Bourbaki

Theorem 1

Tag 09D3, permanent. Cite it as [Bourbaki, Tag 09D3].

Integration, chapter VII, § 1. Construction of a Haar measure

On every locally compact group, there exists a left (resp. right) Haar measure, and, up to a constant factor, there exists only one.

A) Existence. — Set K(G)=K\mathscr{K}(G) = \mathscr{K}, K+(G)=K+\mathscr{K}_+(G) = \mathscr{K}_+, K+∗=K+−{0}\mathscr{K}_+^* = \mathscr{K}_+ - \{0\}. If C is a compact subset of G, we denote by K+∗(C)\mathscr{K}_+^*(C) the set of f∈K+∗f \in \mathscr{K}_+^* with support in C. For f∈Kf \in \mathscr{K} and g∈K+∗g \in \mathscr{K}_+^*, there exist numbers c1,…,cn≥0c_1, \ldots, c_n \geq 0 and elements s1,…,sns_1, \ldots, s_n of G such that f≤∑i=1nciγ(si)gf \leq \sum_{i=1}^n c_i \gamma(s_i)g: for, there exists a nonempty open set U in G such that inf⁡s∈Ug(s)>0\inf_{s \in U} g(s) > 0, and the support of ff can be covered by a finite number of left-translates of U. Let (f:g)(f : g) be the infimum of the numbers ∑i=1nci\sum_{i=1}^n c_i for all systems (c1,…,cn,s1,…,sn)(c_1, \ldots, c_n, s_1, \ldots, s_n) of numbers ≥0\geq 0 and elements of G such that f≤∑i=1nciγ(si)gf \leq \sum_{i=1}^n c_i \gamma(s_i)g. Then:

(i) (γ(s)f:g)=(f:g)(\gamma(s)f : g) = (f : g) for f∈Kf \in \mathscr{K}, g∈K+∗g \in \mathscr{K}_+^*, s∈Gs \in G; (ii) (λf:g)=λ(f:g)(\lambda f : g) = \lambda(f : g) for f∈Kf \in \mathscr{K}, g∈K+∗g \in \mathscr{K}_+^*, λ≥0\lambda \geq 0; (iii) ((f+f′):g)≤(f:g)+(f′:g)((f + f') : g) \leq (f : g) + (f' : g) for f∈Kf \in \mathscr{K}, f′∈Kf' \in \mathscr{K}, g∈K+∗g \in \mathscr{K}_+^*; (iv) (f:g)≥(sup⁡f)/(sup⁡g)(f : g) \geq (\sup f)/(\sup g) for f∈Kf \in \mathscr{K}, g∈K+∗g \in \mathscr{K}_+^*; (v) (f:h)≤(f:g)(g:h)(f : h) \leq (f : g)(g : h) for f∈Kf \in \mathscr{K}, g∈K+∗g \in \mathscr{K}_+^*, h∈K+∗h \in \mathscr{K}_+^*; (vi) 0<1(f0:f)≤(f:g)(f0:g)≤(f:f0)0 < \frac{1}{(f_0 : f)} \leq \frac{(f : g)}{(f_0 : g)} \leq (f : f_0) for f,f0,gf, f_0, g in K+∗\mathscr{K}_+^*; (vii) let f,f′,hf, f', h be in K+\mathscr{K}_+ with h(s)≥1h(s) \geq 1 on the support of f+f′f + f', and let ε>0\varepsilon > 0; there exists a compact neighborhood V of e such that, for every g∈K+∗(V)g \in \mathscr{K}_+^*(V),

(f:g)+(f′:g)≤((f+f′):g)+ε(h:g). (f : g) + (f' : g) \leq ((f + f') : g) + \varepsilon(h : g).

The properties (i), (ii), (iii) are obvious. Let f∈Kf \in \mathscr{K}, g∈K+∗g \in \mathscr{K}_+^*; if f≤∑i=1nciγ(si)gf \leq \sum_{i=1}^n c_i \gamma(s_i)g with the ci≥0c_i \geq 0, then sup⁡f≤∑i=1ncig(si−1s)\sup f \leq \sum_{i=1}^n c_i g(s_i^{-1}s) for some s \in G, therefore sup⁡f≤(∑i=1nci)sup⁡g\sup f \leq (\sum_{i=1}^n c_i) \sup g, whence (iv). Let us now prove (v); let f∈Kf \in \mathscr{K}, g,hg, h in K+∗\mathscr{K}_+^*; if f≤∑i=1nciγ(si)gf \leq \sum_{i=1}^n c_i \gamma(s_i)g and g≤∑j=1pdjγ(tj)hg \leq \sum_{j=1}^p d_j \gamma(t_j)h (ci≥0,dj≥0,si,tjc_i \geq 0, d_j \geq 0, s_i, t_j in GG), then f≤∑i,jcidjγ(sitj)hf \leq \sum_{i,j} c_i d_j \gamma(s_i t_j)h, therefore (f:h)≤∑i,jcidj=(∑ici)(∑jdj)(f : h) \leq \sum_{i,j} c_i d_j = (\sum_i c_i)(\sum_j d_j); thus (f:h)≤(f:g)(g:h)(f : h) \leq (f : g)(g : h). Applying (v) to f0,f,gf_0, f, g on the one hand and to f,f0,gf, f_0, g on the other, one obtains (vi). Finally, let f,f′,hf, f', h be in K+\mathscr{K}_+ with h(s)≥1h(s) \geq 1 on the support of f+f′f + f', and let ε>0\varepsilon > 0. Set F=f+f′+12εhF = f + f' + \frac{1}{2} \varepsilon h; the functions φ,φ′\varphi, \varphi', that coincide respectively with f/Ff/F and f′/Ff'/F on the support of f+f′f + f' and are zero outside it, belong to K+\mathscr{K}_+; for every η>0\eta > 0, there exists a compact neighborhood VV of ee such that ∣φ(s)−φ(t)∣≤η|\varphi(s) - \varphi(t)| \leq \eta and ∣φ′(s)−φ′(t)∣≤η|\varphi'(s) - \varphi'(t)| \leq \eta for s−1t∈Vs^{-1} t \in V. Then let g∈K+∗(V)g \in \mathscr{K}_+^*(V); for every s∈Gs \in G,

φ⋅γ(s)g≤(φ(s)+η)⋅γ(s)g; \varphi \cdot \gamma(s)g \leq (\varphi(s) + \eta) \cdot \gamma(s)g;

for, this is obvious at the points where γ(s)g\gamma(s)g is zero, therefore outside of sVsV; and in sVsV, φ≤φ(s)+η\varphi \leq \varphi(s) + \eta; similarly, φ′⋅γ(s)g≤(φ′(s)+η)⋅γ(s)g\varphi' \cdot \gamma(s)g \leq (\varphi'(s) + \eta) \cdot \gamma(s)g. This stated, let c1,…,cnc_1, \ldots, c_n be numbers ≥0\geq 0 and s1,…,sns_1, \ldots, s_n elements of GG such that F≤∑i=1nciγ(si)gF \leq \sum_{i=1}^n c_i \gamma(s_i)g; then

f=φF≤∑i=1nciφ⋅γ(si)g≤∑i=1nci(φ(si)+η)⋅γ(si)g f = \varphi F \leq \sum_{i=1}^n c_i \varphi \cdot \gamma(s_i)g \leq \sum_{i=1}^n c_i (\varphi(s_i) + \eta) \cdot \gamma(s_i)g

and similarly for f′f'; consequently

(f:g)+(f′:g)≤∑i=1nci(φ(si)+φ′(si)+2η)≤(1+2η)∑i=1nci (f : g) + (f' : g) \leq \sum_{i=1}^n c_i (\varphi(s_i) + \varphi'(s_i) + 2\eta) \leq (1 + 2\eta) \sum_{i=1}^n c_i

since φ+φ′≤1\varphi + \varphi' \leq 1. Applying the definition of FF, then (ii), (iii) and (v), one infers that

(f:g)+(f′:g)≤(1+2η)(F:g)≤ (f : g) + (f' : g) \leq (1 + 2\eta)(F : g) \leq
(1+2η)[((f+f′):g)+12ε(h:g)]≤ (1 + 2\eta)[((f + f') : g) + \frac{1}{2}\varepsilon(h : g)] \leq
(((f+f′):g)+12ε(h:g)+2η((f+f′):h)(h:g)+εη(h:g) (((f + f') : g) + \frac{1}{2}\varepsilon(h : g) + 2\eta((f + f') : h)(h : g) + \varepsilon\eta(h : g)

and, if η\eta has been chosen so that η[2((f+f′):h)+ε]≤12ε\eta[2((f + f') : h) + \varepsilon] \leq \frac{1}{2}\varepsilon, one obtains (vii).

As V runs over the set of compact neighborhoods of e, the K+∗(V)\mathscr{K}_+^*(V) form a base of a filter B\mathscr{B} on K+∗\mathscr{K}_+^*. Let F\mathscr{F} be an ultrafilter on K+∗\mathscr{K}_+^* finer than B\mathscr{B}. On the other hand, let us fix f0∈K+∗f_0 \in \mathscr{K}_+^* and let us set, for f∈K+∗f \in \mathscr{K}_+^* and g∈K+∗g \in \mathscr{K}_+^*,

Ig(f)=(f:g)(f0:g). I_g(f) = \frac{(f : g)}{(f_0 : g)}.

By (vi), lim⁡g,FIg(f)=I(f)\lim_{g,\mathscr{F}} I_g(f) = I(f) exists in the compact space [1/(f0:f),(f:f0)][1/(f_0 : f), (f : f_0)].

By (iii), I(f+f′)≤I(f)+I(f′)I(f + f') \leq I(f) + I(f'). By (vii), I(f)+I(f′)≤I(f+f′)+εI(h)I(f) + I(f') \leq I(f + f') + \varepsilon I(h) for all ε>0\varepsilon > 0 provided hh is ≥1\geq 1 on the support of f+f′f + f'; it follows that I(f+f′)=I(f)+I(f′)I(f + f') = I(f) + I(f'). By Ch. II, §2, No. 1, Prop. 2, I is extendible to a linear form on K\mathscr{K}; this linear form is a nonzero positive measure on G, left-invariant by (i); this is the sought-for left Haar measure. Passing to the opposite group, one deduces from this the existence of a right Haar measure.

B) Uniqueness. — Let μ\mu be a left Haar measure, ν\nu a right Haar measure. Then νˇ\check{\nu} is a left Haar measure. We are going to show that μ\mu and νˇ\check{\nu} are proportional. This will prove that any two left Haar measures are indeed proportional.

Let f∈Kf \in \mathscr{K} be such that μ(f)≠0\mu(f) \neq 0. By Lemma 1, the function DfD_f defined on G by the formula

Df(s)=μ(f)−1∫f(t−1s)dν(t) D_f(s) = \mu(f)^{-1} \int f(t^{-1}s) d\nu(t)

is continuous on G. Let g∈Kg \in \mathscr{K}. The function (s,t)↦f(s)g(ts)(s, t) \mapsto f(s)g(ts) is continuous with compact support in G×GG \times G. By Ch. III, §4, No. 1, Th. 2,

μ(f)ν(g)=(∫f(s)dμ(s))(∫g(t)dν(t))=∫dμ(s)∫f(s)g(ts)dν(t)=∫dν(t)∫f(s)g(ts)dμ(s)=∫dν(t)∫f(t−1s)g(s)dμ(s)=∫g(s)[∫f(t−1s)dν(t)]dμ(s)=μ(g⋅μ(f)Df), \begin{align*} \mu(f)\nu(g) &= \left( \int f(s) d\mu(s) \right) \left( \int g(t) d\nu(t) \right) \\ &= \int d\mu(s) \int f(s)g(ts) d\nu(t) = \int d\nu(t) \int f(s)g(ts) d\mu(s) \\ &= \int d\nu(t) \int f(t^{-1}s)g(s) d\mu(s) \\ &= \int g(s) \left[ \int f(t^{-1}s) d\nu(t) \right] d\mu(s) = \mu(g \cdot \mu(f) D_f), \end{align*}

whence

ν(g)=μ(Df⋅g). \nu(g) = \mu(D_f \cdot g).

This proves, first, that DfD_f does not depend on ff. For, if f′∈Kf' \in \mathscr{K} is such that μ(f′)≠0\mu(f') \neq 0, then Df⋅μ=Df′⋅μD_f \cdot \mu = D_{f'} \cdot \mu, therefore Df=Df′D_f = D_{f'} locally almost everywhere for μ\mu, hence everywhere, since DfD_f and Df′D_{f'} are continuous and the support of μ\mu is GG. We may therefore write Df=DD_f = D. The formula (16) gives

μ(f)D(e)=νˇ(f). \mu(f)D(e) = \check{\nu}(f).

The formula (19) may be extended by linearity to the functions f∈Kf \in \mathscr{K} such that μ(f)=0\mu(f) = 0. We have D(e)≠0D(e) \neq 0 since νˇ≠0\check{\nu} \neq 0. This indeed establishes the proportionality of μ\mu and νˇ\check{\nu}.

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Cites

Transcribed from the printed volume. Not translated.