Bourbaki

§ 7. Convergence of the Hausdorff series (real or complex case)

In this paragraph we assume that KK is one of the fields R\mathbf{R} or C\mathbf{C} with its usual absolute value. Recall that a normable algebra over KK is a (not necessarily associative) algebra over KK with a topology T\mathscr{T} with the following properties: (1) T\mathscr{T} can be defined by a norm: (2) the mapping (x,y)↦xy(x, y) \mapsto xy of A×AA \times A into AA is continuous. A normed algebra over KK is an algebra AA over KK with a norm such that ∥xy∥≤∥x∥ ∥y∥\|xy\| \leq \|x\|\ \|y\| for all x,yx, y in AA. We denote by gg a complete normable Lie algebra over KK. We choose a norm on gg and a number M>0M > 0 such that

∥[x,y]∥≤M∥x∥ ∥y∥for x,y in g. \|[x, y]\| \leq M \|x\|\ \|y\| \quad \text{for } x, y \text{ in } g.

1. Continuous-polynomials with values in gg

Let II be a finite set and let P(gI;g)P(g^I; g) (resp. P^(gI;g)\hat{P}(g^I; g)) be the vector space of continuous-polynomials (resp. formal power series with continuous components) on gIg^I with values in gg. Recall (Differentiable and Analytic Manifolds, R, Appendix) that P(gI;g)P(g^I; g) has a graduation of type NI\mathbf{N}^I and that P^(gI;g)\hat{P}(g^I; g) is identified with the completion of the vector space P(gI;g)P(g^I; g) with the topology defined by the filtration associated with the graduation of P(gI;g)P(g^I; g). Moreover, P(gI;g)P(g^I; g) is a graded Lie algebra with the bracket defined by [f,g](x)=[f(x),g(x)][f, g](x) = [f(x), g(x)] for f, g in P(g^I; g), x \in g^I; this Lie algebra structure can be extended by continuity to P^(gI;g)\hat{P}(g^I; g) and makes it into a complete Hausdorff filtered Lie algebra.

By proposition 2 of § 6, no. 3, there exists one and only one continuous Lie algebra homomorphism ϕI:u↦u~\phi_I : u \mapsto \tilde{u} of L^(I)\hat{L}(I) into P^(gI;g)\hat{P}(g^I; g) mapping the indeterminate of index i to pri\mathrm{pr}_i for all i∈Ii \in I, since pri∈P(gI;g)\mathrm{pr}_i \in P(g^I; g). It follows that u~∈P(gI;g)\tilde{u} \in P(g^I; g) for u∈L(I)u \in L(I); more precisely, when u∈L(I)u \in L(I), u~\tilde{u} is just the polynomial mapping (ti)↦u((ti))(t_i) \mapsto u((t_i)) of § 2, no. 4. On the other hand, clearly ϕI\phi_I is compatible with the multigraduations of L(I)L(I) and P(gI;g)P(g^I; g). If u=∑v∈NIuvu = \sum_{v \in \mathbf{N}^I} u_v, where uv∈Lv(I)u_v \in L^v(I) for v∈NIv \in \mathbf{N}^I, then

u~=∑v∈NIu~v,where u~v∈Pv(gI;g). \tilde{u} = \sum_{v \in \mathbf{N}^I} \tilde{u}_v, \quad \text{where } \tilde{u}_v \in P_v(g^I; g).

Let u=(uj)j∈Ju = (u_j)_{j \in J} be a finite family of elements of L^(I)\hat{L}(I), let v∈L^(J)v \in \hat{L}(J) and let w=v∘uw = v \circ u (§ 6, no. 3). We write u~=(u~j)j∈J∈g\tilde{u} = (\tilde{u}_j)_{j \in J} \in \mathfrak{g}. Then

(2)

v~∘u~=(v∘u)∼. \tilde{v} \circ \tilde{u} = (v \circ u)^{\sim}.

This follows by extending by continuity formula (7) of § 6, no. 3 and from Differentiable and Analytic Manifolds, R, Appendix, no. 6.

2. Group germ defined by a complete normed Lie algebra

Let H=∑r,s≥0Hr,s∈L^(U,V)H = \sum_{r,s \geq 0} H_{r,s} \in \hat{L}(U, V) be the Hausdorff series (§ 6, no. 4, Definition 1). We shall show that the corresponding formal power series

(3)

H~=∑r,s≥0H~r,s∈P^(g×g,g) \tilde{H} = \sum_{r,s \geq 0} \tilde{H}_{r,s} \in \hat{P}(g \times g, g)

is convergent (Differentiable and Analytic Manifolds, R, 3.1.1).

We introduce the following formal power series η∈Q[[U,V]]\eta \in \mathbf{Q}[[U, V]]

(4)

η(U,V)=−log⁡(2−exp⁡(U+V)) \eta(U, V) = -\log(2 - \exp(U + V))
(5)
=∑m≥11m(exp⁡(U+V)−1)m = \sum_{m \geq 1} \frac{1}{m} (\exp(U + V) - 1)^m
(6)
=∑m≥11m∑r1,…,rms1,…,smUr1Vs1r1!Ur2Vs2r2!⋯UrmVsmsm!. = \sum_{m \geq 1} \frac{1}{m} \sum_{\substack{r_1, \ldots, r_m \\ s_1, \ldots, s_m}} \frac{U^{r_1} V^{s_1}}{r_1!} \frac{U^{r_2} V^{s_2}}{r_2!} \cdots \frac{U^{r_m} V^{s_m}}{s_m!}.

Hence

(7)

η(U,V)=∑r,s≥0ηr,sUrVs, \eta(U, V) = \sum_{r,s \geq 0} \eta_{r,s} U^r V^s,

where

ηr,s=∑m≥11m∑r1+⋯+rm=rs1+⋯+sm=sri+si≥11r1!…rm!s1!…sm!. \eta_{r,s} = \sum_{m \geq 1} \frac{1}{m} \sum_{\substack{r_1 + \cdots + r_m = r \\ s_1 + \cdots + s_m = s \\ r_i + s_i \geq 1}} \frac{1}{r_1! \ldots r_m! s_1! \ldots s_m!}.
Now let uu and vv be two positive real numbers such that u+v<log⁡2u + v < \log 2; then 0≤exp⁡(u+v)−1<10 \leq \exp(u + v) - 1 < 1; the series derived from (5) and (6) by substituting uu for UU and vv for VV are convergent and the above calculations imply that
∑r,s≥0ηr,survs=−log⁡(2−exp⁡(u+v))<+∞. \sum_{r, s \geq 0} \eta_{r,s} u^r v^s = -\log(2 - \exp(u + v)) < +\infty.
Let r,s≥0r, s \geq 0 and let ∥H~r,s∥\| \tilde{H}_{r,s} \| be the norm of the continuous-polynomial H~r,s\tilde{H}_{r,s} (Differentiable and Analytic Manifolds, R, Appendix, no. 2).

Lemma 1

∥H~r,s∥≤Mr+s−1ηr,s. \| \tilde{H}_{r,s} \| \leq M^{r+s-1} \eta_{r,s}.
Let ri,sir_i, s_i be in N\mathbf{N} for 1≤i≤m1 \leq i \leq m, with sm=1s_m = 1; we write r=∑iri,s=∑isir = \sum_i r_i, s = \sum_i s_i and consider the following element of L({U,V})L(\{U, V\}):
Z=((∑i=1m−1(ad U)ri(ad V)si)(ad U)rm)(V). Z = \left( \left( \sum_{i=1}^{m-1} (\mathrm{ad}\ U)^{r_i} (\mathrm{ad}\ V)^{s_i} \right) (\mathrm{ad}\ U)^{r_m} \right)(V).
Then Z~=f∘p\tilde{Z} = f \circ p, where ff is the following (r+s)(r+s)-linear mapping of gr+sg^{r+s} into gg:
(x1,…,xr,y1,…,ys)↦(ad(x1)∘⋯∘ad(xr)∘ad(y1)∘⋯∘ad(ys))(ys) (x_1, \ldots, x_r, y_1, \ldots, y_s) \mapsto (\mathrm{ad}(x_1) \circ \cdots \circ \mathrm{ad}(x_r) \circ \mathrm{ad}(y_1) \circ \cdots \circ \mathrm{ad}(y_s))(y_s)
and where pp is the following mapping of g2g^2 into gr+sg^{r+s}:
(x,y)↦(x,…,x⏟r,y,…,y⏟s); (x, y) \mapsto (\underbrace{x, \ldots, x}_{r}, \underbrace{y, \ldots, y}_{s});
hence ∥Z~∥≤∥f∥≤Mr+s−1\| \tilde{Z} \| \leq \| f \| \leq M^{r+s-1} (Differentiable and Analytic Manifolds, R, Appendix). Applying these inequalities to the various terms on the right hand side of formula (9) of § 6, no. 4, we obtain:
∥(Hr,s′)∼∥≤Mr+s−1r+s∑m≥11m∑r1+⋯+rm=rs1+⋯+sm=sr1+s1≥1,…,rm−1+sm−1≥11r1!…rm!s1!…sm!. \|(H'_{r,s})\sim\| \leq \frac{M^{r+s-1}}{r+s} \sum_{m \geq 1} \frac{1}{m} \sum_{\substack{r_1 + \cdots + r_m = r \\ s_1 + \cdots + s_m = s \\ r_1 + s_1 \geq 1, \ldots, r_{m-1} + s_{m-1} \geq 1}} \frac{1}{r_1! \ldots r_m! s_1! \ldots s_m!}.
A similar argument gives
∥(Hr,s′′)∼∥≤Mr+s−1r+s∑m≥11m∑r1+⋯+rm−1=r−1s1+⋯+sm−1=sr1+s1≥1,…,rm−1+sm−1≥11r1!…rm−1!s1!…sm−1! \|(H''_{r,s})\sim\| \leq \frac{M^{r+s-1}}{r+s} \sum_{m \geq 1} \frac{1}{m} \sum_{\substack{r_1 + \cdots + r_{m-1} = r-1 \\ s_1 + \cdots + s_{m-1} = s \\ r_1 + s_1 \geq 1, \ldots, r_{m-1} + s_{m-1} \geq 1}} \frac{1}{r_1! \ldots r_{m-1}! s_1! \ldots s_{m-1}!}

whence, by (8)

∥H~r,s∥<ηr,sMr+s−1r+s≤ηr,sMr+s−1, \|\tilde{H}_{r,s}\| < \eta_{r,s} \frac{M^{r+s-1}}{r+s} \leq \eta_{r,s} M^{r+s-1},

which proves the lemma.

Proposition 1

The formal power series* H~\tilde{H} is a convergent series (Differentiable and Analytic Manifolds, R, 3.1.1); its domain of absolute convergence (Differentiable and Analytic Manifolds, R, 3.1.4) *contains the open set

Ω={(x,y)∈g×g∣∥x∥+∥y∥<1Mlog⁡2}. \Omega = \left\{ (x, y) \in g \times g \mid \|x\| + \|y\| < \frac{1}{M} \log 2 \right\}.

Let u,vu, v be two real numbers >0> 0 such that u+v<1Mlog⁡2u + v < \frac{1}{M} \log 2; then (Lemma 1)

(12)M∑r,s≥0∥H~r,s∥urvs≤∑r,s≥0ηr,sMr+survs=−log⁡(2−exp⁡M(u+v))<+∞ \text{(12)} \quad M \sum_{r,s \geq 0} \|\tilde{H}_{r,s}\| u^r v^s \\ \leq \sum_{r,s \geq 0} \eta_{r,s} M^{r+s} u^r v^s = -\log(2 - \exp M(u + v)) < +\infty
by (9).

Let h:Ω→gh : \Omega \to g denote the analytic function (Differentiable and Analytic Manifolds, R, 3.2.9) defined by H~\tilde{H}, that is by the formula

(13)h(x,y)=∑r,s≥0H~r,s(x,y)=∑r,s≥0Hr,s(x,y)for (x,y)∈Ω. \text{(13)} \quad h(x, y) = \sum_{r,s \geq 0} \tilde{H}_{r,s}(x, y) = \sum_{r,s \geq 0} H_{r,s}(x, y) \quad \text{for } (x, y) \in \Omega.

This function is called the Hausdorff function of gg relative to MM (or simply the Hausdorff function of gg if no confusion can arise). Note that Hr,s(U,−U)=0H_{r,s}(U, -U) = 0 if r+s≥2r + s \geq 2 and hence

(14)h(x,−x)=0for ∥x∥<12Mlog⁡2. \text{(14)} \quad h(x, -x) = 0 \quad \text{for } \|x\| < \frac{1}{2M} \log 2.

Similarly

(15)h(0,x)=h(x,0)=xfor ∥x∥<1Mlog⁡2. \text{(15)} \quad h(0, x) = h(x, 0) = x \quad \text{for } \|x\| < \frac{1}{M} \log 2.

Proposition 2

Let

Ω′={(x,y,z)∈g×g×g∣∥x∥+∥y∥+∥z∥<1Mlog⁡32}. \Omega' = \left\{ (x, y, z) \in g \times g \times g \mid \|x\| + \|y\| + \|z\| < \frac{1}{M} \log \frac{3}{2} \right\}.

If (x,y,z)∈Ω′(x, y, z) \in \Omega', then

(16)(x,y)∈Ω,(h(x,y),z)∈Ω,(y,z)∈Ω,(x,h(y,z))∈Ω \text{(16)} \quad (x, y) \in \Omega, \quad (h(x, y), z) \in \Omega, \quad (y, z) \in \Omega, \quad (x, h(y, z)) \in \Omega
and
(17)h(h(x,y),z)=h(x,h(y,z)). \text{(17)} \quad h(h(x, y), z) = h(x, h(y, z)).

Let (x,y,z)∈Ω′(x, y, z) \in \Omega'; clearly (x,y)∈Ω(x, y) \in \Omega and (y,z)∈Ω(y, z) \in \Omega. Moreover:

∥h(x,y)∥≤∑r,s∥Hr,s∥∥x∣r∥y∣s, \|h(x, y)\| \leq \sum_{r, s} \|H_{r,s}\| \|x|r\|y|s,

and hence by (13)

∥h(x,y)∥≤−1Mlog⁡(2−exp⁡M(∥x∥+∥y∥)). \|h(x, y)\| \leq -\frac{1}{M} \log(2 - \exp M(\|x\| + \|y\|)).

Now M(∥x∥+∥y∥)<log⁡32−M∥z∥M(\|x\| + \|y\|) < \log \frac{3}{2} - M\|z\|; we write u=exp⁡(M∥z∥)u = \exp(M\|z\|); then 1≤u≤321 \leq u \leq \frac{3}{2} and

M(∥h(x,y)∥+∥z∥)<−log⁡(2−exp⁡(log⁡32−M∥z∥))+M∥z∥=−log⁡(2−32u)+log⁡u=log⁡2u24u−3=log⁡(2+2(u−1)(u−3)4u−3)≤log⁡2. \begin{align*} M(\|h(x, y)\| + \|z\|) &< -\log(2 - \exp(\log \frac{3}{2} - M\|z\|)) + M\|z\| \\ &= -\log\left(2 - \frac{3}{2u}\right) + \log u = \log \frac{2u^2}{4u - 3} \\ &= \log\left(2 + \frac{2(u - 1)(u - 3)}{4u - 3}\right) \leq \log 2. \end{align*}

We see similarly that (x,h(y,z))∈Ω(x, h(y, z)) \in \Omega.

We now prove (17). In the Lie algebra L^({U,V,W})\hat{L}(\{U, V, W\}),

H(H(U,V),W)=H(U,H(V,W)) H(H(U, V), W) = H(U, H(V, W))

by Proposition 4 of § 6, no. 5. By no. 1, formula (2), we therefore have in P^(g×g×g,g)\hat{P}(g \times g \times g, g) the relation

H~∘(H~×Idg)=H~∘(Idg×H~). \tilde{H} \circ (\tilde{H} \times \mathrm{Id}_g) = \tilde{H} \circ (\mathrm{Id}_g \times \tilde{H}).

By Differentiable and Analytic Manifolds, R, 3.1.9, there exists a number ε>0\varepsilon > 0 such that formula (17) is true when ∥x∥,∥y∥\|x\|, \|y\| and ∥z∥\|z\| are ≤ε\leq \varepsilon. But the functions (x,y,z)↦h(h(x,y),z)(x, y, z) \mapsto h(h(x, y), z) and (x,y,z)↦h(x,h(y,z))(x, y, z) \mapsto h(x, h(y, z)) are analytic functions on Ω′\Omega' with values in gg (Differentiable and Analytic Manifolds, R, 3.2.7). As Ω′\Omega' is connected and they coincide in a neighbourhood of 0, they are equal (Differentiable and Analytic Manifolds, R, 3.2.5).

The above results imply:

Let α\alpha be a real number such that 0<α≤13Mlog⁡320 < \alpha \leq \frac{1}{3M} \log \frac{3}{2}. Let

G={x∈g∣∥x∥<α}, G = \{x \in g \mid \|x\| < \alpha\},

Θ={(x,y)∈G×G∣h(x,y)∈G} \Theta = \{(x, y) \in G \times G \mid h(x, y) \in G\}
and m:Θ→Gm : \Theta \to G be the restriction of hh to Θ\Theta. Then: (1) Θ\Theta is open in G×GG \times G and mm is analytic. (2) x∈Gx \in G implies (0,x)∈Θ(0, x) \in \Theta, (x,0)∈Θ(x, 0) \in \Theta and m(0,x)=m(x,0)=xm(0, x) = m(x, 0) = x. (3) x∈Gx \in G implies −x∈G-x \in G, (x,−x)∈Θ(x, -x) \in \Theta, (−x,x)∈Θ(-x, x) \in \Theta and

m(x,−x)=m(−x,x)=0. m(x, -x) = m(-x, x) = 0.

(4) Let x, y, z be elements of G such that (x, y) ∈ Θ, (m(x, y), z) ∈ Θ, (y, z) ∈ Θ and (x, m(y, z)) ∈ Θ. Then m(m(x, y), z) = m(x, m(y, z)).

In other words (Chapter III, § 1), if we write −x = σ(x), the quadruple (G, 0, σ, m) is a Lie group germ over K.

3. Exponential in Complete Normed Associative Algebras

In this no. we denote by A a complete normed unital associative algebra (General Topology, Chapter IX, § 3, no. 7). Then \|x.y\| ≤ \|x\|. \|y\| for x, y in A.

Let I be a finite set and let P^(AI;A)\hat{P}(A^I; A) be the vector space of formal power series with continuous components on AIA^I with values in A (Differentiable and Analytic Manifolds, R, Appendix, no. 5) with the algebra structure obtained by writing

f.g=m∘(f,g)for f,g in P^(AI;A), f.g = m \circ (f, g) \quad \text{for } f, g \text{ in } \hat{P}(A^I; A),

where m:A×A→Am : A \times A \to A denotes multiplication on A. Arguing as in no. 1 and using Proposition 1 of § 5, no. 1, we define a continuous homomorphism of unital algebras u↦u~u \mapsto \tilde{u} of A^(I)\hat{A}(I) into P^(AI;A)\hat{P}(A^I; A) mapping the indeterminate of index i to pripr_i; this homomorphism extends the Lie algebra homomorphism of L^(I)\hat{L}(I) into P^(AI;A)\hat{P}(A^I; A) defined in no. 1. If u=∑vuvu = \sum_v u_v with uv∈Av(I)u_v \in A^v(I) for v∈NIv \in \mathbf{N}^I, then u~=∑vu~v\tilde{u} = \sum_v \tilde{u}_v, where u~v\tilde{u}_v is the polynomial mapping (ti)i∈I↦uv((ti))(t_i)_{i \in I} \mapsto u_v((t_i)).

Let u=(uj)j∈Iu = (u_j)_{j \in I} be a finite family of elements of A^(I)\hat{A}(I), let v∈A^(J)v \in \hat{A}(J) and write w=v∘uw = v \circ u (§ 5, no. 1). Then

(v∘u)∼=v~∘u~. (v \circ u)^{\sim} = \tilde{v} \circ \tilde{u}.

This follows by extending by continuity formula (2) of § 5, no. 1 and from Differentiable and Analytic Manifolds, R, Appendix, no. 6.

In particular we take I = {U}, identify A and A(U)A^{(U)} and consider the images e~\tilde{e} and l~\tilde{l} of the series e(U)=∑n≥1Un/n!e(U) = \sum_{n \geq 1} U^n/n! and l(U)=∑n≥1(−1)n−1Un/nl(U) = \sum_{n \geq 1} (-1)^{n-1} U^n/n in P^(A;A)\hat{P}(A; A). Then ∥U~n∥≤1\|\tilde{U}^n\| \leq 1 for ∥x1…xn∥≤∥x1∥…∥xn∥\|x_1 \ldots x_n\| \leq \|x_1\| \ldots \|x_n\| for x1,…,xnx_1, \ldots, x_n in A. Therefore the radius of absolute convergence of e~\tilde{e} (resp. l~\tilde{l}) is infinite (resp. ≥1\geq 1).

We shall denote by eAe_A (resp. lAl_A) the analytic mapping of A into A (resp. of B into A, where B is the open unit ball of A) defined by the convergent series e~\tilde{e} (resp. l~\tilde{l}) and we shall write exp⁡A(x)=1+eA(x)\exp_A(x) = 1 + e_A(x) (for x∈Ax \in A) and

log⁡A(x)=lA(x−1) \log_A(x) = l_A(x - 1)

(for x∈A,∥x−1∥<1x \in A, \|x - 1\| < 1). Then

exp⁡Ax=∑n≥0xnn!(x∈A) \exp_A x = \sum_{n \geq 0} \frac{x^n}{n!} \quad (x \in A)

(20) log⁡Ax=∑n≥1(−1)n−1(x−1)nn(x∈A,∥x−1∥<1).\log_A x = \sum_{n \geq 1} (-1)^{n-1} \frac{(x-1)^n}{n} \quad (x \in A, \|x-1\| < 1).

As (e∘l)(U)=(l∘e)(U)=U(e \circ l)(U) = (l \circ e)(U) = U (cf. § 6, no. 1), by (18) e~∘l~=l~∘e~=IdA\tilde{e} \circ \tilde{l} = \tilde{l} \circ \tilde{e} = \mathrm{Id}_A. Therefore (Differentiable and Analytic Manifolds, R, 3.1.9) (21) exp⁡A(log⁡A(x))=x(x∈A,∥x−1∥≤1)\exp_A(\log_A(x)) = x \quad (x \in A, \|x-1\| \leq 1) (22) log⁡A(exp⁡A(x))=x(x∈A,∥x∥<log⁡2)\log_A(\exp_A(x)) = x \quad (x \in A, \|x\| < \log 2) for ∥x∥<log⁡2\|x\| < \log 2 implies ∥exp⁡A(x)−1∥≤exp⁡∥x∥−1<1\|\exp_A(x) - 1\| \leq \exp \|x\| - 1 < 1.

Finally we consider A as a complete normed Lie algebra. Then

∥[x,y]∥=∥xy−yx∥≤2∥x∥⋅∥y∥. \|[x, y]\| = \|xy - yx\| \leq 2\|x\|\cdot\|y\|.
Proposition 1 of no. 2 implies that the domain of absolute convergence of the formal power series H~\tilde{H} contains the set
Ω={x,y)∈A×A∣∥x∥+∥y∥<12log⁡2}. \Omega = \{x, y) \in A \times A \mid \|x\| + \|y\| < \frac{1}{2} \log 2\}.
Hence H~\tilde{H} defines an analytic function h:Ω→Ah : \Omega \to A. Then h(x,y)=∑r,sHr,s(x,y)h(x, y) = \sum_{r,s} H_{r,s}(x, y) (cf. § 3, no. 1, Remark 4).

Proposition 3

For ∥x∥+∥y∥<12log⁡2\|x\| + \|y\| < \frac{1}{2} \log 2,* (23) exp⁡Ax⋅exp⁡Ay=exp⁡Ah(x,y).\exp_A x \cdot \exp_A y = \exp_A h(x, y). It follows from (18) and the relation euev=eu+ve^u e^v = e^{u+v} that

m∘(1+e~,1+e~)=(1+e~)∘H~ m \circ (1 + \tilde{e}, 1 + \tilde{e}) = (1 + \tilde{e}) \circ \tilde{H}
in P^(A×A;A)\hat{P}(A \times A; A). We therefore deduce from Differentiable and Analytic Manifolds, R, 3.1.9 that (23) is true for (x,y)(x, y) sufficiently close to (0,0)(0, 0), whence the proposition follows by analytic continuation (Differentiable and Analytic Manifolds, R, 3.2.5).

Exercises

See the exercises for § 7.

Transcribed from the printed volume. Not translated.