Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Sai Nallani — 16 March 2024 Ross Problemset

§1 Problem 1
A polynomial is integral when it has integer coefficients. The square of 2 is a solution to
the integral polynomial equation x2 −2 = 0. A number is rational when it can be expressed
as ab for integers a and b (with b ̸= 0). A number is irrational when it is not rational.

(a) Suppose c is a non-square integer. (That is, c ̸= n2 for any n.) Explain

why c is not rational. Similarly, if c is a non-cube integer, does it follow

that 3 c is irrational?
We will approach this by first looking at this lemma.

Lemma 1.1
a
A rational number q = b ∈ Q has an integral square (q 2 ∈ Z) if and only if
b | a (b divides a).
Proof. If b | a, we can rewrite a = bk for some integer k. Therefore, we have
that q = ab = k. Since q 2 = k 2 ∈ Z, we are done with this direction of the
proof.
 a 2  a 2
If q 2 = ∈ Z, we have that = n for some integer n. We will use
b b
the Fundamental Theorem of Arithmetic to proceed. Let us write a and b as
follows:
a = pa11 pa22 . . . pr ar
b = pb11 pb22 . . . pr br
The set {p1 , p2 , . . . , pr } is the union of all primes in the factorizations of a and
b. In other words, if p ∈ {p1 , p2 , . . . , pr }, then p | a or p | b. Furthermore, we
also say that ai , bi ≥ 0. Since we have that b2 | a2 , we must have that for all
1 ≤ i ≤ r =⇒ 2ai ≥ 2bi . Why? Assume that there exists an i such that
1 ≤ i ≤ r and 2ai < 2bi . This means that, by the properties of exponents,

a2 p2a1 2a2
1 p2 . . . pi
2ai . . . p 2ar
r p12a1 p2 2a2 . . . pr 2ar
= = =n
b2 p2b 1 2b2
1 p2 . . . pi
2bi . . . p 2br
r p2b1 2b2
1 p2 . . . pi
2bi −2ai . . . p 2br
r

But since 2bi −2ai > 0, we get that pi divides p2a1


1 p2 2a2 . . . pr
2ar (multiply both

sides by p2b1 2b2 2br ). By Euclid’s lemma, p should divide one of p2aj
1 p2 . . . pr i j
2aj
where 1 ≤ j ≤ r, j ̸= i. Again by Euclid’s lemma, since pi | pj = pj pj . . . pj ,
| {z }
2aj times
we have that pi | pj . However, this is a contradiction because pi ̸= pj and
pi ̸= 1, and pj is prime, meaning its divisors are only 1 and itself. Therefore,
our original assumption that 2ai < 2bi for some i must be false.
Since we know that 2ai ≥ 2bi , we have that ai ≥ bi for all 1 ≤ i ≤ r. By the
Fundamental Theorem of Arithmetic, b | a. Now, we are done.


Assume that c is a rational number. That is, assume
√ a
c=
b
a
Furthermore, we can add the restriction that b ∤ a. Otherwise, b is an integer and

1
Sai Nallani — 16 March 2024 Ross Problemset

c would not be a non-square integer. for integers a, b. Squaring yields


 a 2
c=
b
By our lemma, the rational number ab has an integral square. This means that b | a.
But this contradiction because then ab would be an integer, making c a perfect

square. Therefore, c must be irrational. We can generalize the second part of
Lemma 1.1 a bit more:

Proposition 1.2 (bn | an ⇐⇒ b | a)


We will prove that bn | an if and only if b | a.
Proof. Using the same reasoning as before, write the prime factorizations of a
and b:
a = pa11 pa22 . . . pr ar
b = pb11 pb22 . . . pr br
The set {p1 , p2 , . . . , pr } is the union of all primes in the factorizations of a and
b. In other words, if p ∈ {p1 , p2 , . . . , pr }, then p | a or p | b. Furthermore, we
also say that ai , bi ≥ 0.
Assume b | a. Then we have that ai ≥ bi for all 1 ≤ i ≤ r. Otherwise, similar to
the argument in Lemma 1.1, we will have a contradiction. Then, by multiplying
by n, we have that nai ≥ nbi for all 1 ≤ i ≤ r. Since this is exactly what it
means, in terms of prime factorizations, for bn to divide an , we are done.
Assuming bn | an , we have that nai ≥ nbi =⇒ ai ≥ bi . This is what it means,
in terms of prime factorizations, for b to divide a. Therefore, we are done.


3
Now, assume that c is rational. This means that
√3
a
c=q=
b
a
for some rational q = b ∈ Q. By taking the third power on both sides, we have that

a3
c= =⇒ b3 | a3
b3
Since b3 | a3 , we have that b | a. This means that there exists an integer k such
3
that bk = a. This implies that c = ab3 = k 3 . This is a contradiction because c

is a non-cube integer. Therefore, our original assumption is incorrect and 3 c is
irrational.
√ √
(b) Find an integral polynomial equation that has α = 3 + 5 as a solution.
Show that α is irrational.
A great way to get rid of radicals
√ is to multiply √by something called a conjugate.
√ √
The conjugate of x = a + b is x = a − b. Conjugates are nice because
x · x = a − b ∈ Z! Conjugates are helpful when trying to find integral polynomials.

For example, to get the integral polynomial that has a root of a, we can just
√ √
consider P√(x) =√(x − a)(x + a) = x2 − a. To get the integral polynomial with
root α = 3 + 5, we can play around with multiplying linear terms of x and
conjugates of α. √
(x − α)(x + α) = x2 − (8 + 2 15)

2
Sai Nallani — 16 March 2024 Ross Problemset

Seems like we don’t have an integral polynomial yet. What we can do is make our
algebra a bit easier by using the substitution u = x2 − 8. Now, we have that

(x − α)(x + α) = u + 2 15

Multiplying (u + 2 15) by some other term to yield an integral polynomial is a
bit easier than playing with two square roots. By our previous example, we can
consider √ √
(u + 2 15)(u − 2 15) = u2 − 60
Substituting back in for u = x2 − 8, we get the integral polynomial P (x) =
x4 + 16x2 + 4 having α as a root! Just as a recap,

P (x) = (x − α)(x + α)(x2 − α2 ) = x4 + 16x2 + 4


√ √
α is √
irrational.
√ Here’s why. Assume that α = 3 + 5 is rational. This means
that 3 + 5 = ab for some integers a and b. Then, by rearranging and squaring,
we get
 a √ 2 a2 a√
3= − 5 = 2 +5−2 5
b b b
a2 √
 
1
a (3 − 2 − 5) = 5
−2 b b
2b2 + a2 √
= 5
2ab

This is a contradiction because √we have 5 to be rational; since 5 is a non-square
integer, by our results from (a), 5 is irrational. Therefore, our original assumption
that α was rational was incorrect and α is irrational. We are done.

(c) Let a√and b be integers. Find an √integral polynomial equation that has √
√ √ √
a + b as a solution. Must a + b be irrational? If a ̸= b, must a − b
be irrational?
√ Using the same technique as for the previous question, consider,

for α = a + b, the polynomial:

P (x) = (x − α)(x + α) = x2 − α2

We know that α2 = a + b + 2 ab, so we can rewrite P (x) as

P (x) = x2 − (a + b) + 2 ab

Let us do the substitution u = x2 − (a + b); we have P (x) = u + 2 ab. How do we
get rid of the square-root? By using difference of squares to our advantage.
√ √ √
Q(x) = (u + 2 ab)(u − 2 ab) = u2 − (2 ab)2 = u2 − 4ab

Now, we can expand our integral polynomial u2 into x4 − 2x2 (a + b) + (a + b)2


Considering Q(x) again, we have

Q(x) = x4 − 2x2 (a + b) + (a + b)2 − 4ab =


√ √
(x − α)(x + α)(x2 − (a + b) + 2 ab)(x2 − (a + b) − 2 ab)
√ √
If a and b are perfect squares, a + b will√be rational. What if a and b are
√ n
non-square integers? Let us assume that a + b = m is rational, for some rational

3
Sai Nallani — 16 March 2024 Ross Problemset

n
number m ∈ Q, and non-square integers a, b. Naturally, we might try the same
process as for the previous question:
√ n √
b= − a
m
 n √ 2 a2 2a √
b= − a = 2 − a+a
m b b
An important property of the rational numbers are that they are closed under
addition and multiplication, which is essentially the same as saying two rational
numbers make another rational number when added or multiplied.
√ √ √ √
Let us analyze a − b now. If a and b are squares of integers, a − b will be
rational. Assuming that a and b are non-square integers,√we proceed similarly as
√ n
before. Assume, for the sake of contradiction, that a + b = m for integers n, m.
Consider the following manipulations,
√ √ n
a− b=
m
√ n √
a= + b
m
n2 n√
a= 2 =2 b+b
m m
n 2
Subtracting the rationals on the right side, take the rational such that q = a− m 2 −b.
2n
√ ′ m
We have that q = m b. Next, take the rational such that q = q · 2n . We have

that b = q ′ ∈ Q. This is a contradiction because √ b is non-square, and non-square

integers have irrational roots. Therefore, a − b must be irrational if a and b are
non-square.

Lemma 1.3 (Q is closed)


Consider the rational numbers q1 = ab11 and q2 = ab22 . Since q1 + q2 = a1 bb21+a
b2
2 b1
,
and the sum is a ratio of two integers, it is also a rational number. Furthermore,
q1 · q2 = ab11 ba22 , and the product is a ratio of two integers, showing that it is a
rational number.
Since the product and sum of two rational numbers is another rational number,
we have shown that Q is closed.

√ √
Why is this so helpful? We have already seen it in the question with 3+ 5; let’s
look at our equation for b again.

a2 2a √
b= − a+a
b2 b
a2 2a √
b−2
−a=− a
b b
Using the fact that the rationals are closed, we know that there is a rational q such
2
that q = b − ab2 − a.
2a √
q=− a
b
b √
q·− = a
2a

4
Sai Nallani — 16 March 2024 Ross Problemset

b
Since − 2a is a rational number, multiplying it by q also forms another rational

number. Let’s called that number q ′ . We have that a = q ′ ∈ Q. However, this
is a contradiction because a is non-square, so√its square root must be irrational!

Therefore, our original assumption that a + b is rational must’ve been false, so
it must be irrational.
√ √ √
(d) Formulate some generalizations. √As a starting
√ point,
√ is β = 3+ √5+ √7 irrational?

What about numbers like γ = 3 2 − 2 3 − 3 5 + 6 and δ = 3 5 − 2? We will
prove some basic lemmas.

Lemma 1.4
Let q ∈ Q and x ∈ R − Q where R − Q are the set of irrational numbers.
a) q + x is irrational
̸ 0.
Assume q + x is rational. Then, q + x = a/b for integers a, b with b =
This implies that x = a/b − q which implies that x is rational due to the
closure of the rational numbers. Hence, x + q is irrational.
b) qx is irrational
Assume qx is rational. Then, qx = a/b for integers a and b. Then,
x = a/b · 1/q, which implies that x is rational due to the closure of the
rationals. We have a contradiction; hence, qx must be irrational.

Let aP
1 , . . . , an be positive integers that are all non-square. Then, we conjecture

that ni=1 ai is irrational. We proved the cases 1 and 2, so now it is time to
consider the general case. Let us try using induction. We have the base cases
covered already,
Pn √ so assume for all n, and positive non-square integers a1 , . . . , an , we
have that i=1 ai is irrational. Consider the case n + 1. We need to prove that
n+1
X √
ai
i=1

is irrational. Assume that it is rational, and subtract n terms.


n
√ n X√
a1 = − ai
m
i=2
Pn √
Let x = i=2 ai . Squaring both sides yields

n2 n
a1 = − 2 x + x2
m2 m
n n 2
Despite the rationality of x2 , we get that m 2
2 − 2 m x + x is irrational. Therefore,
we have a contradiction because we have the integer a1 is irrational. Therefore, our
original assumption is wrong and the induction hypothesis is correct. We conclude
the sum of irrational square roots is also irratinoal.
Some questions I would like to prove, but could not are:
√ √ √
Question 1.5. Let αn = a1 + a2 + · · · + an . What is the least k such that a
polynomial of degree k has integer coefficients and αn as a root?

5
Sai Nallani — 16 March 2024 Ross Problemset

Question 1.6. When is


√ √ √
n1
a1 + n2
a2 + · · · + nk
ak

irrational?

6
Sai Nallani — 16 March 2024 Ross Problemset

§2 Problem 2

(a) Figure out how A’s magic trick works, and write up a clear, mathematical
explanation on how to perform it. Let us first imagine a diagram of
the transformations that our number goes through. Say we pick a non-
negative integer n, and we are playing the game with B. Here are the
transformations n goes through.
$    %
    3 3n
3n 3 3n 2 2
n → 3n → → →q=
2 2 2 9
Woah, that seems a little complicated, but let’s break it down. We start with the
non-negative integer n, and then we multiply it by three, then we divide by two
and round down (that is the purpose of the floor function). Then, we apply the
same multiplying by three, dividing by two, and flooring the number. Finally, we
take the quotient when our transformed number is divided by 9 by the final floor
at the end.
A gets three pieces of information; did we round the first time we divided by two,
did we round the second time we divided by two, and the final quotient. Let us
denote as follows:
We rounded = 1
We did not round = 0

Looking at the transformation diagram again, the bolded values are the information
that A gets.
$    %
       3 3n
3n 3 3n 2 2
n → 3n → , r1 → , r2 → q =
2 2 2 9

As shown above, r1 , r2 , q are the pieces of information that A gets. We know that
r1 , r2 = 0, 1, because they are the variables showing if we rounded or not, and q

7
Sai Nallani — 16 March 2024 Ross Problemset

is the final quotient that A gets. If A’s trick always works, there is a one-to-one
correspondence between (r1 , r2 , q) ⇐⇒ N0 . In other words, since A can figure out
what unique number you have based on those three bits of information, there is a
surjection between the two sets. In more fancy terms, there is a surjection f , which
is A’s magic number guessing function, such that

f : {0, 1} × {0, 1} × N0 → N0

where N0 is the set of all non-negative integers.


Why is it a surjection? First of all, since all non-negative numbers work, f can be
used to produce values in the entire codomain N0 . If there is a number n ∈ N0 such
that f (r1 , r2 , q) ̸= n for all r1 , r2 , q, then A’s trick doesn’t work for n. Therefore,
since there is no number in N0 such that the function doesn’t point to it at all
inputs, we know that it is surjective. We don’t know that it is an injection yet
because it could be the case that one of the variables does not matter. With our
new way of looking at the problem, we might think about what affects the order
which we round. Since the thing that causes the rounding is dividing by 2 twice,
it’s natural to think about the divisibility of our number n by 4. If our number is
divisible by four, we do not even need to round. So, it is natural to consider what
happens to the inputs when our number is n ≡ 0, 1, 2, 3 (mod 4), where (mod 4)
is kind of like asking ”how far is n from a multiple of four”. Let us analyze the
cases when n = 4k + q, where 0 ≤ q ≤ 3, for all integers k ∈ Z.
a) n = 4k The transformations are as follows:

4k → 12k → (⌊12k/2⌋ = 6k, 0) → (⌊18k/2⌋ = 9k, 0) → q = k

Note that we never have to round down, because 4k is divisible by two both
times.
b) n = 4k + 1 The transformations are as follows:

4k + 1 → 12k + 3 → (⌊(12k + 3)/2⌋ = ⌊6k + 1 + 1/2⌋ = 6k + 1, 1)

→ (⌊3/2(6k + 1)⌋ = ⌊(18k + 3)/2⌋ = ⌊9k + 1 + 1/2⌋ = 9k + 1, 1) → k


Note we rounded both times.
c) n = 4k + 2 The transformations are as follows:

4k + 2 → 12k + 6 → (⌊(12k + 6)/2⌋ = 6k + 3, 0)

→ (⌊18k + 9⌋ 2 = ⌊9k + 4 + 1/2⌋ = 9k + 4, 1) → k


We did not have to round the first time, but we rounded the second time.
This is because 4k + 2 = 2(2k + 1), and dividing by two yields us the number
2k + 1, which is odd.
d) n = 4k + 3 The transformations are as follows:

4k + 3 → 12k + 9 → (⌊(12k + 9)/2⌋ = ⌊6k + 4 + 1/2⌋ = 6k + 4, 1)

→ (⌊(18k + 12)/2⌋ = 9k + 6, 0) → k
We rounded the first time, but we did not have to round the second time.
Using the function f , A can use his trick.

8
Sai Nallani — 16 March 2024 Ross Problemset

The beauty of all the cases is that we can characterize what to do with the final
quotient k depending on our rounding information. The final quotient is always k,
so to recover the original number, we multiply by four and add the remainder. The
remainder depends on the rounding. For example,

f (0, 0, k) = 4k

This makes sense, because in the first case we did not round both time and we got
the final quotient to be k. So to recover the original number, we just multiply it by
four to get 4k. Similarly, we get the following equations:

f (1, 1, k) = 4k + 1

f (0, 1, k) = 4k + 2
f (1, 0, k) = 4k + 3
Let us see what happens in the 19 case. As shown in the original dialogue, we know
that B rounds the first time, but does not round the second time. Furthermore, the
final quotient is 4. Using A’s magical (just kidding, it is just modular arithmetic)
function, we can evaluate f (1, 0, 4) = 4(4) + 3 = 19.

(b) What variants of this trick can you come up with using the same
principles? Can you change the numbers 2, 3, and 9 in the trick,
or maybe the operations involved? What about the information you ask
for? A general idea to get a trick is to make sure you have enough information to
create a surjection between the natural numbers and the information we get. Let
us try to generalize the transformations.

n → an → ⌊an/b⌋ → a/b ⌊an/b⌋ → ⌊(a/b ⌊an/b⌋)/c⌋

Our example was just the case where a = 3, b = 2, c = 9. First, let us assume that
gcd(a, b) is coprime. This is to avoid some of the cases where we will not have to
round because an is divisble by b.
I believe that considering the cases for n (mod b2 ) would be effective for analyzing
other variants of the trick.

9
Sai Nallani — 16 March 2024 Ross Problemset

§3 Problem 3

I will start by elaborating on the structure of fashionable numbers. Generating a few


1
operation, we get 1, 12 , 23 , 67 , 34 , 35 , 11
6 3 12 5

numbers using the a+b , 4 , 13 , 6 , . . . . Since the one
and only starting number in the set of fashionable numbers is 1, it is clear that every
fashionable number n except 1 can be decomposed into fashionable numbers a and b such
1
that n = a+b . This is because we generate the fashionable numbers with 1. This also
implies that every fashionable number can be broken down to the starting number 1. For
example, the starting number 23 = 1+1 1 . Let’s quantify this a little more and give it
1+1
1
some more notion. Consider the sets A1 = {1} and Ai+1 = { a+b : a, b ∈ Ai } ∪ Ai . Note
that
A1 ⊂ A2 ⊂ A3 ⊂ ...
1
Since all a, b ∈ Ai can be composed with 1s, a+b ∈ Ai+1 can also be composed with all 1s.
Now, we might be interested in how complex the fraction is. We can define something
called depth. In layman’s terms, it is how deep down the denominators get. For example,
1
1 has a depth of 0; 1+1 = 1/2 has a depth of 1; and 1+1 1 has a depth of 2. This property
1+1
is really easy to identify with our notation of the fashionable numbers with the Ai ; the
depth of a fashionable number n is k − 1 where k is the least positive integer such that
n ∈ Ak . Finally, we can observe that

[
F = Ai
i=1

is the set of all fashionable numbers with the starting set {1}.

(a) Find the smallest and largest fashionable numbers, and prove that
they are the smallest and largest. By generating the fashionable numbers,
we conjecture that 1 is the largest fashionable number and 1/2 is the smallest.

10
Sai Nallani — 16 March 2024 Ross Problemset

We will prove every fashionable number is between 1/2 and 1. More specifically,
we will prove that for all i > 0 ∈ Z, all n ∈ Ai are 1/2 ≤ n ≤ 1. For the base case,
consider i = 1. A1 = {1}, and 1/2 ≤ 1 ≤ 1, so the base case is satisified. Assume
that for all i, n ∈ Ai =⇒ 1/2 ≤ n ≤ 1. Note that Ai+1 = {1/(a + b) : a, b ∈ Ai }.
Since (for all a, b ∈ Ai ) 1/2 ≤ a, b ≤ 1 implies that 1 ≤ a + b ≤ 2, taking the inverse
1
yields that 1/2 ≤ a+b ≤ 1. Since all n ∈ Ai+1 are in the form n = 1/(a + b) for
some a, b ∈ Ai , this means that all n ∈ Ai+1 are between 1 and 1/2. Since 1/2 and
1 are also fashionable numbers, we know that 1/2 and 1 are the least and greatest
fashionable numbers respectively.

(b) Which numbers are fashionable? Can you characterize the set of all
fashionable numbers? All numbers that can be represented as a complex fraction
with only 1s are fashionable. Most of my thinking for this question lies before part
(a) where I defined how fashionable numbers are generated.

Conjecture 3.1 (All rational numbers between 1/2 and 1 inclusive are fashionable.).
Running some tests on Jupyter notebook, I have not found a number between 1/2
and 1 such that it is not a fashionable number for sufficiently large i (for Ai ). If
q=m n ∈ [1/2, 1], then m/n = 1/(a + b) for some a, b ∈ F . Rearranging, we get that
n = m(a + b). Maybe using the fact that fashionable numbers are composed with a
bunch of 1s, we can show that this is true. I have yet to find a counter example,
causing me to conjecture this.

Furthermore, all fashionable numbers of depth 3 or lower can be represented by a


bunch of 1s and a finite set as follows:
1
1 1
1 1 + 1 1
a1 +a2 + a3 +a4 a5 +a6 + a7 +a8

where ai = 1, 2! This is a different way than using our previous method of taking
the union of a bunch of sets in order of depth. Instead of recusion on the sets, we
can instead use the following recursion to generate the fashionable number of depth
n.

11
Sai Nallani — 16 March 2024 Ross Problemset

Definition 3.2. Let Sn = {a1 , a2 , . . . , a2n−1 } (where ai = 1, 2) and let g : {0, 1} ×


S → P (S) where g(0, Sn ) = {a1 , a2 , . . . , a2n−2 } and g(1, Sn ) = {a2n−2 +1 , . . . , a2n−1 }
(P (S) is the power set of S, or the set of all subsets of S. Essentially, g splits Sn
into half. Finally, let
1
fn (Sn ) =
fn−1 (g(0, Sn )) + fn−1 (g(1, Sn ))

and f1 ({ak }) = a1k . This is a fancy way of making the fact that fashionable numbers
of depth n can be represented as a bunch of 1s.
This is a different way of generating fashionable numbers because we can iterate
n−1
over the 22 different possible sets of Sn .

(c) Word on the street is that in 2025, the fashion rules will be the same
except for ”1 is fashionable”. Explore how the set of fashionable numbers
changes if we begin
√ with a different starting set of numbers. For example,
what if 2 and 3 are the given fashionable numbers? Our existing structure
for fashionable numbers can be nicely extended to any starting set of fashionable
numbers! We just need to modify our A1 to equal the set of starting numbers
that we want. Generalizing the case we have thought about, when we only have 1
fashionable number in our set of fashionable numbers, it is fairly easy to set up a nice
structure for them! Let A1 = {a1 }, where a1 is the only starting fashionable number.
Then we have that A2 = {1/(2a1 )}, and continuing with the same construction of
fashionable numbers as before, we can take the grand union. Similar to how we got
all fashionable numbers to be decomposed into 1s, we can decompose all fashionable
numbers into a1 s (except the top numerator). For example, for a fashionable
number of depth 2 and starting number a1 , we can represent it as follows:
1
1 1
s1 +s2 + s3 +s4

where si = a1 or 2a1 . Similar to the recursive definition in Definition 3.2, we can


replace the 1s and 2s with a1 s and 2a1 s. However, I was not able to find a nice
pattern with bigger sets of starting numbers.

12

You might also like