SOLUTIONS TO SELECTED EXERCISES: SHEET 3 Exercise 1 a: Consider the map Ψ : GL+ n (R) → (0, ∞) × SLn (R) given by: 1 1 Ψ(g) := (det g) n , (det g)− n g Ψ is continuous and hence Ψ−1 ⊆ (0, 1] × B ⊆ GL+ n (R) is measureable n2 whenever B ∈ B SLn (R). As GL+ (R) ⊆ R is open, C(B) = {0} ∪ Ψ−1 ⊆ n 2 (0, 1] × B is measureable in Rn . Left invariance of the measure: mSLn (R) (B) := mRn2 C(B) 2 follows immediately from C(gB) = gC(B) and the fact that on Rn an element g ∈ SLn (R) acts – after an appropriate permutation of the standard basis – like a block-diagonal n2 × n2 -matrix with n copies of g along the diagonal. b: We use the coordinate system φ : Aff 1 (R) 3 (a, b) 7→ (a, a−1 b) ∈ R× × R. 1 On R× × R we define the measure dν(α, β) := |α| dα dβ and we claim that −1 (φ )∗ ν is a left-Haar measure on Aff 1 (R). Let f ∈ Cc Aff 1 (R) and denote ψ(α) := xα, then for left-translation – indicated by subscript – follows: ! Z Z f x y ◦ φ−1 (α, β) ( 1) −1 dβ dα (φ )∗ ν f( x y ) = |α| 1 R× R ! Z Z f ◦ φ−1 xα, β + (xα)−1 y = dβ dα |α| R× R Z Z f ◦ φ−1 (xα, β) (trans. inv.) = dβ dα |α| R× R ! Z Z −1 f ◦φ ψ(α), β 0 0 (ψ = x) = |ψ (α)| dα dβ |ψ(α)| R R× ! Z Z f ◦ φ−1 α, β × × dα dβ = (φ−1 )∗ ν(f ) (ψ(R ) = R ) = |α| × R R and thus we have indeed found a left Haar measure. For right translation – indicated by superscript – follows: ! Z Z (x y) xy f 1 ◦ φ−1 (α, β) −1 ( ) (φ )∗ ν f 1 = dβ dα |α| R× R ! Z Z f ◦ φ−1 xα, x−1 β + x−1 y = dβ dα |α| R× R ! Z Z f ◦ φ−1 xα, x−1 β (trans. inv.) = dβ dα |α| R× R 1 2 SOLUTIONS TO SELECTED EXERCISES: SHEET 3 Z Z (subst. β 7→ xβ) |x| = R× = |x| (φ (as above) R −1 ! f ◦ φ−1 xα, β dβ dα |α| )∗ ν(f ) −1 Hence follows ∆Aff 1 (R) ( x y1 ) = |x| First look at . Exercise 5 v) = det d0 φAdg−1 v . Fix some δ < δ0 such that: ρ(Ad−1 g U := Bδ ∪ Adg−1 Bδ ∪ Adg Bδ ⊆ Bδ0 Note that for v, w ∈ U : φAdg−1 v (w) = (Adg−1 ◦φv ◦ Adg )(w) In particular: −1 ρ(Ad−1 det(d0 Adg )−1 ρ(v) g v) = det(dv Adg −1 ) As Adg−1 is linear, det(dv Adg−1 ) = det(d0 Adg−1 ) = det(d0 Adg )−1 and thus: ρ(Ad−1 g v) = ρ(v) We calculate: µ(Bg) =µ(g −1 Bg) = XZ k = XZ k Bδ 1g−1 Bg∩Pk gk exp(v) dv Bδ 1B∩Pk0 gk0 exp(Adg v) ρ(v) dv XZ 1 1B∩Pk0 gk0 exp(v) ρ(Adg−1 v) dv = |det dAdg | Adg Bδ k Z X 1B∩Pk0 gk0 exp(v) ρ(v) dv = det dAdg−1 k Adg Bδ = det dAdg−1 µ(B) where we used that for a fixed measure on g, the Haar measure constructed in the above manner does not depend on the choice of the neighbourhood U = Bδ of 0. As this is by itself an interesting fact, we give the construction of the Haar measure for neighbourhoods of 0 which are not necessarily open balls around 0. Let V ⊆ g an open neighbourhood of 0 such that exp(V ) ⊆ G is open and exp V is a diffeomorphism onto its image. Denote U := exp(V ). As exp and p : G × G → G are continuous, there is some open neighbourhood A ⊆ g of 0 such that exp(A) exp(A) ⊆ U and hence we obtain a well-defined product on A × A: ∗ :A × A → g v ∗ w := log exp(v) exp(w) By the same argument, we can assume that A is chosen so that the above holds and exp(u) exp(v) exp(w) ∈ U for all u, v, w ∈ A. We note that det(d0 φ0 ) = 1 as φ0 = id. As v 7→ det(d0 φv ) is continuous, we can assume without loss of generality that A is chosen so that for v, w ∈ A holds det(dv φw ) > 0. Lemma 0.1. Let v0 ∈ A, f ∈ Cc (G), then: Z Z f exp(v0 ) exp(v) ρ(v) dv = ∗A v0 ∗A f exp(v) ρ(v) dv SOLUTIONS TO SELECTED EXERCISES: SHEET 3 3 Proof. If v0 , v, w ∈ A, then φφv0 v (w) = (φv0 v)∗w = (v0 ∗v)∗w = v0 ∗(v ∗w) = φv0 ◦ φv (w), hence d0 φφv0 v = dv φv0 d0 φv and thus ρ(φv0 v) = det(dv φv0 )ρ(v), yielding: Z Z f exp(v0 ) exp(v) ρ(v) dv = f exp(φv0 v) ρ(φv0 v) det(dv φv0 ) dv A ZA = f exp(v) ρ(v) dv v0 ∗A Given this local definition of the measure, we define: XZ m(B) := 1B∩Pk gk exp(v) ρ(v) dv ∀B ∈ BG A k≥1 where we chose a symmetric neighbourhood O ⊆ g of 0 such that O ∗ O ∗ O ⊆ A, (gk )k∈N ∈ GN and (Pk )k∈N ∈ BGN such that: G [ G= Pk = gk exp(O) k≥1 k≥1 Pk ⊆gk exp(O) ∀k ≥ 1 We check that this is independent of the choice of the partition and the sequence. Assume that (Ql )l∈N ∈ BGN and (hl )l∈N ∈ GN are chosen similarly. Assume that Pk ∩ Ql 6= ∅, then there are vk , wl ∈ O such that: gk exp(vk ) = hl exp(vl ) ⇒ h−1 l gk ∈ exp(O ∗ O) Beppo-Levi implies: m(B) = XZ XZ O vk,l ∗O 1B∩Pk ∩Ql hl exp(v) ρ(v) dv A XZ l≥1 1B∩Pk ∩Ql hl exp(v) ρ(v) dv X Z k,l≥1 = 1B∩Pk ∩Ql hl exp(vk,l ) exp(v) ρ(v) dv X Z k,l≥1 = X Z k,l≥1 = 1B∩Pk gk exp(v) ρ(v) dv O k≥1 = A k≥1 = 1B∩Pk gk exp(v) ρ(v) dv 1B∩Ql hl exp(v) ρ(v) dv A where in the second to last step we used that for v ∈ A holds hl exp(v) ∈ Pk iff v = vk,l ∗ w for some w ∈ O, which can be checked easily using the definitions of A and O. Left-invariance of this measure follows by the same argument as in class: if (Pk )k∈N is a partition corresponding to O ⊆ A and the sequence (gk )k∈N , then (g −1 Pk )k∈N is a partition corresponding to O ⊆ A and the sequence (g −1 gk )k∈N . Finally we note that the measure is independent of the specific choice of O and A for some fixed choice of measure on g. In order to see this, assume that A1 , O1 and A2 , O2 are two distinct choices as above. Let exp(O1 ) ∩ exp(O2 ) =: U . The measures defined by (A1 , O1 ) and (A2 , O2 ) agree on U without vanishing which follows from the fact that we can choose U as an element of both partitions defining the measures. 4 SOLUTIONS TO SELECTED EXERCISES: SHEET 3 References 1. Alexander S. Kechris, Classical Descriptive Set Theory, Springer, 1995.