Bourbaki

Lemma 6

Tag 0348, permanent. Cite it as [Bourbaki, Tag 0348].

Théories spectrales, chapter IV, § 4. Opérateurs partiels

Let g1g_1 and g2g_2 be μ\mu-measurable functions on X.

a) The partial operator mg1+mg2m_{g_1}+m_{g_2} is closable and its closure is the multiplication operator mg1+g2m_{g_1+g_2};

b) One has mg1∘mg2⊂mg1g2m_{g_1}\circ m_{g_2}\subset m_{g_1g_2};

c) Suppose that g1g_1 is bounded. Then one has mg2∘mg1=mg1g2m_{g_2}\circ m_{g_1}=m_{g_1g_2}. Moreover, the domain of mg2m_{g_2} is contained in the domain of mg1g2m_{g_1g_2}, and mg1∘mg2m_{g_1}\circ m_{g_2} is the restriction of mg1g2m_{g_1g_2} to dom(mg2m_{g_2}).

It is elementary that mg1+g2m_{g_1+g_2} is an extension of mg1+mg2m_{g_1}+m_{g_2}; the latter operator is therefore closable, and mg1+mg2‾⊂mg1+g2\overline{m_{g_1} + m_{g_2}}\subset m_{g_1+g_2}.

Let f∈L2(X,μ)f\in \mathscr{L}^2(X, \mu) be such that the function h=(g1+g2)fh= (g_1+g_2)f belongs to L2(X,μ)\mathscr{L}^2(X, \mu). For every integer n⩾1n\geqslant 1, let XnX_n denote the set of x∈Xx\in X such that ∣g1(x)∣+∣g2(x)∣⩽n|g_1(x)|+|g_2(x)|\leqslant n, and let φn\varphi_n be the characteristic function of XnX_n. We have φnf∈\varphi_nf\in dom(mg1+mg2m_{g_1}+m_{g_2}). Since (φnf)(x)(\varphi_nf)(x) tends to f(x)f(x) for every x∈Xx\in X, and since for every n∈Nn\in \mathbf{N} one has ∣φnf∣⩽∣f∣|\varphi_nf|\leqslant |f| and ∣(g1+g2)φnf∣⩽|(g_1+g_2)\varphi_nf|\leqslant ∣(g1+g2)f∣=∣h∣|(g_1+g_2)f|=|h|, with h∈L2(X,μ)h\in \mathscr{L}^2(X, \mu) by hypothesis, Lebesgue's theorem (INT, IV, p. 137, § 3, No.o7^o7, Theorem 6) implies that the sequence of pairs of classes in L2(X,μ)×L2(X,μ)L^2(X, \mu)\times L^2(X, \mu) of (φnf,(g1+g2)φnf)(\varphi_nf,(g_1+g_2)\varphi_nf), which belong to the graph of mg1+mg2m_{g_1}+m_{g_2}, tends to the pair of classes of (f,h)(f, h) in L2(X,μ)L^2(X, \mu). The closure of mg1+mg2m_{g_1}+m_{g_2} is therefore indeed equal to mg1+g2m_{g_1+g_2}.

It is elementary that mg1∘mg2⊂mg1g2m_{g_1}\circ m_{g_2}\subset m_{g_1g_2}. Suppose that g1g_1 is bounded, so that mg1m_{g_1} is a continuous linear mapping on L2(X,μ)L^2(X, \mu). Then the domain of mg2∘mg1m_{g_2}\circ m_{g_1} is the set of classes of functions f∈L2(X,μ)f\in \mathscr{L}^2(X, \mu) such that g2(g1f)∈L2(X,μ)g_2(g_1f)\in \mathscr{L}^2(X, \mu), that is to say the domain of mg1g2m_{g_1g_2}. Hence mg2∘mg1=mg1g2m_{g_2}\circ m_{g_1}=m_{g_1g_2}.

It is likewise elementary that dom(mg1∘mg2m_{g_1}\circ m_{g_2}) == dom(mg2m_{g_2}) is contained in dom(mg1g2m_{g_1g_2}), and that the restriction of mg1g2m_{g_1g_2} to this space is equal to mg1∘mg2m_{g_1}\circ m_{g_2}.

One must beware of believing that mg1∘mg2=mg1g2m_{g_1}\circ m_{g_2}=m_{g_1g_2} in general (exercise 10 of IV, p. 347). Nevertheless, one has the following partial result:

Bourbaki sets the proof directly after the statement, and the corpus does not separate them, so what is above is the statement and its proof together.

Cited by

Nothing.

Cites

Nothing.

Machine translation of the English, by gpt-5.4, against glossary version 34. Not checked by a person.