The Lie subalgebra of a subgroup #
For a subgroup K of a Lie group G the derivations whose one-parameter subgroup never leaves
K,
𝔨 = {X : 𝔤 | ∀ t : ℝ, lieExp (t • X) ∈ K},
are the candidate Lie algebra of K. TauCeti.Lie.lieSubalgebraOfSubgroup K is the Lie
subalgebra of 𝔤 they generate, defined for every subgroup, and this file proves that for a
closed K the generating set is already a Lie subalgebra, so that the generation is inert and
membership is exactly the defining condition
(TauCeti.Lie.mem_lieSubalgebraOfSubgroup).
Addition and the bracket produce the new element as a limit of elements of K, each using a
different piece of the Layer 0/1 machinery; scalars involve no limit at all:
- scalars are free:
t • (c • X) = (t * c) • Xreparametrises the exponential line ofX, so the one-parameter subgroup ofc • Xis a reparametrisation of that ofXand lands inKon the nose; - addition is the Trotter product formula
TauCeti.Lie.tendsto_lieExp_smul_mul_lieExp_smul_pow, which exhibitslieExp (t • (X + Y))as a limit of products of elements ofK; - the bracket is the infinitesimal adjoint. If
XandYboth generate one-parameter subgroups insideKthen so doesAd (lieExp (s • X)) Y, by the conjugation formulaTauCeti.Lie.conj_lieExp; the curves ↦ Ad (lieExp (s • X)) Ytherefore runs inside𝔨and starts atY, and its derivative at0is⁅X, Y⁆byTauCeti.Lie.hasDerivAt_Ad_lieExp_smul_apply_zero, the derivation-level form of the Layer 1 theoremTauCeti.Lie.hasDerivAt_tangentAd_mulInvariantExp_smul_apply_zero. A difference quotient of a curve in a linear subspace stays in that subspace, so⁅X, Y⁆lies in its closure.
Why the definition generates, and closedness #
The definition of lieSubalgebraOfSubgroup is total in K, as the roadmap specifies it, because
it takes the Lie subalgebra generated by 𝔨 rather than asserting that 𝔨 is one. That the
generating set is itself a subalgebra is true for an arbitrary subgroup — for X, Y ∈ 𝔨 the
subgroup generated by the two one-parameter subgroups is arcwise connected, hence an immersed Lie
subgroup of K whose Lie algebra contains X + Y and ⁅X, Y⁆ — but that argument is the
subalgebra ↔ immersed-subgroup correspondence, which the roadmap schedules two layers later
(Layer 4, exists_immersedLieSubgroup_lieSubalgebra) and which is not available here.
What is available without it is the closed case: only the scalar closure survives with no
topological hypothesis, since both the Trotter formula and the difference quotient produce the new
element as a limit of elements of K, never as an element of K on the nose.
TauCeti.Lie.isClosed_setOf_forall_lieExp_smul_mem turns those limits back into membership, and it
is the only place closedness is used. So for a closed K the generation collapses
(TauCeti.Lie.coe_lieSubalgebraOfSubgroup), which is the hypothesis under which the next Layer 2
target — the closed-subgroup theorem, promoting K to an embedded Lie subgroup with
Lie(K) = 𝔨 — is stated.
Main definitions #
TauCeti.Lie.lieSubalgebraOfSubgroup: the Lie algebra of a subgroup.
Main results #
TauCeti.Lie.lieExp_smul_add_mem,TauCeti.Lie.lieExp_smul_smul_memandTauCeti.Lie.lieExp_smul_lie_mem: the three closure properties, stated on the defining condition itself so that they can be used without producing the bundled subalgebra.TauCeti.Lie.hasDerivAt_Ad_lieExp_smul_apply_zero: the adjoint orbits ↦ Ad (lieExp (s • X)) Yhas derivative⁅X, Y⁆at the origin. This is the derivation-level infinitesimal adjoint, the analytic input to bracket closure.TauCeti.Lie.finiteDimensional_lieSubalgebraOfSubgroup: finite-dimensional subgroup models.TauCeti.Lie.lieSubalgebraOfSubgroup_le_iff: the universal property of the generated subalgebra, for an arbitrary subgroup.TauCeti.Lie.lieSubalgebraOfSubgroup_induction: induction on the generated subalgebra, for an arbitrary subgroup.TauCeti.Lie.mem_lieSubalgebraOfSubgroupandTauCeti.Lie.coe_lieSubalgebraOfSubgroup: for a closed subgroup, the Lie algebra ofKis exactly𝔨.TauCeti.Lie.isClosed_setOf_forall_lieExp_smul_mem: the candidate Lie algebra is a closed subset of𝔤.TauCeti.Lie.Ad_mem_lieSubalgebraOfSubgroup: the Lie algebra ofKis invariant under the adjoint action of the normalizer ofK.TauCeti.Lie.lieSubalgebraOfSubgroup_le_comap_Ad_sub_id: forgin the normalizer ofK,Ad g - 1mapsLie(K)into itself, in the form thatSubmodule.mapQconsumes.TauCeti.Lie.lieSubalgebraOfSubgroup_inf: for closed subgroups, the Lie algebra of an intersection is the intersection of the Lie algebras.TauCeti.Lie.lieSubalgebraOfSubgroup_map_conj: the Lie algebra of a conjugate of a subgroup is the adjoint image of its Lie algebra.
References #
- Lie groups and the Lie algebra correspondence roadmap, Deliverable A, Layer 2, "The Lie subalgebra of a subgroup".
- J. Hilgert and K.-H. Neeb, Structure and Geometry of Lie Groups, Springer (2012), Section 9.1.
- J. M. Lee, Introduction to Smooth Manifolds, 2nd edition (2013), Theorem 20.12.
The three closure properties #
The zero derivation generates the trivial one-parameter subgroup, which lies in every subgroup.
Scalar closure. Rescaling a derivation reparametrises its one-parameter subgroup, so the condition defining the Lie algebra of a subgroup is invariant under scalars. No hypothesis on the subgroup is needed.
Additive closure, from the Trotter product formula: lieExp (t • (X + Y)) is the limit of
the powers (lieExp ((t / n) • X) * lieExp ((t / n) • Y)) ^ n, each of which lies in K, so a
closed K contains the limit.
The candidate Lie algebra of a closed subgroup is closed. It is an intersection over the
time parameter of preimages of K under the continuous maps X ↦ lieExp (t • X).
The adjoint action of an element of the normalizer of K preserves the defining condition:
conjugating the one-parameter subgroup of Y by g ∈ N(K) is the one-parameter subgroup of
Ad g Y, and stays inside K.
The derivative of the adjoint orbit #
The infinitesimal adjoint, on derivations. The adjoint orbit s ↦ Ad (lieExp (s • X)) Y
has derivative ⁅X, Y⁆ at s = 0.
This is TauCeti.Lie.hasDerivAt_tangentAd_mulInvariantExp_smul_apply_zero, the same statement for
the tangent space at the identity, transported across the canonical isometric identification
TauCeti.Lie.leftInvariantDerivationLinearIsometryEquivModelVectorSpace of
LeftInvariantDerivation I G with the model space E. Bracket closure below is the only consumer
so far: it is what makes ⁅X, Y⁆ a limit of difference quotients of a curve in 𝔨.
Bracket closure #
Bracket closure. The curve s ↦ Ad (lieExp (s • X)) Y takes its values in the candidate
Lie algebra of K, starts at Y and has derivative ⁅X, Y⁆ at s = 0; its difference quotients
therefore lie in the linear span, and the bracket is their limit.
The Lie subalgebra #
The Lie algebra of a subgroup. For a subgroup K of a Lie group G, the Lie subalgebra of
𝔤 generated by the derivations whose one-parameter subgroup lies entirely in K. It is the
candidate Lie K, identified with the honest Lie algebra of K by the closed-subgroup theorem.
The generating set is itself a Lie subalgebra — for a closed K this is
coe_lieSubalgebraOfSubgroup below, and for an arbitrary K it is a Layer 4 statement not
available here — so generating is inert; it is what makes the definition total in K. See the
module docstring.
Instances For
The model given by the Lie algebra of a subgroup is finite-dimensional.
A derivation whose one-parameter subgroup lies in K lies in the Lie algebra of K. This
direction needs no hypothesis on K.
The universal property of the Lie algebra of a subgroup. It is the least Lie subalgebra
containing every derivation whose one-parameter subgroup lies in K, so a Lie subalgebra contains
it as soon as it contains those derivations. No hypothesis on K is needed; together with
mem_lieSubalgebraOfSubgroup_of_forall this is the elimination rule for an arbitrary subgroup.
Induction on the Lie algebra of a subgroup. A property that holds for every derivation
whose one-parameter subgroup lies in K and for 0, and is preserved by addition, scalar
multiplication and the bracket, holds on all of Lie(K). This is LieSubalgebra.lieSpan_induction
for the generating set, exposed across the module boundary, where the definition does not
unfold.
The Lie algebra of a closed subgroup is exactly 𝔨: the three closure properties make the
defining condition a Lie subalgebra, so generating it adds nothing.
The Lie algebra of a closed subgroup is exactly 𝔨, as sets.
The one-parameter subgroup generated by an element of the Lie algebra of a closed subgroup K
lies in K.
The exponential of an element of the Lie algebra of a closed subgroup K lies in K.
The Lie algebra of a closed subgroup is a closed subset of 𝔤. This is what makes the
closed-subgroup theorem's complement argument available: 𝔨 is a closed linear subspace of a
finite-dimensional space, hence has a linear complement on which the exponential is transverse.
Monotonicity. A larger subgroup has a larger Lie algebra.
The Lie algebra of an intersection is the intersection of the Lie algebras. For closed
subgroups both sides say that the one-parameter subgroup of X lies in each factor, so the
statement is an interchange of the two quantifiers. Closedness is what collapses each generated
subalgebra to that defining condition; without it only ≤ is available, from
lieSubalgebraOfSubgroup_mono. This is how a subgroup cut out by two independent conditions —
such as the special unitary group, unitary and of determinant one — has its Lie algebra computed
one condition at a time.
The Lie algebra of the whole group is the whole Lie algebra.
The Lie algebra of K is invariant under the adjoint action of the normalizer of K.
Conjugating by g ∈ N(K) carries the one-parameter subgroup of X to the one of Ad g X, which
therefore also stays in K. This is the bundled form of lieExp_smul_Ad_mem, the statement that
makes the adjoint orbit of the bracket proof run inside 𝔨. No hypothesis on K is needed:
Ad g is an automorphism of 𝔤 preserving the generating set, hence preserves the subalgebra it
generates.
Ad g - 1 preserves the Lie algebra of K for g ∈ N(K), in the form that lets it
descend to the quotient 𝔤 ⧸ Lie(K) through Submodule.mapQ.
The Lie algebra of a conjugate of a subgroup is the adjoint image of its Lie algebra.