Algebra A Solutions A1 Suppose a, b, c > 0 are integers such that abc − bc − ac − ab + a + b + c = 2013. (A.1) Find the number of possibilities for the ordered triple (a, b, c). Solution. Subtracting 1 from both sides and factoring, we obtain 2012 = (a − 1)(b − 1)(c − 1). (A.2) Writing a0 = a − 1 and similarly with b0 , c0 , it suffices to count the ordered integer triples (a0 , b0 , c0 ) such that a0 , b0 , c0 ≥ 0 and a0 b0 c0 = 2012. Immediately, we require a0 , b0 , c0 > 0. Let τ (n) denote the number of positive divisors of n. The prime factorization of 2012 is 22 · 5031 , so the number of possible triples is X τ (d) = τ (1) + τ (2) + τ (22 ) + τ (503) + τ (2 · 503) + τ (22 · 503) d|2012 d>0 which simplifies to 1 + 2 + 3 + 2 + 4 + 6 = 18 . A2 Find the number of pairs (n, C) of positive integers such that C ≤ 100 and n2 + n + C is a perfect square. Remark. This problem was inspired by the “New Year’s Problem” of the 2008-2009 Wisconsin Math Talent Search. Solution. We can write n2 + n + C = (n + m)2 for some integer m ≥ 1. Expanding and rearranging, C = m2 + (2m − 1)n, so we have reduced the problem to counting pairs (m, n) of positive integers such that m2 + (2m − 1)n ≤ 100. This is 10 X m=1 $ 100 − m2 2m − 1 % = 99 + 32 + 18 + 12 + 8 + 5 + 3 + 2 + 1 + 0 which simplifes to 180 . 1 (A.3) A3 Let x1 = √ 10 and y1 = √ 3. For all n ≥ 2, let √ xn = xn−1 77 + 15yn−1 √ yn = 5xn−1 + yn−1 77 (A.4) (A.5) Find x65 + 2x45 − 9x45 y52 − 12x25 y52 + 27x25 y54 + 18y54 − 27y56 . Solution. Let fn = x2n − 3yn2 . We factor x6n + 2x4n − 9x4n yn2 − 12x2n yn2 + 27x2n yn4 + 18yn4 − 27yn6 = (x2n − 3yn2 )3 + 2(x2n − 3yn2 )2 = fn3 + 2fn2 . (A.6) Moreover, f1 = 1, and by induction, we can prove fn = 2fn−1 for all n ≥ 2. In particular, f5 = 24 , so the answer to the problem is 212 + 3 · 28 = 4096 + 512 = 4608 . A4 Suppose a, b are nonzero integers such that two roots of x3 + ax2 + bx + 9a coincide, and all three roots are integers. Find |ab|. Solution. Write x3 + ax2 + bx + 9a = (x − r)2 (x − s) (A.7) for some integers r, s. Expanding and equating coefficients, we require 2 r s = −9a + 2rs = b 2r + s = −a r2 (A.8) Combining the first and third constraints gives r2 s − 18r − 9s = 0. Solving this quadratic in r, we find that 36(s2 + 9) must be a perfect square, because r, s are integers. Therefore, s2 + 9 = d2 for some d > 0, whence (d + s)(d − s) = 9. We cannot have s = 0 because this would imply a = 0 by the first constraint in the display, so the only solutions for s are ±4. Then (r, s) = ±(6, 4), so |ab| = |(2r + s)(r2 + 2rs)| = 1344 in both cases. 2 A5 Suppose w, x, y, z satisfy w + x + y + z = 25 wx + wy + wz + xy + xz + yz = 2y + 2z + 193 (A.9) (A.10) The largest possible value of w can be expressed in lowest terms as w1 /w2 for some integers w1 , w2 > 0. Find w1 + w2 . Remark. This problem was inspired by USAMTS Year 21, Round 4, Problem 2. Solution. We solve the problem in greater generality. Suppose W, X, Y, Z ≥ 0 satisfy the modified system W +X +Y +Z =A (A.11) W X + W Y + W Z + XY + XZ + Y Z = B (A.12) Expanding (X − Y )2 + (X − Z)2 + (Y − Z)2 ≥ 0, we obtain (X + Y + Z)2 ≥ XY + XZ + Y Z. Therefore, B = W (X + Y + Z) + XY + XZ + Y Z ≤ W (A − W ) + (A − W )2 . 3 (A.13) After rearranging, we get the following quadratic inequality in W : 2W 2 − AW + (−A2 + 3B) ≤ 0, (A.14) √ whence the largest possible value of W is ≤ (A + 9A2 − 24B)/4. But this value is actually attained at X = Y = Z, as we can see by solving the system where W + 3X = A and 3W X + 3X 2 = B. Setting (W, X, Y, Z) = (w − 1, x − 1, y + 1, z + 1), we find that the original problem is equivalent to the problem in W, X, Y, Z with (A,√ B) = (25, 193 + 25 − 2) = (25, 216). Thus, the largest possible value of w = W + 1 is (25 + 441)/4 + 1 = 46/4 + 1 = 25/2, giving an answer of 25 + 2 = 27 . A6 r Suppose the function ψ satisfies ψ(1) = 2+ q real x. Determine the greatest integer less than 3 2+ √ Q100 2 and ψ(3x) + 3ψ(x) = ψ(x)3 for all n n=1 ψ(3 ). 3 Solution. In order to compute ψ(33 ), we can choose ψ to be any function that satisfies the stated conditions. In particular, if α is any real number and ψ(x) = 2 cos(αx), then ψ(3x) = 2 cos(3αx) = 2(4 cos3 (αx) − 3 cos(αx)) = (2 cos(αx))3 − 3(2 cos(αx)) = ψ(x)3 − 3ψ(x). By applying the half-angle formula twice to 2 cos(π/4) = r 2 cos(π/16) = q 2+ 2+ (A.15) √ √ 2, we find 2 (A.16) so we can take α = π/16 = 2π/32. Since 38 = (34 )2 ≡ (−15)2 ≡ 1 (mod 32), we find that ψ(3n+8 ) = ψ(3n ) for all n. Now, 4 Y 2 cos(3n π/16) = 4 sin(π/8) cos(π/8) = 2 sin(π/4) = √ 2 (A.17) n=1 √ √ √ Q Q n 12 2 = 4096 2. We and similarly 8n=5 2 cos(3n π/16) = 2. Therefore, 100 n=1 ψ(3 ) = 2 check that 5792 is the largest integer whose square is ≤ 2 · 40962 . A7 Evaluate s r q √ 2013 + 276 2027 + 278 2041 + 280 2055 + . . . (A.18) Remark. This problem was inspired by a theorem of Ramanujan. He famously posed the similar problem of evaluating s r q √ 1 + 2 1 + 3 1 + 4 1 + ... (A.19) in the Journal of the Indian Mathematical Society, with no response after six months. A good, albeit somewhat romanticized, place to read about Ramanujan is [?]. Solution. We prove that a+N +n s = r (a + n)2 + aN + N q (a + n)2 + a(N + n) + (N + n) (a + n)2 + . . . 4 (A.20) for all a ∈ R and N, n ∈ Z≥0 with n | N . The result holds for N = 0 and any a, n, and if it holds for an arbitrary triple (a, N, n), then squaring both sides and simplifying, r N + a + 2n = q (a + n)2 + a(N + n) + (N + n) (a + n)2 + . . . (A.21) so it holds for (a, N + n, n), whence induction finishes the proof. Our original problem is the case (a, N, n) = (7, 276, 2). The answer is a + N + n = 285 . A8 The author thanks Will Sawin for his contribution to this solution. Let S be the set of permutations of {1, 2, . . . , 6}, and let T be the set of permutations of S that preserve compositions: i.e., if F ∈ T , then F (f2 ◦ f1 ) = F (f2 ) ◦ F (f1 ). (A.22) for all f1 , f2 ∈ S. Find the number of elements F ∈ T such that if f ∈ S satisfies f (1) = 2 and f (2) = 1, then (F (f ))(1) = 2 and (F (f ))(2) = 1. Solution. We can, and will, write the elements of S in “cycle” notation: For example, f = (1, 2, 4)(3, 5) means f (1) = 2, f (2) = 4, f (3) = 5, f (4) = 1, f (5) = 3, f (6) = 6. We write ι for the identity, or null, permutation. Right-to-left composition of permutations can be expressed as left-to-right1 composition of cycles: For example, (1, 2)(1, 2) = (1, 2)2 = ι and (1, 2)(1, 3) = (1, 2, 3). In particular, every permutation can be decomposed into a composition of transpositions, not necessarily in a unique way. Moreover, all the transpositions are generated by the 6 transpositions (1, 2), (1, 3), . . . , (1, 6). We want to first describe T completely. For all f ∈ S, there is a uniquely associated element Ff ∈ T defined by Ff (g) = f −1 ◦ g ◦ f . However, these do not account for all of T . To get the others, it is necessary to describe S in more detail: 6! We can classify the elements of S according to “cycle type.” For example, there are 2·4! = 15 6! transpositions, or type-2 cycles; there are 3·3! = 40 type-3 cycles, such as (1, 2, 3); there 6! are 2·2·2·3! = 15 type-2, 2, 2 cycles, such as (1, 2)(3, 4)(5, 6); and so on. The main thing to notice is that there are as many transpositions as type-2, 2, 2 cycles, which suggests that a suitable bijection between them could extend to an element of T , since the transpositions generate T . 1 We introduce this change in the direction of composition because it makes the results from cycle composition more clear. 5 Let S2 be the set of transpositions, and let S2,2,2 be the set of type-2, 2, 2 cycles. By the associativity of composition of permutations, a bijection S2 → S2,2,2 extends to an element F0 ∈ T if and only if it preserves compositions on S2 , i.e., F0 (f1 ◦f2 ) = F0 (f1 )◦F0 (f2 ) for all f1 , f2 ∈ S2 . The composition of two transpositions is either ι, a type-3 cycle, or a type-2, 2 cycle. The composition of two type-2, 2, 2 cycles is either ι; a type-2, 2 cycle, if they share exactly one transposition; or a type-3, 3 cycle, if they share no transpositions. Therefore, it suffices to find a bijection S2 → S2,2,2 that sends disjoint transpositions to 2, 2, 2-cycles that share exactly 1 transposition, and transpositions with 1 number in common to type-2, 2, 2 cycles that share no transpositions, such that in the latter case, the induced map on the compositions is well-defined: i.e., F0 ((1, 2)(1, 3)) = F0 ((2, 3)(1, 2)) = F ((1, 3)(2, 3)). These conditions are met if, for example, we let F0 be defined by (1, 2) (1, 3) (1, 4) (1, 5) (1, 6) ... 7→ 7 → 7→ 7→ 7→ (1, 2)(3, 6)(4, 5) (1, 6)(2, 4)(3, 5) (1, 3)(2, 5)(4, 6) (1, 5)(2, 6)(3, 4) (1, 4)(2, 3)(5, 6) ... (A.23) (Our choice of the image of (1, 2) follows the choice in [?].) It is useful to note that F0 ◦ F0 is the identity permutation on elements of S. We can check that a given element of T takes the form Ff for some f ∈ F if and only if it sends isolated transpositions to isolated transpositions. Therefore, F0 does not arise from an f in this way. However, we claim that T = {Ff : f ∈ S} ∪ {Ff ◦ F0 : f ∈ S}. (A.24) First, observe that if C is the set of all elements of S of a given cycle type, then any element of T , being invertible, must map C bijectively onto another set of the same form, i.e., consisting of all elements of another given cycle type. Such a set is called a conjugacy class of elements of S. But one verifies that the only conjugacy classes of size 15 in S are S2 and S2,2,2 . Altogether, if we take an arbitrary G ∈ T , then either G keeps the transpositions in place or exchanges them with the type-2, 2, 2 cycles; in the former case, G = Ff for some f , and in the latter case, G ◦ F0 keeps transpositions in place, so G ◦ F0 = Ff for some f , whence G = G ◦ F0 ◦ F0 = Ff ◦ F0 . This proves the claim. Finally, we can compute the answer to the problem. Let S 0 be the set of g ∈ S such that g(1) = 2 and g(2) = 1. If Ff maps S 0 onto itself, then by computation, any cycle notation for f must either contain (1, 2), or else cannot contain any cycles involving either 1 or 2. There is a bijection between the possibilities for f in the former case and those in the latter case; moreover, in the latter 6 case, counting by cycle type shows there are 1 + 6 + 8 + 3 + 6 = 24 possibilities, so that in total, there are 48 possibilities. Conversely, any such f yields an Ff that maps S 0 onto itself. By the example above, we know we can choose F0 such that F0 ((1, 2)) = (1, 2)(3, 6)(4, 5). Thus, any element Ff ◦ F0 such that Ff sends any one of (1, 2), (3, 6), or (4, 5) to (1, 2) will map S 0 onto itself. So there are also precisely 3(48) = 144 elements of the form Ff ◦ F0 that map S 0 onto itself. Altogether the answer to the problem is 48 + 144 = 192 . Remark. For further information about this solution, we advise the reader to consult [?]. For more about the mathematical objects at work in the problem, we recommend the beautiful exposition in Ch. 1-3 of [?]. In particular, S and T have the structure of groups; S is S6 , the symmetric group on 6 letters, and T = Aut(S6 ), its automorphism group. References [He] I. N. Herstein. Abstract Algebra. 3rd Ed., Prentice Hall (1996). [JR] G. Janusz, J. Rotman. “Outer Automorphisms of S6 .” Amer. Math. Mon., Vol. 89, No. 6 (1982), 407-410. [Ka] R. Kanigel. The Man Who Knew Infinity: a Life of the Genius Ramanujan. New York: Charles Scribner’s Sons (1991). ISBN 0-684-19259-4. 7