Knapp Study Group

Week 1 Problems

Nekoma
June 13, 2023

1 Quick Checks
1. What is the GCD of 26 and 80? What about the GCD of 8x2 + 10x + 2 and 48x + 12 (which are
polynomials with integer coefficients)?

Solution: Easy method. 26 is a semi prime, 26 = 2 × 13. 2|80, 13 ∤ 80. Thus gcd(26, 80) = 2.

Less easy method.

80 = 26 × 3 + 2
26 = 2 × 13 + 0
and thus the gcd is 2.

For the polynomials, one can do long division, but that’s tedious. Observe that

8x2 + 10x + 2 = 2(x + 1)(4x + 1)

and
48x + 12 = 12(4x + 1)

Now 2(4x + 1) divides both, and the remains are x + 1 and 6 respectively, which have nothing in
common. Thus the gcd is 2(4x + 1).

1
2. For each of the following parts, answer the question “Does a divide b?”
a) a = 5, b = 10
b) a = 10, b = 5
c) a = 10, b = 10
d) a = 0, b = 0
e) a = 5, b = 0

Solution:

a) Yes
b) No
c) Yes
d) Yes

e) Yes

Page 2
3. What is φ(7), where φ is the Euler φ function?

Solution: φ(7) = 6 since 7 is a prime number, anything below it won’t have anything common with
it, and 1 has gcd 1 with everything by default.

Page 3
4. If x is a positive integer which gives a remainder of 2 when divided by 3 and a remainder of 2 when
divided by 4, what are the possibilities for what x can be?

Solution: We are given that x = 2 mod 3 and x = 2 mod 4. Then x = 4k + 2 for some k ∈ N0
which says that

4k + 2 = 2 mod 3 ⇒ 4k = 0 mod 3 ⇒ k = 0 mod 3 ⇒ k = 3ℓ

for some ℓ ∈ N0 . Plugging this in, we have x = 12ℓ + 2, and thus the possibilities for x are all those
numbers such that dividing by 12 leaves a remainder of 2.

Page 4
5. Let    
1 2 3 4 5 1 2 3 4 5
τ1 = , τ2 =
5 3 4 2 1 3 4 5 2 1
be permutations. Give a cycle decomposition of τ1 and find all σ such that σ ◦ τ1 = τ2 . Compute the
sign of τ1 .

Solution: The cycle decompositions are (following orbits):

τ1 = (15)(234), τ2 = (135)(24)

Now, if σ ◦ τ1 = τ2 , then σ = τ2 ◦ τ1−1 . We have

τ1−1 = ((15)(234))−1 = (51)(432) = (15)(243)

Thus
σ = (135)(24)(15)(243) = (1)(2)(345) = (345)

The sign of τ1 is (−1)2+3 = (−1)5 = −1

Page 5
6. Perform row reduction on the matrix
 
1 2 4 1 0 0
2 4 7 0 1 0
0 1 4 0 0 1

with real entries. List out the elementary matrices for each of the steps you perform. What is the inverse
of  
1 2 4
2 4 7?
0 1 4

Solution: For later purposes, we denote

 
1 2 4
A = 2 4 7
0 1 4

and  
1 0 0
I = 0 1 0
0 0 1

Then we are row reducing (A|I). We have

     
1 2 4 1 0 0 1 2 4 1 0 0 1 2 4 1 0 0
2 4 7 0 1 0 → 0 0 −1 −2 1 0 → 0 1 3 −2 1 1
0 1 4 0 0 1 0 1 4 0 0 1 0 1 4 0 0 1
   
1 0 −2 5 −2 2 1 0 0 9 −4 −2
→ 0 1 3 −2 1 1 → 0 1 0 −8 4 1
0 0 1 2 −1 0 0 0 1 2 −1 0

This is the row reduced matrix.

Observe that this is in the form (I|B) where

 
9 −4 −2
B = −8 4 1
2 −1 0

Now, we want to find the inverse of A. Does A even have a inverse? Yes, for we see that A was row
reduced to I, and hence it does have an inverse. Then there exists C (the inverse of A such that
AC = I. By the above row reduction,

AC = I ⇒ IC = B ⇒ C = B

and thus the inverse of A is  

9 −4 −2
B = −8 4 1
2 −1 0

Page 6
2 Problems

7. Prove by induction that if S is a set of positive integers with no least element, then it is empty (this is
called the well-ordering principle). Next, prove that there does not exist a sequence of positive integers
a1 , a2 , a3 , . . . such that a1 > a2 > a3 > . . . (this fact is used often enough that a proof invoking it is
called a proof by descent).

Solution: Proving the contrapositive seems to be simpler. We show that any non-empty set S of
positive integers (natural numbers henceforth) contains a least element.

To that end, if 1 ∈ S, then S has a least element since 1 is the smallest natural number. So we
assume that the statement holds whenever the set S contains a natural number between 1 and n.

Now suppose S contains a natural number between 1 and n + 1. If it has no number between
1 and n, then it S must contain n + 1 by the supposition, and since it contains no smaller numbers,
n + 1 is the least element in S. On the other hand, if S contains a number from between 1 and n,
then the inductive hypothesis kicks in and says that there is a least element in S.

Thus by induction, the proof is complete.

The question then asks us to show that there does not exist a strictly decreasing sequence of natural
numbers. Well if it did, then considering this as a set, this has a least element, say an . But then
since an+1 < an , an cannot be the least, and we have hit a contradiction.

Page 7
8. The trace of an n × n matrix A is defined as
n
X
Tr A = Akk .
k=1

Prove that for any two n × n matrices A, B with real entries,

Tr AB = Tr BA.

Solution: Lets see. First we have X

(AB)ij = Aik Bkj
k

Then

X
Tr AB = (AB)kk
k
XX
= Akl Blk
k l
XX
= Akl Blk
l k
XX
= Blk Akl
l k
X
= (BA)ll
l
= Tr BA

Page 8
9. If p is a polynomial with real coefficients (i.e. an element of R[x]) and z is a complex root of p (identifying p
with its corresponding polynomial in C[x]), then prove that z is as well. Assuming that every nonconstant
polynomial with real coefficients has a complex root, prove by strong induction that every nonconstant
polynomial with real coefficients can be written as a product of linear and quadratic factors.

Pn
Solution: Let z be a root of p(x) = i=0 ai xi = Then

n
X n
X n
X n
X n
X
0 = p(z) = ai z i ⇒ 0 = ai z i = ai z i = ai z i = ai z i = p(z̄)
i=0 i=0 i=0 i=0 i=0

and thus z̄ is also a root of p.

Now, to show that every nonconstant polynomial can be factored into real linear and quadratic
terms, we use strong induction as asked. We induct on the degree of the polynomial. Since it is
non-constant, the base case is for degree 1, a linear polynomial, say p(x) = ax + b, a, b ∈ R. This is
already just a real linear term, so the base case is true. Then let the induction hypothesis be that
the statement is true for polynomials upto degree d − 1, d ≥ 1.

Then for a degree d polynomial q(x), we start by noting that we have a complex root z (told to
assume in the question). If the root is purely real, then z ∈ R and (x − z) is a real linear factor of q.
Then q(x)/(x − z) is another polynomial with real coefficients that is of degree d − 1, and thus by the
induction hypothesis can be written as a product of real linear and quadratic terms. Multiplying
that by (x − z), we see that q(x) can also be written similarly.
If z is complex, then z̄ is also a root, and hence both (x − z) and (x − z̄) are factors of q, and thus so
is (x − z)(x + z̄) = (x2 − z z̄) = (x2 − |z|2 ) which is real quadratic. Then q(x)/(x2 − |z|2 ) is another
polynomial with real coefficients that is of degree d − 2, and thus by the induction hypothesis can
be written as a product of real linear and quadratic terms. Multiplying that by (x2 − |z|2 ), we see
that q(x) can also be written similarly.

Since in both cases, the statement holds for d when it holds for all natural numbers < d, thus by
strong induction, the statement holds for all natural numbers, and thus holds for all polynomials.

Page 9
10. Prove the binomial formula for matrices: if A, B ∈ Mn×n (R) (i.e., they are n × n matrices with real
entries) such that AB = BA, then
p
p
X p!
(A + B) = Ak B p−k .
(p − k)!k!
k=0

A more concise way to write this is that

p  
p
X p
(A + B) = Ak B p−k .
k
k=0

Solution: Oh beauty, here’s a combinatorial proof. The proof will go through for any commuting
elements x, y of a ring with identity. So I will show this for that. This is just a special case of that
with the ring being that of matrices with real entries.

To that end, we look at (x + y)p . We are multiplying (x + y) with itself p times. Any term in this
expansion has a choice of picking either a x or a y from each of the terms. Say we picked x from k
terms and y from the rest p − k terms. Since x and y commute, whatever the ordering of choices of
x and y, we can swap the elements to bring all the x at the beginning, so that the term looks like
lke xk y p−k . How many such terms are there? Exactly as many terms as there are ways to select
the k x terms, or equivalently, choosing k terms from the p terms (we choose x from these k terms).
p!
This can be done in kp = k!(p−k)!


ways, and thus in the expansion, the whole term reads
 
p k p−k
x y
k

Collecting all terms, we have

p  
X p k p−k
(x + y)p = x y
k
k=0

Page 10
11. Make sure you understand the proofs of the division algorithm (1.1) and the proof of correctness of
Euclid’s algorithm (1.2). (You do not have to write anything down for this part.) Next, prove that
computing the GCD of positive numbers a and b where a ≥ b using Euclid’s algorithm takes at most
C log a steps for some constant C, as long as a is sufficiently large. A step here means either addition,
subtraction, multiplication, or computing a division and remainder. We express this by saying that
Euclid’s algorithm takes O(log a) time. (Note: Interestingly, the worst-case scenario happens when a
and b are consecutive Fibonacci numbers!)

Solution: The note is the hint! We start with the following claim.
Claim. If Euclid’s algorithm takes n steps for the pair (a, b) with a > b, then a ≥ Fn+2 and b ≥ Fn+1
(where Fn is the n-th Fibonacci number, starting with F1 = 0, F2 = 1).

Proof. If n = 1, then, b | a. The smallest possible such pair is (2, 1) which exactly matches the
bounds F2 = 1, F3 = 2. So we can assume that this holds when it takes n − 1 steps.

For n steps, the first step is

a = bq + r
and then we calculate gcd(b, r) which takes n − 1 steps and hence by the induction hypothesis,
b ≥ Fn+1 and r ≥ Fn . Then
a = bq + r ≥ b + r ≥ Fn+1 + Fn = Fn+2
and thus the statement still holds.

By induction, the statement holds for any n ∈ N.

A second claim just for completion.

Claim. We have
ϕn = Fn+1 ϕ + Fn
where ϕ is the golden ratio. As a corollary, ϕn ≤ Fn+1

Proof. The easiest proof is by induction. But what if someone doesn’t know the relation already?
Here’s a proof for that.

We have ϕ2 = ϕ + 1. Thus any power of ϕ can be written as a linear term in ϕ. Let

ϕn = an ϕ + bn
Then
an+1 ϕ + bn+1 = ϕn+1 = ϕn ϕ = (an ϕ + bn )ϕ = an ϕ2 + bn ϕ = an (ϕ + 1) + bn ϕ = (an + bn )ϕ + an

which gives us
an+1 = an + bn
bn+1 = an
Using the fact that a1 = 1, a2 = 1 and b1 = 0, b1 = 1, we have
an = Fn+1 , bn = Fn
and hence the result follows. Since everything here is non-negative, we have

ϕn ≤ Fn+1

Page 11
With this in hand, we go back to analysing the number of steps. If the algorithm takes n steps for
(a, b), then b ≥ Fn+1 ≥ ϕn . But then

ln b
ln b ≥ n ln ϕ ⇒ n ≤ = O(ln b)
ln ϕ

Page 12