<<

Embedding Quadratic Fields into Algebras

Rose Mintzer-Sweeney, Alexander Schlesinger, Katherine Xiu

August 4, 2016

1 / 24 Overview

1 Background and Definitions

2 p-adic Analysis

3 Limits and a Conjecture

4 Sieving and Bounding

5 Further Questions

2 / 24 Quadratic Fields

Definition A quadratic field is the set

√ √ Q( k) = {a + b k : a, b ∈ Q}

with the usual addition and multiplication. Here k is a squarefree other than 0 or 1.

3 / 24 i 2 = m, j2 = n, ij = −ji. Here m, n are again squarefree other than 0 or 1.

ij is often written as k.

Rational Quaternion Algebras

Definition A is the set of all elements

m, n = {a + bi + cj + dij : a, b, c, d ∈ Q} Q with component-wise addition and multiplication defined by

4 / 24 j2 = n, ij = −ji. Here m, n are again squarefree integers other than 0 or 1.

ij is often written as k.

Rational Quaternion Algebras

Definition A quaternion algebra is the set of all elements

m, n = {a + bi + cj + dij : a, b, c, d ∈ Q} Q with component-wise addition and multiplication defined by i 2 = m,

4 / 24 ij = −ji. Here m, n are again squarefree integers other than 0 or 1.

ij is often written as k.

Rational Quaternion Algebras

Definition A quaternion algebra is the set of all elements

m, n = {a + bi + cj + dij : a, b, c, d ∈ Q} Q with component-wise addition and multiplication defined by i 2 = m, j2 = n,

4 / 24 Here m, n are again squarefree integers other than 0 or 1.

ij is often written as k.

Rational Quaternion Algebras

Definition A quaternion algebra is the set of all elements

m, n = {a + bi + cj + dij : a, b, c, d ∈ Q} Q with component-wise addition and multiplication defined by i 2 = m, j2 = n, ij = −ji.

4 / 24 ij is often written as k.

Rational Quaternion Algebras

Definition A quaternion algebra is the set of all elements

m, n = {a + bi + cj + dij : a, b, c, d ∈ Q} Q with component-wise addition and multiplication defined by i 2 = m, j2 = n, ij = −ji. Here m, n are again squarefree integers other than 0 or 1.

4 / 24 Rational Quaternion Algebras

Definition A quaternion algebra is the set of all elements

m, n = {a + bi + cj + dij : a, b, c, d ∈ Q} Q with component-wise addition and multiplication defined by i 2 = m, j2 = n, ij = −ji. Here m, n are again squarefree integers other than 0 or 1.

ij is often written as k.

4 / 24 Example   A more unusual example is 2,−3 . Here i 2 = 2, j2 = −3, and (ij)2 = 6. Q

Examples of Rational Quaternion Algebras

Example   The classic example is −1,−1 , the Hamiltonian . Here Q i 2 = j2 = (ij)2 = −1.

5 / 24 Examples of Rational Quaternion Algebras

Example   The classic example is −1,−1 , the Hamiltonian quaternions. Here Q i 2 = j2 = (ij)2 = −1.

Example   A more unusual example is 2,−3 . Here i 2 = 2, j2 = −3, and (ij)2 = 6. Q

5 / 24 For fixed k, we are interested in knowing for which (m, n) there is such an embedding.

Embeddings

Definition √   We say ( k) embeds into a quaternion algebra m,n if there exists an Q Q √   injective ring φ : ( k) ,→ m,n . Q Q

6 / 24 Embeddings

Definition √   We say ( k) embeds into a quaternion algebra m,n if there exists an Q Q √   injective ring homomorphism φ : ( k) ,→ m,n . Q Q

For fixed k, we are interested in knowing for which (m, n) there is such an embedding.

6 / 24 Example

√ −1, −1 Q( −1) → √ Q a + b −1 7→ a + bi

Example

√ 2, −3 Q( −1) → √ Q a + b −1 7→ a + b(i + j)

Examples of Embeddings

Let k = −1.

7 / 24 Example

√ 2, −3 Q( −1) → √ Q a + b −1 7→ a + b(i + j)

Examples of Embeddings

Let k = −1. Example

√ −1, −1 Q( −1) → √ Q a + b −1 7→ a + bi

7 / 24 Examples of Embeddings

Let k = −1. Example

√ −1, −1 Q( −1) → √ Q a + b −1 7→ a + bi

Example

√ 2, −3 Q( −1) → √ Q a + b −1 7→ a + b(i + j)

7 / 24   The embedding exists iff ∃ω ∈ m,n such that ω2 = k. Q √ Given ω, the corresponding map is φ : a + b k → a + bω. The other direction is similar.

An Equivalent Condition

There is an equivalent, easier to check condition for the existence of φ.

8 / 24 √ Given ω, the corresponding map is φ : a + b k → a + bω. The other direction is similar.

An Equivalent Condition

There is an equivalent, easier to check condition for the existence of φ.   The embedding exists iff ∃ω ∈ m,n such that ω2 = k. Q

8 / 24 The other direction is similar.

An Equivalent Condition

There is an equivalent, easier to check condition for the existence of φ.   The embedding exists iff ∃ω ∈ m,n such that ω2 = k. Q √ Given ω, the corresponding map is φ : a + b k → a + bω.

8 / 24 An Equivalent Condition

There is an equivalent, easier to check condition for the existence of φ.   The embedding exists iff ∃ω ∈ m,n such that ω2 = k. Q √ Given ω, the corresponding map is φ : a + b k → a + bω. The other direction is similar.

8 / 24 Write ω = xi + yj + zij. We can check that

−k = −ω2 = mx2 + ny 2 + mnz2

Clearing denominators, we see that there exists ω such that ω2 = k iff

kW 2 − mX 2 − nY 2 + mnZ 2 = 0

for some W , X , Y , Z ∈ Z not all zero.

Reduction to Quadratic Forms

One can check that for any ω such that ω2 = k, Re(ω) = 0.

9 / 24 −k = −ω2 = mx2 + ny 2 + mnz2

Clearing denominators, we see that there exists ω such that ω2 = k iff

kW 2 − mX 2 − nY 2 + mnZ 2 = 0

for some W , X , Y , Z ∈ Z not all zero.

Reduction to Quadratic Forms

One can check that for any ω such that ω2 = k, Re(ω) = 0. Write ω = xi + yj + zij. We can check that

9 / 24 Clearing denominators, we see that there exists ω such that ω2 = k iff

kW 2 − mX 2 − nY 2 + mnZ 2 = 0

for some W , X , Y , Z ∈ Z not all zero.

Reduction to Quadratic Forms

One can check that for any ω such that ω2 = k, Re(ω) = 0. Write ω = xi + yj + zij. We can check that

−k = −ω2 = mx2 + ny 2 + mnz2

9 / 24 kW 2 − mX 2 − nY 2 + mnZ 2 = 0

for some W , X , Y , Z ∈ Z not all zero.

Reduction to Quadratic Forms

One can check that for any ω such that ω2 = k, Re(ω) = 0. Write ω = xi + yj + zij. We can check that

−k = −ω2 = mx2 + ny 2 + mnz2

Clearing denominators, we see that there exists ω such that ω2 = k iff

9 / 24 Reduction to Quadratic Forms

One can check that for any ω such that ω2 = k, Re(ω) = 0. Write ω = xi + yj + zij. We can check that

−k = −ω2 = mx2 + ny 2 + mnz2

Clearing denominators, we see that there exists ω such that ω2 = k iff

kW 2 − mX 2 − nY 2 + mnZ 2 = 0

for some W , X , Y , Z ∈ Z not all zero.

9 / 24 Theorem (Hasse-Minkowski) Given an integer quadratic form Q(x), Q(x) = 0 has non-zero integer solutions if and only if Q(x) = 0 has non-zero real solutions, and

Q(x) = 0 has non-zero solutions in Zp for every prime p. Checking whether there are real solutions is easy.

Hasse-Minkowski

There are well-developed tools for handling a problem like this. We depend on two main theorems.

10 / 24 Q(x) = 0 has non-zero real solutions, and

Q(x) = 0 has non-zero solutions in Zp for every prime p. Checking whether there are real solutions is easy.

Hasse-Minkowski

There are well-developed tools for handling a problem like this. We depend on two main theorems. Theorem (Hasse-Minkowski) Given an integer quadratic form Q(x), Q(x) = 0 has non-zero integer solutions if and only if

10 / 24 Q(x) = 0 has non-zero solutions in Zp for every prime p. Checking whether there are real solutions is easy.

Hasse-Minkowski

There are well-developed tools for handling a problem like this. We depend on two main theorems. Theorem (Hasse-Minkowski) Given an integer quadratic form Q(x), Q(x) = 0 has non-zero integer solutions if and only if Q(x) = 0 has non-zero real solutions, and

10 / 24 Checking whether there are real solutions is easy.

Hasse-Minkowski

There are well-developed tools for handling a problem like this. We depend on two main theorems. Theorem (Hasse-Minkowski) Given an integer quadratic form Q(x), Q(x) = 0 has non-zero integer solutions if and only if Q(x) = 0 has non-zero real solutions, and

Q(x) = 0 has non-zero solutions in Zp for every prime p.

10 / 24 Hasse-Minkowski

There are well-developed tools for handling a problem like this. We depend on two main theorems. Theorem (Hasse-Minkowski) Given an integer quadratic form Q(x), Q(x) = 0 has non-zero integer solutions if and only if Q(x) = 0 has non-zero real solutions, and

Q(x) = 0 has non-zero solutions in Zp for every prime p. Checking whether there are real solutions is easy.

10 / 24 Theorem (Hensel’s Lemma) n Let P(x) be an integer . Suppose P(x0) = 0 mod p and 0 P (x0) 6= 0 mod p. Then ∃x˜0 ∈ Zp such that P(x ˜0) = 0.

So, we can lift solutions modulo p into solutions in Zp.

Hensel’s Lemma

Checking whether there are p-adic solutions is also tractable.

11 / 24 So, we can lift solutions modulo p into solutions in Zp.

Hensel’s Lemma

Checking whether there are p-adic solutions is also tractable. Theorem (Hensel’s Lemma) n Let P(x) be an integer polynomial. Suppose P(x0) = 0 mod p and 0 P (x0) 6= 0 mod p. Then ∃x˜0 ∈ Zp such that P(x ˜0) = 0.

11 / 24 Hensel’s Lemma

Checking whether there are p-adic solutions is also tractable. Theorem (Hensel’s Lemma) n Let P(x) be an integer polynomial. Suppose P(x0) = 0 mod p and 0 P (x0) 6= 0 mod p. Then ∃x˜0 ∈ Zp such that P(x ˜0) = 0.

So, we can lift solutions modulo p into solutions in Zp.

11 / 24 Legendre Symbol

Recall that: Definition For an odd prime p and an integer x, the Legendre symbol is defined by

 0 if p | x x   = 1 if ∃y s.t. x ≡ y 2 mod p . p −1 otherwise

12 / 24 1 (Conditions from R-analysis) If m < 0 and n < 0, then k < 0.   2 k (Conditions from Zp, p 6= 2) If p - k is an odd prime and p = 1, then  −m n  p · p If p | m, p | n, then p = 1.

 n  If p | m, p - n, then p = 1.  m  If p - m, p | n, then p = 1. 3 (Conditions from Z2) Omitted for the sake of brevity.

Conditions

Theorem (Ehrman, Mintzer-Sweeney, Schlesinger, Sheydvasser, Xiu 2016) √   There exists an embedding ( k) ,→ m,n iff the following conditions Q Q are met

13 / 24   2 k (Conditions from Zp, p 6= 2) If p - k is an odd prime and p = 1, then  −m n  p · p If p | m, p | n, then p = 1.

 n  If p | m, p - n, then p = 1.  m  If p - m, p | n, then p = 1. 3 (Conditions from Z2) Omitted for the sake of brevity.

Conditions

Theorem (Ehrman, Mintzer-Sweeney, Schlesinger, Sheydvasser, Xiu 2016) √   There exists an embedding ( k) ,→ m,n iff the following conditions Q Q are met

1 (Conditions from R-analysis) If m < 0 and n < 0, then k < 0.

13 / 24  −m n  p · p If p | m, p | n, then p = 1.

 n  If p | m, p - n, then p = 1.  m  If p - m, p | n, then p = 1. 3 (Conditions from Z2) Omitted for the sake of brevity.

Conditions

Theorem (Ehrman, Mintzer-Sweeney, Schlesinger, Sheydvasser, Xiu 2016) √   There exists an embedding ( k) ,→ m,n iff the following conditions Q Q are met

1 (Conditions from R-analysis) If m < 0 and n < 0, then k < 0.   2 k (Conditions from Zp, p 6= 2) If p - k is an odd prime and p = 1, then

13 / 24  n  If p | m, p - n, then p = 1.  m  If p - m, p | n, then p = 1. 3 (Conditions from Z2) Omitted for the sake of brevity.

Conditions

Theorem (Ehrman, Mintzer-Sweeney, Schlesinger, Sheydvasser, Xiu 2016) √   There exists an embedding ( k) ,→ m,n iff the following conditions Q Q are met

1 (Conditions from R-analysis) If m < 0 and n < 0, then k < 0.   2 k (Conditions from Zp, p 6= 2) If p - k is an odd prime and p = 1, then  −m n  p · p If p | m, p | n, then p = 1.

13 / 24  m  If p - m, p | n, then p = 1. 3 (Conditions from Z2) Omitted for the sake of brevity.

Conditions

Theorem (Ehrman, Mintzer-Sweeney, Schlesinger, Sheydvasser, Xiu 2016) √   There exists an embedding ( k) ,→ m,n iff the following conditions Q Q are met

1 (Conditions from R-analysis) If m < 0 and n < 0, then k < 0.   2 k (Conditions from Zp, p 6= 2) If p - k is an odd prime and p = 1, then  −m n  p · p If p | m, p | n, then p = 1.

 n  If p | m, p - n, then p = 1.

13 / 24 3 (Conditions from Z2) Omitted for the sake of brevity.

Conditions

Theorem (Ehrman, Mintzer-Sweeney, Schlesinger, Sheydvasser, Xiu 2016) √   There exists an embedding ( k) ,→ m,n iff the following conditions Q Q are met

1 (Conditions from R-analysis) If m < 0 and n < 0, then k < 0.   2 k (Conditions from Zp, p 6= 2) If p - k is an odd prime and p = 1, then  −m n  p · p If p | m, p | n, then p = 1.

 n  If p | m, p - n, then p = 1.  m  If p - m, p | n, then p = 1.

13 / 24 Conditions

Theorem (Ehrman, Mintzer-Sweeney, Schlesinger, Sheydvasser, Xiu 2016) √   There exists an embedding ( k) ,→ m,n iff the following conditions Q Q are met

1 (Conditions from R-analysis) If m < 0 and n < 0, then k < 0.   2 k (Conditions from Zp, p 6= 2) If p - k is an odd prime and p = 1, then  −m n  p · p If p | m, p | n, then p = 1.

 n  If p | m, p - n, then p = 1.  m  If p - m, p | n, then p = 1. 3 (Conditions from Z2) Omitted for the sake of brevity.

13 / 24 Conjecture For any fixed k,

 n √   m,n o # SN ∩ (m, n): Q k ,→ lim Q = 0. N→∞ #SN

Long-Term Behavior

The next logical question is to determine how often these conditions are met. Define

 2 SN = (m, n) ∈ Z : 1 < m, n < N, m, n squarefree

14 / 24 Long-Term Behavior

The next logical question is to determine how often these conditions are met. Define

 2 SN = (m, n) ∈ Z : 1 < m, n < N, m, n squarefree

Conjecture For any fixed k,

 n √   m,n o # SN ∩ (m, n): Q k ,→ lim Q = 0. N→∞ #SN

14 / 24 Note: we are using 1 < m, n < N for simplicity, but the same proof applies to −1 > m, n > −N.

Our Theorem

Theorem (EMSSSX 2016) For any fixed k,  n √   o # S0 ∩ (m, n): k ,→ m,n N Q Q lim 0 = 0, N→∞ #SN where

0  2 SN = (m, n) ∈ Z : 1 < m, n < N, m, n squarefree and coprime .

15 / 24 Our Theorem

Theorem (EMSSSX 2016) For any fixed k,  n √   o # S0 ∩ (m, n): k ,→ m,n N Q Q lim 0 = 0, N→∞ #SN where

0  2 SN = (m, n) ∈ Z : 1 < m, n < N, m, n squarefree and coprime .

Note: we are using 1 < m, n < N for simplicity, but the same proof applies to −1 > m, n > −N.

15 / 24 0 6 3 2 One can check that #SN ≈ π2 N for large N. To show that the limit is 0, we find an upper bound for the numerator that is o(N2). We do this by focusing exclusively on the p-adic conditions on n coming from m.

Our Approach

How can one prove a result like this?

16 / 24 To show that the limit is 0, we find an upper bound for the numerator that is o(N2). We do this by focusing exclusively on the p-adic conditions on n coming from m.

Our Approach

How can one prove a result like this? 0 6 3 2 One can check that #SN ≈ π2 N for large N.

16 / 24 We do this by focusing exclusively on the p-adic conditions on n coming from m.

Our Approach

How can one prove a result like this? 0 6 3 2 One can check that #SN ≈ π2 N for large N. To show that the limit is 0, we find an upper bound for the numerator that is o(N2).

16 / 24 Our Approach

How can one prove a result like this? 0 6 3 2 One can check that #SN ≈ π2 N for large N. To show that the limit is 0, we find an upper bound for the numerator that is o(N2). We do this by focusing exclusively on the p-adic conditions on n coming from m.

16 / 24 Let Y m˜ = p.

 k  p =1 p|m

Definitions for a Sieve Approach

√   Recall that there is an embedding ( k) ,→ m,n only if Q Q

k  n  If p | m, p n, and = 1, then = 1 - p p

17 / 24 Definitions for a Sieve Approach

√   Recall that there is an embedding ( k) ,→ m,n only if Q Q

k  n  If p | m, p n, and = 1, then = 1 - p p

Let Y m˜ = p.

 k  p =1 p|m

17 / 24 Note that n satisfies the conditions for an embedding if and only if

n modm ˜ ∈ A(m ˜)

Thus an upper bound for the numerator of our limit is given by X 1{n modm ˜∈A(m ˜)} 0 (m,n)∈SN

Definitions for a Sieve Approach

Definition The admissible class is the set

x  A(m ˜) := {x ∈ /m˜ : = 1, ∀p | m˜} Z Z p

18 / 24 Thus an upper bound for the numerator of our limit is given by X 1{n modm ˜∈A(m ˜)} 0 (m,n)∈SN

Definitions for a Sieve Approach

Definition The admissible class is the set

x  A(m ˜) := {x ∈ /m˜ : = 1, ∀p | m˜} Z Z p

Note that n satisfies the conditions for an embedding if and only if

n modm ˜ ∈ A(m ˜)

18 / 24 Definitions for a Sieve Approach

Definition The admissible class is the set

x  A(m ˜) := {x ∈ /m˜ : = 1, ∀p | m˜} Z Z p

Note that n satisfies the conditions for an embedding if and only if

n modm ˜ ∈ A(m ˜)

Thus an upper bound for the numerator of our limit is given by X 1{n modm ˜∈A(m ˜)} 0 (m,n)∈SN

18 / 24 Specifically,

m˜ #A(m ˜) ≈ , 2ω(m ˜) where ω(x) is the of distinct prime factors of x.

Approximating the Size of the Admissible Class

This upper bound is useful because we have a good approximation of the size of the admissible class A(m ˜).

19 / 24 Approximating the Size of the Admissible Class

This upper bound is useful because we have a good approximation of the size of the admissible class A(m ˜). Specifically,

m˜ #A(m ˜) ≈ , 2ω(m ˜) where ω(x) is the number of distinct prime factors of x.

19 / 24 Therefore, if we show that

X 1 = o(N), 2ω(m ˜) m

An Upper Bound

Thus,

X X N 1{n modm ˜∈A(m ˜)} ≈ (#A(m ˜)) 0 m˜ (m,n)∈SN m

20 / 24 An Upper Bound

Thus,

X X N 1{n modm ˜∈A(m ˜)} ≈ (#A(m ˜)) 0 m˜ (m,n)∈SN m

Therefore, if we show that

X 1 = o(N), 2ω(m ˜) m

20 / 24 So, we split our sum into two pieces: one where ω is close to the average, and one where it is not.

Definition

  1 loglog(N) B(N) = m < N : ω(m ˜) − loglog(N) < 2 4 B(N)c = {m < N : m ∈/ B(N)}

So,

X 1 X 1 X 1 = + . 2ω(m ˜) 2ω(m ˜) 2ω(m ˜) m

Splitting the Sum

Note that, on average, ω(x) ≈ log log(x).

21 / 24 Definition

  1 loglog(N) B(N) = m < N : ω(m ˜) − loglog(N) < 2 4 B(N)c = {m < N : m ∈/ B(N)}

So,

X 1 X 1 X 1 = + . 2ω(m ˜) 2ω(m ˜) 2ω(m ˜) m

Splitting the Sum

Note that, on average, ω(x) ≈ log log(x). So, we split our sum into two pieces: one where ω is close to the average, and one where it is not.

21 / 24 So,

X 1 X 1 X 1 = + . 2ω(m ˜) 2ω(m ˜) 2ω(m ˜) m

Splitting the Sum

Note that, on average, ω(x) ≈ log log(x). So, we split our sum into two pieces: one where ω is close to the average, and one where it is not.

Definition

  1 loglog(N) B(N) = m < N : ω(m ˜) − loglog(N) < 2 4 B(N)c = {m < N : m ∈/ B(N)}

21 / 24 Splitting the Sum

Note that, on average, ω(x) ≈ log log(x). So, we split our sum into two pieces: one where ω is close to the average, and one where it is not.

Definition

  1 loglog(N) B(N) = m < N : ω(m ˜) − loglog(N) < 2 4 B(N)c = {m < N : m ∈/ B(N)}

So,

X 1 X 1 X 1 = + . 2ω(m ˜) 2ω(m ˜) 2ω(m ˜) m

21 / 24 B(N) is tightly controlled, so it is easy to show that the sum over B(N) is o(N). We can use a result of Granville and Soundararajan (2006)1 to show that #B(N)c is small. It follows that the sum over B(N)c is also o(N). We are done!



An Outline of How to Control the Sums

Why is this splitting useful?

1A. Granville and K. Soundararajan, Sieving and the Erd˝os-KacTheorem 22 / 24 We can use a result of Granville and Soundararajan (2006)1 to show that #B(N)c is small. It follows that the sum over B(N)c is also o(N). We are done!



An Outline of How to Control the Sums

Why is this splitting useful? B(N) is tightly controlled, so it is easy to show that the sum over B(N) is o(N).

1A. Granville and K. Soundararajan, Sieving and the Erd˝os-KacTheorem 22 / 24 It follows that the sum over B(N)c is also o(N). We are done!



An Outline of How to Control the Sums

Why is this splitting useful? B(N) is tightly controlled, so it is easy to show that the sum over B(N) is o(N). We can use a result of Granville and Soundararajan (2006)1 to show that #B(N)c is small.

1A. Granville and K. Soundararajan, Sieving and the Erd˝os-KacTheorem 22 / 24 We are done!



An Outline of How to Control the Sums

Why is this splitting useful? B(N) is tightly controlled, so it is easy to show that the sum over B(N) is o(N). We can use a result of Granville and Soundararajan (2006)1 to show that #B(N)c is small. It follows that the sum over B(N)c is also o(N).

1A. Granville and K. Soundararajan, Sieving and the Erd˝os-KacTheorem 22 / 24 An Outline of How to Control the Sums

Why is this splitting useful? B(N) is tightly controlled, so it is easy to show that the sum over B(N) is o(N). We can use a result of Granville and Soundararajan (2006)1 to show that #B(N)c is small. It follows that the sum over B(N)c is also o(N). We are done!



1A. Granville and K. Soundararajan, Sieving and the Erd˝os-KacTheorem 22 / 24 How small of an upper bound on the number of embeddings can we obtain using sieve methods? Can we find a lower bound on the number of embeddings?

Further Research

How fast does this limit goes to 0?

23 / 24 Can we find a lower bound on the number of embeddings?

Further Research

How fast does this limit goes to 0? How small of an upper bound on the number of embeddings can we obtain using sieve methods?

23 / 24 Further Research

How fast does this limit goes to 0? How small of an upper bound on the number of embeddings can we obtain using sieve methods? Can we find a lower bound on the number of embeddings?

23 / 24 Acknowledgements

We would like to thank Max Ehrman and Senia Sheydvasser for their wonderful mentorship and expert guidance. We would like to thank Sam Payne, Jos´eGonz´alez,and Michael Magee for organizing the SUMRY program. Finally, we would like to thank MathFest for having us!

24 / 24