Download as pdf or txt
Download as pdf or txt
You are on page 1of 35

Art of Problem Solving

Group Theory (3111)


Christian Hokaj

Friday
Jun 17, 2022 - Sep 16, 2022
7:30 - 9:30 PM ET (4:30 - 6:30 PM PT)

Overview
Week 12 (Sep 2) Class Transcript - Field Extensions
< Go back to the class overview page
Copyright © AoPS Incorporated. This page is copyrighted material. You can view and print this page for your own use, but you
cannot share the contents of this file with others.
Display all student messages • Show few student messages • Hide student messages
CH328 2022-09-02 19:30:18
Group Theory Week 12: Field Extensions

CH328 2022-09-02 19:30:51


Group Theory Week 12: Field Theory Week 2: Field Extensions

CH328 2022-09-02 19:30:58


Hi everyone!

ScaryDragon 2022-09-02 19:31:12


GREETINGS

Hulk23 2022-09-02 19:31:12


MOOO

HotSoup 2022-09-02 19:31:12


hello

MasterInTheMaking 2022-09-02 19:31:12


chocolate lava cakes!

just4fun 2022-09-02 19:31:12


hi

MathWizard20 2022-09-02 19:31:12


Hi!

CH328 2022-09-02 19:31:15


This week we will be studying polynomials, and how to use them to build fields.

CH328 2022-09-02 19:31:37


Now, polynomials don't form a field like we saw last week. So a lot of what we're doing today will really be ring theory in disguise.

CH328 2022-09-02 19:31:58


But you have to understand some ring theory to understand field theory. That's a theme which doesn't go away even if you get
deeper into fields than we will in this course.

ScaryDragon 2022-09-02 19:32:10


Why not call it ring theory week 1?

CH328 2022-09-02 19:32:43


Group Theory Week 12:Field Theory Week 2: Field Extensions Sort of Ring Theory? Week 1: Field Extensions

CH328 2022-09-02 19:32:50


Hulk23 2022-09-02 19:32:57
Looks right

ScaryDragon 2022-09-02 19:32:57

CH328 2022-09-02 19:33:00


A key thesis will be that the set of polynomials k[x] over a field behaves algebraically much like the set Z of integers.

CH328 2022-09-02 19:33:32


A lot of what we prove today will feel like stuff we already know over z!

CH328 2022-09-02 19:33:37


Z

CH328 2022-09-02 19:33:40


Let's recall that a field k is a set with two binary operations +, ⋅, such that (k, +) and (k ∖ {0}, ⋅) are groups, and such that the
operations interact via distributivity.

CH328 2022-09-02 19:33:44


We've actually seen a few examples of fields-from-polynomials already.

CH328 2022-09-02 19:33:55

Last week, we investigated the following fields, among others: Q[x], with x2 = 5x − 1, F3 [√2], and the non-field F7 [√2].

CH328 2022-09-02 19:34:02


What do all these fields have to do with polynomials?

Ilikeapos 2022-09-02 19:35:18


They are the "rings of remainders" when divided by a polynomial

shadowcharmer 2022-09-02 19:35:18


Each has a set k[x] of polynomials

The_goat 2022-09-02 19:35:18


They're built like a polynomial because we raise sqrt2 to powers and use F for coefficients

just4fun 2022-09-02 19:35:18


Root of polynomials give the additional degree of freedom?

CH328 2022-09-02 19:35:29


In all these cases, we started with a known field k, took a polynomial f with coefficients from k, and added to k a root of f !

CH328 2022-09-02 19:36:03


That gave the elements a very nice structure.

CH328 2022-09-02 19:36:17


In these cases above, we added a root to the polynomial x2 − 5x + 1, or x2 − 2, as the case may be.

CH328 2022-09-02 19:36:21


We saw another field, too, last week. The field with four elements (call it F4 ) had additive group C22 = ⟨1, a⟩, and multiplication
table as below:

CH328 2022-09-02 19:36:22

⋅ 0 1 a a + 1

0 0 0 0 0

1 0 1 a a + 1

a 0 a a + 1 1

a + 1 0 a + 1 1 a

CH328 2022-09-02 19:36:35


(Notice this is not Z/4Z, which isn't a field.)

CH328 2022-09-02 19:36:44


How did we view F4 in these terms of adding a new root to some polynomial with coefficients in a smaller field, too?

DottedCaculator 2022-09-02 19:37:34


a is the root of a 2
= a + 1

niheyang 2022-09-02 19:37:34


2
a = a + 1

Ilikeapos 2022-09-02 19:37:34


We took the ring of remainders of the polynomial x2 + x + 1 with all coefficients viewed modulo 2

niheyang 2022-09-02 19:37:34


a is the root of a2 − a − 1 = 0

CH328 2022-09-02 19:37:36


We didn't add a square root of an integer this time, but a can be characterized as a root of x2 − x − 1.

CH328 2022-09-02 19:37:42


Depending on how far you've gotten with last week's homework, you've seen even more examples, like Q[√7].
3

CH328 2022-09-02 19:38:13


(And how far you are in the textbook. By the way, if you like this stuff, you should definitely read the chapters that aren't a part of
the assigned reading. There's a lot of awesome stuff in there. But it's okay if you don't have time. That's why they aren't assigned.)

CH328 2022-09-02 19:38:19


And there's another familiar example of this process from algebra and precalculus - what is it, and what polynomial are we adding
a root to there?

Math5K 2022-09-02 19:39:20


creates x
2
x + 1 = 0 = i, −i

CH328 2022-09-02 19:39:22


Right!

CH328 2022-09-02 19:39:42


A big story in algebra and precalculus is that we don't have real solutions to x2 + 1 = 0.

CH328 2022-09-02 19:39:54


How do we handle that? We invent the number i and add it to R!

CH328 2022-09-02 19:40:00


And we get the complex numbers! We have C = R[√−1]. Here, we adjoin a root of x2 + 1 to R.

CH328 2022-09-02 19:40:08


So, we should be getting convinced that "adding a root to a polynomial" is almost our only idea about how to extend a field.

CH328 2022-09-02 19:40:15


Remember that a field K is an extension of k if k is a subfield of K; we write K/k and say "K over k."

CH328 2022-09-02 19:40:25


So field extensions, at least many of them, are all about solving polynomials.

CH328 2022-09-02 19:40:37


But we also saw last week that we can't just add whatever we want...

CH328 2022-09-02 19:40:50

Last week, we found that F3 [√2] is a field, but F7 [√2] is not. Can we explain why, in terms of the polynomials whose roots we're
adding?

Ilikeapos 2022-09-02 19:41:53


The polynomial x2 − 2 is not irreducible modulo 7
DottedCaculator 2022-09-02 19:41:53
x − 2 has a root in F7 but not in F3
2

YaoAOPS 2022-09-02 19:41:53


The polynomial can't have roots in the field

Math5K 2022-09-02 19:41:53


2 is a quadratic residue mod 7, but not mod 3

CH328 2022-09-02 19:41:55


In all both cases, the polynomial is x2 − 2.

CH328 2022-09-02 19:41:57


Over F7 , this polynomial already has roots, namely, 3 and 4!

CH328 2022-09-02 19:42:05


In F7 [x], we have x2 − 2 = (x − 3)(x − 4), as you can easily check.

CH328 2022-09-02 19:42:24


If this were a field, we would have to have √2 = 3 or √2 = 4, by the zero product property.

CH328 2022-09-02 19:42:35


While it's not a field, F7 [√2] is a perfectly decent object with an addition and multiplication (a ring).

CH328 2022-09-02 19:42:40


But it has the non-fieldish property that √2 − 3 and √2 − 4 are both nonzero, while their product is zero.

CH328 2022-09-02 19:42:48


Generally, if a, b ≠ 0 and ab = 0 in a ring, we call a and b zero divisors.

CH328 2022-09-02 19:43:12


And fields have no zero divisors.

CH328 2022-09-02 19:43:18


OK, I think we're ready to try to put this construction in a broader context. Any questions first?

CH328 2022-09-02 19:44:08


Alright, here comes the key construction, in general. Remember that, for any field k, we have the polynomial ring k[x].

CH328 2022-09-02 19:44:16


Suppose we have a polynomial f in k[x], and we want to extend k to a new field (or ring) K with a root
n
= a0 + a1 x + ⋯ + an x

of f , also called x.

CH328 2022-09-02 19:44:28


What new equation are we planning to require to hold in K?

HotSoup 2022-09-02 19:45:41


n
a0 + a1 x + ⋯ + an x = 0

Math5K 2022-09-02 19:45:41


n
a0 + a1 x+. . . +an x = 0

CH328 2022-09-02 19:45:44


We're planning to require a0 + a1 x + ⋯ + an xn = f = 0 in K. This is a straightforward generalization of what we did with
Q[x] satisfying x − 5x + 1 = 0.
2

CH328 2022-09-02 19:46:04


The only tricky thing is where I wrote f = 0 instead of f (x) = 0. The reason for this is simple but subtle: we can't actually think
about f as a function whose domain includes x, because x is just a symbol in k[x]!

CH328 2022-09-02 19:46:16


So we're really just setting this string of symbols, a0 + ⋯ + an x ,
n
equal to the identity of a certain group.

CH328 2022-09-02 19:46:34


Now for the big question: do we know a way to "impose equations" in a group?

mbbchess 2022-09-02 19:47:33


Presentation

shadowcharmer 2022-09-02 19:47:33


Yes, relators

HotSoup 2022-09-02 19:47:33


presentations

Math5K 2022-09-02 19:47:33


find a presentation of the group

CH328 2022-09-02 19:47:38


It was related to presentations!

CH328 2022-09-02 19:48:29


What did we do to a group without any relations to give it relations? We could just stick the relator into the presentation, but what
operation on the group did that correspond to?

Ilikeapos 2022-09-02 19:49:46


Quotienting it?

HotSoup 2022-09-02 19:49:46


quotienting

CH328 2022-09-02 19:49:52


Right, that's what quotients did!

CH328 2022-09-02 19:49:59


For any field k and any f ∈ k[x], we can construct a ring, k[x]/(f ), whose additive group is the quotient of k[x] by a certain
subset called (f ), and whose multiplication is "modulo f ."

CH328 2022-09-02 19:50:16


I used scare quotes because we haven't yet said what, exactly, "modulo f " means.

CH328 2022-09-02 19:50:22


In the process of figuring that out, we'll also figure out what subset we'll mean by (f )!

CH328 2022-09-02 19:50:32


Let's see if you can help me figure it out.

CH328 2022-09-02 19:50:39


The key analogy to keep in mind is that polynomials over a field behave a whole lot like integers in many ways. In particular,
constructing k[x]/(f ) is going to be a lot like constructing Z/nZ.

CH328 2022-09-02 19:50:47


With that in mind, when do you think two polynomials g, h ∈ k[x] are "congruent modulo f ?"

Math5K 2022-09-02 19:51:34


f |g − h

Ilikeapos 2022-09-02 19:51:34


When g − h is divisible by f in the polynomial sense

The_goat 2022-09-02 19:51:34


g - h = nf

MathWizard20 2022-09-02 19:51:34


g − h can be divide by f to get another polynomial.

CH328 2022-09-02 19:51:37


For integers, we have a ≡ b (mod n) if a − b is a multiple of n.

CH328 2022-09-02 19:51:42


We can use precisely that definition for polynomials! Say g ≡ h (mod f ) if g − h is a multiple of f , that is, when g − h = tf for
some polynomial t ∈ k[x].

CH328 2022-09-02 19:51:59


(Note, though, that it makes sense to let t be any polynomial, not just an element of k.)

CH328 2022-09-02 19:52:03


In particular, which polynomials are equivalent to 0, modulo f ?

The_goat 2022-09-02 19:53:07


Multiples of f

HotSoup 2022-09-02 19:53:07


tf for some polynomial t

Ilikeapos 2022-09-02 19:53:07


The polynomial multiples of f

CH328 2022-09-02 19:53:10


All the multiples of f are! (That is, polynomial multiples of f )

CH328 2022-09-02 19:53:18


So what subgroup are we quotienting k[x] by to get the additive group of k[x]/(f )?

CH328 2022-09-02 19:53:42


(Hint: it's not going to be just the subgroup generated by f ! Because then we only get k-multiples of f , not polynomial multiples of
f)

Ilikeapos 2022-09-02 19:54:28


Multiples of f

Math5K 2022-09-02 19:54:28


the polynomial multiples of f

The_goat 2022-09-02 19:54:28


0 residue class mod f

CH328 2022-09-02 19:54:31


We're quotienting by the subgroup (f ) = {gf ∣ g ∈ k[x]} formed by all multiples of f !

CH328 2022-09-02 19:54:53


So this is much bigger than the subgroup generated by f : it contains gf for every g ∈ k[x], not just nf for n ∈ Z and not just nf
for n ∈ k . Such a subset of a ring is called an ideal, although we'll rarely if ever use this word.

CH328 2022-09-02 19:55:09


Though maybe we should... ideals are very important in ring theory! Ideals are to ring theory what subgroups are to group theory.

CH328 2022-09-02 19:55:35


Let's get some practice working with these quotients.

CH328 2022-09-02 19:55:37


Find a representative of the lowest possible degree for the equivalence class of g = x
7
− x
5
+ 2x
3
− x + 1 in F5 [x]/(x2 − 3).

CH328 2022-09-02 19:55:43


What is a basic "move" we can make to reduce the powers of x we see in g?

Math5K 2022-09-02 19:56:31


replace x2 with 3

Ilikeapos 2022-09-02 19:56:31


Replace xk with 3xk−2

CH328 2022-09-02 19:56:34


Since x 2
− 3 = 0, that is, x2 = 3, we can replace any occurrence of x2 with 3.
CH328 2022-09-02 19:56:57
(Some of you have noticed that we could also use polynomial division... that's great, too! We'll get to that later. )

CH328 2022-09-02 19:56:59


In particular, what degree should our representative get down to?

Ilikeapos 2022-09-02 19:57:35


1

niheyang 2022-09-02 19:57:35


1

MasterInTheMaking 2022-09-02 19:57:35


1

The_goat 2022-09-02 19:57:35


Order 1

mbbchess 2022-09-02 19:57:35


1

HotSoup 2022-09-02 19:57:35


1?

shadowcharmer 2022-09-02 19:57:35


1 or 0

Hulk23 2022-09-02 19:57:35


1

The_goat 2022-09-02 19:57:35


Degree 1

Math5K 2022-09-02 19:57:35


1

just4fun 2022-09-02 19:57:35


1

CH328 2022-09-02 19:57:37


We see g will definitely be equivalent to some polynomial of degree at most 1!

CH328 2022-09-02 19:57:40


What's our answer?

2
g ≡ x + (mod x − 3)

Ilikeapos 2022-09-02 19:59:02


2
g ≡ 3x + 1 (mod x − 3)

niheyang 2022-09-02 19:59:02


2
g ≡ 23x + 1 (mod x − 3)

Hulk23 2022-09-02 19:59:02


2
g ≡ 23x + 1 (mod x − 3)

HotSoup 2022-09-02 19:59:02


2
g ≡ 23x + 1 (mod x − 3)

Math5K 2022-09-02 19:59:02


2
g ≡ 23x + 1 (mod x − 3)

MasterInTheMaking 2022-09-02 19:59:02


2
g ≡ 23x + 1 (mod x − 3)
The_goat 2022-09-02 19:59:02
2
g ≡ 3x + 1 (mod x − 3)

DottedCaculator 2022-09-02 19:59:02


2
g ≡ 23x + 1 (mod x − 3)

just4fun 2022-09-02 19:59:02


2
g ≡ 23x + 1 (mod x − 3)

CH328 2022-09-02 19:59:04


We get

3 2
g ≡ 3 ⋅ x − 3 ⋅ x + 6x − x + 1

≡ 23x + 1
2
≡ 3x + 1 (mod x − 3).

Wow, so g is just 3x + 1! Interesting.

CH328 2022-09-02 19:59:19


But we want k[x]/(f ) to be more than just a group!

CH328 2022-09-02 19:59:21


Why is multiplication in k[x]/(f ) well-defined?

CH328 2022-09-02 19:59:29


Let's start to think this through. What are the elements of k[x]/(f ), again?

Ilikeapos 2022-09-02 20:00:16


Cosets of remainders (upon polynomial division) modulo f

Math5K 2022-09-02 20:00:16


the residue classes mod the polynomial f

The_goat 2022-09-02 20:00:16


Residue classes mod f

MasterInTheMaking 2022-09-02 20:00:16


equivalence classes of polynomials defined by multiples of f ?

CH328 2022-09-02 20:00:18


They're the cosets g + (f ), for g ∈ k[x]. For simplicity, we'll also write [g] for g + (f ).

CH328 2022-09-02 20:00:24


Now, what do you think we'll propose for the multiplication operation on k[x]/(f )? That is, what should [g][h] be?

CH328 2022-09-02 20:00:30


(Don't think too hard... )

Hulk23 2022-09-02 20:01:14


[gh]

smileapple 2022-09-02 20:01:14


[gh]

niheyang 2022-09-02 20:01:14


[gh]

DottedCaculator 2022-09-02 20:01:16


[gh]

CH328 2022-09-02 20:01:17


Ideally, we want to define [g][h] = [gh] ! Or, in coset notation, (g + (f ))(h + (f )) = gh + (f ). That's a nice simple definition,
and it's exactly analogous to how we define the addition in the quotient group k[x]/(f ) : [g] + [h] = [g + h].

CH328 2022-09-02 20:01:31


Ha, Ideally. I didn't even notice I did that.

MasterInTheMaking 2022-09-02 20:01:42


ba dum tsss

CH328 2022-09-02 20:01:44


What do we have to check to see whether the definition [g][h] = [gh] actually gives a well-defined binary operation?

Math5K 2022-09-02 20:02:45


check that [g][h] can only reduce to one residue class

MasterInTheMaking 2022-09-02 20:02:45


any element in g multiplied by an element in h gives the same result

CH328 2022-09-02 20:02:48


We have to prove that, if [g] = [g ]

and [h] ′
= [h ], then [gh] = [g h ].
′ ′

CH328 2022-09-02 20:03:10


(Otherwise, our binary operation wouldn't be well-defined: it would associate more than one output with one pair of inputs!)

CH328 2022-09-02 20:03:22


We can make this more concrete by writing some of these things explicitly.

CH328 2022-09-02 20:03:26



[g] = [g ] means g = g

+ t1 f and similarly, we have h ′
= h + t2 f , for some t1 , t2 ∈ k[x].

CH328 2022-09-02 20:03:29


What do we do now?

Ilikeapos 2022-09-02 20:04:32


We can expand (g ′

+ t1 f )(h + t2 f )

Hulk23 2022-09-02 20:04:32


Multiply them

HotSoup 2022-09-02 20:04:32


multiply and show that g ′ h′ ≅ gh

smileapple 2022-09-02 20:04:32


′ ′ ′ ′ ′ ′
gh = (g + t1 f )(h + t2 f ) = g h + g h + (t1 + t2 + t1 t2 f )f

mbbchess 2022-09-02 20:04:32


Calculate [g ′ h′ ]

shadowcharmer 2022-09-02 20:04:32


Compute gh and g ′ h′

The_goat 2022-09-02 20:04:32


Multiply and then solve for h'g'

CH328 2022-09-02 20:04:34


We multiply!

CH328 2022-09-02 20:04:36


We have

′ ′ ′ ′ ′ ′ 2
gh = (g + t1 f )(h + t2 f ) = g h + t1 f h + t2 f g + (t1 + t2 )f .

How does this help?


Math5K 2022-09-02 20:05:24
Everything except g'h' is a multiple of f

just4fun 2022-09-02 20:05:24


You can divide with f.

smileapple 2022-09-02 20:05:24


reduces to g ′ h′ 'modulo' f

The_goat 2022-09-02 20:05:24


We see that it's the same residue class as gh cuz all the other terms are multiples of f

niheyang 2022-09-02 20:05:24


we can merge all the terms with f , and it equals g ′ h′ + t3 f .

Hulk23 2022-09-02 20:05:24


Combine the terms with f?

Ilikeapos 2022-09-02 20:05:24


Everything that's not g ′ h′ is a multiple of f

just4fun 2022-09-02 20:05:24


Can factor with f.

CH328 2022-09-02 20:05:26


Setting t3 = t1 h′ + t2 g ′ + (t1 + t2 )f , this shows gh ′ ′
= g h + t3 f , so gh ′
≡ g h

(mod f ), as desired!

CH328 2022-09-02 20:05:31


Also, does our proof of well-definedness remind you of anything?

HotSoup 2022-09-02 20:06:07


we did the same thing with Z/nZ

CH328 2022-09-02 20:06:17


It's like the proof that multiplication mod n works properly, from number theory.

Ilikeapos 2022-09-02 20:06:21


Normal subgroup

The_goat 2022-09-02 20:06:21


Normal subgroups

Math5K 2022-09-02 20:06:21


a similar proof for groups

CH328 2022-09-02 20:06:41


It's also similar to the proof that a quotient group is a group when we have a normal subroup.

CH328 2022-09-02 20:06:51


Both of these are actually just examples of a more general, abstract proof from ring theory.... But that's leave that to the side for
now.

CH328 2022-09-02 20:06:54


Next, let's see what this result will buy us. Last week and earlier, we were a bit nervous about commutativity and associativity of
multiplication, as well as distributivity, in some of the fields we were working with.

CH328 2022-09-02 20:07:02


Prove that k[x]/(f )'s multiplication is associative, commutative, and distributive.

CH328 2022-09-02 20:07:07


For instance, let's think about commutativity. How can we try to prove that [g][h] = [h][g] in k[x]/(f ), for all g, h ∈ k[x]?

Ilikeapos 2022-09-02 20:08:22


This is because polynomial multiplication is commutative and taking modulo f doesn't affect that
DottedCaculator 2022-09-02 20:08:22
multiplication of polynomials is commutative

CH328 2022-09-02 20:08:24


We could take representatives of each coset and multiply... but we don't even have to do that.

CH328 2022-09-02 20:08:29


By definition, [g][h] = [gh] = [hg] = [h][g].

CH328 2022-09-02 20:08:38


We know multiplication in k[x] is commutative!

CH328 2022-09-02 20:08:43


That is, k[x]/(f ) inherits commutativity of multiplication straight from k[x].

CH328 2022-09-02 20:08:52


Are the proofs of associativity or distributivity going to be any harder?

Math5K 2022-09-02 20:09:21


no

smileapple 2022-09-02 20:09:21


no

HotSoup 2022-09-02 20:09:21


dont think so

The_goat 2022-09-02 20:09:21


Shouldn't be

MathWizard20 2022-09-02 20:09:21


No?

Hulk23 2022-09-02 20:09:21


Nope

MasterInTheMaking 2022-09-02 20:09:24


not really I think

CH328 2022-09-02 20:09:25


No! Both these properties also follow directly from their analogues in k[x].

CH328 2022-09-02 20:09:28


So, we're now completely happy that k[x]/(f ) is a ring. Soon, we'll start to think about when it's a field!

CH328 2022-09-02 20:09:36


Next, our earlier examples gave an idea of how we can concretely find nice representatives for elements of the quotient.

CH328 2022-09-02 20:09:41


Let's describe the unique best representative of the equivalence class [g] ∈ k[x]/(f ), given some g ∈ k[x].

CH328 2022-09-02 20:10:17


This is like what you did on your last writing problem to show SO(2) ≅ SE(2)/T . But now with a different group.

CH328 2022-09-02 20:10:21


OK, so g is some polynomial, of degree m, say, and f is a polynomial of degree n by which we've quotiented.

CH328 2022-09-02 20:10:26


What kind of polynomial do we think will give the "best" representative of [g]?

HotSoup 2022-09-02 20:11:24


some polynomial of most degree n − 1

Ilikeapos 2022-09-02 20:11:24


The one with least degree
Math5K 2022-09-02 20:11:24
the one with deg(g) < deg(f)

skim146 2022-09-02 20:11:24


the polynomial of the least degree (also reduced mod n if the field has additive group Z/nZ)

The_goat 2022-09-02 20:11:24


remainder of g when divided by f?

MasterInTheMaking 2022-09-02 20:11:24


the one with lowest degree

DottedCaculator 2022-09-02 20:11:24


polynomial of degree <n

CH328 2022-09-02 20:11:27


In the examples we've had, including F4 , Q[x]/(x2 − 5x + 1), and so on, we've always worked with representatives of minimal
degree.

CH328 2022-09-02 20:11:29


We only need degrees less than n = deg(f ).

CH328 2022-09-02 20:11:35


In fact, we've essentially already proven that [g] has such a representative, by the algorithm from above: replace big powers with
smaller powers, using f , until you get stuck. And you get stuck when xn (or any larger power of x) doesn't appear in your
representative anymore.

CH328 2022-09-02 20:11:53


So [g] has a representative of degree less than n. But is this representative unique?

CH328 2022-09-02 20:12:26


(unqiue among representatives with degree less than n, I mean. )

Hulk23 2022-09-02 20:12:41


Yes

MasterInTheMaking 2022-09-02 20:12:41


hopefully yes

smileapple 2022-09-02 20:12:41


by the division algorithm for polynomials it should be

The_goat 2022-09-02 20:12:41


Should be by division lemma

CH328 2022-09-02 20:12:58


We sure hope so! And some of you have recognized that this is essentially just division!

CH328 2022-09-02 20:13:03


Suppose [g] = [h] and deg(h) < n.

CH328 2022-09-02 20:13:07


Since [g] = [h], we know g − h ∈ (f ), that is, g − h = tf for some t(x) ∈ k[x].

CH328 2022-09-02 20:13:13


Therefore, we have g = tf + h, with deg(h) < deg(f ) = n.

CH328 2022-09-02 20:13:24


If you didn't see the division at first... do you see it now?

shadowcharmer 2022-09-02 20:13:51


Yes

Hulk23 2022-09-02 20:13:51


yeah
CH328 2022-09-02 20:13:55

CH328 2022-09-02 20:14:03


This is just like our usual division theorem for integers... but for polynomials now!

CH328 2022-09-02 20:14:14


Prove that if f , g ∈ k[x] then there exist unique q, r ∈ k[x] such that g = qf + r and deg(r) < deg(f ).

CH328 2022-09-02 20:14:25


We already know that such a q and r exist. We have our "replace xn " algorithm for finding r, and it implies that there is some q
somewhere, though we don't keep track of it.

CH328 2022-09-02 20:14:40


Many of you earlier recognized that polynomial long division works, too!

CH328 2022-09-02 20:14:43


The same algorithm you're used to works over any field, and it produces precisely the desired q and r.

CH328 2022-09-02 20:14:55


However, as we hinted above, we now have two different algorithms for finding r. (And at least one for q.) It's not obvious that
these algorithms lead to the same result.

CH328 2022-09-02 20:15:00


How could we prove that the remainder r in the division theorem is unique?

HotSoup 2022-09-02 20:16:18


sps two remainders exist r1 and r2 and make a contradiction

The_goat 2022-09-02 20:16:18


Assume there's 2 and show they're the same

just4fun 2022-09-02 20:16:18


Start with defining generic polynomials with specified degrees.

MathWizard20 2022-09-02 20:16:18


We can use a similar proof as the Divison Therom form Integers.

CH328 2022-09-02 20:16:20


Frequently, to prove a certain thing is unique, a good approach is to assume we have two such things, and prove they must be
equal.

CH328 2022-09-02 20:16:23


So, suppose g = q ′ f + r′ and g = qf + r, where r, r′ satisfy the condition of the division theorem, so they have degree strictly
less than n = deg(f ).

CH328 2022-09-02 20:16:31


Let's subtract these two equations. We have

′ ′
0 = (q − q )f + (r − r ),

or (q − q ′ )f ′
= r − r.

CH328 2022-09-02 20:16:41


Let's look at degrees on both sides. We know deg(f ) = n, so what's the degree of (q − q ′ )f ?

DottedCaculator 2022-09-02 20:17:30


≥ n

HotSoup 2022-09-02 20:17:30


≥ n

smileapple 2022-09-02 20:17:30


≥ n

The_goat 2022-09-02 20:17:30


At least n

Leo.Euler 2022-09-02 20:17:30


the degree is more than n

MasterInTheMaking 2022-09-02 20:17:30


greater than or equal to n

CH328 2022-09-02 20:17:33


The degree of a product is the sum of the degrees. So deg((q − q ′ )f ) ′
= deg(q − q ) + n.

CH328 2022-09-02 20:17:39


What about the right-hand side? What can we say about deg(r − r′ )?

smileapple 2022-09-02 20:18:16


< n

Hulk23 2022-09-02 20:18:16


<n

MasterInTheMaking 2022-09-02 20:18:16


less than n

The_goat 2022-09-02 20:18:16


< n

Math5K 2022-09-02 20:18:16


less than n

YaoAOPS 2022-09-02 20:18:16


< n

Ilikeapos 2022-09-02 20:18:16


< n

CH328 2022-09-02 20:18:18


By assumptions, r and r′ have degree less than n, so we can say deg(r − r′ ) < n.

CH328 2022-09-02 20:18:24


So we have two equal polynomials, one of degree deg(q − q ′ ) + n and one of degree less than n.

CH328 2022-09-02 20:19:09


That's impossible! Unless q = q

, in which case deg(q − q ′ ) + n isn't actually at least n.

CH328 2022-09-02 20:19:26


Either way, we can conclude q − q ′ = 0 from the fact that deg(q − q ′ ) + n = deg(r − r ).

CH328 2022-09-02 20:19:30


This implies r = r′ since r − r′ = (q − q )f ,

so that's the proof!

CH328 2022-09-02 20:19:38


Alright, good work, team. In summary, we've proven that every element [g] of k[x]/(f ) may be uniquely represented by a
polynomial of degree less than deg(f ), namely, the remainder in g ÷ f . You can find that representative however you want!

CH328 2022-09-02 20:19:49


For instance:

CH328 2022-09-02 20:19:51


Compute the remainder when g(x) = x
7
− x
5
+ 2x
3
− x + 1 is divided (over Q) by f (x) = x
2
− 3 .

CH328 2022-09-02 20:19:59


We could do this the hard way by long division. What is the easy way?
just4fun 2022-09-02 20:20:41
Let x^2=3

MasterInTheMaking 2022-09-02 20:20:41


plug in 3 for x
2

HotSoup 2022-09-02 20:20:41


we just proved the answer is unique so use our replacing algorithm

Math5K 2022-09-02 20:20:41


we found the representative of [g] before

Math5K 2022-09-02 20:20:41


it was 23x + 1

CH328 2022-09-02 20:20:44


This is equivalent to what we did a few minutes ago! We know the remainder in g ÷ f is just the smallest-degree representative of
[g] in Q[x]/(x2 − 3).

CH328 2022-09-02 20:20:47


The remainder is 23x + 1, no long division necessary!

CH328 2022-09-02 20:20:51


Essentially, the process we used to find the remainder earlier is a bit easier than full long division because we don't bother with
finding the quotient.

CH328 2022-09-02 20:21:00


Let’s now try to figure out when k[x]/(f ) is a field, in general.

CH328 2022-09-02 20:21:21


We already know that it's a (commutative) ring (with an identity)... what's left?

YaoAOPS 2022-09-02 20:21:55


inverses

just4fun 2022-09-02 20:21:55


Inverse?

The_goat 2022-09-02 20:21:55


Inverses?

Ilikeapos 2022-09-02 20:21:55


Multiplicative inverses

skim146 2022-09-02 20:21:55


inverses

CH328 2022-09-02 20:21:57


We have to see when every nonzero element of k[x]/(f ) has a multiplicative inverse. Everything else here is done!

CH328 2022-09-02 20:21:59


Let's try an example, first.

CH328 2022-09-02 20:22:05


Let K = Q[x]/(x3 + x + 1). Does x have a reciprocal in K?

CH328 2022-09-02 20:22:14


What are we trying to do, concretely?

CH328 2022-09-02 20:22:34


(By the way, we'll define "reciprocal" as "multiplicative inverse" in this class; they'll mean the same thing. )

The_goat 2022-09-02 20:23:24


Find an element that multiplies to 1
Ilikeapos 2022-09-02 20:23:24
Find a polynomial f (x) such that xf (x) − 1 is divisible by x3 + x + 1

HotSoup 2022-09-02 20:23:24


find a polynomial t such that xt = 1 in K

skim146 2022-09-02 20:23:24


find another polynomial f such that xf = 1

niheyang 2022-09-02 20:23:24


find a polynomial, so it multiplied by x divided by x3 + x + 1 has remainder 1

Math5K 2022-09-02 20:23:24


find f(x) such that x
3
+ x + 1|f (x)x − 1

CH328 2022-09-02 20:23:26


We're trying to find f such that xf
3
∈ Q[x] ≡ 1 (mod x + x + 1).

CH328 2022-09-02 20:23:31


More concretely, that means there's some g ∈ Q[x] such that

3
xf + g(x + x + 1) = 1.

CH328 2022-09-02 20:23:43


Note this is an equation in good old Q[x], not in K (just yet) so things are familiar.

CH328 2022-09-02 20:23:46


Hmm. Does trying to solve this equation for f and g remind you of something?

HotSoup 2022-09-02 20:24:28


bezout's lemma

DottedCaculator 2022-09-02 20:24:28


euclidean algorithm, bezout

YaoAOPS 2022-09-02 20:24:28


bezouts

Ilikeapos 2022-09-02 20:24:28


Bezout's Lemma

CH328 2022-09-02 20:24:32


It reminds me a lot of trying to solve, say, 7x + 25y = 1 for x and y in Z!

CH328 2022-09-02 20:24:36


This is Bezout's lemma: such an equation can be solved because 7 and 25 are relatively prime.

CH328 2022-09-02 20:24:55


What do we do to realize Bezout's Lemma and find explicit x and y in Z?

YaoAOPS 2022-09-02 20:25:52


gcd?

Ilikeapos 2022-09-02 20:25:52


Euclidean Algorithm

HotSoup 2022-09-02 20:25:52


euclidean algortihm

CH328 2022-09-02 20:25:53


We use Euclid's algorithm for the gcd and then we reverse it! (Some people call this the "extended Euclidean Algorithm" when you
do the reversing, too.)

CH328 2022-09-02 20:26:02


Remember how that goes? We repeatedly divide with remainder:

CH328 2022-09-02 20:26:05

25 = 3 ⋅ 7 + 4

7 = 1 ⋅ 4 + 3

4 = 1 ⋅ 3 + 1

Reversing the work shows

1 = 4 − 1 ⋅ 3

= 4 − 1 ⋅ (7 − 1 ⋅ 4)

= 2 ⋅ 4 − 1 ⋅ 7

= 2 ⋅ (25 − 3 ⋅ 7) − 1 ⋅ 7

= 2 ⋅ 25 − 7 ⋅ 7.

CH328 2022-09-02 20:26:15


So, let's see whether we can pretend Q[x] is Z well enough to solve xf + g(x
3
+ x + 1) = 1 via Euclid's algorithm.

CH328 2022-09-02 20:26:24


What's the first step in Euclid's algorithm for x and x3 + x + 1?

Hulk23 2022-09-02 20:27:05


Divide x^3+x+1 by x

Ilikeapos 2022-09-02 20:27:05


3 2
x + x + 1 = x(x + 1) + 1

The_goat 2022-09-02 20:27:05


find remainder

YaoAOPS 2022-09-02 20:27:05


divide x3 + x + 1 by x?

Math5K 2022-09-02 20:27:05


3 2
x + x + 1 = x(x + 1) + 1

CH328 2022-09-02 20:27:08


In Euclid's algorithm, we divide the big thing by the small thing with remainder, then repeat until we get stuck.

CH328 2022-09-02 20:27:10


So, first, we divide x3 + x + 1 by x in Q[x].

CH328 2022-09-02 20:27:14


We have x3 + x + 1 = x(x2 + 1) + 1, so the quotient is x2 + 1 and the remainder is 1.

CH328 2022-09-02 20:27:53


What do we learn about gcd(x, x3 + x + 1) from this?

YaoAOPS 2022-09-02 20:28:15


gcd is 1

The_goat 2022-09-02 20:28:15


It's 1

HotSoup 2022-09-02 20:28:15


3
gcd(x, x + x + 1) = 1

romans58 2022-09-02 20:28:15


It equals 1

MasterInTheMaking 2022-09-02 20:28:15


its 1

DottedCaculator 2022-09-02 20:28:15


3
gcd(x, x + x + 1) = 1

MathWizard20 2022-09-02 20:28:15


3
gcd(x, x + x + 1) = 1

CH328 2022-09-02 20:28:22


The remainder is 1, so the GCD is 1.

CH328 2022-09-02 20:28:25


Incidentally, the GCD of two polynomials f and g is any polynomial of maximal degree that divides both f and g. You can see that
this is only well-defined up to multiplying by a constant, since every constant divides every polynomial, but it's otherwise a well-
behaved notion.

CH328 2022-09-02 20:28:38


And what's x−1 in K?

Ilikeapos 2022-09-02 20:29:46


2
−x − 1

CH328 2022-09-02 20:29:48


We have x(x2 + 1) − (x
3
+ x + 1) = −1, so x−1 = −x
2
− 1.

CH328 2022-09-02 20:29:53


OK! In hindsight, maybe we could've guessed that one. But in general, these things are quite tricky to guess.

CH328 2022-09-02 20:30:11


Let's do a less trivial example after the break. We resume at 8:35 EST (5:35 PST).

CH328 2022-09-02 20:35:21


Back to the math! Any questions before we resume?

MasterInTheMaking 2022-09-02 20:35:49


have you ever seen a rick roll in a mathematical paper?

CH328 2022-09-02 20:35:53


I have not!

CH328 2022-09-02 20:36:21


There's a rumor that someone once put a note in his PhD thesis that he'll give the first 5 people who e-mail him about the note
$50... _

CH328 2022-09-02 20:36:46


I'm not sure if that's true, though.

CH328 2022-09-02 20:37:02


Back to field theory/ring theory/polynomials!

CH328 2022-09-02 20:37:06


Still taking K = Q[x]/(x3 + x + 1), can we find the reciprocal of x2 − x + 3?

CH328 2022-09-02 20:37:17


This one is harder, but it's the same idea. What was that idea, again?

Hulk23 2022-09-02 20:38:01


Euclidean

The_goat 2022-09-02 20:38:01


Euclid's division algorithm
Ilikeapos 2022-09-02 20:38:01
Repeated Euclidean Algorithm

HotSoup 2022-09-02 20:38:01


division with remainder & reversed EA

smileapple 2022-09-02 20:38:01


2 3
(x − x + 3)f + g(x + x + 1) = 1

YaoAOPS 2022-09-02 20:38:01


bbezouts?

CH328 2022-09-02 20:38:03


We're trying to solve f ⋅ (x
2
− x + 3) + g ⋅ (x
3
+ x + 1) = 1, so we need the (extended) Euclidean Algorithm!

CH328 2022-09-02 20:38:13


What's the first step?

Hulk23 2022-09-02 20:39:13


Divide x^3+x+1 by x^2-x+3

The_goat 2022-09-02 20:39:13


Divide and get remainder

romans58 2022-09-02 20:39:13


Divide x3 + x + 1 by x2 − x + 3

just4fun 2022-09-02 20:39:13


Divide

CH328 2022-09-02 20:39:16


We do division with remainder on x3 + x + 1 and x2 − x + 3. I won't make you do it because if you've gotten this far into group
theory I'm sure you can divide polynomials.

CH328 2022-09-02 20:39:28


We get

3 2
x + x + 1 = (x + 1)(x − x + 3) + 2x − 2.

I'll help with the gory computations; you just help me decide what to do. What now?

Ilikeapos 2022-09-02 20:40:52


Divide x2 − x + 3 by 2x − 2

The_goat 2022-09-02 20:40:52


Divide x^2-x+3 by 2x-2

smileapple 2022-09-02 20:40:52


divide x by 2x − 2
2
− x + 3

CH328 2022-09-02 20:40:54


We divide again! We find

1
2
x − x + 3 = ( x) (2x − 2) + 3.
2

CH328 2022-09-02 20:40:59


We can stop with that remainder of 3. How come?

Ilikeapos 2022-09-02 20:41:38


It's a constant

The_goat 2022-09-02 20:41:38


Divide by it?

Hulk23 2022-09-02 20:41:38


At the end we'll just divide the inverseby 3?

CH328 2022-09-02 20:41:57


If we can solve f ⋅ (x
2
− x + 3) + g ⋅ (x
3
+ x + 1) = 3, we can just divide by 3 to finish! Remember, every polynomial is
divisible by every constant. Q is a field! Its elements have inverses.

CH328 2022-09-02 20:42:14


The fancy way to say this is "3 is a unit of Q[x]."

CH328 2022-09-02 20:42:24


OK, so what's next?

CH328 2022-09-02 20:43:07


(Remember we're looking for the inverse of x2 − x + 3 )

Ilikeapos 2022-09-02 20:43:36


"Roll back" our algorithm to find the quotient and remainder

DottedCaculator 2022-09-02 20:43:36


reverse the steps to get an equation in terms of x2 − x + 3 and x3 + x + 1

CH328 2022-09-02 20:43:39


We reverse our work. First,

1
2
3 = (x − x + 3) − ( x) (2x − 2).
2

Now what?

HotSoup 2022-09-02 20:44:36


keep plugging and moving upwards

CH328 2022-09-02 20:44:40


Now we substitute for 2x − 2 by reversing the first step:

1
2 3 2
3 = (x − x + 3) − ( x) (x + x + 1 − (x + 1)(x − x + 3)).
2

And now?

HotSoup 2022-09-02 20:45:27


rearrange factors

skim146 2022-09-02 20:45:27


1
distribute the − 2 x

smileapple 2022-09-02 20:45:27


collect stuff

DottedCaculator 2022-09-02 20:45:27


distribute

CH328 2022-09-02 20:45:30


Now we gather like terms! This gives

1 1
2 3
3 = (1 + ( x) (x + 1)) (x − x − 3) − ( x) (x + x + 1).
2 2
So what's our solution?

HotSoup 2022-09-02 20:47:05


1 2 1 1
x + x +
6 6 3

romans58 2022-09-02 20:47:05


1 1 1 2
+ x + x
3 6 6

The_goat 2022-09-02 20:47:05


divided by 3

CH328 2022-09-02 20:47:07


Dividing by 3 on both sides shows that

1 1 1 1 1
2 −1 2
(x − x − 3) = + x(x + 1) = x + x +
3 6 6 6 3

in Q[x]/(x3 + x + 1).

CH328 2022-09-02 20:47:09


Any questions about that process?

CH328 2022-09-02 20:47:42


Not bad! We know how to compute in k[x]/(f ) now.

CH328 2022-09-02 20:48:08


It's unfortunately not as easy as my favorite modular inverse algorithm "guess unless n > 20 "

CH328 2022-09-02 20:48:29


When is k[x]/(f ) not a field?

CH328 2022-09-02 20:48:35


For instance, F7 [x]/(x2 − 2) isn't a field. What's an element with no multiplicative inverse?

CH328 2022-09-02 20:49:15


Well, other than 0. That never has a multiplicative inverse.

YaoAOPS 2022-09-02 20:49:37


x-3?

CH328 2022-09-02 20:49:40


Right!

CH328 2022-09-02 20:49:41


Earlier, we observed that x2 − 2 = (x − 3)(x − 4) in F4 [x].

CH328 2022-09-02 20:49:45


Therefore, in F7 [x]/(x2 − 2), we have (x − 3)(x − 4) = 0 while x − 3, x − 4 ≠ 0.

CH328 2022-09-02 20:50:06


Therefore, one can show x − 3 and x − 4 don't have multiplicative inverses here.

CH328 2022-09-02 20:50:10


In general, what property of f would make you sure k[x]/(f ) fails to be a field in the same way as F7 [x]/(x2 − 2) ?

Math5K 2022-09-02 20:51:00


f is reducible

DottedCaculator 2022-09-02 20:51:00


f is reducible

CH328 2022-09-02 20:51:04


If f factors , so f = gh , then [g][h] = [0] in k[x]/(f ).

CH328 2022-09-02 20:51:08


Of course, if g or h is a constant, then this is trivial. So let's assume neither g nor h is a constant.

CH328 2022-09-02 20:51:37


Is this enough to prove k[x]/(f ) is not a field?

Leo.Euler 2022-09-02 20:52:36


yes

just4fun 2022-09-02 20:52:36


yes

YaoAOPS 2022-09-02 20:52:36


yes

The_goat 2022-09-02 20:52:36


Yes?

MasterInTheMaking 2022-09-02 20:52:36


yes

Math5K 2022-09-02 20:52:36


yes

CH328 2022-09-02 20:52:38


Yes! If f = gh, then both g and h have degree less than n.

CH328 2022-09-02 20:52:47


We've proven that every polynomial has a unique representative of degree less than n in k[x]/(f )!

CH328 2022-09-02 20:52:55


So g can't be 0.

CH328 2022-09-02 20:52:59


Therefore, if f factors as a product of non-constant factors, then k[x]/(f ) has zero divisors, and is not a field.

CH328 2022-09-02 20:53:05


Such an f is called reducible.

CH328 2022-09-02 20:53:08


We've proven that k[x]/(f ) is not a field if f is reducible.

CH328 2022-09-02 20:53:21


We might also conjecture that, if f does not factor as a product of two nonconstant polynomials in k[x], then k[x]/(f ) is always a
field.

CH328 2022-09-02 20:53:31


Let's prove it!

CH328 2022-09-02 20:53:34


Why is K = k[x]/(f ) a field, if f ∈ k[x] is irreducible?

CH328 2022-09-02 20:53:40


To be clear, we say f is irreducible if it's not reducible, so that it can't be factored into non-constant factors. To avoid trivialities, we
also require that f be of positive degree (much as we don't let 1 be a prime number.)

CH328 2022-09-02 20:54:02


What do we need to do?

Hulk23 2022-09-02 20:54:39


There's always an inverse

Math5K 2022-09-02 20:54:39


show each element has an inverse

Hulk23 2022-09-02 20:54:39


Show there's always an invesre

CH328 2022-09-02 20:54:45


We need to explain why [g] ∈ K has a multiplicative inverse if [g] ≠ [0].

CH328 2022-09-02 20:54:48


We know [g] has a unique representative with degree less than deg(f ), so we may as well assume deg(g) < deg(f ).

CH328 2022-09-02 20:54:59


Finding g −1 means to find t and s in k[x] such that tg + sf = 1 in k[x]. For then gt ≡ 1 (mod f ), as desired.

CH328 2022-09-02 20:55:02


And how do we propose to do that?

Ilikeapos 2022-09-02 20:55:36


Euclidean Algorithm

HotSoup 2022-09-02 20:55:36


the usual way with the extended EA

CH328 2022-09-02 20:55:42


Exactly with what we've been doing!

CH328 2022-09-02 20:55:48


With Euclid's algorithm! When does Euclid's algorithm say we can solve this equation for t and s if we briefly imagine that t and s
are integers?

MasterInTheMaking 2022-09-02 20:56:28


relatively prime

HotSoup 2022-09-02 20:56:28


if gcd(g, f ) = 1

mbbchess 2022-09-02 20:56:28


When they are relatively prime

Math5K 2022-09-02 20:56:31


when gcd(g, f ) = 1

CH328 2022-09-02 20:56:33


It says we need gcd(g, f ) = 1.

CH328 2022-09-02 20:56:39


If f is irreducible, can we be sure g and f have no common factors?

CH328 2022-09-02 20:56:44


(Here, as in the definition of irreducibility, a factor of a polynomial is a non-constant factor.)

Ilikeapos 2022-09-02 20:57:12


Yes

YaoAOPS 2022-09-02 20:57:12


we can

HotSoup 2022-09-02 20:57:12


yes, or else f would be reducible

The_goat 2022-09-02 20:57:12


Yes

just4fun 2022-09-02 20:57:12


yes.
MasterInTheMaking 2022-09-02 20:57:12
yes. If they do have a common factor, that factor shows f is reducable

CH328 2022-09-02 20:57:15


Yes, we can! Any factor of g is of too small a degree to be a factor of f .

CH328 2022-09-02 20:57:19


So, assuming Bezout's lemma works as well for k[x] as for Z, we're home free!

CH328 2022-09-02 20:57:23


Do you think Bezout's lemma will work for k[x]? That is, do you think we can always solve tg + sf = 1 in k[x] if g and f have no
common factors of positive degree?

Math5K 2022-09-02 20:58:06


yes

YaoAOPS 2022-09-02 20:58:06


yes

romans58 2022-09-02 20:58:06


yes

The_goat 2022-09-02 20:58:06


I would hope so

Ilikeapos 2022-09-02 20:58:06


Yup

HotSoup 2022-09-02 20:58:06


yes

just4fun 2022-09-02 20:58:06


yes?

CH328 2022-09-02 20:58:09


Yes, it will! Remember how we proved Bezout's lemma back in Week 3?

CH328 2022-09-02 20:58:11


We proved that ⟨a, b⟩ = ⟨d⟩, for any integers a, b ∈ Z and d = gcd(a, b) defined as the smallest positive element of ⟨a, b⟩.

CH328 2022-09-02 20:58:20


An analogous proof will work for k[x]. The key difference is that we have to take the "smallest" element of the ideal
(f , g) = {tg + sf ∣ t, s ∈ k[x]}, which is bigger than just a subgroup.

CH328 2022-09-02 20:58:27


Otherwise, the proof is essentially identical to the one we discussed earlier in the course.

CH328 2022-09-02 20:58:34


In short, Bezout's lemma works great in any ring that has a nice division algorithm.

CH328 2022-09-02 20:58:40


And there we have it! Minus the tedious detail of checking that the Bezout's lemma proof doesn't actually depend on the fact that
our ring consists of integers, we have proven that k[x]/(f ) is a field if and only if f is irreducible.

CH328 2022-09-02 20:58:49


Good stuff!

CH328 2022-09-02 20:58:53


Who's got questions?

CH328 2022-09-02 20:59:31


Here's a quick bit of practice with these ideas. As for integers, if f , a ∈ k[x] we say f ∣ a when a ÷ f has remainder 0. In other
words, f is a factor of a (assuming f is not constant.)

CH328 2022-09-02 20:59:36


Prove that if f is an irreducible polynomial and f ∣ ab, then f ∣ a or f ∣ b . (Euclid's Lemma for Polynomials)

CH328 2022-09-02 20:59:49


Note that f ∣ ab means [ab] = 0 in k[x]/(f ). How do we finish?

Ilikeapos 2022-09-02 21:00:42


ZPP

YaoAOPS 2022-09-02 21:00:42


no zero divisors

smileapple 2022-09-02 21:00:42


$[ab]=[a]\implies [a]=0or=0$ by the ZPP

Math5K 2022-09-02 21:00:42


Then [a] = 0 or = 0

HotSoup 2022-09-02 21:00:42


zero product property

CH328 2022-09-02 21:00:44


By the ZPP for k[x]/(f ), which is a field, this means [a] = 0 or [b] = 0. That is, f ∣ a or f ∣ b !

CH328 2022-09-02 21:00:50


What this tells us is that irreducible polynomials are a lot like prime integers. They satisfy Euclid's lemma and Bezout's lemma in
the same way, and they're the things you can quotient by to get fields.

CH328 2022-09-02 21:01:09


Homomorphisms

CH328 2022-09-02 21:01:19


Last week, we established a few facts about field homomorphisms: they are always injective, and there's always a homomorphism
to a field k from either Q or one of the fields of prime order giving the minimal subfield.

CH328 2022-09-02 21:01:43


With the fields k[x]/(f ) constructed, we can get a much more flexible understanding of certain homomorphisms.

CH328 2022-09-02 21:02:08


Are there any homomorphisms K → C, where K = Q[x]/(x
3
+ x + 1)?

CH328 2022-09-02 21:02:20


First off, let's check that K is a field. We didn't actually finish this earlier. What do we need to show?

YaoAOPS 2022-09-02 21:03:01


no roots in Q

Ilikeapos 2022-09-02 21:03:01


has no rational roots
3
x + x + 1

just4fun 2022-09-02 21:03:01


x^3+x+1 is irreducible.

The_goat 2022-09-02 21:03:01


x^3+x+1 is irreducible

CH328 2022-09-02 21:03:05


We need to show that f (x) = x
3
+ x + 1 is irreducible over Q.

CH328 2022-09-02 21:03:39


We should be a little careful in general; "irreducible" is not the same as "no roots" in general, because, for instance, a quartic might
split into two irreducible quadratics.

CH328 2022-09-02 21:03:59


But why does it suffice to show x3 + x + 1 has no rational roots here?
DottedCaculator 2022-09-02 21:05:07
if it factors it must be linear x quadratic

shadowcharmer 2022-09-02 21:05:07


x
3
+ x + 1 has degree 3

romans58 2022-09-02 21:05:07


If it factors one of the factors has to be linear, meaning it has a rational root

Ilikeapos 2022-09-02 21:05:07


If it splits at all, there must be at least one linear factor

smileapple 2022-09-02 21:05:07


its degree three

HotSoup 2022-09-02 21:05:07


the cubic always has a linear factor which has a root

CH328 2022-09-02 21:05:15


Right, a cubic polynomial can only split as the product of a linear and quadratic polynomial or the product of three linear
polynomials. In any case, any factorization of a cubic will have a linear factor, which means a rational root!

CH328 2022-09-02 21:05:32


So this works for quadratics and cubics, but not higher degree polynomials.

CH328 2022-09-02 21:05:48


Does f have any linear factors (in Q[x])?

DottedCaculator 2022-09-02 21:06:27


no

Ilikeapos 2022-09-02 21:06:27


No

smileapple 2022-09-02 21:06:27


no

The_goat 2022-09-02 21:06:27


No

Math5K 2022-09-02 21:06:27


no because -1 and 1 are not roots

HotSoup 2022-09-02 21:06:27


no, it has no rational roots

YaoAOPS 2022-09-02 21:06:27


rational roots theorem

CH328 2022-09-02 21:06:30


No. By the Rational Root Theorem, the only possible roots are ±1. None of these solve f .

CH328 2022-09-02 21:06:34


Thus this polynomial is irreducible, so K is a field.

CH328 2022-09-02 21:06:41


Back to our main problem. If φ : K → C is a homomorphism, where might φ send x?

DottedCaculator 2022-09-02 21:07:26


α where α is a root of x3 + x + 1 = 0

smileapple 2022-09-02 21:07:26


a (complex) root of x3 + x + 1??

CH328 2022-09-02 21:07:29


Since x3 + x + 1 = 0 in K, we must have φ(x3 + x + 1) = φ(x)
3
+ φ(x) + 1 = 0 in C.

CH328 2022-09-02 21:07:38


So, φ must send x to a root of x3 + x + 1 in C.

CH328 2022-09-02 21:07:57


(Note, by "homomorphism" here, we mean a homomorphism of fields, which is stronger than just a homomorphism of additive
groups.)

CH328 2022-09-02 21:08:01


Are there any other choices to make, in choosing φ?

HotSoup 2022-09-02 21:08:48


no

Math5K 2022-09-02 21:08:48


no

mbbchess 2022-09-02 21:08:48


No

CH328 2022-09-02 21:08:49


No! Once we choose φ(x), we know φ(a + bx + cx2 ) = a + bφ(x) + cφ(x) ,
2
for any a + bx + cx2 ∈ K.

CH328 2022-09-02 21:08:51


How many possibilities are there for φ(x)?

DottedCaculator 2022-09-02 21:09:32


3

MasterInTheMaking 2022-09-02 21:09:32


3

just4fun 2022-09-02 21:09:32


3

skim146 2022-09-02 21:09:32


3

YaoAOPS 2022-09-02 21:09:32


3

The_goat 2022-09-02 21:09:32


3?

Ilikeapos 2022-09-02 21:09:32


3

skim146 2022-09-02 21:09:32


by the fundamental theorem of algebra

CH328 2022-09-02 21:09:34


There are three, by the Fundamental Theorem of Algebra. Here they are in the complex plane:

CH328 2022-09-02 21:09:36


CH328 2022-09-02 21:09:52
By the way, two of these roots appear to be complex conjugates!

CH328 2022-09-02 21:09:54


We see x3 + x + 1, like any cubic, has a real root r. If we factor out x − r from x3 + x + 1, the quotient is a quadratic with real
coefficients, so its complex roots are conjugates of each other!

CH328 2022-09-02 21:10:00


Anyway, we have three proposals for φ. Let's pick any of these three complex roots of x3 + x + 1 and call it α. Then
2 2
φ(a + bx + cx ) = a + bα + cα .

CH328 2022-09-02 21:10:06


What do we need to check now for φ to be a homomorphism?

Hulk23 2022-09-02 21:10:54


φ(ab) = φ(a)φ(b)

just4fun 2022-09-02 21:10:54


phi(ab)=phi(a)phi(b)

The_goat 2022-09-02 21:10:54


Homomorphic property

Leo.Euler 2022-09-02 21:10:54


it should preserve multiplication

DottedCaculator 2022-09-02 21:10:54


,
φ(a) + φ(b) = φ(a + b) φ(a)φ(b) = φ(ab)

Ilikeapos 2022-09-02 21:10:54


If it respects the additive and multiplicative structure of the field

just4fun 2022-09-02 21:10:54


phi(gh)=phi(g)phi(h)

CH328 2022-09-02 21:11:02


We need to check that it preserves both addition and multiplication.

CH328 2022-09-02 21:11:22


Addition is just... bashing.

CH328 2022-09-02 21:11:26


2 ′ 2 ′ ′ ′ 2 ′ ′
φ((ax + bx + c) + (a x + b x + c )) = φ((a + a )x + (b + b )x + (c + c ))

′ 2 ′ ′
= (a + a )α + (b + b )α + (c + c ).

2 ′ 2 ′ ′
= [aα + bα + c] + [a α + b α + c .]

2 ′ 2 ′ ′
= φ(ax + bx + c) + φ(a x + b x + c ).

CH328 2022-09-02 21:11:46


Just apply the definition of φ and do some reorganizing.

CH328 2022-09-02 21:11:56


Now for products. We have

2 ′ 2 ′ ′ ′ 4 ′ ′ 3 ′ ′ ′ 2 ′ ′ ′
(ax + bx + c)(a x + b x + c ) = aa x + (ab + a b)x + (ac + bb + a c)x + (bc + b c)x + cc .

What can we use to easily simplify this, in K?

The_goat 2022-09-02 21:12:52


Reduce x^4 and x^3

HotSoup 2022-09-02 21:12:52


reducing powers/long division

Math5K 2022-09-02 21:12:52


use x3 = −x − 1

CH328 2022-09-02 21:12:56


We use x3 + x + 1 = 0, so x3 = −x − 1 and x4 = −x
2
− x. That makes

′ 4 ′ ′ 3 ′ ′ ′ 2 ′ ′ ′
aa x + (ab + a b)x + (ac + bb + a c)x + (bc + b c)x + cc

′ 2 ′ ′ ′ ′ ′ 2 ′ ′ ′
= aa (−x − x) + (ab + a b)(−x − 1) + (ac + bb + a c)x + (bc + b c)x + cc

′ ′ ′ ′ 2 ′ ′ ′ ′ ′ ′ ′ ′
= (−aa + ac + bb + a c)x + (−aa − ab − a b + bc + b c)x + −ab − a b + cc .

CH328 2022-09-02 21:13:01


We want to show that, if we apply φ to this product, we get φ(ax2 + bx + c)φ(a x
′ 2
+ b x + c ).
′ ′

CH328 2022-09-02 21:13:07


That is, we want to show that

2 ′ 2 ′ ′ ′ ′ ′ ′ 2 ′ ′ ′ ′ ′ ′ ′ ′
(aα + bα + c)(a α + b α + c ) = (−aa + ac + bb + a c)α + (−aa − ab − a b + bc + b c)α + −ab − a b + cc .

I claim we have absolutely no more work to do. How come?

CH328 2022-09-02 21:14:05


(Think about what we've done so far. )

Leo.Euler 2022-09-02 21:14:35


We replace x in the above expression with α

The_goat 2022-09-02 21:14:35


It looks the same as multplication in K

DottedCaculator 2022-09-02 21:14:35


we can reduce the product with α and get the same thing
3
+ α + 1 = 0

CH328 2022-09-02 21:14:38


Of course this identity holds in C! We just proved it holds in a field in which x3 + x + 1 = 0, and we know nothing else about x.

CH328 2022-09-02 21:14:45


In C, we know α3 + α + 1 = 0, and also some more things about α. But just the first fact is enough to prove this identity, in the
very same way as we did in K !
CH328 2022-09-02 21:14:53
We conclude that there are exactly three homomorphisms K → C, given by sending α to any of the three roots of x3 + x + 1 in
C.

CH328 2022-09-02 21:15:02


So, there are three isomorphic copies of K in C.

CH328 2022-09-02 21:15:22


Of course these three fields intersect, in Q. But let's make sure they're distinct.

CH328 2022-09-02 21:15:24


For instance, say α is the real root of x3 + x + 1 in C and β is one of the complex roots.

CH328 2022-09-02 21:15:43


How could we argue that β is not in Q[α] ⊂ C?

CH328 2022-09-02 21:16:54


(If α is real, and every element in Q is real... )

Math5K 2022-09-02 21:17:27


β has a nonzero imaginary part

Ilikeapos 2022-09-02 21:17:27


β is not real

just4fun 2022-09-02 21:17:27


then beta can't be there bc its complex?

HotSoup 2022-09-02 21:17:27


β is not real

MasterInTheMaking 2022-09-02 21:17:27


the isomorphism sends those numbers to real numbers

CH328 2022-09-02 21:17:29


Well, Q[α] is a subset of the real numbers! And β is non-real.

CH328 2022-09-02 21:17:38


OK, but could Q[β] contain β¯?

CH328 2022-09-02 21:17:45


This is a tricky one! Think about isomorphism.

just4fun 2022-09-02 21:18:25


No.

Leo.Euler 2022-09-02 21:18:25


no

CH328 2022-09-02 21:18:34


It can't! Though it's more subtle to prove.

CH328 2022-09-02 21:18:36


Since Q[α] and Q[β] are both isomorphic to K , they are isomorphic to each other. Since Q[α] contains only one root of
x
3
+ x + 1, the same must be true of Q[β]!

CH328 2022-09-02 21:18:43


So the three subfields of C isomorphic to K, namely Q[α], Q[β], and Q[β¯], really are distinct.

CH328 2022-09-02 21:18:54


Incidentally, it seems a bit funny that Q[α] and Q[β] are isomorphic, when one is a field of real numbers and the other is not.

CH328 2022-09-02 21:19:00


This is creepy but true. From field theory's perspective, whether a number is real or not is not inherently meaningful.
CH328 2022-09-02 21:19:12
It only becomes meaningful in fields that are closed under complex conjugation!

CH328 2022-09-02 21:19:23


Roughly speaking, we can generalize this to say that homomorphisms k[x]/(f ) → K are given by choosing a root of f in K.

CH328 2022-09-02 21:19:34


This is a bit of an oversimplification, though! Can you see any issues with this idea?

CH328 2022-09-02 21:20:19


(Think about how k and K might differ which could cause problems...)

romans58 2022-09-02 21:21:04


What if they have different characteristic

DottedCaculator 2022-09-02 21:21:04


k might not output elements of K

CH328 2022-09-02 21:21:16


Well, for instance, such a homomorphism gives a homomorphism k → K. Suppose k and K have different characteristic.

CH328 2022-09-02 21:21:28


We can't have that kind of a homomorphism! If k is of characteristic p but K is of different characteristic q, then we're in big
trouble trying to map k to K, since any homomorphism will try to map p = 0 ∈ k to p ≠ 0 in K.

CH328 2022-09-02 21:21:36


It turns out that the best way to handle this is to suppose that k is a subfield of K already. Then it becomes the case that a
homomorphism φ : k[x]/(f ) → K which agrees with the inclusions of k on either side is uniquely determined by a choice of a
root of f in K.

CH328 2022-09-02 21:21:47


That's a bit of a mouthful, but with the assumptions straight, the proof is identical to our example above!

CH328 2022-09-02 21:21:55


We'll use this theorem a lot: if K is an extension of k, then a homomorphism k[x]/(f ) → K is uniquely determined by choosing a
root of f in K. Any questions about that?

CH328 2022-09-02 21:22:44


Snakes will get to your questions if you had any!

CH328 2022-09-02 21:22:53


Let's introduce one more piece of terminology.

CH328 2022-09-02 21:22:55


The degree of an extension

CH328 2022-09-02 21:23:00


A part of what we've proven above says that every element of k[x]/(f ) may be written uniquely in the form
a0 + a1 x + ⋯ + an−1 x
n−1
, assuming f has degree n.

CH328 2022-09-02 21:23:09


Therefore, there are n elements of k[x]/(f ), namely {1, x, … , xn−1 }, such that every element of k[x]/(f ) may be written
uniquely as a sum of products of those elements with coefficients from k.

CH328 2022-09-02 21:23:20


In general, if K/k is a field extension and K contains elements b0 , b1 , ⋯ bn−1 such that every element of K may be uniquely
written in the form a0 b0 + ⋯ + an−1 bn−1 for a0 , … , an−1 ∈ k, then we say that the degree of K over k is n, and we say
{b0 , … , bn−1 } is a basis for K over k.

CH328 2022-09-02 21:23:55


If you've seen some linear algebra elsewhere or in the textbook, this is equivalent to thinking of K as an n-dimensional k-vector
space.

CH328 2022-09-02 21:24:06


If you haven't, that's fine!

CH328 2022-09-02 21:24:15


We'll use the notation [K : k] for the degree of K over k.

CH328 2022-09-02 21:24:19


For instance, what is the degree of C/R?

HotSoup 2022-09-02 21:25:03


2

Ilikeapos 2022-09-02 21:25:03


2

DottedCaculator 2022-09-02 21:25:03


2

Math5K 2022-09-02 21:25:03


2

romans58 2022-09-02 21:25:03


2

smileapple 2022-09-02 21:25:03


2

just4fun 2022-09-02 21:25:03


2

CH328 2022-09-02 21:25:07


It's 2; a basis is {1, i}. This is a rephrasing of the familiar fact that every complex number can be written in terms of two real
numbers: its real and imaginary parts.

CH328 2022-09-02 21:25:12


What is the degree of Q[x]/(x3 + x + 1) over Q?

MasterInTheMaking 2022-09-02 21:26:02


3

Leo.Euler 2022-09-02 21:26:02


3

The_goat 2022-09-02 21:26:02


3?

Ilikeapos 2022-09-02 21:26:02


3

shadowcharmer 2022-09-02 21:26:02


3

DottedCaculator 2022-09-02 21:26:02


3

smileapple 2022-09-02 21:26:02


3?

Hulk23 2022-09-02 21:26:02


3?

CH328 2022-09-02 21:26:04


It's 3! A basis of this field is {1, x, x2 }, so that every element of this field can be uniquely written as a + bx + cx2 for
a, b, c ∈ Q.

CH328 2022-09-02 21:26:10


Let ω ∈ C be a nontrivial cube root of unity. What is the degree of Q[ω] over Q?

CH328 2022-09-02 21:26:14


Any guesses?

CH328 2022-09-02 21:26:48


It's tempting to say that every element of Q[ω] can be written uniquely as a + bω + cω2 , with a, b, c ∈ Q, much like our previous
example.

CH328 2022-09-02 21:27:03


However... are we sure about that? Certainly we can write them all this way. But...

The_goat 2022-09-02 21:27:33


Actually 2

Math5K 2022-09-02 21:27:33


2

romans58 2022-09-02 21:27:33


Oh wait I think its 2

The_goat 2022-09-02 21:27:33


2
w = −1 − w

Math5K 2022-09-02 21:27:33


w
2
= −w − 1 , so we only need 1 and 2

just4fun 2022-09-02 21:27:33


2

shadowcharmer 2022-09-02 21:27:33


2?

CH328 2022-09-02 21:27:36


It's not unique! We know ω2 = −ω − 1.

CH328 2022-09-02 21:27:44


So it's 2. We have ω2 = −1 − ω, so that a + bω + cω2 = a − c + (b − c)ω.

CH328 2022-09-02 21:28:20


The key is that x3 − 1 with root ω is not the smallest degree polynomial with root ω.

CH328 2022-09-02 21:28:33


Is there a polynomial f (x) ∈ Q[x] such that Q[ω] ≅ Q[x]/(f )?

CH328 2022-09-02 21:28:49


First, what must be the degree of f ?

Math5K 2022-09-02 21:29:22


2

The_goat 2022-09-02 21:29:22


2

DottedCaculator 2022-09-02 21:29:22


2

romans58 2022-09-02 21:29:22


2

MasterInTheMaking 2022-09-02 21:29:22


2

CH328 2022-09-02 21:29:25


If Q[x] is to be a field (and it is), then f must have degree 2 to make the extension Q[ω]/Q have degree 2.
CH328 2022-09-02 21:29:32
So we want a quadratic that has ω as one of its roots! What's such a quadratic?

CH328 2022-09-02 21:29:46


(It may help to think about x . )
3
− 1

The_goat 2022-09-02 21:30:00


2
x + x + 1

DottedCaculator 2022-09-02 21:30:00


2
ω + ω + 1 = 0

Math5K 2022-09-02 21:30:00


2
x + x + 1

niheyang 2022-09-02 21:30:00


2
x + x + 1

smileapple 2022-09-02 21:30:00


2
x + x + 1

MasterInTheMaking 2022-09-02 21:30:00


x^2 + x + 1

CH328 2022-09-02 21:30:03


We know ω3 − 1 = 0, and this difference of cubes factors as (ω − 1)(ω2 + ω + 1) = 0. Since ω ≠ 1, we have that
2
ω + ω + 1 = 0.

CH328 2022-09-02 21:30:06


Therefore we find that Q[ω] ≅ Q[x]/(x
2
+ x + 1).

CH328 2022-09-02 21:30:12


Hmm. That's a lot of example field extensions that have turned out to be of the form k[x]/(f ).

CH328 2022-09-02 21:30:16


Do we have any field extensions that aren't like that?

CH328 2022-09-02 21:30:39


(Think back to last week. )

CH328 2022-09-02 21:31:25


We built one field containing Q[x]...

CH328 2022-09-02 21:32:20


It's okay if you don't remember; it's quite different from the ones we saw today!

CH328 2022-09-02 21:32:23


Recall k(x), the field of rational functions, which contains k as a subfield.

CH328 2022-09-02 21:32:43


This extension can't k(x) be of the form k[x]/(f ). Does anyone see why?

Math5K 2022-09-02 21:33:42


there are infinitely many rational functions, while finitely many residue classes in k[x]/(f )

CH328 2022-09-02 21:33:46


The degree of f would have to be infinite! Indeed, k(x) contains all of k[x] as a subring--we can't possibly give a finite basis.

CH328 2022-09-02 21:33:52


By the way, in general, suppose you write y = a0 x0 + ⋯ + an−1 xn−1 , with the aj all in some field k.

CH328 2022-09-02 21:33:56


Then we say you've written y as a k-linear combination of x0 , … xn−1 . That's very important terminology for linear algebra.
CH328 2022-09-02 21:34:30
So infinite field extensions can give us ones of new forms.

CH328 2022-09-02 21:34:43


Whether or not finite ones can have new forms depends on the characteristic.

CH328 2022-09-02 21:34:55


Every finite-degree extension K/Q is of the form Q[x]/(f ) for some f .

CH328 2022-09-02 21:35:27


This fact is called the primitive element theorem, and in fact it's true for any finite extension K/k of characteristic-zero fields.

CH328 2022-09-02 21:35:38


Things are a little trickier in characteristic p.

CH328 2022-09-02 21:36:18


Summary

CH328 2022-09-02 21:36:20


This week we explored field extensions much more thoroughly, through the lens of polynomial rings k[x].

CH328 2022-09-02 21:36:25


Specifically, we found that if we quotient k[x] by the ideal (f ) = {gf ∣ g ∈ k[x]}, then the result, k[x]/(f ), is a field if and only if
f is irreducible, which means f cannot be factored nontrivially.

CH328 2022-09-02 21:36:28


We also introduced the notion of the degree of an extension K/k : it's the size of a subset S of K such that every element of K
may be written uniquely as a linear combination of elements of S. Such an S is called a basis for K/k.

CH328 2022-09-02 21:36:43


See you all next week, to solve a bunch of problems that stumped Euclid!

© 2023 Art of Problem Solving


About Us • Contact Us • Terms • Privacy

Copyright © 2023 Art of Problem Solving

You might also like