The weights of the spinor module #
A polarization of a quadratic space (V, Q) splits it as W ⊕ W' ⊕ L with W and W'
isotropic and in perfect QuadraticMap.polar-pairing, and TauCeti.spinAction makes the exterior
algebra S = ⋀·W a module over CliffordAlgebra Q. This file diagonalizes S.
Fix a basis w : ι → W. The pairing turns each coordinate functional of that basis into a vector
w' i of W', and the Clifford bivector
H i = bivector Q (w i) (w' i) = ⅟2 (ι (w i) * ι (w' i) - ι (w' i) * ι (w i))
is a quadratic element of the Clifford algebra, so it lies in the Lie subalgebra
CliffordAlgebra.quadraticLieSubalgebra Q that realizes 𝔰𝔬(V, Q). These elements commute
(TauCeti.SpinPolarizationData.lie_diagonalBivector_diagonalBivector), and they act diagonally
on the exterior basis of S: on the basis vector indexed by
a finite set s of indices — the wedge of the w i for i ∈ s — the element H i acts by
spinWeight K s i = if i ∈ s then ⅟2 else -⅟2.
So the weight of that basis vector is the half-integer sign vector ½(±1, …, ±1), the signs
recording which basis vectors of W occur. Half-integrality is recorded literally, in
TauCeti.spinWeight_add_self_of_mem and TauCeti.spinWeight_add_self_of_notMem: twice a weight
is ±1.
Two facts pin the diagonalization down. The simultaneous eigenspace at the eigenvalue tuple
spinWeight K s is exactly the line spanned by the corresponding exterior basis vector
(TauCeti.spinWeightSpace_spinWeight), and, over a ring without zero divisors, every other tuple
of eigenvalues has zero simultaneous eigenspace
(TauCeti.spinWeightSpace_eq_bot_of_notMem_range). Hence, for a finite index type, the tuples
that do occur are exactly the sign vectors (TauCeti.range_spinWeight), and, when K is
nontrivial, there are 2 ^ l of them on an index type of cardinality l
(TauCeti.ncard_range_spinWeight); their lines exhaust
S (TauCeti.iSup_spinWeightSpace_eq_top).
The cancellation behind the first of those needs no field: two distinct sign vectors differ in
some coordinate by ±1, a unit, so a coefficient annihilated by that difference vanishes over any
commutative ring in which 2 is invertible.
What is deliberately absent: the elements H i are only shown to commute, and no Lie subalgebra
of CliffordAlgebra.quadraticLieSubalgebra Q is exhibited as a Cartan subalgebra, so "weight"
below always means a tuple of simultaneous eigenvalues for the indexed family H, not a linear
character of a Cartan subalgebra. No ordering of ι is used to single out a Borel, so no weight
is called highest here; the elements H i are not compared with a matrix model of 𝔰𝔬(2l + 1)
or 𝔰𝔬(2l); and the abstract Bₗ/Dₗ root data are not mentioned. Those comparisons are the
rest of the layer this file opens.
Main definitions #
TauCeti.SpinPolarizationData.diagonalBivector: the Clifford bivectorH iof a basis vector ofWand its dual vector inW'.TauCeti.spinWeight: the sign vector½(±1, …, ±1)attached to a finite set of indices.TauCeti.spinWeightSpace: the simultaneous eigenspace of the elementsH iat a prescribed tuple of eigenvalues.
Main results #
TauCeti.SpinPolarizationData.lie_diagonalBivector_diagonalBivector: the diagonal bivectors commute.TauCeti.SpinPolarizationData.spinAction_diagonalBivector_basis: the diagonal bivectors act diagonally on the exterior basis, by⅟2or-⅟2.TauCeti.SpinPolarizationData.spinAction_two_smul_diagonalBivector_basisandTauCeti.SpinPolarizationData.spinAction_diagonalBivector_sub_diagonalBivector_basis: twice a diagonal bivector acts by±1, and a difference of two of them by a difference of two weights. These are the two combinations that a coroot of an orthogonal Lie algebra realizes.TauCeti.spinWeightSpace_spinWeight: each weight space is the line spanned by its exterior basis vector.TauCeti.spinWeightSpace_eq_bot_of_notMem_rangeandTauCeti.spinWeightSpace_ne_bot_iff: no other tuple of eigenvalues occurs.TauCeti.iSup_spinWeightSpace_eq_top: the weight spaces exhaust the spinor module.TauCeti.range_spinWeight: for a finite index type the weights are exactly the sign vectors, and, whenKis nontrivial,TauCeti.ncard_range_spinWeightcounts them: there are2 ^ lof them on an index type of cardinalityl.TauCeti.ncard_image_spinWeight_evenandTauCeti.ncard_image_spinWeight_odd: on a nonempty index type the two parities of sign vector are equinumerous,2 ^ (l - 1)each. On the empty index type they are not: the only sign vector is the empty one, which is even, so the even count is2 ^ (0 - 1) = 1and the odd one is0, which is why the odd theorem asksl > 0. These are the weights of the two half-spin summands; the identification isTauCeti/RepresentationTheory/Spin/HalfSpin/Weight.lean.
References #
- W. Fulton and J. Harris, Representation Theory: A First Course (1991), §20.1: the weights
½(±1, …, ±1)of the spin moduleS = ⋀·W. - H. B. Lawson and M.-L. Michelsohn, Spin Geometry (1989), Chapter I §5.
- Spin-representations roadmap, Layer 5, "Weights of the spin module".
The sign-vector weights #
The weight of the exterior basis vector indexed by a finite set s: the half-integer
sign vector ½(±1, …, ±1) whose i-th sign records whether i occurs in s.
Instances For
The weights are half-integral: twice the weight at an occupied index is 1.
The weights are half-integral: twice the weight at a vacant index is -1.
Distinct index sets have weights that differ by a unit in some coordinate. This is the cancellation making the weight spaces one-dimensional over an arbitrary commutative ring.
Distinct index sets carry distinct weights.
The values of TauCeti.spinWeight are exactly the sign vectors ½(±1, …, ±1), when the
index type is finite. Together with TauCeti.spinWeightSpace_ne_bot_iff this says the tuples of
eigenvalues occurring in the spinor module are exactly those sign vectors;
TauCeti.ncard_range_spinWeight counts them.
There are exactly 2 ^ l sign vectors on a finite index type of cardinality l: distinct
index sets carry distinct weights, and there are 2 ^ l index sets.
The sign vectors with an even number of + signs number 2 ^ (l - 1) on an index type of
cardinality l. For l > 0 that is half of the 2 ^ l sign vectors of
TauCeti.ncard_range_spinWeight, the odd ones of TauCeti.ncard_image_spinWeight_odd being the
other half; for l = 0 it is no half but the single empty sign vector, which is even, and
2 ^ (0 - 1) = 1 counts it. These are the weights of the even half-spin summand, by
TauCeti.setOf_spinWeightSpace_ne_bot_and_le_spinPlus_eq_image_spinWeight_even.
The sign vectors with an odd number of + signs number 2 ^ (l - 1), the other half of
TauCeti.ncard_image_spinWeight_even, on a nonempty index type: the empty index type has no sign
vector of odd parity. These are the weights of the odd half-spin summand, by
TauCeti.setOf_spinWeightSpace_ne_bot_and_le_spinMinus_eq_image_spinWeight_odd.
The diagonal bivectors #
The i-th diagonal bivector of a polarization with a chosen basis of its isotropic
summand: the Clifford bivector of the i-th basis vector and its dual.
Equations
- P.diagonalBivector b i = CliffordAlgebra.bivector Q ↑(b i) ↑(P.dualVector b i)
Instances For
The diagonal bivectors are quadratic elements of the Clifford algebra, so they lie in the Lie
subalgebra that realizes 𝔰𝔬(V, Q).
The diagonal bivectors commute. Of the four polar values entering the bracket of two Clifford bivectors only the two Kronecker pairings survive, and off the diagonal they vanish too; on the diagonal the two surviving contributions cancel.
The diagonal action on the exterior basis #
Creation after annihilation at the same index is the occupation projection: it fixes the exterior basis vectors containing that index and kills the others.
Annihilation after creation at the same index is the complementary vacancy projection, by the creation–annihilation anticommutator and the Kronecker pairing.
The diagonal bivectors act diagonally on the exterior basis of the spinor module: the i-th
one multiplies the basis vector indexed by s by ⅟2 when i ∈ s and by -⅟2 otherwise, so
that basis vector is a weight vector of weight TauCeti.spinWeight K s.
Twice a diagonal bivector acts on the exterior basis by ±1, according to whether its
coordinate is occupied: the spin weights are half-integral, so their doubles are units. This is
the combination realized by the short coroot of an odd orthogonal Lie algebra.
A difference of two diagonal bivectors acts on the exterior basis by the difference of the
two spin weights, which is 0 or ±1. This is the combination realized by the long coroot of
an orthogonal Lie algebra.
The weight-space decomposition #
The weight space of the spinor module at a tuple χ of prescribed eigenvalues: the
simultaneous eigenspace of the operators H i, the i-th one acting with eigenvalue χ i.
Equations
- TauCeti.spinWeightSpace Q P b χ = ⨅ (i : ι), ((TauCeti.spinAction Q P) (P.diagonalBivector b i)).eigenspace (χ i)
Instances For
Membership in a weight space, unfolded: a spinor lies in it exactly when every H i scales
it by the prescribed eigenvalue χ i.
The exterior basis vector indexed by s lies in the weight space at spinWeight K s.
The diagonal bivectors are diagonal in the exterior basis, read on coordinates: applying H i
scales the t-th coordinate of any spinor by the i-th entry of the weight of t.
Each weight space of the spinor module is a line, spanned by the exterior basis vector carrying that weight. Distinct sign vectors differ somewhere by a unit, which is what forces every other coordinate of a weight vector to vanish.
No tuple of eigenvalues other than a sign vector occurs in the spinor module, over a ring without zero divisors.
The weights of the spinor module are exactly the sign vectors: a tuple of eigenvalues has
a nonzero simultaneous eigenspace precisely when it is one of the TauCeti.spinWeight K s.
The weight spaces exhaust the spinor module: the lines carrying the sign vectors span
S = ⋀·W.