THEORY OF NUMBERS HOMEWORK 5 SOLUTIONS Computations (1) Compute the following which are related to the Möbius function: (a) Compute the values of the MöbiusP function µ(n) for n P = 1, · · · , 20 (b) In class, we showed that σ(n) = d|n d and τ (n) = d|n 1. Use the Möbius inversion formula to recover the functions f (n) = n and g(n) = 1 for n = 60. Solution: (a) The following table lists the values of the Möbius function: µ(1) = 1 µ(8) = 0 µ(15) = 1 µ(2) = −1 µ(9) = 0 µ(16) = 0 µ(3) = −1 µ(10) = 1 µ(17) = −1 µ(4) = 0 µ(11) = −1 µ(18) = 0 µ(5) = −1 µ(12) = 0 µ(19) = −1 µ(6) = 1 µ(13) = −1 µ(20) = 0 µ(7) = −1 µ(14) = 1 (b) The following table lists the values of the Möbius function, τ , and σ for the divisors of 60 µ(1) = 1 τ (1) = 1 σ(1) = 1 µ(2) = −1 τ (2) = 2 σ(2) = 3 µ(3) = −1 τ (3) = 2 σ(3) = 4 µ(4) = 0 τ (4) = 3 σ(4) = 7 µ(5) = −1 τ (5) = 2 σ(5) = 6 µ(6) = 1 τ (6) = 4 σ(6) = 12 µ(10) = 1 τ (10) = 4 σ(10) = 18 µ(12) = 0 τ (12) = 6 σ(12) = 28 µ(15) = 1 τ (15) = 4 σ(15) = 24 µ(20) = 0 τ (20) = 6 σ(20) = 42 µ(30) = −1 τ (30) = 8 σ(30) = 72 µ(60) = 0 τ (60) = 12 σ(60) = 168 Then, the Möbius inversion formula gives the following: µ(1)τ (60) + µ(2)τ (30) + µ(3)τ (20) + µ(4)τ (15) + µ(5)τ (12) + µ(6)τ (10) + µ(10)τ (6) + µ(12)τ (5) + µ(15)τ (4) + µ(20)τ (3) + µ(30)τ (2) + µ(60)τ (1) 1 = 12 − 8 − 6 + 0 − 6 + 4 + 4 + 0 + 3 + 0 − 2 = 1 and µ(1)σ(60) + µ(2)σ(30) + µ(3)σ(20) + µ(4)σ(15) + µ(5)σ(12) + µ(6)σ(10) + µ(10)σ(6) + µ(12)σ(5) + µ(15)σ(4) + µ(20)σ(3) + µ(30)σ(2) + µ(60)σ(1) = 168 − 72 − 42 + 0 − 28 + 8 − 12 + 0 + 7 + 0 − 3 = 60 Therefore, the values of the Möbius inversion formula are f (60) = 60 and g(60) = 1. (2) For all primes p between 1 and 20 and for all 1 ≤ n < p, compute the Legendre symbol np . Solution: The primes between 1 and 20 are 2, 3, 5, 7, 11, 13, 17, 19. Then the Legendre symbols are as follows: 1 1 1 2: =1 3: =1 5: =1 2 3 5 2 2 = −1 = −1 3 5 3 = −1 5 4 =1 5 1 1 1 =1 11 : =1 13 : =1 7: 7 11 13 2 2 2 =1 = −1 = −1 7 11 13 3 3 3 = −1 =1 =1 7 11 13 4 4 4 =1 =1 =1 7 11 13 5 5 5 = −1 =1 = −1 7 11 13 6 6 6 = −1 = −1 = −1 7 11 13 7 7 = −1 = −1 11 13 8 8 = −1 = −1 11 13 9 9 =1 =1 11 13 2 17 : 1 =1 17 2 =1 17 3 = −1 17 4 =1 17 5 = −1 17 6 = −1 17 7 = −1 17 8 =1 17 9 =1 17 10 = −1 17 11 = −1 17 12 = −1 17 13 =1 17 14 = −1 17 15 =1 17 16 =1 17 19 : 10 11 = −1 1 =1 19 2 = −1 19 3 = −1 19 4 =1 19 5 =1 19 6 =1 19 7 =1 19 8 = −1 19 9 =1 19 10 = −1 19 11 =1 19 12 = −1 19 13 = −1 19 14 = −1 19 15 = −1 19 16 =1 19 3 10 =1 13 11 = −1 13 12 =1 13 17 =1 19 18 = −1 19 (3) If a and n are relatively prime, then the order of a mod n is the smallest integer x such that ax ≡ 1 mod n. For n = 1, · · · , 15 and 1 ≤ a < n, compute the order of a mod n. Hint: For example, the order of 4 mod 9 is 3 since 4 6≡ 1 mod 9, 42 ≡ 16 ≡ 7 6≡ 1 mod 9, and 43 ≡ 64 ≡ 1 mod 9. Solution: 1: o(1) = 1 2: o(1) = 1 3: o(1) = 1, o(2) = 2 4: o(1) = 1, o(3) = 2 5: o(1) = 1, o(2) = 4, o(3) = 4, o(4) = 2 6: o(1) = 1, o(5) = 2 7: o(1) = 1, o(2) = 3, o(3) = 6, o(4) = 3, o(5) = 6, o(6) = 2 8: o(1) = 1, o(3) = 2, o(5) = 2, o(7) = 2 9: o(1) = 1, o(2) = 6, o(4) = 3, o(5) = 6, o(7) = 3, o(8) = 2 10 : o(1) = 1, o(3) = 4, o(7) = 4, o(9) = 2 11 : o(1) = 1, o(2) = 10, o(3) = 5, o(4) = 5, o(5) = 5, o(6) = 10, o(7) = 10, o(8) = 10, o(9) = 5, o(10) = 2 12 : o(1) = 1, o(5) = 2, o(7) = 2, o(11) = 2 13 : o(1) = 1, o(2) = 12, o(3) = 3, o(4) = 6, o(5) = 4, o(6) = 12, o(7) = 12, o(8) = 4, o(9) = 3, o(10) = 6, o(11) = 12, o(12) = 2 14 : o(1) = 1, o(3) = 6, o(5) = 6, o(9) = 3, o(11) = 3, o(13) = 2 15 : o(1) = 1, o(2) = 4, o(4) = 2, o(7) = 4, o(8) = 4, o(11) = 2, o(13) = 4, o(14) = 2 Proofs (4) (a) Suppose that f is a multiplicative function with f (1) = 1. Show that X µ(d)f (d) = (1 − f (p1 ))(1 − f (p2 )) · · · (1 − f (pk )) d|n where n = pa11 · · · pa22 · · · pakk is the prime factorization for n. (b) Use the formula above to find a simple formula for X µ(d)τ (d) d|n Solution: 4 P (a) Let G(n) = d|n µ(d)f (d), then we first show that G is multiplicative. Let m and n be relatively prime. Then X G(mn) = µ(d)f (d) d|mn = X µ(d1 d2 )f (d1d2 ) X µ(d1 )f (d1)µ(d2 )f (d2 ) d1 |m,d2 |n = d1 |m,d2 |n = X µ(d1 )f (d1) · d1 |m X µ(d2 )f (d2) d2 |n = G(m)G(n) Where the second inequality follows from the fact that m, n are relatively prime and is a proof we did in class. The second equality is because µ and fQare multiplicative. Since G is multiplicative, if n = pa11 · · · pakk , then G(n) = G(pai i ). Therefore, if we can show that for any prime p and a ≥ 1, G(pa ) = 1 − f (p), then the result follows. P Consider G(pa ) = d|pa µ(d)f (d). Now, if d|pa , then d must also be a power of p since all prime divisors of d would then divide p. In addition, the power of d must be less than a by the divisiblity relation. Therefore d is of the form pb where P 0 ≤ b ≤ a. Substituting this into the expression above, we have a G(p ) = 0≤b≤a µ(pb )f (pb ). For all b ≥ 2, µ(pb) = 0 since pb is not a product of discinct primes. For b = 1, µ(p1 ) = −1 since p is a product of one distinct prime, and for b = 0, µ(p0 ) = 1 since µ(1) = 1. Therefore, this sum simplifies to f (1) − f (p). In the statement of the problem, f (1) = 1, so this equals 1 − f (p), completing the proof. (b) Since τ (1) = 1 as 1 has only one divisor and τ is multiplicative, the result above applies. Moreover, for any prime p, τ (p) = 2 since there P are only two ak a1 positive divisors for p, 1, p. Therefore, if n = p1 · · · pk , the sum d|n µ(d)τ (d) = Q Q Q (1 − τ (pi )) = (1 − 2) = (−1) = (−1)k . (5) (a) Show that if b is a positive integer not divisible by the prime p, then b 2b (p − 1)b + +·+ =0 p p p (b) Let a, b be integers not divisible by p. Show that either one or all three of the integers a, b, ab are residues mod p. Solution: (a) This question is only true for p an odd prime because when p = 2, the sum is 1 = 1. Assume now that 2 p is an odd prime and that b = 1. In this case, the which is the sum over 1, · · · , p−1. We showed in class sum is p1 +· · ·+ (p−1) p that half of 1, · · · , p − 1 are residues and half of 1, · · · , p − 1 are nonresidues. Therefore, half of the symbols above are 1 and half are −1, so that the sum is 0. 5 Now, consider an arbitrary b. We showed in the proof of Fermat’s little theorem that {b, 2b, · · · , (p − 1)b} consists of elements congruent to {1, · · · , (p − 1)} mod p. In particular, there is a bijection between the two sets via congruence. Therefore, mod p, the set {b, 2b, · · · , (p − 1)b} is the same as the set {1, · · · , (p − 1)} just possibly in a differentorder. Then, since Legendre symbols b do not change under congruency the sum p + · · · + (p−1)b is the same sum b up to reodering and we know the second sum to be 0. as p1 + · · · + (p−1) p (b) We do this by a case-by-case analysis. If a, b are nonresidues, then ab = p b a = (−1)2 and so ab is a residue and there is one residue in {a, b, ab}. p p If a is a residue, butb is a nonresidue (or equivalently a is a nonresidue, but ab b b is a residue), then p = ap = (1)(−1) = (−1), so ab is a nonresidue p and there is one residue in {a, b, ab}. If a, b are residues, then let x, y be square roots, so x2 ≡ a mod p and y 2 ≡ b mod p. Then xy is a square root of ab since (xy)2 ≡ x2 y 2 ≡ ab mod p. Therefore all three of {a, b, ab} are residues. Note: We could also do the last case by using the same formula as above. Challenge (6) Show that if the equation φ(n) = k, where k is a positive integer, has exactly one solution n then 36|n. Hint: Prove this by contradiction, i.e., assume that 2 ∤ n or 4 ∤ n and construct a number m different from n such that φ(n) = φ(m) contradicting the assumption that there is only one solution. Solution: We prove this using a case-by-case analysis and contradiction: Let n be the solution to φ(n) = k. Assume first that 2 ∤ n, then φ(2n) = φ(2)φ(n) = 1 · φ(n) = φ(n) = k, which contradicts the assumption that φ(n) = k has only one solution. Assume now that 2|n. If 2|n, but 4 ∤ n, then n = 2m and 2 ∤ m. Then k = φ(n) = φ(2m) = φ(2)φ(m) = 1 · φ(m) = φ(m), which contradicts the assumption that φ(n) = k has only one solution. Assume now that 4|n. If 3 ∤ n, then n = 2l m where l ≥ 2, 3 ∤ m and 2 ∤ m. Then φ(n) = φ(2l m) = φ(2l )φ(m) = (2l − 2l−1 )φ(m). On the other hand, φ(3 · 2l−1 m) = φ(3)φ(2l−1)φ(m) = 2(2l−1 − 2l−2 )φ(m) = (2l − 2l−1 )φ(m), which contradicts the assumption that φ(n) = k has only one solution. Assume now that 4|n and 3|n. If 3|n, but 9 ∤ n, then n = 2l · 3m where l ≥ 2, 2 ∤ m, and 3 ∤ m. Then φ(n) = φ(2l )φ(3)φ(m) = (2l − 2l−1 )(3 − 1)φ(m) = (2l+1 − 2l )φ(m). On the other hand, φ(2l+1 m) = φ(2l+1)φ(m) = (2l−1 − 2l )φ(m), which contradicts the assumption that φ(n) = k has only one solution. Therefore, 4|n and 9|n which implies that 36|n. 6