Bourbaki

Theorem 1 (von Neumann)

Tag 034G, permanent. Cite it as [Bourbaki, Tag 034G].

Théories spectrales, chapter IV, § 4. Opérateurs partiels

Let uu be a closed symmetric operator on a complex Hilbert space E and (E+,E−)(E_+,E_-) its deficiency pair.

a) For every boundary condition bb for uu, the partial operator ubu_b is a closed symmetric extension of uu;

b) The mapping b↦ubb\mapsto u_b is a bijection of the set of boundary conditions for uu onto the set of closed symmetric extensions of uu;

c) For every boundary condition bb for uu, one has

Im(ub+iu_b+i) == Im(u+iu+i)⊕\oplus Im(bb)

Im(ub−iu_b-i) == Im(u−iu-i)⊕\oplus Ker(bb)∘^{\circ};

d) For every boundary condition bb for uu, the deficiency pair of ubu_b is (Ker(bb), Ker(b∗b^*)).

Let bb be a boundary condition for uu and I⊂E+I\subset E_+ its initial subspace. The restriction of u∗u^* to Γb∣I⊂E+⊕E−\Gamma_{b|I}\subset E_+\oplus E_- is the continuous linear mapping defined by x+b(x)↦ix−ib(x)x+b(x)\mapsto ix-ib(x) for x∈Ix\in I. The graph of ubu_b is the direct sum of the graph of uu and the graph of this linear mapping; it is therefore closed. The closed operator ubu_b is an extension of uu such that ub⊂u∗u_b\subset u^*.

Let γ1\gamma_1 and η1\eta_1 be elements of I. Consider the elements

γ=γ1+b(γ1),η=η1+b(η1) \gamma =\gamma_1+b(\gamma_1),\eta =\eta_1+b(\eta_1)

of the graph Γb∣I\Gamma_{b|I}. We calculate

⟨γ∣u∗(η)⟩=i(⟨b(γ1)∣η1⟩−⟨γ1∣b(η1)⟩)+i(⟨γ1∣η1⟩−⟨b(γ1)∣b(η1)⟩) \langle \gamma |u^*(\eta )\rangle =i(\langle b(\gamma_1)|\eta_1\rangle - \langle \gamma_1|b(\eta_1)\rangle )+i(\langle \gamma_1|\eta_1\rangle - \langle b(\gamma_1)|b(\eta_1)\rangle )
=i(⟨b(γ1)∣η1⟩−⟨γ1∣b(η1)⟩) =i(\langle b(\gamma_1)|\eta_1\rangle - \langle \gamma_1|b(\eta_1)\rangle )

since bb is isometric on I. It follows that one has

⟨u∗(γ)∣η⟩=⟨γ∣u∗(η)⟩(12) \langle u^*(\gamma )|\eta \rangle =\langle \gamma |u^*(\eta )\rangle \tag{12}

for all γ\gamma and η\eta in Γb∣I\Gamma_{b|I}.

Let xx and yy be elements of dom(uu),γ,\gamma and η\eta elements of Γb∣I\Gamma_{b|I}. Then

⟨x+γ∣ub(y+η)⟩=⟨x∣u(y)⟩+⟨x∣u∗(η)⟩+⟨γ∣u(y)⟩+⟨γ∣u∗(η)⟩ \langle x+\gamma |u_b(y+\eta )\rangle =\langle x|u(y)\rangle +\langle x|u^*(\eta )\rangle +\langle \gamma |u(y)\rangle +\langle \gamma |u^*(\eta )\rangle
=⟨u(x)∣y⟩+⟨u(x)∣η⟩+⟨u∗(γ)∣y⟩+⟨u∗(γ)∣η⟩ =\langle u(x)|y\rangle +\langle u(x)|\eta \rangle +\langle u^*(\gamma )|y\rangle +\langle u^*(\gamma )|\eta \rangle
=⟨ub(x+γ)∣y+η⟩ =\langle u_b(x+\gamma )|y+\eta \rangle

where we have used the fact that uu is symmetric and that γ\gamma and η\eta belong to the domain of u∗u^*, as well as formula (12). The operator ubu_b is therefore a closed symmetric extension of uu. Assertion a) is proved.

The mapping b↦ubb\mapsto u_b is injective. In fact, a partially isometric mapping of E+E_+ into E−E_- is uniquely determined by its initial subspace and by its restriction to the latter. Now the domain of ubu_b determines the initial subspace I of bb and the graph of the restriction of bb to I.

Let us prove that the mapping b↦ubb\mapsto u_b is surjective. Let ww be a closed symmetric extension of uu. One has w⊂w∗⊂u∗w\subset w^*\subset u^*. The domain of ww is therefore a subspace of dom(u∗u^*) containing dom(uu), and ww is the restriction of u∗u^* to this subspace. Let G be the intersection of the domain of ww and of E+⊕E−E_+\oplus E_-. It is a closed subspace of Eu∗E_{u^*} (lemma 7) and one has dom(ww) == dom(uu)⊕G\oplus G by proposition 26.

Let x∈E+x\in E_+ and y∈E−y\in E_- be such that x+y∈Gx+y\in G. Since ww is symmetric, and is the restriction of u∗u^* to dom(ww), we have

⟨x+y∣u∗(x+y)⟩=⟨u∗(x+y)∣x+y⟩ \langle x+y|u^*(x+y)\rangle =\langle u^*(x+y)|x+y\rangle

This equality is equivalent to

⟨x∣ix⟩+⟨x∣−iy⟩+⟨y∣ix⟩+⟨y∣−iy⟩ \langle x|ix\rangle +\langle x| -iy\rangle +\langle y|ix\rangle +\langle y| -iy\rangle
=⟨ix∣x⟩+⟨ix∣y⟩+⟨−iy∣x⟩+⟨−iy∣y⟩ =\langle ix|x\rangle +\langle ix|y\rangle +\langle -iy|x\rangle +\langle -iy|y\rangle

that is, to ∥x∥2=∥y∥2\|x\|^2=\|y\|^2. In particular, the canonical projection p+:G→E+p_+: G\rightarrow E_+ is injective and, if I denotes its image, the space G is the graph of an isometric mapping b0b_0 of I into E−E_-. For xx in I and yy in E−E_- such that x+y∈Gx+y\in G, we have

∥x+y∥u∗2=∥x∥u∗2+∥y∥u∗2=2(∥x∥2+∥y∥2) \|x+y\|^2_{u^*}=\|x\|^2_{u^*}+\|y\|^2_{u^*}= 2(\|x\|^2+\|y\|^2)
=4∥x∥2=4∥p+(x+y)∥2 = 4\|x\|^2= 4\|p_+(x+y)\|^2

Thus, considering G as a closed subspace of the Hilbert space Eu∗E_{u^*}, the mapping x↦12p+(x)x\mapsto \frac{1}{2}p_+(x) of G into I is an isometry. In particular, I is closed in E+E_+ (lemma 8 of I, p. 107). There then exists a unique boundary condition bb for uu whose initial space is I and which coincides with b0b_0 on I. We have dom(ww) == dom(ubu_b), whence w=ubw=u_b. This establishes assertion b).

Let us prove assertions c) and d). Let bb be a boundary condition for uu and I its initial subspace. For x0∈x_0\in dom(uu) and y∈Iy\in I, we have

ub(x0+y+b(y))−i(x0+y+b(y))=(u−i)(x0)−2ib(y) u_b(x_0+y+b(y))-i(x_0+y+b(y)) = (u-i)(x_0)-2ib(y)

We have Im(u−iu-i) =E−∘= E^{\circ}_- (prop. 25). Since the image of bb is contained in E−E_-, this formula shows that Im(ub−iu_b-i) == Im(u−iu-i)⊕\oplus Im(bb). By loc. cit., we thus have Ker(ub∗+iu^*_b+i) =E−∩= E_-\cap Im(bb)∘=^{\circ}= Ker(b∗b^*).

Analogously, one verifies that Im(ub+iu_b+i) == Im(u+iu+i)⊕I\oplus I and hence (loc. cit.) that Ker(ub∗−iu^*_b-i) =I∘∩E+== I^{\circ}\cap E_+= Ker(bb).

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.