98
Have Yourself a Merry Little Christmas (Theorem) Kate Thompson October 11, 2017

Have Yourself a Merry Little Christmas (Theorem)

  • Upload
    others

  • View
    13

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Have Yourself a Merry Little Christmas (Theorem)

Have Yourself a Merry Little Christmas(Theorem)

Kate Thompson

October 11, 2017

Page 2: Have Yourself a Merry Little Christmas (Theorem)

Personal Story Time:

I How I became a number theorist.

I Why I love number theory.

I I’m teaching number theory this spring! (even thoughtypically offered in fall only) 21-441

Page 3: Have Yourself a Merry Little Christmas (Theorem)

What odd primes can be written as a sum of two (integer)squares?

3? No.5? Yes.Data?

→ 3← 12 + 22 = 5 → 7←→ 11← 22 + 32 = 13 42 + 12 = 17→ 19← → 23← 52 + 22 = 29→ 31← 62 + 12 = 37 52 + 42 = 41

What’s going on here?

Page 4: Have Yourself a Merry Little Christmas (Theorem)

What odd primes can be written as a sum of two (integer)squares?

3?

No.5? Yes.Data?

→ 3← 12 + 22 = 5 → 7←→ 11← 22 + 32 = 13 42 + 12 = 17→ 19← → 23← 52 + 22 = 29→ 31← 62 + 12 = 37 52 + 42 = 41

What’s going on here?

Page 5: Have Yourself a Merry Little Christmas (Theorem)

What odd primes can be written as a sum of two (integer)squares?

3? No.

5? Yes.Data?

→ 3← 12 + 22 = 5 → 7←→ 11← 22 + 32 = 13 42 + 12 = 17→ 19← → 23← 52 + 22 = 29→ 31← 62 + 12 = 37 52 + 42 = 41

What’s going on here?

Page 6: Have Yourself a Merry Little Christmas (Theorem)

What odd primes can be written as a sum of two (integer)squares?

3? No.5?

Yes.Data?

→ 3← 12 + 22 = 5 → 7←→ 11← 22 + 32 = 13 42 + 12 = 17→ 19← → 23← 52 + 22 = 29→ 31← 62 + 12 = 37 52 + 42 = 41

What’s going on here?

Page 7: Have Yourself a Merry Little Christmas (Theorem)

What odd primes can be written as a sum of two (integer)squares?

3? No.5? Yes.

Data?

→ 3← 12 + 22 = 5 → 7←→ 11← 22 + 32 = 13 42 + 12 = 17→ 19← → 23← 52 + 22 = 29→ 31← 62 + 12 = 37 52 + 42 = 41

What’s going on here?

Page 8: Have Yourself a Merry Little Christmas (Theorem)

What odd primes can be written as a sum of two (integer)squares?

3? No.5? Yes.Data?

→ 3← 12 + 22 = 5 → 7←→ 11← 22 + 32 = 13 42 + 12 = 17→ 19← → 23← 52 + 22 = 29→ 31← 62 + 12 = 37 52 + 42 = 41

What’s going on here?

Page 9: Have Yourself a Merry Little Christmas (Theorem)

What odd primes can be written as a sum of two (integer)squares?

3? No.5? Yes.Data?

→ 3← 12 + 22 = 5 → 7←→ 11← 22 + 32 = 13 42 + 12 = 17→ 19← → 23← 52 + 22 = 29→ 31← 62 + 12 = 37 52 + 42 = 41

What’s going on here?

Page 10: Have Yourself a Merry Little Christmas (Theorem)

Theorem (Fermat’s Christmas Theorem, 1640)

Let p be an odd prime. There exist integers x and y such that

x2 + y2 = p

if and only if p ≡ 1 (mod 4); i.e., if and only if p = 4n + 1 forsome n ∈ N.

First proof?

I Euler, ≈ 1747.

I Lagrange (and Gauss), ≈ 1775− 1800

I Dedekind, ≈ 1875− 1890

I Minkowski, ≈ 1900

I Zagier, 1990

I A. David Christopher, 2016

Page 11: Have Yourself a Merry Little Christmas (Theorem)

Theorem (Fermat’s Christmas Theorem, 1640)

Let p be an odd prime. There exist integers x and y such that

x2 + y2 = p

if and only if p ≡ 1 (mod 4); i.e., if and only if p = 4n + 1 forsome n ∈ N.

First proof?

I Euler, ≈ 1747.

I Lagrange (and Gauss), ≈ 1775− 1800

I Dedekind, ≈ 1875− 1890

I Minkowski, ≈ 1900

I Zagier, 1990

I A. David Christopher, 2016

Page 12: Have Yourself a Merry Little Christmas (Theorem)

Theorem (Fermat’s Christmas Theorem, 1640)

Let p be an odd prime. There exist integers x and y such that

x2 + y2 = p

if and only if p ≡ 1 (mod 4); i.e., if and only if p = 4n + 1 forsome n ∈ N.

First proof?

I Euler, ≈ 1747.

I Lagrange (and Gauss), ≈ 1775− 1800

I Dedekind, ≈ 1875− 1890

I Minkowski, ≈ 1900

I Zagier, 1990

I A. David Christopher, 2016

Page 13: Have Yourself a Merry Little Christmas (Theorem)

Theorem (Fermat’s Christmas Theorem, 1640)

Let p be an odd prime. There exist integers x and y such that

x2 + y2 = p

if and only if p ≡ 1 (mod 4); i.e., if and only if p = 4n + 1 forsome n ∈ N.

First proof?

I Euler, ≈ 1747.

I Lagrange (and Gauss), ≈ 1775− 1800

I Dedekind, ≈ 1875− 1890

I Minkowski, ≈ 1900

I Zagier, 1990

I A. David Christopher, 2016

Page 14: Have Yourself a Merry Little Christmas (Theorem)

One direction is “easy”.

(An odd prime p = x2 + y2) ⇒ (p ≡ 1 (mod 4)).

Proof.For n ∈ Z, n2 ≡ 0, 1 (mod 4). So, x2 + y2 ≡ 0, 1, 2 (mod 4).Thus, if p = x2 + y2 is odd, then p ≡ 1 (mod 4).

The other direction will not be a two line proof.

Page 15: Have Yourself a Merry Little Christmas (Theorem)

One direction is “easy”.

(An odd prime p = x2 + y2) ⇒ (p ≡ 1 (mod 4)).

Proof.For n ∈ Z, n2 ≡ 0, 1 (mod 4). So, x2 + y2 ≡ 0, 1, 2 (mod 4).Thus, if p = x2 + y2 is odd, then p ≡ 1 (mod 4).

The other direction will not be a two line proof.

Page 16: Have Yourself a Merry Little Christmas (Theorem)

One direction is “easy”.

(An odd prime p = x2 + y2) ⇒ (p ≡ 1 (mod 4)).

Proof.For n ∈ Z, n2 ≡ 0, 1 (mod 4). So, x2 + y2 ≡ 0, 1, 2 (mod 4).Thus, if p = x2 + y2 is odd, then p ≡ 1 (mod 4).

The other direction will not be a two line proof.

Page 17: Have Yourself a Merry Little Christmas (Theorem)

One direction is “easy”.

(An odd prime p = x2 + y2) ⇒ (p ≡ 1 (mod 4)).

Proof.For n ∈ Z, n2 ≡ 0, 1 (mod 4). So, x2 + y2 ≡ 0, 1, 2 (mod 4).Thus, if p = x2 + y2 is odd, then p ≡ 1 (mod 4).

The other direction will not be a two line proof.

Page 18: Have Yourself a Merry Little Christmas (Theorem)

We will use one tool with some frequency:

If p ≡ 1 (mod 4), there is an integer 1 ≤ u ≤ p − 1 withu2 ≡ −1 (mod p).

Just so we’re clear:

I If p = 5, then u =2, 3.

I If p = 13, then u =5, 8

I If p = 101, then u = 10, 91

Not to get side-tracked, but for those interested this follows fromWilson’s Theorem.

Page 19: Have Yourself a Merry Little Christmas (Theorem)

We will use one tool with some frequency:

If p ≡ 1 (mod 4), there is an integer 1 ≤ u ≤ p − 1 withu2 ≡ −1 (mod p).

Just so we’re clear:

I If p = 5, then u =2, 3.

I If p = 13, then u =5, 8

I If p = 101, then u = 10, 91

Not to get side-tracked, but for those interested this follows fromWilson’s Theorem.

Page 20: Have Yourself a Merry Little Christmas (Theorem)

We will use one tool with some frequency:

If p ≡ 1 (mod 4), there is an integer 1 ≤ u ≤ p − 1 withu2 ≡ −1 (mod p).

Just so we’re clear:

I If p = 5, then u =2, 3.

I If p = 13, then u =5, 8

I If p = 101, then u = 10, 91

Not to get side-tracked, but for those interested this follows fromWilson’s Theorem.

Page 21: Have Yourself a Merry Little Christmas (Theorem)

We will use one tool with some frequency:

If p ≡ 1 (mod 4), there is an integer 1 ≤ u ≤ p − 1 withu2 ≡ −1 (mod p).

Just so we’re clear:

I If p = 5, then u =

2, 3.

I If p = 13, then u =5, 8

I If p = 101, then u = 10, 91

Not to get side-tracked, but for those interested this follows fromWilson’s Theorem.

Page 22: Have Yourself a Merry Little Christmas (Theorem)

We will use one tool with some frequency:

If p ≡ 1 (mod 4), there is an integer 1 ≤ u ≤ p − 1 withu2 ≡ −1 (mod p).

Just so we’re clear:

I If p = 5, then u =2, 3.

I If p = 13, then u =5, 8

I If p = 101, then u = 10, 91

Not to get side-tracked, but for those interested this follows fromWilson’s Theorem.

Page 23: Have Yourself a Merry Little Christmas (Theorem)

We will use one tool with some frequency:

If p ≡ 1 (mod 4), there is an integer 1 ≤ u ≤ p − 1 withu2 ≡ −1 (mod p).

Just so we’re clear:

I If p = 5, then u =2, 3.

I If p = 13, then u =

5, 8

I If p = 101, then u = 10, 91

Not to get side-tracked, but for those interested this follows fromWilson’s Theorem.

Page 24: Have Yourself a Merry Little Christmas (Theorem)

We will use one tool with some frequency:

If p ≡ 1 (mod 4), there is an integer 1 ≤ u ≤ p − 1 withu2 ≡ −1 (mod p).

Just so we’re clear:

I If p = 5, then u =2, 3.

I If p = 13, then u =5, 8

I If p = 101, then u = 10, 91

Not to get side-tracked, but for those interested this follows fromWilson’s Theorem.

Page 25: Have Yourself a Merry Little Christmas (Theorem)

We will use one tool with some frequency:

If p ≡ 1 (mod 4), there is an integer 1 ≤ u ≤ p − 1 withu2 ≡ −1 (mod p).

Just so we’re clear:

I If p = 5, then u =2, 3.

I If p = 13, then u =5, 8

I If p = 101, then u =

10, 91

Not to get side-tracked, but for those interested this follows fromWilson’s Theorem.

Page 26: Have Yourself a Merry Little Christmas (Theorem)

We will use one tool with some frequency:

If p ≡ 1 (mod 4), there is an integer 1 ≤ u ≤ p − 1 withu2 ≡ −1 (mod p).

Just so we’re clear:

I If p = 5, then u =2, 3.

I If p = 13, then u =5, 8

I If p = 101, then u = 10, 91

Not to get side-tracked, but for those interested this follows fromWilson’s Theorem.

Page 27: Have Yourself a Merry Little Christmas (Theorem)

We will use one tool with some frequency:

If p ≡ 1 (mod 4), there is an integer 1 ≤ u ≤ p − 1 withu2 ≡ −1 (mod p).

Just so we’re clear:

I If p = 5, then u =2, 3.

I If p = 13, then u =5, 8

I If p = 101, then u = 10, 91

Not to get side-tracked, but for those interested this follows fromWilson’s Theorem.

Page 28: Have Yourself a Merry Little Christmas (Theorem)

So, in what time remains we will prove:

(p ≡ 1 (mod 4))⇒ (p = x2 + y2)

using

I Algebraic techniques (mimic-ing Dedekind)

I Geometric techniques (a la Minkowski)

I Analytic techniques (local-global principle)

Page 29: Have Yourself a Merry Little Christmas (Theorem)

So, in what time remains we will prove:

(p ≡ 1 (mod 4))⇒ (p = x2 + y2)

using

I Algebraic techniques (mimic-ing Dedekind)

I Geometric techniques (a la Minkowski)

I Analytic techniques (local-global principle)

Page 30: Have Yourself a Merry Little Christmas (Theorem)

So, in what time remains we will prove:

(p ≡ 1 (mod 4))⇒ (p = x2 + y2)

using

I Algebraic techniques (mimic-ing Dedekind)

I Geometric techniques (a la Minkowski)

I Analytic techniques (local-global principle)

Page 31: Have Yourself a Merry Little Christmas (Theorem)

Algebra Idea: Use Arithmetic in the Gaussians

The Gaussian integers are simply

Z[i ] = {a + bi : a, b ∈ Z, i2 = −1}.

Why should this have anything to do with the sum of two squares?

(x + iy)(x − iy) = x2 + y2

So essentially, saying that p = x2 + y2 is somehow equivalent tofactoring p in the Gaussians.

Page 32: Have Yourself a Merry Little Christmas (Theorem)

Algebra Idea: Use Arithmetic in the Gaussians

The Gaussian integers are simply

Z[i ] = {a + bi : a, b ∈ Z, i2 = −1}.

Why should this have anything to do with the sum of two squares?

(x + iy)(x − iy) = x2 + y2

So essentially, saying that p = x2 + y2 is somehow equivalent tofactoring p in the Gaussians.

Page 33: Have Yourself a Merry Little Christmas (Theorem)

Algebra Idea: Use Arithmetic in the Gaussians

The Gaussian integers are simply

Z[i ] = {a + bi : a, b ∈ Z, i2 = −1}.

Why should this have anything to do with the sum of two squares?

(x + iy)(x − iy) = x2 + y2

So essentially, saying that p = x2 + y2 is somehow equivalent tofactoring p in the Gaussians.

Page 34: Have Yourself a Merry Little Christmas (Theorem)

Algebra Idea: Use Arithmetic in the Gaussians

The Gaussian integers are simply

Z[i ] = {a + bi : a, b ∈ Z, i2 = −1}.

Why should this have anything to do with the sum of two squares?

(x + iy)(x − iy) = x2 + y2

So essentially, saying that p = x2 + y2 is somehow equivalent tofactoring p in the Gaussians.

Page 35: Have Yourself a Merry Little Christmas (Theorem)

So suppose p ≡ 1 (mod 4). Then, by our fact, there is an integerm with m2 ≡ −1 (mod p). That means that m2 + 1 ≡ 0 (mod p)or p|(m2 + 1).

Factor m2 + 1 in the Gaussians as (m + i)(m − i). So what wehave now is

p|(m + i)(m − i).

One fun fact if p|ab and p is prime, then p|a or p|b. What we’regoing to first argue, then, is that p is not prime in Z[i ]. Thatmeans we’re going to show that while p|(m + i)(m − i), p - m + iand p - m − i .

Page 36: Have Yourself a Merry Little Christmas (Theorem)

So suppose p ≡ 1 (mod 4). Then, by our fact, there is an integerm with m2 ≡ −1 (mod p). That means that m2 + 1 ≡ 0 (mod p)or p|(m2 + 1).

Factor m2 + 1 in the Gaussians as (m + i)(m − i). So what wehave now is

p|(m + i)(m − i).

One fun fact if p|ab and p is prime, then p|a or p|b. What we’regoing to first argue, then, is that p is not prime in Z[i ]. Thatmeans we’re going to show that while p|(m + i)(m − i), p - m + iand p - m − i .

Page 37: Have Yourself a Merry Little Christmas (Theorem)

So suppose p ≡ 1 (mod 4). Then, by our fact, there is an integerm with m2 ≡ −1 (mod p). That means that m2 + 1 ≡ 0 (mod p)or p|(m2 + 1).

Factor m2 + 1 in the Gaussians as (m + i)(m − i). So what wehave now is

p|(m + i)(m − i).

One fun fact if p|ab and p is prime, then p|a or p|b. What we’regoing to first argue, then, is that p is not prime in Z[i ]. Thatmeans we’re going to show that while p|(m + i)(m − i), p - m + iand p - m − i .

Page 38: Have Yourself a Merry Little Christmas (Theorem)

Suppose that p|m + i . Then there would exist integers a, b so that

p(a + bi) = m + i .

Among other things, this means there’s an integer b so thatpb = 1. Einen. Kleinen. Problemo. So p - m + i .

Similarly, you can conclude that p - m − i and therefore p is notprime in the Gaussians.

Now we still have to show that p is a sum of two integer squares.

Page 39: Have Yourself a Merry Little Christmas (Theorem)

Suppose that p|m + i . Then there would exist integers a, b so that

p(a + bi) = m + i .

Among other things, this means there’s an integer b so thatpb = 1. Einen. Kleinen. Problemo. So p - m + i .

Similarly, you can conclude that p - m − i and therefore p is notprime in the Gaussians.

Now we still have to show that p is a sum of two integer squares.

Page 40: Have Yourself a Merry Little Christmas (Theorem)

Suppose that p|m + i . Then there would exist integers a, b so that

p(a + bi) = m + i .

Among other things, this means there’s an integer b so thatpb = 1. Einen. Kleinen. Problemo. So p - m + i .

Similarly, you can conclude that p - m − i and therefore p is notprime in the Gaussians.

Now we still have to show that p is a sum of two integer squares.

Page 41: Have Yourself a Merry Little Christmas (Theorem)

Since p composite, there exist nonzero integers x1, x2, y1, y2 with:

I gcd(x1, y1) = gcd(x2, y2) = 1

I |y2| ≤ |y1|such that

p = (x1 + y1i)(x2 + y2i)

= p + 0i = (x1x2 − y1y2) + i(x2y1 + x1y2)

y1 = −(x1y2)/x2, so x2|x1 or (x1/x2) ∈ Z.

p = x1x2 +

(x1y2x2

)y2

=x1x2

(x22 + y22 )

Since x22 , y22 ∈ N, we know x1 = x2, and we’re done.

Page 42: Have Yourself a Merry Little Christmas (Theorem)

Since p composite, there exist nonzero integers x1, x2, y1, y2 with:

I gcd(x1, y1) = gcd(x2, y2) = 1

I |y2| ≤ |y1|

such that

p = (x1 + y1i)(x2 + y2i)

= p + 0i = (x1x2 − y1y2) + i(x2y1 + x1y2)

y1 = −(x1y2)/x2, so x2|x1 or (x1/x2) ∈ Z.

p = x1x2 +

(x1y2x2

)y2

=x1x2

(x22 + y22 )

Since x22 , y22 ∈ N, we know x1 = x2, and we’re done.

Page 43: Have Yourself a Merry Little Christmas (Theorem)

Since p composite, there exist nonzero integers x1, x2, y1, y2 with:

I gcd(x1, y1) = gcd(x2, y2) = 1

I |y2| ≤ |y1|such that

p = (x1 + y1i)(x2 + y2i)

= p + 0i = (x1x2 − y1y2) + i(x2y1 + x1y2)

y1 = −(x1y2)/x2, so x2|x1 or (x1/x2) ∈ Z.

p = x1x2 +

(x1y2x2

)y2

=x1x2

(x22 + y22 )

Since x22 , y22 ∈ N, we know x1 = x2, and we’re done.

Page 44: Have Yourself a Merry Little Christmas (Theorem)

Since p composite, there exist nonzero integers x1, x2, y1, y2 with:

I gcd(x1, y1) = gcd(x2, y2) = 1

I |y2| ≤ |y1|such that

p = (x1 + y1i)(x2 + y2i)

= p + 0i = (x1x2 − y1y2) + i(x2y1 + x1y2)

y1 = −(x1y2)/x2, so x2|x1 or (x1/x2) ∈ Z.

p = x1x2 +

(x1y2x2

)y2

=x1x2

(x22 + y22 )

Since x22 , y22 ∈ N, we know x1 = x2, and we’re done.

Page 45: Have Yourself a Merry Little Christmas (Theorem)

Since p composite, there exist nonzero integers x1, x2, y1, y2 with:

I gcd(x1, y1) = gcd(x2, y2) = 1

I |y2| ≤ |y1|such that

p = (x1 + y1i)(x2 + y2i)

= p + 0i = (x1x2 − y1y2) + i(x2y1 + x1y2)

y1 = −(x1y2)/x2, so x2|x1 or (x1/x2) ∈ Z.

p = x1x2 +

(x1y2x2

)y2

=x1x2

(x22 + y22 )

Since x22 , y22 ∈ N, we know x1 = x2, and we’re done.

Page 46: Have Yourself a Merry Little Christmas (Theorem)

Since p composite, there exist nonzero integers x1, x2, y1, y2 with:

I gcd(x1, y1) = gcd(x2, y2) = 1

I |y2| ≤ |y1|such that

p = (x1 + y1i)(x2 + y2i)

= p + 0i = (x1x2 − y1y2) + i(x2y1 + x1y2)

y1 = −(x1y2)/x2, so x2|x1 or (x1/x2) ∈ Z.

p = x1x2 +

(x1y2x2

)y2

=x1x2

(x22 + y22 )

Since x22 , y22 ∈ N, we know x1 = x2, and we’re done.

Page 47: Have Yourself a Merry Little Christmas (Theorem)

What this problem really was using was:

I Unique factorization of elements into primes.

Believe it or not, this doesn’t hold in greater generality. And thisbecomes a delicacy in higher-level algebraic number theory.

As a parting example, consider instead Z[√−5]. Then

6 = 2 · 3 = (1 +√−5)(1−

√−5) are two “distinct” factorizations.

So if, say, you wanted to try to generalize and say what primes areof the form x2 + 5y2, you’d have problems using these techniques.

Page 48: Have Yourself a Merry Little Christmas (Theorem)

What this problem really was using was:

I Unique factorization of elements into primes.

Believe it or not, this doesn’t hold in greater generality. And thisbecomes a delicacy in higher-level algebraic number theory.

As a parting example, consider instead Z[√−5]. Then

6 = 2 · 3 = (1 +√−5)(1−

√−5) are two “distinct” factorizations.

So if, say, you wanted to try to generalize and say what primes areof the form x2 + 5y2, you’d have problems using these techniques.

Page 49: Have Yourself a Merry Little Christmas (Theorem)

Geometric Proof

The main tools here are:

I Lattices

I Minkowski’s Convex Body Theorem (MCBT)

A lattice formally-speaking is a discrete subset of Rn, and given agenerating set {v1, v2, ..., vn} any point in the lattice is someZ-linear combination of the vi ’s.

Informally, you can think of Zn ⊆ Rn, or any “graph paper onsteroids” series of dots in the plane.

Page 50: Have Yourself a Merry Little Christmas (Theorem)

Geometric Proof

The main tools here are:

I Lattices

I Minkowski’s Convex Body Theorem (MCBT)

A lattice formally-speaking is a discrete subset of Rn, and given agenerating set {v1, v2, ..., vn} any point in the lattice is someZ-linear combination of the vi ’s.

Informally, you can think of Zn ⊆ Rn, or any “graph paper onsteroids” series of dots in the plane.

Page 51: Have Yourself a Merry Little Christmas (Theorem)

Geometric Proof

The main tools here are:

I Lattices

I Minkowski’s Convex Body Theorem (MCBT)

A lattice formally-speaking is a discrete subset of Rn, and given agenerating set {v1, v2, ..., vn} any point in the lattice is someZ-linear combination of the vi ’s.

Informally, you can think of Zn ⊆ Rn, or any “graph paper onsteroids” series of dots in the plane.

Page 52: Have Yourself a Merry Little Christmas (Theorem)

A lattice tiles the plane (informally, think literally of graph paper).A fundamental domain is a basic tile. That is, given thegenerating set {v1, ..., vn} the fundamental domain is then-dimensional parallelogram whose sides are given by those vectors.

Example: If you’re thinking of Z2 ⊆ R2, a fundamental domaincould be a unit square.

Theorem (Minkowski’s Convex Body Theorem)

Let L ⊆ Rn be a lattice with fundamental domain D. Let B ⊆ Rn

be convex, and centrally-symmetric. If

vol(B) > 2nvol(D)

then B contains a nonzero lattice point in its interior.

Page 53: Have Yourself a Merry Little Christmas (Theorem)

A lattice tiles the plane (informally, think literally of graph paper).A fundamental domain is a basic tile. That is, given thegenerating set {v1, ..., vn} the fundamental domain is then-dimensional parallelogram whose sides are given by those vectors.

Example: If you’re thinking of Z2 ⊆ R2, a fundamental domaincould be a unit square.

Theorem (Minkowski’s Convex Body Theorem)

Let L ⊆ Rn be a lattice with fundamental domain D. Let B ⊆ Rn

be convex, and centrally-symmetric. If

vol(B) > 2nvol(D)

then B contains a nonzero lattice point in its interior.

Page 54: Have Yourself a Merry Little Christmas (Theorem)

A lattice tiles the plane (informally, think literally of graph paper).A fundamental domain is a basic tile. That is, given thegenerating set {v1, ..., vn} the fundamental domain is then-dimensional parallelogram whose sides are given by those vectors.

Example: If you’re thinking of Z2 ⊆ R2, a fundamental domaincould be a unit square.

Theorem (Minkowski’s Convex Body Theorem)

Let L ⊆ Rn be a lattice with fundamental domain D. Let B ⊆ Rn

be convex, and centrally-symmetric. If

vol(B) > 2nvol(D)

then B contains a nonzero lattice point in its interior.

Page 55: Have Yourself a Merry Little Christmas (Theorem)

What does this have to do with p ≡ 1 (mod 4)⇒ p = x2 + y2?

If you think about Q(x , y) = x2 + y2 and what the level sets are forthat function, they are circles centered at the origin. In particular,those are convex, centrally-symmetric bodies. And if you think ofyour lattice as containing integer points, then what you’d really liketo show is that x2 + y2 = p has an integer point on the boundary.There are just a few issues:

1. Z2 is too big of a lattice.

2. MCBT shows there’s a nonzero lattice point inside a region,not on the boundary.

Page 56: Have Yourself a Merry Little Christmas (Theorem)

What does this have to do with p ≡ 1 (mod 4)⇒ p = x2 + y2?

If you think about Q(x , y) = x2 + y2 and what the level sets are forthat function, they are circles centered at the origin. In particular,those are convex, centrally-symmetric bodies. And if you think ofyour lattice as containing integer points, then what you’d really liketo show is that x2 + y2 = p has an integer point on the boundary.

There are just a few issues:

1. Z2 is too big of a lattice.

2. MCBT shows there’s a nonzero lattice point inside a region,not on the boundary.

Page 57: Have Yourself a Merry Little Christmas (Theorem)

What does this have to do with p ≡ 1 (mod 4)⇒ p = x2 + y2?

If you think about Q(x , y) = x2 + y2 and what the level sets are forthat function, they are circles centered at the origin. In particular,those are convex, centrally-symmetric bodies. And if you think ofyour lattice as containing integer points, then what you’d really liketo show is that x2 + y2 = p has an integer point on the boundary.There are just a few issues:

1. Z2 is too big of a lattice.

2. MCBT shows there’s a nonzero lattice point inside a region,not on the boundary.

Page 58: Have Yourself a Merry Little Christmas (Theorem)

What does this have to do with p ≡ 1 (mod 4)⇒ p = x2 + y2?

If you think about Q(x , y) = x2 + y2 and what the level sets are forthat function, they are circles centered at the origin. In particular,those are convex, centrally-symmetric bodies. And if you think ofyour lattice as containing integer points, then what you’d really liketo show is that x2 + y2 = p has an integer point on the boundary.There are just a few issues:

1. Z2 is too big of a lattice.

2. MCBT shows there’s a nonzero lattice point inside a region,not on the boundary.

Page 59: Have Yourself a Merry Little Christmas (Theorem)

What does this have to do with p ≡ 1 (mod 4)⇒ p = x2 + y2?

If you think about Q(x , y) = x2 + y2 and what the level sets are forthat function, they are circles centered at the origin. In particular,those are convex, centrally-symmetric bodies. And if you think ofyour lattice as containing integer points, then what you’d really liketo show is that x2 + y2 = p has an integer point on the boundary.There are just a few issues:

1. Z2 is too big of a lattice.

2. MCBT shows there’s a nonzero lattice point inside a region,not on the boundary.

Page 60: Have Yourself a Merry Little Christmas (Theorem)

To fix these issues, we alter the lattice and do not consider Z2.Instead, we consider the lattice L generated by {(p, 0), (u, 1)},where 1 ≤ u ≤ p is an integer such that u2 ≡ −1 (mod p).

Consider any vector v ∈ L. Then

v = a(p, 0) + b(u, 1)

= (ap + bu, b)

|v | = p(a2p + 2abu) + b2(u2 + 1)

≡ 0 (mod p).

That is, every vector in this lattice has a norm divisible by p.

Page 61: Have Yourself a Merry Little Christmas (Theorem)

To fix these issues, we alter the lattice and do not consider Z2.Instead, we consider the lattice L generated by {(p, 0), (u, 1)},where 1 ≤ u ≤ p is an integer such that u2 ≡ −1 (mod p).

Consider any vector v ∈ L. Then

v = a(p, 0) + b(u, 1)

= (ap + bu, b)

|v | = p(a2p + 2abu) + b2(u2 + 1)

≡ 0 (mod p).

That is, every vector in this lattice has a norm divisible by p.

Page 62: Have Yourself a Merry Little Christmas (Theorem)

To fix these issues, we alter the lattice and do not consider Z2.Instead, we consider the lattice L generated by {(p, 0), (u, 1)},where 1 ≤ u ≤ p is an integer such that u2 ≡ −1 (mod p).

Consider any vector v ∈ L. Then

v = a(p, 0) + b(u, 1)

= (ap + bu, b)

|v | = p(a2p + 2abu) + b2(u2 + 1)

≡ 0 (mod p).

That is, every vector in this lattice has a norm divisible by p.

Page 63: Have Yourself a Merry Little Christmas (Theorem)

And now, in a second moment of brilliance, we don’t consider thelevel set associated to x2 + y2 = p. Instead, we look at the circlex2 + y2 = 2p.

We want to use Minkowski, and we can see fairly easily the volumeof our ‘B’ is 2πp. But what is the volume of the fundamentaldomain? Since the lattice is generated by (p, 0) and (u, 1)

vol(D) = det

∣∣∣∣(p u0 1

)∣∣∣∣ = p.

And since 2πp > 22p, MCBT tells us that there is a nonzerolattice point inside our B.

Page 64: Have Yourself a Merry Little Christmas (Theorem)

And now, in a second moment of brilliance, we don’t consider thelevel set associated to x2 + y2 = p. Instead, we look at the circlex2 + y2 = 2p.We want to use Minkowski, and we can see fairly easily the volumeof our ‘B’ is 2πp. But what is the volume of the fundamentaldomain? Since the lattice is generated by (p, 0) and (u, 1)

vol(D) = det

∣∣∣∣(p u0 1

)∣∣∣∣ = p.

And since 2πp > 22p, MCBT tells us that there is a nonzerolattice point inside our B.

Page 65: Have Yourself a Merry Little Christmas (Theorem)

And now, in a second moment of brilliance, we don’t consider thelevel set associated to x2 + y2 = p. Instead, we look at the circlex2 + y2 = 2p.We want to use Minkowski, and we can see fairly easily the volumeof our ‘B’ is 2πp. But what is the volume of the fundamentaldomain? Since the lattice is generated by (p, 0) and (u, 1)

vol(D) = det

∣∣∣∣(p u0 1

)∣∣∣∣ = p.

And since 2πp > 22p, MCBT tells us that there is a nonzerolattice point inside our B.

Page 66: Have Yourself a Merry Little Christmas (Theorem)

But now we’re done. Every lattice point has norm equal to someinteger multiple of p. We’re looking in a region where all pointsinside have norm strictly less than 2p. So, our nonzero latticepoint has norm equal to some multiple of p bounded between 0pand 2p (exclusive). Therefore, we must have some element ofnorm p, which gives us a solution to

x2 + y2 = p.

Page 67: Have Yourself a Merry Little Christmas (Theorem)

There are a few restrictions to these geometric methods:

I The reliance on MCBT that the region B be convex andcentrally symmetric (let alone of some kind of finite volume).For instance, this would prevent you from using MCBT on thelevel sets associated to x2 − y2.

I Beyond that, in higher variables, generalizations of thesegeometric techniques only really work well when you have aneven number of variables. So, trying to use this even forx2 + y2 + z2 is more difficult that you would (want to) think.

Page 68: Have Yourself a Merry Little Christmas (Theorem)

There are a few restrictions to these geometric methods:

I The reliance on MCBT that the region B be convex andcentrally symmetric (let alone of some kind of finite volume).For instance, this would prevent you from using MCBT on thelevel sets associated to x2 − y2.

I Beyond that, in higher variables, generalizations of thesegeometric techniques only really work well when you have aneven number of variables. So, trying to use this even forx2 + y2 + z2 is more difficult that you would (want to) think.

Page 69: Have Yourself a Merry Little Christmas (Theorem)

There are a few restrictions to these geometric methods:

I The reliance on MCBT that the region B be convex andcentrally symmetric (let alone of some kind of finite volume).For instance, this would prevent you from using MCBT on thelevel sets associated to x2 − y2.

I Beyond that, in higher variables, generalizations of thesegeometric techniques only really work well when you have aneven number of variables. So, trying to use this even forx2 + y2 + z2 is more difficult that you would (want to) think.

Page 70: Have Yourself a Merry Little Christmas (Theorem)

Analytic Proof: Local-Global Theorem

The idea behind the local-global principle (aka the Hasseprinciple) is this: if you have a polynomial P(x1, x2, ..., xn) (overthe integers, with integer coefficients) satisfying

P(x1, x2, ..., xn) = 0

then ‘obviously’ you have a solution to

P(x1, x2, ..., xn) ≡ 0 (mod n)

for all n ∈ N.

But when can you move backwards? That is, when does asolution (mod n) for all n imply a solution in Z?

Page 71: Have Yourself a Merry Little Christmas (Theorem)

Analytic Proof: Local-Global Theorem

The idea behind the local-global principle (aka the Hasseprinciple) is this: if you have a polynomial P(x1, x2, ..., xn) (overthe integers, with integer coefficients) satisfying

P(x1, x2, ..., xn) = 0

then ‘obviously’ you have a solution to

P(x1, x2, ..., xn) ≡ 0 (mod n)

for all n ∈ N.

But when can you move backwards? That is, when does asolution (mod n) for all n imply a solution in Z?

Page 72: Have Yourself a Merry Little Christmas (Theorem)

This greatly depends upon the polynomial. Very famously, in 1951Selmer showed that while

3x3 + 4y3 + 5z3 = 0

has nontrivial solutions (mod n) for all n, the only solution overthe integers is the trivial x = y = z = 0.

Lucky for us, P(x , y) = x2 + y2 − p satisfies the Hasse principle.That is, if P(x , y) = 0 has a nontrivial solution (mod n) for all n,then it is guaranteed to have a solution in Z.

Page 73: Have Yourself a Merry Little Christmas (Theorem)

This greatly depends upon the polynomial. Very famously, in 1951Selmer showed that while

3x3 + 4y3 + 5z3 = 0

has nontrivial solutions (mod n) for all n, the only solution overthe integers is the trivial x = y = z = 0.

Lucky for us, P(x , y) = x2 + y2 − p satisfies the Hasse principle.That is, if P(x , y) = 0 has a nontrivial solution (mod n) for all n,then it is guaranteed to have a solution in Z.

Page 74: Have Yourself a Merry Little Christmas (Theorem)

Still, it’s a daunting task to check that there’s a solution (mod n)for all n. One thing we use to simplify life is the ChineseRemainder Theorem (CRT–typically covered in AlgebraicStructures once you’ve gotten to rings).

We won’t state it, but what it does for you is it tells you you onlyhave to look (mod pk) for all primes p and all powers k (sonot–strictly speaking– (mod n) for all n).

Page 75: Have Yourself a Merry Little Christmas (Theorem)

Still, it’s a daunting task to check that there’s a solution (mod n)for all n. One thing we use to simplify life is the ChineseRemainder Theorem (CRT–typically covered in AlgebraicStructures once you’ve gotten to rings).

We won’t state it, but what it does for you is it tells you you onlyhave to look (mod pk) for all primes p and all powers k (sonot–strictly speaking– (mod n) for all n).

Page 76: Have Yourself a Merry Little Christmas (Theorem)

Again, though...there are infinitely many primes! And infinitelymany prime powers!

Consider the following: a solution (mod pk) implies a solution(mod p`) for all ` ≤ k (concretely, for instance, a solution(mod 9) implies a solution (mod 3)). Again, though, it’d beGREAT to work backwards. Wouldn’t it be amazing if for somepower of k , a solution (mod pk) implied a solution (mod p`) for` ≥ k?

Again, we luck out. The most important tool in computationalnumber theory: Hensel’s Lemma (think of it as a numbertheorist’s Newton’s method).

Page 77: Have Yourself a Merry Little Christmas (Theorem)

Again, though...there are infinitely many primes! And infinitelymany prime powers!

Consider the following: a solution (mod pk) implies a solution(mod p`) for all ` ≤ k (concretely, for instance, a solution(mod 9) implies a solution (mod 3)). Again, though, it’d beGREAT to work backwards. Wouldn’t it be amazing if for somepower of k , a solution (mod pk) implied a solution (mod p`) for` ≥ k?

Again, we luck out. The most important tool in computationalnumber theory: Hensel’s Lemma (think of it as a numbertheorist’s Newton’s method).

Page 78: Have Yourself a Merry Little Christmas (Theorem)

Again, though...there are infinitely many primes! And infinitelymany prime powers!

Consider the following: a solution (mod pk) implies a solution(mod p`) for all ` ≤ k (concretely, for instance, a solution(mod 9) implies a solution (mod 3)). Again, though, it’d beGREAT to work backwards. Wouldn’t it be amazing if for somepower of k , a solution (mod pk) implied a solution (mod p`) for` ≥ k?

Again, we luck out. The most important tool in computationalnumber theory: Hensel’s Lemma (think of it as a numbertheorist’s Newton’s method).

Page 79: Have Yourself a Merry Little Christmas (Theorem)

Applying Hensel just to x2 + y2 − p = 0 for p ≡ 1 (mod 4) primesays to guarantee a solution (mod qk) for all primes q it sufficesto show a solution exists:

I (mod q3) for q = 2.Not bad: p ≡ 1 (mod 4)⇒ p ≡ 1, 5 (mod 8). 02 + 12 ≡ 1(mod 8) and 22 + 12 ≡ 5 (mod 8).

I (mod q) for q odd.Slightly more delicate, but it’s one of CMU’s favorite kinds ofproofs: contradiction.

Page 80: Have Yourself a Merry Little Christmas (Theorem)

Applying Hensel just to x2 + y2 − p = 0 for p ≡ 1 (mod 4) primesays to guarantee a solution (mod qk) for all primes q it sufficesto show a solution exists:

I (mod q3) for q = 2.

Not bad: p ≡ 1 (mod 4)⇒ p ≡ 1, 5 (mod 8). 02 + 12 ≡ 1(mod 8) and 22 + 12 ≡ 5 (mod 8).

I (mod q) for q odd.Slightly more delicate, but it’s one of CMU’s favorite kinds ofproofs: contradiction.

Page 81: Have Yourself a Merry Little Christmas (Theorem)

Applying Hensel just to x2 + y2 − p = 0 for p ≡ 1 (mod 4) primesays to guarantee a solution (mod qk) for all primes q it sufficesto show a solution exists:

I (mod q3) for q = 2.Not bad: p ≡ 1 (mod 4)⇒ p ≡ 1, 5 (mod 8). 02 + 12 ≡ 1(mod 8) and 22 + 12 ≡ 5 (mod 8).

I (mod q) for q odd.Slightly more delicate, but it’s one of CMU’s favorite kinds ofproofs: contradiction.

Page 82: Have Yourself a Merry Little Christmas (Theorem)

Applying Hensel just to x2 + y2 − p = 0 for p ≡ 1 (mod 4) primesays to guarantee a solution (mod qk) for all primes q it sufficesto show a solution exists:

I (mod q3) for q = 2.Not bad: p ≡ 1 (mod 4)⇒ p ≡ 1, 5 (mod 8). 02 + 12 ≡ 1(mod 8) and 22 + 12 ≡ 5 (mod 8).

I (mod q) for q odd.

Slightly more delicate, but it’s one of CMU’s favorite kinds ofproofs: contradiction.

Page 83: Have Yourself a Merry Little Christmas (Theorem)

Applying Hensel just to x2 + y2 − p = 0 for p ≡ 1 (mod 4) primesays to guarantee a solution (mod qk) for all primes q it sufficesto show a solution exists:

I (mod q3) for q = 2.Not bad: p ≡ 1 (mod 4)⇒ p ≡ 1, 5 (mod 8). 02 + 12 ≡ 1(mod 8) and 22 + 12 ≡ 5 (mod 8).

I (mod q) for q odd.Slightly more delicate, but it’s one of CMU’s favorite kinds ofproofs: contradiction.

Page 84: Have Yourself a Merry Little Christmas (Theorem)

And that basically finishes the analytic proof.

1. The Hasse principle applies to x2 + y2 − p = 0, so if solutionsexist (mod n) for all n, then a solution exists in Z.

2. The Chinese Remainder Theorem says instead of looking(mod n) for all n, you can just look at prime powers.

3. Hensel’s Lemma tells you the prime powers you really careabout are 8 (where there’s a solution) and q (for q odd, whereagain there’s a solution).

4. QED

Page 85: Have Yourself a Merry Little Christmas (Theorem)

And that basically finishes the analytic proof.

1. The Hasse principle applies to x2 + y2 − p = 0, so if solutionsexist (mod n) for all n, then a solution exists in Z.

2. The Chinese Remainder Theorem says instead of looking(mod n) for all n, you can just look at prime powers.

3. Hensel’s Lemma tells you the prime powers you really careabout are 8 (where there’s a solution) and q (for q odd, whereagain there’s a solution).

4. QED

Page 86: Have Yourself a Merry Little Christmas (Theorem)

And that basically finishes the analytic proof.

1. The Hasse principle applies to x2 + y2 − p = 0, so if solutionsexist (mod n) for all n, then a solution exists in Z.

2. The Chinese Remainder Theorem says instead of looking(mod n) for all n, you can just look at prime powers.

3. Hensel’s Lemma tells you the prime powers you really careabout are 8 (where there’s a solution) and q (for q odd, whereagain there’s a solution).

4. QED

Page 87: Have Yourself a Merry Little Christmas (Theorem)

And that basically finishes the analytic proof.

1. The Hasse principle applies to x2 + y2 − p = 0, so if solutionsexist (mod n) for all n, then a solution exists in Z.

2. The Chinese Remainder Theorem says instead of looking(mod n) for all n, you can just look at prime powers.

3. Hensel’s Lemma tells you the prime powers you really careabout are 8 (where there’s a solution) and q (for q odd, whereagain there’s a solution).

4. QED

Page 88: Have Yourself a Merry Little Christmas (Theorem)

“Obviously” the downside to the analytic techniques is that they’rehard! But good news/bad news: they generalize most easily (atleast to other polynomials, and to higher numbers of variables).

For those really intrigued, in higher dimensions the buzz-word is“modular forms”.

Page 89: Have Yourself a Merry Little Christmas (Theorem)

“Obviously” the downside to the analytic techniques is that they’rehard! But good news/bad news: they generalize most easily (atleast to other polynomials, and to higher numbers of variables).

For those really intrigued, in higher dimensions the buzz-word is“modular forms”.

Page 90: Have Yourself a Merry Little Christmas (Theorem)

Where Do We Go From Here?

It may be surprising in the first place that a theorem firstannounced in 1640 had a new proof published as late as 2016.This is one increasingly “big deal” in number theory: showing thatcertain techniques work is crucial–especially for those trying tosolve still-open problems.

But still, what generalizations could there possibly be of this 1640result?

Page 91: Have Yourself a Merry Little Christmas (Theorem)

Where Do We Go From Here?

It may be surprising in the first place that a theorem firstannounced in 1640 had a new proof published as late as 2016.This is one increasingly “big deal” in number theory: showing thatcertain techniques work is crucial–especially for those trying tosolve still-open problems.

But still, what generalizations could there possibly be of this 1640result?

Page 92: Have Yourself a Merry Little Christmas (Theorem)

You could see the Christmas Theorem as saying “Withfinitely-many exceptions (p = 2), all the primes represented byx2 + y2 can be determined by congruence conditions.” Viewingthings that way, a natural question to ask would be “Does thishold for polynomials beyond x2 + y2?”

Theorem (Clark, Hicks, Parshall, T., 2013)

There are precisely 2779 polynomials of the form ax2 + bxy + cy2

(where b2 − 4ac < 0) for which–with finitely many (explicit)exceptions–the primes represented can be determined completelyby congruence conditions (also made explicit).

Page 93: Have Yourself a Merry Little Christmas (Theorem)

You could see the Christmas Theorem as saying “Withfinitely-many exceptions (p = 2), all the primes represented byx2 + y2 can be determined by congruence conditions.” Viewingthings that way, a natural question to ask would be “Does thishold for polynomials beyond x2 + y2?”

Theorem (Clark, Hicks, Parshall, T., 2013)

There are precisely 2779 polynomials of the form ax2 + bxy + cy2

(where b2 − 4ac < 0) for which–with finitely many (explicit)exceptions–the primes represented can be determined completelyby congruence conditions (also made explicit).

Page 94: Have Yourself a Merry Little Christmas (Theorem)

Where else could you go?

I Up the number of variables.

I Can you say how often a prime would be represented?

I Forget about primes. Look at full congruences classes. Lookat integers to avoid as well as integers to represent. Look atarithmetic progressions.

I Change the playing field (almost literally). Instead ofconsidering polynomials over Z, what about polynomials inZ[i ] (or other rings of integers)? What about over polynomialsrings (going meta and looking at polynomials of polynomials)?

Page 95: Have Yourself a Merry Little Christmas (Theorem)

Where else could you go?

I Up the number of variables.

I Can you say how often a prime would be represented?

I Forget about primes. Look at full congruences classes. Lookat integers to avoid as well as integers to represent. Look atarithmetic progressions.

I Change the playing field (almost literally). Instead ofconsidering polynomials over Z, what about polynomials inZ[i ] (or other rings of integers)? What about over polynomialsrings (going meta and looking at polynomials of polynomials)?

Page 96: Have Yourself a Merry Little Christmas (Theorem)

Where else could you go?

I Up the number of variables.

I Can you say how often a prime would be represented?

I Forget about primes. Look at full congruences classes. Lookat integers to avoid as well as integers to represent. Look atarithmetic progressions.

I Change the playing field (almost literally). Instead ofconsidering polynomials over Z, what about polynomials inZ[i ] (or other rings of integers)? What about over polynomialsrings (going meta and looking at polynomials of polynomials)?

Page 97: Have Yourself a Merry Little Christmas (Theorem)

Where else could you go?

I Up the number of variables.

I Can you say how often a prime would be represented?

I Forget about primes. Look at full congruences classes. Lookat integers to avoid as well as integers to represent. Look atarithmetic progressions.

I Change the playing field (almost literally). Instead ofconsidering polynomials over Z, what about polynomials inZ[i ] (or other rings of integers)? What about over polynomialsrings (going meta and looking at polynomials of polynomials)?

Page 98: Have Yourself a Merry Little Christmas (Theorem)

Thank you.