8 THEORY OF EQUATIONS it is absolutely impossible to express, ‘by means of a formula involving only rational operations and root extractons, the roots of an equation of degree higher than fourth, when the coeficents are left arbitrary. ‘The proot of this impossibility belongs to higher parts of algebra and ‘cannot be attempted in this course. ‘As to the algebraic solution of ‘cubic and biquadratic equations, ‘the theory is comparatively: simple and oot be explained in this chapter and again resumed from a higher point of view in Chap. XII. 2. Cerdan’s Formules. There fs no loss of generality in taking the general cubic ‘equation in the form f@attatthete=0 since division by the coefficient of 2* does not modify the roots of the equation. By introduction of a new unknown ‘this equation can be Sinplifed, moreover, so that it will not contsin the second power of the ‘unknown. To this end we set gaytk with & still arbitrary. By ‘Taylor's formula jor rote Ov Bye aod J) =< Balt + bk+o, | fk) = BB+ Ik +b, jar" @ = thts, ser" = ‘To get rid of the term involving y* it suffices to choose & so that Bkea=0 or k=—3 C9 it follows that, after substituting Since then aay-f : the proposed equation is transformed into vtpytg=0 ay where peo-S @ 0 Be 208 eCUBIC AND BIQUADRATIC EQUATIONS 85 ‘A cubic equation of the form (1) can be solved by means of the following devioe: We seek to satisty it by setting yruty, ‘thus introducing two unknowns uand.v. On substituting this expression into (1) and arranging terms in a proper way, u and v have to satisfy the equation wot (p+3w)(u-+o) +9=0 @ with two unknowns. This problem is indeterminate unless another rela. tion between u and vis given. For this relation we take Bu+p =0 or ‘Then, it follows from (2) that: woiteng, 0 that the solution of the cubic (1) can be obtained by solving the system of tivo equations wtta-g wake @) ‘Taking the cube of the latter equation, we have wee ZB @ ‘and so, from equations (3) and (4), we know ‘the sum and the product of the two unknown quantities w and v*, These quantities are the roots of the quadratic equation a eta-m 0. ‘Denoting them by A and B, we have then ~94./842, ae at VE +e 4 ./f4 2. Ba-g- Varn where we are at liberty to choose the square root as we please. Now owing to the eymmetry between the terms uw’ and o* in the system (3) wwe can set wed, f=LY 86 THEORY OF EQUATIONS If some determined value of the cube root of A is denoted by VA, the three possible values of u will be unVE, un0V, ue ata where ritivs ——F isan imaginary cube root of unity. As to it will have also three values: 2=VB, veoVB, veutVE but aot every one of them can be associated with the three possible values of u, since u and » must eatisty the relation wok Uf VB stands for that cube root of B which ‘satisfies the relation WA-VB=- P ‘then, the values of » that can be associated with unVE, veoVd, unaT v= VB, v=o, Hence, equation (1) will have the following roots: na VE+VB, nod + ov, n= OVE +0VB. ‘These formulas are known as Cardan’s formulas after ‘the name of the Italian algebraist Cardan (1501-1576), who was the first to ‘publish them. It must be remembered that ~/A can be taken arbitrarily among ‘the three possible cube roots of A, but +/B must be so chosen that VA. VE =-2. will be VB. 8. Discussion of Solution. In discussing Cardan’s formulas we shall suppose that p and g are real numbers. Then, the nature of the roots will be shown to depend on the function A= 4p 4279 OF Ba wv but will qua ‘Ther 0 ths HenceCUBIC AND BIQUADRATIC EQUATIONS 87 Obviously, A will be positive, zero, or negative. Supposing fist A to be positive, the square root FP [a ata" Vi vill be real, and we shall take it positively. ‘Then, A and B will be real, and by VA we shall mean the real cube root of A. Since p is real and VE.YB~-2, VB will be the real cube root of B. Hence, equation (1) has a real root n= VALVB, Dut the other two roots, wn oVE + 0 V ~~ ZEEE pg VEE ho VE +08 ~~ MAEVE ygVE- VB will be imaginary conjugates since A and B are not equal and conse- quently Va - VB x0. Example 1. Let the proposed eubio equation be Bhat 200, int, it must be transformed by the substitution eau-k ‘The resulting equation for y (best found by synthetic division) is Ky - 945 =0, 0 that pe-h Hence, and VE = KVB+IBVE, VE = 4% —16V8, n= fF IE + VB— VD), no 5 (VF Vea) + 8 (ver - YeTETA), bm — (VFI + Von) ~ 28 (Seep - YATE).88 TREORY OF EQUATIONS Cairespondingly, th roots of the proposed equation are a= f+ IVE + V5 — 18S - 1), an -LOMTIE + VEIT 42) 4 (VETS — Vx —15V8), 5 (VTE + VBE + 2) - (ers a — V2 = 158). equation Ste-20 as, however, an ftegal root Land the remizing two rots, 145, are iaginary. On comparison with the expressions obtained from Cardan's formulas we discover the rather curious fact that V6 4 15V5 + W265 —15VE = 4, although both eube roots are irrational numbers. The explanation of this fellows from a comparison of the imaginary roots. ‘This comparison gives for the difference of the same cube roots ‘W285 + 15V8 - V2 15V5 = 23, VB PVE = 243, We 15V8-2- vi. Te follows that 26 + 153 and 26 — 153 are cubes of numbers 2-+ V3 and 2~ V3._ Such simplification of cube roots occurs always when the eubic equation hss a rational root but not otherwise, Example 2. To solve the equation Hen Hore the preliminary transformetion is not necessary and Cardan's formulas can be ‘pplied directly. We have whence 0. Peo gen, An1¢R ‘Consequently, the real root is VR - BAT eth agian rots ae ~LWVBF - VE?) 2 RFI + WUD.tt iss Seether tannin ical tat a me ava CUBIC AND BIQUADRATIC BQUATIONS 89 ‘To caleulate these roots appraximataly, use con be made of any of the wellinowa bsndbooks thst contain tables of squares, cubes, square roots, and cube root. From such tables itis found that VBB+1= 6.201502, -V2B—1 = 42015008, and further WVBR TT = 124s0800, VVB—1 = 16250815, ‘Hence, the real root of the proposed equation is approximately 1.8460840 — 1.6250615 = 0,2210225 ‘with the degree of approximation allowed by seven-place tables, To case A= 0 AnBa-h and the roots of the equation vtryte | 4, alle = ait. Rb wn db nil ee eee Bee eee oO Fo0 is trivia. Problems Solve the cubic equations: Lt 62~6=0, 22 1d — 0, Sa 0r—5 = 0. 42418 —6=0, : 5 2 $643 =0. 6 2 Bt 5 =O, 1. 82 — G2 =0. 8 2+ bot — 35 = 0. OP 48r+ oot =O, 10. + 62-4 G+ 5-0. LL het + 10h — 245 = 0, 18 Sat 4 12 + 1020 — AT = 48. #-242=0, Mot 8e~2=0, . 15 P+ Got + 8 +B =0. 36 Sot + 12st 4 G2 — 8 =. a Show that a. Vee a-VVe—9=1 1 VTE VET a2 19, VIG +10 - Vis — 10 = 2. 20, VER Van — Vas — Va = 23, 2, What isthe outer radius of a pbeciea!elill of thickness 1 in, if the volume of the ebell is equal to the volume ofthe hollow space inside? 22. Bolve Prob. 21.f the volume ofthe bellow space ee the volume ofthe shel. 23. An open box has the shape of s cube with each edge 10 in. long. If the capacity ofthe box is 600 cu. in, what isthe thickness of the walls? ‘The walls ae supposed ta be uniformly thick.a 90 THEORY OF EQUATIONS 4, Ineducible Case. We return now to the discussion of the general solution and consider what happens when A<0. A curious phe- ‘nomenon occurs here, for in this case ff. Pa; (A 4°27 108 is purely imaginary and both numbers are imaginary, 50 that the roots of equation (1) in Sec: 2 are expressed through the cube roots af imaginary numbers, and yet all three of them are real. To see this let VA n=otbi be one of the eube roots of A.’ Since B is conjugate to-A, the number a— bi will be one of the cube roots of B, and it must be taken equal to WB in order to satisfy the condition VE.VB=-2. ‘Thus, VA=atbi, VWE=0-bi, end from Cardan’s formulas it follows that the roots v= 2a, te = (at bio + (a— bit = — a bVv3, ¥s= (@ bia + (a — bio = — at bVB are real and, moreover, unequal. It is clear that ys > ys. If ak = us, wwe should have b=-cv3, 0 that WA = all ~iV3). But then A= al(1~ V8)! = — Bat would be real, which is not true. Similarly, it is shown that 1 > ys. Example, To olve the equation v-wyti=0.2 3 4 y : 3 wien cbt CUBIC AND BIQUADRATIC EQUATIONS 91 snd the real roots are presented in the form nara, neove+ uve, ne OVE + OV. ‘These expressions are not suitable for direct celoulation beeause of the cube roots of imaginary numbers, If we try to find Yenothi lgebraicaly, we are led to the solution of the to simultaneous equations . 1 va cr a Solving fr 8 a the it and substituting wed ee BET ooo oH) 4, whence ee we, We ra ‘Equating the two expressions for #, we have the equation 2a 2k Ge” Gia = TF ‘which efter duo simplifeations becomes a)’ + Be) ~ 24(2)* $1 = 0. Setting = = at, we have for = a cubic equation Bf ot We +10, hich by the substitution 2 = y — 1 ie transformed into My 2 = 0; or, setting y = — 8s, into B-& +190, ‘But this is the same equation thet we wanted to solve. Consequently, we did not sdvance a step in trying to find o and b by an algebraic process, ‘The fact that the zeal roots of a eubie equation ; vtpytgn0 a +E <0 sre presented in 1 form involving the cube roots of imeginary numbers puzaled the old algebraists for a long time, end this caso was called by them carus trreductbili, irreducible case. We know now thet, for instance, when p and ¢ are rational num- bers, but among the three real roots of an equation vtmwtgn092 THEORY OF EQUATIONS ‘none is rational, itis absolutely impossible to express any of these roots in a form {involving only real radicals of any kind. 5. Trigonometric Solution. In spite of the algebraic difficulties inherent in the irreducible case, it is possible to present the roots in a form suitable for numerical computation by extracting the cube root of =-44s,/-£-E Ann gtiy ga ‘rigonometricelly. ‘The square of the modulus of 4 is -C)-W-#-€2) <2) on (F whence or by its tangent -v=E nO on the condition that ¢ is taken in the first or second quadrant according as q is negative or positive, Having found p and ¢, we can take a= Vilom $4 isn f)= JF P(x 8+ ising), B= FP (ocs co = 608 120° + 4 sin 120°, the roots ys, vs, ys will’be given by and ‘Then, sinceCUBIC AND BIQUADRATIC EQUATIONS In practice it is more convenient to present ys and ys thus: any coe (aor - 2), ne 2B 00s (0 +§): Beample 1. Yor the equation voy ti=0 wehave { hence, $ = 120° and a 2ee4O", p= Dom 20, w= Doe 8, 3 Approximate values of the roots ean be found dizetly from the trigonometse tables 3 and xe 4 n= 1.5s20ssse62, | wm ~ La7asss2416, a 1 = 0.347 2969554. 3 Example 2 To cove the equation voto For this equation peter nes sod 2 1 4 tava Further computation was made with six-place tables of logarithms, The calcula ‘ions and results ean be presented as follows: ¢ log 27 = LAS1964 og 7 = 0848008 tog ow $ = 99nE128 — 10 log V37 = 0.715682 Jog 3 = 0.477121 tegtang = 9256518 “Top gS OST $= 10'E8'90"2 log VIE = 0.183088 Hib 88TS20 tog 2 = 0201020 tog cos (00° +) = o.7508 — 10 s log 2V% = 0.485018 0.485018 be (= = n= soi =r Taam boas (08) = ora — 10