Posted on

Table of Contents

Q1

Prove that k[x]k[x] has infinitely many irreducible polynomials.

Proof:

Consider [f1(x),f2(x),...,fn(x)][f_1(x),f_2(x),...,f_n(x)] with fn(x)f_n(x) as the largest irreducible polynomial.

Then we have a new polynomial f(x)=f1(x)f2(x)...fn(x)+1f(x)=f_1(x)f_2(x)...f_n(x)+1.

Obviously, f(x)f(x) have higher order and cannot be divided by any of the previous polynomials.

Q2

For prime p1,p2,p3,...,pnZp_1,p_2,p_3,...,p_n \in \Z, consider a set of rational number r=a/b,a,bZr=a/b,a,b\in\Z and ordpiaordpib\text{ord}_{p_i}a \geq \text{ord}_{p_i}b. Prove that this set is a ring, and p1,p2,...pnp_1,p_2,...p_n are the only primes.

Proof:

First, prove addition:

Consider r1=a/br_1 = a/b and r2=c/dr_2=c/d. We need to prove r=r1+r2r=r_1+r_2 is still in set: r=ab+cd=ad+bcbd r=\frac{a}{b}+\frac{c}{d} = \frac{ad+bc}{bd} That is to prove: ordpiad+bcmin(ordpiad,ordpibc)min(ordpia+ordpid,ordpic+ordpib) \text{ord}_{p_i}{ad+bc} \geq \min(\text{ord}_{p_i}{ad},\text{ord}_{p_i}{bc})\geq \min(\text{ord}_{p_i}{a}+\text{ord}_{p_i}{d},\text{ord}_{p_i}{c}+\text{ord}_{p_i}{b}) Since ordpiaordpib\text{ord}_{p_i}{a}\geq \text{ord}_{p_i}{b} and ordpicordpid\text{ord}_{p_i}{c}\geq \text{ord}_{p_i}{d}

We shall see: ordpiad+bc>ordpibd \text{ord}_{p_i}{ad+bc} > \text{ord}_{p_i}{bd} Then prove multiplication: r=r1r2=acbd r = r_1r_2 = \frac{ac}{bd}

ordpiac=ordpia+ordpicordpib+ordpidordpibd \text{ord}_{p_i}{ac} = \text{ord}_{p_i}{a}+\text{ord}_{p_i}{c} \geq \text{ord}_{p_i}{b} + \text{ord}_{p_i}{d} \geq \text{ord}_{p_i}{bd}

So it is a ring.

For a prime in this set, we have abcedf    abcd or abef\frac{a}{b}|\frac{ce}{df} \iff \frac{a}{b}|\frac{c}{d}\ \text{or}\ \frac{a}{b}|\frac{e}{f}.

So by all means, a,ba,b should both be prime if b1b\neq 1.

Then ordpia=1\text{ord}_{p_i}a = 1 when a=pia=p_i and so does b. If b1b\neq 1, ordbb=1>ordba=0\text{ord}_bb = 1 > \text{ord}_ba = 0 that will make a/ba/b not in this set. So all primes should be p/1p/1.

Q3

Prove there are infinitely many prime number.

Proof:

If there are finite many prime: p1,p2,p3,...,plp_1,p_2,p_3,...,p_l. Then for n=p1p2p3...pln=p_1p_2p_3...p_l, φ(n)=1\varphi(n)=1.

However: φ(n)=ni=1l(11/pi)=1 \varphi(n)=n\prod_{i=1}^l(1-1/p_i)=1 That is i=1l(11/pi)=i=1l1/pi \prod_{i=1}^l(1-1/p_i)=\prod_{i=1}^l 1/p_i That's a contradiction.

Q4

If a is a non-zero integer, then for m>nm>n that (a2n+1,a2m+1)=1 or 2(a^{2^n}+1,a^{2^m}+1)=1\ \text{or}\ 2.

Proof: a2m1=(a2m1+1)(a2m11)=(a2m1+1)(a2m2+1)(a2m21)=...=...(a2n+1)(a2n+1) a^{2^m}-1 = (a^{2^{m-1}}+1)(a^{2^{m-1}}-1) = (a^{2^{m-1}}+1)(a^{2^{m-2}}+1)(a^{2^{m-2}}-1) = ...= ...(a^{2^{n}}+1)(a^{2^{n}}+1) So we know, if some prime pa2n+1p|a^{2^n}+1, then it must be pa2m1p|a^{2^m}-1.

Assume an integer c1c_1 such that c1p=a2n+1c_1p=a^{2^n}+1

That is, a2m+1=c2p+2a^{2^m}+1=c_2p+2.

So our question is actually (c2p+2,c1p)(c_2p+2,c_1p)

By the Euclidean algorithm, that is (cp,2)(cp,2)

So when cpcp is odd, that is, a2ma^{2^m} is even, meaning aa is even, GCD=1GCD=1.

When cpcp is even, that is, aa is odd, GCD=2GCD=2.

Q5

Prove there are infinitely many primes.

Proof:

Consider 22n+12^{2^n}+1; for a prime, there is only one n allowing it to p22n+1p|2^{2^n}+1.

So since nn are infinity many, there must be infinitely many primes.

Q7

Prove n!npn!p1/(p1)\sqrt[n]{n!}\leq\prod_{p|n!}p^{1/(p-1)}

Proof:

First: ordpn!=n/p+n/p2+n/p3+... \text{ord}_pn!=\lfloor n/p\rfloor+\lfloor n/p^2\rfloor+\lfloor n/p^3\rfloor+...

ordpn!n(1/p+1/p2+1/p3...)=n/(p1) \text{ord}_pn!\leq n(1/p+1/p^2+1/p^3...)=n/(p-1)

So, we have: n!pn!pn/(p1) n! \leq \prod_{p|n!}p^{n/(p-1)} That is: n!npn!p1/(p1) \sqrt[n]{n!}\leq\prod_{p|n!}p^{1/(p-1)}

Q8

Prove there are infinitely many primes.

Proof:

By Q7, the left side can go to infinity when nn\to \infin

So the right side should have infinity many terms, that is, infinitely many primes.

Q9

Show that a multiplicative function is completely determined by its value on prime powers.

Proof:

A function is multiplicative if f(ab)=f(a)f(b)f(ab)=f(a)f(b) whenever (a,b)=1(a,b)=1.

So: f(x)=f(ipiai) f(x)=f(\prod_i^{\infin}p_i^{a_i})

Q10

Prove that if f(x)f(x) is multiplicative, then g(x)=dxf(d)g(x)=\sum_{d|x}f(d) is also multiplicative.

Proof:

Consider n=abn=ab where a=pSapa=\prod_{p\in S_a}p and b=pSbpb=\prod_{p\in S_b}p . Here, Sa,SbS_a,S_b are two set of primes unique to each other, also in length of la,lbl_a,l_b.

Then: g(a)=i=0laf((lai))=i=0lap(lai)p g(a) = \sum_{i=0}^{l_a}f({l_a \choose i}) = \sum_{i=0}^{l_a}\prod_{p \in {l_a \choose i}} p

g(b)=i=0lbf((lbi))=i=0lbp(lbi)p g(b) = \sum_{i=0}^{l_b}f({l_b \choose i}) = \sum_{i=0}^{l_b}\prod_{p \in {l_b \choose i}} p

So g(a)g(b)=i=0la+lbf((la+lbi))=i=0la+lbp(la+lbi)p=g(ab) g(a)g(b) = \sum_{i=0}^{l_a+l_b}f({l_a+l_b \choose i}) = \sum_{i=0}^{l_a+l_b}\prod_{p \in {l_a+l_b \choose i}} p = g(ab)

Q11

Prove ϕ(n)=ndnμ(d)/d\phi(n)=n\sum_{d|n}\mu(d)/d .

Proof:

By Q9 and Q10, we can first prove μ(d)/d\mu(d)/d multiplicative.

By definition:

μ(ab)=(1)ν(ab)=(1)ν(a)+ν(b)\mu(ab) = (-1)^{\nu(ab)} = (-1)^{\nu(a)+\nu(b)} as along as (a,b)=1(a,b)=1.

So μ(ab)ab=μ(a)aμ(b)b \frac{\mu(ab)}{ab} = \frac{\mu(a)}{a}\frac{\mu(b)}{b} As same as in Theorem 2, f(x)=xf(x) = x is multiplicative: n=dnϕ(d)    ϕ(d)=dnμ(d)nd=ndnμ(d)/d n=\sum_{d|n}\phi(d) \implies \phi(d) = \sum_{d|n}\mu(d)\frac{n}{d} = n\sum_{d|n}\mu(d)/d

Q12

Find formulas for:

  • dnμ(d)ϕ(d)\sum_{d|n}\mu(d)\phi(d)
  • dnμ(d)2ϕ(d)2\sum_{d|n}\mu(d)^2\phi(d)^2
  • dnμ(d)/ϕ(d)\sum_{d|n}\mu(d)/\phi(d)

By Q11, they are all multiplicative: μ(ab)ϕ(ab)=(1)ν(ab)abpab(11/p)=μ(a)ϕ(a)μ(b)ϕ(b) \mu(ab)\phi(ab) = (-1)^{\nu(ab)}ab\prod_{p|ab}(1-1/p) = \mu(a)\phi(a)\mu(b)\phi(b) The rest two are the same.

So the result is determined by its primes powers (as in Q9)

Assume G1(n)=dnμ(d)ϕ(d)G_1(n)=\sum_{d|n}\mu(d)\phi(d), and n=i=1lpiain=\prod_{i=1}^{l}p_i^{a_i} G1(n)=G1(p1a1)G1(p2a2)...G1(plal)=i=1lμ(pi)ϕ(pi) G_1(n) = G_1(p_1^{a_1})G_1(p_2^{a^2})...G_1(p_l^{a^l})=\prod_{i=1}^{l}\mu(p_i)\phi(p_i)

G1(n)=(1p1)(1p2)(1p3)...(1pl) G_1(n)=(1-p_1)(1-p_2)(1-p_3)...(1-p_l)

Assume G2(n)=dnμ(d)2ϕ(d)2G_2(n)=\sum_{d|n}\mu(d)^2\phi(d)^2 G2(n)=i=1lμ(pi)2ϕ(pi)2=(p11)2(p21)2...(pl1)2 G_2(n) =\prod_{i=1}^{l}\mu(p_i)^2\phi(p_i)^2 = (p_1-1)^2(p_2-1)^2...(p_l-1)^2 Assume G3(n)=dnμ(d)/ϕ(d)G_3(n)=\sum_{d|n}\mu(d)/\phi(d) G3(n)=i=1lμ(pi)/ϕ(pi)=11p111p2...11pl G_3(n)=\prod_{i=1}^{l}\mu(p_i)/\phi(p_i)=\frac{1}{1-p_1}\frac{1}{1-p_2}...\frac{1}{1-p_l}

Q13

Show σk(n)=dndk\sigma_k(n)=\sum_{d|n}d^k formula and prove that it is multiplicative.

Proof:

First prove dkd^k is multiplicative. That (ab)k=akbk(ab)^k=a^kb^k. So σk(n)\sigma_k(n) is multiplicative.

So σk(n)=i=1lpi(ai+k)=ni=1lpik \sigma_k(n)=\prod_{i=1}^lp_i^{({a_i}+k)}=n\prod_{i=1}^lp_i^k

Q15

Prove that dnμ(n/d)ν(d)=1\sum_{d|n}\mu(n/d)\nu(d)=1 and dnμ(n/d)σ(d)=1\sum_{d|n}\mu(n/d)\sigma(d)=1

Proof:

By Q10 use same way to prove that if f(n)f(n) is multiplicative, then h(n)=dnμ(n/d)f(d)h(n)=\sum_{d|n}\mu(n/d)f(d) is also multiplicative.

Then dnμ(n/d)ν(d)=i=1l(μ(piai/p1ai)ν(piai)+μ(piai/piai1)ν(piai1))=i=1l(ai+1ai)=1 \sum_{d|n}\mu(n/d)\nu(d)=\prod_{i=1}^{l}(\mu(p_i^{a_i}/p_1^{a_i})\nu(p_i^{a_i})+\mu(p_i^{a_i}/p_i^{a_i-1})\nu(p_i^{a_i-1})) = \prod_{i=1}^{l}(a_i+1-a_i) = 1

dnμ(n/d)σ(d)=i=1l(σ(piai)σ(piai1))=i=1lpiai+11piai+1pi1=i=1lpiai=n \sum_{d|n}\mu(n/d)\sigma(d)=\prod_{i=1}^{l}(\sigma(p_i^{a_i})-\sigma(p_i^{a_i-1}))=\prod_{i=1}^{l}\frac{p_i^{a_i+1}-1-p_i^{a_i}+1}{p_i-1}=\prod_{i=1}^{l}p_i^{a_i}=n

Q16

Prove that ν(n)\nu(n) is odd iff n is a square.

Proof:

For n=i=1lpiain=\prod_{i=1}^{l}p_i^{a_i} ν(n)=i=1l(ai+1) \nu(n)=\prod_{i=1}^{l}(a_i+1) If ν(n)\nu(n) is odd, then every term of (ai+1)(a_i+1) must be odd, that is, aia_i is even. So nn is a square.

Q17

Prove that σ(n)\sigma(n) is odd iff n is a square or twice a square.

Proof:

Still, assume n=i=1lpiain=\prod_{i=1}^{l}p_i^{a_i} σ(n)=i=1lbj=0aipibj \sigma(n)=\prod_{i=1}^{l}\sum_{b_j=0}^{a_i}p_i^{b_j} If ν(n)\nu(n) is odd, then every term shall be odd.

For each term, if p2p\neq 2, then pp must be odd. So there are ai+1a_i+1 odd number (pbp^b) summing up. To keep this term to be odd, ai+1a_i+1 must be odd, that is, aia_i is even, and n is a square.

If p=2p=2, this term must be odd, so nn could be twice a square.

Q18

Prove that ϕ(m)ϕ(n)=ϕ((m,n))ϕ([m,n])\phi(m)\phi(n)=\phi((m,n))\phi([m,n])

Proof:

Consider m=pm2pnpa,n=pmpn2pbm=p_m^2p_np_a, n=p_mp_n^2p_b. Therefore, (m,n)=pmpn,[m,n]=pm2pn2papb(m,n)=p_mp_n,[m,n]=p_m^2p_n^2p_ap_b. And pm,pn,pa,pbp_m,p_n,p_a,p_b are relatively prime. ϕ([m,n])=pmϕ(pm)pnϕ(pn)ϕ(pa)ϕ(pb) \phi([m,n]) =p_m\phi(p_m)p_n\phi(p_n)\phi(p_a)\phi(p_b)

ϕ((m,n))=ϕ(pmpn)=ϕ(pm)ϕ(pn) \phi((m,n))=\phi(p_mp_n)=\phi(p_m)\phi(p_n)

ϕ(m)=ϕ(pm2pnpa)=pmϕ(pm)ϕ(pn)ϕ(pa) \phi(m) = \phi(p_m^2p_np_a)=p_m\phi(p_m)\phi(p_n)\phi(p_a)

ϕ(n)=ϕ(pn2pmpa)=pnϕ(pm)ϕ(pn)ϕ(pa) \phi(n) = \phi(p_n^2p_mp_a)=p_n\phi(p_m)\phi(p_n)\phi(p_a)

lhs=ϕ(pm)2ϕ(pn)2ϕ(pa)ϕ(pb) \text{lhs} = \phi(p_m)^2\phi(p_n)^2\phi(p_a)\phi(p_b)

rhs=ϕ(pm)2ϕ(pn)2ϕ(pa)ϕ(pb) \text{rhs} = \phi(p_m)^2\phi(p_n)^2\phi(p_a)\phi(p_b)

Q19

Prove that ϕ(mn)ϕ((m,n))=(m,n)ϕ(m)ϕ(n)\phi(mn)\phi((m,n))=(m,n)\phi(m)\phi(n)

Proof: lhs=ϕ(pm3pn3papb)ϕ(pmpn)=pm2pn2ϕ(pm)2ϕ(pn)2ϕ(pa)ϕ(pb) \text{lhs} = \phi(p_m^3p_n^3p_ap_b)\phi(p_mp_n)=p_m^2p_n^2\phi(p_m)^2\phi(p_n)^2\phi(p_a)\phi(p_b)

rhs=pmpnpmϕ(pm)ϕ(pn)ϕ(pa)pnϕ(pm)ϕ(pn)ϕ(pa)=pm2pn2ϕ(pm)2ϕ(pn)2ϕ(pa)ϕ(pb) \text{rhs}=p_mp_np_m\phi(p_m)\phi(p_n)\phi(p_a)p_n\phi(p_m)\phi(p_n)\phi(p_a)=p_m^2p_n^2\phi(p_m)^2\phi(p_n)^2\phi(p_a)\phi(p_b)

So lhs=rhs\text{lhs}=\text{rhs}

Q20

Prove that dnd=nν(n)/2\prod_{d|n}d=n^{\nu(n)/2}

Proof:

For each dnd|n, there must be a n/dnn/d|n.

So there are ν(n)\nu(n) dividers, and there are ν(n)/2\nu(n)/2 pairs of dn/dd*n/d.

So the result is nν(n)/2n^{\nu(n)/2}

Q21

Define Von Mangoldt function Λ(n)=logp\Lambda(n)=\log{p} if n is a power of pp, otherwise 0. Prove that dnμ(n/d)logd=Λ(n)\sum_{d|n}\mu(n/d)\log{d}=\Lambda(n).

Proof:

First, assume n=i=1lpiain=\prod_{i=1}^{l}p_i^{a_i} dnΛ(d)=i=1la=1ailogpi=log(i=1lpiai)=log(n) \sum_{d|n}\Lambda(d)=\sum_{i=1}^{l}\sum_{a=1}^{a_i}\log{p_i}=\log{(\prod_{i=1}^{l}p_i^{a_i})}=\log(n) Then apply the Möbius inversion: Λ(n)=dnμ(n/d)logd \Lambda(n)=\sum_{d|n}\mu(n/d)\log{d}

Q22

Show that the sum of all integers tt such that 1tn1\leq t\leq n and (t,n)=1(t,n)=1 is 1/2nϕ(n)1/2 \cdot n\phi(n).

Proof:

For each (t,n)=1(t,n)=1, there shall be a (nt,n)=1(n-t,n)=1 and obviously 1ntn1 \leq n-t \leq n.

So we can pair up all relatively primes, the number of pair shall be 1/2ϕ(n)1/2 \phi(n) unless t=n/2t=n/2. And each pair, on average, shall be 1/2n1/2*n

So the sum of tt shall be 1/2nϕ(n)1/2*n\phi(n)

Q23

For a function f(x)f(x), define ψ(n)\psi(n) as the number of f(j),j=1,2,...,nf(j), j=1,2,...,n such that (f(j),n)=1(f(j),n)=1. Prove that ψ(n)\psi(n) is multiplicative and ψ(pt)=pt1ψ(p)\psi(p^t)=p^{t-1}\psi(p).

Proof:

As same as the proof for ϕ(n)\phi(n), we can consider a series f(1)n,f(2)n,...,f(n)n\frac{f(1)}{n},\frac{f(2)}{n},...,\frac{f(n)}{n}

We shall have same result: dnψ(d)=n \sum_{d|n}\psi(d)=n So by Möbius inverse: ψ(n)=dnμ(d)nd \psi(n)=\sum_{d|n}\mu(d)\frac{n}{d} That is ψ(n)/n\psi(n)/n is multiplicative. So ψ(n)\psi(n) is multiplicative by Q10.

For ψ(pt)\psi(p^t), the number of term remains dptd|p^t the same, so ψ(pt)=ptdpμ(p)=pt1ψ(p)\psi(p^t)=p^t\sum_{d|p}\mu(p)=p^{t-1}\psi(p)

Also, ψ(pa)=11/p=ψ(p) \psi(p^a)=1-1/p=\psi(p) So by multiplicative: ψ(n)n=i=1lψ(pai)piai=pnψ(p)p \frac{\psi(n)}{n}=\prod_{i=1}^{l}\frac{\psi(p^{a_i})}{p_i^{a_i}}=\prod_{p|n}\frac{\psi(p)}{p}

Q25

Riemann Zeta Function ζ(s)=n=11/ns\zeta(s)=\sum_{n=1}^\infin 1/n^s. Prove the formal identity (Euler's Identity) ζ(s)=p(1(1/ps))1\zeta(s)=\prod_{p}(1-(1/p^s))^{-1}

Proof:

Consider each n: n=ppk n=\prod_{p}p^k So the Zeta function actually are summing all different compositions of prime powers. ζ(s)=pk1pks \zeta(s)=\prod_p\sum_k\frac{1}{p^{ks}} For each prime term, this is a sum of a geometric series: k1pks=(11ps)1 \sum_k\frac{1}{p^{ks}}=(1-\frac{1}{p^s})^{-1} So: ζ(s)=p(1(1/ps))1 \zeta(s)=\prod_p(1-(1/p^s))^{-1}

Q26

Prove some identities.

  • ζ(s)1=n=1μ(n)ns\zeta(s)^{-1}=\sum_{n=1}^{\infin}\frac{\mu(n)}{n^s}
  • ζ(s)2=n=1ν(n)ns\zeta(s)^2=\sum_{n=1}^{\infin}\frac{\nu(n)}{n^s}
  • $ \zeta(s)\zeta(s-1)=\sum_{n=1}^{\infin}\frac{\sigma(n)}{n^s}$

Proof:

Assume two function: F(s)=n=1f(n)ns,G(s)=n=1g(n)ns F(s)=\sum_{n=1}^\infin\frac{f(n)}{n^s},G(s)=\sum_{n=1}^\infin\frac{g(n)}{n^s} Consider Dirichlet convolution: F(s)G(s)=n1m1f(n)g(m)(mn)s=mn=1fg(mn)(mn)s F(s)G(s)=\sum_{n-1}^{\infin}\sum_{m-1}^{\infin}\frac{f(n)g(m)}{(mn)^s}= \sum_{mn=1}^{\infin}\frac{f*g(mn)}{(mn)^s} Let f(s)=1f(s)=1 , then F(s)=n=11ns=ζ(s)F(s)=\sum_{n=1}^{\infin}\frac{1}{n^s}=\zeta(s).

Assume h(s)=1g(s)h(s)=1*g(s) H(s)=n=1h(n)ns=F(s)G(s)=ζ(s)n=1g(n)ns H(s)=\sum_{n=1}^{\infin}\frac{h(n)}{n^s}=F(s)G(s)=\zeta(s)\sum_{n=1}^\infin\frac{g(n)}{n^s}

h(n)=dng(d) h(n)=\sum_{d|n}g(d)

Let g(n)=μ(n)g(n)=\mu(n): h(1)=μ(1)=1 h(1)=\mu(1)=1

h(n)=dnμ(n)=0 h(n)=\sum_{d|n}\mu(n) = 0 So: H(s)=1=ζ(s)n=1μ(n)ns H(s)=1=\zeta(s)\sum_{n=1}^{\infin}\frac{\mu(n)}{n^s}

ζ(s)1=n=1μ(n)ns \zeta(s)^{-1}=\sum_{n=1}^{\infin}\frac{\mu(n)}{n^s}

Let g(n)=n0g(n)=n^0: h(n)=dnd0=ν(n) h(n)=\sum_{d|n}d^0=\nu(n) Now, G(s)=ζ(s)G(s)=\zeta(s) n=1ν(n)ns=ζ(s)2 \sum_{n=1}^{\infin}\frac{\nu(n)}{n^s} = \zeta(s)^2 This can be generalize to g(n)=nαg(n)=n^\alphah(n)=dndα=σα(n) h(n)=\sum_{d|n}d^\alpha = \sigma_\alpha(n) When α=1\alpha=1, its σ(n)\sigma(n)n=1σ(n)ns=ζ(s)ζ(s1) \sum_{n=1}^{\infin}\frac{\sigma(n)}{n^s} = \zeta(s)\zeta(s-1) More generally: n=1σα(n)ns=ζ(s)ζ(sα) \sum_{n=1}^{\infin}\frac{\sigma_\alpha(n)}{n^s} = \zeta(s)\zeta(s-\alpha)

Q27

Prove that 1/n\sum1/n where n is square free integers, diverges.

Proof:

The sum is actually: n=1μ(n)n>n=1μ(n)n=ζ(1)1 \sum_{n=1}^{\infin}\frac{|\mu(n)|}{n} > \sum_{n=1}^{\infin}\frac{\mu(n)}{n} =\zeta(1)^{-1}

ζ(1)1=p<n(1+1/p) \zeta(1)^{-1} = \prod_{p<n}(1+1/p)

Take log: logp<n(1+1/p)=p<nlog(1+1/p)<p<n1/p \log{\prod_{p<n}(1+1/p)}=\sum_{p<n}\log{(1+1/p)} < \sum_{p<n}1/p So it diverge.