4. Projection Lemmas
Let \mathcal{A} be a unital C^*-algebra in this chapter. Here we collect relevant
results about (self-adjoint) projections in \mathcal{A}. Recall that an element p in a
C^*-algebra is a projection if p^2 = p^* = p. By the Spectral Mapping Theorem (or using
the full Gelfand Duality), a self-adjoint element a \in \mathcal{A} is a projection if and
only if the spectrum of a is contained in \{0, 1\}.
For all projections p \in \mathcal{A}, p \le 1.
Lean code for Lemma4.1●1 theorem
Associated Lean declarations
-
IsStarProjection.le_one[complete]
-
IsStarProjection.le_one[complete]
-
theoremdefined in Mathlib/Algebra/Order/Star/Basic.leancomplete
theorem IsStarProjection.le_one.{u_1} {R : Type u_1} [Ring R] [PartialOrder R] [StarRing R] [StarOrderedRing R] {p : R} (hp : IsStarProjection p) : p ≤ 1
theorem IsStarProjection.le_one.{u_1} {R : Type u_1} [Ring R] [PartialOrder R] [StarRing R] [StarOrderedRing R] {p : R} (hp : IsStarProjection p) : p ≤ 1
We can get this by Gelfand duality (cf. Sakai 1.2.3) and Lemma 3.3.
Let \mathcal{A} be a C^*-algebra and p, q \in \mathcal{A} be projections. Then
p - q \ge 0 if and only if qp = pq = q.
Lean code for Lemma4.2●1 theorem
Associated Lean declarations
-
IsStarProjection.le_tfae[complete]
-
IsStarProjection.le_tfae[complete]
-
theoremdefined in Mathlib/Analysis/CStarAlgebra/Projection.leancomplete
theorem IsStarProjection.le_tfae.{u_1} {A : Type u_1} [NonUnitalCStarAlgebra A] [PartialOrder A] [StarOrderedRing A] {p q : A} (hp : IsStarProjection p) (hq : IsStarProjection q) : [p ≤ q, q * p = p, p * q = p, IsStarProjection (q - p), IsIdempotentElem (q - p)].TFAE
theorem IsStarProjection.le_tfae.{u_1} {A : Type u_1} [NonUnitalCStarAlgebra A] [PartialOrder A] [StarOrderedRing A] {p q : A} (hp : IsStarProjection p) (hq : IsStarProjection q) : [p ≤ q, q * p = p, p * q = p, IsStarProjection (q - p), IsIdempotentElem (q - p)].TFAE
By Corollary 3.2, if p - q \ge 0 then
q(p - q)q = qpq - q^3 = qpq - q \ge 0. By Lemma 4.1, we have p \le 1,
and employing Corollary 3.2 we obtain qpq \le q1q = q. Since
qpq \ge q and qpq \le q we have qpq = q, which implies that q(p - q)q = 0. By the
C^*-property of the norm, we have \|(p - q)^{1/2} q\|^2 = \|q(p - q)q\| = 0 and so
(p - q)^{1/2} q = 0 and therefore (p - q)q = (p - q)^{1/2}(p - q)^{1/2} q = 0. It
follows that pq = q, and taking adjoints, qp = q. Conversely, if qp = pq = q, one
easily checks that p - q is a projection, and so its spectrum is contained in
\{0, 1\} and it is positive.
For all projections p, q \in \mathcal{A} such that q \le p, p - q is a projection.
Lean code for Corollary4.3●1 theorem
Associated Lean declarations
-
IsStarProjection.le_tfae[complete]
-
IsStarProjection.le_tfae[complete]
-
theoremdefined in Mathlib/Analysis/CStarAlgebra/Projection.leancomplete
theorem IsStarProjection.le_tfae.{u_1} {A : Type u_1} [NonUnitalCStarAlgebra A] [PartialOrder A] [StarOrderedRing A] {p q : A} (hp : IsStarProjection p) (hq : IsStarProjection q) : [p ≤ q, q * p = p, p * q = p, IsStarProjection (q - p), IsIdempotentElem (q - p)].TFAE
theorem IsStarProjection.le_tfae.{u_1} {A : Type u_1} [NonUnitalCStarAlgebra A] [PartialOrder A] [StarOrderedRing A] {p q : A} (hp : IsStarProjection p) (hq : IsStarProjection q) : [p ≤ q, q * p = p, p * q = p, IsStarProjection (q - p), IsIdempotentElem (q - p)].TFAE
The next result is specific to W^*-algebras. Sakai obtains it from the fact that the
projections of a W^*-algebra form a complete lattice
(Proposition 8.2.4, via the correspondence between
\sigma-closed one-sided ideals and projections in
Proposition 8.2.3). The proof of
Theorem 7.2.6 only needs suprema of directed sets, for which the
following direct argument suffices.
In a W^*-algebra M, the supremum of a nonempty directed set of projections is a projection.
Lean code for Lemma4.4●1 theorem
Associated Lean declarations
-
Ultraweak.isStarProjection_of_isLUB[complete]
-
Ultraweak.isStarProjection_of_isLUB[complete]
-
theoremdefined in LeanOA/Ultraweak/Normal.leancomplete
theorem Ultraweak.isStarProjection_of_isLUB.{u_1, u_2} {M : Type u_1} {P : Type u_2} [CStarAlgebra M] [PartialOrder M] [StarOrderedRing M] [NormedAddCommGroup P] [NormedSpace ℂ P] [CompleteSpace P] [Predual ℂ M P] {d : Set M} (hne : d.Nonempty) (hd : DirectedOn (fun x1 x2 ↦ x1 ≤ x2) d) (hdp : ∀ q ∈ d, IsStarProjection q) {e : M} (he : IsLUB d e) : IsStarProjection e
theorem Ultraweak.isStarProjection_of_isLUB.{u_1, u_2} {M : Type u_1} {P : Type u_2} [CStarAlgebra M] [PartialOrder M] [StarOrderedRing M] [NormedAddCommGroup P] [NormedSpace ℂ P] [CompleteSpace P] [Predual ℂ M P] {d : Set M} (hne : d.Nonempty) (hd : DirectedOn (fun x1 x2 ↦ x1 ≤ x2) d) (hdp : ∀ q ∈ d, IsStarProjection q) {e : M} (he : IsLUB d e) : IsStarProjection e
In a W⋆-algebra, the supremum of a nonempty directed set of projections is a projection.
Let e be the supremum of the directed set (q_\alpha) of projections; it is the
\sigma-limit of the q_\alpha by Theorem 7.1.4, hence is
self-adjoint by Lemma 5.2. Since q_\alpha \le e \le 1, we
have 0 \le q_\alpha (1 - e) q_\alpha \le q_\alpha (1 - q_\alpha) q_\alpha = 0, and so
(1 - e) q_\alpha = 0 by the C^*-identity, i.e., e q_\alpha = q_\alpha. Multiplication is
separately \sigma-continuous (Lemma 2.2.6), so
q_\alpha = e q_\alpha \to e^2, but also q_\alpha \to e. Hence e^2 = e.