An introduction to quaternion algebras
Benjamin Linowitz
Department of Mathematics Oberlin College An Introduction to Quaternion Algebras
One of the many nice features of the complex numbers is that their multiplication can be described using the geometric operations of stretching and rotation.
The importance of the complex numbers in two-dimensional geometry was recognized very early.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
William Rowan Hamilton sought a number system which would play an analogous role in three-dimensional geometry.
“Every morning in the early part of the above-cited month [October 1843], on my coming down to breakfast, your (then) little brother William Edwin, and yourself, used to ask me: Well, Papa, can you multiply triplets? Whereto I was always obliged to reply, with a sad shake of the head: No, I can only add and subtract them.”
– Hamilton (in a letter to his son)
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Hamilton never succeeded in finding a three-dimensional real division algebra.
In fact, we now know that no such algebra exists!
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
In more modern language, Hamilton was looking for an associative multiplication law on
∼ 3 D = R + Ri + Rj = R
where i 2 = −1 and every nonzero element of D has an inverse.
Suppose that there are a, b, c ∈ R such that
ij = a + bi + cj.
If we multiply on the left by i we get −j = ai − b + cij. Substituting our formula for ij into this expression yields
0 = (ac − b) + (a + bc)i + (c2 + 1)j.
2 But this means that c = −1, a contradiction as c ∈ R.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Hamilton did, however, find a four dimensional real division algebra.
Theorem (Hamilton, 1843) The R-algebra H with basis {1, i, j, ij} and defining relations
i 2 = −1 j2 = −1 ij = −ji
is a four-dimensional division algebra. William Rowan Hamilton (1805-1865)
Hamilton was so excited by this discovery that he carved these relations into the stone of the Brougham Bridge.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Modern vector analysis was pioneered by Willard Gibbs.
In his treatise Elements of Vector Analysis he introduced the now standard notation of dot products and cross products, and related them to quaternionic multiplication via the beautiful formula
vw = −v · w + v × w,
where v, w ∈ Ri + Rj + Rk ⊂ H.
Gibbs didn’t consider the quaternion product to be a “fundamental notion” in vector analysis though, and argued for a vector analysis that would apply in arbitrary dimension.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
This rivalry between notations flared into a war in the late 1800’s between the ‘quaternionists’ and the ‘vectorists’.
Willard Gibbs (1839-1903)
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
The most prominent quaternionist was James Clerk Maxwell.
“The invention of the calculus of quaternions is a step towards the knowledge of quantities related to space which can only be compared, for its importance, with the invention of triple coordinates by Descartes. The ideas of this calculus, as distinguished from its operations and symbols, are fitted to be of the greatest use in all parts of science.” James Clerk Maxwell (1831-1879)
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
There was also Hamilton’s student Peter Tait:
“Even Prof. Willard Gibbs must be ranked as one the retarders of quaternions progress, in virtue of his pamphlet on Vector Analysis, a sort of hermaphrodite monster, compounded of the notation of Hamilton and Grassman.”
Peter Tait (1831-1901)
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
On the vectorist side was Lord Kelvin:
“Quaternions came from Hamilton after his really good work had been done; and, though beautifully ingenious, have been an unmixed evil to those who have touched them in any way, including Clerk Maxwell.”
William Thomson (Lord Kelvin, 1824-1907)
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Ultimately, the superiority and generality of vector notation won out.
Certain useful fragments of Hamilton’s quaternions do remain in standard calculus courses however.
For example, we still use the “right-hand rule” i × j = k in multivariable calculus.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Let’s write (−1, −1, R) in place of H.
This notation suggests a number of ways to generalize H.
For instance, let (1, 1, R) be the R-algebra with basis {1, i, j, ij} and defining relations
i 2 = 1 j2 = 1 ij = −ji.
0 1 1 0 Then (1, 1, ) =∼ M ( ) via i, j 7→ , . R 2 R 1 0 0 −1 0 −1 1 0 Also, (1, −1, ) =∼ M ( ) via i, j 7→ , . R 2 R 1 0 0 −1
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
More generally, we can define (a, b, R) to be the R-algebra with basis {1, i, j, ij} and defining relations
2 2 ∗ i = a j = b ij = −ji a, b ∈ R .
∼ 2 2 × Because (a, b, R) = (ax , by , R) for all x, y ∈ R , we conclude that ∼ (a, b, R) = H if a, b < 0 and ∼ (a, b, R) = M2(R) otherwise.
Thus (a, b, R) is either a division algebra or else M2(R).
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
There are other ways that we could have generalized (−1, −1, R).
If F is a field of characteristic zero and a, b ∈ F ∗ we can define the generalized quaternion algebra (a, b, F ).
Let F = Q and consider the Q-algebra (−1, −1, Q).
As (−1, −1, Q) ( (−1, −1, R), we see that (−1, −1, Q) is a division algebra.
∼ As before we also see that (1, 1, Q) = M2(Q).
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Recall that if F = R,(a, b, F ) is a division algebra or else M2(R).
Theorem (Wedderburn) For any field F , if the F -algebra (a, b, F ) is not a division algebra then ∼ (a, b, F ) = M2(F ).
Joseph Wedderburn (1882-1948)
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Note that the case F = R is special in that in general there will not be a unique quaternion division algebra over F .
Theorem (Classification of quaternion algebras over Q) Let A = (a, b, Q) be a quaternion algebra over Q. Then the set of primes p such that (a, b, Qp) is a division algebra over Qp is finite with an even number of elements. Conversely, for any finite set S of primes with even cardinality there exists a unique (up to isomorphism) quaternion algebra (a, b, Q) such that (a, b, Qp) precisely when p ∈ S.
Consequence: There are infinitely many quaternion division algebras over Q.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
A “complex” example: F = C.
Let A be a quaternion algebra over C.
By the fundamental theorem of algebra, A cannot be a division algebra.
∼ Then by Wedderburn’s theorem, A = M2(C).
Thus M2(C) is the only quaternion algebra over C.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Over the finite field Fq :
× Let D be a quaternion division algebra over Fq and let D act on D by conjugation.
When we count orbits with the class equation we get X X q4 − 1 |D×| = q4 − 1 = q − 1 + [D× : C (d)×] = q − 1 + . D q2 − 1 d d
Therefore there exists m ∈ Z such that q4 − 1 = q − 1 + m(q2 + 1), a contradiction.
Thus M2(Fq) is the only quaternion algebra over Fq. Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
The reduced norm:
Suppose that A = (a, b, F ) is a quaternion algebra over a field F with char(F ) 6= 2.
Given a quaternion α = t + xi + yj + zk we define an involution on A by α = t − xi − yj − zk.
Note that αα = t2 − ax2 − by 2 + abz2.
When A = (−1, −1, F ) this simplifies to the pleasing
αα = t2 + x2 + y 2 + z2.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
The reduced norm of α ∈ A is defined to be nr(α) = αα.
Note that nr(·) is multiplicative. If α, β ∈ A then
nr(αβ) = αβ = (αβ)(αβ) = αββα = αnr(β)α = ααnr(β) = nr(α)nr(β).
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
An application of the reduced norm:
Theorem (Euler, 1749)
2 2 2 2 2 2 2 2 (a1 + a2 + a3 + a4)(b1 + b2 + b3 + b4) =
2 (a1b1 − a2b2 − a3b3 − a4b4) + 2 (a1b2 + a2b1 + a3b4 − a4b3) + 2 (a1b3 − a2b4 + a3b1 + a4b2) + 2 (a1b4 + a2b3 − a3b2 + a4b1) Leonhard Euler (1707-1783)
Euler’s “four squares identity”’ shows that the product of two sums of four squares is again a sum of four squares.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
An easy proof of Euler’s four squares identity:
Let α, β ∈ (−1, −1, R) be given by
α = a1 + a2i + a3j + a4k,
and β = b1 + b2i + b3j + b4k.
2 2 2 2 2 2 2 2 Then nr(α) = a1 + a2 + a3 + a4 and nr(β) = b1 + b2 + b3 + b4.
The identity now follows from the fact that nr(α)nr(β) = nr(αβ) and the fact that in (−1, −1, R) the reduced norm of an element (i.e., αβ) is always the sum of four squares.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Definition: An order of an algebra over Q is a subring which is also a finitely generated Z-module containing an Q-basis of the algebra.
Example 1: Z[i] is a quadratic order of the Q-algebra Q(i).
Example 2: M2(Z) is a maximal order of M2(Q).
Example 3: The subring
Z[i, j] = Z + Zi + Zj + Zk of the Q-algebra (a, b, Q) is an order provided that a, b ∈ Z.
Remark: If F is a number field with ring of integers OF then all of this generalizes to F -algebras by replacing Q with F and Z with OF . Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
An application to number theory
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Theorem (Lagrange, 1770) Every natural number can be expressed as the sum of four squares.
Joseph-Louis Lagrange (1736-1813)
We’ll use quaternion orders to give a short proof of this theorem.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Let A = (−1, −1, Q).
We’ve already seen that Z[i, j] = Z + Zi + Zj + Zk is an order of A.
Z[i, j] is not a maximal order of A.
√ √ This is analogous to the fact that Z[ −3] is an order of Q( −3) but is not√ the maximal order (i.e., ring of integers) since this is Z[(−1 + 3)/2].
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
√ √ Our analogy with Z[ −3] and Z[(−1 + 3)/2] is not an accident.
Indeed, the element α = i + j + k satisfies α2 + 3 = 0, so it is natural to define −1 + i + j + k ω = , 2 which also satisfies ω2 + ω + 1 = 0.
It turns out that the order
O = Z + Zi + Zj + Zω = Z[i, j] + Z[i, j]ω
is the unique maximal order of A which contains Z[i, j].
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Unit groups:
× ∼ Z[i, j] = {±1, ±i, ±j, ±k} = Q8, the quaternion group of order 8.
× S O = Q8 (±1 ± i ± j ± k)/2, is a group of order 24 isomorphic to SL2(F3).
Easy lemma 1: For all β ∈ O there exists γ ∈ O× such that βγ ∈ Z[i, j].
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Lemma 2: Let p be a prime. There exists π ∈ O such that nr(π) = p.
Proof: If p = 2 then we have nr(1 + i) = 12 + 12 = 2. So suppose p is odd.
∼ ∼ Then O/pO = (−1, −1, Fp) = M2(Fp) and there exists a left ideal I mod p ⊂ O/pO with dimFp (I mod p) = 2.
The preimage of I in O must be principal, hence there exists β ∈ (−1, −1, Q) such I = βO.
There’s a theory of norms for ideals of orders, which says that N(I ) = |O/I | = p2 and that N(βO) = nr(β)2.
Therefore nr(β) = p.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
We now prove Lagrange’s theorem. Let n be a natural number.
It suffices to find an element of α ∈ Z[i, j] with nr(α) = n.
By multiplicativity of the reduced norm it suffices to consider the case n = p is prime.
By Lemma 2 there exists β ∈ O such that nr(β) = p.
× By Lemma 1 there exists γ ∈ O such that α := βγ ∈ Z[i, j].
× × But nr(γ) ∈ Z , and because Z = {±1} and nr(γ) must be a sum of four squares and hence positive, we conclude that nr(γ) = 1 and nr(α) = p.
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Geometric applications
Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras
Theorem (Siegel, 1944)
Given a positive integer d, let hd , Rd denote the class number and regulator of the unique quadratic order of discriminant d. Then
X π2 h R = X 3/2 + O(X log X ). d d 18ζ(3) d Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras On the other hand, if one counts real quadratic fields by regulator instead of discriminant then one can obtain separation. Theorem (Sarnak, 1982) As X → ∞ we have X e2X h ∼ . d 2X Rd Something that makes Sarnak’s proof especially novel is that the main ideas arise from differential geometry. Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras (Rough!) Sketch of Proof Consider the modular surface S = H/SL2(Z). The lengths of closed geodesics on S correspond to regulators of real quadratic fields. These lengths appear with multiplicity equal to the class number. The Prime Geodesic Theorem in this context gives an asymptotic formula for the number of closed geodesics with length less than X . Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras Moral: The lengths of geodesics of certain hyperbolic surfaces encode interesting number theoretic information. Given a Riemannian manifold M, let LS(M) denote the length spectrum of M; that is, LS(M) is the set of lengths of closed geodesics on M. The following information is all determined by LS(M): dim(M) scalar curvature volume(M) arithmeticity Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras Let A = (a, b, Q) be a quaternion algebra over Q with at least one of a, b positive. Let O be a maximal order of A. ∼ Then O ⊂ (a, b, Q) ⊂ (a, b, R) = M2(R). Let Γ(1) denote the subgroup of SL2(R) consisting of those elements of O with reduced norm 1 (equiv., determinant 1). Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras As a subgroup of SL2(R), the group Γ(1) acts on the complex upper half plane H by fractional linear transformations: az + b z 7→ . cz + d The resulting quotient space X (1) = H/Γ(1) is a Shimura Curve. ∼ X (1) is a hyperbolic surface which is compact when A =6 M2(Q). ∼ When A = M2(Q) the resulting Shimura curve is simply the modular surface. Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras Theorem (Vign´eras, 1980) There exist Shimura curves X (1) and Y (1) which are not isomorphic yet satisfy LS(X (1)) = LS(Y (1)). √ 1. Vign´eras’Shimura curves are defined over Q( 10) and arise from the same quaternion algebra. 2. They are also isospectral with respect to the Laplace operator (by the Selberg Trace Formula). Benjamin Linowitz An introduction to quaternion algebras An Introduction to Quaternion Algebras Geometrically, Shimura curves arising from the same quaternion algebra are commensurable. Theorem (Reid, 1992) If X (1) and Y (1) are Shimura curves satisfying LS(X (1)) = LS(Y (1)) then X (1) and Y (1) are commensurable and arise from the same quaternion algebra. Idea of proof: Lengths of geodesics on Shimura curves arise from maximal subfields of the associated quaternion algebra, and a quaternion algebra over a number field is determined by its maximal subfields. Benjamin Linowitz An introduction to quaternion algebras