove that..3 2ms (4m) . -f ( . 5.. x a are any numbers.1) p I ( I rp . . Hence state a rule for the divisibility of a number by 73 or 137.1. Prove that. m . show that (i) \p ..that 1.1. . (4m -1) .1 is divisible by 65520. where x is any number.. If the congruence. prove that J\ T ~/S(mod 73 or 137). I mp . a p ~~ 1 == I ( mod p) 9 which is Fermat's theorem. 3. 8.. if n is a prime greater than 13. 6..

4. . If # 2 is the sum of the squares. which is said to be of the rth degree. [Use the last example to show that the sum of the squares of the numbers not greater than n and not prime to it is 3 U \ a~ ab + '") + 2 n + 6^ a ~ a+ ~->-l 7. if p is a prime. 2a6.. 2-.. 26... . Roots of a Congruence. with integral coefficients.1. .. such that for each pair. respectively... . If s l9 s 2 . . If x Q is a root of the congruence f(x) ^0(mod n). be f ( B ) together with groups (C). 1. (D). of the numbe rs less than n and prime to it. . 7i/a .1 (modp) are 1. 13. 26c.. together. 36c. the roots of the congruence x^" 1 . show that where a. (C). show that t he sum of the rth powers of the numbers not greater than n and not prime to it is 14. .. . Let n~a p b Q c r .. 2. taken 3. 5 3> denote the sums of the rth powers of the numbers in the groups (A).... 6.I I Ad 11 l_j 12.. 3. etc.. 3a. Ex.. c. a and a'. c.. 3a6. ( A ) 6c. Arrange the numbers 2. 6. . In other words. c.. n/bc . are the different prime factors of n. 6. formed in a similar way from the combinatio ns of a. any value of x which satisfies the condition f(x)=Q(modn) is called a solution or root of the congruence f(x) =0. and conside r the groups a. E\ r ery square number is of the form 5n. THE LINEAR CONGRUENCE 429 Ex. n/ab . . If f(x) is a polynomial of the rth degree in x. 36. a 1 a&. b f . 2a. .... 4. aa f ~\ (mod 17). 15 in pairs. a6 1 6. the congruence x z ==2(mod5) . (B). and $ 3 the sum of the cubes.. and when we say that a congruence has r roots we mean that it has r distinct roots. where a. p . . ... It follows from Format's theorem that. 3. n/6 .. or 5wl> thus no square can be congruent with 2 to the modulus 5. Two roots will be regarded as distinct only when they are incongruent with respect to the modulus. so also is any number congruent with x to the modulus n. are different primes.

there are g incongruent solutions of ax^b(mod n).1) a are divided by n. No such value exists unless 6 is > divisible by g. then a'x-b' is divisible by n' and the given congruence is equivalent to a'xs-fe'fmod n'). To solve ax = b(mod n) when a is prime to n.b is divisible by gn'. let afn be expressed as a simple continued fraction with an even number of quotients. Since a' is prime to ri. NOTE. a. the expressions a/6 (mod n) and a7&'(modn) are equivalent. it is generally easier to proceed as in Exx. 2a.. . then 6/a(mod n) has no meaning : in every other case it has infinitely many values. . in some order or other.ny Q = 1 . or the product of small primes. and it is easy to prove the following : (i) If a = #' and bz=b'(modn). since a is prime to n. we may replace any coefficient by any number congruent with it to the modulus n. Hence there is just one value of x such that ax^b(modn) and 0^x<n. oc. The Linear Congruence.. these values are all congruent (mod n) : if a is not prime to n. If a and n have a common divisor which does not divide 6. Thus t he congruence 13x= 17 (mod 5) is identical with 3#=2 or with 3a. For if the terms of the arithmetical progression 0. and g is the greatest common divisor of these numbers. (3) The expression 6/a(mod n) is used to denote any solution of ax = 6 (mod n).. n-1. so that a' is prime to ri. (n . which are the g distinct roots of a# = 6(mod n). the remainders are the numbers 0..= -3 or with z= . and since a is prime to n. there is no solution. and let y /x g be the last convergent but one. provided that b is divisible by g. the last congruence has one solution a<n'. a + ((7 l)n'.1 (mod 5). n=gri.. the congruence ax H= b(mod n) has just one distinct root. a + 2n'.has no solution. 2. they are all congruent (mod n/g) where g is the greatest common divisor of a and n. Therefore ax E= 6 = abx Q . which is the required solution. 430 SIMULTANEOUS CONGRUENCES Such expressions possess many of the properties of ordinary fractions. If this condition is satisfied and b=gb'. in dealing with the congruence /(#)=0 (mod n). (2) // a is not prime to n.. We require values of x such that ga'x . namely. 1. . If b is not divisible by g. (1) If a is prime to n. . 1-3 below. For let a~ga' . and (/solutions <w. (ii) am/bm(mod mn) is equivalent to a/6 (mod n). If a is small. a-t-n'. Then ax . x^=bx Q . so that ax Q == 1 (mod n). 8. If a is prime to n. It is obvious that.

so that z==6(mod 7) and x~G.17 . 3. (i) It is required to find x so that x-a(moda) and x==b(mod /?). 5(mod 43) and &(mod 43 )~ ~ 5 ' Therefore x= -\^. where I is the L. 20 (mod 21).55= 31 (mod 43). and then the congruences are equivalent to the single congruence x = x (mod I). we replace the first two by x = x 1 (modi!) as in the last section. 1. which satisfies the last congruence.50. 13. of a.. y. Let ^ be the greatest common divisor of a and j8. Otherwise. t where x t = a + at/j. a solution always exists .M.a is divisible by g.(mod 43)~ . O o IA O O Ex. the congruence is equivalent to 5#=2(mod 7). the first three congruences are equivalent to x = x 2 (mod m).. where x 2 is any solution (found as before) and m is the L.~ 2 4. (ii) Since 3 is a divisor of 15.M. therefore #~ . of a and j8. 6 . 6. THEOREM ON FRACTIONS 431 and therefore of a. Thus a solution ^ of the given congruences exists if.C. j8. of I and y. therefore as -Hf?= ^s ~ == 22 (mod 157). (ii) To find x so as to satisfy a number of relations of tJie form xz=a(mod<x. Ex. We have -).l==6(mod 7). 5~ . Continuing thus..). the greatest common divisor of a. It should be observed that solutions always exist when a is prime to /?.(iii) ak/bk(mod n) is equivalent to a/b (mod n).. y. (i) Hero 6#==2-7== -5 (mod 7). Moreover. and only if. it will be seen that the given congruences are together equivalent to xz==g(mod L) where is any solution and L is the L. If 6 .C. x = b(mod j8).a is not divisible by g there is no solution. Taking this with the third. t. 2. j8. if k is prime to n. Solve 36* = 7 (tnod 157). x^c(mody). (4) Simultaneous Congruences. We have 12*== ?= ^|?== . 21. Solve (i) 5ar=2 (wod 7) ..C.17 . . so that the general solution is x = x l + afilg . there is just one value y l of y less than /?/</. thus 1 =43 . and the general value of y is given by y = y l + f}/g . (ii) ISzssG (mod 21).M. We have x = a +ocy where y is given by a + ocy = 6 (mod j8) or a2/ = fe~a(mod j8). Sotoe 1 7z= 1 1 (mod 43).. /?. 2 .and the third convergent is 5/2. Ex. . ..

. 3. :==-.. = 7 + ...... 2. giving y^-Wmod 12) =s-. any value of x which satisfies the se conditions is a solution. In this case the solution can be found more easily by the method of (3).(mod 4) si (mod 4). The fourth convergent is ~ . If n is the product of factors a.. the fraction m/n can be expressed uniquely in the form m a B y A 7 . We have 140 4 . (5) We can apply these methods to solve a linear congruence when the modulus is a composite number... Thus 20 and x = 59 -f 140J.~ (mod 12) ... 1. 8 = 1. c 9 d. thus _. = 5 + 9ys 2 (mod 12). .9 (1 + 4$) 14 + 362.. Ex.1 +4t/== . Driven that 35x^4 (mod 9) and 55x=2(mod 12)..l(mod 4).. 6.l(mod 5). .. .1 -f 202= .. 1<7 2 + 1 ~\r 2* 4" 2 o Therefore x~ ~ (mod 140) == 59 (mod 140).= . which are prime to one another.4 (mod 7) and z = -^r. find the general value of x.+ y + + .if <*> A y. . Ex. and x=. 19 .2 (mod 12)... Therefore x = . 17 1 1 t/ 1 8 and z==y-(mod 7) = -=-4 (mod 7)... We have x=~ (mod 9)==-^r== 5 (mod 9).+ 7 fc. Z.1 (mod 5)..(mod 5)= ...140 . oo 1 &o 3 1 Hence a. as in the next example.1 -. so that y = 5z. Thus we have #== (mod 4)== . A Theorem on Fractions. y o Thus y = l+4t and a. A) ... Hence x= . = 5 4. Ex. and 59 ... which is the general solution.. Solve 19*== 1 (mod 140).. 19 9...(mod 7)= 3 (mod 7).= ....... and m is prime to n... 5 and 7 : and since these numbers are prime to one another. and therefore 19#==1 for the moduli 4.5.7. are prime to one another. which is the general solution..

A. Again. 'j8. Let a' bed . Express Yr~~7\ ^ e f orm . we can obtain a relation of the form m x u z w ... it will be seen that a . ac . (-3)(-2) .J ___ L _..1)( .I j ~ I .. we find that i + f +y+^r = l^^. 1 fractions and k an integer. ( . 1 ly = 101 (mod 5)...1) . = 1. ( . z. 4?t s 1= 12.. w == 3( mod 11). . 4. where a. it = 3. ZEE! (mod 7). which is prime to a'._ __ _i. x' can be found so that a'x + ax' = m. w_x ' 1 -I OUJ. . y^/3 + kJ). 4 .4^-3. he . ?/. for any common factor of these two would divide m.. u are determined by 5. } . so that jfe=-l. i! is prime to a. y. 5 . .+ -+. .VA T 7 :i 7 aa a a n a bed . . n a b c I where x.7. etc. 4. Similarly j8 has just one value less than 6.I 432 THE GENERAL CONGRUENCE Proceeding in this way. lu ~ 101 (mod 11). if we multiply each side of the equality (A) by n. with regard to the moduli a. y^2(mod5)..5 .. j8 .l)x = 1. j . )8. k are positive integers and a<a.+ A% the fraction* being positive p roper 4 O / J. I. and then mfn can be put in the required form by writing x^a + frjcr. Here x. 2 . Moreover. Thus we have m _x x f . Z^m(moda).. = w(mod 6). etc. etc....n a b c I v ' where a. jS<6. 1 . 1 ... Since 6c. 6. a? = 1 (mod 4). w are integers. y = 2.. a has just one positive value less than a.. . then a is prime to a'. Ex. y. Ilz=sl01(mod7). z. .7. liars 101 (mod 4). are the least positive residues of x. x' is prime to a'.+ .. respectively . Hence the expression found for m/n is unique. and therefore integers x. With a? = l. and so on. y == 1 -4.

97(3). In this article /(re). a polynomial of lower degree than /(x). we obtain an identical relation of the form whore the remainder R'(x) is of lower degree than <p(x). 4. This is expressed by writing Again.2 and x 4 in succession is 12 5+342 shown on the right. . in which the modulus p is a prime. 1.2~ 5(x -2)(x -4) (x + l)(mod 7). and leads to an identical congruence of the form f(x) = Q(x) . Suppose that any coefficient which occurs in the division is replaced by any number congruent with it (mod n). Prove that 2. (1) // MN-O(modp) where p is a prime. Division (mod 7) by x . (2) Division (mod n). CONGRUENCES 433 // f(x) is divisible (mod n) by (#-a) r . In the last line 19 is replaced by 5 \ __4 5-1+1 + Q and 21 by 0.. If /(x) is divided by <p(x). (1) If every coefficient of a polynomial /(x) is divisible by n. . Ex. The analogy between congruences and equalities is carried a step further by the following theorems.1 and -6 by + 1. The process modified in this way is called division (mod n). For since MN is divisible by the prime p. If 72 (x) ^0 (mod w). which is also written as /(x) H^ 99 (x) (mod n).4x . The reckoning shows that 20 + 9 or* + 3z 2 .. Congruences to a Prime Modulus.<p(x) r= 0(mod n). so that the congruence holds when x ~2. 11.2===0(mod 7). In the third line 13 is replaced by 10-2+2 . where R(x) is of lower degree than <p(x). or N ^Q(mod p). 4 or 6. The General Congruence. and we say that f(x) is identically congruent with zero to the modulus n. then /(x) is divisible by n for all values of x.10. either M or N is so divisible. /(x) is said to be identically congruent with (p(x)(modri) if /(x) . then either M = Q(modp).\x . the congruence f(x)==Q(modn) is said to have r roots equal to a. (p(x)+R(x)(modn). then f(x) is said to be divisible (mod n) by (p(x). G are roots of 5x* + 3# 2 . will denote polynomials with integral coefficients.

.F. with the modification that any coefficient may be replaced by any number congruent with it (mod p) . 2. Solve Here 4z 2 4-43.C.{ I p ..(2) If a is a root of f(x)z=Q(modp).1.r-f IE= 2. For any common divisor of x ..) r (x .1 -f 1} ^ 0(m od p).. Putting x=oc we have 72= therefore /(.. a r . then the congruence (p(x)^0(mod p) has r roots. p . hence every solution of /(#)=EO(mod^) is a solution of x p ~ l -I ==0(modjp).. (5) // f(x)'^Q(modp) is a congruence of the n-th degree with n roots and <p(x) is a factor of 'f(x) of degree r./? which is independent of x. 3. (3) Iff(x) is divisible (mod p) by (x~a) r and by (x-fly where a then it is divisible (mod p) by (x . a r .C.<y. By a modified H. 4.. 2 x .ft)* . thus completing the analogy referred to at the head of this article. process be carried out for the functions f(x) and x p ~ l . 3. 3.. This follows at once from the preceding theorems.. . . (x a) -f R. Let the H. Hence a r ^0(mody) for r<p-] 9 which is Lagrange's theorem (Art. 1.a p _.. . or (2^-fl) 2 ^ -3s 4. (4) If f(r) is a polynomial of degree n. 6) . then f(x) is divisible (mod p) by x-a.1. then the last divisor R(x) is the highest common di visor ( mod p) of f(x) and x p ~ l -l. (6) // p is a prime. Let /(a. also I p . Let the roots of the congruence be oc v 2 . 12.. Hence x~ 2 or 4 (mo d 7). . 2.. This is an immediate consequence of (4). p . .1 .1 .. Reduction of a Congruence. 434 ROOTS OF CONGRUENCES In treatises on the Theory of Numbers the notion of congruence is extended to complex numbers.1 + I =0(mod p). 2.-f4==0.1 =0(mod p) are 1.3 4. Ex.r) is divisible (mod p) by x-oc.)^0(mod p) be a congruence of the nth degree to a prime modulus p. . process we can find a congruence R(x) ^0(mod p) of degree r of which the roots are ot v <x 2 .a and x /? is a divisor of a . where some of these may be multiple roots. any root which occurs r times being counted as r distinct roots.. the congruence f(x)^Q(rnod p) cannot have more than n roots. For the roots of x**.a 2 ar p . where R is independent of x. it follows that a^".F. and2. p .1. it follows from Fermafs theorem that the roots of the congruence x p ~ l ~ 1 ^0(mod p) are 1. . which is Wilson's theorem. For by di vision (mod p) we have f(x) ^Q(x) . and therefore Hence if a r denotes the sum of the products r together of 1 .

Solve (i) 16 ^^ 31 (mod 1217) .2# 3 .F. Ex. it is found that the last divisor is x + 2. process for the left-hand side and re 6 -!. (ii) 2:c==3 (mod 7) and 3*^5 (mod 11) . BICYCLE-GEAR PROBLEM 435 . show that a -b is even and 5. Now afe3{mod7) has no solution.286a + 3646 . if (i) 29x= 1 (mod 13) . 3z=4 (mod 11). Hence the common divisors (mod p) of u and v are the same as those of v and w.1 == 3 (x + 2) 2 ( x~ .x . SJunv that the congruence 3# 4 -2x* ~4x 2 -x-l==Q(inod 7) has only one distinct root and solve it.385a + 1766 . w au + bv where neither a nor 6 is divisible by p. 2. 1. EXERCISE XLIV 1. If #= a (mod 22) and x =: b (mod 40).77c (mod 1001). Solve (i) 78z= 1 (mod 179) . prove that z== . Let u. prove that x. 3. If x~a (mod 16) b (mod 5)==c (mod 11). It follows that R(x)^0(modp) is the congruence referred to above. hence -2 is the only distinct root.= 13 (mod 179). (iv) 26#=35 (mod 1901). By division (mod 7) we find that 3z 4 . and 7z~3 (mod 20). 6. (iii) 8z= 1 (mod 7). v be any two consecutive remainders in the process. If x==a (mod 7)== 6 (mod 11) c (mod 13). If w is the next remainder.3) (mod 7). showing that . (H) 30#=~31 (mod 1861) .2 is the only distinct r oot.560c (mod 880). 4.4z 2 . (ii) 78a. Carrying out the modified H. Find the form of x.C.The truth of this statement depends on the following. (iii) 15*= 28 (mod 1009) .

15c (mod 105). z. where x. . Express 1001 in the form 7 2<13.35 4. prove that x== .216 4. 2 < 1 1 . k are positive 9. z. If x = a (mod 3)^6 (mod 5)=== c (mod 7). 8. 11 k.7. 2/ . O 11 where x. y.4.~ k. ?/ < 5. Express -1 / OU in the form integers and < 32. k are positive integers and #<7. y.

4. find the form of n. second and third powers are of the forms 3^4-1. 'Prove that the congruence 3# 4 . of which the first.4 and 12m. If n 2 and (n + l) 2 are of the forms llm 4.3# .4s 3 . A bicycle with no free wheel has 19 teeth in the back chain. and find them. 12.^ c 17. Find four consecutive numbers divisible by 5. Prove that the congruence incongruent solutions. (mod 11) has four ~0 (mod 29) has only one distinct .. If ?i 2 . and the circumference of the back wheel of the bicycle is known to be almost exactly 7 feet. 9. 13. 14. and find them. 121 links in the chain. Find tfie form of numbers. Prove the following rule to find p IG> ^ . find the form of n.10. 18. Bicycle gear as a revolution counter.4. and find it. If # 2 4-4oH-2 is divisible by 23. and n 2 ~ 3 is divisible by 11.2 is divisible by 7.a.4^ (mod 7) has only tw o incongruent solutions..wheel. Prove that the congruence solution. 11. 2 4. 11 respectively. 16. 4^4-1 and 5>i4-l respectively. find the form of x. 15. 7. 62 teeth in the front chain.wheel.

~g 2 . 2 1 . [If x is the number of revolutions of the back wheel. (v) If n-l occurs as a residue. are equal. 0.1. and conversely. Denote the residues by r . (ii) If r M = l.15.. and so r t r Q .. the least positive residues of successive terms of the geometrical progression a. 9. r^ n ^ at least two. 5.1. Having found r 4 = 16.wheel. Mark a tooth .4 in the back chain. w-r 3 ..2.7. 1 .13.16. then /z = 0(modJ). At the finish count these again. etc.the number of revolutions (R) of the back wheel (and so the distance travelled) in any journey of not more than 10 miles.J CHAPTER XXVII RESIDUES OF POWERS OF A NUMBER. Residue. etc.21086) 7502. of the numbers r . and conversely.. ag~.10.= 3025a -f 21086(mod 7502).2.. and therefore u.17. . and if t is the number of terms in the recurring period. Hence. 4..18. 19-4. It follows that ag t ~a 9 ag t+l ^ag.12.13. -g 3 . . etc. and denote it by 6. so is every residue. then /A==v(mod J). r l9 r 2 . hence ^* = l(modn). Thus for the modulus 19..4. a tooth B in the front chain.. etc. Hence show that #==49a(mod 62) and #=51&(mod 121). for they are congruent (modn) with ~gr. 1.8. r 9 = 18 = 19 .10. Take the increase from A to (7 or 121 minus the decrease.3.9.15.6.14. (iv) ^^^^(mod n). Therefore ag s ^ag* +t (mod n) .. 7. say r s and r s+t . (iii) If r M = /v.14.. 2 2 . and. the number of terms in the period is the least index for which <7 e ~l(mod n). Then the number E is the remainder in the division (3025a 4. are Index.12. . and a link C in the chain.11. Here J = 18 = <p(19).. recur . 19z=a(mod 62) and 19#==&(mod 121). the subsequent residues are n-r v n-r 2 . 5.16.. It should be noticed that : (i) No two residues in the period are equal. the residues of 2. we have r 5 = 16 x 2 = 13 (mod 19). r v r 2 .17. Each of these is one of the <p(ri) numbers <n and prime to it. Again.wheel.11. ag. etc. and the sul>sequent residues are 19-2. ag i+2 ~ag 2 . since a is prime to n. RECURRING DECIMALS Residues of Powers of a Number. and denote it by a. r t+l r^ f^^^* e ^ c -> where t^<p(n). 3. t^cp(n) and is independent of a.18. and every number not divisible by 19 is congruent . Take the increase from A to B or 62 minus the decrease. 6. At the start count the teeth or links from A to B and from A to C in the direction of turning.8.. Moreover. // a and g are prime to any modulus n. . for since a and g are prime to n. 1 .

for b<t.... if b r = a 8 (r<t... and since b is prime to n.9.. so that t may be equal to q>(n).12.. Also they are differen t from one another and from those of the set (A).8.1 . are Index.. then g r =g s ( mod n) and g r ~ 5 = 1 .8.5.. consider the residues of bg. .. If some number b of the set S does not occur in the set (A). where 0<6<Z. 1..... which is a divisor of 99(13) ~ 12. bg 2 .. Hence..... > Otherwise 2t<(p(n)> and then some number c of the set S does not occur in (A) or (B). 5 2 ..12. for b is prime to n.. bg t " l 9 denoting them by 6 ..5. the residues of 5.. 2.. let <p(w)=a + &. then bg r zzbg s .. & A .s ) according as s r...10. If the sets (A) and (B) include all the numbers in /S. If the set (A) includes all the numbers in S 9 then t q>(n)....5...... Here = 4. if t<cp(ri).. Again. . 5 1 .. ( A ) Each of these is one of the q>(ri) numbers less than n and prime to it. 8. 1.12. RESIDUES OF POWERS 437 2...5. Also no two of them are equal ..8...12. For if b r = b s where s<r<t.. // g is prime to n and t is the least index other than zero for which g i ^\(modn} 9 then t = (p(ri) or is a divisor of <p(n)... and by the preceding t^(p(n). Residue... In both cases b would belong to the set (A).... We then consider the residues of cg 9 cg l 9 cg 2 9 . For #o (n ) = l(mod n).. it is a divisor of q>(n).. . for r-s<t. t then 2t = <p(ri).. which is not the case. b t ^ .11. .4. which is contrary to supposition... bg l . Let the least positive residues of g^ 9 g l 9 g 2 .3....1... For the modulus 13. for if a r ~a s where s<r<t..7. Second proof. (B) These are all included in the set S. 6 2 . 6... If t<y(n). Let S denote the set of numbers less than n and prime to it. g r ~ s ~l 9 which contradicts the hypothesis that t is the least number such that g*~l.. then Therefore g b ^l(modn) 9 and consequently 6 = 0.. 0. cgr*. g*~ l be o> a i> a 2> a t-i .1...1.(mod 19) with some power of 2..s<t) 9 then bg r ^g s 9 and therefore b^g s ~ r or 6~<jr*-( r ... .

6. 2 7 .. Illustration. Every number loss than 13 occurs. 2 U are written down round the circumference of a circle.. 6 2 s2 10 . counting . primitive roots exist .. 2 2 . the first. 12. 6 1 . 2. the process can be continued until the numbers in S have been arranged in groups of t numbers. 7. every number prime to p is congruent with some power of g. and are equal to 2. .... so that the period of residues includes every number less than p.. the others can be found as in the following illustration. we can write down the residues (mod 13) of powers of any number.. In the margin. Thus 3t^<p(n)i and.. the period is incomplete and g may be called a subordinate root... 4.. . seventh and eleventh numbers in order round the circle. there are four primitive roots of 13.. Hence.. the reason being that 5 is prime to 12.. 11. without further reckoning.. 9.. they are congruent with 2 1 . we can show that these numbers are all included in S 9 and that they are different from one another and from those in the sets (A) and (B). 2 M is congruent (mod 13) with a primitive root.. 5. If tp-l... for the present we assume this to be the case. Let p be an odd prime.. 11. Hence if g is a primitive root of p. then g is said to belong to the index t. for any odd prime modulus.. are found by starting with 1 and taking eveiyffth number. and so 2 is a primitive root of 13. 6 2 . 438 PRIMITIVE ROOTS 3. for the modulus 13 the least positive residues of 2. 10. Every number less than 13 appears.. and. ^ . Hence t is a divisor of (p(n). g any number not divisible by p. Since <p(] 2) = 4. Since cp(p) = p-l. 7. the residues of 6. if 3t<(p(n). Incidentally.. It can be proved that. When one has been found.. the period is said to be complete and g is called a primitive root of the modulus p. etc. 8. Thus for powers of 6. 2 l ..denoting them by CQ> ^ ^ . 2 11 .. 2 5 . (C) As be'fore.... t is equal to or is a divisor of p-1... 3. 6. It t is a divisor of p-1... this proof establishes the truth of Euler's generalisation of FermatVtheorem.. since 6-2 5 . and t the least index for which g t s 1 (mod p). if JJL is any number less than 12 and prime to it. in the same way. The residues are 1.. Hence 6 is a primitive root.. fifth.. An Odd Prime Modulus: Primitive Roots.

. 5 2 . rjn. (v)). the least positive residues of g. we have the sequence ^m> ^m+e* ^m+2e> These residues recur. 10m. 10 2 m.. 5. Thus 5 is a subordinate root . /* 3 . Thus if p = 19 and # = 2 (See Art.. a s . 5. for g x ^l{modp) if x = ef..... Starting with any residue r m and taking every eih residue... t the least index for which #' == 1 (mod y) and r fl . Again. . then that is to say. the sum of the residues which form the period is divisibk by p. all the fractions m/n have the same period. where t is the least index other than zero for which 10* = 1 (mod n) . . Hence. then * = 18 and 5. have the same period. suppose that e is a factor of t (not t itself) and = ef.then (1) If n is prime to 10. and they form a sub-period containing f residues . r^n. 1. r v r^ . RECURRING DECIMALS 439 4. after four counts we reach the number. and for no smaller value of x.. then 2 = 4 and Again. 12. are the least positive residues (mod n) of m. it belongs to the index 4.. g l . Thus if jo = 13 and </ = 5. g 2 . Since 3 is the greatest common divisor of 9 and 12 and 12 = 3 .from 1. Recurring Decimals. . 5 1 .. by Arts. Also ^ + W. we have 5 = 2 9 ... g any number >1 and not divisible by p. For the remainders r v r 2 . d t) then r 8 /n = 0'd s+l a s+2 .. the remainders recur. and the period is 1.+W2 + -. . . 4. from which we started.. or every third number in the reverse order.. Also if m/n = O-a^ .. so that all the fractions m/n..-+W(/-i)=^ Hence the sum of the residues forming the sub-period is divisible by p. and the residues of 5. m/n is equal to a pure recurring decimal with a period of t figures.. so that t=*(p(n) or is a divisor of cp(n). . a t a l . . are obtained by counting every ninth number. in the process of expressing m/n as a decimal. with a period of t figures. 8. Let m<n and prime to n. If p is an odd prime. 1 and 2. It follows that (i) If t = y(n) and m is any number <n and prime to it. .

or t is a divisor of p-l. and let Then F(t)^0(mod p). for 10 T isnot si. and l/p has a period of t figures.. the last of which is not zero. + lO(*-V* = (lQ(p-Ur _ 1)/(10 T . iyy\ (2) // . This follows from the reasoning in the last part of Art. the fractions m/n can be arranged in e groups of t fractions. and therefore 10* T -1 is not ^O(modj) 2 ). according as 10 is or is not a primitive root of p. or a divisor of pt. 440 DECIMAL PERIODS (iii) // np. then t=p-I. If r<.F = <1. such that the periods are the same for fraction in the same group and different for those in different groups. not 2 or 5. then <p(n)p r ~ l (p 1) and t is equal to or is a divisor of p r ~ l (p-l). m/ (3) // F= <1. then F . and therefore Hence T is pt. any prime except 2 or 5. n and 10 are prime to one another.l)F(r).. For 10*F is an integer with not more than k digits.(ii) If (jp(n)=et where e>l. Now 10 PT . for lO'sl(mod^). The only known cases in which Ijp 2 has a period of t figures are when p = 3 or 487. and k is the greater of the two ^ JJL. where m is prime to 10. For let T be the number of figures in the period of l/p 2 . where m. or a divisor of t. (iv) If p is a prime. v. we have 1 + 10 T + 10 2T + . 2. since p is a prime. Hence F(T) = 10<*-V T = l(modp). and. The case in which n=p 2 deserves further consideration.1 . Consequently r is not <t and T = t or pt.* If np r . where r is t.(10 T .1) E=0(mod p). NOTE. then l/p 2 has a period of t or pt figures. T = t or pr. then F is a terminating decimal with k figures.

. 10 2 w. In practice.. //. respectively.. after r steps of division. . where n is prime to 10 and m prime to n. therefore 10 e s 1 (mod p a ) and. the following theorem is useful. t". then IO k F = m'/n. 5. Ex. .. where m' is prime to n and less than 10*. 59. and since p a . q. . Similarly a r+2 = 9-a 2 . 19.. 9. 9-a 2 . we can easily express any fraction as a decimal by using the theorem of Ch.. are prime to one another . l/q b . . etc. .. the division need not be carried beyond . Again. .. . if r is any common multiple of f.. 10m. since $' is the least index such that 10*' = 1 (mod p a ). where t is the number of figures in the period of l/n. hence a r+l = 7(10-0^ +/) = 9 . 29 . For let m. Hence the subsequent residues are congruent with -m l9 -m 2 . where / is a positive proper fraction. the subsequent figures of the decimal are 9-a x . v. 9-a 3 .. 61. etc. 97. where n is a prime or a power of a prime... Therefore IQ k F = I+f.. It follows that t is the least common multiple of t'. r.. t'" 9 etc. f. etc. . l/p a . Let m/n=0' a^a^ .C. etc. (4) // F = -~j~ where p. and are equal to n m l9 n w 2 .. are different primes other than 2 and 5 and m is prime to p. . t is the L... . t" '. 47.. q.. 17. where / is an integer of not more than k digits and /is a pure recurring decimal with a period of t figures. etc. XXVI. t must be a multiple of t' . then m r n. t'". .M. we have 10 T ^1 for each of the moduli p*.is a mixed recurring decimal. . t". a r+l is the greatest integer in 10(n-m)/n.. (f. To express 1/73 as a decimal. and the period contains 2r figures. 23. Moreover. For let k be the greater of the two jit. . then 10* s 1 (mod n). USEFUL THEOREM 441 For let n = p a (f . Given a table showing the periods of decimals equivalent to fractions 1/n. . oft'. ". then if t. are the numbers of figures in the periods of the decimal equivalents of F. etc.m = m (mod n) . m ly m 2 ..%. be the least positive residues of m. 1. with a period of t figures. r. * The primes less than 100 of which 10 is a primitive root are 7. Therefore t is a common multiple of t' . t'. . . tfie remainder n-m occurs. etc. t" . similarly. t is a multiple of t".

54... for since the remainder g \ ^AQ / . 2 2 . /(6)-3 3 .. 22.. 46.. the corresponding figures in the decimal being ~510 9-0... 10 2 ..... 44 . 101..271. For the modulus 19... Wh at are the primitive roots of 19? . the fractions ??f/73 can be arranged in 72/8=9 groups such th at the period is the same for those in the same group. 2. 13.. 11 . choos e any number loss than 73 not included in the set (A). 10 1 . A third group can be found in the same way by choosing a number less than 73 and not included in the sets ( A).41 . 7 1 .. given the least positive residues of 2... (A) and one group is obtained by giving m these values. By the preceding. 27.the stage shown in the reckoning. 29.. 72.. the subsequent remainders are 73-10.. etc. 73 . 9 are given in the following table : /(4)-3 2 . Thus -^Q 1/73-0-01369863. 11 . 51.. etc.. 137. (B) A second group is obtained by giving m these values.. The remainders for 2 /73 are congruent (mod 73) with 2x1... 442 RESIDUES EXERCISE XLV NOTE...... /(7)-3 2 . etc. The prime factors of /(0 = 10*-1 for { = 1... 53. 7 2 . /(8) = 3 2 . . and 7. 239 .. say 2. the number of figures in the period is 8 and (p (73) 72. /(9)^3 4 . etc.. 19. the comple te set of remainders for 1/73 is 1. 270 73-27.. and are therefore 2.. (B) and so on. 1. Hence for values of m less than 73.7. 4649. write down those of 10. 101 .. 37 .. 1).. 63. To get a second group. (Art.. 333667.. 10. 2 x 10.. etc...... .. 2 x 27. 20.... 71.37. etc. 3.. 9-1.. 2 1 . /(5)-3 2 .9136 72(=73-l) occurs.. Again. 11 .

find the least positive residues of 3.1 or +1 (modp) according as t is even or odd.. 10 1 . g 1 . t is even.. and we write . etc. For the modulus 17.. etc. g 2 . t is the number of figures in th e period of l/n or l/p . p is a prime not 2 or 5. are t he least positive residues of </. If ---O-atGU . a^_ l9 . prove that V) **. If l/n= O'(iia 2 .. r t _ l ^g s (modp) 9 where 8 = + I +2 + .1). * n p 4..a t is the number of which the digits are a l9 a 2 .. . [The decimal = a& 2 .. g belongs to the index t. Prove that O-a^ .. . . . and r . 3 2 . a t x r a^a^ .. (ii) /wr.... then r and a t are the least positive integers x 9 y 8uchthat lte-ny = l and a^z . r t _^ .. 6 r ...or ~ = 0'a 1 aotf 3 . ^/(lO* . If p is an odd prime. n n 6....] 1 7^-1 5. f y j. 3.. a t a^a^ . show that theorem (ii) becomes Wilson's theorem. (modn or p)... aA t + ---^ + t + . g^2. ./ j. r . J .1) where a 1 a 2 . [(i) If t is even... a t and r is the remainder immediately preceding the remainder 1 in the division. 10 2 .I (mod p). Verify when #^13... .. r l9 r 29 etc.. prove that (i) p 1 pccurs as a residue if.] RECURRING DECIMALS Here n is prime to 10.. and = 0-6i6 2 .. W hat are the primitive roots of 17 ? 3.2. . a r . +(t ... 5.. g f ! 2 ~ . and only if.. etc. . are the least positive residues of 10 . (iii) If g is a primitive root of p. (ii) v^o .. 3 1 . r 19 r 29 etc..

4 x 6 + 3 =27.. Put 9 to the right of 6.. 36 is 9.. a^ . put 7 to the left of 6 and carry 2. Put 3 to the right of 2. 8. put to the left of 7. a 2r . 4 x 7 + 2 = 30. put 9 to the left of 2. Hence the following : Put down 3 and divide by 4 as below : "076923. 9.] RECURRING DECIMALS 443 7. (i) Put down 3 (the last figure of the period) and multiply by 4 as follows : 076923 4 4x3 =12. . 8.where the a's are digits of the numbers indicated. prove that and give an example to show that such relations do not hold if p is a composite number. a^_i/r=a 1 a 2 . a r a r+l .. 30 is 7 and 2 over. . . which is the first figure of the period. If p is a prime and 1/p = a^ . a t . 27 is 6 and 3 over. 79. and the figures recur. 59. Justify the following : If A C= 9408 ~0 .. Express -^ as a decimal by using the equality 10 x 4 . Put 7 to the right of 0. [l/ft^O-a^ . put 2 to the left of 3 and carry 1. If the process is continued. 4 into 9 is 2 and 1 over.. . . 4x9 =36. 8 are p = 17. 43. a t a^a z . explain why no figure has to be carried when a l is multiplied by r. Having found r and a t9 all the figures in the period may be found as in Ex. put 6 to the left of 9 and carry 3. rln Q-a t a l .. 19. a^! . Cases in which l/p can be quickly expressed as a decimal by the methods of Ex. 4x2 + 1= 9. 29.. 23. 7. 4 into 12 is 3. 89.. the figures recur and tV=0'076923. (ii) Or thus in Ex. Put 6 to the right of 7. 4 into 3 is 0. .13 x 3 = 1. Put 2 to the right of 9. 4 4 4 4 into into into into 3 is 0.... 10.

. 14. [Express ~ or ~. Prove that gV^O-0 12345670. 10 show that j& =0-032258064516129.37 4 . _ 11.] fit \jTt 15.as a decimal. Find n' when n~ 41 and when n~123. prove that the values of n for which the decimal equivalent of 1/n has a period of t figures are as follows : n 1 3. Show that for values of m<41 the fractions r/i/41 can be arranged in 8 groups such that all the fractions in any group have the same period. the remainder 2 occurs. By the method of Ex.then tV = -0588235294 . If n is a prime or a power of a prime. 444 BINOMIAL CONGRUENCES 12. prove that a number n f can be found such that every digit in the product nn' is 1. What are the values of m for the group to which 2/41 belongs ? 13. where jB=4J.A. If n is any number prime to 10. . (7=4J5. [The division shows that 10* =588 x 17 + 4 . according as n is or is not prime to 3.9 3 27.] 2 F B. and show that the values of m for which m/81 has this period are given by m 1 (mod 9). etc.C. [After 6 steps of division.

j0-3. 59. which is usually successful. If 1/67 is converted into a decimal fraction. 47. y = 23. = 3. 19. XL VI. 37. \Ve find that a value of g is il2 . 79. Primitive Roots (continued). Ex. 1. as remainders in the process. but in the example below we give a method. = 2. = 7. that g 2 10(mod 67) will lead to a primitive root. 17. converting 12/67 into a decimal fractio n. less than 100. 11. 5.101 5 | 6 41. we find that all the numbers less than 67 are included .] 16. See note at the head of this Exercise. W hen these are interpolated between the remainders in the reckoning for 1/67.1. and its . we obtain. and. 137 9 333667 [The values of n for which 1/n has a period of t figures are included among the divisors of 10* . 3 suggests that t his period consists of every alternate figure of the period of some primitive root. p = ll. and. the residues of the odd powers of 12. 89. 43. 13. with a period of 33 figures. 3. 4649 73. 67. thus 12 is a primitive root. 29. g . 31. p. 0-5.for any prime modulus. 61. 53. 73. Show that the only prime p for which the decimal equivalent of l/p has (i) a period of 10 figures is 9091. 13 239. founded on the assumption of Art. 83. 79.271 | 7. Find the primitive roots of 67. g. explaining why only 5 and 6 steps of division respectively are necessary. = 6. Also express 1/9091 and 1/9901 as decimals. consequently. p = 7. and (ii) a period of 12 figures is 9901. For convenience in solving the congruences in Ex. the reckoning shows that 10 is a s ubordinate root of modulus 67. 6. Art. p = I . The orthodox methods of shortening the reckoning are given in another volume . we here give the least primitive root. Theprocessof finding a primitive root is one of trial.

etc. Also... (256) a = (-2) .e. Let x^g s (mod p). 2. a single value of A.. 2 27 . less than p-l. 33. 32. U. we can find r. 53.1 . 61. # 15 =31(wod 37). which is the only solution. by Chap. has d solutions. etc. or is not.. The Congruence x n ^a(mod p). if x n z=a. hence 5=3. x s= a x ( mod p). Hence... (3) 2 s5(mod 23). then 2 15S ==2 9 . 25. 30. Ex. such that If A has this value. 17.let g be a primitive root of p. 23. and 15$~9(mod 36) .. then g ns = a^g r (mod p). Solve the congruence. together with their periods . 3... 2 15 . 27 (mod 36). since the period of g contains all the numbers less than y. x 13 == 21 (mod 23). 62. 15.period is 1. these being the residues of 12 fc .. according as r is.. and z^2 3 . 21.. 8). Let d be the highest common divisor of n and p .... 3.. 49.31. 1. From this. tFSE OF PRIMITIVE ROOTS 445 7. 17 E= 1 (mod 22) . thus we find that other primitive roots are 61.. and. (2) // n is not prime to p-l. XXV. 51. (x v ~ l ) k = x (mod p ) . such that g r z=a(modp). and. as in Art.. 8. For. where 5 has any value which satisfies (A). i.... Here x* ~ 21* (mod 23) .. 12.. there is a single solution.. divisible by d .. (1) Ifn is prime to p. Solve the congruence. x nX = x . EXERCISE XLVI .. Then. 10. and 13 . where k is prime to 66. 3. since n is prime to j>-l.. all the other primitive roots can be immediately writte n down. It is supposed that p is an odd prime. if x=2 s . 36. and thus.. there are several solutions. 7. hence we have z:EE21 17 ~(~2) 17 ~(-2) . and we find that 2 9 =.. 41. the solutions of # n E=a(mod p) are given by x^g s (modp). and therefore also the congruence x n ^a. 2. 6. 13. for d is a factor of n and p 1 (Chap. or there is no solution. there is.. XXVI. then x nA = a A . Ex. or no solution. 32. A primitive root of 37 is 2. (A) The congruence (A).I. 7.. and therefore ns^r(mod p-l) ..

a. f(x + A)>0. 6. a: 21 = 2 (mod 31). 2. It may be convenient not to remove the multiple roots. which proves the theorem. that a primitive root of 237 will satisfy one at least of the congruences. z 3 =l(mod 19). 1-6 : 1. * 7 ==8(mod 13). then Hence if f(h). <7== 10. The usual procedure is as follows : (i) We find an interval which contains all the roots. 8. 12. 1. by a process of approximation. 7. g a = 10. we find smaller and smaller intervals which contain the root. This depends on the following theorem : // h is a number such that f(x) and all its derivatives are positive when x = h. without calculation. x 5 =2(mo& 17). f M (h) are all positive. Find the general solution in integers of # 2 = 19z/-f-5.Solve the congruences in Exx. is given in the next article. VI. . for that is to say /(x)>0 for x^h.. XVIII. and VI. 11 = 5 (mod 31). no algebraical solution of the general equation of the fifth or higher degree has been discovered. Thus we are only concerned with irrational roots. (ii) We separate the roots. 5. 4. For if n is the degree of f(x). (iii) Taking any interval which contains a single root. In fact. Find three solutions in integers of # 8 = 73t/4-3. CHAPTER XXVIII NUMERICAL SOLUTION OF EQUATIONS 1 . If any rational or multiple roots exist they can be found and removed from the equation (Ch. that is to say we find intervals each of which contains a single root or a multiple root. f'(h). The Problem under consideration is to find the actual or approximate values of the real roots of any equation witli numerical coefficients. ^=31(mod41). Newton's Method of finding an Upper Limit to the Roots. 3. but which yields closer limits. . 9. Show. This question is quite distinct from that of finding an algebraical solution. A method due to Newton which involves more calculation. Ways of doing this have been given in Ch. 14). 2.. 4 s 10. then h is an upper limit to the roots of f(x) =0.

It should be observed that if we have found a number h r such that are all positive. /"(a) =2 (3s -10). . the integral part of which is therefore 1 1 . Separation of the Roots. Fourier's rule. This follows from the above theorem.NOTE. In applying this theorem. . is incomplete.Six + 1609^0. 2. while that of Sturm definitely separates the roots.. 3. 1. Ex. 1. then these functions are all positive for x>h r . z 3 -20z 2 . 8 and 12. * 4 -4x 3 -3* + 23-0.110=0. p. so that SEPARATION OF ROOTS 447 We then find the least integer ^ such that f^ n "^(x)>0 when xh^ If this value of x does not make f^ n ~^(x)>0 9 we find by trial a greater integer h 2 such th at Continuing thus. (Cf. (1) If a is a root of /(x)=0. 5. z 3 ~2z 2 -51z. we suppose that the coefficient of the highest power of x inf(x) is positive. Ex. Find the integral part of the greatest root of f(x)=x*-10x*-Ilx -100 = 0.) Here f'(x) -3rr 2 -20*. 1. 1-5 find an upper limit to the roots by Newton's method. The least integral values of x for whieh f"(x). f^(x).11. /'(a*) and f(x) are positive are respectively 4. Subsidiary Theorems./ n -^(). for the derivatives of f^(x) are / (r+l) (z). though often convenient. /'"(z)=6. as the variable x increases through the value a. 91. We require the following theorems. we can find the least integer h which makes f(x) and all its de rivatives positive. but its application may be very laborious.. 3. EXERCISE XLVII In Exx. the functions /(x) and/'(#) have opposite signs just before x=<x and the same signs just after x=a. 4. Rules for this purpose were given by Fourier and by Sturm. Thus 12 is the least integer which exceeds the greates t root.

. which is not zero. Denote the modified remainders by / 2 (z). just before x = a... fi( x )> fz( x )> "-fr( x ) are called Sturm's functions and f^x). In the modified process it is essential that such multipliers should be positive..r /(a + A)-A/'(a) + /"(a) + .For since /(a)=0.. is an r-multiple root of f(x)=Q. of f(x) and fi(x) be performed with this alteration : The sign of each remainder is to be changed before it is used as a divisor : the sign of the last remainder is also to be changed. .. then... In the ordinary H. / 3 (#)..... and the result follows by applying theorem (1) to every two consecutive terms....C. (B) Suppose that /^(a) is the first term of the sequence /'(a).. we have 7>2 7. which proves the theorem. Sturm's functions are connected by the equations . Sturm's Functions..... all the terms of the sequence (C) except the last vanish.. For./ (1 .. the signs of /(a -h h) and /'(a + A) are respectively the signs of the first terms in (A) and (B) which do not vanish ..F. process we can multiply (or divide) any remainder by any constant... /"(*). as the variable x increases through the value a.. when z = a. and all the functions have the same sign just after X QL. 448 STURM'S FUNCTIONS (2) If a. Let f(x) be a polynomial and fi(x) its first derivative.C.. (A) a) + . 4. Hence /(a + h) and /'(a -f h) have the same or opposite signs according as h ^ 0.(A) . this function has the same sign just before and just after X OL....... /'>(*) (C) are alternately + and -.. or .F. It is important to observe that since /^(aJ^O and f^(x) is continuous at #=a. + /W(a) + . Let the operation of finding the H.. . .. . For sufficiently small values of h.. they are therefore the same as the signs of A r /* r) (a) and h r ^ l f^((x).. f r ( x ) \ then /(#). /'(*). /"(a).and -f. It follows that as x increases through the value a.. r changes of sign are lost in the sequence (C). .1) *. f r (x) are known as the auxiliary functions.+ 1 /0(a) + . the signs of the functions /(*)... f 2 (x).

. STURM'S THEOREM 449 Proof for the case in which f(x)^0 has no repeated root. Thus if <7 S = 0. A similar remark applies if one of the y's is zero. f m ^(x) and f m+ i(x) are continuous at x=a.+i (a). etc . f r (x).. and consequently f r (x) is independent of x. Moreover. these quotients being functions of x. and they have the same signs just after x=a. the signs of / m _i(^). (iii) If any term of (S) except the first is zero. Hence. 5. are the quotients in the process just described. Suppose now that x increases continuously from x a to x b. Sturm's Theorem. if this happened.-i (a) =-/. / (a) stands for /(a). . one change of sign is lost in the sequence (S). then /s-i(*H -/mfc). (ii) If a is a root of f m (x) where m = l. (ii) No two consecutive terms of the sequence (S) can vanish for the same value of x. or r-1.. no one of Sturm's functions can change its sign unless x passes through a value which makes that function vanish.where q v q 2 .-/w+iMIf ra = l. Suppose that x passes through a value a which makes just one of Sturm's functions vanish. A(*).. (S) when x=a over the number of changes of sign in the sequence when x~b. MX). all the subsequent terms including f r (x) would vanish. For f(x) and f^x) have opposite signs just before x=a../(z) and/ x (x) have no common factor. 2. Thus if / s (^)==0. (i) If a is a root of /(x) = 0. for. (i) Since /(x) = has no repeated root. . so that each of these is of invariable sign near #=a. For / rn (a)=0. b are any real numbers (a<6).. fz(x)./ m (x). no change of sign is gained or lost. . for these functions are polynomials. Iff(x) is a polynomial and a. the terms which precede and follow it have opposite signs.f 3 (x) . the number of distinct roots of f(x) = which lie between a and b (any multiple root which may exist being counted once only) is equal to the excess of the number of changes of sign in the sequence of Sturm's functions /<*). It is important to notice that the relation between f(x) and fi(x) is essentially different from that connecting f^x) and/ 2 (x) . then /. and therefore f m -i(oc)=-. Equations (A) lead to the following conclusions. just before and also just after #=a.. As x varies.

(iii) Suppose that one x a\ then.Thus. f r (x).. 450 ESSENTIAL PROPERTIES Remarks. we arrive at a function f s (x) such that all the roots of / s (. at any stage. as ar increases. no two of them can be consecutive functions. process need not be continued./i (). This proves the theorem. the H. /j(x).V + \ and from this the sign of a'" may often be found without much calculation. .fm+i( x ) are either -f -f ./ r (^)> if / TO (a) = 0.i change gained or lost. and no change is gained or lost in the sequence / 1 (x). . Thus one change is lost in the sequence (S). and we can use /(x). . labour may be saved by the following considerations..( a' -^ . showing that no change of sign is lost or gained. the last f unction f r (x) is independent of x and its sign only is required.r)=0 are imaginary. Thus. and under no other circumstance is ..F. f s (x). a'". (ii) If.C. For TO-1 (a) and / m+1 (a) have opposite signs. relative to numbers on the scale of integer s. Ex* 1. no change of sign is gained or lost. If f(x) is not one of them. Let x have a value such that f r -i(x) = 9 then f r (x)= -fr-zfa).. a!'x + 6". If x passes through a value a which makes more than one of Sturm's functions vanish. Find the number and position. (i) If there is no repeated root. In applying Sturm's theorem. then by the preceding. then/ of Sturm's functions as f m (x) vanishes when the number of changes of sign in the sequence we may regard the sign of/ rn (a) as either + or ... and f s (x) has this property.. If f(x) is one of them. one change of sign is lost between f(x) and/ 1 (a?). a change of sign in the sequence (S) is lost whenever x passes through a root of f(x)=0. in counting /()./ 2 (x). of the real roots of . if the last three functions are a'x 2 + b'x + c'. instead of the complete set of functions. For the essential property of the last function is that it should remain of invariable sign for all real values of x.or + or h-f or 1-. then a'" = .

99+ 104+ 301 43 3569 .F.4+ 7 43 4-12.9# 2 -4z + 7.8+ 28+ 12=/(s) 4.69=/ a (z) 83 .5727 3569-8213 -4257+4472 + 12943 -4257 + 7128+ 6831 .288. process is as follows : -99 /(*) = 4 .21 172 172 .276 43. and the modified H.72.16 + 84 + 48 -12 + 27+ 12.4+ 7 -3-4+ 21+ 12 4 .9.Here f l (x) 4x* .172+ 301 .C.12 .5976 .387.9.

/ 4 (*)= +48074. then/ 4 (s) = When x=-^> / 2 (x) is negative.2237 . MODEL SOLUTIONS 451 The number of changes of sign in the sequence of Sturm's functions for various values of x are shown below. x -oo -2 -1 2 3 oo /i(*) -f .191 / 4 (*)= + 48074 43 2237 Thus / 2 (aO-43a. If / 3 (#) =0. f s (x) = S3x-191. Thetef ore / 4 (x) is positive. The sign off^x) is found more easily by Remark (i).5727 83 32) 2656. 2 -72:r-69.6112 185671-475341 185671+427267 83 .-f -f -f 4- .

1 and and two between 2 and 3. Jf or 15 2 <4 . 2. 24. the signs of Sturm's functions are : -oo 1 2 oo /(*) />(*) Number of changes of sign/ ' " showing that there is one real root between 1 and 2.x . 8 .C. x. Find the number and position of the real roots of (i) Taking f(x) ~x* . proce ss is 5**-l Also.1. . . f s (x)^ -fi(x). . when / 2 ()=0. and the modified H. therefore For various values of x.+ + Number o* .1. and that the other roots ar e imaginary. and we need not pr oceed further. X -oo -2 -1 1 2 oo . . } 4 4 3 2200 changes of signj Hence the equation f(x)=0 has one root between -2 and -1.F. Ex. The roots of / 2 () =0 are imaginary. one between . we have f^x) 5x* .

as x increases through any root a of <f>(x) a change of sign is lost in the sequence (S'). as in Art.. /?. tl*i(x) = q. fc). ... ... -^(a)..F.-/r(*).. ./<*) /(*) .. Also for any value of x. therefore ^(x) and <f> L (jr) have the same sign when x = oc. .+ + + + -f -f + + . by Art. etc. + + Thus there is one real root between . /(*)> fl(*)> /a(*). they have the same signs for values of x near x=oc. Sturm's Theorem (Multiple Roots)... $%(x}.l (x-y) r ... Suppose that /(*) = (*-)"(* -W*-y) r where p..2 and . It follows that the number of real roots of (f>(x)^0 in the interval (a.1 . another between 1 and 2. are all different (re al or imaginary). 4 by u. 452 ALL ROOTS REAL 6.... Whence..... we have the identities </>(x) = q l i/J l (x)-ifj 2 (x). Dividing equations (A) of Art. It remains to consider the case in which the equation f(x) = has repeated roots.. of f(x) and /^z)... and since these functions are continuous. the other roots being imaginary... Again. . 5. it follows that no change of sign is gained or lost as x passes through a value which makes one of the set ^(x). Hence.j/i 2 (x)-fa(x). (S) are divisible by u..... . Hence u is the H..1.. the number of changes of sign in the sequence (S) is the same as that in the sequence (S'). are positive integers and a.l (x-p)*. 5. ^ (a) =p(oc .iff r (x) vanish. q.. and tfj r (x) is independent of x.C./3)(a ....y) . and all of Sturm's functions. where a<b is equal to the excess of the number of changes of sign in the sequence (S') when x = a over the number of changes of sign when . then where u = (x-a)*.. Denote the quotients by so that <f>(x) = (x-<x)(x~l3)(x-'y) . y.. r.

XII... Again. if / 3 (z) = 0. Application of Sturm's Theorem. then / 4 (z)= -/ 2 (z)..3H 2 ) (3aJ . In proving this we require the identity (Ch. THE BIQUADRATIC 453 7. as x passes from .3a 2 # 2 / 3 z 3 + 3Hz + G 3H 2 . Conditions for the Reality of all the Roots... For. n changes of sign will be lost in the sequence of Sturm's functions if and only if both of these conditions hold.. XII. the number of Sturm's functions is n -f 1.... In general.. 9.x = 6. and (ii) the leading coefficients of all these functions must be positive. The necessary and sufficient conditions that the roots of f(x) = Q may be all real and different are : (i) the number of Sturm' s functions must be n -f 1 ...3G(GI)L + KL 2 = (a 2 / . For the biquadratic.. therefore ) = 3H(GI) 2 ... f 3 (z)=-(2HI-3aJ)z-GI..HI) = 9a 4 /J 2 .. (6)) 2 + 4# 3 = a 2 (#/-aJ) .oo to + oo .2HI) 2 + 3G 2 I (3aJ ... That is.12a 3 HI*J + a 2 H 2 (U 3 . the number of distinct roots of f(x) in the interval (a. 10 (6))..... 6) is equal to the excess of the number of changes of sign in the sequence (S) when x~a over the number of changes of sign when x = 6..27J 2 ) + 3I(G 2 + 4:H 3 )(3aJ-HI)* = a 2 H 2 (4/ 3 . Let/(x) be a polynomial of degree n with its leading coefficient positive.. (A) Sturm's process is -Hz G f 3 (z)=-Lz-GI where L = 2HI-3aJ. where K=a 2 I-3H 2 (Ch.27 J 2 ) ... 10. Sturm's functions are /() and f z (z)=-3Hz 2 ~3Gz-K.

Hence if A is positive and both H and 2HI . For..* + W* + *& + * 3H 2 z 3 + 9H 3 z +3GH 2 3H 2 z 3 + 3GHz 2 +HKz f 2 (z)=-3Hz 2 -3Gz-K * ./(*)> .... two roots are real and two imaginary.3GHz 2 + (12/P ..... and an r-multiple root is counted r times. . AW.. In particular if N ..N') .. / 2 (x).. . where a.oo to -f oo .(*). and if N N' = no real root lies between a and b. If A is negative.... 454 FOURIER'S THEOREM 8. H and 2HI . all the roots are real If A is positive and at least one of the two. (?) when x a and when x = 6.. XII. f n ( x ) be its successive derivatives. b are any real numbers.. all the roots are imaginary. we may therefore take / 4 (z)=/ 3 -27/ 2 .. in the first case four changes of sign are lost in the sequence of Sturm's functions. Fourier's Theorem.. of which a is less than 6. is positive. Let R be th e number of real roots of f(x)~0 which lie between a and b. there is just one real root in the interval.N' = 1 .. Let f(x) be a polynomial of the n-th degree and let fi(x). /.. * The identity (A) is used in these steps. then N^N f and R^N-N'i also (N .R is an even number or zero..3a J are negative. All this has been proved more easily in Ch. and in the second case no change of sign is lost..3aJ.. as x passes from . Let N. N f respectively denote the number of changes of sign in the sequence /(*)..a 2 HI)z + 3H 2 G -3GHz 2 -3G 2 z -KG a 2 ) -a 2 (2HI-3aJ)z-a 2 GI Disregarding square factors...

but neither f m -i(x) nor / m + r (x) is zero. / 6 (*) = 240. and if any are lost.. two changes are lost. 3. (1) Suppose that x passes through a root a of /(x) = 0. Let x increase continuously from x = a to x = b. r changes of sign are lost in the sequence f(x). Proof. (ii) Suppose that when x=a' the r functions all vanish.23* 2 + 38* ~ 4 -0. 1. (ii) If a is an r-multiple root. under this heading. their number is even. Then if r is odd there is a loss of r . 3. This has been proved in Art. there is no gain or loss of changes of sign when / m _i(a') and f m+ i(at') have opposite signs : but if these have the same sign. By Newton's method we find that all the roots lie in the interval ( . Here / x (x) = 2 (5^ -f 14* 3 . . (i) If / m (a') = 0. f t (x) =2(20** +42* 2 . For various values of x the signs of Fourier's functions are as below.fi(*).4Qx* . no change of sign can he gained or lost in the sequence (F) unless x passes through a value which makes one of its terms vanish. DISADVANTAGES 455 Ex. (i) If a is an unrepeated root. As x varies.') nor f m +i(a) is zero. / 4 (a?) =8 (30* + 21). but neither f m -i(<x.7.(See Art. . 4)..120* -23).60* 2 .) (2) Suppose that x passes through a value a' which makes one or more of the functions /j(x).. and they have the same sign just after x=a'.:fr(*).In this connection /(x) and its derivatives are called Fourier's functions. no change of sign is ever gained. while if r is even there is a loss of r changes of sign in either case.1 or r + 1 changes of sign according as / m -i(a') and f m + r (&) have opposite signs or the same sign. / 2 (x).23z + 19). For f m (x) and f m+ i(x) have opposite signs just before x=a'. Thus. Apply Fourier's method to separate the roots of f(x) =2^ + Ix* ./ n (x) vanish. a change of sign is lost between /(x) and fi(x).

X -7 -6 -2 -1 1 3 4 /(*) . 4.+ + + fl( x ) + -f ." + + .-f -f /.(*) + 4.+ . + /(*) -f + .

f(x) is positive. in which case they can be separated by further trials. 0-5) and (0-5.+ /!(*) + + 4. for sufficiently small values of A. We may therefore put a + h instead of a and b -h instead of 6. or none at all. 1). if real. Suppose that (as in the last example) all the roots of an equation are accounted for except two.7.-f + -f 4+ Number of | changes of sign/ 5443 3 1 1 These results show that there is one root in each of the intervals ( . If we fail to do this the possibilities are : (i) The roots may be very nearly equal. / 2 (#). This difficulty is met by observing that. 1). we find that when x =0-5. Hence there is a root in each of the intervals (0.h. e te wh en re =0-5. Subdividing the last interval. It should be noticed that we need not find the signs oif l (x). Remarks. as stated above. must lie in a certain interval (a. (2) Disadvantage of Fourier's Method. . the question arises as to what sign is to be attached to one of Fourier's functions which may happen to vanish when x=a or b. the number of roots between a and b (a<b) is the same as the number of roots between a + h and b .2. 6). . which. (3. and either two roots or none at all in the interval (0. . 4). for it lias been shown that either two roots. lie between and 1. ( .6). where A->0.1 ). it may require a large number of trials to separate them. If these roots are nearly equal. (1) In applying Fourier's theorem.

Use Sturm's theorem to show that the equation x z + 3Hx + G=Q has three real and distinct roots if and only if G* + 4# 8 <0. 3z 4 -10z 2 -6a. 2-4 use Fourier's method. then by the second mean value theorem. z 3 -20z 2 . J -4a:-f 11=0. and if we begin by testing for equal roots it would be better to use Sturm's process. and must therefore have at least s imaginary roots.101 =0. Imaginary Roots. If it is known that a certain interval contains no root of f(x) = 0. In Exx. 5-11 use Sturm's method.10z + l=0.3x a . n changes of sign are lost in the sequence of Fourier's functions.s real roots. 9. b) where a. Similarly N' is the number of changes of sign in the equation obtained by diminishing the roots of f(x)=0 by b. Newton's Method of approximating to a Root. (iii) The roots may be imaginary. EXERCISE XLVIII 1. find intervals (a. 13.3z 4 .3 Ix +1609 = 0. and that. we obtain Sudan's statement of Fourier's theorem. the transformed equation is so that N is the number of changes of sign in this equation.46* . As x increases from . 2z 3 -15z 2 -8a.oo . then the equation cannot have more than n .3x-f-23=0. z 5 .oo to 4. 8. s 3 + 2z 2 -51z + 110=0. 6. + 166=0. 7. 10. (ii) Newton's method of finding an upper limit to the roots. From Fourier's theorem deduce (i) Descartes' rule of signs .2x . as x increases through the interval. 3. a.] 9.* ~4* 3 . 456 NEWTON'S METHOD (3) If the roots of f(x) are diminished by a.4.(ii) The roots may be equal.+16=0. x 5 + 3x 4 + 2s 3 .24z 3 + 95s 8 . Let a 4.2 =0. 12. 4. If e<0-55. s changes are lost. If N and A 7 ' arc defined in this way. and in Exx. a^-3a. x*~x*-4x* + 4z + l=0. In Exx. jS). 11. . Suppose that the equation f(x)~Q has a single unrepeated root in a small interval (a. b are consecutive integers which contain the real roots of the following equations. 2z 3 + llz 2 . and no matter how many trials are made. prove that the equation # 4 -f 4# 8 -f 6x 2 -4x + e=0 has two real and two imaginary roots. [/ = e + 7. 2.h be the actual value of the root. 5. so that f'(x)^Q within the interval. J = . 2-11. we may not be able to discover whether this is or is not the case.

hence. this is the case if | 1 1<| h |. /?).a is a small number of the first order.1 5. and ai = l-46. /( ai ) =0-0193. =a -/()//'(). then f(x)=x 2x -3=0. so that each of these functions retains the same sign for all values of x with which we are concerned. This method is often employed in cases where the interval (a. that if k is the greatest value of f"(x) in the interval. . NOTE. /(. In the preceding. and f(x). that is if This condition is satisfied if /(a) and/ 7 ' (a) have the same sign. and . it follows that t has the same sign as h. f(x) may be any function of x such that f(x). jS). the work proceeds as follows. /(a)//' (a) =0-04. and f'(x)-x zx (2 + 2 log e x). /'( ai ) = 8-324. . By repeating the process. and then it is important to know under what circumstances a t 11 is a closer approximation to the root than a. the approximation can be carried to any required degree of accuracy. f'(x) and /"(a?) are continuous in the interval (a. Starting with a . and <x 2 = l-4577 . if /? . Put a = x*. and a 3 = l-457669. Now oq is certainly a better approximation than a if o^ lies between a and a + A. Fourier's Rule.+0-375 when 3 = 1-5. /'(a) =9-487. the error in taking o^ as the value of the root is of the second order of smallness. the value of a required is (1-457669)* = 2-12480. and h 2^0 according as /(a) and /'(a) have different signs or the same sign. such that a A/a = 3. so that *=-/()//'( + W Hence if t -/(a)//'(a). Now where 0<0 1 <1. Also /(ar)=-2 when z = l. /(a) =0-375. unless k/f'(<x) is large.For f(x) increases with x or decreases as x increases according as /"()0. /'(a a ) =8-262. the value of the root is ocj -f e where Thus. Find an approximate value of a. where 0<0<1. Hence FOURIER'S RULE 457 It follows./(a + h) =/(a) + hf (a) + &*/" (a + 0h) . We shall assume that neither f'(x) nor f"(x) is zero for any value of x in the interval (a. and since a 1 =a + /.6. /(a 1 )// / (a 1 )--0-0023. i .)= 0-000255. I. f(*jlf '(**)= 0-000031. jS) is not very small. Ex.

458 VARIOUS CASES The various cases which can arise are illustrated in Fig. +. in those on the right they have opposite signs and Fourier's rule does not apply. 66. We shall now show that a a is a better approximation than a.Thus we have Fourier's rule. then a x is a better approximation than a. 66. (i)-(iv). and /(a) and f"(a) have the same siyn. It is to be noted that the tangent to the graph of f(x) at the point where #=a cuts the x-axis at the point a x . provided that the interval is sufficiently small. . which is as follows : // neither f (x) nor f"(x) is zero near the point a. In the figures on the left./(*)- a a a a cn+h FIQ. /(a) and /"(<*) have the same sign.

. Hence t and KI is a better approximation than a- .. J5te.. 2. /"(2) = 12. Let /(x)r. th-n f'(x)=3x* -2.The true value of the root = a 4.. and therefore <x x is a better approximation than a if |J<|A|.3) is 12. in cases where Fourier's rule does not apply. Moreover. ' .. and a t is a closer approximation than a if ... and we draw the following conclusion : If neither f (x) nor f"(x) is zero in the interval (a. Taking a 2 and applying Newton's method. that is if f"( JLA2 / \ or if | <2 If /" (x) ^t in the interval (a. /"() -60:.. the root lies between 2 and 3.x 3 -2a.\f"(<x + 0h) \. (C) In this case. by (A).. EVALUATION OF ERROR 459 This is certainly the case if ..-5.| j3 -a | < ' ^ .. /J) and is a better approximation than a. Here /(2) and /"(2) have opposite signs. ^3). Since /(2)<0</(3). we can find a positive number I such that 1 1 \<. and Fourier's rule does not apply. / / (2)=:10. if Z|/(a)|<2{/'(a)} 2 . Also /(2)= -1.h =! + . ai =2 -=^-= 2-1. But /i<0-l. Find the real root of X s = 2x + 5 to nine significant figures. that is.and the least value of |/"(#)| in the interval (2.. | h \<\ t\. .

-0-000016557. Hence the error in taking 2-09455 as the root is numerically less than 0-00002. 460 NEARLY EQUAL ROOTS Ex. and so the root lies between 2-09457 and 2-09453. . giving ai =a -/(a)//' (a) =2-094551483 . . J Area donkey can graze over segment A BB' + segment CBB' ... * This approximation might have been obtained graphically.. /'(2-1) -11-23. and therefore /"(a . and BO the value of the root to nine significant figure s is 2-09455148.* we have /(2-1) =0-061. Hence ai ---2-l -Pl -^? =2-094569 . and consequently | | <~ (0-007 )2^|~|< 0-00003. /"(#) increases with a:. LOAB^x' measured in radians. 67 AB is the stretched rope. e is negative. Also | \<l (0-00002) 2 ^<10~ n . J jL'2ij Also the value of h is equal to that of in the first approximation. Thus 2a' + *= w . Again.. 24. . A donkey is tethered to a point A in the circumference of a circular fiel d of radius a by a rope of length L Find the value of I in order that he may be able to graze over exactly half of the field.. In Fig. taking a -2-1. Z 11 Moreover. c is negative.Moreover... 2o B. taking a 2-09455. 3. Next. and /'(a) = 11-1 1607. Moreover. we find that /(a).... / // (2-l)-12-6. LAOB=x.C. so that | h | < 0-007. A.

approximately. But the roots are separated by a root of /' (x) ~ 0.1) 0. Hence we have . therefore /"(a) i s positive.tan x } . Equating this to JaV. Consequently /'(a) is small. by Art. 6=/'(a). giving Newton's rule. j8).) have the same sign and. by Fourier's rule. Further. If x increases from to ir/2. Neglecting squares of /?. Nearly equal Roots. and Ifa ^2 sin 35 25' = 1-16. Taking a ^= 1-239. f"(x) increases from ?r/2 to oo . therefore / / '(x)<34 if x<ac.\ sin 2x'). then where a=/(a). so that /' (x) is zero for some value of x in the interval. etc.70 50 X . In a question of this kind it is important to start with a really good approxima tion. /'(a) ^-5-578. Suppose that the equation /(z) = has two nearly equal roots within the small interval (a. and then by measurement Z_^O# = 71^l-239 radians (nearly).\ sin 2x) + l*(x' . we have a + 6A==0. \ ** j and f f/ (x)~ n sec x(\ + 2 tan 2 x)-2 tan x sec a:. and the next approximation is i ^ -/(a)// 7 (a) = 1-239 -. and let oc + h be the true value of one of them. | c |<J(0-003) a -^<0-00004.tan x -f r. HORNER'S METHOD 461 This equation has nearly equal roots. and is therefore approximately the same as (a + |6A) 2 = 0. we can show that /"(a)<34 . Hence.0-003 -= 1-236. Thus the following results are correct to the last figure : L AGE = 1-236 radians . to be found graphically or otherwise. 9 (B). Hence. We draw the figure so that the shaded area s are nearly equal. and the term ch 2 cannot be neglected in comparison with bh. /(a) ^0-018. it will be found that x is given by f(x)~x. Hence /7a/6. ( I r sec x . Hence /(a) and / " (a.= a 2 (x . if a x + is the true value of x. very nearly. /' (x) tan x ( .sec x . Neglecting cubes of h. 10. (?-/"(). so that Newton's rule does not apply. we have a + 6A + cA 2 = 0. a z is a closer approximation than a. we find that.

This follows from the fact that <(#) = has a root between q and q + l. we avoid the use of decimals by multiplying the roots of (B) by 10. is the most convenient way of finding approximate values of the irrational roots of equations of the type /(z) = where f(x) is a polynomial. <f>(q) and <f>(q + I) have the same signs. The root is evolved as a decimal. To find the negative roots of f(x) = 0. obtaining the equation <f>(x) = a Q x n + lOV? 71 . namely q-rs .l + . Decrease the roots of (A) by p. and there is no root between and q. 11. which is nearly the same as (a + |M) 3 = 0. 7 3a b c . .. we find the positive roots of /( . s . In finding the value of q. In what follows. is definitel y given by Newton's method. the figures of which are obtained in succession.. Any rational roots which may exist can also be found in the same way.. One great merit of the method is the very concise and orderly way in which the reckoning can be arranged. j8).. It is only applicable when 2a/b and b/2c are nearly equal. 2. the value of one of them being a-fA. (The case of two nearly equal roots will be considered later. or preferably figure by figure as in Ex.. T = "2c fcPP 1 " *-)' This result is used in Homer's method (Art. where p is a positive integer. . we are only concerned with positive roots. q is the greatest number such that if the roots of (C) are decreased by q y the sign of the last term of the transformed equation is the same as that of th>e la st term of (C).l b 1 x + W n b n = Q (C) This has a single root between and 10. Ex...x) =0. The process described in this article. namely 0-qrs .. XXVIII. we have a -f bh -f ch 2 + dh 3 .. after a certain stage. . where q. 3) to suggest a trial value of h..i 2a b f . due to Horner. (A) and suppose that this has a single root a in the interval (p y p f 1 ). * --y---. so that <(0). .. Consider the equation f(x) = a x n -f a^~ v 4.-35 (nearly). Jn the same way. Homer's Method. and let the transformed equation be a<p n + b l x n . r. -h a n ^x -f a n 0. 462 HORNER'S METHOD FOR CUBE ROOT In practice the value of q is suggested and. 1 1 .. . are the figures in the decimal part of the root : then p can be found as in Ch. Moreover.+b n _ l x + b n = (B) This has a single root between and 1. neglecting A 4 ..1 * ..) Let (x^p-qrs . if the equation has three nearly equal roots in the interval (a..+I0 n . 2 below. Hence.

. 0+ 0-111(4 (a) 4 + 16. Ex.47 8+48 12 120+4800-47000(8 (6) 128 + 5824. so obtaining an equation of which r-s . is a root.5.408000(0 (c) 14400 +69120000 -408000000(5 (d) 14405 +69192025 . -f(p)!f'(p} -&/&_! (approximately).For. Having found the correct value of y.111=0 (A) The work is arranged below in a form suitable for explanation. Continuing as above.62039875 14410 +69264075 14415 144. regarding p as an approximate value of a. we can find as many figures as we choose in the decimal representation of a.62039875(8 () 47. In practice it wo uld be arranged as in Ex. we decrease the roots of (0) by q.15 + 6926407... I. we have 0-qrs . The explanation is given on the opposite page. When a certain number of figures have been found. about as many more can be found by a contracted process explained in the next example. multiply the roots of the resulting equation by 10..408 136 + 6912 144 1440 +691200 .Q 48 u y 11=0-8. 2. Find to ten significant figures the real root of /(*)= 3 . _ O ..

by Newton's method . 4-80<a<4-81. the resulting equation is (B) If a4-fA. EXPLANATION OF PROCESS 463 NOTE.9 692. 4-8<a<4-9. EXPLANATION OF THE SUCCESSIVE STEPS (a) Decreasing the roots by 4.. 4-805 . 4-8. -0-000000004 ....7 6928.1 1.2 6. etc. Also that for the values 4. Denoting the root by a. . -0-062 .9 .3 692884 692897 69289. with possibly an error in the last figu re.8 69.44 6927560 6928713 + 692871. f(x) = -47. the successive steps show that 4<a<5. 4-8058955338 of x.6619395 6619395 ( 9 (/) 383439 383439 (5 (j 36990 ( 5 2345 ( 3 267 ( 3 59 ( 8 4 Hence the required root is 4-8058955338... . -0-408....

(f) This step is similar to (e)... and this proves to be correct..... (c) Here the roots of (D) are multiplied by 10..... the sign of the last term would be c hanged.. two figures from.. (E) and the next figure of the root is 8..62039875 = 0... ... we should cut off three figures fro m the coefficient of x 3 . We have to multiply the roots of (E) by 10 and decrease those of the resulting equation by 8... (e) The equation found in (d) is ar 3 + 1441 5.... we write down the last term of (E) as it stands. we have Hence we may expect that the next two figures of the root are and 5. tJiat of x 2 . (C) Taking 9 as a trial figure.. and thus a root of (f> (x) lies between and 9........ for decreasing the r oots of (C) by 9 changes the sign of the last term ... and the process is simply contracted divisio n. i nstead of introducing O's.... In the case of an equation of higher degree.. (d) The roots of the equation obtained in (c) are multiplied by 10. Trying 8......T 2 + 69264075z . this shows that <f> (0) and <^ (9) have opposite dgns........ the resulting equation is (f) (a?) ^o.. and so on... NOTE.. 3 + 120a.. and those of the resulting equat ion iecreasod by : if they were decreased by 1... and we find that 5 is the next figure.. and neglect the coefficient o f a: 3 .. . (g) Only the last two terms remain. and decreasing the roots by 8 . (D) Denoting the root of (C) by 8 -f k.. we find that this is the correct figure....... the coefficient of x.. 2 +4800* -47000-0 .(b) Multiplying the roots of (B) by 10. 30 is the next figure of the root... ... In doing this the important figures will retain the same relative position if.......... we find that this is too large. In the multiplication allowance should be mod 1 J r the figures cut off...408 .. the resulting equation is s 3 + 144s 2 + 6912&. cut off one fi gure from.= .

x .x* . 1 -1 9 -1 -2000 89 -1110 10 2 19 279 | -428000 9 29 | 341 -32621000 6 31 40700 -4139264 8 33 43931 -321124 . Find to nine significant figures the real root of x* . 2.2000 =0.464 TWO NEARLY EQUAL ROOTS Ex. In practice. and the first step is to decrease the roots by 10. It is easily seen that the root lies between 10 and 20. the work is arranged thus : |f =0-7.

2 ) 34568 ( 7237 .38 1135 180 37 .8 3776 477571 3782 37.88 47759.4 4776.6 350 4724300 -34568 359 368 4746956 476964.8 3770 477267.

81 -167 12 . 3. NOTE.120000 ( 2 .390625 . to nine significant figures.5712 . Find approximate values of the two roots of 3s 4 .The required root is 12-9686724. The reckoning is as follows.61X 3 + 127z 2 + 220* -520-0 which lie between 2 and 3. and the explanation is on the opposite page.. Ex. 3-61 +127 +220-520 (2 (a) + 254+ 92 + 17 .(6) -55 -49 -43 -37 3-370 -16700 +92000 -364 -17428 +57144 -358 -18144 +20856 -352 -18848 -346 3-3460-1884800 + 20856000-57120000 (5 (c) .1902025 + 1 1345875 . It will be seen that Newton's approximation gives 1 more than the correct figure until we come to the 8.3445 .

0 + 1993..8 .( .(d) SEPARATION OF ROOTS 465 For the smaller root : -194.6-641 (0 (e) 199.6(3 .5 + 175000 -390625(4 -19376 + 97496 .62 (3 (g) For the greater root : -194.6.4 -19362.19390 + 19930 -19404 . (h) -18.04 + 1993.9' -149.7 ..641 ..9 +1582 (9 .(/') ...167 f 79 1..641(9 (') + 247 +1582 -1499 -1....62 + 187 18..4 79 (4 .-3430-1919175 + 1750000 -3415-1936250 -3400 -3..3 -641 (3 (/) + 193 .

one change of sign being lost. (6) Suspecting two nearly equal roots.6 =0..e. beginning at the stages (/ ).. and the other between 2-2549 and 2-255 . the signs are +----.. this has two roots between and 1. and the separation of the roots is a matter of trial. EXPLANATION OF THE SUCCESSIVE STEPS (a) Decreasing the roots by 2. the resulting equation is 3a^ . and the fig ure 2 is suggested by . Here the rule of Art. Hence.37z 3 .-1. 20856 nr ~ 20856--' and 2rT8848T ' 5 -Thus the roots lie between 2-254 and 2-255. the sig ns of the terms are +-. Thus one root lies between 2-2540 and 22541. The positive root of 2x 3 . (e) and (e f ). . and when they are decreased by 3. 2.. and 2-2549942... So far.. (/'). (c) and (d).. If the roots are decreased by 3. where the last figure in each case is not to be relied upon. If at (e) we decrease the roots by 1 (instead of by 0). But when the roots arc decreased by 2.12 =0. The figure 5 in (c) is sugges ted by 2(-5712) __ . it will be found that again the sign of the las t term is . we have no assurance that 2 is the proper figure.. 8.8 5(2 Thus the roots are 2-2540333. the two roots lie between 2-2 and 2-3. The same is true if. Immediately after the roots have been separated. The two positive roots of x* . as at (e'). the sign of the last term of the resulting equation i s .. and the roots are separated. 10 fails.167z 2 + 92z . i.3x .+ +. Decreasing the roots by 2. by Art. the signs of the terms are + . then.. the proper figure is suggested by Newton's rule. we apply the rule of Art. we decrease the roots by 9. so that two changes of sign are lost. 466 APPROXIMATION TO ROOTS EXERCISE XLIX Find by Newton's method to five significant figures : 1.+ . Similar remarks apply to these steps. 10.Ix -f 7 ^0. By the question.

6..3 1#+ 1609 = in the interval (13. 20.2x* .0. The two roots of 2* 4 -27z 3 + 25z 2 -j-179:r-275---0 in the interval f2.2).782 = in the interval (4. The two roots of 2x* . 11.3). The three real roots of x* .12z 4 4.2x 3 . 2 -f 12^ + 9^0 in the interval (1.5). 19. 18. 3) ..6Lr 3 + 127* 2 -f 220* . 3 4.Wx z .1 10.15)..3x = 8002. 22. 10. [Here x is given by f(x)~x~ l ~log. x* -= 9-9995 (approx. it will be found that x 2-5 is a good approximation.207* .10z 2 ~ 6x + 16 = in the interval (1. 6). 12. Five appli- . . The two roots of 2* 3 + ll# 2 .. showing that x -2-50616 (nearly). The cube root of 4129. The two roots of # 3 -2(Xr 2 .10r 3 .13s 2 llx + 133 --0 in the interval (3.Sx + 166 =0 in the interval (5. and that/(z)<0 if 0<o?<2-177. show that the equation f(x)-Q has just one solution other than zero. (ii) x 5 + toe* .59z 3 + 150* 2 -f 201* = 207. If f(x) = e x . Verify that for this val ue of x. 9.97 = 0.34x + 40 = in the interval (1.10z-f 1=0 in the interval (0. 17.2x rr 962. 16. 14. Show also that f(x)<0 if 0<x< 1-260. The greatest and least roots of 3s 4 .llx 2 -32x-f -40 = in the interval (1. the solution being a? =1-260 . where the last figure is not to be relied on.).177 .520 = 0.Find the roots indicated below correct to seven places of decimals by Horner's method.1 . 4).1). The two roots of x* ./(x)=0 has just one positive solution given by x = 2.2o. 4. The two roots of 3x 4 .2). 13. (in) x 5 4. 21.2).r--0. Find an approximate solution of a:* =10. . 3.2). The real root of a. If f(x)~e x -e~ A -4x. The two roots of Ix* . The greatest root of the following : (i) x 5 + 4# 4 . The two roots of x* + 5$x*-388x + 511=0 in the interval (2. 7. By drawing the curves y~xr l and t/=log 10 #. The two roots of 2z 4 -16z 3 -17a a + 392. show that the equation . The two roots of 3z 4 . The positive root of 2z 4 -f 5a. 3 + 4z 2 -f.10* 2 . 4 -4^ 3 ~ 12o. 5.1 12x 2 . 15.2#. 8. The two roots of 3a.1 =. 3 + 29* ..3z 4.

... and cannot be considered here...... In particular.. y) =0.. as generally happens.. in ascending powers of x...... then y is called an algebraic function of x... It is required to approximate to the value of y.. Let f(y + h) = f(y) + A/...... If two variables x and y are connected by an equation represented by /(#....... But if. if f(x. let y = a-fa where a is 0(x). The following process is of great practical use... we are unable to express y explicitly in terms of x............ y and /(x..... however.... we are not justified in assuming without further enquiry that y is really a function of x in the sense hitherto understood. y) = does define y as a single... Some exercise in tracing curves from their equations will...or multiple-valued function of x..a.. .cations of Newton's process give x 2-50616. etc..... A Rule for Approximations... convince the student that... y)=0..........1 CHAPTER XXIX IMPLICIT FUNCTIONS. Implicit Functions. and so defines ?/ as a two-valued function of x... Suppose that J/ = ^ + ^/(?/) *~ x2 </>(y) +x?i/j(y) + .... 2... y} is a polynomial in x...... at any rate in a large number of cases. 2 )... Taking as a simple case the equation this gives y = x t j(1i ... (A) where x is small and/(y)... . the equation f(x........ J The first approximation is obtained by putting x=0. and affords excellent practice in finding approximate values. The general form of a curve represented by an algebraic equation can generally be found by pure algebra.... we say that y is an implicit/unction of x... <(?/). (0) 468 INVERSION OF SERIES For the second approximation. do not involve x.* then ... (y) + ..... The general question involves difficult analysis. CURVE TRACING 1. giving y.

.... in equation (A) substitute the result of the rth approximation in xf(y).... ... the (r-2)th approximation in x 3 ifj(y) y the first approximation in the term containing x r .. Ex.. the second approximation is y = a + xf(a)....e....... ... (E ) which gives <Ac iAird! approximation....y = a + xf(a+<x..... terms containing XOL.. etc. neglecting small quantities of the fourth order..... ... x 2 .. Next let y = a-ha-fj8 where j8 is 0(x 2 ).... y==a + xf(a + a) + x 2 <f> (a) ... . we shall find that . he deduced that 1 a 8 1. to find the equation giving the next approximation. 1..... namely Thus jS-x 2 {/(a) .} + x 2 {</>(a) + ... (F) which gives the fourth approximation.3 z 5 ........ . i... then y = a + xf(a -fa -f j3) + 2 <(a -ha + ]8) 4.. 2 )~i to prove that sin~ l x=xPutting x sin ?/. Similarly... if y is 0(x 3 )....... and neglecting small quantities of the second order.) + 2 <(a-fa) + . .. (D) so that oc xf(a). Newton used the expansion of (1 ~a....} + ..... = a-f x {/(a) -fa/! (a) + .. and neglecting small quantities of the third order./ x (a) + ^(a)}.... Hence the following Rule : Having found r successive approximations to the value of y..... the (r-l)th approximation in x 2 <f>(y)... and neglect terms containing powers of x higher than x r ...

... is small...3 -f ^ ..U'..... If fj. (C) If y is small.. 2.. so is sin y. * This means ' of the same order as a...... so is y(y ...... this gives and so on...... write equation (A) in the form 7 2/... therefore y is nearly equal to or 1... 7......... Ex. find approximate values of y which satisfy the equation If fj... is small... and the first approximation is Substituting y for sin y in the second term of (C) and neglecting higher terms....l)(y + 1)......... II. (E) Following the rule.....) ROOTS OF A CUBIC 469 To do this. the third approximation is given by siny^y-^^ and neglecting the powers of y higher than (F) The fourth approximation is given by Expanding and neglecting powers of y higher than y 7 ..... Wri te the equation (A) ..<B.. + (A) 1^..... the second approximation is y-%y* ..g .' (Ch.

For the second. substitute p? for y 3 in (A). Theorem.... giving y^-iM-lf* 1 -^ that is.... put Y ~y . q are real numbers (Ch..i^ .....^/LI.... the roots of y 3 -y+^ = are where + '" + \n V2 Write the equation in the form y*-y = 2x.... (ii) If y ...ift 8 . The last example suggests the possibility of expressing the roots of equation (A) as power series in //..... y =fi + (p... + /r 1 ) 3 . 5.. // /^ 2 <4/27.f M 2 ..... If all the roots of ax 3 -h 36x 2 -f 3cx + d Q are real.. the first approximation is 2/ = /x..1 is small.(i) If y is small. (A) by a transformation of the form x=p + qy. For the third. then equation (A) may be written The first approximation is Y = ... then approximately This result also follows from the preceding. XII.Jft)= ... (iii) Similarly we can show that if y + 1 is small.. Further approximations can be found in the same way. giving y -p + ^ + 3^ 5 + 12/z 7 .. 470 ROOTS OF A CUBIC 3.. Ex.^ + 3/z 6 ) 3 . 4. Cubic Equation with Three Real Roots... giving y = //. y = /x -f (p... F= -J/i-|( -i/x-f/x a ) 2 -i(-^) 3 . The second is y = .J/i -f ( ...p? + 3/A For the fourth.J/t -f /x 2 . where p. 2/ = 1 . 1)... (A) . giving 2/=/ For the third. the equation can be reduced to y 3 -t/+/x = where /x 2 <4/27 ..1.... 4.. . for the sum of the roots of equatio n (A) is zero.

. .. Differentiating again. (^-k)y n+2 + (2n + l)xy n ^ + (n^^)y n ^O . by Taylor's theorem..&)i .. (i)J/ (yJo 885 . Hence from (E) in succession..j = v/9.3uvy = u* + v 3 .^)^w 2 + xu^x* .. Similarly. these values are given by x 2 x n where y n = ' / r: and ^)o (^i)o (^2)0* are the values of y.. ^/S = w/9. d n u d n v Writing u n = -=-^ . (C)* * An alternative proof of this result is indicated in Ex.... giving (x 2 . then y 3 ....... and if n .1 ) we have (y^o^l. we have therefore (x 2 .. XXIII. 6 of the next Exercise .. Assuming that the values of y can be expressed as power series in x. (E) The possible values of (y) are 0. and differentiating with regard to x.&)*. 1. and similarly (x 2 . v n = -7-^ .v/3. where & = 1/27.&)% = .. y l9 j/ 2 . ax n when cc = 0..jfc) w 2 4. we have for n = 0...fc)i. Thus we may take w 8 = x + (x 2 . 2... (yn +2 )o = 3(3w + l)(3 M -l)(y n ) ..... 2 .....fc) = u/ 3 .. (x 2 i) v 2 + xt...x^i = (x 2 .. .......where x= -/z/2 and x 2 <l/27. (Of course... 11).......1 * since yi* 2 /^ 2 . 1. (D) Putting #=0 and = 1/27.(a. we have (x 2 . This equation will be identical with (A) if w 3 + v 3 = 2x and 3uv ...1 .... x is to be made zero after differentiation.. Since y u + v.. EXPRESSION AS POWER SERIES 471 Differentiating n times by Leibniz' theorem (Oh.... Let y = u + v.) The assumption will be justified if the series is convergent...k)~% = Wj/3. it follows that (x 2 -k)y 2 + xy l ~y/ . v 3 = a? .

^^x 7 + . Therefore y = l -F(-x). which includes this. y are both small. prove that approximately y=x + ax n + na 2 x m ~ l + f n (3n . 7. . we have Hence the series u 2n . (ii) // (y) =l. 5. If 8. then (y 1 ) i =l. the root corresponding to (y)o ss O is given by y = F(-x)-F(x). If y=l+xa v and x is small. Putting x= -/x/2. we have the results stated above. . ..6 . the sign being + or .^ tan 7 as + .2 . x 3 a 4. ?/ are both small.(3n-3) 3 .Thus y=-l+F(x) where . t are positive integers (>s> t) and A is small.. . . EXERCISE L 1.^ - for such values of x as make the series convergent. prove that tan x . If y=x + ay* + by* and a. .J tan 3 x + -| tan 6 a: .x + f x 3 + ^-a.. . L. . then .1 ) a s a: 3n . . then approximately y = 1 -f xa -f x z a 2 log a + f a^a 3 (log a) 2 . . Denoting the series by % 4.Hence the numerical value of (y n ) Q is the same as before.u z + u 3 + . A more general theorem.approximately ?/ = x + a* 2 + (6 + 2a 2 )x 3 + 5a (b + a 2 ) z 4 . 5 4. the equation y*-y* + (-OA = has a root nearly equal to l-X-(-l+8 + t)-(-l+2s + t)(-l+s + 2t). are convergent if x 2 <l/27. (n + l)(n + 3). If y x + ay n where x. 3. Eu 2n+v and consequently also u n . 2. is given in Ex. (iii) Since the sum of the roots of (A) is zero. 472 APPROXIMATIONS AS POWER SERIES 4.according as n is odd or even. Having given that x = tan x .

12 (t/- (ii)y. (v) If m = 5 and A. vhere (t/) is a value of y obtained by putting #=0 in (A). -^~ .= where y 1 ~-~. and then L + ? VP Mn-^ K * + | 3 'H "^^ 16'F I L. dii d 2 y are the corresponding values of . If y*-y=2x. and (y^. (t/ 2 ) . one value of (y) Q is 0. . show that Hence prove that equation (B) holds if x 2 <k. y t = ~. show that 2 -l) = (32/ 2 ~l) 3 24*. 6 =^. . v m ~x-*/(x 2 -k).I . . 1. (iv) If m is odd. u>x dx (iii) Using Ch. Ex. is sufficiently small. . then (i) x and y are connected by the equation m ' ' ~ 2 ~ li (ii) If a. 11. A/f g . . m being a positive integer. . equation (A) becomes y 6 and the values of (y) Q are 0. XXIII.. Deduce that 7. If y=u + v where u m =x + ^(x 2 -k).ii li 6.

In the examples which follow. . ' F 68 If a = and ^ = 0. At such a point there are two tangents.. Except in special cases (see (iii)). So in this case the curve has two parallel asymptotes.. or node.. . DOUBLE POINTS 473 4. each meeting the curve in three consecutive points. Any other straight line through D meets the curve in two consecutive points. is an asymptote.. XVII. so that near P the curve is indefinitely close to the tangent.. and the line meets the curve in one point at infinity.(D) where a curve crosses itself is called a double point.. 20). If P moves to an infinite distance along a branch of the curve. If P is a point on the curve and Q a point on it near P. tangents and asymptotes are found by a process of approximation.. the line meets the curve in two points at infinity and.. It may happen that an infinite branch has no asymptote.. If three roots are equal. If w=0 is the equation to a curve of the nth degree and y =/*# -f v that of any straight line. the abscissae of the points of intersection are given by an equation of -+ . the line meets the curve in * two coincident points ' and. and say that this is the equation to a curvilinear asymptote.. We then find the simplest equation which nearly represents the branch at oo . one root of equation (A) is infinite. (i) If this equation has two equal roots. this limiting position is called an asymptote to the curve.Thus all the roots are real.y 5 -f f i/ 3 . (A) obtained by substituting px + v for y in u = 0. (ii) If a = 0.. Intersections of a Straight Line and Curve. the tangent at P is defined as the limiting position of the secant PQ as Q moves up to P. except in special cases (see (iv)). the line meets the curve in three coincident points.) If the double point is at infinity. and they can be expressed as power series in x if ! &' ( vi) If m = 7 and F f = |.. (Fig. 5.. At such a point. except in special cases (see (iii)). 69. the curve crosses the tangent.^i/ = 2x. the equation is y 1 .... Thus as Q moves to P.. the point of contact is a point of inflexion (Ch.-fa n = 0... the tangents are parallel straight lines... Tangents and Asymptotes. (iii) Any point. the distance of Q from the tangent at P becomes indefinitely small. touches the curve. each meeting the curve in .. and the tangent at P tends to a limiting position.

in Fig.a) . 474 CURVE TRACING If the tangents at a double point coincide. move the origin to that point. FIG. proceed as follows.at which the tangents are imaginary. as shown in Fig. the point is called $ cusp. Curve Tracing. 0). (iii) Find the approximate form of the curve near some of these points. given by x = 6 -f m 2 . Near (0. 0) (if on the curve) x. and. (iv) Fig.three points at infinity. 69. symmetry with regard to either axis. of the oval in Fig. the points (if any) where it cuts the axes. To find the general shape of a curve represented by a given equation. (i) Consider any symmetry which the curve may possess. 71. e. 71. To find the form near any other point. it is important to be able to account for all the n points in which any straight line meets the curve. 71. and at a point P. 71. may be considered as the limiting form. such as A. Any straight line parallel to the* asymptotes meets the curve in two points at infinity. is called a conjugate point. is a sketch of the curve whose equation is y* = (x-a)*(x-b). y are small.g. 70. which is a sketch of the curve whose equation is 6. A is the only point on the curve to the left of the line y = b . FIG. AP meets the curve in two coincident points at A . on eliminating y t we have (x a) 2 (x -b ~ m 2 ) = . Therefore. (i). (ii) Mark any points on the curve which can be found easily. The quickest way of finding the tangent at any point is to find the value . The equation to any straight line AP through A is y = m(x . The conjugate point at A. In tracing a curve. There are two kinds of cusp.g. A point on a curve. or in opposite quadrants. e. if no other point in its immediate neighbourhood is on the curve. 0) and B(b. and thus A may be regarded as a double point. when a a'. remembering that imaginary points occur in pairs. The axis of x meets the curve in A (a. (ii). hence. (i). but not necessarily of the same order.

2. (vi) By considering the value of ~ . find the tangents (if any exist) which dx 70 are parallel to the axes.e. JS) .of T. where m and n are positive integers.) (v) Find the form of the curve near infinity. In doing this. APPROXIMATIONS TO FORM 475 (iv\ Find any regions in which there is no part of the curve. we cannot say on which side of the tangent the curve lies. or discover any peculiarities which the point may possess. but not necessarily of the same order. y x x f O y - y O x FIG. as in Ex. (Exx. it is better to use a process of approximation. at a point of inflexion. without further different! ition. Notice that. The reader should be able to draw the 'parabolas' represented by y m = x n . 1. and the approximate form of any ' parabolic ' branches. i. we find the (linear) asymptotes (if there are any). 1.But. So. For example. for a/etu points. 72. where one or both of x and y are large.

73.* ' 7.1 )y = (x 2 . . even w hen the latter can be done. 1) are on the curve. y 476 APPROXIMATE FORM AT INFINITY Near (0.] 2 H B. FIG. so that the student can verify the conclusions arrived at by the methods just de scribed.y + x (A) Near (0. Trace the curve ( if . 0). 0). Thus in each case the curve can be ' plotted ' to any degree of acc uracy required . ~y for x. [See Ex. showing that the curve is above the tangent in the first quadrant and below it in the third quadrant. (0.. Ex. (2. (2. y~4=x. 0} is a point of inflexion. or again. obtained by putting (4#) 3 for y 3 in (A). 0). will be obvious. x 2nd approx. equation (A) becomes 2 7 + 3 Y 2 + 7 s = (x* 4) x.1. 7 and in Exercise LI have been chosen so that one o f these methods can be used. when only the characteristic form of the curv e is required. y ~ 4x + 63^ 3 . This is the equation to the tangent at (0. putting Y = y . x and y are small and of the same order. 1st approx. 0). Thus (0. We can then find any number of points on the curve and the gradients at these points.4) x (A) The equation is unaltered by writing ~x. the advantages of the above method over the method of plotting. The points (0. we may be able to express x and y as functions of some variable /. However.. and gain confidence in using them.d y dx 2 '' (vii) It may be possible to find y explicitly in terms of x or x in terms of y . Hence there is symmetry in opposite quadrants. 1).C. A. y. Write the equation y 3 . Examples on Curve Tracing. * The examples given in Art. 1. 1 ). 22 of Exercise LI.x 3 .

(approx.*)/(! . y . Thus the curve is as given in -Fig. The tangent is parallel to Ox where x=2/\/3=l-16. on the curve y~x-.1 3 ). and the tangent is parallel to Oy. 2 J = . Near (2.x= This shows that. giving y~-x~Q.x 0.Gz 2 (approx. 1). which shows that the curve is below the tangent.). 2 7 -f. 1 ). 75.2x. it will bo seen that any number of points on the curve can be found by giving any values to t in the equations y = tx. Near (00.2# .. showing that the curve is below the asymptote on the right and above it on the left. which is the equation to the tangent at (0.y \ FIG. 2nd approx. . as :r~> oo . From (A). which is therefore an asymptote. Again. and on the asymptote y = x. 74. putting X=x-2 9 we find that t/ = . where y~ dil/s/3.(3x 2 . dyjdx . the gradient of the tangent is 4. If l</<4. 1st approx.6# 2 .0). showing that the equation to the tangent is y -S(x -2) and that the curve is below it... giving y = 1 . and approximately 2/ 3 -a^~0.8z .).4z.3 ( . Therefore. the corresponding point is imaginary. x 2 = (4 .2#) 2 . thus no part of the curve lies in the space swept out by a line turning round from the position y = x to the position y = 4x. x and y are both large and of the same order . the curve becomes indefinitely close to the line y .1 ).4x or y = I . at each of the points ( 2.0-58.4)/ (3?/ 2 .00). Again. y Putting y ~tx in equation (A). 2nd approx.

y for y . hence no part of the curve li es between the lines x = l. 2. x 2. x . 2 2x 4 showing that y -x is a tangent at (0..-. Taking y =x as a first approximation. 2^2 2 I 2 Near (1. Write the equation in the form * 2 )=0 (A) N ear (0. Also y is imaginary if 1 <x<2 . y '" 3 2 if ' . Trace the curve (x-l)(x~2)y 2 = 2x z (A) The equation is unaltered by writing .^ ~ showing that x . the curve is approximately given by y z -x 2 = or y~x. Again.x) .1 =0 and x .{i " 1} y3 .~r~ (#-")?/ (1 -2)# 9^2 O 92 ^ear (2. ~ " . henco the curve is symmetrical with regard to Ox. from (A) 2 (y . . 0) and that the curve is above the tangent . writing equation (A) in the form x 2 (y z .2 =~0 are asymptotes. oo ). FIG. r . 0) which is a double point. the curve approaching the first from the left and the second from the right.1 .x) = (3*2/ 2 .re V ) j(y 4. 76. Similarly y + x~Q is a tangent at (0. 0).2 .2) = I near (oo >s /2).ASYMPTOTES PARALLEL TO AXES 477 Ex. GO ). giving as a second approximation 1 3z 3 3 . -.

478 FOLIUM OF DESCARTES 8. the line y . by retaining only certain terms in its equation. or near infinity. \i FIG.. /. -^2). The same is true for the line a. Also. . where /<! or >2. .2y 2 . as in Fig. In the next example we consider the possibility of finding the approximate form of a curve near the origin. The asymptotes parallel to Ox can be found by equating to zero the coeffic ient of the highest power of y in (A) : similarly for those parallel to Oy. for all va lues of k except 1. 78. 77. y + N/2 ^ - We can now sketch the curve.8 I 1 I i x y*' tes.k meets the curve in two finite points and in two points at oo . NOTE. it will be seen that 3 and near (oo. the H J2 1 ' I I \.

3 -)-?/ 3 0. Thus near ( oo. showing that y -f x = .3o z .. as a first approximation ?/-ha: = 0.... we get a. Similarly. 0) is represented by x 2 .Ex.a is an asymptote...... we consider the equation t/ 2 ..2 -^-^ = ...a 4. for if y is O(x)... This represents the curve near (oo .. thus xy ~x(x-\-a) f " 2 (the last step being obtained by long division).Say = 0. x is small and 0(y 2 ) ..a for y...... oo).... Hence the third approximation i s a 3 y + x ~ .1.. (A) Omitting the first term. . Disregarding the factor y... 0). y -f x 3a .. A third approximation is obtained by substituting .a. we have y 3 -3axy=0. oo ) .. Trace the curve x* + y*-3axy-Q (a>0) .... Hence (B) approximately represents the curve or part of the curve near (0..... and can be neglected in comparison with the other terms of (A). by omitting the second term. and that the curve ..Sax = .. 79... and can therefric be neglected in comparison with the other terms in (A). (B) If y is small. . we find that part of the curve near (0. then Saxy is O(x 2 )... . O FIG.x ... Omitting the third term. so that r* is 0(?/ 6 )..* _ /V.. SAJ For a second approximation.....

or. Consider the curve represented by t/ = where u is a polynomial in x. cx r y s . -a \ o \ \ \ -a FIG. 81. y The curve is symmetrical with regard to the line y x. 9. A rule which goes by the name of * Newton's Parallelogram ' enables us to choose these terms so that v = may represent the curve. be any terms . . then s = 3a//(l -f J 8 ). bx^tf 1 .g. < = 1.. y. if y tx. when such parts exist. near the origin or near infinity. ORDER OF MAGNITUDE OF TERMS 479 H K Fio. so that any number of points can be found by giving different values to t . e.approaches it from above at both ends.. xy = 3a/2. Newton's Parallelogram. part of it. Further. Let axy n . 80. Let v be the expression obtained by retaining only certain terms of u. for its equation is unalte red if x and?/ are interchanged.

J5. the rule fails . whose coordinates referred to any axes are (m. for the corresp onding terms cannot be of the same order. 2nd. s). B according as C is on the same side of AB as 0. (See Ex. MHO. and therefore y is O^te-wO/fr-fl)). 3 give the form of the curve near the origin.) Fio. in the diagram. (p. Ex. NOTE. and let the angle xHA=0. 1.. say A. then y n ~~ (l is 0(x p ~ m )... 480 EXAMPLES OF CURVE TRACING . 2 give the form near infinity. Rule. In the following examples the points representing the 1st.. cutting Ox in H. n). therefore the terms represented by A. Thus the term C is of an order lower or higher than the terms A. Represent these by points A. Join AB. 2. 3 and 2. The figure so formed is 'Newton's Diagram.. The terms 1. 3rd. . . 82. B are 0(x H ). . If the straight line AB passes through 0. If all the other points lie on the same side of AB as 0. then v = represents the curve u Q (or part of it) near infinity..' Suppose that ax m y n .. Draw CK parallel to AB to meet Ox in K . then v = approximately represents the curve u = (or part of it) near (if on the curve). B and any points which may happen to lie on the line AB. or on the opposite side. Hence the following rule. 8. terms o f the equation are marked 1. Choose any two of these. q). .of u... then y is 0(x. . Let v be the expression whose terms are those represented by A. If all the other points lie on the side of AB remote from the origin 0. B. (r. B. then. The terms 1. C. Art. 3. Mark on Newton's diagram points A. w-ncot0 = 0#. bxVy* are small or large numbers of the same order . C. for similar reasons. the term cx r tj s represented by C is 0(x os )..ie ) and ax m y n is 0(x m ~ ncoie ). representing all the terms of the polynomial u. 1. .

ax 2 y -y*^0. (ax . a z x 2 ~ax z yQ and ax 2 y-l-y* 0. (ax . 3. 0). giving v-a=0 and '5 : 3 4 ^2 3 1 . Thus near (0. Write the equation in the form. 5. The form near the origin is given by terms 1. and near ~^ infinity by the terms 1. Near infinity.6//) 2 =0. 2. 2.by) 2 = ax 2 ij -f y 4 .Ex.2abxy + b 2 y 2 . 4 and 4. a 2 x 2 .

(iv)]. 84. 3 . y ^x + x 3 .) 1. 83. 84. [Fig. Near (0. 1 ). near oo . [Fig. 84. 4.-f 1). EXERCISE LI Trace the curves represented by the following equations and work out the accompanying details.|z 2 .00). oo ). Near (GO .] Ne.0). near (a.0). 3 + 2/ 3 ^a 2 x. [Fig. 3 + 1/ 3 . (For hints for plotting. y*-y + x=Q. t/ 3 = 3aa. 0). (iii)]. oo ). . 22. 5. 0).$jc~ 3 ~. [Fig. (v)]. y~ -x* . near (00.3ax 2 . -f !# 2 ) . The tangent is perpendicular to Ox at the points 1 ^ . 2 . (ii)].ar (0. Near (0.\x .0). a. [Fig. Near (0. 2. 3. ?/ 2 ^a. Near (0. y == 1 . see Ex.00). a.FIG. 84. (x-a)y*=a*x. (i). = . Near (0. ?/ 2 ^a. 84. near (00. y*=a z x. y = (a. 2 (2.

closer approximations being y = Tangents are parallel to Ox where y 3x. 1st approx. x .2 -a(>/5 + 1) I'D approx. 84.. i. [Fig. y= ( a 6. y y Near (00.e... 7. a.aij 2 ..0)..y / . 3rd approx. 0). (viii)].. (i). to the parabola (y-x-a) 2 ~ax...' a 2 \ dy ... (vii)]. y r ... x=?* Near (00. 2nd approx.. near (oo. x(y-x) 2 =a 2 y. (vi)]. (A) ... i. 482 EXAMPLES OF CURVE TRACING 9. -1.. 3 -3i/ 2 ^0 and t/^fz-fz 2 ..00). If -~ 0. Near (0. y a (# ~a)~tf a (a. [Fig. Near (00. 1st approx. 84. (A). is on the curve. oo ). y=x*Jax + a. Near (0. but the curve passes through no real point near O : consequently O is a * conjugate point. + a).. Near (00.00).00). [Fig...e. 0). a 3 Near (a. 84.. # 4 -f2i/ 3 0. y x la* V^T" This shows that the curve becomes indefinitely close to that represented by (A)..00). y~xa.] Near (0..3*y 2 + 2y* = 0. 8. x(y-x) 2 =. This is therefore a parabolic asymptote. 85. x*^a*y.. oo ). x* . x = a-{ . a). [Fig. Near (0.

. .. 3 x z ' With y = # as 1st approx. Near (00..... which meets the curve at one other point.. [Fig 85.. 12 Near (oo... it gives the general trend of the curve......... ax*-y* 0.00). y-^ .] Near (0..T + ..... showing that the curve (A) is below (B) and that (y -f I) 3 -4 3# 2 ~ is a parabolic asymptote.... (iii).00).. x Near (00.. If y tx % then x = t*(3-2t). ?/ 3 . [Fig..f.....in the diagram .. 2) are on the curve.. Near (00.. (ii). y~^ . 12.. ax(y-x) 2 = y*.. ............ 10. a 2 a 2 x 3 11. ( . the 2nd is ?/ z ^ ....] .. the tangent 2 2*^/2 beuig 3x .. showing that the point O is a cusp. show that approximately x 2 a*x a The equations y = are represented by ------.......2?/ 0...0). 85. (A) The points (0...... 85. Near (0.. 3x~2y^0. ( B ) 03 & Equation (A) is represented by ----.... y=-r$--i and D tt x' A -a z y=0....) . 3)... by solving for y and expanding in descending powers of x. y~ -3 s X s ... O O 2nd approx..in the diagram.... x'-a'sy-iy^O.. [Fig. 3x 2 + y*~Q (represented by ----..3). The asymptote 2/ 3 cuts the curve again at ( . (3x~2y)*-3x*y-y*^0..0). Mark points on the curve corresponding to < = 1.] Near (0.. (B) 2nd approx... viz.. -A -1..0) (0..2^ y3 2nd approx.00). (iv).2. 0).. but (B) shows that it is not a parabolic asymptote.. y-x=Q and ax^y*.I....-^ xJ...-J).f .

85* 484 POINTS OF INFLEXION 18. 1316. 17. are curves whose equations are given in Exx.1 or else 0<t/<4. xt/ 2 . coincide. These are of the form xy*-2ky~px* + qx* + rx + $* (See Exx. (viii). (B) touch at the point ( . for value s of x other than zero. 15. 85.f ). and. s. 13-16 employ the methods of the previous examples to find first and second approximations to the curve at infinity . Show that every curve whose equation is xy z . and of the points where the curv es cut the asymptotes.IQy^x* . zz/ a -242/^:c 3 -10:r 2 ~5a:+150.15x 2 4. (vi). then. d. (ii) In Exx. A.2ky = px* -f. NOTE. J5. xy* . 18. 1706).b) (x -c)(x~d). B. p.1 and y 4 touch the curve. By solving for x. C.k) 2 =p (x -a)(x. q t r. show that y< . showing that the curves (A). or only one. (i) Show that. Also show that the lines y = . then (3.85. Fig.qx 2 -f rx + s has three asymptotes. trace the curves. * Curves having an equation of this form belong to the first of four classes int o which Newton divided cubic curves (Enumeratio Linearum tertii ordinis.2ky px* + qx 2 + rx + s. Diagrams (v) and (vi) show that a cusp ia of a higher order of singularity than a node . and then separate again. c.I2y = x. which correspond to x=a.MX . 16.r -2*/) 2 -0. . . a^ a -8y = x 8 -8a. owing to a change in the coefficients. / (viii) FIG. 14. Z>. if xy 2 .If 3z 2 + ?/ 3 ^0.) 13. by the help of the poi nts. 17. (vii).f . A.104. b. y is given by an equation of the form (xy . (v). where p^Q. C t D. according as p is positive or negative. while (vii) and (viii) show the * change of partners ' that genera lly occurs when two of the points.14z a + 60x . 8 -3aJ-8. approach one another.

. express x in terms of y. CHAPTER XXX INFINITE PRODUCTS 1. (B).. whose equation is (y*-l)y~(x 2 -4:)x 9 has three points of inflexion which very nearly lie on the line x 64y. Ex.. plot x~ -t/ 3 . as follows : For For For For Ex.. (B). 20.. and ( -^ ) = 6z 2 + 6o# -f 6.2cb' + da'). and putting ~0.. the . (C) respectively by a'. (B) d 2 )/ Differentiating again. 6. Exx.. 3. [At a point of inflexion y~-~2x* + 3ax 2 + bx.... and (C). 475). Show that the curve (p... so that 21.. [From (i). is a sequence of numbers. and # = ?/.. the result is obtained by addition... a n . If a v a 2 . 2. by differentiation. express y in terms of x. use y=te.2bb' + ca') x + (be' .. and add corresponding abscissae. (A) lie on the straight line whose equation is %y(<i'c f .6" 2 ) -:= (ac' .Points of Inflexion 19. 12. Plot some of the curves in Examples 1-12. a nd y* = (4p*-l)/(p*-l). Exx.. Show that the points of inflexion of the cubic curve y(a'x z + 2b'x + c') = ax* + 3bx z + 3cx + d .. Multiplying equations (A).. real or complex. 8. 7. we have ct>x (C) The coordinates of a point of inflexion satisfy equations (A). 1. at a point of inflexion. 9.. Prove that the points of inflexion on the curve y*~x z (x 2 + 2ax + b) are determined by the equation 2x*-'+ Qax 2 + 3 (a 2 -f b)x + 2ab =0.1 dx \flte/ 22. [If p=dy/dx 9 show that. -2(a'x + b') 9 a'x 2 + 2b'x + c'. 11.. or x=ty. 4. 5. # 2 = (4p 6 -p 4 )/(^ 6 -l)... 10....

or if P n ->0 when no factor of P n is zero. to zero or oscillates according as log P n tends to a finite limit. If P n tends to a finite limit P. or simply = Ua n . | u n |<1.. or that it converges to P .. // u l9 u 2 . to oo .. to . as regards convergence..oo . to oo . w 3 . It is generally convenient to write an infinite product in the form A necessary (but not sufficient) condition for convergence is that w n ->0. and if this limit is not zero a n = P n /P fl _ 1 ~>l. and so.. a n will be denoted by P n or by II^ia r . the infinite product Ua n and the infinite series 2 log a n both converge. the infinite product is said to diverge or to be divergent. except when a factor of P n is zero. The reader may ask why we say that the product diverges (to zero) when P w ->0. If P n -> a finite limit. a n . Consequently there will be no loss of generality in supposing that | u n |<1 for every n. as n-> oo . then m can be . The simplest and at the same time the most important case is when all the us have the same sign. the infinite product is said to oscillate.. . The reason is as follows. The character of P. and P n tends to a finite limit.. The series 2u n is convergent or divergent.oo or oscillates. we say that the infinite product a 1 a 2 a 3 .. if a > a 2> .. different from zero. If P n -> oo or . Proof. will not be affected by removing any number of factors from the beginning of the product. each of the infinite products is convergent or divergent according as the series Eu n is convergent or diverge nt. For logPn^logaj-f log a 2 + . both diverge or both oscillate. -flog a n .product a x a 2 . 486 CONVERGENCE 2. then P n .. to oo = /7"~5a n . are real and positive. and we write P = a 1 a 2 a 3 . then a n ->l. are positive and less than unity. Suppose that Zu n is convergent and that its sum is s . is convergent and that its value is P. after a certain stage. 1 ~> the same limit.. Thus we write . If P n oscillates. According to the definition. Definition. Theorem.. Thus if Ua n is convergent..

. XIV. Therefore P n and Q n tend to positive limits as n->oo . Moreover.found so that u m + u m+l + u m+2 + .2. First suppose that P is convergent. and consequently a positive number k exists such that for every n. by Ch. and both of the products 77(1 -f w n ) and 77(1 ~u n ) are convergent. ABSOLUTE CONVERGENCE 487 Also. . (!-)> 1 .. to oo <1. Therefore P n ->oo and Q n ->0. Hence there is no loss of generality in assuming that s<l and consequently s n <s<l for every n. 1 6 n 1~6 g n = (l -U.. By Ch. an integer m must exist such that for all values of p that is to say. so that 77(1 +u n ) diverges to oo and 77(1 u n ) diverges to zero. General Condition for Convergence. none of which is zero. we have P n = (l+u 1 )(l-fw 2 ).) ..(l+w w )< T -<.2. Also the omission of the first m-l factors of a product does not affect its convergency.)(l -. Also P n increases and Q n decreases as n increases. < for ^ = 1. real or complex. The necessary and sufficient condition for the convergence of the infinite product is as follows : Corresponding to any positive number 6. 1.. Let (a n ) be a sequence of numbers. (A) Proof. By Ch. Then P n tends to a limit which is not zero. Next suppose that Zu n is divergent. we can find m such that for y = l. 6.. 1... however small. XV..> 1 -*. Q n >0..3. XIV. 3.3.

3. the series Zu n is absolutely convergent. then and l. 3.. Also by (B).... ....) 4. if condition (A) holds. u 2 '...2. . Then. .. 6.Therefore _rn+p i ~ fa so that condition (A) holds. by Art. or moduli. so that |P n |>some fixed positive number when n>m.... is said to be absolutely convergent when the product is convergent. Theorem. XV... If u l9 u 2 .(l+t4+p)-l|<6 for ^ = 1. w 3 . we have | P m+J) |>(1 -e) . the infinite product P = (l +*!)(! +^(1 +113). and therefore IP -P I IP I -IP \< IP I I * m+p L m I ~~ I * m+p \ \ f m \ ^ fc \ * m\ Hence | P m+p P m \<ie'. Hence P n -> a limit different from zero and the product P is convergent. (B) Taking e<l.<p<l + . ... and only if. where c is positive and as small as we choose. u 3 '. Let P and P' be defined as above and suppose that P' is convergent. P m+p and P m have the same sign.. Consequently P n cannot tend to zero. | P m |. are real or complex numbers whose absolute values... An absolutely convergent infinite product (P) is convergent. Hence the product 77(1 +u n ) is absolutely convergent if. Conversely. we can find m so that |(l+t4+i)(l+t4+a)-. Absolute Convergence..... are denoted by w/.. (See Ch..

488 EXPANSION OF PRODUCT 6. q.. so that P n = (l +xu 1 )(l+xu%) .1 1->0. Since P is absolutely convergent. . as m tends to oo . . (1 -f xu n )^l +px + p 2 x* + .. We shall prove that if the factors of an absolutely convergent infinite product (P) are rearranged in any way. . and therefore QJPm = a v a Q .a * where p. a s . u n taken r together. the product remains absolutely convergent and its value is unaltered. and x are any numbers.. real or complex... 3. 3 that | a p a a ......3..Now (1 This follows on expanding each side and remembering that for any numbers a.. We shall prove that if the series Eu n is absolutely convergent. so that every v is a u and every u is a v. +p n x n . Let a r = 1 + u r . Derangement of Factors.(l+w m+P )-l|< for 2> = 1. and therefore Q n /P m -* 1 . c. P = Q. Thus P m and Q n tend to the same limit.. . . it is possible to find n so that all the factors of P m =a 1 a 2 .2. It remains to be shown that P = Q.. s are all >w. Let p r be the sum of the products of u ly u 2 . Expansion of an Infinite Product as a Series... Consider the infinite products where P is absolutely convergent and Q is formed by rearranging the factors of P. b n . and consequently the product Q is absolutely convergent. P is convergent. it follows from Art. Also.. 6. 6. Therefore Zv n is absolutely convergent.. that is to say. so does n. a m are included among the factors of Q n = b^b 2 . Consider the infinite product where u l3 w 2 .. w 3 . This series remains absolutely convergent when its terms are rearranged. and since P is convergent.. . For any suffix w. then . by Art. so is the series Su n .. Hence. Therefore |(l+t/ m+1 )(l+w m ^ 2 ). b r = 1 + v r ..

and p r becomes an absolutely convergent series. and consequently Z r '->0. #>/-> a limit l r ' such that If r->oo .. to oo. .... taken r together . .. (1 +z* n + 1 therefore (1 +zx n+l )(l -\-p l z+p 2 z 2 + . -f-p n 2" Equating the coefficients of z r . Let w/ = | u r | .. which is convergent . p f -> a finite limit l r for r = l. w n '. and erefore p r '<s' r j\r.. it follows that.. then s' r / I r->0. p r f becomes an infinite series. and since P n ->P. JE'ar. (ii) For all values of x. to oo . | c 2 I? e ^ c Therefore Q n -P n ->0. w 2 '. and it follows that Since Eu n is absolutely convergent. Zu^ converges to a sum s'. therefore (1 +zx*)(l +zx?) .. Let Q n = l + I l x + I 2 x 2 + . Also p r ' increases with n. ^t'nc? ^e coefficient of z r in the expansion of Assume that In this identity put zx for 2.. so that l r f is finit e. Hence p r -> a finite limit l r) and we may write Z r = (l +e r )Pr where e r ->0. P = (l +zw 1 )(l -f xu z )(l +xiiz) . I r occurs in the expansion. +l n x n . however great * may be. 2. Thus as n->oo . and so that j>/ is the sum of the products of W/. then where m is the greatest of | l |. n. x' = | x | .. to oo = 1 + ^x + Z 2 x 2 -f . Each of these products multiplied by 1 r occurs in the expansion of GENERAL COEFFICIENT 489 therefore p r ' ... for all values . therefore. ...(i) As n->oo . 1. as n->oo ..

1 in succession for r. 1..x n ~ r + l all n *> tend to zero as ?i->oo.. +log (1 +w n ). therefore the infini te product is convergent. to 00 = l+f where I The series x 4.. is absolutely convergent. are real and ~I<u r <l for every r. (A) where I is tV> smaller and h the greater of the numbers 1 and 1 -f x. XIX. (4). 1 a* 2 1 x~ if z>-l and ^0. // | x | <1. ... By Ch.. 7...x 2 -f jr 3 . Also x n .r+1 ^ r Putting r ~.+u n = r m ..y . .. 2... 2. Let P w -(l + w i)(l +^). .. When the factors are real and positive. and denoting its value by P. by the theorem just proved. . then ~ r-<x-log(l 4-x)<. show that for all values of z.(1 +u n ).. -f log (1 +ti w ) = J mfn .. 7. 1. . both diverge or both oscillate. x n ~\ ..1 -x n . Then log P n = log (1 +1^) + log (1 +w 2 ) + . n ~\-x Whence by multiplication 490 LOGARITHMIC METHOD Ex. therefore Z r has the value stated above. r . 7/2 4-7/2 4. As explained in Art. Let u m+l +u M+2 + ..2.. _L 77 2 <% m+I^m+2^~'^ U n ~ s m^ log (1 +w m+1 ) + log (1 +ti ro+2 ) + . the last assumption involves no loss of generality.1 . n . where u l9 t/ 2 .. . the theory of infinite products can be made to depend on that of infinite series by the use of logarithms... . Hence the infinite product P and the infinite series 27 log (1 +u n ) both converge. Another Method. where Z r lim p r and p r is the same as in Ex....

.r _ 4 ^-^ ?i5 2 "if m>n m > n 2~L ' where L is the least and H the greatest of the numbers 1 . diverges to 0.n.. or oscillates. Show that . then since r. or oscillates. Hence by the general principle of convergence. <l-)(l+i)(l-i)(l+-J-). for sufficiently large values of n. and when the logarithm of a complex number has been defined and its properties investigated.u ly 1 -f w 2 . d+t)(H 2. diverges to .o o . 1.2> . EXERCISE LII Obtain the results in Exx. (l+i)(l 6. however great. by addition we find that ^w_. (i) Suppose that u n 2 is convergent... This includes all possible cases. and therefore where c is positive and as small as we please. l+u n . then P diverges to 0.-i. Hence (a) If u n converges or oscillates between finite limits. P converges.GO . 5 m>n <2Lc. the same method can be applied when the factors of P are complex numbers. then P may converge..n^**Vn it follows that. w w for x in (A). then by the general principle of convergence we can find m so that for all values of n. .jn ^. 1 4. 1-5 by considering the value of P n in each case. w m 4. \P*ni = l and 5. diverges to oo .. (b) If u n diverges to + o> or oscillates so that its upper limit is -f. (i- 4.. EXAMPLES OF INPINITE PRODUCTS 491 (ii) Suppose that Zu^ diverges. as u n converges. for any positive value of A T .n~^.Putting ti^+i. diverges to oo .

C... Show that the infinite products are convergent for all values of x. Remembering that Jim --.A.. ^ . . Show that (1 + a:) (1 +-)( 1 4-^ ). . A X Hence if | x \ < 1.] 11.H. Show that sin x = 2 n sin ~ cos | cos ^ . 492 EXPANSION OP PRODUCTS 10.. ~ B.1. diverges to oo or to according as O. show that is exactly divisible by ( 1 .|<l and \zx\<\.tf r + . . V ^ 2 /^ 3 / 9. If n. cos ^ ..d CTe ? '~ (!-*)(! -*). 6. . ( 1 ~ x r ). deduce that 0->o ^ xx x sin a. at ~ 1 + *iZ + * 2 Z 2 + .. If |a. 1. to oo ^v.x 2 ) . show that where a _(* -*")(* ~* +i ) . r are positive integers.(l [Proceed as in Art. Ex...x) ( 1 .. 8. A C 7. to 00 . (l+x)(l+x z )(l+x*) .

) = ! + tjzx + t 2 z z x z -f .. where ^ r =a. r /(l -a: 2 )(l -a. to oo where * r ^z r /(l -x)(l -a: 2 ) .] 13.. show that 14. k. If (l+zx)(l+zx*)....tO 00 .. COMBINATIONS AND DISTRIBUTIONS 1.(l+zx* n .l )^l+p l z+p 2 z*+. ==1 where / ^ --____ . where (ii) . k are all equal to unity. we obtain the number of the . CHAPTER XXXI PERMUTATIONS.. r ). The sum of these products is the coefficient of i x r in the expansion of When a... is denoted by H.)(l -2o. (1 -a.. Hence by equating the coefficients of z r . (l-X^t^X. c. show that for all values of z.. If \x\<\ and \zx\<\... . show that .. . 6. (1 -x* r ). If | a. (l+zx)(l+zx*)(l+zx 5 ).+p n z n .t^.=l+^2 + ^ 2 2 2 -}-. If | x | < 1. 3 )(l -zx 5 ) . | < 1 and | zx \ < 1... (1 -sa. when each may be taken as often as we please. 6..12.. The number of combinations of n things r together. Combinations with Repetitions. 4 ) .. to oo J . .. put zx for z and show that (l-zz)(l+t l z + t 2 z* + ... show that . . x r\ 15.... and is the same as the number of homogeneous products of degree r which can be formed with n letters a. [Having explained why such an expansion is possible. .

"^ 2... .. Combinations and Permutations of n Things. = k.products. and P the number of permutations... equal to unity. bb . (1 +cx + c 2 z 2 -h . c. //... taken k together.6's... . therefore // . of x r in (1 + x + 2 -f . of n letters aa .. 2.. to/factors . 494 DISTRIBUTIONS (2) P is \k times the coefficient of x k in the product / x x 2 x\ L x x 2 tf\ For the number of permutations involving Aa's. . of x r in (1 -x)~ n . and so on. such that A -f/z + v + . v. C coefficient of x* in (1 ~x)~ 5 PI 1 4 coefficient of x* i .. 1. . + c r x r ) . b.. cc ... /x. Whence the result follows. 3... For the sum of the combinations in question is the coefficient of x k in the product . 1. chosen (with repetitions) from 0.. .. q the number ofVs... LetC be the number of combinations... Find the number of combinations (C) and the number of permutations (P) of the letters of the ivord parallel taken four together.. +*)(! +Z-fZ 2 -f . and their number is the coefficient of x k in the expression (l+X+X 2 + . p is the number of cCs. is Hence where the summation is to include every set of values of A. Thus #? = coefft. re's. . Ex.) n = coefft... -f X r ) which is obtained from the preceding by putting a. then (1) C is equal to the coefficient of x k in the product where f is the number of different letters.. r the number ofc's. +X p )(l -h# + 2 +.. Therefore P/ \k is the coefficient of x k in the above product. We have by the preceding. not all different.

) Ex. and so blank lot . (iv) Blank lots may. a ' blank lot ' being a lot ' which contains none of the things. and so the parcels ar e. in the second. In how many ways can five books a. it is a case of distribu tion in groups. (B) in Ex. the groups are said to be different. (C) are different. 3 ? Such distributions as (A). Ex. the result is called a distribution. 1 are identical. A number of different cases arise. or may not. 1. Further blank lots are admissible. < . 3.whence we find that C22. Because (A). 6. 3. they are called parcels. In the first case the classes are called groups . it is implied that each person gets at least one book. c. in the second case they are said to be indifferent. c c c Because such arrangements as (A) and (B) are different. thus : (i) Some of the things may be alike. <% rf. (C) are different. 6. If n things are divided into r lots of classes. 2. d y e on three masts ? ARRANGEMENT IN GROUPS 495 Denoting the masts by 1. 2. Further. For example. have to be considered. e be divided among three peo ple denoted by 1. 1. or they may be all different. 4. P 286. Distributions. (ii) The order of the things in each class may. for all the flags may be flown on only one or only two masts. be admissible. it may. 2. Ex. be necessary to consider the order in which the classes stand. consider what is implied in the following questions. Also the distributions (A). (iii) When a distribution has been made. so it is a case of distri bution in parcels. In the first case the classes are said to be different (meaning that they have to be distinguished from one another) . or may not. How many different signals can be made by flying five flags a. (See Art. 123 123 123 (A) : d e <B) * < {C . said to be different. the following arrangements give different signals . or may not.

5.. 496 PARCELS OF UNLIKE THINGS 5.(r + n-l) or |wC*l}. blank groups being admissible.. (r + n.r .I / \r -1 =r(r + l)(r + 2) .1 marks of partition. This can be done in C"~\ Hence the required number is | n C^l*. . 1. so the distribution is in parcels. 6.. according as blank groups are or are not admissible.* Since r . How many different signals can be made by flying five different flags on three masts ? The required number = 3.) 4. The number of ways in which n different things can be arranged in r different groups is r(r + l)(r + 2). is indicated by a |j bed \ef\g.l) r . and consider the distributions in which blanks are allowed. k be written in a. Distribution in Parcels. and so t he parcels are indifferent. blank lots being admissible.4. The total number of these is r n . 3. In any distribution. a 2 . none of the groups Arrange the letters in all (b) In every such arrangeout of the (n-1) spaces ways. 2.. denote the parcels by a r . (2) The number of ways in which n different things can be distributed into r different parcels.. there being no blank lots..6. For each of the n things can be assigned to any one of the r parcels.7= 2520. the number of different arrangements is \n-\. groups. . can be obtained as follows : (a) possible orders. . Ex. Arrangement in Groups. * Thus one arrangement of 7 letters in 5 groups. the second blank. Proof. 5. row in any order. In how many ways can jive books be tied up in three bundles ? Here the order of the books in a bundle does not matter.) Ex. (i) Let n letters a. (See Art. can be obtained thus : place among the letters r . is r n . This can be done in In ways. 1. ment. (ii) All the arrangements of the letters in r being blank.1 of the things are alike. and arrange the n + r-l things (consisting of letters and marks) in all possible orders. a r .. Also the bundles are not to bo distinguished from one another. All the arrangements of the letters in r groups.1). (See Art. (1) The number of ways in which n different things can be distributed into r different parcels. place (r-1) marks of partition in (r-l) between the letters.s are not allowed.. Proof. Ex. is which is In times the coefficient of x n in the expansion of (e x .5. Ex.

} ft/ 1W| . . this gives the result in question. and it is obvious that the coefficients are formed as in a binomial expansion. Hence the number of distributions in which no one of x assigned parcels is blank is r n _ (j*(r .. a x is blank. a 2 . . a 2 are not blank is J Of these last.x) n . 1...1 + Cf ~ 2 . that is to say.E~ l ) u x . + ( . A complete proof is as follows : Let u x denote the number of distributions in which no one of the parcels a t . a 2 is blank is J Therefore the number in which a. + (. . .The number in which a x is blank is (r .. Therefore the number of distributions in which no one of <x v is blank is u x . where E~ l denotes the operation of changing r into r-1.2) n -.l) n = (1 . .1 )*E. u x+l = (1 .Cf E. Ex. This number is represented by E~ l u x . this gives the result in question.E~ l )r n .*. ft .X ) r n and when x = r.(r o) .l) n + Of (r . Of these.1 in u x (which is a function of r). In how many ways can five different books be distributed among three pers . the number in which a x+1 is blank is obtained by changing r into r .l) x (r . Also we have u { . a 3 is blank is J Therefore the number in which^ i> 2> a a are n k blank is J This process can be continued as far as we like.r n . the number in which] . 1 1 ! > r n -2 r-l) tt + r-2 n . .1 ) n .. . the number in which] . > (T 1) Z(r A) 4. U11 . Therefore the number in which o^ is] not blank is Of these last.E~ l u x . When x = r. > (r-1 n -(r-2". PARCELS OF LIKE THINGS 497 Hence w x = (l -E~ l ) x r n = (1 ..(r .

// blank lots are allowed. can be chosen from 1.. If there are no blank lots. I 5 = 150.. the summation to include every such selection. 3. the number of distributions is the same as that of n + r things of the same sort into r parcels with no blank lots. where no parcel is to contain more than k things and there are no blank lots. . 1 5 ) -. For such a distribution can be effected thus : put one of the n h r things into each of the r parcels.3 . of x 11 in (x + x 2 + x >3 f . A:. .. . any such distribution can be effected as follows : place the n things in a row and put marks of partition in a selection of r 1 out of the n l spaces between them. x k . 8. 4. Proof.= 2. and this is the required number. tied HJ> in three bundle ? The required number ~ (3 5 .x k ) r or of (l-x k + l ) r (l-x). In how many ways can five different books be.. are any terms selecte d from x. blank lots being allowed. /x. 2 of the next section.ons.. according as blank lots are or are not admissible. Ex.. ma ke a total score of 1 1 ? The required number is equal to tho co. . may be denoted by 2x x x^x v . which is therefore the number of such distributions. . 6 . The coefficient of x n in the product is therefore the number of ways in which r numbers A.. 2 s + 3 . + a.r . This can be done in C n ~~^ ways.3 . x v . For the product of r factors. is the coefficient of x n in the expansion of (x + x 2 + x* + . . . In hoiv many icays can three persons. . Distributions of things of the same sort into indifferent parcels are considered in Art. 2...x r . 498 DERANGEMENTS Ex. These theorems can also be proved as in Ex. 2 5 + 3 . 1. Hence the number in question is /Tn-fr i O r _i .. 2. if each person is to have at least one book ? The required number ^3 5 . repetitions being allowed.. x 2 . each throwing a single die once. x^. O v) (3) The number of ways in which n things of the same sort can be distributed into r different parcels is C>i{~ 1 or C^Ii. . . (4) The number of ways in which n things of the same sort can be distributed into r different parcels. where X A . such that their sum may be n. and distribute the remaining n things into r parcels.~>. each of which is x -f x 2 -f x 3 -f ..

I |w-2 74-1 +|tt-2.) 3 ^co. . 6. .. neither a l nor a 2 is in situ is J Of these last. .The total number of arrangements is \n\ and of these The number in which a. Derangements.). \ \n~\. of* 8 in (^* 2 + 9 ~ * 8 )(l"3z 6 ) = 45 -18 -=27. The number of distributions of n like things into r parcels is the coefficient o f x n in (1 +x + x z -f . of x 11 in u. 3 -|. Ex. stands first.o.!tt-2. . the number of ways in which they can be deranged so that. These expressions are respectively equal to (1 . ac cording as blank lots are or are not allowed. . . Of these last. is Therefore the number in which a l is not in situ is n In 1. etc. .x)" r and x r (\ -x)~ r . to GO } r or in (x -f x 2 -f a. 3 (l -^) 3 (1 -a. no one of them occupies its original place is Proof. .. the number in which a 3 *\ is in situ is . If n things are arranged in a row.3 (l ^co. Any change in the order of the things in a group is called a derangement.. of x 8 in (l-o:). 2.3 ^c'. to oo ) r . the number in which a 2 l i i . Apply this method to prove the theorems of the la^t section. is in situ is J I Therefore the number in which] .] l I 1Mi. is in situ. Denote the group of things by a j? a 2 > a n.e.

and we obtain the result in question. a%... Of these..2 -. + (. This number is represented by E~~ l v x .(\-E~iy\n = (i .. .. that is to say. (i) Tho expression (A) is sometimes called ^n(> factorial n. Therefore *he number of arrangements in which no one of a {9 fl 2 .1 yE-*) | n . ^ . Hence the number of arrangements in which no one of x assigned letters is inaitu is n-Cf JB-l+Cfjn -2 -.. a 3 is ?.j>i-Cf l'/&-l+0 |n--3-. If x = n.?i . NOTE. we have Hence 1 . and we obtain the result in question.. a x is in situ is v x -E~*v x ... it follows that v x =.si^/ is This process can be continued as far as we like. and it is obvious that the coefficients are formed as in a binomial expansion. as in the preceding.inone or aj. the last term is ( -I) 77 .[)*~[n-. THEOREM OF WHITWORTH 499 A complete proof is as follows : Let Vy denote the number of arrangements in which no one of a } . .. and is denote d . the last term is (-1)". <Wi^O "-">.1 + <7/?.+(-!)* |n -a.r. y \n--3 w-1+3 ln-2. (which is a function of n). .* Also. . a 2 .1... .+(.Therefore the number in which no] . the number in which a x ^ l is in situ is obtained by changing n into n .c 'f A'. If = ??..1 in uj.. a x is in situ. where E~~ l denotes the operation of changing n into n .

. 3.2 + 1^-1. . then 3. . Ex. shortly.. and we say that 3211 is a 4-partition of 7. 2 2. I -1^2. 5 have been written do wn. can be calculated thus : Begin with 1 and multiply successively by 1. then 2. then 4. followed by all the partitions of 3 (that is. The complete set is . 3. a p-partition of n.211. 2. Partition of Numbers. | 3=3. . A General Theorem. and finally r letters occur in N r sets. if the selection contains p numbers. Any selection of the numbers 1. 2. then the number of events free from all the conditions is as stated above. .. Choice and Chance. + . It is easily shown that 1 1 n is the integer nearest to \n Ic. . .GIN. 4.] YN T . 111) . The proof is exactly similar to that in the last article. every two of the conditions are simultaneously satisfied in N 2 of the events. * Whitworth. Or more generally : * If there are N events and r possible conditions such that every single condition is satisfied in N l of the events. stated by Whitworth * as follows : If there are N sets of letters and. followed by the partitions of 4 which contain no part greater than 2 (that is.. j | 2=0. Thus 7=34-2 + 1+1. if out of r assigned letters a.. each letter occurs in N t sets.by 1 1 n. 11) . 3... 7. sets.. [[4^4. c. 2. 2. (ii) Tho values of || n for n 1. explain how to find all the partitions of 6. . eac all the letters i 500 CONSTRUCTION OF PARTITION-TABLE 8. Take first 6 . I -1^0. and finally al l the r conditions are simultaneously satisfied in N r of the events. 1. 1111). 111111. increasing any even product by 1 and decreasing every odd product by 1 before the next multiplication. lastly..24 1 -9. is called a partition of n . . 3.. and so on. then the number of sets free from all the s N ~ C[N t + CIN Z . Siij}posing that all the partitions of 1. . followed by all the partitions of 2 (that is. each combination of two letters occurs in N h combination of three letters occurs in N% sets. 73.. (1) In this section we consider a number as formed by the addition of other numbers. 2. Thus || I a. it is called a partition of n into p parts or. 21. p. + ( . 3. 6. of which the sura is n. In Arts. 4. (with repetitions). then 5 followed by 1 . 5 and 6 we have instances of a theorem.

. m) : call the set of spaces (n.... 1. p.. .. .. 2.. 77? f2 ^ 77? +77' 2 l . 1. or p parcels. 33. 2). 222.. 3111. we have 77r i -77 1 n . .. 2.. ..+ 11 p ij whence the result (A) follows... 3. 0. . . 3). and therefore 77^^ 0. * Hp is subsequently denoted by P(n. Also 77^ = 1. 111111... and so on to any extent. 3.. We have 77? = 1. 0.v . The third column is 1.. .. starting from the top. denote the nth space of the mth row by (n. In the table on page 501.. Therefore Changing n into n . 21111.. 77S = i(w-l) or \n. we have TTH-I n n -p rjn-p rjn-v rjn-p llp^l^lll -Til* -r 11$ + .. etc. mid using (2). terms of the nth column. 77^= 77? + 77^+ 77*. the n-th diagonal. .. 0. and the fourth diago nal 1. *).... two. (n + 2. 1. .6. If P>ln> then II%= n"1\. according as n is odd or even.. (2) // Up is the number of p-partitions of n* then n$=n$-i + n$.. 0..1 and p into p-1.... CLASSES OF PARTITIONS 501 The second column is 1. .. 42. . Hence the numbers in the (n + l)th diagonal are the sums of one. 51.. 3...... (n + 1. therefore the third diagonal is 1. 2. 1). 77^-1 = 1. 2. This can be done by placing one of the elements in each of the p parcels and then distributing the remaining n ~p elements into 1 or 2 or 3 . blanks being inadmissible... (3) Construction of a table of values of 77^.. 321. 411. We can fill up the first and second rows and the first and second diagonals.. . For in this case p^np. three. (A) To make the p -partitions of n is to distribute n elements into p indifferent parcels. 2211.

TABLE OF ^-PARTITIONS OF n (i. VALUES OF P(n y p. *)) tt~ 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 .e.

19 20 p=l 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 .

1 1 1 2 1 1 2 2 3 3 4 4 5 5 6 6 7 7 8 .

8 9 9 10 3 1 1 2 3 4 5 7 8 10 12 14 16 19 21 .

24 27 30 .33 4 5 __ 1 1 1 2 1 3 2 5 3 6 5 9 7 11 10 15 .

13 18 18 23 23 27 30 34 37 39 47 47 57 54 70 64 84 6 1 1 2 3 .

5 7 11 14 20 26 35 44 58 71 90 7 1 1 2 3 .

5 7 11 15 21 28 38 49 65 82 8 1 1 2 3 .

5 7 11 15 22 29 40 52 70 9 1 1 2 3 5 .

7 11 15 22 30 41 54 10 1 1 2 3 5 7 .

Thus : P(n. If the parts are to be unequal.<7 means that no part exceeds q. <p. and <y means that the number of parts does not exceed p.. <4.. *) is the number of jo-partitions of n.*) .. .. *.... ^p. and <.. ^p. in general... none of which exceeds q. ^q) is the number of partitions of n into unequal parts. then Q is used instead of P. An asterisk means that no restriction is placed on the number or the nature of the parts. P(n. p. Q(n. that is 23.. p means that there are p parts .. p.4.. (C) .. the greatest of which is q. ) where the number and the nature of the parts are respectively indicated in the first and second spaces following n. P(n.. *) is the number of partitions of n into p or any smaller number of parts. 502 CONJUGATE PARTITIONS (5) The values of For instance.... we have P(n.*HP(14.. *) can be found from the table on p...*). the is the sum of the the 4th number in P(n.11 15 22 30 42 (4) Various classes of partitions of n may be considered : the number of those in any particular class is denoted by a symbol of the form P(n. and is therefore the llth diagonal. In the first space. and. 501. q means that the greatest of the parts is q .. In the second space. as the case may be. q) is the number of partitions of n into p parts... number of partitions of 10 into not more than 4 parts first four numbers in the 10th column.p.*) = P(n+p. That is to say that P(10.

..... For consider the product Any terms out of the first. y........... there corresponds a partition of n into q parts of which the greatest is p. j8.............. of which the greatest is q... (1 -a*)..1 .<y.(6) We have the following relations : and Q(n.. x 3y ..... x q& where a. ^q) is the coefficient of x n in the expansion of (1 ....... (7) Conjugate Partitions. 1 Clearly 431 can be derived from 3 2 2 1 by the same process......... are any of the numbers 0. we have 3 2 2 1.p-l. .X )~ l (l -x*).... </th factors may be represented by ce.... Consider a partition of 8 into 3 parts of which the greatest is 4.. second. we have a partition of n q into p l parts.q)^P(n. (F) P(n........ (E) P(n..p. Consequently.. (D) For consider any partition of n into p parts. and these are caljed conjugate partitions.. 3.. q....p)... *..... EULER'S USE OF SERIES 503 (ii) P(n.*) .*... <y) . .. 2.p)... g) = P(n. and if we sum the i ] 1 1 columns instead of the rows. Also from every partition of the latter kind we can derive one of the former..*) = P(.....l (l -x 3 ). we have P(n..... (i) Q(n. <y) is the coefficient of x n in the expansion of For this coefficient is the number of ways in which n can be obtained by adding a selection of the numbers 1..... If we remove the part q..... P(n. (G) Thus by 5. a partition of 8 111 into 4 parts of which 3 is the greatest.*. ..... 1... <p) .. <p.. none of which exceeds q... (H) (8) Euler's use of Series in the Enumeration of Partitions. .q) = Q(n-q. <jj).1 .. say 431.......*) = P(n.. q... y . x 2/3 .... This may be written as in the margin. .. .. *.. .q...p.p........ Thus to every partition of n into p parts of which the greatest is q.... 2.

3. to oo =l -Z). Hence by (ii). 2. 4. tO 00 = . *) . XXX. *. Q(n.. . From this number we must subtract the number of partitions with a part greater than 9. 2. 2. 11) 1. .1 . 3. 4.l (I -zx 2 )~ l .. Ex. These partitions are 10. /Au^ ^Ae number of partitions of n into p unequal parts can be found from the table of values of P(n. is 11. Or thus.*) is the coefficient of x n ~W+ l > in the expansion of (i . Ex. ^q) is the coefficient of z p x n in the expansion of (1 +zx)(l +zx*)(l +Z3 3 ) .. and similarly <2(17. 3...p. 2. *)=P(n~l-p(p + i). .l (l -x 2 )~ l (l -x 3 ). 3. 9 ? Wo havo 0(17..3.2. The rest follows from Ch. whence by (5). 1 ..3.. Hence the required number is 11-2 = 9.. 11. Thus the total number of ways of forming 17 by adding four of the numbers 1. (iv) It follows that Q(n. 504 EXAMPLES OF SELECTIONS (v) The reader should have no difficulty in verifying the following (1 -zx). Hence the result follows.p. we have a + 2/3 + 3y 4. . 1. *)=P(n~^p(p.x ).4.. .. 1 . Q(n.1 . *)--=ll. repetitions being allowed. . )~ l .*)=P(11. 3. and their number is 2. -f qO = tt. G(17. (1 +z&). *) in (3).p. q. 4. . p. 10) = C(7..3. *) = P{4. 4.. and is therefore the coefficient of n -iP(p-fi) in the expansion of For the coefficient of z p x n in the first product is the number of ways in which n can be obtained by adding a selection of p of the numbers 1.. 2. (iii) Q(n 9 p.1 = l + EP(n.1).. <10)=0(7.q. L In how many ways can 11 be made up by adding four of the numbers 1. *) = !. ..If the product of these terms is x n . *)=P(17 -^4. <j>)... p. (1 -2^).. 3. (1 ~x p )~ l . 4. 6.1 1~ZX~ 1 \-ZX 2 ). Hence the coefficient of x n in the product is the number of ways in which n can be made up by adding any selection of the numbers 1.

[Number = coefficient of z 3n in ( 1 . 2n+1 ) 3 (l -#)~ 3 . of which r are alike. a v occurs more than once is In.1. Prove that the number of ways in which a candidat e can get 3n marks is i(w + l)(57i 8 + 10n + 6). two cf s and one e is 71. .] 12. In the expansion of (a x +a 2 + + O w where n is any positive integer not greater than p 9 prove that the coefficient ofany term in which none of the numb ers a lf a 2 . and (i) one or more of the masts need not be used . 3. four 6's. The number of ways hi which 2n things of one sort. In how many ways can a white balls. The total number of permutations of n things taken 1. . Find the number of ways of displaying 5 flags on 4 masts if all the flags are to be used. if one or more of the bags may remain empty ? 8. Show that the number of different selections of 5 letters which can be made from five a's. in n .x 2n+1 ) ( 1 . 2. In how many ways can 5 rings be worn on the 4 fingers of one hand ? 10.EXERCISE LIII 1. 4. 6 black balk and c red balls be put in m different bags. the maximum mark for each of three papers is n . 2n of another sort and 2n of a third sort can be divided between two persons so that each may have 3n things is 3n 2 + 3n+ 1.. giving not less than 2 marks to any question ? 11.x n+l ) 3 ( 1 . 1.] 9.x)~ 4 . 2 n ~ 1 different ways. three c's. 5.. In an examination. Show that a selection of (n . (ii) all the masts are to be used. In how many ways can an examiner assign 30 marks to 8 questions. 2. n are alike of another sort and n are alike of a third sort. or n togethe r is the integer nearest to e in . of which n are alike of one sort. blue and green balls in 286 different ways. and that 84 of these contain balls of all four colours. 3.. white. Show that a selection of 10 balls can be made from an unlimited number of red. [Number of selections of 3ft things = coefficient of x? n in (1 -a. 6.1) things can be made from 3n things. Find the number of ways in which n things. can be arranged in circular order.. for the fourth paper it is 2n. /I 1 x .

the number of solutions in integers (excluding unity) of the equation is n m . 2 Welshmen and 4 men of other nationalities (all different). . . Any one of the sequences occurs in \n-l of these . ) 1 L L_ \n + l n + l. d come together is [4-CJ. A party of 10 consists of 2 Englishmen. Show that the number of arrangements of the letters abed which involve none of the sequences ab.2) 2 n ~~ 1 . cd is 11.. in how many ways can each of n persons take a . (i) In how many ways can all the letters be put into th wrong envelopes ? (ii) In how many ways can 2 letters be rightly placed and 3 letters wrongly placed ? 14.. In how many ways can they stand in a row so that no two men of the same nationality are next one another ? In how many ways can they sit at a round table ? 18. 16.2* [2 = 8. The number of arrangements of the n terms a l9 <z 2 .2 .] 20.2 3 + Cl . and so on..n + 2 /* J EXAMPLES OF ARRANGEMENTS 505 13.. a m arc distinct prime numbers other than unity. Explain this.. There are 5 letters and 5 directed envelopes.I . . Show that if a l9 a 2 .[ + -f . b nor c. Show also that the number of solutions in which at least one x is unity is [Solution follows from the general theorem of Art..[Total number=e I n . \n-2-.F"-a -) w (l -x)~ z = coefficient of x n in 2 n (l -x)~ z -n . 2 a 3> a i ja r+i occurs is [n-CT [w-l + r:. The number of arrangements of the letters abed in which neither a. Verify by writing down the arrangements. 2 Scotsmen. a n in which no on e of r sequences a^. any two in I n . 17. Given n pairs of gloves. 7.. be.] 19. + (-l) r C r r \n-r. and write down the arrangements in question. 2 n ~ 1 (l ~x)~\ etc. The number of combinations n together of 3n letters of which n are a and n are b and the rest unlike is (n 4. [For the total number of arrangements is \n.] 15. [Number == coefficient of x n in (1 ~x n ^) 2 (l + x) n (l ~x)~\ or in (2 .

(ii) Zero values are excluded. A necklace is made up of 3 beads of one sort and tin of another. (i) If n things are arranged in circular order. Show that TIlj'^S/* 2 . beads . the number of arrangements of the CM. 10. those of each sort bein# similar. 26. y y z. zero values excluded. the number of such sets of three is -J(n-2)(n-3)(n-4). the number of ways of select ing three of the things no two of which are next each other is ira(n-4)(n-5). y. (iii) No variable may exceed 10 . 25. (v) x. 506 PARTITIONS 23. Show that the total number of possible arrangements of the beads is 3n 2 -I 3/i -f 1. w are included.right-handed and a left-handed glove without taking a pair ? 21. of x 4-7/4-z \-u. (ii) If the n things are arranged in a row. z. zero included. Verify when tt-9. of x n in 24. [The required number is co. 22.2# + 3z~n is i(*--l)(" where c~--H or J according as n is or is not a multiple of 3. The number of independent solutions in unequal positive integers.2p is the integer nearest to -&$p*(Zp ~ 3). (iv) Each variable is an odd number. Find the number of positive integral solutions of x + y -f-z-h w>~20 under the following conditions : (i) Zero values of x. w have different values (zero excluded). [Having put on the three beads. Prove that the number of positive integral solutions (zero values excluded) of the equation x 4.

^q) is cqiioi to the coefficient of x n ~ p in the expansi on of Hence show that Q(*. *). There arc n points in a plane. 4. Show that the total number of partitions of n is P(2/i. i>-12 were given by Dudeuey GEOMETRICAL PROBLEMS 507 3. Ve rily if n=4. *. 4). *). 28. *). *)=P(2n* r?.1. 8. (3). excluding the (ra-f n) points. Show that 7 3 (2ri-fr. the number of points of intersection is ^mn(m. no three being collinear except p of them. *). 32. no two of them being parallel and no three meeting in a point. no three being collinear. to tind P(25. If n straight lines of indefinite length are drawn in a plane. (iii) P(12. which are collinear. How man} 7 triangles can be drawn with their vertices at three of the points ? 2. ~~ * The results hi Exx. If m points in a straight line are joined in all possible ways to n points in another straight line. 18. Show that EXERCISE LIV * 1. Find the values of (i) />(18. There are n points in a plane. then (i) the number of points of intersection is ^n(n 1) . 5 *). n.1). then.in the spaces between the three (which are indifferent) is 27. 30. . Show that the number of n-sided polygons with their vertices at these points is -J j n . 31.l)(n. Hence use the table of Art.P. Show that P(n. <5. In how many ways can 20 be expressed as the sum of four unequal positive integers ? 29. (ii) P(12. p. n + r. Find the total number of partitions of 10.

b. the number of those which are isoscel es or equilateral is i(3c .. in general. Show that. 3. show that < u n ~-u. acco rding as r? is odd or even. Each side of a triangle is an integral number of inches.0) rerion s.1) if c is odd and -J-(3c . . prove that the greatest value of m(m ~ l)(3w. The number of squares formed on a piece of squared paper by m horizontal lines and n vertical lines (m<n) is %m(m . 2.l)(3>i m l). show that the number of different triangles is J (c + 1 ) 2 or -J c (c + 2). Prove that the number of different (i.1) is -i(a + l)(a-l)(a-3) or ia(a-l)(a-2). Of the triangles in the last question.2). 4. \. Of the triangles in the last question.I) 2 or \n(n . Show that the number of ways in which three numbers in arithmetical progression can be selected from 1. then v n v n ~\ + u n-i where v n is the required number and u n is the s ame as in Ex. prove that this number is Ar(a + l)(a -])(. b.e. n are positive integers such that m^n and m + n a. [(ii) If u n is the number of parts. no side exceeding c inches. n is i(n .] 5. c are integers and If c is given. 9.n.] 6. .(n) the plane is divided into \ (/t 2 + n-f 2) parts. 11..2) if c is even. l _ l 4. c inches where a. What are the exceptional cases ? [If no three planes have a common line of intersection and no four meet in a point. according as c is odd or even.3) or Aa(a-l)(a-2). 7. (ii) If a straight rods of indefinite length are placed so as to form the greate st possible number of squares. according as a is odd or even.congruent) triangles win ch can be so formed is or according as c is odd or even. (i) If M. 12. 8. The sides of a triangle are a. where a is constant. 10. the number of those which are isoscel es or equilateral is J(3c 2 -fl) or Jc 2 . non. according as c is odd or even. n planes divide space into J-(?i 3 + 5?j.-m . according as a is odd or even.

saying that some event is unlikely. then the chance that the event happens is a/ (a + 6). if a = 0. This implies that (at any rate in certain cases) probability can be measured and can be compared with certainty. both of them being used to denote measure of prol)ability.2K B. We are constantly forming rough estimates of probabilities. are placed in a bag. First Principles. B. . the odds in favour of the event are as a : b.A. A complete knowledge of these enables us to say that the event is certain to happen or certain to fail. K> all of the same sort. and the chance that one of the n balls is drawn is nx. it is certain to fail. EXCLUSIVE EVENTS 509 If b = 0. If an event can happen in a ways and fail in b ways. CHAPTER XXXII PROBABILITY 1. and that one of them is drawn. Thus if x is the chance that A is drawn. In what follows the words probability and chance are synonymous. The following definition is a statement of Laplace's First Principle. is drawn is also x. in other words it is equally likely that A. and so nx = 1 and x = l/n. and there is nothing to lead us to believe that one of these ways should occur rather than another. and its probability is 1 .. .. the odds against the event are as b : a. In speaking of the probability of an event. and this is always done. that is to say. very likely or certain to happen.C. and the probability of its happening is 0. the chance that any other ball. Thus the probability of drawing A is 1/n. say 5. Measurement of Probabilities. The degree of doubt depends on our knowledge of the controlling conditions. which must be regarded as a degree of probability. the chance that the event fails is 6/(a + 6). or K will be drawn. Suppose that n balls A. all the ways are equally likely. doubt is implied as to whether it will or will not happen. Thus there is nothing to prevent us from choosing certainty as the unit of probability. Our notions with regard to probabilities therefore lead us to assert that the chance that either A or B is drawn is 2x. . But it is certain that one ball will be drawn. There is nothing to favour the drawing of one ball more than another.. likely.. Definition. B. the event is certain to happen.

the chance that the card is a spade is i | . In the case of mutually exclusive events. the chance of its happening is the sum of the chances of its happening in the several different ways. It can also be stated as follows : If an event can happen in several different ways. the c hance of drawing white is J . if th e pennies are denoted by a.' Ex. . then Events of a type to which this definition does not directly apply will be considered later. . c. the event is as likely to happen as to fail. where the letters denote integers. the odds are 3 to 1 against their falling alike. then a can fall head or tail. it is an even cha?ice that they will fall all alike. 6. Hence there are 8 equally likely cases of which 2 are favourable.. Out of g equally likely ways. or all tails. q are the respective chances of the happening and failing of an event. i. 2 /<7. the chance that ' ace ' turns up is -J-. (ii) // a six-faced die is thrown. or 2 heads and a tail.' As a matter of fact. Examples. the ctiance that one or other of them occurs is the sum of the chances of the separate events.e. . the events can happen in a ly a 2 . A man of much practical experience in the laying of odds said : ' // you toss three pennies simultaneously. For consider n events of which the chances are ajg. -f a n . the chance that the card is ace of spades is -^>. So they fall all alike in 2 out of 4 possible cases. . This statement is known as Laplace's second principle. the chance that the card is an ace It is easy to make a mistake as to the meaning of ' equally likely. one or other of them can happen in a l + a 2 -f . (i) // a coin is tossed. the chance of ' heads ' is J-. or 2 tails and a head. 1 For they must fall all heads. a n /g res pectively. all heads or all tails.If a = 6. and since the events cannot concur. (iii) // a ball is drawn from a bag containing 3 white. and so for b and c. . and we say that there is an even chance of its happening. a. Exclusive Events. the odds against black are as 3 : 2. For. If p. .. (iv) // a cord is drawn from a well-shuffled pack. 2. Events are said to be mutually exclusive when the happening of any one of them precludes the happening of any other. .. a n ways respectively. and 2 black balls. one only of which can occur on the same occasion. 1.

chance of throwing 11 is . . 10 in the expansion of . 1. (iii) less than eleven. i. 2). (5. (1. . 1). together with th e thirty throws (1. 1).. Therefore t is the sum of the coefficients of 3 .. These are made up of the six throws (1. (6. We count (5. 27 1 . 4 . . (iii) Let / be tho number of cases in which less than 11 is thrown. the chance of throwing doublets 3% . (2.. 2. what is the chance of throwing (i) four -five-six. In a single cast with two dice.. . Ex. Ex. those in favour of the event are (6. . These cases are those in which exactly 3. (4. chance of throwing four-five -six -^ ~Q^ o t5u (ii) Let s be the number of cases in which 11 is thrown. therefore the chance of throwing 7 is -/ .. (3. .'. then s^coeflt. (2. a. the chance of throwing five-six: -3%. 1). Thus the chance of throwing two aces^g . against two numbers of which the sum is 7 ? Out of 36 possible cases. e. 6).. and the odds against throwing 7 are as 5 : 1.. than ten ? The number of possible cases is 6 3 . so theie are 6x6 equally likely cases. 4). -f a n )/g. (5. which is the sum of the chances of the events. 2 ). In a single cast with two dice. 5. what is the chance of throwing (i) two ac es. 4. i 3 1 . and call it five-six. 2).e. (ii) eleven. (iv) more.. what are the odds against throwing 7. The chance of this is therefore (a 4 -f a 2 -f . (ii) doublets. 6). 3).ways. 5) as the same throw. In a single cast with three dice. (i) Among those four -jive-six occurs 1 3 times . and the number of these is 6. 6). 5).*. (6. of x 11 in (x f x' 2 +x* + .J~. 3. 510 CALCULATION OF PROBABILITY Ex.. .-. (ui) five-six (i. 6). . 5). (2. x~ } . (6. 10 are thrown. one die turns up five and the other six) ? Any face of either die may turn up. + :r 6 ) 3 ^27 .

Events are said to be independent when . balls are drawn from a bag containing G white and 4 black balls. 6 ) 3 (l -x)~* = coefficient of x 1 in (1 -z)~ 4 (l -3rc 8 ) ^coefficient of x 7 in ( 4x + -.'. and 2 black balls are drawn ? Five balls can be selected out of ten in CJ ways.=. 6 = 120-12^108. B has a choice of 9 places.coefficient of x in x 3 (l -o. (iv) The total number of possible cases is made up of those in which more than 1 0 are thrown and those in which less than 11 are thrown . chance of throwing less than 11 is ^. + # 6 ) 3 . chance of throwing more than 10 1 -J ==i. B) sitting together ? A having taken his place. 2 of which are next to A. and this is the number of poss ible cases. a player stakes on throwing more than 10. 6. 1. What arc the odds ag ainst two specified ptrsons (A. t. Four cards are drawn from a pack of 52 cards..3* ways . Independent Events.'. and this is the number of favourable eases . C% way s. hence.i -iv C *'t 10 the required chance = ^ T ^p .) Ex.(x + x. . 6V \ I . Five.s> .">TC 5 ^L 3. INDEPENDENT EVENTS 511 Ex. 5. What is the chance that one of each suit is drawn ? Four cards can be selected from 52 in C% z ways . .4 are as 7 : 2. Wh at is the chance that 3 white.. 2 . (In the old game of passe-dix.+ x* . which is the number of favourable cases. hence the odds against B sitting next to . ^4 parti/ of ten take. therefore .3* 28561 Ex. Also three white balls and two black balls can bo selected in C\ . This is the number of possible cases. their seats at a round table. 4. One of each suit can be selected in 1.

and this may occur in a^a* ways. and these are all equally likely. Or thus. . and there are the following possibilities : A A A A } and A 2 may both l may happen and A may happen and A l l and A 2 may both happen.Pi)(l . Suppose that A L can happen in a L ways and fail in 6 t ways. chance of throwing ace with a single die in two trials ? The chance of not throwing ace in one trial is -|. The probability of the concurrence of several independent events is the product of their separate probabilities.p 2 ). and this may occur in aj> } ways.(^ ) 2 = ^Q. Chance of at least one success yg + 2 ^ f J^~. Also suppose that A 2 can happen in a 2 ways and fail in 6 2 ways. both do not fail) is 1 -(1 -^(l -pzl^ 512 INTERDEPENDENT EVENTS Ex. Hence also the chance that one and only one of the events happens is PiO -ft) +^ 2 (1 -PiH Pi +Pz. of which the probabilities are p l and p 2 respectively. 2 fail. the chance of throwing at least one ace 1 . I. the chance that A 2 happens and A 1 fails is p 2 (l -pi) . fail. What is the.the happening of any one of them does not affect the happening of any of the others. ^ 2 -=a 2 /(a 2 f 6 2 ). the chance that both A l and A 2 fail is (1 .ZpiP* Also the chance that at least one of the events happens (i. fail. The total number of ways in which both A l and A 2 are concerned is (a l +b 1 )(a 2 -\-b. and this may occur in 6 t 6 2 ways. Laplace's third principle. It is sufficient to consider two independent events A t and A 2 . It should also be noticed that the chance that A l happens and A 2 fails is p t (l -p 2 ) . all equally likely. all of which are equally likely. 2 ).e. and this may occur in a 1 6 2 ways. Hence the chance th at both A 1 and A 2 occur is # 1 a 2 /(a 1 + 6 1 )(a 2 4-6 2 )j which is equal to P ip%. . the chance of not throwing ace in two trials is () a . Chance of success in both trials 3^. Then Pi = ai/(<*>i + b 1 ).'. Chance of success in the first trial and failure in the second \ . Chance of failure in the first and success in the second f -|.

Also if p 3 is the probability of a third event A 3 after A l and A% have happened. 2 d would fo would ce Ex. of throw Art.(|) n . another contains 5 whi black balls. Similarly the chance of choosing the second bag and drawing a white is \ f. te . explain why the following reasoning is inc orrect : The chance of throwing ace in the first trial is ^ and the chance of ace in the second trial is J. (f) w =4.'. if the reasoning were correct. then the probability that both A and A 2 happen is p } p 2 . In how many throws with a single die will it be an even chance that ace t urns up at least once ? The chance against ace turning up in n throws is (|) n . .Ex. we have l-() n = . iv the that it is white ? The chance that the first bag is chosen is i. It should be noticed that. 3. and the chance that it turns up at least once is 1 . Thus the events are not mutually exclusive and oes not apply. i. 4. 2. If a bag is chosen at random and a ball is drawn from it. For the reasoning of the last article holds good under these conditions. the chance of choosing the first bag and drawing a white is \ \ . the chance of drawing a white from a bag chosen at random is FREQUENCY 513 . which is obviously false. Interdependent Events.. n = 3-8 nearly. and the chance of drawing a white ball from it is . then the probability that A v A 2 and A 3 will happen is Pip^p^ and so on for any number of dependent events. Ex. If two events A v A 2 are such that p is the probability of A ly and p 2 is the probability of A 2 on the supposition that A l has happened. Hence it is less than an even chance that aco turns up once in 3 trials. therefore the chance of ace in two trials is -|. /.and 3 hat is chance One bag contains 3 white balls and 2 black balls . .". 1. it llow that the chance of throwing ace in 6 trials would be -f . and mor e than an even chance that it turns up in 4 trials.e. If then n is the required number. /. The throwing of ace in the first trial does not exclude the possibility ing ace in the second trial. Referring to the last example. that ace rtainly occur in 6 trials .

C*/C\ ~iy. 5. 2. 1 only applies to events of a restricted type . the chance o f its being white is -^. 4 . he must fail at the first and second trials and succeed at th e third . the required chance 10 -^ f -f -f '5. Suppose the balls drawn in the order indicated by wwwbb. will be a boy is about 0-51. If all the balls were put into one bag and a ball were drawn. of the children born in Europe are boys. In the second caw. . about to be born. Ex. Five balls are drawn. He continues doing this until he draivs a spade. ' in the long run/ For example : (i) It is an observed fact that about 51 per cent. (ii) exactly three trials ? The chance of success at any particular trial is \. as in Art. one by one. It is obvious that the definition in Art. 4 . in particular.NOTE. from a bag containing 6 white and 4 bla ck balls. Hence we say that the probability that a child. . 2.10 9 8 7 6Also the chance that 3 white and 2 black balls are drawn in any other particular order is obtained from the preceding by merely altering the order of the numerat ors. What is the chance tliat 3 white and 2 black balls are drawn ? First method.' such as are used by Insurance .'. which is equal to the result previously obtained. .'. It makes no difference to the result if we suppose the 5 balls dra wn simultaneously. 3. Another Way of estimating Probabilities. (ii) According to ' tables of mortality. A person draws a card from a pack. required chance (J) 2 ~-fij. that he will have to make (i) at least three trials. or is believed to occur. There are other cases in which probability is estimated by considering the frequency with which the event occurs. The number of orders in which this can occur is the number of permutations of th e letters wwwbb. Ex. Second method. . to games of chance. the required chance -=C'* . The chance that this occurs is easily seen to be _6_. Hence. What is the chance. and shuffle the pack. replaces it. required chance = (2) 2 ~~Q.*. and is therefore j 5 / 1 3 2 10 . he will h ave to fail at the first anct second trials . In the first case. which is not the same as before. . 3.

Companies. Next suppose that /1's ticket gives him a chance p t of receiving a p a chance p 2 of receiving a 2 . A man's expectation of life. This sum is the sum of the expectations arising from his chances of securing the separate sums. where N is a large number.188 men 50 years old. In general. 1. if an event has happened in pN trials out of a total of N trials. subject to risk. so that events need not be independent. This value is also called the expected value of P or the expectation with regard to P. where N is large. If a quantity P can assume the values P l9 P 2 . Suppose that a person (A) has a ticket in a lottery which gives him a chance p of a prize of a. If the lottery were held N times. If the lottery were held N times. and this is called his expectation. and this has been verified experimentally. 1). dying before the end of the n . jo 3 . we may expect him to get the prize about pN times. and so on. The average value of a quantity P. . we may expect him to receive a l on about pfl occasions. It should be observed that every p denotes the chance that P has the corresponding value. Thus in a large number of throws we may expect ace to turn up in about J of the total number of trials.. where N is a large number. If a six-faced die is thrown a large number of times. Ex. If p r is the ehanre that he will survive r years. we say that p is the probability that the event will happen on a fresh trial. Expectation. 514 EXPECTATION This notion of probability is not at variance with that contained in Laplace's first principle (Art. is the average value which P assumes in the long run. the average or the expected value of P is ^1^1+ j> 2 P 2 + p 3 P 3 -f. Altogether he may be expected to get about N(p l a l -f p 2 & 2 + ) Thus we may say that on an average he gets (p l a l 4-jt) 2 tf 2 -f ) f r a single lottery. Definitions. P 3 . and the chances that it has these values are respectively p v p 2 . 70... and so on. .. <2 2 on about p 2 N occasions.. This is merely a generalisation of what has been said in the case of a lottery. . Thus we may say that on an average he receives pa for a single lottery.. we expect one face to turn up about as often as another. and is called his expectation. is usually taken to mean tho average number of years which men of his age survive. 6. receiving pNa. .. out of 81.552 live to the age of 60. and from this point of view we say that the chance of throwing ace is |. We therefore say that the chance that a man of age 50 will live another 10 years is 70552/81188 = 0-869.

SUCCESSIVE EVENTS 515 Of these the number of ways in which 2 white balls can be drawn is C\ 3.r successes and r failures where r is the integral part of (n + 1) q . 5 = 15. which are mutually exclusive. Hence his chances of receiving 20s. Thus the event in question can happen in C ways. (2) We have C^ = C^ r . I f he is to receive 10s. .ext year.. For the chance of its happening r times and failing n . 2 black balls can be drawn is Cf = 10. . therefore the chances that the event happens exactly n times. 2. . A person draws 2 balls from a bag containing 3 white and 5 black balls.*. Now the number of different orders in which these things can occur is the number of different arrangements of r things of one sort and n . for every white ball which he draws and Is. 1 white and 1 black ball can bo drawn is 3 . 1 + p 2 . Successive Events. and 2s.C> n ~V + .. so that in the most probable case the ratio of the number of successes to the number of failures .. are respectively 3/28. . therefore the required chance is Cp r q n ~ r . are respectively the first. and the chance that it happens in any one of them is p r q n ~ r . third. If nq is an integer. that is of (n + ])q.r things of another sort. A closer e stimate will be given in another volume. expectation = (/ g .. 2 +p 3 .r times in a specified order is p r q n ~ r . exactly (n 2) times.. his expectation of life is roughly p l . lls. terms in the expansion (p + q) n =p n + C^p n ~ l q 4. . q be the respective chauces of the happening and of the failing of an event at a single trial. which is equal to C n r : . second. 7.. 10/28. 15/28.20 + |f 11 + Jf 2) =88.r. Let p.. exactly (n~l) times.. 9d. . Hence the most probable case is that of n . then : (1) The chance of its happening exactly r times in n trials ix C"p r q n ~ r . what is his expectation ? The number of ways in which 2 balls can be drawn is C\ 28. for every black b all. then r = nq and n r = np. Ex.. This number is I n/ \r \n . (3) The greatest term in the expansion is that in which r is the integral part of . 3 4-.

and the chances that these things occur are the successive terms of the first expression. or exactly (n . . A and B throw alternately with a single die. or in just (r + 1) trials. The chance of this is p C f j.+ Cr~l P^f^ which is the second of the expressions in question. the probability that the ratio of the number of successes to the number of trials differs from p by less than any assigned number.. is to receive 1. 1. chance that ace turns up (i) exa ctly 3 timej. A having the first throw. what is the. 2. 46 fe. or exactly r time s.. The proof of this is beyond our present scope. Ex. 9.is p : q.r -f + Vr-lP'q + ^iP^ + . or in just (r + 2) trials. we may use Stirling's theorem. // a die is thrown 5 times. however small. can be made as near to certainty as we choose.2) times. Or. which states that 516 PROBLEM OF POINTS (5) If u nfT is the chance that the event happens at least r times in n trials. What are their expectations ? . (4) It may be useful to state an important theorem due to Bernoulli : If the chance of an event happening on a single trial is p and if a number of independent trials are made. Th e person who first throws ace.1) out of the preceding (r + s -I) trials. That these two expressions are equal may be proved as in Exercise XXXV. (ii) at least 3 times ? Let x and y be the required chances. For instance. Ex. then (i) it must happen exactly n times. then and or by the second formula. and therefore Un. it must happen in the last of these and also in (r ..1 ^r-igS^Cy+*~ 1 ^ r r/ . we shall prove that (i) u nt r = p n + C% p n ~ l For if the event happens at least r times. . (ii) it must happen exactly r times in just r trials.+*. or in just n trials. by making the number of trials sufficiently great. Now that it may happen exactly r times in just (r + s) trials. and requires a use of approximate values of large factorials.. or exactly (n-1) times.

' A and B play a scries of game* which cannot b e drawn. i . ^ B chance==1 . Let the chances that A. J^a.. j8 things of another sort (J5). placed in a row.65 + . 5 \ 4 ! ! ! G r. throws are i 5 i /sy i /r>y i 6' 6*6' \Qj *6' \Qj '6"" . second.. If this is done six times.. and p. the card is replaced and the pack shuffled. C. 2 diamonds and 2 black cards ? . B occurs P times and C occurs y times (where a + p -f y = n) is Jl [*[(/ that is to say. what is the chance that the cards drawn are 2 hearts. A card is drawn from a pack. 4. therefore p + q + r = l.respectively. Consider the case of an experiment which at any trial must produce one of three results denoted by A. q are their respective chances of winning a single game. and y things of a third sort (C). CHANCE UNDER DIFFERENT CONDITIONS 517 (6) Several Alternatives. and the required chance is obtained by substituting these numbers for r and n in either of the formulae in (5). . What is t he chance that A wins m games before B wins n games ? Here A must win at least m games out of m + w-1. B. ' The Problem of Points. B. q.The chances that the stake is won at the first. Also the chance that in n trials A occurs a times. the chance is the term containing p a q^r y in the expansion of For the chance that A may happen a times. B j8 times and C y times in a specified order is p a qPr Y .' = . third. 3. Now the number of orders in which these things can occur is the number of arrangements of a things of one sort (A). So far.= - Hence. This number is In/ la I j8 \y.. the expectations of A and 5 are -ff and ^. . Hence the result follows. C occur at any trial be p. r respectively.. at each trial there have been two alternatives the event may succeed or it may fail. One of the three results has to occur. Ex.

2 3. we have ' __ p4 . if p' is the chan ce that 2 honours belong to A and 2 to the other players. 12 26 .At any trial. the chance that a heart or a diamond is drawn is ^=4 . third card dealt is a diamond is . Find th e chance that it consists of 2 hearts. The chance that the first card dealt is a heart is ^f . Let p be the chance that. of the 3 remaining honours. Hence by the preceding. 5. then ^_/T3 . . 3J. fourth card dealt is a diamond is ^f . and the chances of these possibili ties are respectively ^3 an d Y^. . 3j9 . 12. 2 diamonds and 2 black cards. 3 . 12. or is not. 6. 25 5~2~ * 5 1" * 50 * 40 * 48 * 4 7" The required chance is obtained by multiplying this number by [6/[2 [2 1_2 . The card turned up is. . Hence the chance that 2 hearts. 3 8 . Find the. chance that in a game of whist the dealer (A) has exactly two h onours. and is therefore equal to 518 INVERSE PROBABILITY Ex. 1 1 . 12 belo ng to A and 39 to the other players. P U 2 51 50 49 48 > . second card dealt is a heart is -^-f . A ' hand ' of six cards is dealt from a pack in the ordinary way. an honour . Of the remaining 51 cards. = 512* Ex. 2 diamonds and 2 black cards are dealt in this o r in any specified order is j^3 . 39. 1 is among the 12 cards an d 2 are among the 39. fifth card dealt is a black is ff. 1. where the card turned up is not an honour. sixth card dealt is a black is f-f . also the chance that a black card is drawn is ^. In the first case A certainly has one honour. P ^I 51 50 49In the second case.

therefore Ci + e 2 + <?3 = l and with similar values for # 2 Qv It is to be noticed that p l is the probability that. a ball is drawn from it and is found to be white. . and so for p^ p z . B.p z ' Ps> where Pi^ 1 /(a 1 + 6 1 ). It is required to find the probabilities Q l9 Q%. therefore any one of the (a^ -|..a t x a^y a 3 z black .Q 2 :Q3 = <*>& : a z y : a 3 z ^p l '. the probabilities in question remain unchanged: ABC white . C contains v)hite and black balls. Choose these so that x(a l + 6j) ^y(a 2 -t.b l 6 2 6 3 A bay is chosen at random. the required chance^ -$p +-sp f = 8. B. If the numbers of the balls are altered as below. and so for the other possibilities. C respectively. Ex. the event will occur on th e supposition thai the ball comes from A..6 2 ) ^ z(a 3 + 6 3 ). C l9 actually led to the event ? This question is said to be one of inverse probability.a l a 2 a s black . Q% that the ball came from A. If the ball which was drawn came from A. Probability of Causes. P 2 = Now a white ball is drawn.. z are any numbers whatever. Each of three bags A... The three bags now contain the same number of balls. Therefore Qi'. the numbers o f which are as follows : A B C white ..a^y + a 3 z) white balls is as likely to be drawn as another. What is the probability that a specified cause. then it belonged to the group a x x of white balls.b v x b^y b^z when #./. 1. ?/. (1) Suppose that an event has happened which must have arisen from one of a certain number of causes C l9 C 2 . PROBABILITY OF FUTURE EVENTS 519 . .

NOTE. Q s . Q 3 . Then the probability Q r of the existence of the cause C r . estimated before the event took place. Suppose that an event has occurred which must have been due to one of the causes.. It is usual to call P 19 P 2 . so also for (? 2 . Then any one of the white balls is as likely to be drawn as another.. P n the a priori probabilities of the e . m z such as B. P 3 have similar meanings. C bag respectively. . Alter the numbers of the balls as in the preceding. and m 3 such as C. 2. C n . except that there are ra x bags such as A cont aining a x white and b t black balls. as in Ex.. Q l + Q z -f <? 3 = 1. Let p r be the probability of the event on the assumption that the cause C r exists. If PI is the chance estimated before the event that an A bag will be chose n and P 2 . . Al so Q lt Q 2t Q 3 are the chances that the ball came from an A y B. etc. C l9 C 2 . therefore and as before. Let P r be the probability of the existence of the cause C r . Gl + 02 whence the result follows. therefore Q x : Q 2 : Q 3 =P l p l : P 2 P 2 * PsPa and Qi^PiPiKP&i + PsP* + ?&*). therefore Gi = ^iPil ("hPi + m *Pz + 3#j) with similar values for Q 2 . The same as the preceding. and that from an . NOTE. (2) A General Statement. 1. estimated after the event has occurred. Observe that P J p l is the chance that a white ball is drawn. we have P l : P 2 : PS^W! : W 2 : m 3 . is given by For an argument similar to that in the last two examples shows that Qi : % : 3 : = and since the event has happened. and Q l is the chance that it c omes from one of the m l groups of ax balls.Ex.4 bag..

1. The product P r p r is the antecedent probability that the event will occur. face downwards. and of these it is equally likdy that 0. $#. . and using a notation simila r to the above. The argument depends on the assumption (1) that Q lf $ 2 >"' ar ^ proportional t o PiPit P 2 Pz> which is justified in (1) In particular. . 1. we have p _ _p _i . 520 A POSTERIORI PROBABILITY Ex.. Two cards are drawn and are found to be spades. A pack of cards is counted.. 4 .. A ball is drawn and is found to be white. the number of white balls may be 0. the odds in favour of it s having occurred from the first cause are as P 1 p l : P^Pf The way in which these principles are applied to determine the Probability of Fu ture Events is illustrated in Ex. . There are two possible hypotheses : . 5 are white. 5.5 2X Hence the required chance liEx. C 5 . What is the chance that this is the only white ball ? There are 6 possible hypotheses . and that from the rth cause. 2. 3. C lt ..\ ^ . . Q n the a posteriori probabilities. A bag contains 5 balls. if an event is due to one of two causes. and it is found that one card is missing. /.. 2^ ~~J J 1 . 2. p 2 5 . Denoting these possibilities by 6 Y . What are the odds again st the missing card being a spade ? The 4 event ' is that two spades are drawn. 3.xistence of the causes and Q l9 Q 2 . what is the chance that a second drawing will give a white ball ? ~6. p l 5 .u. 4. the required chance = Q l P l p l --.^ 5 73 . 4. nm i ^n _. 1.i 2 _5. 2. // ut <Ae last example the ball which was drawn is replaced. ^5 5 > 1 1+2 + 3 + 4 + 5 1 g ~5~~~" = 2 . ana p Q . 3.

. _ * * __ .. all possible numb ers being equally likely... P z = %. m-g-rH-1. . (r + q.. .. are that the number of white balls are m q. then Q r u r v T jS t \p \q \m-\-\ where a^irVr-i ^ (by Ch.. The chances p l9 p 2 of the event under the hypotheses C\.. . (n-r+p-l). p. Denote these by C lf C 2 .. the chance that another drawing will give a white ba ll is The possible hypotheses. p 2 .. another drawing gives a whit e ball.. 1. m q lf . we have ^ /?m r-r+l nq \-r-\\rm _/j 7/ . are the a posteriori probabilities of C lt C 2 . C m _ q _^ r If p L . A bag contains m balls which are either white or black.. . _ . The a priori probabilities P 19 1\ of C 19 C 2 are P l J... If p white and q black balls have been drawn in p + q succ essive trials without replacement.1). . C r . 10. . . all equally likely. and A is independent of r. 4. . If Q 19 Q& . . (C 2 ) The missing card is not a spade. with hypothesis <7 r . Ex. Pr'-^p '^q l^p^q ^ u r v r> where u r r(r + l)(r + 2) . C 2 are Mil 1312 * 5TO> P'2. _ __ _.(Cj) The missing card is a spade. are the antecedent probabilities of the event under these h ypotheses. XXI.) VALUE OF TESTIMONY 521 If p r ' is the chance that. m-q-p-r+l n-l -r n^ f -. = 5~l ' 5~0 The odds against the missing card being a spade are as Ex.. ...

Thus /S" can be obtained from S by writing n . Such an application is open to adverse criticism. (ii) are P t 1/n. the chance that they both say it is true is pp'. the procedure is as follows.l)/n. and A a nd B both assert tJiat it is white. If p v p 2 are the chances of the event under (i) and (ii). 1. namely : (i) that to each witness there pertains a constant p (his credibility].1 for n and p for ^ . and the chance that another drawing gives a white ball is r r~r*QrPr' = S'l(m-q-p)S.. what are the odds in favour of the truth of a statement which A a nd B concur in making ? The 4 event ' is the agreement of A and B in making a certain statement. If we are prepared to make these assumptions. the chance that they both say it is true is (1 -p)(l -p')> Thus the antecedent probabilities of the event on the hypotheses (i). If it is false. of (i).r m-q-p m-q-p* thus p n l =0. It rests on two assumptions which can hardly be justified. If it is true. one of which is ivhite. Value of Testimony. A bag contains n balls.1) balls remain in the bag. (ii) it is false. (ii) are pp' and (\-p)(\-p') t and the odds in favour of the truth of the statement are as pp' : (1 -p)(l -p') Ex. (n-r +p).. 2. the chance = 9. What are the odds in favour of its being white ? The ' event ' is the agreement of A and B in making a statement. The a priori probabilities P 19 P 2 . In the case of (ii).1 . (ii) it is false. P^ (n. The chance that A should choose this ball and wrongly assert that it was drawn from . (n . we have p ~pp'. where S' = 2 r r ^\(n-l ~r)(n-r) . The possibiliti es are (i) it is true. which measures the average frequency with which he speaks the truth. The probabilities that A and B speak the truth are p. and one of these is white. and hen ce -_ T"T and j . Ex. The theory of probability has been used to estimate the value of the testimony of witnesses. (ii) that the statements of witnesses are independent of one another in the sense required in the theory of probability. p f respectively. A ball is drawn from the bag. If p is the probability that a statement made by A is true and p / has a similar meaning for B. The possibilities are : (i) the statement is true .

P . is (a-b) 2 ja z . where b<a. Q. OB. Two points P. ^f . PB~y. O P B' B FIG. P lies in A'B'. hence Pt =(l-p)(l-p')(n-l). Let AP = x. Hence PQ>cuJ% if. 2. PB^xy. and the odds in favour of a white ball having been drawn are as 522 POINTS TAKEN AT RANDOM 10. the ordinates at these points are equal to a . B' are the mid-points of OA. The following statements are axiomatic : (i) If a point is taken at random on a given straight line AB. Q. If A'. and the number of favourable cases by PQ. the chance that it falls on a particular segment PQ of the line is PQ/AB. 86. Ex. Sh ow that the chance that PQ>b. Take a point P at random in AB. that is \. The chance that B should do the same thing is (1 p') (n. Or we may say that the total number of cases is represented by AB. Geometrical Applications. show that it is an even chance that Let AB be a diameter of a circle with centre and radius a. then x + y~2a. Q are taken at random on a straight line OA of length a.1) . A as in the order 0. the chance that the point falls on a is a/S. and all values of x between and 2a are equally likely.the bag is (lp)(n-l). Ex. Draw the ordinate PQ. L Given that x + y-2a where a is constant and that all values of x between and 2a are equally likely. then PQ 2 = AP . P. and only if. The points are as likely to fall in the order 0. Hence the chance that z//>f a 2 is A'tt'/AB. (ii) If a point is taken at random on an area S which includes an area o-.

PA. Complete the figure as in the diagram. where OL-^PR^b. 4 red and 5 black balls. what is the chance that (i) they are both red ? (ii) one is red and the other black ? 2. A bag contains ra white and n black balls. two are bla ck and two are white or red. A OAA' CALCULATION OF CHANCE 523 EXERCISE LV 1. the favourable cases are those in which Q lies in EA. by the area of the shaded rectangle.. i. 4 red and 5 black balls. If p+q balls are drawn without replacement.e. and the total number of favourable cases by the area of the trianglq LMA' . a rim finisim+n is C p . We may therefore suppose that Q is to the right of P.) 4.'. 87. Hence the total number of cases is represented by area of the triangle OAA'. the number of cases in which the distance of P from lies between x and x + $x and Q is in PA.^/C p+g . In the last question what is the chance that at least three of the four balls . If two balls are drawn from a bag containing 2 white. Draw OA ' at right angles to OA and equal to it. is represented by Sx. 3. Of these. what is the chance that exactly two of them are black ? (i. If four balls are drawn from a bag containing 3 white. the chance that these consist of p white and q black balls . If 8x is small. . the required chance = FIG. A.e. and their number is represented by the upper part of the shaded rectangle cut off by LM.

B are two. Prove also that this expression is inialtered by interchanging m and n. form single rank. 8. king. (16.] 7. There are C a n possible cases. queen. as five-five-six. two are chosen at random . show that . knave (of the same or of different suits) ? 11. m balls arc drawn simultaneously and replaced . From a bag containing a balls.C. find the chance that (i) all the /'s are together.' Show that the chance that the same letter is chosen on both occasions is |. prove that the chance that their sum is odd is yf 6. Find the chance that two specified men are next one another in the same four. A party of n men of whom A.} If two squares are chosen at random on a chess board. The favourable cases consist of selections of r balls out of the m already drawn. Of these 14 are favourable.] 12. (iii) all different ? Explain why the sum of these probabilities is unity.] B. then n balls are drawn. 524 INDEPENDENT EVENTS 15. A letter is chosen at random out of ' assinine ' and one is chosen at random out of i assassin. with selectio ns n-r out of a-m not already drawn. [(ii) Any particular throw of this sort. (ii) two alike and the third different. (ii) exactly 12. and verify the truth of this statement. 13. In a throw with three dice what is the chance that the dice fall (i) all alike.drawn from the bag are black ? 5. Show that the chance that exactly r balls are common to the two drawings is 'C'^~/( '^. A squad of 4w men form fours. (ii) no two 's are together. If the letters of ' attempt ' are written down at random. what is the chance that they are ace. What is the chance that (i) A 9 B are next one another. A. (iii) less than ^2 ? 9. occurs in 3 cases. What is the chance of throwing 10 in a single cast with three dice ? 8. Out of 20 consecutive numbers. (iii) that not more than m men are between them ? 14. In a single cast with four dice what is the chance of throwing (i) two doublets. [Consider the second drawing. (ii) that exactly m men are between them. How many times must a pair of dice be thrown that it may be at least an even chance that double six may occur at least once ? 10. If four cards are drawn from a pack. 5 throws of this kind. an d there are 6 . [The number of possible cases is 7 .

The fact that Ti the probabilities. chance .] 17..] 23. If v4's chance of winning a single game against B is f .. 3.+ .2 + 24. (iii) a lower card than B. (ii) at least two aces ? [For (ii). the number of favourable cases is 4.. A draws a card from a pack of n cards marked 1. n. A bag contains n tickets numbered 1. A bag contains n cards marked 1. and show that p ~". Find the chance that A draws (i) the same card as B . number of favourable cases -C\ . pi are the probabilities of four independent events. find the chance that (i) two and not more than two of the events happen . If p l9 pz< p 3 . 6*2 8 +48. (ii) four games (at le ast) out of the first six. 2. 2. etc. and let^/ be the chance of drawing a white ball from a bag containing (a + a') white and (b + b f ) black balls. 3..] 21.1 n ^n n n } J 20. .. Verify that the sum of these chances is unity. . 3. n . (ii) the chance that they have contact at a corner is yj^. (ii) at least two of the events happen. If A is to draw all the cards in succession and is to receive 1 shilling for every card which comes out in its proper order.{-.(i) the chance that they have a side common is y\. another contains a' white and b' black balls. Find p. What is his expectation ? 22. on which the several expectations depend. are not independent does not affect the argument. [The expectation arising from each ball is ~ of a shilling. 112 n 1 [For (ii). prove that his expectation is 1 shilling. p r according as a-\-b ~a' ~b' and ab' a'b have the same or opposite signs.+ 1. 2. . the card is replaced in the pack and then B draws a card.+ . 18. p'. C -\-C\ .4. .. n. and is to receive a number of shillings equal to the product of the num bers drawn. [(i) Out of 64 x 63 possible cases. Let p be the chance of drawing a white ball from a bag chosen at random. One bag contains a white balls and 6 black balls .. A person draws two tickets at once... What is the chance that a hand of five cards contains (i) exactly two aces . (ii) a higher card than B . find /Ts chance of winning (i) two games (at least) out of the first three . 19.3 + 36.

a 2 x 2 + . whence it follows that A and B have equal chances of winning the set. For 2 of t^em. 10 in ( .5} + (n . What is the chance that the second can be subtracted from the first without ' borrowing ' ? 26. three. the chance that the server wins a game is always p. Ibhe number of favourable cases = 2(n . B. 27. (ii) if this card is not an honour . 2.. Denoting /Ts chance of winning the set by x. (Huyghens. Hence.. the chance that he loses is q. C can sit in n . In a set of tennis between A and 7J. + x 6 ) 2 . In a game of whist.) ^2 f Three players A . (iii) before the last card dealt is turned up. (ii) if this card is not an honour .] \ JL x / GAMES OF SKILL 525 25. 31. they proceed to play deuce and vantage games and A has the service. The decimal part of the logarithms of two numbers taken at random are found to seven places. Two persons each make a single throw with a pair of dice. C can sit in n-5 places . (iii) before the last card is turned up.] 29.. Out of n persons sitting at a round table. C. B. The score is five all. then the chance = Q A ~ 2 ( a * 2 + + a n 2 ) : t>D f\ X* \ 4 and this is equal to {coefficient of a. Prove that the chance that no two of these three are sitting next one another is (n-)(n ~ 5)/(n ~l)(n-2). for n . are chosen at random.5 of them. . find the chance that each of the four players should have an honour : (i) if the card turned up is an honour . Three cards are drawn from a bag containing n cards marked 1. g.6 places.6). show that x=pq + (p 2 + q*)x. . In a game of whist.5) (n . Find the chance that (i) the three cards form a sequence . C of equal skill engage in a match consisting of a .+ a 12 x 12 . A and B play with two dice on the condition that A wins if he throws 6 before B throws 7 and B wins if he throws 7 before A throws 6. (iii) they involve no sequence. Show that A's chance is to J5's chance as 30 : 31. . 13 can sit in ri-~3 places. A.24. What is the chance that the throws are equal ? [If (x + x 2 + . find the chance that each party has two honours : (i) if the card turned up is an honour .. [Relative to A. 3. (ii) they contain a sequence of two ..j }/36 2 . 30. n.

2. The sides of a rectangle are chosen at random. except that each ball is as. B. 13 plays O. OACB. and it is equally likely that 1. A bag contains 3 balls. What is the chance that all are white ? 36. each less than a given length a. (i) What is the chance that this is the only white ball ? (ii) What is the chance that another drawing will give a white ball ? 34. 2 or all of them ar e white. The chances that A . p. and so on. B. A plays B. p". [Proceeding as in Ex. P l9 P 2 ^3 are ^ ne terms in t he expansion of (i + i) 3 .] 526 GEOMETRICAL PROBLEMS 35. we see that P .] 38. A ball is drawn and is found to be white. C speak the truth are respectively p.series of games in which . all such lengths being equally likely. 1. likely as not to be white. [First show that if ^4 loses the first game. Two balls are drawn. The rest proceeds as in Ex. and the arc AB. A circular disc of diameter c is thrown on . A floor is paved with rectangular bricks. and it is not known how many of these are white. each of length a and breadth b. [Draw a square. C plays A . Show that the chance that the diagonal is less than a is 7T/4. B assert to have happened and C denies ? Geometrical 37. his chance of winning the match is ^ : if he wins the first game. in rotation. A bag contains 5 balls.-1 plays 13. What are the odds in favour of an event actually having happened which (i) all three assert to have happened ? (ii) A. his chance is JJ. p'. the favourable cases are when P lies in the quadrant GAB.] 33. If ONPM is one of the rectangles. p. The same question as the last. 520. 520. whose side is a . Prove that the chances which A . and these are white. The player who first wins two consecutive games wins the match. C respectively have of winning the match are as 12 : 20 : 17.

Choose Q on one semicircle on PP' . Show that the chance that the greater of the parts AP. the chance that P'Q' lies entirely within PQ is (a-b)la. R. [(ii) Let AP==x. 89. P' must lie in LM. the chance that their distance apart is greater than the radius of the circle is -|. 39. ^g=rz. P'Q' is less than c is c*/a&. P.c . 90. the chance that they do not lie on the same semicircle is J. On a straight line AB of length a + b. Q. If two points are taken at random on the circumference of a circle. a or b.the floor. k. PB is at least k times the smaller is 2/(k+l). 40. of lengths a. 41. A point P is taken at random in a straight line AB. Hence. etc. number of favourable cases- Jo total number ic . (ii) If c is less than. [The unfavourable cases are represented by the thick arc in Fig. If <2=c.) If PQs. 6 respectively. two segments PQ 9 P'Q'.] 42. the chance t hat the common part of PQ. (Fig. and P'Q' overlaps PQ towards B. 89. If three points. are measured at random. [Choose P.] one brick is FIG. Show that the chance that it falls entirely on (a-c)(b-c)lab. (i) If a>6. c. are taken at random on the circumference of a circle. E must lie on P'Q'.

and N is not a perfect square. it follows that any simple recurring continued fraction is equal to a quadratic surd. and is of special importance in the theory : it includes the forms \/AjB and JN. then It follows that every positive quadratic surd or its reciprocal is of one of . its value is of the form (JN b^)/r v where N. XXIV. it is to be observed that : (i) There is no loss of generality in assuming that N .x)dx =-| c 2 . UNb 1 )/r 1 can be expressed as a simple recurring continued fraction. Total number of cases ab. In every case it is assumed that (i) The surd in question is greater than unity.b^ is divisible by r v For (JNb 1 )/r l = (jNr 1 *b 1 r l )/r 1 * and Nrf-farJ* is divisible by r^ 2 . Jo total number of favourable cases c 2 . For if ^V-^i 2 =^ 2 where r 2 is a positive integer. 33. This will be called the normal type. except that 6 t may be zero.] CHAPTER XXXIII CONTINUED FRACTIONS (2) EXPRESSION OP A QUADRATIC SURD AS A SIMPLE CONTINUED FRACTION 1. hence (^/N b 1 )/r 1 can always be replaced by an expression of the same form for which the above condition holds. In considering this theorem. (ii) We need only consider surds which are greater than unity. 2. . the number of favourable cases = I (c . etc. We shall prove that conversely any positive number of the form. In other words. Surds of the Form (JNb l )/r v From Oh. (ii) N ~b^ is divisible by r v (A) The type (JN + bJ/Tt where b l *<N. b l9 r } are positive integers. Types of Quadratic Surds.four types considered in the next article.P L M Q FIG. 91.

2a 1 6 1 ).. (C) and therefore b 2 <JN< b 2 4. r 2 are positive integers. and since N -b^ 2 is divisible by r p so also is N -b 2 2 ...................528 SURDS OF NORMAL TYPE To express (^N + b^)/^ as a simple continued fraction...... Hence 6 2 is positive.. Finally....... _ n ... we form the equations .. therefore a 2 is a positive integer.. .. Again... it will be seen that 6 3 andr 3 are positive integers. . . where a n is an integer such that (F) ... 6 2 is a positive number less than ^/-ZV.... and so for n = 3....... For taking n = 2. 6...r x ....... Continuing the process.... Reasoning as before. (G) leading to By steps similar to the above... &!?!<*/ N + b^af. 3. by the preceding (*JN + b 2 )/r 2 >l... 5. This is contrary to the supposition that 6 2 < 0.a .U-1 1 W"i f where a a is the integral part of (^/N -i-b l )/r li and kj^'Vi-fri* r I r 2 = N-b^ 2 .....b^ is divisible by r 2 . 4.... r is a positive integer. . and consequently r 2 is positive..L + r!. Further.. for the nth complete quotient is ..... but 6 1 < N /^ r > therefore 6 1 <r l and consequently 6 1 <a 1 r 1 ... therefore N 6 2 2 >0...... (B) We shall first show that 6 2 .^) 2 = N-b*-r l (afa . we form the equations for n = 2.. the fraction is periodic. we can show that every a... it follows that N /A r <r 1 ... Since ^ is the integral part of (*JN + b l )/r l . we have N-b 2 * = N. ............. Also N ... (D) If we suppose that 6 2 <0. Hence r 2 is a positive integer.. and it is obviously an integer...(a x f j .

Here the second complete quotient is a surd of normal type. N/57-7 -/ X or instance.. all the succeeding a's recur and in the same order : the same is true for the fe's and r's. We form the equations r z 0l+ __.and for every n. />/57 + 7 1 1 . where a x .= 1 s/57-7 / 4 3+l-t-l+l+l4-7 and dim r ~~p _ __ _ . are the integral parts of the fractions on the left and. From this stage. Hence the fraction (JN + b n )/r n cannot have more than 22V distinct values. (B) The lype (*JN -ftj)'?*! where /> 1 2 <A T .1 1 1 1 T. a 2 .. for every n.= 1 / = --. . "w-f-l " * } n ~~ a n r n> r n r n+ 1 ~ ^n 4-1 ~ .% 2 2 / -1 (C) The form (bi + JN)/r i where b 2 >N. N/57-7 / 2 .. i-t i --. b n <JN y therefore r n = (b n + b n+l )/a n <2JN. and one of the complete quotients must occur again. REDUCTION TO NORMAL TYPE 529 It will now be shown that in the process of expressing a positive quadratic surd of any other type as a simple continued fraction. Thus the continued fraction is periodic. a stage must occur at which the complete quotient is a surd of normal type. Whence it ivill follow that in every case the continued fraction is periodic. .

1. .As in (A). In practice we replace the written work involving surds by an easy mental process. it has an acyclic part consisting of a single quotient. Express (\/37 -f 8)/9 as a simple continued fraction. if JN> 6 -f r. that is. from Chap.N /A r + 6)/r< . Now (n + 1) being even. (ii) i/" ( .JN>b + r. as in the next example. Here the second complete quotient is a surd of type (C)./iV + &)/r . are all greater than ^N . arid b l9 b 2 .. (iii) if b > JN. then (i) if b<>JN<b + r.. then (rx-b) 2 = N and the second root of this quadratic is ( .b n f t > 0. ^=9. and since also JN . it follows that A r ~^ 4 . and b n +. For if x = (JN + b)/r . Method of Reckoning. and (iii) if (~JN + b)fr>0.JN>a n r n . r n are all positive.. 6 n41 is a decreasing sequence of integers .. The various quantities are now found in succession from the equations . and therefore r n+ 1 is negative. I. that is. the fraction has no acyclic part. the acyclic part consists of a single quotient . Let the surd (JN + b)/r be expressed as a simple recurring fraction. Ex. and by hypothesis b^ > S /JV.. n cannot be even . the fraction has no acyclic part and these conditions are equivalent to 6< N /2V<6-hr. 1 >0. 6 r . . 33. Consequently x /-ZV + 6 n+l >0. if A>B. Theorem. (D) The type (b l ~JN)/r 1 where 6 1 2 >^V. hence. the acyclic part contains one or more quotients. for in that case we should have and therefore b n+l >JN. Here a i integral part of the surd = l. it has an acyclic part of one or more quotients. Hence a stage must occur at which This being so. the simple continued fraction equivalent to v A/B has a single non-recurring quotient. it follows that (i) t/ -1<( . 530 MENTAL PROCESS OF RECKONING 3. Suppose that 6 15 6 2 .N /2V + 6)/r<0. the (n + l)th complete quotient may be written and since ^w+i^^ 7 ? ^i s or ^^ ie next quotient is a surd of normal type according as 6 W a. every r is an integer and every b is an integer or zero. 4. then r ly r 2 . .. XXIV. Hence n is odd. if b>JN. 1 0. (ii) if . In particular. also ^ 8.

fe n = fl n. 7 .-i r n-i -&-n r n = ( N ~ W r n-i> n = integral part of giving the table : n 1 2 3 4 6 6 b 8 i 3 4 o 3 r 9 4.

3. for 3<<s/37< 3 -f 7. We have A/ -= = -5 = l . from this stage. is a repetition of the part between the dotted line s..3 4 7 a 1 1 1 3 2 1 The reckoning. marks the beginning of the recurring period. 3. (\/37+3)/7.M. It will be shown in the last article of t his chapter that.C. p rocess.. the above reckoning can be replaced by a G. for values of n = l. . and s/37+8 1111 9 " + i+I+3V2+"" Jf -XNOTE. CYCLE OF QUOTIENTS 631 . 5. and it is expressed as a simple continued fraction by constructing equations of the type . where N = AB. if continued. (1) It is supposed that A. r l = B. . >/A/B as a Continued Fraction. The surd is \ /> jD 7*j therefore of the type discussed in Art. The third complete quotient. B are positive integers and that A>B. 2. 6 a = 0.

6 m -6 n <r n _ 1 and 6 m -6 n <^n (n^2) (D) (3) The Cycle of Quotients.a 2 4. 6 n <\/^ r <^n + r n ( n ^2) (C) For any suffix w. 2. it has been shown in Art.a 3 -I. All this may be expressed by writing M_N/A[ _L_JL _L _J !_ \ B r l l a 2 -i. K -f . therefore. <. 111 TT I l 1 o J Hence a c .2a x + * " ' * * ._!== a 3 . Summary. therefore JN + b n >r n .V#Ai ~ i. as in Art.b n *)/r n = (JN .b w therefore ./#<& + r nHl (B) Again. etc.. (iii) for the rest of the cycle.Therefore by Ch. 3. an( i b n+l = leading to A/"S = As in Art. . the other root of this equation being . . 31. a l . 3 that a 1 is the only non-recurring quotient. = 2a t -f ^c+ a 2"^" a c-fl + and therefore a c+l 2a l9 a c = a%. b m <*JN.where a n is an integer such that a n < (JN + b n )fr n < a n + 1. (ii) the last partial quotient of the cycle is 2a l . Every complete quotient is greater than 1. (2) The following inequalities are required. the partial quotients equidistant from the beginning and end are equal." * a 3 -f. XXIV. every a. then /-^ 111 * * Now JNfr^-a^ is a root of r^ (x + a^^N. Let c be the number of elements in the cycle. Also.. the continued fraction equivalent to (*JN + b n )/r n has no acyclic part. It has been shown that \I^AjB can be expressed as a simple recurring fraction in which (i) there is a single non-recurring element.. (A) Also r n _! = (# .b n ) (JN -f b n )/r n >JN. . hence from (B) and (C). if n>l. 6 and r is a positive integer and the continued fraction is periodic.

(5) Calculation of the Quotients. Subtracting and using the hypothesis. b c ^ and is reciprocal. etc. i. This matter is settled by the following theorem. and the reciprocal parts of the a and r cycles correspond.b m _ l )/r m __ l are less than unity. a m are respectively equal to b n+} . The character of the recurrence and the reciprocity of the three sequences is exhibited below. a n . a 2 .. ?' c = /' 2 . 532 CALCULATION OF QUOTIENTS (4) The b and r Cycles. Hence. a 3 .. 1. a 3 . both (6 rn _ 1 .. and therefore b c + m b m and r c +m r m f r ^^2. . where the recurring periods are enclosed in brackets. 6 3 . . b c 6 3 . Therefore the right-hand side o . (D). we have and so on . .. . no further calculation is necessary. b c+ l = b 2 . 1 .e.and we shall call the sequence.. and is reciprocal after the first term. In finding the values of 6 n . r w .. r c . z/ we can teH w/ien the middle of the b cycle has been reached... l + b m and n+1 r n+1 = 6 w+2 + 6 n4 . 2.. we find that Now by the inequalities in (2).. ci c _ m ^a 2+m for m = 0. a n . (JN + b c+m )/r e+m = (JN 4.b m )/r m . i . a 2 . . hence. tt 2 a a 2 . From this stage all the complete quotients recur in the same order . Let c be the number of elements in the cycle. . r rn . so that jjf \ j * * The recurrence is due to the fact that the second complete quotient appears again as the (c + 2) th. For rn+i = (N-b$ + i)lr n = (N-b m *)!r m = r m _ r Also ^m-i r m-i := =b m _.e. T Q . r n . It should be noticed that the a and 6 cycles correspond.b n + 2 )/r n+l and (6 n42 . (i) the cycle of Vs is 6 2 . i. o a l a 2 & 3 . f c -i = r^ etc. If b m . the reciprocal part of the cycle of quotients. Kemembering that a c+1 = 2a 1 . (ii) the cycle of r's is r v r 2 . then b m and b n+l are equidistant from the ends of the b cycle.

for its last term is 2a x = 2. Ex. ^ m _ 2 a m-2 are respectively equal to 6 n+3 . Because 6 4 = 6 5 11. therefore a 4 is the inid-term of the reciprocal part of t he a cycle. we have the following : (i) If two consecutive b's.. In this case c is odd. and so on. ^ = 11. n+2 . a. (ii) // r n = r n+l and a n =^a n+l (so that two consecutive r's are equal and likewise the corresponding a's). CONVERGENTS 533 Putting m = n and m = n + 1 in succession. then r n . namely b m and b m + v are equal.* (ii) For /s/61 we have rj_ = l.. table is 60 [757546 r [I 12 3 4 9 5 5 .. r n+l and a n .f (A) is numerically less than unity. and consequently a w _i = a n+ i and b m _ l = b n+ 2 We can show by a similar argument that 6 TO _ 2 . Henco N 187. t hus V 17 111111 II" + 4~+ 8+ r+ 8"+ IT 2* x. 2 . In this case c is even. a x =7 and the 6. then r m and a m are the mid-terms of the reciprocal parts of the r and a cycles. Express ^/^i ana V^l as simple continued fractions. a l ~l t and proceeding as in Art.. (i) A/T^J. This cycle can therefore be completed. 4. r. we construct the following table by a mental process : 6 [11 13 11 11 r [11 6 3 22 a 1 [4 8 1 8 4 2].= V. 1. r n4 . . . and b n+l is the mid-term o f the b cycle.. a n+1 are the midterms of the reciprocal parts of the r and a cycles.

. Also the last quotient of the cycle is 14..-}... (j? n . therefore = *PnPn=i ?i zq n + ? w -i and Equating rational and irrational parts.= 05.J and let p r /q r be the rth convergent... 11111111111 Therefore v . j.. Suppose that where N = AB. then IN 1 11 /A 7 + h V t.... a 7 are the mid-terms o f the reciprocal part of the a cycle. Find two positive integral solutions of 11 x z .J+ 2+ 2+ 1+ --. Seeing that a e a 7 and r 6 r 7 . by equation (B). we conclude that a 6 . we find that r 3 r 6 = 3 . we find that Bp n *-Aq* =(l)r w+1 ..J+ _ .. 534 CONTINUED FRACTION FOR Solving for b n+1 and r n+1 and putting N = AB and r^B.. 1)..... V W-Hl .. If for some value of n it happens that (.as a continued fraction ( (5).. Also . r l ^B. then.l) n r n+l = M. q n ) is a soluti on of Ex..... Expressing ^/^f. * * (6) Relations connecting the p's and q's..a7 [1431221341 14].+ . (B) (7) Integral Solutions of Bx 2 -Ay 2 = M..^ B t ~ wHere 21 == ~~~ i ^i o-f a n + z r^. Ex.4+ . 1...17?/ 2 3.

...1)/2.37) are solutions.4). Also a n is the mid-term of the reciprocal part of the ' a 3 cycle. and so if r n = 2 we have a w >(a 1 H-6 fl )/2-l>6 n -l>6 n . (A) p n *-Nq n * =(-l)r n+1 .. ^N as a Continued Fraction... .. (B) CALCULATION OF CONVERGENTS 535 (3) Integral Solutions of x 2 -Ny 2 = M. and therefore (5. (i) In the cycle o/Vs. one member and only one is equal to 1. 1^4/^4=46/37.. c is even and a n is the mid-term of the reciprocal part of the a cycle. JAe number (c) of elements in the cycle belonging to JN is even. In particular.1... therefore 6 n = 6 w+1 ... namely r.... (46.lqi = 5/4.. (2) Equations (A) and (B) at the top of the page become ^?n?n-l-^n^n-l = (~l) W &4-l .. remembering that 6 2 = a 1 and using the equations _r ... But by the preceding. if c is the number of elements in the cycle belonging to JN.. (ii) // r n >2. Also a n is the integral part of (^i + b n )jr n ... therefore a n = b n . a n <6 n .. If./V ar + y --J f m r w+l 1V ~ we have a m = integral part of (JN + 6J/r TO . Also o = &H-&... For in this case n>l and so that (iii) // r n = 2.. so that a n = (a 1 + a n )/2 or (a l ^ a n . ^n a n ^b n ....% + & m = 6 2 + b m . 6. it is found that (-l) n r n+1 = M. therefore a^b n ... for some value of w. and therefore a n ^a l or a^-l... the necessary and sufficient condition that r n+l may be equal to 1 .. q n ) is a solution of x 2 -Ny* = M.. For suppose that r m = l... therefore Therefore ( N /^H-6 m+ i)/ i m ^i = ( N /^-r62)/ r 2> and ^tn^l marks the beginning of a new cycle of r's... Additional theorems are the following.. (1) Here r x = l and all the conclusions of the last article hold. then (p n . Hence by (5). and a n = a x or a l .. Finally a n is the integral part of (^i + n )/2. For a x is the integral part of <JN and x /^V>6 n ...... (i). then. Also 2a n = 6 n + 6 n+1 .

solutions are (p tc yjtc) when c is even and.. hence. _. where c is the number of elements in the cycle belonging to <JN.l) m q n+l * Hence any common factor of X and Y is a factor of q n and of q n+l . and to find p n /q n we need only calculate the (/'s and then use the first of equations (C) . Ex.1. the method giving no solution when c is even..5) is a solution. 1. q 5 ~ 5 and (18. (ii)/or the equation x* Ny 2 = I. we find that p n ~i = r n q n ~b n+l q n ^ and Nq n _ l ^r n p n ~-b n ^p n _ l . If we multiply equations (A). changing n into n 1 in equation (B) and using equation (A). 3. Also > 6 18. solutions are (p(2*-i)c> ?(2*-i)c) when c is odd . (B) firstly by q n and q n -i> and secondly by p n and p n _i> we obtain by addition Again....(PM C > ?2fc) when c is odd ._ ?c 2+ a m+ f i f 11 111 q n+l where ~ ~ -. Xq m -Yp m = (-I)-iq n and Xft^.Yp m . (i) // m + n = c. (4) Calculation of Convergents.13^/ 2 .. The following formulae are useful in this connection..^(.." ~ therefore If ^ is the numerator and Y the denominator of the last fraction. 2. . (D) In general. Find a positive integral solution of x 2 . .is that n = tc where = 1. the p's are considerably larger than the <?'s. (5) The Convergent p c /q c . . it follows that (i) for the equation x 2 -Ny 2 = l. We find that * * so that c = 5. then For .

.. (H) icA w equivalent to the two equations For using equations (E).. and.. 536 HALF-CYCLES (ii) In particular.. which are the results in question. so also are X and Y.and since these numbers are prime to one another.. that p m P c <^ r q m q C ' Tlie samc results follow in a similar way when c is odd... (F) // c is even and n \c. (iv) // N is a prime and c is even and n = ^c > then Pc + JNq e = l(p n +JNq n )*.. Pn + JNqn = (-I) m (P*-JNq m ){p e + JNq e ). (K) ... l>c=P*!7n+.. ...... n 2 .. First suppose that c is even...Nq n q e = ( .. p m q c -q m p c -(p m q m ^-q m p m ^)q n -=(-l)q r .1) M .1). (J) which is equivalent to p c = -J.. 25.... and the mid-term of the reciprocal part o f the a cycle is a^ or a. Also jt.W5Wi and ? c = ? n 2 + ?JM-i .. 2..... If m is even.... JN<p c /q c <p m /q m and p m p c >Nq n q e . and therefore (PmPc ~ N<lmqc) 2 -Pn 2 It remains to consider the sign of p m p c ... (iii) // m -f n = c... Thus the fractions pjq c and X/Y are in their lowest terms... p n p c ... if c is odd and n = |(c . ..p n q n .... it follows from Ch.....# (PmVc ~ VmPc? = (?> w 2 ~ ^?m 2 ) (^ 2 ~ W)Now p m 2 -Nq m ^(-lY"r m ^ V-AV-f-lJ'VHi. XXIV... we have r^^^r^^.. ....1.. Also we have the identity (PmPc ~ %n?c) 2 ..Nq m q c = ( .. and thus p m p c ...Nq m q c ..2Vy n 2 = ( ..........l)p n and p n q c ~ q n p c = ( l) n q n .I) w 7> n . therefore Pc/Vc = (Pn* + NqM> n <ln .. The convergent pjq^ precedes pjq c . It should be noticed that these results hold if c is replaced by tc... and therefore p c = X and q c = Y..(p n 2 + Nq n 2 ) and q c ..... For by equations (I). since mtn = c. If m is odd..

y . ^ 1+1+2+1+1+8 # * so that for \/21. V 2p w 7 n / V ?c 2 p w ? n ?c the sign being determined by the fact that p n <>JNq n > according as n is even or odd. the last fraction is in its lowest terms. MULTIPLE-CYCLES 537 Therefore p n * .Nq n 2 = ( .l) n . p c and q c may be found by using equation (K) and noting that p c /q c is in its lowest terms. 713 9. C^Q 2/ . and c is the number of quotients in the cycle. c = 6./T) 2 _ JV0 2 2 /-n 21 ^2 \7 2\ 9 and (2s_^L) ^(A) -A^-.. therefore Removing fractions and equating rational and irrational parts. p. #3 2.4 55 _ r ~ o n o = fo anc * therefore j9 6 . 2 and q c =p n qn Hence..t>5. 534. and the rest follows by (1). and p 81+21. Since N is a prime and p n is prime to q n .L * _L I . Vfi^ .Nq n ^ = ( . we have Again. and th e latter alternative is impossible. 2 or ( .(i) V Pn/q n is the n-th convergent of <yiV=ai+ J_. NOTE. Hence p n 2 ._ 9 J_.-i.l) n . When N is not a prime.l) n . Thus we find that >/21 = 4 + -i. ^ iw (6) The Convergent Pn-fec/qn+<c. (iii)._L. then Pn+tc =PnPtc + N( lnqtc and q n + tc =p n q tc + q n p tc (L) For the (c + l)th quotient is 2a x and the corresponding complete quotient is a x + N /A r . hence ( * ~ Nq " L ( -1)* ~. 1_ 11 ^_ . g 6 = 12..

we find that Pte + >/# fee = (Pc + V^c) 2 . Hence the smallest solution of the equatiort in question ). therefore any common factor of -X and y divides p n and y n . (Q ) Ex.. 1. ...% H.1.. by means of equations (I)..... L* Find a solution in positive integers of x* .. and generally...61g n a = ... equations (L) are equivalent to the single relation Putting t = 1 and n = c. Hence by equations (F). (P) and therefore p n+tc + JNq n+tc = (p n + JNq n ) (p c + JNq c y . q n ^ tc ^Y... Using equations (0) and observing that p n * . Hence Xj Y is in its lowest terms : so also is p n+t c/<ln+tc..Therefore p n+tc = X.Sly 2 = 1. In Ex. 21 + 453 .. 3c.. ( N ) in particular... # 533. As a special case we have P(s+t)c=PscPtc + N q sc q tc and q( S + t )c=Psc<Itc + y8cPtc'> ... Pn+tc ( a l + Pn/<In)Ptc + Ptc~l PnPtc + Afyngtc If JC is the numerator and Y the denominator of the last fraction.._p n ? n the last quotient being the (te + l)th. q n = q* + g c 2 . From this.... which are the equations in question. P 2 tc-Ptc 2 + Nq tc 2 and q^ ic ^^p tc q tc .... Pu -Ms +Pe?6 = 164 . and p 6 ~164.. the number of quotie being 11. in succession. p....29718...... ^61 is expressed as a continued fraction. P3C + V-%c = (Pc + J^Vc) 3 -. . The mid-elements of the cycle are the 5th and 6th.. e 58. p tc + JNq tc (p c + JNq c ) ( .21 2 + 58 2 = 380 5.. 2c. (0) 538 ABBREVIATED CALCULATION (ii) Since JN is irrational.. q 22 q 6 ~2l.... which are prime to one another. for every positive t. 58 . . nts in the cycle is (p 22 .. .. 2? 6 ^453. Ptc X-N qtc Y= Pn ( Ptl ?-Nq t *) and PtcY -q te X = q n ( Pt -Nq t *).. Now Ptc 2 -^9tc 2=: il ^7 (5)> (iii)....

....... Rule.. ....A 7 ?? 2 -= J-/1... . The Cycle belonging to z = ( N /N + b)/r... r' are both even and N. Suppose that where r' is a positive integer..... which is identical with (B) if q = ry....... 7.. Since N =6 2 -f-rr'. r..... In both cases.. (A) the latter pair being used only when r....... In case (B). 226153980)....... Hence N b 2 is divisible by 4.Pn ^Pn 2 + 61 ^u 2 = 2 ^ii 2 + 1 = 1766319049.... b are odd....... This fraction consists of one or more of the cycles belonging to (JN + b)/r. * Thi8 problem was set by Fermat as a challenge to the English mathematicians of his time...... .. 3)..... (D) For equation (C) is the same as qz 2 (p .............JN in the given surd............... . then the simple continued fraction corresponding to z has no acyclic part (Art. p-q' = 2by....... Proof........... p' ' r'y. (B) x /2.... q 22 = 2p n q u = 226153980.. (F ) where x is rational...................... (H) .... from (E) and (F)...... Express the result as a simple continued fraction with an even or an odd number of quotients according as the sign on the right of the chosen equation is + or ......... 77 are odd integers such that 2 .. y are positive integers such that x 2 ~-Ny 2 = 1.. where p y q y p'........... y = 77/2 where . q f are positive integers ... r f are both even.... for g= ry. q' =x-b\j ... we iind that p = x + by. Substitute x/y for .......... (B) (1) We shall prove that this equation can be written in the form z~(pz f p')/(qz + q'). GENERAL RULE 539 From (D)... Let (x> y) be a solution in positive integers of any one of the equations x 2 -Ny*= 1 or 4... (C) such that pq'-p'q=l and q'^q .r 2 . (G) and p/q = (x/y -f b)/r ... (E) where y may be any rational.. the surd (JN + b)/r is the positive root of rs 2 -2fe-r' = .. ......q')z p' = 0. and the cycle can be calculated as follows.... Thus the least solution is (1766319049. (E) we have -Kp + q') 2 ^Ny 2 I =. All possible cases are included in the following : (A) x....... p' = r'y.. and if A T is odd so also is 6.

It remains to be shown that q'^q. This will be so if x/y^.b + r. Three cases must be considered : (i) // x 2 - Ny 2 = ~ 1 or * - Afy 2 = - 4, then xly<^JN<b + r and q'<q. (ii) // x 2 - Ny 2 ^=I, the least value of y is 1, and

Now N<(b + r) 2 , therefore N**:(b + r) 2 - 1. Hence x/y^\/N -i-l^b + r and (f^q* (iii) // ~/2, y^rj/2 where z -Nr) 2 --^i, the least value of 77 is land xA/<V^TT. Let N=^(b + r) 2 - k so that A: is a positive integer, then r/==r(26 + r) - & where r, /' are even. Therefore A; is divisible by 4, and 4 is its least value. Hence x/7/<VA r Tl<6-fr and q'^q. (2) It follows that Up 11 -=a x + ----- ..- , ............................... ( J ) (7 2 -f a n where n is even or odd according as pq f -p'q = l or -1, then p'/q' is the convergent immediately preceding p/q. (Ch. XXIV, 12.) Now let z f be determined by 111 z = a l + - ... -7 , 2 + a n -f z then z = pz ' + P' and ^jtP'^^tP', ? 2/ + 2' J 2 ' + ?' 2 21 + ?' ' Hence 2;'=^ and the fraction (I) consists of one or more of the cycles belonging to z. The rule can be used (as in Ex. 2) to express any quadratic surd as a continued fraction, beginning at the stage where recurrence commences. 2M B- C - A

540 EXAMPLES OF USE OF RULE Ex. 1. Express (*JQ2l +21)/18 as a simple continued fraction.

Since 21< s /621<21 + 18, there is no acyclic part. Also 18r' = 621 -21 2 = 180,

giving r' =. 10. Thus r, r' are even and 6 is odd, also 25 2 -621. I 2 ^4. Using the rule, we substitute 25/1 for ^/621 in the given surd, and express the result as a simple continued fraction with an even number of quotients, thus 25 + 21 23 _ 9 1 1 1 ~~T8~~""" + r+ r+ 4' s/621 +21 n 1 1 1 and ^ - - = 2 + v - - . 18 1+1+4 * # Ex. 2. Express (^13-+ 7)/4 as a simple continued fraction. Since 7> N /13, there is an acyclic part. The reckoning on the right b 7, 1 shows that the second complete quotient is (/s/13 + l)/3, and since r 4, 3 l<*yi3<J \ 3, we can apply the rule. We have a 2, 1 3 2 -13.1 2 ^-4 and 18 2 - 13 . 5 2 - - 1. Since r --3, an odd number, the first equation is useless. Taking the second, we substitute 18/5 for jy/13 in (;s/13 + l)/3, and express the result as a simple c ontinued fraction with an odd number of quotients, thus 18 \1 23 , 1111 ___ L I ] ._ j^ _ ~ 1 -j ___ - ___ - 5/3 15 1 + 1 + 6 + 1* This is the cycle belonging to the given surd, and s/13 + 7 L JL _L _L 1 4 " + 1 | IT IT 6+ I'

EXERCISE LVI 1. Verify some of the following results, where the values of 6, r, a are given as far as the middle of the first cycle : fb. 0.3. 1.2 Cb. 0.4. 2. 3. 3 (i) s/13 -j r . 1 . 4 . 3 . 3 (ii) v/19 \ r . I . 3 . 5 . 2 . 5

I a . 3 . 1 . 1 . 1 I a . 4 . 2 . 1 . 3 . 1 Cb . 0.5. 1.4. 5. 5 T6.0. 6. 1.5. 4. 5. 5 , (Hi) ^31 j r . 1 . 6 . 5 . 3 . 2 . 3 (iv) ^43^ r . 1 . 7 . 6 . 3 . 9 . 2 . 9 I a . 5 . 1 . 1 . 3 . 6 . 3 [ a . 6 . 1 . 1 . 3 . 1 . 5 . 1 f 6 . . 9.1.8. 7.7 (v) s/91 \ r . I . 10 . 9 . 3 . 14 . 3 [ a . 9 . 1.1.5. 1.5 {b . . 10 . 8 . 7.5.9. 7 . 8 . 10 r . 1. 9.5. 12.7.4. 15.3. 3 a. 10. 2.3. 1.2.4. 1.6. 6

GENERAL CASES 541 r 6 . . o . a ("6. O.a-l.a-1 (vii) N/^Tl J r . 1 . 1 . 1 (viii) \l~tf~- 1 J r . 1 . 2(a - 1) . 1 l.a.2rt.2a [a.a-1. 1 . 2(a-~l) __ fb.Q.a.a f b . .a-l.a-2.a-2 (ix) VaM- 2 J r . 1 . 2 . 1 (x) Va 2 - 2 -| r . I . 2a - 3 . 2 . 2a - 3 I a . a . a . 2a I a . a - 1 . 1 . a - 2 . 1 (a>2) r 6 . . a . a p.O. a .a -2. 2 (xi) Va(a4 1H r . 1 . a . 1 (xii) Va^Ht -I r . 1 . 4 . a . a la.a.2.2a La.a.i(a-l). 1 -1 (a> 1 and odd) r 6 . . a 4 1 . a - 2 . a . a (xiii) *Ja(a 4 4) ^ r . 1 . 2a - 1 . 4 . a . 4 (a> 1 and odd) la.a + 1. 1 . -l(a - 1) .2 . (-!) 2, In certain cases, a solution of one of the equations x- - Ny 2 ~- 1 or i4 can be written down at once ; thus if A T = a~ 1 , then a 2 - N 1 2 - T 1 ; if A = a 2 J- 2, then (a 2 : j- 1 ) 2 - A T a 2 1 ; if A 7 = a (a 4- 1 ), then (2a -f 1 ) 2 - A . 2 2 -= 1 ; if A ^ a 2 i 4, then a 2 --A.I 2 ~ T 4. Apply this to write down the integral solutions of x 2 - A T ?/ 2 1 when N = 5l, 47, 56.

3, Find a solution of x 2 -109?/ 2 -4, and apply the H.C.F. process to show that v/109 + 7 JL _JL A JL J_ * __. ^ *- Y^ 2+ T+ 94 14 2 * ^- * 4, Obtain the following :

I4l4l4l42 ; ^ 6 1+ 84 24 14 34 2 ;

*

/17_ 9 111111 V3 = " + 24 F + ff IT 24*4*

5. Find two positive integral solutions of (i) 5^ 2 ~13t/ 2 ^7, (ii) 5x 2 -13i/ 2 ^ -7, (iii) 3x 2 - 17i/ 2 ^7. 6. Find positive integral solutions of x 1 ~~ 621?/ 2 ~4 and of a; 2 621z/ 2 1. Referring to Art. 7, verify that if 623/25 and 7775/312 are substituted for v/62 1 in the fraction ( N /62l421)/18, the first gives two cycles and the second three cycles of the fraction.

542 SUNDRY THEOREMS 7. If A=:(2ra + 3) 2 -4, prove that -l 111 _ ~~- JYI -I __ ___

___ __ 2+ m-f 1+ 4(w-f-l)+ 1 ' * # 8, If a is odd, find a positive integral solution of (i) x 2 -(a 2 + 4)y 2 = a; (ii) x 2 -(a 2 + 4)*/ 2 -^ -a.

9. Prove that r n ^2a lf and that if r n ~2a l9 then a w 1 and a n is the midterm of the reciprocal part of the a cycle. Show also that if A T is an odd prime and r n = 2a l9 then A r must be equal to 3. [If r w = 2fl,, n <(N/-V-f'6J/r w <], therefore 6 n + 6 n , 1 =a n r n = 2a 1 , so that b n ~b n+l . Hence a n is the mid-term of the reciprocal part of the a cycle. If N is an odd prime, we must have a n a.i or a t 1. Hence a x = 1 or 2, and N = 3, 5, 7. On trial it is found that JV 3.] 10. If r n = r nn = a 19 then a n = a w}1 = l, so that a n , a n+1 are the midterms of the reciprocal part of the a cycle. [a n ~ I(a l + b n )la l ~\ or 2, according as b n ^a l or b n = a l . If 6 n a x , &n+i = 2! -6 n ffj, so that 6 n ~ & w ,i. This is impossible. Hence a n = l and 4i = J(ffi + &nu)/i = l-] 11. If aj is an odd prime, or a power of such a number, and r n a^ then =!. [For r n r nn X-bn +l9 r n -ir n N-b n 2 , therefore b n and 6 nfl are solutions of a; 2 =JV(mod Gfj), and since each <!, and a l is an odd prime, b n ~\~b n+1 ~ a l9 therefore a n = (b n + b n t l }Ir n ^aja^ --!.] 12. If, after a certain stage, the simple continued fractions which are equivale nt to two irrationals z and z' are identical, prove that z' (pz+p')/(qz + q') where pq'~p'q=^:l where p, p', q 9 q' are integers. Show also that the converse is true.

13. If the number (c) of elements in the cycle belonging to \/N is even, (i) prove that Pc/zf ( Jci2 = (Pc-^^)l^c^ a^ c rding as c is or is not divisible by 4. (ii) By taking N = 21 9 illustrate the fact that equations (J) of Art. 6, (5), ( iv), are not necessarily true unless N is a prime, arid verify the equations for N = 19. 14. If the number (c) of elements in the cycle belonging to *JN is odd : (i) Prove that N is the sum of two squares which are prime to one another. Also, by taking A' 205, show that the converse of this is not generally true. (ii) If ra = J (c 1 ) and n J (c -f 1), prove that

and verify when iV = 29.

[(i) A T is a factor of p r 2 + L]

560

MISCELLANEOUS EXERCISES (A) L Show that the product of any two numbers, as 19 and 35, can be found as on the right.* In the left-hand column, each number is the jg quotient when the preceding number is divided by 2. Any 9 70 remainders are disregarded. In the right-hand column, 4 each number is twice the preceding. 2 The numbers 140 and 280 which are opposite even numbers 1 on the left are crossed out, and the sum of the other numbers 1 9 x 35 ~ 665 on the right is 19 x 35. [The reckoning shows that 19x35^(1 +2 + 2 2 2. Show that n 8 is the sum of n consecutive odd numbers, and find tjie middle or the two middle numbers. 3. Take any sum of money less than 12, say 9 8s. 7d. Reverse the order of pence and pounds, obtaining 7 8s. 9d. The difference between these sums is 1 19s. lOd. Reversing the order of pence and pounds, we obtain 10 19s. Id. Adding the last two sums, we have 12 18s. lid. Prove that the result will be the same whatever sum less than 12 is chosen. 4. If a/b is a proper fraction and q l9 q 2 , ... q n are the quotients and r t , r 2 , ... r n the remainders when b is divided by a, r 19 r 2 , . . ?"n~i> respectively, show that <^

q l where R n = ( - 1 )

. tfn b (ii) For some value of n, R n = 0, so that a/b can always be expressed in the form 1111 I h 9 p q r s

where p, q, ... are positive integers, their number being finite. (iii) Express in this way f-f and J-f . 5. Prove that a -f b -f c + d and a f 6 - c - d are factors of 2 (a 4 -4- 6 4 + c 4 -f d*) - (a 2 + 6 2 + c 2 -f d 2 ) 2 4- Sabcd, and find the remaining factors. 6. Solve

7. If x, y, z are real, prove that (x 2 -f y 2 + z 2 )/ (yz + zx + xy) cannot lie between 1 and 2. 8. (i) Find the simplest equation with integral coefficients which has -Vf + s/f and -Vf-Vf among its roots. What are the other roots of the equation ? (ii) Prove that 4(z 2 + #4 l)*-21x*(x+ l) 2 ^(z- l) 2 (2a; + l) 2 (z + 2) 2 . *Known as the Russian peasants' method of multiplication.

544 MISCELLANEOUS (A) 9. If #-ft/ + 2 = o; 2 -f ?/ 3 4-2 2 -2, show that

10. If a~\~b + c + d and a; f ?/ | 2 | t = 0, prove that the results of rational ising the equations *Jax -f *Jby + *Jez 4- V^ ~- 0, v/fa: 4- *Jay + \'dz -f >/cJ 0, are equivalent. State the two other equations which lead to equivalent results. 11. Find the equation whose roots arc given by %=*<-*, 2 + (ai a + s* a + S3 a ), (i=l, 2, 3), where x l9 x 2 , x% are the roots of x s ~ x 2 -f 4 ~ 0. 12. Solve the equation

being given that it has a pair of roots whose sum is zero. 13. If ^ _-f ------- f / + ------ and oH-6-f c }-d 0, prove that x nid x -me x f tno x + nm the only finite value of x is m(nc + bd)/(a + b). 14. If (ca' -ac') z <A(ab' -a'b)(bc/ -b'c), prove that 6 2 >ac and &' 2 >a'c'.

15. If a< 6 < c < r/, and a - c> ^ /> - r/, the roots of (x - a)(x - c) ^k(x - b)(x-d) are real for all values of k. 16. If l + m + n 0, show that the expression a 2 / 2 -I- b 2 m' z + c' 2 /i 2 2brmn ~ 2canl - 2ablm is positive if a&c(a + & + c) is positive.

17. Show that the expression (ex ~-az) z - (ay -bx)(bz -cy) is the product of two linear functions of x, y, z, and that, when a, b, c are real, the coefficien ts in the linear functions are real if 6 2 5? 4ac. 18. If the cubic derived from the equation abed - + --, -- -- -f ----- -,

x-\-a x + b x + c x + d has a pair of equal roots, then either one of the numbers a, b is equal to one o f the numbers c, d or 1111 -+T=-+1abed 19. Express in factors (be - a 2 ) (ca - b 2 ) + (ca- b 2 ) (ab - c 2 ) -f (ab ~ c 2 ) (be - a 2 ). 20. Show that if a, b, c are real, the roots of

are real. 21. Eliminate x, y, z from the equations : x y z __ x y z y z x z x y

y

MISCELLANEOUS (A) 545

22, Prove that 10 n - (5 + */l7) n - (5 - \/17) n is divisible by 2 n+1 . 23. (a) Solve (x- l) 3 + (z-2) 3 + (x -3) 3 + ... -f (x-n)^-O. (6) If a:, i/, z are unequal and

show that each product is equal to - 1 and that yz + zx -f xy ~ 3 (a: -f y 4- 2) -f 9 = and xyz-2(x + y + z) + 9~0.

24. By the method of partial fractions, prove that, if a, 6, c, ... ft are /& nu mbers, then the sum of n fractions of the type

(a-b)(a~~c)(a-d) ... (a-k) is zero if m is an integer less than n- I and is unity if m n - I . 25. If n is a positive integer, 21 24 2 9 - ... to n terms

\n + 2\n-2 + 4 jn-f3|n-3 _ _ , . * - '

26. Under what conditions are the following equations consistent ?

27. Show that, if the roots of x* - ax 1 -f 6x - c- are not in A.P., then there are in general three transformations of the form x y-\-\ such that the transformed cubic in y has its roots in G.P. 28. (a) Rrove algebraically that, if a, 6, a, j8 are real numbers, then

the positive root being indicated by the radical sign. (b) If a and b are positive and a-f-6=-l, show that \ 2 / 1 \2

29. Examine, for different positive or negative values of x and p, the conver-

gence or divergence of r\ r x " r-\ rN/(^ + 1 )-V w (1)27-; (11)^^5 -- . 30. Show how to find the nth term and the sum of n terms of a recurring series of which the scale of relation is of the form ^ n -f jm n _ 1 + ^ n _ a = ^ anc * of which the first four terms are known. Apply the method to the series (i) 1,2,5, 14..., (i>=-4, q = 3) (ii) 1,2,5, 12.... b--2, ^=-1) 31. (a) Show that 2Z'6 2 c 2 (6 2 -- c 2 ) (a 4 - 6 2 c 2 ) - 27 (a 2 6 2 4- a 2 c 2 - a 4 ) [(6 4 - c 2 a 2 ) (a 2 - 6 2 ) -f (c 4 - a 2 6 2 ) (c 2 - a 2 )]. (b) Eliminate x, y, z from x*y 4- 2 z = a 3 , y 2 z 4- 2/ 2 a: = 6 8 , z z x + z 2 y = c 3 , JM/Z = rf 3 .

546 MISCELLANEOUS (A) 32. (a) Show that any root of the number n can be expressed in the form

where each of the quantities a, 6, c, ... is n, 1 or n~ l . (b) If # is less than >/A r , but greater than %JN 9 show that

* AT *u ' -^ . # 1S nearer to N s than is either - - or x.

. 33. Show that ___ a z n(a i -a y ) (a-a 2 )" 2 ~ (a-a 3 )- 2 Write out the terms of the product in the numerator, and give the resulting expression its correct sign. 34. Discuss the reality of the roots of x 4 + 4z 3 - 2x* - I2x + a = 0, for all real values of a.

[Denoting the equation by/(#)0, consider the roots of f'(x)~Q.] 35. Solve the equation 46u; 3 -f 72# 2 -f 18&- 11=0, by expressing the left-hand side as the sum of two cubes. 36. If ( 1 + x) n 1 + c^x {- c z x z + ...+ c n x n , show that

[Denoting the expressions by ^ and v, show that f-==^-, etc.] ttic ct^c 37. Prove that there is only one set of real values of x, y t z which satisfy th e equation

and find them. [Writing a, 6, c, d, for (l-x), (x-y), (y-z)> z, we have a+b + c\-d = l and o + 6 a + c 2 4-d a = i ; hence Z (a-b) 2 = 0.] 38. If oj is a fifth root of 2 and a; = eo + w a , prove that

39. Prove that, if r 2 p 2 ^, the product of a pair of roots of the equation x 1 ~\-px* 4- qx 2 4- rx + 5 is equal to the product of the other pair. Solve x* - # 3 - 16# 2 - 2x + 4 = 0. 40. Obtain the roots of

in the form 41. Prove that a; 4 "* 1 + 1 = (3 + . 77 r 2 " 1 (a: 2 + 2o: cos r a -1), where n is a positive integer and (4?i-f l)a = 27r.

MISCELLANEOUS (A) 547 42. If three real numbers satisfy the equations x + y + z = 5 and yz 4- zx 4- xy = 8, prove that none of them is less than 1 nor greater than 2^. 43. Find a general formula for all the positive integers which, when divided by 5, 6, 7, will leave remainders 1, 2, 3 respectively, and show that 206 is the leant of them.

44. Show that a determinant of order n can be expressed as one of "order n-l as follows.

If

a l^l c l ^1

, then Jrt 1 "- 2 =

[Multiply A by

a t ...0 -&! tfi ...0 -c, a 1 ...0

-^ ... j 45. Show that, if n is a positive integer, the number of solutions of the equati on

where ^ and 7? 2 are positive integers or zero, is equal to N or N + 1 according as r is or is not equal to 1, where N denotes the quotient and r the remainder when n is divided by 6. ^3^3 3 3 a 2 3 3 6 2 46. Prove that 7 rr^-TT or ~rr > a > b being positive, (a-f r>) 4 4 4 4 4

47. If n is a positive integer, 3.4 n(n-\) 4.5 w(n-l)(n-2) JL ~\~ OIL ~\~ ~ ~ * ~ r ^r ~"7^

1 .2

48. If a, 6, c, d are real and

I 3

1.2

b 2 - ac ad -be ad -be r 2 - bd '

show that each ratio = ( ~ ) and that a*c + bd* + a*d* 0, the real values of the \a/ roots being taken in all cases. 49. Prove that the condition that (ax + by -\-cz) (bx + cy + az)(cx -\-ay + bz) can be expressed in the form P* + Q* where P and Q are linear functions of x 9 y> z is a 3 + 6 3 + c 3 ^3a6c; and find values of P and Q. 50. If x i9 x 2 , ... x n are positive numbers such that Zx=n, prove that Zx m ^n if

548 MISCELLANEOUS (A) 51. If r is an odd integer and q r =^ ~> , 7, - - } > prove that -

to i(r-f 1) terms is equal to |. a^-f tf 2 # (r + 1 ) a r+l - 2na r - (r - 1 ) fl^ - 0.

/ } 4, # n 52. If ( j -I -l-a^-f tf 2 # 2 -h... + # r r -f ... , show that

Also show that

53. Sum the series 27* ft ;

54. Examine the convergence of th series whose wth term is x n /(x* n +x n +1) for any value of x. 55. If m is a positive integer, show that

56. Prove that, as the real variable x changes steadily from - oo to -h oo f the function (x - a) 2 ' r -"V-6 1 ' where a and 6 are real and r/ ,/;, assumes twice over al] values except those in an interval of length 4(6 a) ; and locate this range of values precisely.

a a 3 a 4 - 1 |-0, 6 fr 1 b* - 1 r, c 3 ^* - 1

57. Prove that, if

and a, ft, c are all different, then 58. Prove that, in the expansion of (1 \-x) n (I x)' '* in ascending powers of # , where n is a positive integer, the sum of the first n -t- r coefficients is + 3)+4r(r i- 1)}.

59. The graph of y~--- , --- -. has a turning point at P(2, -1). Find (x l)(x 4) the values of a and 6, and show that y is a maximum at P. Sketch the curve. 60. Prove that, if x ^ 100, then x^ -%((x+ 1 ft i- (x - 1 ft] -= *x~ (nearly), the approximation being correct to at least eight decimal places. 61. If f'(x) is positive, show that f(x) is increasing. Prove that 2x + x cos x - 3 sin x >0 if 0< x< 4 .

62, Prove that

T75 + 5.

11.31.3.5 and that ^ -f - - 4- - ^ : 4- . . . to infinity = 1 . 44.64.6.8

MISCELLANEOUS (A) 549 63. Solve the equations : y* + z*-x(y + z)^a,-} )~ b > .

[Show that 6 + c - a = 2 (x 2 - yz), etc.] 64. By expanding (1 -2x-3x 2 )~ 1 in two different ways, prove that 2 + (n - 1) . 3 . 2-' +

65. If ax + by + cz = 1 and a, 6, c are positive, show that the values of x, y, z, for which - H h - is stationary, are given by x y z Show that this is a true maximum or minimum, if xyz>Q> 66. Show that the equation tan x~ I + x has an infinite number of real roots, and find graphically the approximate value of the smallest positive root. 67. Assuming that a;{log(l + a;)}~ 1 can be expanded in ascending powers of a:, f~\d the first four terms of the expansion. By help of this result prove that a capital sum accumulating at compound interest at r per cent, per annum will be increased ten-fold after /23026 \ ( h 1-15 j years approximately.

68. If x y~y 2 -\-2y 3 -y* 9 and it is assumed that y can be expanded in ascending powers of x 9 show that 69. Draw graphs of _2(z 2 ~6z) 2 2(x*-$x) 70. By expanding in two ways, prove that

71. Show that (a --- #)(4-3o: 2 ), where a is a positive constant, has one maximum value and one minimum value. Find these values and show that the / 1\ 3 difference between them is J ( a + - } . What is the least value of this differe nce V a/ for various values of a ? 72. Prove that x* - ISx 2 + 4dx + 9 = has four real roots, if d* ^ 1728.

550 MISCELLANEOUS (A) 73. If r*x^aR + bW, where and the quantities denoted by all the letters are real and positive, prove that jc, ^ 2^ab. 74. Prove that, if is small, one root of x* + x 2 -(I+2c*)x-I^O is approximately 1 f ie 2 , and find corresponding approximations to the other roots. 75. If , /?, y are the roots of x 3 - ax 2 4- bx c = 0, the area of the triangle, of which the lengths of the sides are a, /?, y, is

If the triangle is right- angled, show that a(4ab -a 3 - 8c)(a 2 - 26) ^ 76. If n is a positive integer, prove that 2 n n n (i\ __ < _<3 n - 1 when

containing p and q lines respectively. while if equal parts are allowed. and has the value 6 when x = 0. 79.71 71 (ii) P. no two lines being coincident or parallel. 80. Find the condition that ax + b/x can take any real value for real values of x.2bx + c)/(a 7 x 2 -f 2b f x + c') which has tm i ni ng values 3 and 4 when x = 2 and 2 respectively. rc n <(^. there are 75 ways. show that log y lies between 2 f . Prove that all the roots of the equation are real if 8< k< 11... Express f .m /a + x\ . Show that the number 30 may be divided into three unequal parts in 61 ways.2 2 . 6.(x . where a. 85. I _ 1 CITlH 2 AT/ 3 ~~3 for positive values of x. 6 are unequal positive quantities.a] (x ~ b)/ (x -c)(x-d) in terms of y where y~(x~ d)f (x-c). Find a function (ax 2 H. divide the plane into pq + 2p + 2q-l parts. Show that two pencils of straight lines in one plane.when n /?/ 1 \ y 2 -~ 77. 82. MISCELLANEOUS (A) 551 86. m are positive and m>l. . not scoring more than 4 off any ball ? 84. _. 83. If y is positive.j and 78. If a. In how many ways can a batsman make 14 runs in 6 balls.3 3 . show that 81.. Find the minimum values of a m + a . Prove that the equation has no real roots if r 4 <s 3 .

<u Z 94. 88. Show that an approximate solution of x log x 4. as x. 3 2/3 90. y tend to zero along the curves x ~y (i) y=x-x 2 . 3 ...-a. where e is small. where a>b>c.(a-c)..and hence.x . Prove that 7 2n . Show that a sum of n pence can be made up of pennies. . If (x l -x then 4 I #. If a. show that f can take all real values if (c-a)(c~b)(d~a)(d-b) is negative. Find the limits of -. or otherwise. 87.to oo =2^6. c are real.1 = e.48n .1 is divisible by 2304. 92. 89. Find the maximum and minimum values of and draw a rough graph of the curve. 6. prove that square root of a 2 + 6 2 + c 2 -6c-ca-o6 is v/3 greater than |(2a-6~c) or than -. halfpennies and farthings in (n + I) 2 different ways.~ . (ii) f/^a. Prove that if a r = 1 -f r~ . Li l then 1 + o a i + b"" 2 a 2+--. -f y. j3 I y3 91._ y l I -c). (iii) y-^x-x*. & 93.+ -^ -f .

If x>2. Prove that. as far as the fifth power of x.. when n is large.. Expand {2a. and show that for small values of x it leads to a good approximation for e x . prove that.. 97. Show that one root of the equation is approximately 1-0103. is p + q -r + 1.. prove that 552 MISCELLANEOUS (A) 98. 99 . If a. 101. prove that ab ^ a p !p + b < //qt 100. and p and q are positive rationals such that 1 /p + 1 /q I .(n + r-1) n+ ^ ^+ _ j +. if n is a positive integer. 102. the coefficient of x n in (l+x + x . 96.. w 2 17.95. and find the other roots correct to two places of decim als. prove that n n(n + l) n n(n-l) ^^ " t "'"" *~*~ + " and deduce that n(n-l) (r-l) n(n-l)(n-2) (r-l)(r-2) _n(n + l) . + V(9 + 3z 2 )}/(3 _#) i n ascending powers of x. Deduce that ei = 1-2840 . 6 are positive._ _ . p things being of one sort and q of another. Find approximations to the roots of where a is small. 103. 1 H ) Show that e is given approximately by with an error of about 0-46 per cent.. If 3w n -7w n _ 1 + 5w n _ 2 ~u n _ 3 and w ~l> ^i 8. By means of the expansions of e x and log (1 +x). If r>p>q. show that the number of combinations of p+q things r at a time..

MISCELLANEOUS (A) 553 111.. 6. show that an approximate root of the equation is given by x a{l+%p log e a + ^n 2 log e a ( I + -J. c 2 are all greater than 1 or all less than 1. fc 2 . 107.. If x + y + z~xyz and x 2 =yz. c are real and then a 2 . the value of log e (log e 1-01) correct to five plac es of decimals. Solve 109.-. without using tables. Show that the result of eliminating x and y from the three equations 1. Show that the coefficient of a 371 * 1 in the expansion of -.z ) n is 1 104. Find. then y and z may have any real values and x 2 ^ 3. . 113. Prove that log e {log.|a.. ( 1 + *)*} = . If a. 108.\x + ^-a: 2 . If # 3 ~^ a 3 where ^ is small. having given log e 10 2-302585.~ in a series of ascending powers of # is (x + 4)(x +&) 112.11 1 11 x-a y-a a 9 x-b y-b b 9 and is 110. Find to five places of decimals the real root of x 3 + x + 1 = 0. The number of ways of selecting r pairs from n different things is 105..log e a)}. Having given that prove that 106. 3 .

. Find the sum of the terms after the nth in the expansion of (1 + #)/(! ~x) 2 in ascending powers of x. . 120. .. and the G. Show that the 2n-th convergent of . by suitable choice of x. Find the coefficient of z n in the expansion in ascending powers of z of l+xz _ . Prove that ( 1 + x^ = 1 -f x + x* + f* 3 + V* 4 + 115. An examination consists of three papers. (ii) ? 2 n^ 2 n + ( .xz z 1 f/l + V4a?+l\) n + a /l-Vto-fl\ n+2 l H 2 j/ -(2. which is greater than 2n. Prove that the ratio of this sum to the sum of the corresponding terms in the expansion of (l-x)" 1 can be made equal to any given number A. to each of which m marks are assigned as a maximum. .M. 3434 118.M. of their squares by any of the numbers. 1 show that . ~f a r x r + ..l) n and ? 2 n-i/4 = (<7 2n -#an-i)/5 119. with a like result for the q'a. The difference between the A. If the infinite product ' 2/ V 2V V 2V is expressed in the form a -f c^x + a 2 x 2 -f . [Prove that (i) p 2 n 1 Ijp2n-z + 12# 2n _ 4 . 117.l)a r+ i = 4(a r + a r _j). Show that there are f (m + l)(w + 2) ways in which this may occur. then (2 r +* . Explain clearly why the restriction upon A is necessary. and the A. One candidate obtains a total of 2m marks on the three papers. / . 12 n + 4( ..l) n }. .M.M. of a set of nearly equal numbers is to a first approximation equal to the quotient of half the difference between the square of their A. . 116.----.i 114.z . differs from the 43+4-3+ value of the continued fraction by 13/{9 .

. 554 MISCELLANEOUS (A) 122. Prove that the number of ways in which n prizes may be distributed among q people so that everybody may have one at least is 123. n axe b. if n > 1 .-2) (n-2)(n-3)(7t-4)^ 3 I + wo. H j^r o: 2 H r. . 121. b are positive integers. is divisible by (x . 126. where a^b^c.o.6. Prove that where m and ?i are positive integers.i + 3CJU + . Prove that when x is sufficiently small 1 1 1 11 Jog (1 +#)~~ # 2 12* 24* Also. Show that th e sum will be repaid in A years.&2 . Show that the number of combinations taken n together of 3n letters of which n are a. and the rest unlike is ( % 4. v .. and that the q uotient is where p r = m(m + n -r. where A is the integer equal to or next greater t han 125.c ~ a.26'Jt.ft)y 4. and partly to repay capital. prove that 2.-2/ 2 = 3. x + (c + a)y + caz ..1) (r^m). . If x + (b+c)y + bcz = b~c. the probability that ^(a 2 f 6 2 ) is a posi tive integer is gV 127. and that this sum is equal to 2 w ~ 1 (n + 2). long are placed in a passage..a . and repays b at the end of each year.. x + (^ 4. A man owes a sum of a. and 10 ft. Ladders 15 ft. # | (n + 2) \n " " ' L__ 'L 124. > r ~(r-f-l)n (r<m). . partly to pay the interest at 100?* per cent. each with its top . 128. If a.-x M . + nC? + n + l. .I) 2 .x 3 + .

( . if they gave one to the monkey. the nuts could be divided into 7 equal lots..a . giving i/ = 0-5761287. an infinite sequence of positive integers. ^ = ^(100. . ) [If the width of the passage is x ft. 6^7-880644. (Sunday Times. and the tops of the ladders are a ft... Prove that the least value of N is 7 8 .1) ~ . . (iv) If a is a prime to b and a<6.6 is unaltered by the set of operation s corresponding to what each robber did. In the night a robbor woke and decided to take his share. A party consisting of 7 robbers ami a monke}' had stolon a number N of nuts.ere left.7.c n (mod6)... Hence the values of N are given by 2. The other robbers woke in succession. a b 5 Hence if 6 5(t/-|-l).+ ! = ?> and a 2 -& 2 -125.6 of i\ satisfies the numerical conditions . from the ground. q n are the quotients and r l9 r 2 r s> r n the remainders when ac L9 r L c z . the remaining nuts could be divided into 7 equal lots. Ci<. r n _^r n respectively. nc 3 . .. fur 6 . from the ground..155929. If the point where the ladders cross is 5 ft.. and they found that. show that -. then . which were to be divided equally among the robbers. If q l9 g 2 ... and y 4 4-2?/ 3 + 5i/ 2 --2?/-1^0. c 3 . show that 0<z/<l.6 = 5764795.~ .I .6 2 )^ 6... He found that.. are divided b y 6. .. </ 3 . and b ft..1. b C L c^ qr^ qca. In the morning they divided what v. . and each gave one to the monkey and took one-seventh of the remaining nuts.against a wall and its foot against the opposite wall. He gave one to the monkey and took his share..] MISCELLANEOUS EXERCISES (B) 1.6. found that the same thing was possible. then . Thus ..^ 1 r where E n = -.c b (iiiJV-ar^ . Let a/b be a proper fraction and c l9 r 2 ..7 ~-7 = . find the width of the passage. [The value . if he gave one to the monkey..7 . .

(iv). half the remainder and half an apple to a second. then \ \y 8. Prove that. Show that 8 ~ 3 2 " 3" 2 " .a ^? 1 + _?? . Express the following fractions in the form : ft + J'. to show that e is irrational.. Sum the series n 2 + 2(w-l) 2 -f3(rc-2) 2 -f. a man sells half the contents of the cask and half an apple to one customer. 3 3.. 5. (iii) J.| 6 C L c^'o c. and so on.5. Use Ex. always selling half of what are left and half an apple. (2n + l) ' (i) J . 3.to infinity . where n is a positive integer.C. 4. z are positive and x + y + z = l. if x> y.... From a cask containing x apples. for some value of n : otherwise the series is convergent. In the last two cases.. 556 MISCELLANEOUS (B) 7.j. 12 n ' 6 where S n is the sum of the squares of the first n positive integers.A.7 3. 2N B. find the value of q n . (h . 6.8 4--? 3 + 4-. c n is divisible by 6. 2. In no . (ii) i ... Prove that $I + TS+-"+ .5 3.n + .JL n 4.5.c 2 c 3 where the series terminates if c x c 2 .

Show that the planes ny-mz = \. prove that (2n . (2n-5) n _ 5 . _ .. If A: r -ri . a+& m + c a P + 6 p + cP a-j-^-t 12. n w _ 3 . . /?. and m m'..u . h-nxp.n < 10. . Z'. II /? = --. + where r is a positive integer greater than unity.a |2r-2 11. cy + e/ Show also that a.y~"n 3 where ad COL -f d Cp -f. ZA + mfji + nv = .. r are positive quantities such that m 13. n n _ B . ^. where ^).d and a.. 6. ex + d ex -a COL f a Ca .. If L TH.= a ---. and if show that / = Z r . and only if.. and if /)9 is not rational.> ~ . _____ -.. a apples are left. If x r = x(x-l)(x-2) . (2n-3) w _3 ... y are the roots of -dx + b aa + 6 -da + 6 x . 9. m TA.>n n ... .(x-r+l).. c. ay + 6 prove that =. . Show that ---. 14. and 10 ou .case is an apple divided. d are connected by the equation a 2 + .. 7i n _ 7 each product being continued as long as the suffixes are greater than zero. w'. s show that k r =kjc r ^ + . Show that z = 2 n (af 1)-1.!)_! . After n sellings.. mx~ly~v y have a common line if. I" and 7W X/ are integers.

MISCELLANEOUS (B) 17. 3.5 n where A r ~( . 15. 20.26 2 6 3 . prove that a (y 4. 19. show that ax b ~ c 4.33 .I) 7 ""' 1 sin 2 cos n ~ 4 . agCj.54x -36-0 is 3 . ^(n ~ 1) or r& -^ .33" .2 + 6 ) + ?/ 2 z 2 + fy/z (y + z) = 0. TIT X" 4. . One root of the equation x 6 . a&x 4. in general.z*x 2 4. (2Z /X -f 5m") a -f (Z /x -f 1) jS can be equal. b. 18. (2Z X -f 5m' -f 1 )a -f Z^. 4.bx c ~ a + cx a ~ b ^a + b + c.x) + c 2 = O t are true for an infinite series of values of x. c 2 -.y8(a is equal to zero. Prove that ~(L-x rt . If a (z 4..9x 5 + 18x* f 9o.Also show that no two of the numbers (2Z + 5m)a + /^. Prove that 2 a4-4-y4-5 a + j3 + y + 8 2(a4-]8)(y4-8) a(y 4.a x c 3 = 26 1 6 3 .8) 4.bzx (z 4. 16. If a.tan. y. 2. find all the roots. a 1 c 2 + a 2 c 1 ~26 1 6 2 .. . z. 3 + 27 x 2 . c are unequal. where y and z are unequal.x) = 0.6 a (z 4. If a 1 yz + b l (y + z) + c 1 ^V. prove that.x + 6) 4.and r = 1.

Find the limiting values of i . 21. 6. (cos x) x3 . if I l9 m l9 n^ . ! 2 -f m^ 4. 23. (cos x) xz .a f 1 4. Show that.x) n 4. 24. n 3 are real quan tities satisfying the six relations. m 3 . q. 1 3 . m 2 . c in terms of p.. Determine the condition that (x n shall be divisible by 1). r.j and are identical if \ 557 a b c a b c p q r ' p z q 2 r 2 ' and find the values of a. n 2 . 22. Show that the first four terms in the expansions of ( 1 4. as x tends to zero through positive or negative values.1 i (cos x) x .according as n is odd or even.^! 2 = / 2 2 -f w a a -h n 2 2 = / 3 -f n 2 r? 3 = l then . / 2 .

If u (b . then 1 ^tt-xsina-tfsin (oc~6 r ) r " =0 _ x' m -2x n cosa + l~~nsina r " =0 x 2 ~-2x cos O r + l 29. Show that the cubic 4z 3 27a 2 (z-f a) has two equal roots.w^a 0. and trace its graph. prove that (i) If u is divisible by 27a 2 .a) n + (a. 26.1) if m<0 or m>16. 27. then n is of the form 3k 1. 30. and r ~(a-f 2r7r)/n.Z!bc. If n and n + 2 are both prime numbers.^ 558 MISCELLANEOUS (B) 25. 31. Prove that.Hbc) 2 . Prove that.c) n + (c.6) n . where k is an integer. show that where A = g 2 + 4th 3 . if n is a positive integer. the equation . show that 28. then u is of the form 3& + 1. Investigate the maximum and minimum values of (x+ l) 5 /(x 5 + 1). where n is a positive integer. if e is small. 4.and 4. Prove that the equation (x+ I)*~m(x 5 + 1) has three real roots if 0<m<16 and only one real root ( .n 3 ? 3 l l m l 4. (ii) If u is divisible by (Ha 2 . If z 3 -j-3/fcz + <7~ [/x(z-hi/) 3 -v(z-f /x) 3 ] for all values of z. prove that (n -2) in -1-2 is divisible by n (n + 2) . If all the roots of the equation are rational and negative.

p. ^4 5 = a 3 2 -f 2^(15 4. .2ia 4 -f 2a 2 a 3 .2a 2 a 4 . show that d*~D. y.Z show that JJL lies between the greatest and least of the numbers a/6. and find approximations to the other two roots. 34. c/d. c. z all tend to a.has a root nearly equal to 1 -f 4e 2 . I are positive numbers arranged in descending order. .. w+1 . Prove that. n. Find the seventh roots of unity and show their positions in the z plane where z x + iy. 6/c. (-!) |^. 36.. 33. find the equation whose roots are z 2 and MISCELLANEOUS (B) 559 35. If z is any one of the imaginary roots. 32. k. If d denotes the determinant a x a 2 a 3 a 4 a a 6 ! # 2 a a a a 4 a 5 a x a a a and D denotes the determinant obtained by changing a r into A r where A ! = a x 2 + 2a 2 a 6 + 2a 3 a 4 . b. k/l. n ) n __ a. ^ 3 = a 2 2 -f 2a 1 a 3 At a 5 2 4. ^4 2 = a 4 2 + 2a x a 2 + 2a 3 a 5 .(y-^)+yP(2-a^) + 2: p (a^-y) 37. prove that ^(yn^g^ + ytn^n^^^gm^^yn) wre ( m . and /xa . . g are positive integers and if x. in the expansion of . If a. Discuss the convergence of the series whose nth terms are (J*)!*-.. If m..

so that as x -> a ^ the ratio of each number to the preceding may tend to infinity : ar 2 . Prove that the coefficient of x m in the expansion of . a. Arrange the following numbers in order. 44. the radius of the base is approximately 2-381 feet. If the volume of a right circular cone is 20 cubic feet. (log a. Prove that --.*)( 1 .*) 43.!<#<!. 2 X . Deduce or otherwise prove that vanishes with or has the same sign as m(m l)x 2 . is given by the least integer contai ning 39. after a certain stage. Find to 6 places of decimals the root of which lies between 5 and 6. in powers of * is (1 . e x .where . Show that the number of distinct sets of three positive integers (none zero) whose sum is the odd integer 2n -f 1.lo8 r . and that in the latter case the last positive term exceeds numerically the sum of the infinite series of negative terms. 1 42.) 2 . when the cone is such that the curved surface is a minimum. Prove that I-x z + ix*-(l+x)e~ x is positive for positive values of x. prove that. 38. and increases as x increases. 560 MISCELLANEOUS (B) 45.*) (1 . 41. Show that the number of distinct sets of three positive integers (none zero) whose sum is an even integer 2n is the integral part of \ri*. . the terms are either all positive or. x x . 40.v 7 ./ n -r v ' . become and remain negative.

z 2 + 2#-f # 2 ~6 2 .a n = A n . Prove that if t/ 2 -f yz + z 2 a 2 . / 4. y. d. 172 50..1 is divisible by 73. Prove that if m is prime and p<m. a*x + 6 3 y -f c 3 ^ + d 3 w = A 3 . ~ are equal. A are all real and unequal. + dw~ A. Solve the equations : x + y + z + w = l. 6. if a b c a b c al bm en where a. . Prove that. Prove that. or otherwise prove. 51. . c are not zero and no two of . z. Prove that 7^=7-.^2 / = 0. at least two and not more than three of x. x z + xy + y z c 2 . then a&c-0. c. y2 -f zx 4. that ln + 1 lrc + 2 |n + 3 ln-f-4 " " 1 [2 [n-2 " |2 |3 n-3" [3 [4 |w^4~~'" 46. w are positive.c 2 2 + ^ 2 t^ = A 2 . if a. then I m n x j+ y + ^o. . I m n 47. If a 1 a 2 ^3 . 6. prove that the sum of the products r together of the a's is greater than 48.|r)2[n-r and deduce.. 49.

and if a. y 2 + 8 2 -fcy8.log n tends to a limit v as n -* o o . and that 1 +3a t 4. 8 f are all different.. Also. . j3 2 + y 2 = /c)5y.1 = 52.6a 2 + . the number of combination s of these r together is equal to the number of combinations 3n ~ r together.n)(2n . If a 2 -fj8 2 = *a0. -f (4w + l)a. )3. the number of combinations is |(n + l)(n + 2) -f (r . MISCELLANEOUS (B) 561 55.r) and the maximum number of combinations is |{3(n-f l) 2 -f 1} or -J(n-f I) 2 .. and prove that KQ v i r - . then 53. 8 2 + 2 ^AcSe. . and if l/-vf* /"I /yi f%\ U )* ' H I'"* ' II I'"' lo g V 1 "~ * " x / a l j -2 M 2-< ~ 5' f 3-* > obtain an expression for t^ n . .+ ~ -f . If of 3n letters there are n a's. Assuming that 1 4. 2 ^ ^ ^ 1 (n n . n = (2n + 1)(2 +f ) n . 56. if n>r>2n + 1. Expand log(l -a: -a: 2 ) as far as the term containing # 4 . \ (li) lim n ( v j' h ~ 2 ^^~3 " + '"" f n~ g -/ 7 * 57.\p-l \m-p + ( -l) p .. y. Prove that 1 1 ttfo" 1 ) n(n-l)(n-2)(n-3) _ 2n-f2r 54. If prove that a r a 2n _ r . prove that (i) lim ^ .1 n ~ 2 1 . n 6's and n c's. according as n is even or odd. H.-..

m. Find hm T-"! 1 . c.2 ~ 0.3x .c n x n . If a.. (ii) when I. where a. (i) when I. must have one of three values. + y + 2 = f 4-^-f # + <y??4-2: = 0. Determine which of these values give maxima and which minima. If * 2 + y a +2 2 = f a + i7 2 -f 2 = 1. 63. the fract ion satisfies this equation... 6. 61. m. find the values of x for whic h (x ~a) l (x ~ b) m (x ..(2x ~ x 2 ) (j* _ ^03 59. and that if x satisfies the equation 2# 3 + 3# 2 . c are positive and . Solve the equations : y + z = a(l -yz). 6. ^ o i . z + x~b(I -zx). where n is a positive integer. Sum the series l+7T + T + T^H~r7 + 7^ + 77.in which a is a ~ a a x equal to . prove t hat . If n points are taken in space so that not more than 3 of them lie in any plane and not more than 2 in any straight line. n are all odd.1 and is repeated any number of times.c) n has maxima or minima. 2 3 8 13 30 55 66. . P 2* 3 8 ' (n + l) a w+lV 2 3 n+1/' 64. show that 3a: 2 ^=(i7-{) 2 and 2: 2 4-f 2 ^!. ra. 62._>o c^-rf* 60.. If (1 + x) n ~Co -\ -cx+ . and a.. n are all even.. Find lim --.to 1 1 |2 3 4 |5 o 562 MISCELLANEOUS (B) . n are integers. 65.1 polygons of n sides. c are real and I.1) straight lines and \ \n . + ".58. show that the points define \n(n . x + y + z -xyz~c(\ ~xy~ yz-zx). Prove that the continued fraction a -----.

. distinguishing cases of absolute and condi tional convergence. given that J log J 2 log J 3 log log 7 = 0-845098 12234 . 5 of the second. 4 of the third. k.n (l w ) 2 for all values of z.4 2 /() -f .. Show that the number of different selections of 6 balls which can be made from the bag is 259. Discuss the reality and equality of the roots of x* . Prove that two lines can be drawn to cut all the four given lines.1) prove that f(a + xw) =f(a) + xAf(a) + -^r^. and that thes e lines cut the line (4) in two points determined by x = y = z k 9 where (k-a)(k~b)(k-c)^(k-a')(k-h')(k-c'). real or complex.J/(o). 71. y = c'. (3) x = c. (2) 3 = 6. Use this to find log 7-063. 68. (4) x=y=z. . Find the equations of the line drawn from the point (k. If Af(a)=f(a + w)-f(a) 9 A 2 f(a) = AJ(a + w) . A bag contains 21 balls of G different colours. I2n ! Z . x(x -. z = a'.2A(3* 2 . 69. etc. and so on. k) to cut the lines (1) and (2). 70.67. The equations of four lines are (1) y = a.4x + 3) + 3A 2 = for different real positive values of A. * = &'. Discuss the convergence of the series E. there being 6 of the first colour.

l) r c . according to the value of n. 9 are drawn at random. Show that the probability that 4 balls of each colour will . 7) f 74. State the critical values of n. A series is such that the sum of the rth term and the (r + l)th is always r 4 .are fractions such that qr-ps \. Prove that (i) the rth term is %r(r . the balls being replaced..'.l)(r 2 . oO 2 73.log 7 -2 = 0-857332 -334 11900 16 log 7-4 = 0-869232 -318 11582 log 7-6 = 0-880814 72. MISCELLANEOUS (B) 563 75. Prove that. 76.+ ^{l-(~ l) r }. Determine ranges of x for which the function (log x)jx (i) increases and (ii) decreases as x increases. wherein n is a given positive number and only positive values of x are considered : (i) The equation n x x has two roots.1) . one root or none according to the value of n. then the denomi% s p r nator of any fraction whose value lies between .( .is not less than q -f s.and . q s Prove that there are two and only two fractions with denominators less than 19 which lie between jf and -J. If n is a positive integer.. r(r 2 1 W^r 2 7^ r (ii) the sum of r terms is -> . Hence prove the following theorems.).r . (ii) The inequality n x >x n is a consequence of either x>n or x<n. . if .. From a bag containing 9 red and 9 blue balls.

A quadratic function of x takes the values y l9 y zt y%> corresponding to three equidistant values of #..(a r +l)-l.. show that.. Establish this result when <^ is small and show that the error is approximately < r> /180. _ . For the continued fraction . 2 . Find the equation determining the values of x for which sin mx/sin. a r things o f an rth kind. 78..a x ) (n . measured in. 77. Hence. sin 2 war/sin 2 x never exceeds m 2 . a n are permuted cyclically. 82. Prove that if 2/i + . Given a l things of one kind. x is stationary. x l a^. show that the number of different groups which can be formed from one or more of the above things is (a 1 + l)(a t + l).Oj + 1) (w .be included is a little less than -J..V 3 >2?/ 2 . and deduce that 2/22/3 2/n_i2/ n 81... etc. 80. if Zn^Vn-i + ^n-z* and if x Q ~I.. then is nearly equal to \TT for all positive values of a and h. prove that (I ^ -f- altered if a l9 a 29 .[(n -M) (n -f 2) (n + 3) . verify by induction that x n+l y n -x n y n+l = ( ~ l) n+1 . a 2 things of another kind.(n . the number of groups having n members is |. the minimum val ue of the function is . is 3 sin (f> / (2 -f cos < ^) provided that <^< OTT.. Hence express approximately the acute angles of a right-angled triangle in terms of the sides and deduce that. 79. if m is an integer. y Q Q. An approximate value of <^. or otherwise.a x + 2)].-----. if c 2 ~a' 2 i fc 2 . When n^2. radians. and show that if r~4 and a l <n<a 2 <a z <a tl . 2/i 1.

3. If -f cz 2 2 = 2 = d. 5. written at random in any order. The digits of a number are 1. Show that (n + l) n ~ l (n 4-2) n >3 n ( ]n ) 2 564 MISCELLANEOUS (B) 84.83. 2. Show that the odds are 115 to 11 against the number being divisible by 11. show that \ w-f 1 86. 9. Prove that sin a sin ^ sin y cos a cos p cos y sin 2a sin 2/J sin 2y sin 3 a sur sin a sin t cos a cos . &z + by 2 prove that 88. 8. 7. If w.x) 3 is 87. n are positive. . The coefficient of x 2n in the expansion of ( 1 -f x*) n / ( 1 . 6. 4. 85.

sin j y sin y COS y 2/2 Z 2 2/3 2 3 ./? + y)./ nn . the number of combinations is Show that the number of combinations in groups of n which can be formed from 3n letters.2) 1 . Prove that if n is a positive integer.4 sin J ()3 . -.+ and show that this sum is 2 n ~ 1 (w + 2). 90. 2 sin (a + j3). Investigate the convergence of the stories whose nth term is 92.a) sin (a . Prove that ^__ n(w-fl) 93.sin (p -y) sin (y . by W 6' r .y) sui ^(y ~ a) sin v(a ~ j8) . show that if r of the letters^are alike. all the letters being unlike. + Vn-s + . 89. x + TT2i" 1T2 . n n(n-l) t(w -!)(*_. 3 91. is <? + VV ./?) sin (a -f.1. of which n are a and n are b and the rest unlike.. If . Denoting the number of combinations of n letters taken r together. the r est being unlike.

each player in turn acts as c banker. . ^ 2 + 5^-f6 and MISCELLANEOUS (B) 565 p 2 2 3 2 94. 96. They are drawn in pairs. ft) ' 99.B)_(A 9 D)_(B. then where A.D)_(C. Also show that if n -7 the chance that a given competitor either wins or is beaten by the actual winner is tV 98. In the game of Craps. z. show that (B. 2 n players of equal skill enter for a tournament. If he throws . v are integers and jtx is prime to v. a) ~ (a. . _ 2n(2n and t 2 + at~b both have rational factors. show that for all real values of x.C)_(C 9 A)_(A. show that (ii) x*= -12(t/~l)+(y-l) 2 approx. If thore are four relations. 2 . d) ~ (b. prove that 95. A fy + B& + Cfy + Djdj = for i = 1. c) (c. the coefficients being real. d) ~ (c. If sin x~xy 2 where x 2 and y ."* V . For the fraction .D) (a~d) ""(ft. Find the chance that two given competitors will play against each other in the course of the tournament. y.I are both small. Discuss the reality of the roots of 16* 4 + 24* a +16^+9 = 0. and the winners of each round are drawn again for the next. [For instance. If ax* + 2hx + b>0 for all real values of x. 97. and solve the equation when k = ^2. . 2 . If he throws 7 or 11 he wins. 100. j= 1. /Lt.' The banker throws with two dice numbered 1 to 6. for all real values of k.

2 oos a oos ft oos y. 566 MISCELLANEOUS (B) 103. . then either *++ !=. in which case 1 U n . aW + b z y 3 + c 2 z*^Q. in the second he loses.a n Hence obtain the general solution of the first equation. If a 2 x* + b*y* + cW = 0. 372. If u n u n _i + au n + bu n _! + c = 0.<) or z = a-c. /? are the roots of unless a = jS. 101 . If *-. Show that the odds against the bank are 251 to 244. 3 or 12. o + x d + y x y o a 104. If rv 1 cosx COS X 1 COS a COS ft COS a 1 COS y cos ft cos y 1 prove that sin 2 y 8in 2 z = cos 2 a -f oos 2 /3 . If he throws any other number. and show that the results so obtained agree with those in Exercise XXXVIII. 15. show that where a. he throws again and continues to throw until either the number he threw first or 7 turns up. In the fir st case he wins.**-?* =*. he loses.2. p. 102.

28. respectively.3* 3 .-c 1 . -21. a. z 4 . and y hrs.496.6* 2 . 4 miles apart. 30. starting from Y at any time between one and four o'clock. and a e z 3 + & 6 */ 3 4.] 107. Show that the odds in favour of the men meeting on the way are as 5 : 4.cV = a 4 ^ 2 -f.. 212. 6. in a game of whist.&x . all times of starting being equally likely. 23. 19. 8.13. A. (ii) T - (p-1) (p-1) EXERCISE II 22. 3z 2 + a. G y3 J -rC'JJ. (i)**'"**. 360. 219. 220. [Suppose that A and B start at x hrs. A man B.-3.-3. EXERCISE III 31.and l. Also the chance that in 25 deals a certain player has a hand void of a suit at least twice is about 0-358.4. If u~ax* + 4bx* + 6cx z + 4:dx + e and t is any root of show that the roots of w = are given by ANSWEES EXERCISE I PAGE 10. 20. Taking axes Ox. 2. 7. 105. The total number of cases is represented by the area of the square.10 . 0-006 . 11. (ii) (x + I) 3 ~(x+ l) + 80 . 3. 28 2 -27 2 . fc^-i^ + fcc'-fta + A. ia-V*' + tfa-tt . 1258. 7.28. J and Y are two places on a road.. 27. 4836. . (iii) 2-276 . 5* 3 -20z 2 + 80a. walks to X at 4 miles an hour. -3z~7. Show that. . (ii) 2-594 .-314. the chance that a hand is void of a suit is about -$. 120. ~^5. (?) 5-394. 1. 3z 3 + o. and the number of favourable cases by the part of the square between the lines *-y= 1.&y + c 4 z 2 . 213. set off OL~3 units along Ox and draw a square on OL. y 2 prove that a 4 z 3 -f 6 4 ?/ 3 -f c 4 z 3 = 0. 10. 6. Oy at right angles. [Chance that a hand is void of a suit 4 .2 =--& 2 = .] 106. walks to Y at 4 miles an hour. 0-002. 22. 5a + l. (i)(x + 5) 3 -13(z + 5) 2 -f56(*4-5). 0-002 . past one. 8 2 -3 2 . A man A y starting from X at any time between one and four o'clock. .

6. a( 21. 22. 15. 4= -*. 3.-2). 7^1 10. ' 9. 2. 3z~4. 14. x* + x-3. 26. 120. 2*4. (1)2/3 + 5^2. 10.7/8. 12. (-l)Ca f ifn=2m. 5.7y + 4) (x . ( x + 3)*(x-3). (* . p* . 32. 3 + 4x 2 . x 2 . 25. 0.^. . 7=3.lla. A = l. ~^.1. y = J(a. (n)|tt + l. 60. 17. Z^l-fz 2 . JC:=:r 3 + l. 6=0. (ii)jf-75y + 7. -( EXERCISE IV . 2z 3 -f 62& 2 y.4)/8. (i) 33.-4y + 2)(a? + 2y-3). a 2 30. (s-1) 2 . A = . 13. 18. (x-3y)(2x + y)(x* + xy + y z ). -n + 7n(n-hl)-67i(7i + l)(n4-2)4 11. 20.Ixy + 4y 2 . 3. 29. 4. 6=-3. if n=2m + l. c = l. 1. 34. -2x* + 3Zx 2 y-12xyz. 31. !>=-*.y . 6. (iii) i/ 3 .s 2 = 0.15?/ 2 -f 257. 3* 2 .-3. 0=1.9. . 20. A=-l. 568 HIGHER ALGEBRA EXERCISE V PAGE 43. a=5. A=-2 ft = l.6x . 18. 16. 16. c = l. 7z 2 .(2a? + y-2)(aj-y + l). EXERCISE VI 50. 27. 7. A. (3a. 8.-2. (-l)(n + l)05. (s-l) a . 23. 24. a = 8.2.=-!. 28.3q z + 3r 2 . 23. 4. a = l. (x-l)*(x + 5). a=3. 3a. JC = i(x + 2).3y 2 + 3z* -w 2 = 0. 21.9. 6-2.6. (i) #i"("+i) . 15.(y-l)(y-2). (i) 2 2 ".I2xyz.

(l-a&c)(l-a 2 ~& 2 -c 2 -f2a&c). and draw AL. hence L OM is a right angle . (iii) (x 2 -2/ 2 . 27. (18 + 0/25. 34. (i)1625. 5. EXERCISE VIII 67. then L. 5. -2. (ii)(. (i) (x 2 + 1/ 2 ) 2 cos 7i tan"" 1 -. 30. -. 6. -4-3i. KX + 4r = (l+* + iy)/(l-*-iy). N represent 1+3. ~(6-c)(c-a)(a-6)(Z l a 2 + 3Z l 6c). 3. O^lLl. If |z| = l. 32. (16)(25)(34) . 33. parallel to BOA. 29. (16)(65)(53)(24). q 51. 13. 26. 7.iy)(x* + y*) . 7T/6. 1. 2. l~2t.18. Geometrically. and N lies on Oy. EXERCISE VII 60.) 2 + 2/ 2 ]. q) and is per- . 2. if x 2 + t/ 2 = l. -(6-fc-2a)(c + a-26)(a + 6-2c).) + 1 sin n( tan~" x L_ \ *> / \ i* (ii) (a:. in L and M : construct OMN similar to OL1. 1-zand (l+z)/(l-z). p^a+f-g-fi. 2. 1. BM 9 parallel to Ox to meet a line through 1. ~7r/3. 2-i. If a=p + iq. J5O13f arerhombi. 31. with L* ION.2 + 2)/8. 3(6H-c-f-2a)(c + a + 26)(a + 6 + 2c). M. EXERCISE IX 78. 24. O D 4.1 . (i) (5 -0/13. i(ltV3). px + qy=p 2 + q 2 . so that BO OA. -4(6-c)(c-a)(a-6)(a + 6 + c). 35. 36. (iii) i( 1 + * cot5 F / y\ ( y 61. 7. If A represents z. 25. -7T/6. Z^O. ^ . (ii) 13/125._ 6. 3. 2. then X = (l -a: 2 -i/ a )/[(l and y-2?//[(l-a. which passes through (p. take any line ^4OJ5. (12)(24)(36)(65) . LOM described in opposite directions.

(ii):r 4 -23x 2 4-59x-52=0. 1(1 \/5).1/2. 6.14z 2 + lla. (i) 4x. 3.3 4 2x*-3x. 4. 17. 8. 1/2.<>. (ii) x 3 -4 3.15 . i. 2/3. 2. 2.0. EXERCISE XII 99. EXERCISE XI 94. -2. 1. . 1/3. 2.llz 2 + 9z .2=0 . 3/2. (iii) 4z 3 . 9/2. 2/3.6. 5. t(3 N/6) . 1.1 N/5). (ii) 2x* + 9x 2 + llx-f 3=0 . -2.9x*+llx -3=0. 10. -2. 2. 2. 1. 11. 5. 3. . V.Sx . 5. 3 J5___^___? ^_ 4 J-+' *-3 x-2 (x-2) 3 (x-2) 3 ' ' x-l i * 2 x ft . 7. 12. <r 9 T 1 ^T 1 \2 . 2/3. 3. -0. 3. 6. 14. 13. (ii) a. 4. (i) z 3 . 9.1/2. .1/2. -1. 7. 6. 6. 3 -75 = 0. . 16. 4. 2. 3-1. (iv) 2z 3 . (i)4z 3 -8z 2 + 8. 3/2.1 = 0. . 2. ANSWERS 569 EXERCISE X PAGE 86.c ^ x i ^ i.pendicular to qx =py. 4 25* 3 + 375z 2 . 3. 11. 7.r-3=0. EXERCISE XIII i i i 2. 3/2. 18. -4. -3. 4. t( . 15. 4.1. 9. 4.11700=0. . 8.1260*. 2. (i) x 3 . -6.1/2. -a. 19. -i(a4~& iv3a 2 + 2a& + 36 2 ). 17.2/3.

10. 1 -ax I -ox (a-b)(a-c) 'x-a' ' (a -b)(a-c) ' x-a ' 18. 15.5.1. _1 1. -I) 3 2(x-fl) 3 JLr 570 HIGHER ALGEBRA PAGE 104. -l + -JL_ + *_ . 17. X4-2 x-3 x-2 | -1) (a: 3 Q j^ _ _____ _i_ ______ 4. x + 2) 3 * ! "' x" 2 -fl" (x 2 -fl) 2 x 2 +~2 x " L(x+l) 2 x 2 + x-fl'(x 2 -fx+l) 2 13 .1.1 1 14. a* 1 . 5. Put x^din Ex.

(3n + 2j" 2 ' llf ~ n(n-f 2) 17. 10.. 4.f2)(3/* J Miw t. 344. 9.5. 19. j 3. 4300.I).2.157) 1 _ Li?_i 5 -Jil^^tl) 36(ft4-l)(rZ'4-2)(tt4-3) * " 2 2.. 1~. 220. ^(47i 2 -f7i-l).. .. where 20. (i) i-/i(6n a -3-l).!)(/<. n(29n 2 + 138r*4.1). pui>x~d. Tzn(n + l)(?i-f 2)(3n + 13). (ii) iw(n . 11. 12. 23. 16. 18. 19. EXERCISE XIV (B) 117.). 2926. l-n(2n + l)(2n. 21. 15. (2w + 2)" 10 * 2. iV/t^j -f. %n(n + l)(n + 2. 715.l)(9w a -9w -2). 14. 286. * ^y/2. 19. 50336. 13.4. x + a + b + c + Z.105. 1.. 3.^. (a~b)(a-c) x-a EXERCISE XIV (A) 116.^. -iVw(" * }) 2 (n + '2). 7-7-7 r. 8. 1. ^ . \ 2. A IV\UIV I I f _ 6..6 . - 140 4(2n + 7)(2ra i i r i i i si . 7.

W^T ii^S). 4. i^ v or a. 1. -|). 27. (i) a 3 + & 3 .2 T *v2). ( t. MX.fc .1 V. I I I ANSWERS 571 PAGE n)' (l-. # 4 + x 2 y 2 + y 4 = (# 2 + xy + y 2 ) (x 2 -xy + y 2 ). EXERCISE XVII 158. (W54^. (1 *N/2. 2. (1. y. 7. 5). a) 2 y. ( \/3. The only solutions are (0. where & = 2 r 28.y)(y .6. 2. 39 2 = 30 2 + 20 2 + 14 2 + 5 2 . 2). where k^=2 n . -2. 131. (f.2 iN/2. 10. ~ . 19. 8.f i>/57~. i. V|7 2 + & 2 + c 2 ). . and that the latter alternative is impossible. . a). 1).iTt'). (i) (0-y)(y-a)(y-|8). 8. 146. ^ ii -Tir.1. (0. (2. tD 2 x4: or -4^36 . in any order. x.. 3. ( . 0).(a 3 + 6 3 ) 2 .a)(a. ( -f. [Remove fractions and show that xy0 or a 2 . q= N/3). 1. ( .3. (iii) 36. (0. 0). 0. (ii) (jB . (ii) .y a fourth root of f .*:) ' 26. [Use the identity. = .i=F iV57).Q| __.r) 3 (I-*) 2 2(1-. 1 EXERCISE XV 129. ( . (a. 1. in any order. (a-iVSdbl^.7). . T 0. 3.(-|.] . 2. 9. (ii) 1. -----. -1.20. 4. -f.^ . -^).0)(a + + y) EXERCISE XVI 143.] 12. IT J-N/2). 5. 3. EXERCISE XVIII 165. 0). -3/2. (3. wy 2 or -* 11. 3: -2: 10.

21. 0). (i. 23. [Put s= 2o 572 HIGHER ALGEBRA PAGE 165. (if. 20.. -3). _ t with ftny ar _ rangement of signs. with any arrangement of signs. 16. [Show that 2x(y . (i. etc. -i).i.] 22. 0.0). the equations are not independent. If Zbc^Q.|. 1TV2). 15. (-!.] n>i . (-3. 19. If 2bc=Q. N/6c). 0). etc. (Voc.2) = 6 2 . 0. = ..i).. -3. 6. 24. (1. (0.. &*=a a (& a + c 2 -a 2 ).=(a6-a6 a c~l)/(a6 2 c 2 -6 2 c + 6). 0. the only solutions are (0. 1). in any order. (a.rt .. (i(a-6+c)VQ...2a)}. x (a z b 2 + a 2 c 2 -b 2 c*)labc. a(ca-b z )(ub-c 2 ) 25. c).1) or = (a + 6 + c)/(a-&-c). = i{l s/(l + 26 + 2c . the upper or lower sign to be taken throughout. or x= . (a. 17. etc. [Show that x(y-z)/bc = . B=^/(c 2 + a*-b 2 ). J(6 + c-a) . 0. 0.1. F VG) where Q = Za*~2Ebc. If J^V(& 2 + c 2 ~a 2 )..1).c 2 .. -1). 0).13. (1.. C= N /( then 2z=J3-hC'. 2z~A+B.1.-!. giving eight solutions. 2y=C+A. .6. . etc. .. dW2...c) or a. 14. 18. etc. (>J2. etc.. (0. where Jfc a =(6 2 + c 2 -a 2 )(c 2 + a a -6 a )(a 2 + 6 2 -c 2 ). a. (1.

1. for (iii).6 2 ) 2 -0. -1. 1. (i) 2 cos 40 or 1-53209. (ii) prove (x 3 + y*)(l -xy}~ 2a 4 -a 5 . 2. i(-lW15). and divide by x-y. i( 5. -2). i[3 9. -1-87939.2a . The values of x. 8. a 3 -f-6 3 + c 3 ^a 2 (6 + c)-f6 2 (c +