Exercises from Section 1.2.4

Tord M. Johnson

March 9, 2014

1. [00] What are 1.1, 1.1, 1.1, 0.99999, and lg 35?

We have:

1.1 = 1 1 1.1 < 2 1.1 = 2 2 1.1 < 1 1.1 = 1 2 < 1.1 1 0.99999 = 0 0 0.99999 < 1 lg 35 = 5 5 = lg 32 lg 35 < lg 64 = 6

2. [01] What is x?

x = x since x is an integer and x 1 < xx.

3. [M10] Let n be an integer, and let x be a real number. Prove that a) x < n if and only if x < n; b) n x if and only if n x; c) x n if and only if x n; d) n < x if and only if n < x; e) x = n if and only if x 1 < n x, and if and only if n x < n + 1; f) x = n if and only if x n < x + 1, and if and only if n 1 < x n.

[These formulas are the most important tools for proving facts about x and x.]

We may prove the various propositions.

Proposition (A). x < n if and only if x < n for all integers n, real numbers x.

Proof. Let n be an integer and x a real number. We must show that x < n if and only if x < n.

If x < n:

x x < x + 1 x < nx + 1 n x < n

If x < n:

x x < x + 1 x < n x < n

Therefore, x < n if and only if x < n. □

Proposition (B). n x if and only if n x for all integers n, real numbers x.

Proof. Let n be an integer and x a real number. We must show that n x if and only if n x.

If n x:

x x < x + 1 n x n x

If n x:

x x < x + 1x 1 x 1 < x n x n < x + 1n x

Therefore, n x if and only if n x. □

Proposition (C). x n if and only if x n for all integers n, real numbers x.

Proof. Let n be an integer and x a real number. We must show that x n if and only if x n.

If x n:

x 1 < x x x n x n

If x n:

x 1 < x x x n x 1 < nx n

Therefore, x n if and only if x n. □

Proposition (D). n < x if and only if n < x for all integers n, real numbers x.

Proof. Let n be an integer and x a real number. We must show that n < x if and only if n < x.

If n < x:

x 1 < x x n < xn x 1 n < x

If n < x:

x 1 < x x n < x n < x

Therefore, n < x if and only if n < x. □

Proposition (E). x = n if and only if x 1 < n x, and if and only if n x < n + 1 for all integers n, real numbers x.

Proof. Let n be an integer and x a real number. We must show that x = n if and only if x 1 < n x, and if and only if n x < n + 1.

If x = n:

x = nn xn x from (b) x = nx nx < n + 1x < n + 1 x 1 < nfrom (a) x 1 < n x n x < n + 1

If x 1 < n x:

x 1 < n xn x from (b) x 1 < n xx < n + 1x < n + 1x nfrom (a) n xx nx = n

If n x < n + 1:

n x < n + 1n x from (b) n x < n + 1x < n + 1x < n + 1x nfrom (a) n xx nx = n

Therefore, x = n if and only if x 1 < n x, and if and only if n x < n + 1. □

Proposition (F). x = n if and only if x n < x + 1, and if and only if n 1 < x n for all integers n, real numbers x.

Proof. Let n be an integer and x a real number. We must show that x = n if and only if x n < x + 1, and if and only if n 1 < x n.

If x = n:

x = nx nx n from (c) x = nn xn 1 < xn 1 < x n < x + 1from (d) x n < x + 1 n 1 < x n

If x n < x + 1:

x n < x + 1x nx n from (c) x n < x + 1n 1 < xn 1 < xn xfrom (d) x n n xx = n

If n 1 < x n:

n 1 < x nx nx n from (c) n 1 < x nn 1 < xn 1 < xn xfrom (d) x n n xx = n

Therefore, x = n if and only if x n < x + 1, and if and only if n 1 < x n. □

4. [M10] Using the previous exercise, prove that x = x.

Proposition. x = x for any real number x.

Proof. Let x be an arbitrary real number. We must show that x = x.

Given

x 1 < x x x + 1 > x x xx < x + 1

and

x 1 < x x x x < x + 1 x x > x 1 x 1 < xx,

by the previous exercise, we have

x = x

as we needed to show. □

5. [16] Given that x is a positive real number, state a simple formula that expresses x rounded to the nearest integer. The desired rounding rule is to produce x when x mod 1 < 1 2, and to produce x when x mod 1 1 2. Your answer should be a single formula that covers both cases. Discuss the rounding that would be obtained by your formula when x is negative.

A simple formula to express x rounded to the nearest integer could be given as

round (x) = x + 1 2.

Note that it satisfies the requirements, as

round (x) = x + 1 2 = x + 1 2 + xx = x + x x + 1 2 = x + x mod 1 + 1 2 = x + x mod 1 + 1 2 = x + 0 = xif x mod 1 < 1 2 x + 1 = xif x mod 1 1 2.

(Note that x mod 1 1 2 implies x is not an integer, otherwise x mod 1 = 0.)

For negative values of x, we find that (in general), round (x) = round (x) except when x mod 1 = 1 2, in which case x is rounded away from zero if positive, towards zero if negative.

6. [20] Which of the following equations are true for all positive real numbers x? (a) x = x; (b) x = x; (c) x = x.

Some, but not all, of the equations are true.

(a)
x = x is true, since for an arbitrary integer n,
x = n n x < n + 1 n2 x < (n + 1)2 n2 x < (n + 1)2 n x < n + 1 x = n.
(b)
x = x is true, since for an arbitrary integer n,
x = n n < x n + 1 n2 < x (n + 1)2 n2 < x (n + 1)2 n x < n + 1 x = n.
(c)
x = x is not true, as can be demonstrated by counterexample. Consider x = 9 4. Then
9 4 = 1 = 1 = 1

but

9 4 = 3 2 = 2.

7. [M15] Show that x + yx + y and that equality holds if and only if x mod 1 + y mod 1 < 1. Does a smiliar formula hold for ceilings?

We may prove the proposition.

Proposition. x + yx + y and equality holds if and only if x mod 1 + y mod 1 < 1.

Proof. Let x and y be arbitrary real numbers. We must show that

x + yx + y

and that equality holds if and only if x mod 1 + y mod 1 < 1.

But

x + y = x + x mod 1 + y + y mod 1 = x + y + x mod 1 + y mod 1 = x + y if x mod 1 + y mod 1 < 1 x + y + 1otherwise.

That is, x + yx + y and equality holds if and only if x mod 1 + y mod 1 < 1, as we needed to show. □

A similar formula holds for ceilings. In particular

x + yx + y

with equality if and only if (x) mod 1 + (y) mod 1 < 1 since

x + y = x y = x + (x) mod 1 + y + (y) mod 1 = xy(x) mod 1 + (y) mod 1 = x + y if (x) mod 1 + (y) mod 1 < 1 x + y 1otherwise.

Note that if x and y are not integers,

(x) mod 1 + (y) mod 1 < 1 x x + y y < 1 x + x + y + y < 1 x + x + 1 + y + y + 1 < 1 x + x 1 + y y 1 > 1 x mod 1 + y mod 1 2 > 1 x mod 1 + y mod 1 > 1.

8. [00] What are 100 mod 3, 100 mod 7, 100 mod 7, 100 mod 0?

We have:

100 mod 3 = 1 since 1 = 100 33(3) 100 mod 7 = 2 since 2 = 100 14(7) 100 mod 7 = 5 since 5 = 100 71007 = 100 + 7(15) 100 mod 0 = 100.

9. [05] What are 5 mod 3, 18 mod 3, 2 mod 3?

We have:

5 mod 3 = 1 since  1 = 5 (3)53 = 5 3(2) 18 mod 3 = 0 since 0 = 18 (3)183 = 18 3(6) 2 mod 3 = 2 since  2 = 2 (3)23 = 2 + 3(0).

10. [10] What are 1.1 mod 1, 0.11 mod .1, 0.11 mod .1?

We have:

1.1 mod 1 = 0.1 since 0.1 = 1.1 11.11 0.11 mod 0.1 = 0.01 since 0.01 = 0.11 0.10.110.1 0.11 mod 0.1 = 0.09 since 0.09 = 0.11 + 0.10.110.1 = 0.11 0.1(2).

11. [00] What does “x y (modulo 0)” mean by our conventions?

x y(mod0)” means x = y, since x y(mod0) is equivalent to asserting x mod 0 = x = y = y mod 0.

12. [00] What integers are relatively prime to 1?

All integers are relatively prime to 1 since for any integer n, gcd (n,1) = 1.

13. [M00] By convention, we say that the greatest common divisor of 0 and n is |n|. What integers are relatively prime to 0?

Given that gcd (0,n) = |n| for any integer n, then only 1 and 1 are relatively prime to 0.

14. [12] If x mod 3 = 2 and x mod 5 = 3, what is x mod 15?

We have:

x 2(mod3) x 3(mod5) 5x 10(mod15) 3x 9(mod15) by Law C (5 3)x (10 9)(mod15) by Law A (3 2)x (9 1)(mod15) by Law A x 8(mod15).

15. [10] Prove that z(x mod y) = (zx) mod (zy). [Law C is an immediate consequence of this distributive law.]

Proposition. z(x mod y) = (zx) mod (zy), z0.

Proof. Let x, y, and z be arbitrary real numbers, z0. We must show that

z(x mod y) = (zx) mod (zy).

But

x mod y = x y x y

if and only if

z(x mod y) = z x y x y = zx zy x y = zx zy zx zy = (zx) mod (zy)

as we needed to show. □

16. [M10] Assume that y > 0. Show that if (x z)y is an integer and if 0 z < y, then z = x mod y.

Proposition. if y|(x z) and 0 z < y, then z = x mod y.

Proof. Let x, y, and z be arbitrary integers such that y|(x z) and 0 z < y. We must show that z = x mod y.

But

x mod y = x y x y = x y x + z z y = x y x z y + z y = x y x z y + z y since y|(x z) = x y x z y + 0 since 0 z < y = x yx z y = x x + z = z

as we needed to show. □

17. [M15] Prove Law A directly from the definition of congruence, and also prove half of Law D: If a b(modulo rs), then a b(modulo r) and a b(modulo s). (Here r and s are arbitrary integers.)

We may prove Law A.

Proposition. If a b and x y, then a ± x b ± y and ax by(modm).

Proof. Let a, b, x, y, and m be arbitrary integers so that a b and x y. We must show that a ± x b ± y and ax by(modm).

For some integer r and s, we have

a = b + mr x = y + ms.

In the case of addition, a + x b + y(modm) since

a + x = b + mr + y + ms = b + y + m(r + s)

for some integer r + s.

Similarly, in the case of subtraction, a x b y(modm) since

a x = b + mr y ms = b y + m(r s)

for some integer r s.

In the case of multiplication, ax by(modm) since

ax = (b + mr)(y ms) = by + ymr bms m2rs = by + m(yr bs mrs)

for some integer yr bs mrs.

Therefore, if a b and x y, then a ± x b ± y and ax by(modm) as we needed to show. □

We may also prove half of Law D, which doesn’t require the assumption that r s.

Proposition. If a b(modrs), then a b(modr) and a b(mods).

Proof. Let a, b, r, and s be arbitrary integers so that a b(modrs). We must show that a b(modr) and a b(mods).

But

a = b + rst

for some integer t, in which case we have that a = b + ru and a = b + sv for integers u = st and v = rt.

Therefore, if a b(modrs), then a b(modr) and a b(mods). □

18. [M15] Using Law B, prove the other half of Law D: If a b(modulo r) and a b(modulo s), then a b(modulo rs), provided that r s.

Proposition. If a b(modr) and a b(mods), then a b(modrs), provided r s.

Proof. Let a, b, r, and s be arbitrary integers so that a b(modr) and a b(mods), with r s. We must show that a b(modrs).

But

a b(modr),

or equivalently, a = b + ru for some integer u. We also necessarily have that ru = sv = 0 + sv for some integer v, or equivalently, that

ru 0(mods)

since

a = b + ru a = b + sv a a = (b + ru) (b sv) 0 = ru sv ru = sv.

By Law B, since r s,

u 0(mods),

or equivalently, u = 0 + sv = sv for some integer v. Substituting sv for u gives us that a = b + rsv, or equivalently, that

a b(modrs).

Therefore, if a b(modr) and a b(mods), then a b(modrs), provided r s. □

19. [M10] (Law of inverses.) If n m, there is an integer n such that nn 1(modulo m). Prove this, using the extension of Euclid’s algorithm (Algorithm 1.2.1E).

Proposition. If n m, there is an integer n such that nn 1(modm).

Proof. Let n and m be arbitrary integers such that n m. We must show that there exists an integer n such that nn 1(modm).

Let d be the greatest common divisor of m and n as computed by Algorithm 1.2.1E. Then there exists two other integers m and n such that

mm + nn = d.

Since n m, we must have that d = 1, and so, nn = 1 + m(m), or equivalently,

nn 1(modm)

as we needed to show. □

20. [M15] Use the law of inverses and Law A to prove Law B.

Proposition. If ax by and a b, and if a m, then x y(modm).

Proof. Let a, b, x, y, and m be arbitrary integers such that ax by, a b, and a m (modm). We must show that x y(modm).

Since a m, by the law of inverses we know there exists an integer a such that

aa 1(modm).

Since a b, from Law A

aa ba 1(modm).

Finally, since ax by(modm), by Law A again, multiplying the congrunce by a yields

aax aby x y(modm)

as we needed to show. □

21. [M22] (Fundamental theorem of arithmetic.) Use Law B and exercise 1.2.1-5 to prove that every integer n > 1 has a unique representation as a product of primes (except for the order of the factors). In other words, show that there is exactly one way to write n = p1p2pk where each pj is prime and p1 p2 pk.

Proposition. Every integer n > 1 has a unique representation as a product of primes (except for the order of the factors).

Proof. Let n be an arbitrary integer such that n > 1. We must show that n has a unique representation as a product of primes (except for the order of the factors).

By exercise 1.2.1-5, we have that n = 1irpi for some arbitrary set of primes pi, 1 i r. We must show that if n = 1jsqj for some other arbitrary set of primes qj, 1 j s, that in fact, r = s and pi = qσ(i) for some permutation σ(i) = j.

Let us assume, however, they are not. That is, let us assume that for all 1 j s, qjp1. Since 1irpi = 1jsqj, we have that

1irpi 0(modp1) 1jsqj 0(modp1).

As all pi and qj are each a set of primes and qjp1, we have that qj p1, allowing us to apply Law B (since qj qj(modp1)) successively until we obtain

1 0(modp1).

But this requires p1 = 1, and since p1 is a prime, a condradiction. Hence, there exists a qj such that qj = p1. We may factor this prime out of our equation as

np1 = 2irpi = 1js jj qj = 1jsqj p1 .

If n was prime, then clearly n = p1 and we have proven there is a unique representation for this trivial case r = s = 1. Otherwise, we may prove by induction on k = r = s that this is so. That is, if we assume that

nk = 1ikpi = 1ikqσ(i)

is a unique factorization, we must show that

nk+1 = 1ik+1pi = 1jsqj = 1ik+1qσ(i)

is too for some integer s = k + 1. But we can make a similar proof by contradiction, assuming that for all 1 j s, qjpk+1. Since 1irpi = 1jsqj, we have that

1irpi 0(modpk+1) 1jsqj 0(modpk+1).

As all pi and qj are each a set of primes and qjpk+1, we have that qj pk+1, allowing us to apply Law B (since qj qj(modpk+1)) successively until we obtain

1 0(modpk+1).

But this requires pk+1 = 1, and since pk+1 is a prime, a condradiction. Hence, there exists a qj such that qj = pk+1. Then

nk+1 = 1ik+1pi = pk+1 1ikpi = pk+1 1ikqσ(i) = qj 1ikqσ(i) = 1ik+1qσ(i).

for

σ(i) = σ(i)if 1 i k j if i = k + 1.

We necessarily have s = k + 1, as both pk+1 and qj are primes and may not be further factored (leading to the inequalities s < k + 1 and s > k + 1, respectively, if they were not). Hence the result as we needed to show. □

22. [M10] Give an example to show that Law B is not always true if a is not relatively prime to m.

An example to show that Law B is not always true if a is not relatively prime to m is for a = 2, b = 0, x = 1, y = 0, and m = 2 so that a = 2⊥̸2 = m. Then

2(1) 0(0)(mod2), 2 0(mod2)

but

10(mod2).

23. [M10] Give an example to show that Law D is not always true if r is not relatively prime to s.

An example to show that Law D is not always true if r is not relatively prime to s is for a = r = s = 2 and b = 0 so that r = 2⊥̸2 = s. Then

2 0(mod2), 2 0(mod2)

but

20(mod2(2)).

24. [M20] To what extent can Laws A, B, C, and D be generalized to apply to arbitrary real numbers instead of integers?

We have that Law A for addition and subtraction hold, as well as Law C.

We may prove Law A.

Proposition. If a b and x y, then a ± x b ± y and ax by(modm).

Proof. Let a, b, x, y, and m be arbitrary real numbers so that a b and x y. We must show that a ± x b ± y.

For some integer r and s, we have

a = b + mr x = y + ms.

In the case of addition, a + x b + y(modm) since

a + x = b + mr + y + ms = b + y + m(r + s)

for some integer r + s.

Similarly, in the case of subtraction, a x b y(modm) since

a x = b + mr y ms = b y + m(r s)

for some integer r s.

Therefore, if a b and x y, then a ± x b ± y and ax by(modm) as we needed to show. □

To see why Law A for multiplication does not hold for the real numbers, consider as an example a = 2 3, b = 1 3, x = 3 5, y = 2 5, and m = 1. Then

2 3 1 3(mod1), 3 5 2 5(mod1)

but

2 5 2 15(mod1).

To see why Law B does not hold for the real numbers, consider as an example a = 2 3, b = 1 3, x = 3 5, y = 9 5, and m = 1. Then

6 15 9 15(mod1), 2 3 1 3(mod1)

but

3 59 5(mod1).

(Note that even the more general relation x y(mod m gcd (a,m))3 5 9 5(mod3) doesn’t hold, assuming Thomae’s function so that gcd (2 3,1) = 1 3.)

We may prove Law C.

Proposition. a b(modm) if and only if an bn(modmn), when n0.

Proof. Let a, b, m, and n be arbitrary real numbers such that n0. We must show that a b(modm) if and only if an bn(modmn).

By exercise 15, we have that

a b(modm) a mod m = b mod m n(a mod m) = n(b mod m) an mod mn = bn mod mn an bn(modmn)

as we needed to show. □

To see why Law D does not hold for the real numbers, consider as an example a = 1 2, b = 1, and r = s = 3 2. Then

1 2 1(mod 3 2), 1 2 1(mod 3 2)

but

1 2 1(mod 9 4).

25. [M02] Show that, according to Theorem F, ap1 mod p = [a is not a multiple of p], whenever p is a prime number.

Proposition. ap1 mod p = [a p] whenever p is a prime number.

Proof. Let a and p be arbitrary integers such that p is prime. We must show that ap1 mod p = [a p].

By Theorem F, we have that ap a(modp). In the case a p, we may apply Law B to deduce that ap1 1(modp), or equivalently, that

ap1 mod p = 1 mod p = 1 = [a p]

since p > 1.

In the case that a⊥̸p, ap, and since p > 1, we have that

ap1 mod p = 0 = 0 mod p = [a p].

Therefore, in either case,

ap1 mod p = [a p]

as we needed to show. □

26. [M15] Let p be an odd prime number, let a be any integer, and let b = a(p1)2. Show that b mod p is either 0 or 1 or p 1. [Hint: Consider (b + 1)(b 1).]

Proposition. a(p1)2 mod p is either 0, 1, or p 1.

Proof. Let a and p be arbitrary integers such that p is prime and let b = a(p1)2. We must show that b mod p is either 0, 1, or p 1.
If b⊥̸p, then clearly b mod p = 0. Otherwise, we need only examine the case b p. By Theorem F, we have that ap a(modp). We may apply Law B to deduce that ap1 1(modp), or equivalently by Law A, that

ap1 1 0(modp).

But

ap1 1 = b2 1 = (b + 1)(b 1)

so that

(b + 1)(b 1) 0(modp).

By canceling out each factor, we obtain

b + 1 0(modp) b 1(modp) b p 1(modp)

and

b 1 0(modp) b 1(modp).

Therefore, b mod p is either 0, 1, or p 1 as we needed to show. □

27. [M15] Given that n is a positive integer, let φ(n) be the number of values among {0,1,,n 1} that are relatively prime to n. Thus φ(1) = 1, φ(2) = 1, φ(3) = 2, φ(4) = 2, etc. Show that φ(p) = p 1 if p is a prime number; and evaluate φ(pe) when e is a positive integer.

We may prove a propery of Euler’s totient function, defined as

φ(n) = |{k : 1 k n k n}|.

Proposition. φ(p) = p 1 if p is a prime number.

Proof. Let p be an arbitrary prime number. We must show that φ(p) = p 1.

Since p is a prime number, its only divisors are 1 and p. That is, k p for 1 k < p but not for k = p. And so

φ(p) = |{k : 1 k p k p}| = |{k : 1 k p 1 k p}| + |{k : k = p k p}| = (p 1) + (0) = p 1

as we needed to show. □

We may evaluate φ(pe) when e is a positive integer by noting there are pe numbers in total, and of those, pe1 are multiples of p such that k⊥̸pe, 1 k pe. (Intuitively, these are k {p,2p,3p,,(pe1)p}.) This yields

φ(pe) = |{k : 1 k pe k pe}| = |{k : 1 k pe}||{k : 1 k pe k⊥̸pe}| = pe pe1 = pe1(p 1) = pe(1 1 p).

28. [M25] Show that the method used to prove Theorem F can be used to prove the following extension, called Euler’s theorem: aφ(m) 1(modulo m), for any positive integer m, when a m. (In particular, the number n in exercise 19 may be taken to be nφ(m)1 mod m.)

Proposition (Euler’s theorem). aφ(m) 1(modm), for any integer m > 0, when a m.

Proof. Let a and m be arbitrary integers such that m > 0 and a m. We must show that aφ(m) 1(modm).

Since a m, we have that

a 1(modm).

Let x1,x2,,xφ(m) be the φ(m) integers such that xi p, 1 i φ(m). For each, we have that

xi xi(modm)

and by Law A, we can multiply each to obtain

axi xi(modm).

Similarly, we may multiply for 1 i φ(m) to obtain

1iφ(m)axi 1iφ(m)xi(modm)

or equivalently

aφ(m) 1iφ(m)xi 1iφ(m)xi(modm).

Finally, as each xi m, we may cancel out the products to obtain

aφ(m) 1(modm)

as we needed to show. □

29. [M22] A function f(n) of positive integers n is called multiplicative if f(rs) = f(r)f(s) whenever r s. Show that each of the following functions is multiplicative: (a) f(n) = nc, where c is any constant; (b) f(n) = [n is not divisible by k2 for any integer k > 1]; (c) f(n) = ck, where k is the number of distinct primes that divide n; (d) the product of any two multiplicative functions.

We may prove that the various functions are multiplicative.

Proposition (A). The function f(n) = nc, where c is any constant, is multiplicative.

Proof. Let f be a function such that f(n) = nc, where c is an arbitrary constant; and let r and s be arbitrary integers such that r s. We must show that f is multiplicative; that is, that f(rs) = f(r)f(s). But

f(rs) = (rs)c = rcsc = f(r)f(s)

as we needed to show. □

Proposition (B). The function f(n) = [k2 n] for any integer k > 1, is multiplicative.

Proof. Let f be a function such that f(n) = [k2 n] for any integer k > 1; and let r and s be arbitrary integers such that r s. We must show that f is multiplicative; that is, that f(rs) = f(r)f(s).

First, we have that k2 n for any integer k > 1. But we also have that n k2, since k2 being a multiple of n would require n to be a square, a possibility precluded by the fact that k2 n for any integer k > 1. Therefore, we have the stronger condition that n k2.

Then

f(rs) = [rs k2] = [k2 1(modrs)] = [k2 1(modr) k2 1(mods)] by Law D = [k2 1(modr)][k2 1(mods)] = [r k2][s k2] = f(r)f(s)

as we needed to show. □

Proposition (C). The function f(n) = ck, where c is any constant and k is the number of distinct primes that divide n, is multiplicative.

Proof. Let f be a function such that f(n) = ck, where c is any constant and k is the number of distinct primes that divide n; and let r and s be arbitrary integers such that r s. We must show that f is multiplicative; that is, that f(rs) = f(r)f(s).

Let kn denote the number of distinct primes that divide n. Since r s, the prime factors of r must be distinct from the prime factors of s, and so we must that have krs = kr + ks.

Then,

f(rs) = crsk = ckr+ks = ckr cks = f(r)f(s)

as we needed to show. □

Proposition (D). The function f(n) = g(n)h(n), where g and h are multiplicative functions, is multiplicative.

Proof. Let f be a function such that f(n) = g(n)h(n), where g and h are multiplicative functions; and let r and s be arbitrary integers such that r s. We must show that f is multiplicative; that is, that f(rs) = f(r)f(s). But

f(rs) = g(rs)h(rs) = g(r)g(s)h(r)h(s) = g(r)h(r)g(s)h(s) = f(r)f(s)

as we needed to show. □

30. [M30] Prove that the function φ(n) of exercise 27 is multiplicative. Using this fact, evaluate φ(1000000), and give a method for evaluating φ(n) in a simple way once n has been factored into primes.

We may prove that Euler’s totient function φ(n) is multiplicative.

Proposition. Euler’s totient function φ(n) is multiplicative.

Proof. Let

φ(n) = |{k : 1 k n k n}|

be Euler’s totient function; and r and s arbtrary integers such that r s. We must show that φ(rs) = φ(r)φ(s).

Let R = {r1,r2,,rφ(r)} be those integers such that rk r, 1 k r; and S = {s1,s2,,sφ(s)} be those integers such that sk s, 1 k s; and define

T = {ti = rs + sr : r R s S}.

T has the following properties:

1.
ti rs. Suppose not. That is, suppose that gcd (rs + sr,rs) > 1 had a prime divisor p. Then, without loss of generality, p divides both r and not s (since r s) and r. But r r, so p = 1, which is a contradiction, similarly for the case p divides s and not r, and so, ti rs.
2.
titj(modrs). Suppose rs + sr rs + sr(modrs). Then rs divides r(s s) + s(r r) and s divides r(s s). But r s, so s divides s s, or equivalently, s s(mods). As s and s are both relatively prime and less than s, s = s. Similarly for rs dividing s(r r) to discover r = r. And so, no two titj(modrs).
3.
n rs ti n ti(modrs). Suppose n rs. Since r s, n can be expressed as n = rv + su for arbitrary integers u and v. We have u⊥̸r and v⊥̸s since rv + su rs and r s. Expressing u as u = r + ra and v as v = s + sb for arbitrary integers a and b, we have n = rv + su = r(s + sb) + s(r + ra) = rs + rsb + sr + sra = rs + sr + rs(a + b), or equivalently, n ti(modrs).

Since each ti rs, no two titj(modrs), and for any integer n rs there is a ti such that n ti(modrs), ij; we have that T = {t1,t2,,tφ(rs)}, those integers such that tk rs, 1 k rs; and that φ(rs) = |R||S| = φ(r)φ(s).

Then

φ(rs) = φ(r)φ(s)

as we needed to show. □

We can use this fact and the result of exercise 27, that φ(pe) = pe pe1 for any prime p, to evaluate φ(1000000) as

φ(1000000) = φ((26)(56)) = φ(26)φ(56) = (26 25)(56 55) = (32)(12500) = 400000.

A method for evaluating φ(z) in a simple way once z has been factored into n primes pi raised to powers wi as 1inpiwi is

φ(z) = 1inφ(piwi ) = 1inpiwi piwi1 = 1inpiwi 1in1 1 pi = z p|z p prime 1 1 p.

31. [M22] Prove that if f(n) is multiplicative, so is g(n) = dnf(d).

Proposition. If f(n) is multiplicative, so is g(n) = dnf(d).

Proof. Let f(n) be a multiplicative function; g(n) = d|nf(d); and r and s arbitrary integers such that r s. We must show that g(rs) = g(r)g(s). But since r s, the divisors d of rs may be partitioned into those that divide r and those that divide s; let these be denoted a and b such that a|r and b|s. Then, as a b since gcd (r,s) = gcd (a,s) = gcd (a,b) = 1,

g(rs) = d|rsf(d) = a|r b|s f(ab) = a|r b|s f(a)f(b) = a|rf(a) b|sf(b) = g(r)g(s).

32. [M18] Prove the double-summation identity

dn cdf(c,d) = cn d(nc)f(c,cd),

for any function f(x,y).

Proposition. d|n c|df(c,d) = c|n d|(nc)f(c,cd).

Proof. Let f(a,b) be an arbitrary function. We must show that

d|n c|df(c,d) = c|n d|(nc)f(c,cd).

But, since n = cdk if and only if n = ck nc = dk for arbitrary integers k and k,

d|n c|df(c,d) = d|nc|df(c,d) = cd|nc|cdf(c,cd) let d = cd = cd|nf(c,cd) = cd|ncd|nf(c,cd) = c|nd|(nc)f(c,cd) = c|n d|(nc)f(c,cd) = c|n d|(nc)f(c,cd)

as we needed to show. □

33. [M18] Given that m and n are integers, evaluate (a) 1 2(n + m) + 1 2(n m + 1); (b) 1 2(n + m) + 1 2(n m + 1). (The special case m = 0 is worth noting.)

We may evaluate the equations.

(a)
We want to evalute
1 2(n + m) + 1 2(n m + 1).

We consider two cases, depending on whether n ± m is odd or even.

Case 1. [n ± m odd] In the case that n ± m is odd,

1 2(n + m) + 1 2(n m + 1) = n + m 1 2 + n m + 1 2 = n + m 1 + n m + 1 2 = 2n 2 = n.

Case 2. [n ± m even] In the case that n ± m is even,

1 2(n + m) + 1 2(n m + 1) = n + m 2 + n m 2 = n + m + n m 2 = 2n 2 = n.

In either case, we have

1 2(n + m) + 1 2(n m + 1) = n.
(b)
We want to evaluate
1 2(n + m) + 1 2(n m + 1).

We consider two cases, depending on whether n ± m is odd or even.

Case 1. [n ± m odd] In the case that n ± m is odd,

1 2(n + m) + 1 2(n m + 1) = n + m + 1 2 + n m + 1 2 = n + m + 1 + n m + 1 2 = 2n + 2 2 = n + 1.

Case 2. [n ± m even] In the case that n ± m is even,

1 2(n + m) + 1 2(n m + 1) = n + m 2 + n m + 2 2 = n + m + n m + 2 2 = 2n + 2 2 = n + 1.

In either case, we have

1 2(n + m) + 1 2(n m + 1) = n + 1.

The special case m = 0 yields

n 2 + n + 1 2 = n

and

n 2 + n + 1 2 = n + 1.

34. [M21] What conditions on the real number b > 1 are necessary and sufficient to guarantee that log bx = log bx for all real x 1?

For real numbers x 1 and b > 1 we have for some arbitrary integer n that

log bx = n n log bx < n + 1 bn x < bn+1 bn x < bn+1 b n log b x < n + 1 log b x = n.

That is, b an integer such that b 2 are necessary and sufficient conditions to guarantee that

log bx = log bx

for all real x 1.

Note that we have the same conditions for ceilings, as

log bx = n n < log bx n + 1 bn < x bn+1 bn < x bn+1 b n < log b x n + 1 log b x = n.

35. [M20] Given that m and n are integers and n > 0, prove that

(x + m)n = (x + m)n

for all real x. (When m = 0, we have an important special case.) Does an analogous result hold for the ceiling function?

We may prove the result for floors.

Proposition. (x + m)n = (x + m)n for integers m, n, n > 0 and real x.

Proof. Let m and n be arbitrary integers such that n > 0, and x an arbitrary real number. We must show that

(x + m)n = (x + m)n.

But for an arbitary integer z

(x + m)n = z z (x + m)n < z + 1 nz (x + m) < n(z + 1) nz m x < n(z + 1) m nz m x < n(z + 1) m nz x + m < n(z + 1) z (x + m)n < z + 1 (x + m)n = z

as we needed to show. □

Note the important special case for m = 0,

xn = xn.

An analogous result holds for ceilings.

Proposition. (x + m)n = (x + m)n for integers m, n, n > 0 and real x.

Proof. Let m and n be arbitrary integers such that n > 0, and x an arbitrary real number. We must show that

(x + m)n = (x + m)n.

But for an arbitary integer z

(x + m)n = z z < (x + m)n z + 1 nz < (x + m) n(z + 1) nz m < x n(z + 1) m nz m < x n(z + 1) m nz < x + m n(z + 1) z < (x + m)n z + 1 (x + m)n = z

as we needed to show. □

36. [M23] Prove that k=1nk2 = n24; also evaulate k=1nk2.

We may prove the sum for floors.

Proposition. 1kn k 2 = n2 4 .

Proof. Let n be an arbitary positive integer. We must show that

1kn k 2 = n2 4 .

We consider two cases, depending on whether n = 2m is even or n = 2m + 1 is odd for an arbitrary integer m. We also will utilize the equality

1kn k 2 = 1kn n + 1 k 2 .

Case 1. [n = 2m] We have the equality

k 2 + n + 1 k 2 = m

since for an arbitrary integer q, if k = 2q is even

k 2 + n + 1 k 2 = 2q 2 + 2m + 1 2q 2 = m + 1 2 = m;

and if k = 2q + 1 is odd

k 2 + n + 1 k 2 = 2q + 1 2 + 2m + 1 2q 1 2 = 1 2 + m = m.

In this case then

1kn k 2 = 1 2 1kn k 2 + n + 1 k 2 = 1 2 1knm = nm 2 = n2 4 .

Case 2. [n = 2m + 1] In this case,

1kn k 2 = 1k2m k 2 + n 2 = (2m)2 4 + m = m2 + m n2 4 1 4.

And so, in either case,

(n 1)2 4 < n2 4 1 4 1kn k 2 n2 4

or equivalently,

1kn k 2 = n2 4

as we needed to show. □

For the second sum, we may prove another result.

Proposition. 1kn k 2 = n(n+2) 4 .

Proof. Let n be an arbitary positive integer. We must show that

1kn k 2 = n(n + 2) 4 .

We consider two cases, depending on whether n = 2m is even or n = 2m + 1 is odd for an arbitrary integer m. We also will utilize the equality

1kn k 2 = 1kn n + 1 k 2 .

Case 1. [n = 2m] We have the equality

k 2 + n + 1 k 2 = m + 1

since for an arbitrary integer q, if k = 2q is even

k 2 + n + 1 k 2 = 2q 2 + 2m + 1 2q 2 = m + 1;

and if k = 2q + 1 is odd

k 2 + n + 1 k 2 = 2q + 1 2 + 2m + 1 2q 1 2 = m + 1.

In this case

1kn k 2 = 1 2 1kn k 2 + n + 1 k 2 = 1 2 1knm + 1 = n(m + 1) 2 = n((n2) + 1) 2 = n(n + 2) 4 .

Case 2. [n = 2m + 1] In this case

1kn k 2 = 1k2m k 2 + n 2 = 2m(2m + 2) 4 + m + 1 = (m + 1)2 (n + 1)2 4 .

And so, in either case,

n(n + 2) 4 1kn k 2 (n + 1)2 4 < (n + 1)(n + 3) 4

or equivalently,

1kn k 2 = n(n + 2) 4

as we needed to show. □

37. [M30] Let m and n be integers, n > 0. Show that

0k<n mk + x n = (m 1)(n 1) 2 + d 1 2 + dxd,

where d is the greatest common divisor of m and n, and x is any real number.

We may prove the result for floors.

Proposition. 0k<n mk+x n = (m1)(n1) 2 + d1 2 + dxd for d = gcd (m,n).

Proof. Let m, n, and d be arbitrary integers such that d = gcd (m,n) and x an arbitrary real number. We must show that

0k<n mk + x n = (m 1)(n 1) 2 + d 1 2 + d x d.

But since for an arbitrary real z, z = x + x mod 1 = x + {x},

0k<n mk + x n = 0k<nmk + x n 0k<n mk + x n .

We have that

0k<nmk + x n = m n 0k<nk + 1 n 0k<nx = m n (n 1)n 2 + 1 nnx = m(n 1) 2 + x.

And so

0k<n mk + x n = m(n 1) 2 + x 0k<n mk + x n .

Then, for n = nd and m = md, we have

0k<n mk + x n = 0k<n mk n + x n = 0i<d nij<n(i+1) mj n + x n = 0i<d 0j<n m(j + ni) n + x n = 0i<d 0j<n mj n + mi + x n = 0i<d 0j<n mj n + x n since mi 0(mod1) = d 0j<n mj n + x n = d 0j<n mj n + xd n = d 0j<n mj + xd n = d 0j<n mj + xd + {xd} n = d 0j<n mj n + xd n + {xd} n = d 0j<n j n + xd n + {xd} n since mjn jn(mod1) = d 0j<n j + xd n + {xd} n = d 0j<n j n + {xd} n + 0j<n xd n = d 0j<n j n + {xd} n + nxd n = d 0j<n j n + {xd} n + xd = d 0j<n j n + {xd} n + 0 since xd 0(mod1) = d 0j<n j n + {xd} n = d 0j<n j n + {xd} n since 0 j + {xd} < n = d 0j<nj + {xd} n = d(n 1)n 2n + dn n {xd} = n d 2 + dn n (xd xd) = n d 2 + x dxd.

For justifications of certain steps, note that

m 0(mod1) i 0(mod1) mi 0(mod1);
m gcd (m,n)(mod n) m d(modn) md d(modnd) m 1(modn) mj j(modn) mjn jn(mod1);

and

0 j n 1 {xd} < 1 0 j + {xd} < n.

And so

0k<n mk + x n = m(n 1) 2 + x n d 2 x + dxd = mn m n + d 2 + dxd = (m 1)(n 1) 2 + d 1 2 + dxd

as we needed to show. □

For ceilings, note that by negating both sides of the equality yields

0k<n mk + x n = 0k<n mk + x n = 0k<n mk x n

and

(m 1)(n 1) 2 + d 1 2 + dxd = (m + 1)(n 1) 2 d 1 2 dxd = (m + 1)(n 1) 2 d 1 2 + dxd

or equivalently since x and m are arbitrary

0k<n mk + x n = (m + 1)(n 1) 2 d 1 2 + dxd.

38. [M26] (E. Busche, 1909.) Prove that, for all real x and y with y > 0,

0k<y x + k y = xy + x + 1(y y).

In particular, when y is a positive integer n, we have the important formula

x + x + 1 n + + x + n 1 n = nx.

Proposition. 0k<y x + k y = xy + x + 1(y y).

Proof. Let x and y be artbitrary real numbers such that y > 0. We must show that

0k<y x + k y = xy + x + 1(y y).

Let n be that unique integer such that n = (1 {x})y so that

n 1 < (1 {x})y n n 1 y < 1 {x} n y n 1 y < 1 + x x n y x + n 1 y < 1 + x x + n y x + n 1 y < x + 1 x + n y n 1 y < x + 1 x n y n 1 < (x + 1 x)y n n 1 < x + 1y xy n

or equivalently

n = x + 1y xy.

Note that for 0 k n 1 < n

x x + k y < x + 1 x + k y = x

and that for n k y 1 < y

x + 1 x + k y < x + 2 x + k y = x + 1.

Then

0k<y x + k y = 0ky1 x + k y = 0kn1 x + k y + nky1 x + k y = nx + ((y 1) n + 1)x + 1 = nx + (y n)x + 1 = nx + yx + 1 nx + 1 = yx + 1 n = x + 1yx + 1y xy = x + 1y + x + 1y + xy = xy + x + 1yx + 1y = xy + x + 1(y y)

as we needed to show. □

______________________________________________________________________________________________________________________________

[Crelle 136 (1909), 42; the case y = n is due to C. Hermite, Acta Math. 5 (1884), 315.]

39. [HM35] A function f for which f(x) + f(x + 1 n) + + f(x + n1 n ) = f(nx), whenever n is a positive integer, is called a replicative function. The previous exercise establishes the fact that x is replicative. Show that the following functions are replicative:

a)
f(x) = x 1 2;
b)
f(x) = [x is an integer];
c)
f(x) = [x is a positive integer];
d)
f(x) = [there exists a rational number r and an integer m such that x = rπ + m];
e)
three other functions like the one in (d), with r and/or m restricted to positive values;
f)
f(x) = log |2sin πx|, if the value f(x) = is allowed;
g)
the sum of any two replicative functions;
h)
a constant multiple of a replicative function;
i)
the function g(x) = f(x x), where f(x) is replicative.

We may prove that numerous functions are replicative.

Proposition (A). f(x) = x 1 2 is replicative.

Proof. Let f be a function defined as

f(x) = x 1 2

and let n be an arbitrary positive integer. We must show that

0kn1f x + k n = f(nx).

But

0kn1f x + k n = 0kn1x + k n 1 2 = 1 x 0kn11 + 1 n 0kn1k 1 2 0kn11 = n x + (n 1)n 2n n 2 = n x + n 1 2 n 2 = n x + n 1 n 2 = n x 1 2 = f(nx)

as we needed to show. □

Proposition (B). f(x) = [x mod 1 = 0] is replicative.

Proof. Let f be a function defined as

f(x) = [x mod 1 = 0]

and let n be an arbitrary positive integer. We must show that

0kn1f x + k n = f(nx).

But for some k = n nx mod n 1, 0 k n 1, we have

x = x + x mod 1 = nx n + nx mod n n = nx n + n k 1 n

if and only if x nk1 n is an integer, or equivalently, if and only if x + k n mod 1 = 0. In the case that such a k exists, then clearly nx and nx is an integer, or equivalently, nx mod 1 = 0, and

0kn1f x + k n = 0kn1 x + k n mod 1 = 0 = 0kn1 kk x + k n mod 1 = 0 + 0kn1 k=k x + k n mod 1 = 0 = 0 + 1 = 1 = [nx mod 1 = 0] = f(nx).

In the case that such a k does not exist, then clearly n x and nx is not an integer, or equivalently, nx mod 10, and

0kn1f x + k n = 0kn1 x + k n mod 1 = 0 = 0 = [nx mod 1 = 0] = f(nx).

In either case, we find that

0kn1f x + k n = f(nx)

as we needed to show. □

Proposition (C). f(x) = [x mod 1 = 0 x > 0] is replicative.

Proof. Let f be a function defined as

f(x) = [x mod 1 = 0 x > 0] = [x > 0][x mod 1 = 0]

and let n be an arbitrary positive integer. We must show that

0kn1f x + k n = f(nx).

But similar to Proposition B, for some k = n nx mod n 1, 0 k n 1, we have

x = x + x mod 1 = nx n + nx mod n n = nx n + n k 1 n

if and only if x nk1 n is an integer, or equivalently, if and only if x + k n mod 1 = 0. In the case that such a k exists, then clearly nx and nx is an integer, or equivalently, nx mod 1 = 0, and

0kn1f x + k n = 0kn1 x + k n > 0 x + k n mod 1 = 0 = 0kn1 kk x + k n > 0 x + k n mod 1 = 0 + 0kn1 k=k x + k n > 0 x + k n mod 1 = 0 = 0kn1 kk x + k n > 0(0) + x + k n > 0 x + k n mod 1 = 0 x + k n mod 1 = 0 = 0 + x + k n > 0 x + k n mod 1 = 0 = x + k n > 0 x + k n mod 1 = 0 = x + k n 1 x + k n mod 1 = 0 = x > 0 x + k n mod 1 = 0 = nx > 0 x + k n mod 1 = 0 = [nx > 0] x + k n mod 1 = 0 = [nx > 0](1) = [nx > 0][nx mod 1 = 0] = f(nx)

since x 0 if and only if x + k n < 1, and so neither it nor any multiple a positive integer such as nx.

In the case that such a k does not exist, then clearly n x and nx is not an integer, or equivalently, nx mod 10, and

0kn1f x + k n = 0kn1 x + k n > 0 x + k n mod 1 = 0 = 0kn1 x + k n > 0(0) = 0 = [nx > 0](0) = [nx > 0][nx mod 1 = 0] = f(nx).

In either case, we find that

0kn1f x + k n = f(nx)

as we needed to show. □

Proposition (D). f(x) = [r ,m : x = rπ + m] is replicative.

Proof. Let f be a function defined as

f(x) = [r ,m : x = rπ + m]

and let n be an arbitrary positive integer. We must show that

0kn1f x + k n = f(nx).

In the case that we may find an integer k such that

x + k n = rπ + m

it must be unique since if we found an integer k such that x + k n = rπ + m, (r r)π can only be an integer if r = r and (m m) can only be a rational if m = m, and so k = k. Then,

x + k n = rπ + m nx + k = nrπ + nm nx = nrπ + nm k nx = rπ + m

for a rational r = nr and an integer m = nm k. Hence

0kn1f x + k n = 0kn1[r ,m : x + k n = rπ + m] = 0kn1 kk [r ,m : x + k n = rπ + m] + 0kn1 k=k [r ,m : x + k n = rπ + m] = 0kn1 kk (0) + [r ,m : x + k n = rπ + m] = 0 + 1 = 1 = [r ,m : nx = rπ + m] = f(nx).

In the case that we may not find such an integer k, even k = 0, then clearly xrπ + m if and only if nxrπ + m for r = nr and m = nm, and in such a case

0kn1f x + k n = 0kn1[r ,m : x + k n = rπ + m] = 0kn1(0) = 0 = [r ,m : nx = rπ + m] = f(nx).

In either case, we find that

0kn1f x + k n = f(nx)

as we needed to show. □

Proposition (E1). f(x) = [r +,m : x = rπ + m] is replicative.

Proof. Let f be a function defined as

f(x) = [r +,m : x = rπ + m]

and let n be an arbitrary positive integer. We must show that

0kn1f x + k n = f(nx).

But we may rely on the proof of Proposition D, and specifically, the determination of the rational r such that r = nr. r > 0 if and only if nr > 0, since n > 0. Hence the result. □

Proposition (E2). f(x) = [r ,m + : x = rπ + m] is replicative.

Proof. Let f be a function defined as

f(x) = [r ,m + : x = rπ + m]

and let n be an arbitrary positive integer. We must show that

0kn1f x + k n = f(nx).

But we may rely on the proof of Proposition D, and specifically, the determination of the integer m such that m = nm k. m > 0 if and only if nr > 0, since n > k 0. Hence the result. □

Proposition (E3). f(x) = [(r +,m + : x = rπ + m] is replicative.

Proof. Let f be a function defined as

f(x) = [r +,m + : x = rπ + m]

and let n be an arbitrary positive integer. We must show that

0kn1f x + k n = f(nx).

But we may rely on the proofs of Proposition E1 and Proposition E2, and specifically, the determination of both the rational r such that r = nr and the integer m such that m = nm k simultaneously. Hence the result. □

Proposition (F). f(x) = log |2sin πx| is replicative, if the value f(x) = is allowed.

Proof. Let f be a function defined as

f(x) = log |2sin πx|

and let n be an arbitrary positive integer. We must show that

0kn1f x + k n = f(nx)

allowing for f(x) = .

Note that we have the equalities

2sin 𝜃 = (ei𝜃 ei𝜃)i = (1 e2i𝜃)ei𝜃iπ2,
0kn1(1 e2iπ(x+kn)) = 1 e2iπnx,

and

0kn1eiπ(x(12)+(kn)) = eiπ(nx12).

Then

0kn1f x + k n = 0kn1 log 2sin π x + k n = log 0kn12sin π x + k n = log 0kn1(eiπx+ k n eiπx+ k n)i = log 0kn1(1 e2iπx+ k n)eiπx+ k niπ2 = log 0kn11 e2iπx+ k n 0kn1eiπx+ k niπ2 = log (1 e2iπnx)(eiπ(nx12)) = log (eiπnx eiπnx)i = log 2sin πnx

as we needed to show. □

Proposition (G). The sum of any two replicative functions is replicative.

Proof. Let f and g be replicative functions, h a function defined as

h(x) = f(x) + g(x),

and let n be an arbitrary positive integer. We must show that

0kn1h x + k n = h(nx).

But

0kn1h x + k n = 0kn1f x + k n + g x + k n = 0kn1f x + k n + 0kn1g x + k n = f(xn) + g(xn) = h(xn)

as we needed to show. □

Proposition (H). A constant multiple of a replicative function is replicative.

Proof. Let f be a replicative function, c an arbitrary real number, g a function defined as

g(x) = cf(x),

and let n be an arbitrary positive integer. We must show that

0kn1g x + k n = g(nx).

But

0kn1g x + k n = 0kn1cf x + k n = c 0kn1f x + k n = cf(xn) = g(xn)

as we needed to show. □

Proposition (I). The function g(x) = f(x x), where f(x) is replicative is itself replicative.

Proof. Let f be a replicative function, g a function defined as

g(x) = f(x x),

and let n be an arbitrary positive integer. We must show that

0kn1g x + k n = g(nx).

Let k be the unique integer such that

{x} + k 1 n < 1 {x} + k n

so that k = n(1 {x}). Then

0kn1g x + k n = 0kn1f x + k n = 0kk1f x + k n + kkn1f x + k n = 0kk1f {x} + k n + kkn1f {x} + k n 1 = 0kk1f {x} + k n + kkn1f {x} + k n n = nkkn1f {x} + k n + k n + 0knk1f {x} + k n + k n = 0kn1f {x} + k n n + k n = f n {x} + k n n = f n{x} + k n = f n{x} + n(1 {x}) n = f n{x} + n n{x} n = f n{x} + n{x} = f n{x}n{x} = f {n{x}} = f {nx nx} = f {nx nx} = f {nx} = f nx nx = g(nx)

as we needed to show. □

40. [HM46] Study the class of replicative functions; determine all replicative functions of a special type. For example, is the function in (a) of exercise 39 the only continuous replicative function? It may be interesting to study also the more general class of functions for which

f(x) + f x + 1 n + + f x + n 1 n = anf(nx) + bn.

Here an and bn are numbers that depend on n but not on x. Derivatives and (if bn = 0) integrals of these functions are of the same type. If we require that bn = 0, we have, for example, the Bernoulli polynomials, the trigonometric functions cot πx and csc 2πx, as well as Hurwitz’s generalized zeta function ζ(s,x) = k01(k + x)s for fixed s. With bn0 we have still other well-known functions, such as the psi function.

n.a.

________________________________________________________________________________

For further results see L. J. Mordell, J. London Math. Soc. 33 (1958), 371–375.; M. F. Yoder, Æquationes Mathematcæ 13 (1975), 251–261.

41. [M23] Let a1,a2,a3, be the sequence 1,2,2,3,3,3,4,4,4,4,; find an expression for an in terms of n, using the floor and/or ceiling function.

We want to find a kind of inverse of the sum of the first n integers. That is, we want to know an = k such that

1ik1i = (k 1)k 2 < n k(k + 1) 2 = 1iki.

Solving for k yields

k 1 < 8n + 1 1 2 k

or equivalently

k = an = 8n + 1 1 2 .

42. [M24] (a) Prove that

k=1na k = nan k=1n1k(a k+1 ak),       if n > 0.

(b) The preceding formula is useful for evaluating certain sums involving the floor function. Prove that, if b is an integer 2,

k=1nlog bk = (n + 1)log bn (blog bn+1 b)(b 1).

Proposition (A). 1knak = nan 1kn1k(ak+1 ak) if n > 0.

Proof. Let n be an arbitrary integer such that n > 0. We must show that

1knak = nan 1kn1k(ak+1 ak).

But

1knak = 1kn(1)ak = 1kn(k k + 1)ak = 1knkak (k 1)ak = 1kn(k 1)ak + 1knkak = (0)a1 2kn(k 1)ak + 1knkak = 2kn(k 1)ak + 1knkak = nan 2kn(k 1)ak + 1kn1kak = nan 1kn1kak+1 + 1kn1kak = nan 1kn1kak+1 kak = nan 1kn1k(ak+1 ak)

as we needed to show. □

Proposition (B). 1knlog bk = (n + 1)log bn (blog bn+1 b)(b 1) if b 2.

Proof. Let b be an arbitrary integer such that b 2. We must show that

1knlog bk = (n + 1)log bn (blog bn+1 b)(b 1).

But from Proposition A, we have

1knlog bk = 1kn1log bk + nknlog bk = 1kn1log bk + blog bnknlog bk = b0k<blog bnlog bk + blog bnknlog bk = 0j<log bn bjk<bj+1log bk + blog bnknlog bk = 0j<log bnj(bj+1 bj) + blog bnknlog bk = 0j<log bnj(bj+1 bj) + (n blog bn + 1)log bn = log bnblog bn 0jlog bn1j(bj+1 bj) + (n + 1)log bn = 0jlog bn1bj+1 + (n + 1)log bn = (n + 1)log bn 1jlog bnbj = (n + 1)log bn (blog bn+1 b)(b 1)

as we needed to show. □

43. [M23] Evaluate k=1nk.

We have by exercise 42

1knk = 1kn1k + nknk = 12k<n2k + n2knk = 1j<n j2k<(j+1)2k + n2knk = 1j<nj((j + 1)2 j2) + n2knk = 1j<nj((j + 1)2 j2) + (n n2 + 1)n = (n n2 + 1)n + 1j<nj((j + 1)2 j2) = (n + 1)nnn2 1j<nj((j + 1)2 j2) = (n + 1)n 1knk2 = (n + 1)nn(n + 1)(2n + 1) 6 = n(n + 1) (n + 1)(2n + 1) 6 = nn (n + 1)(2n + 1) 6 6 = nn (2n + 5)(n 1) 6 .

44. [M24] Show that k0 1j<b(n + jbk)bk+1 = n, if b and n are integers, n 0, and b 2. What is the value of this sum when n < 0?

We may prove the equality for nonnegative n.

Proposition. k0 1j<b n+jbk bk+1 = n if b and n are integers, n 0, and b 2.

Proof. Let b and n be arbitrary integers such that n 0 and b 2. We must show that

k0 1j<b n + jbk bk+1 = n.

By exercise 38 with x = n bk+1 and y = b we have

1j<b n + jbk bk+1 = 0j<bn + jbk bk+1 n bk+1 = 0j<b n bk+1 + j b n bk+1 = nb bk+1 + n bk+1 + 1(b b) n bk+1 = nb bk+1 n bk+1 = n bk n bk+1 .

In the case that n > 0, for some arbitrary k

n bk = 0 0 n bk < 1 n < bk log bn < log bbk log bn < k k log bn + 1.

Then

k0 1j<b n + jbk bk+1 = k0 n bk n bk+1 = 0klog bnn bk n bk+1 + klog bn+1 n bk n bk+1 = 0klog bnn bk n bk+1 + 0 = 0klog bnn bk 0klog bn n bk+1 = n b0 + 1klog bnn bk 1klog bn+1 n bk = n b0 + 1klog bnn bk 1klog bnn bk n blog bn+1 = n b0 + 0 n blog bn+1 = n 0 = n.

In the case that n = 0, then clearly

k0 1j<b 0 + jbk bk+1 = k0 0 bk 0 bk+1 = 0 = n.

Hence, in either case

k0 1j<b n + jbk bk+1 = n

as we needed to show. □

When n < 0, we may find an arbitrary k such that

n bk = 1 1 n bk < 0 0 < n bk 1 n bk log bn log bbk log bn k k log bn.

Then

k0 1j<b n + jbk bk+1 = k0 n bk n bk+1 = 0klog bn1 n bk n bk+1 + klog bnn bk n bk+1 = 0klog bn1 n bk n bk+1 + klog bn 1 + 1 = 0klog bn1 n bk n bk+1 + 0 = 0klog bn1 n bk 0klog bn1 n bk+1 = n b0 + 1klog bn1 n bk 1klog bnn bk = n b0 + 1klog bn1 n bk 1klog bn1 n bk n blog bn = n b0 + 0 n blog bn = n + 1.

45. [M28] The result of exercise 37 is somewhat surprising, since it implies that when m and n are positive integers

0k<n mk + x n = 0k<m nk + x m .

This “reciprocity relationship” is one of many similar formulas (see Section 3.3.3). Show that for any function f, we have

0j<nf mj n = 0r<m rn m (f(r 1) f(r)) + nf(m 1).

In particular, prove that

0j<nmjn + 1 k + 0j<m jn m j k 1 = nm k .

[Hint: Consider the change of variable r = mjn. Binomial coefficients m k are discussed in Section 1.2.6.]

Proposition. 0j<nf mj n = nf(m 1) + 0r<m rn m (f(r 1) f(r)) for any function f and positive integers m and n.

Proof. Let f be any function and m and n arbitrary positive integers. We must show that

0j<nf mj n = nf(m 1) + 0r<m rn m (f(r 1) f(r)).

Note that

mj n = r r mj n < r + 1 nr mj < n(r + 1) nr m j < n(r + 1) m nr m j < n(r + 1) m . nr m j < n(r + 1) m .

Then

0j<nf mj n = 0r<m nr m j<n(r+1) m f mj n = 0r<m nr m j<n(r+1) m f(r) = 0r<m n(r + 1) m nr m f(r) = 0r<m n(r + 1) m f(r) nr m f(r) = 0r<m n(r + 1) m f(r) 0r<m nr m f(r) = 1r<m+1 nr m f(r 1) 0r<m nr m f(r) = nm m f(m 1) + n(0) m f(0 1) + 1r<m nr m f(r 1) 0r<m nr m f(r) = nf(m 1) + 0r<m nr m f(r 1) 0r<m nr m f(r) = nf(m 1) + 0r<m nr m f(r 1) f(r)

as we needed to show. □

Proposition. 0j<nmjn+1 k + 0j<m jn m j k1 = nm k for any function f and positive integers k, m, and n.

Proof. Let f be any function and k, m, and n arbitrary positive integers. We must show that

0j<nmjn + 1 k + 0j<m jn m j k 1 = nm k .

But by the preceding proposition for

f(x) = x + 1 k

we have that

0j<n mjn + 1 k = nm k + 0r<m rn m r k r + 1 k ;

and

r + 1 k = r k + r k 1 r k r + 1 k = r k 1;

so

0j<n mjn + 1 k = nm k + 0r<m rn m r k r + 1 k = nm k 0r<m rn m r k 1 = nm k 0j<m jn m j k 1.

Hence

0j<nmjn + 1 k + 0j<m jn m j k 1 = nm k

as we needed to show. □

46. [M29] (General reciprocity law.) Extend the formula of exercise 45 to obtain an expression for 0j<αnf(mjn), where α is any positive real number.

For any positive real number α, since by the first proposition of exercise 45

mj n = r rn m j < (r + 1)n m ,

we have

0j<αnf mj n = 0r<αm rn m j<(r+1)n m f mj n = 0r<αm rn m j<(r+1)n m f(r) = 0r<αm (r + 1)n m rn m f(r) = 0r<αm (r + 1)n m f(r) rn m f(r) = 0r<αm (r + 1)n m f(r) 0r<αm rn m f(r) = 1r<αm+1 rn m f(r 1) 0r<αm rn m f(r) = αnm m f(αm 1) + (0)n m f(0 1) + 1r<αm rn m f(r 1) 0r<αm rn m f(r) = αnf(αm 1) + 0r<αm rn m f(r 1) 0r<αm rn m f(r) = αnf(αm 1) + 0r<αm rn m (f(r 1) f(r)).

47. [M31] When p is an odd prime number, the Legendre symbol q p is defined to be + 1, 0, or 1, depending on whether q(p1)2 mod p is 1, 0, or p 1. (Exercise 26 proves that these are the only possible values.)

a)
Given that q is not a multiple of p, show that the numbers
(1)2kqp(2kq mod p),       0 < k < p2,

are congruent in some order to the numbers 2,4,,p 1(modulo p). Hence q p = (1)σ where σ = 0k<p22kqp.

b)
Use the result of (a) to calculate 2 p.
c)
Given that q is odd, show that 0k<p22kqp 0k<p2kqp(modulo 2) unless q is a multiple of p. [Hint: Consider the quantity (p 1 2k)qp.]
d)
Use the general reciprocity formula of exercise 46 to obtain the law of quadratic reciprocity, q p p q = (1)(p1)(q1)4, given that p and q are distinct odd primes.

Answers to exercise 47 follow below.

(a)
We may prove the congruence relation in order to deduce that q p = (1) 0k<p22kqp if p⊥̸q, p an odd prime.

Proposition (A). 0<k<p2(1)2kqp(2kq mod p) 0<k<p22k(modp) if p⊥̸q, p an odd prime.

Proof. Let p and q be arbitrary integers such that p⊥̸q and p an odd prime. We must show that

0<k<p2(1)2kqp(2kq mod p) 0<k<p22k(modp).

Let k be an arbitrary integer such that 0 < k < p2. Then

2kq = p 2kq p + (2kq) mod p

or equivalently

(2kq mod p) 2kq(modp).

Since p⊥̸q, q 1(modp), and so

(2kq mod p) 2k(modp).

Also, in the case that 2kq p is even, then so is 2kq mod p, and

(1)2kqp(2kq mod p) 2kq mod p(modp) (1)2kqp(2kq mod p) 2kq(modp);

in the case that 2kq p is odd, then so is 2kq mod p since p is an odd prime, and

(1)2kqp(2kq mod p) 2kq p(modp) (1)2kqp(2kq mod p) 2kq(modp);

and in either case

(1)2kqp(2kq mod p) 2k(modp).

Hence,

0<k<p2(1)2kqp(2kq mod p) 0<k<p22k(modp)

as we needed to show. □

(b)
In order to calculate 2 p, we let q = 2 and p be an odd prime. Then
2 p = (1) 0k<p24kp

has solutions for p = 4n + 1 or 4n + 3 for some integer n; in particular, (1,1,1,1) for p (1,3,5,7)(mod8), respectively; or expressed as the formula

2 p = (1)p+2 4 .
(c)
We may show the relation.

Proposition (C). 0k<p22kqp 0k<p2kqp(mod2) if q odd and p⊥̸q.

Proof. Let p and q be arbitrary integers such that p is an odd prime, p⊥̸q, and q odd. We must show that

0k<p22kqp 0k<p2kqp.(mod2)

But

0k<p22kqp = 0k<p42kqp + p4k<p22kqp = 0k<p42kqp + 12<k(p2)42(p 1 2 k)qp = 0k<p42kqp + 0k<p42(p 1 2 k)qp = 0k<p42kqp + 0k<p4(p 1 2k)qp = 0k<p42kqp + 0k<p4q (2k + 1)qp = 0k<p42kqp + 0k<p4q 1 (2k + 1)qp.

Then, since p⊥̸q and q odd

0k<p22kqp 0k<p42kqp + 0k<p4q 1 (2k + 1)qp 0k<p42kqp + 0k<p4(2k + 1)qp 0k<p2kqp(mod2)

as we needed to show. □

(d)
We may use the general reciprocity formula of exercise 46 to obtain the law of quadratic reciprocity.

Proposition (D). q p p q = (1)(p1)(q1)4 if p and q are distinct odd primes.

Proof. Let p and q be arbitrary integers such that p and q are distinct odd primes. We must show that

q p p q = (1)(p1)(q1)4.

From exercise 46

0k<p2 kq p = p2(q2 1) + 0r<q2 pr q ((r 1) r) = (p + 1)(q 1) 4 0r<q2 pr q = (p + 1)(q 1) 4 q 1 2 0r<q2 pr q = (p 1)(q 1) 4 0r<q2 pr q .

Then, since

0k<p2 2kq p + 0k<q2 2kp q 0k<p2 kq p + 0k<q2 kp q (p 1)(q 1) 4 (mod2)

we have

q p p q = (1) 0k<p22kqp+ 0k<q22kpq = (1)(p1)(q1)4

as we needed to show. □

______________________________________________________________________________________________________________________________

The idea of this proof goes back to G. Eisenstein, Crelle 28 (1844), 246–248; Eisenstein also gave several other proofs of this and other reciprocity laws in the same volume.

48. [M26] Prove or disprove the following identities, for integers m and n:

(a) m + n 1 n = m n ;(b) n + 2 n25 3 = 8n + 24 25 .

Some but not all of the identities may be proven.

(a)
The identity
m + n 1 n = m n

may be disproven by counterexample. Let m = 0 and n = 1. Then

m + n 1 n = 0 + 1 1 1 = 2 1 = 2 = 2 = 2 0 = 0 = 0 1 = m n .

Note, however, that we may prove the identity in the case that n > 0.

Proposition (A). m+n1 n = m n if n > 0.

Proof. Let m and n be arbitrary integers such that n > 0. We must show that

m + n 1 n = m n .

But since n > 0 implies n1 n < 1,

f(x) = c(x) 1 m + n 1 n = m n + n 1 n = m n

as we needed to show. □

(b)
The second identity may be proven.

Proposition (B). n+2n25 3 = 8n+24 25 .

Proof. Let n be an arbitrary integer. We must show that

n + 2 n25 3 = 8n + 24 25 .

But

n + 2 n25 3 = n n25 3 = n + n25 3 = 24n25 3 = 8n 25 = 8n + 24 25

as we needed to show. □

49. [M30] Suppose the integer-valued function f(x) satisfies the two simple laws (i) f(x + 1) = f(x) + 1; (ii) f(x) = f(f(nx)n) for all positive integers n. Prove that either f(x) = x for all rational x, or f(x) = x for all rational x.

We may prove the result.

Proposition. If an integer-valued function f(x) satisfies f(x + 1) = f(x) + 1 and f(x) = f(f(nx)n) for all positive integers n, either (x ) f(x) = x or (x ) f(x) = x.

Proof. Let f be an integer-valued function such that

f(x + 1) = f(x) + 1

and

f(x) = f(f(nx)n).

We must show that either

(x ) f(x) = x

or

(x f(x) = x.

First, we consider the domain of integers. From

f(0) = f(f((1)0)1) = f(f(0))

we may deduce that f(0) = 0. Then, from the inductive hypotheses f(k) = k and f(k) = k for an arbitrary integer k 0, we are able to show that f(k + 1) = f(k) + 1 and f(1 k) = f(k) + 1 respectively, proving by mathematical induction that f(n) = n for all integers n.

Second, we consider the domain of rationals. If f 1 2 0, we have

f 1 2 = f 1 1 2f(12)f 1 2 1 2f(12) = f 1 1 2f(12)f 1 2 f 1 2 = f 1 1 2f(12)f f 1 2 f 1 2 = f(0) = 0;

also, if f 1 n1 = 0, we have

f 1 n 1 = f 1 nf n n 1 = f 1 nf 1 + 1 n 1 = f 1 n 1 + f 1 n 1 = f 1 n 1 + 0 = f 1 n = 0;

and furthermore, if 1 m < n, by induction on m and since mnm n 0,

f m n = f 1 nmf nmm n = f 1 nmf nmm + n n n = f 1 nmf nmm + n nmm n = f 1 nmf n + m nm nm n = f 1 nmf 1 + m n n m n m = f 1 nm 1 + f m n n m n m = f 1 nm 1 + f 1 n m nm n = f 1 nm 1 + 0 = f 1 nm = 0.

Therefore, in the case that f 1 2 0, (x ) f(x) = x.

On the other hand, in the case that f 1 2 > 0, we may define f(x) = f(x) so that f(x + 1) = f(x) + 1 and f(x) = f(f(nx)n), and

f1 2 = f1 2 1 + 1 = f1 2 1 + 1 = f 1 2 + 1 + 1 = 1 f 1 2 0.

Therefore, f(x) = f(x) = x = x; that is, in the case that f 1 2 > 0, (x ) f(x) = x.

Hence, either (x ) f(x) = x or (x ) f(x) = x, as we needed to show.

______________________________________________________________________________________________________________________________

[P. Eisele and K. P. Hadeler, AMM 97 (1990), 475–477.]

[G. Hamel, Math. Annalen 60 (1905), 459–462.]