2.3 Characteristics of the Spectral Seminorms
2.3.1 Basic Properties
Proposition 2.9. The families of seminorms q∆and p∆onLandC, respectively, where the symbols∆denote bounded Borel sets, constitute nets with respect to the inclusion relation. For any∆and∆0we have
∆⊆∆0 ⇒ q∆(L)6q∆0(L), L∈L,
∆⊆∆0 ⇒ p∆(C)6p∆0(C), C∈C.
Proof. For the q∆-seminorms on L the assertion follows from the order relation for operators inB(H)+. Let L belong to the left idealL, then
Q(L∆∗L)6Q(L∆0∗L), which by Definition2.8has the consequence
q∆(L)2= Q(L∆∗L)
6
Q(L∆0∗L)
=q∆0(L)2. This relation extends by continuity of the seminorms to all ofL.
In case of the p∆-topologies, note that for any Borel set ∆ the functional φE(∆), defined throughφE(∆)(.) .
=φ(E(∆).E(∆)), belongs toB(H)∗,1ifφdoes. From this we infer, since moreover∆⊆∆0implies E(∆) =E(∆)E(∆0) =E(∆0)E(∆), that
nZ
Rs
dsx
φ E(∆)αx(C)E(∆)
:φ∈B(H)∗,1o
⊆nZ
Rs
dsx
φ E(∆0)αx(C)E(∆0)
:φ∈B(H)∗,1o for any C∈Cand thus, by (2.17b), that p∆(C)6p∆0(C), a relation which by continuity of the seminorms is likewise valid for any operator in the completionC.
The continuous extensions of the seminorms q∆ and p∆ toL andC, respectively, can be explicitly computed on the subspacesAL andACofA.
2.3 Characteristics of the Spectral Seminorms 15
Lemma 2.10. Let∆denote an arbitrary bounded Borel subset ofRs+1. (i) For any L∈ALwe have
q∆(L) =sup nZ
Rs
dsxω E(∆)αx(L∗L)E(∆)
:ω∈B(H)+∗,1 o1/2
. (2.18a)
(ii) For any C∈ACthere holds the relation p∆(C) =sup
nZ
Rs
dsx
φ E(∆)αx(C)E(∆)
:φ∈B(H)∗,1
o
. (2.18b)
Proof. (i) Note, that we can define a linear subspaceAq0 ofAconsisting of all those operators L0which fulfill
q0∆(L0)2 .
=sup nZ
Rs
dsxω E(∆)αx(L0∗L0)E(∆)
:ω∈B(H)+∗,1o
<∞
for any bounded Borel set∆. On this space the mappings q0∆act as seminorms whose restrictions toLcoincide with q∆ (cf. the Remark following Definition2.8). Now let L∈ALbe arbitrary. Given a bounded Borel set∆we can then find a sequence
Ln n∈N
inLsatisfying
nlim→∞q∆(L−Ln) =0 and lim
n→∞kL−Lnk=0.
The second equation implies
n→∞limkLE(∆)−LnE(∆)k=0,
so that Lebesgue’s Dominated Convergence Theorem can be applied to get for any functionalω∈B(H)+∗,1and any compact K⊆Rs
Z
K
dsxω E(∆)αx(L∗L)E(∆)
=lim
n→∞
Z
K
dsxω E(∆)αx(Ln∗Ln)E(∆) .
According to (2.17c) each term in the sequence on the right-hand side is majorized by the corresponding q∆(Ln)2 and this sequence in turn converges to q∆(L)2 by assump-tion, so that in passing from K to Rs and to the supremum over all ω∈B(H)+∗,1we get
sup nZ
Rs
dsxω E(∆)αx(L∗L)E(∆)
:ω∈B(H)+∗,1o
6q∆(L)2.
This final estimate shows, by arbitrariness of L∈AL and the selected∆, thatAL is a subspace ofAq0 and, from q0∆L=q∆, it eventually follows that for all these L and∆
q∆(L) =supnZ
Rs
dsxω E(∆)αx(L∗L)E(∆)
:ω∈B(H)+∗,1o1/2
.
16 Localizing Operators and Spectral Seminorms
(ii) The proof of the second part follows the same lines of thought. We introduce the subspaceAp0⊆Aconsisting of operators C0satisfying
p0∆(C0) .
=supnZ
Rs
dsx
φ E(∆)αx(C0)E(∆)
:φ∈B(H)∗,1o
<∞
for any bounded Borel set∆and furnish it with the locally convex topology defined by the seminorms p0∆. An arbitrary C∈ACis, for given ∆, approximated by a sequence Cn n∈Nwith respect to the norm and the p∆-topology. As above one has
n→∞limkE(∆)CE(∆)−E(∆)CnE(∆)k=0 and infers
sup nZ
Rs
dsx
φ E(∆)αx(C0)E(∆)
:φ∈B(H)∗,1
o
6p∆(C).
This establishes, by arbitraryness of C∈AC and∆, that AC⊆Ap0, and the equation p0∆C=p∆implies that for these C and∆
p∆(C) =sup nZ
Rs
dsx
φ E(∆)αx(C)E(∆)
:φ∈B(H)∗,1
o .
An immediate consequence of this result is the subsequent lemma, which in some way reverts the arguments given in the concluding remark of the last section in order to establish the Hausdorff property for(L,Tq)and(C,Tp).
Lemma 2.11. Let∆be a bounded Borel set.
(i) For L∈ALwith LE(∆) =0 there holds q∆(L) =0.
(ii) If C∈ACsatisfies E(∆)CE(∆) =0 one has p∆(C) =0.
Next we deal with an implication of the fact, thatLis an ideal of the C∗-algebraA, and clarify the relationship between the seminorms q∆and p∆.
Lemma 2.12. Let∆denote bounded Borel subsets ofRs+1. (i) AL is a left ideal of the quasi-local algebraAand satisfies
q∆(AL)6kAkq∆(L) (2.19)
for any L∈ALand A∈A.
(ii) Let Li, i=1,2, be operators inAL and A∈A, then L1∗AL2 belongs toAC. If in addition the operators Li have energy-momentum transfer inΓi⊆Rs+1 and∆i are Borel subsets ofRs+1containing∆+Γi, respectively, then
p∆(L1∗AL2)6kE(∆1)AE(∆2)kq∆(L1)q∆(L2). (2.20) Proof. (i) For any L∈AL⊆A and arbitrary A∈Athe relation L∗A∗AL6kAk2L∗L leads to the estimate
Z
Rs
dsxω E(∆)αx(L∗A∗AL)E(∆)
6kAk2 Z
Rs
dsxω E(∆)αx(L∗L)E(∆)
2.3 Characteristics of the Spectral Seminorms 17
for anyω∈B(H)+∗,1and thus, by (2.18a) and the notation of the proof of Lemma2.10, to
q0∆(AL) =supnZ
Rs
dsxω E(∆)αx(L∗A∗AL)E(∆)
:ω∈B(H)+∗,1o1/2
6kAksup nZ
Rs
dsxω E(∆)αx(L∗L)E(∆)
:ω∈B(H)+∗,1 o1/2
=kAkq∆(L).
This shows that AL belongs toALand at the same time that the seminorm q0∆(onAq0) can be replaced by q∆to yield (2.19).
(ii) Letφbe a normal functional onB(H)withkφk61. By polar decomposition there exist a partial isometry V and a positive normal functional|φ|withk|φ|k61 such that φ(.) =|φ|(.V). Then
φ E(∆)αx(L1∗AL2)E(∆)
=|φ| E(∆)αx(L1∗)E(∆1)αx(A)E(∆2)αx(L2)E(∆)V 6kE(∆1)αx(A)E(∆2)k
q
|φ| E(∆)αx(L1∗L1)E(∆)q
|φ| V∗E(∆)αx(L2∗L2)E(∆)V for any x∈Rs and the method used in the proof of Proposition2.7 can be applied to get, in analogy to (2.16),
supnZ
Rs
dsx
φ E(∆)αx(L1∗AL2)E(∆)
:φ∈B(H)∗,1o
6kE(∆1)AE(∆2)kq∆(L1)q∆(L2), where we made use of (2.18a). According to the notation introduced in the proof of Lemma2.10this result expressed in terms of the seminorm p0∆ onAp0 reads
p0∆(L1∗AL2)6kE(∆1)AE(∆2)kq∆(L1)q∆(L2),
from which we infer, as in the first part of the present proof, not only that L1∗AL2is an element ofACbut also that p0∆can be substituted by p∆ to give (2.20).
The second part of the above lemma means that the product L1∗L2, defined by two operators L1,L2∈AL, is continuous with respect to the locally convex spaces (cf. [44, Chapter Four, § 18, 3.(5)])(AL,Tuq)×(AL,Tuq)and(AC,Tup).
Corollary 2.13. The sesquilinear mapping on the topological product of the locally convex space(AL,Tuq)with itself, defined by
AL×AL3(L1,L2)7→L1∗L2∈AC,
is continuous with respect to the respective locally convex topologies.
In the special case of coincidence of both operators (L1 =L2 =L) it turns out that p∆(L∗L)equals the square of q∆(L). Another result involving the operation of adjunc-tion is the fact, that this mapping leaves the p∆-seminorms invariant.
Lemma 2.14. Let∆denote the bounded Borel sets inRs+1.
18 Localizing Operators and Spectral Seminorms
(i) For any operator L∈AL there hold the relations p∆(L∗L) =q∆(L)2. whereas the reverse inequality is a consequence of Lemma2.12. This proves the asser-tion.
(ii) Note, thatB(H)∗,1 is invariant under the operation of taking adjoints defined by ψ7→ψ∗withψ∗(A) . for any C∈Ap0 (cf. the proof of Lemma2.10), which is sufficient to establish both of the assertions.
The last statement of this subsection on basic properties of the spectral seminorms establishes their invariance under translations in the s+1-dimensional configuration space.
Lemma 2.15. The subspaces AL andAC of the quasi-local algebraAare invariant under translations. In particular, let∆be a bounded Borel set inRs+1and let x∈Rs+1 be arbitrary, then
2.3 Characteristics of the Spectral Seminorms 19
Therefore the introductory remark in combination with (2.18a) yields for any L∈AL: q0∆ αx(L)2 which, as in the proof of Lemma2.12, establishes the assertions.
(ii) The same argument applies to the seminorm p0∆, so that for C∈AC
The assumed strong continuity of the automorphism groupα(Λ,x):(Λ,x)∈P+↑ acting on the C∗-algebraAcarries over to the locally convex spaces(L,Tq)and(C,Tp); and even the infinite differentiability of (Λ,x)7→ α(Λ,x)(L0) for L0∈L0 is conserved in passing from the uniform topology onL0to that induced by the seminorms q∆. Proposition 2.16. (i) For fixed L∈Lthe mapping
ΞL:P+↑ →L (Λ,x)7→ΞL(Λ,x) .
=α(Λ,x)(L) is continuous with respect to the locally convex space(L,Tq).
(ii) For given C∈Cthe mapping
ΞC:P+↑ →C (Λ,x)7→ΞC(Λ,x) .
=α(Λ,x)(C) is continuous with respect to the locally convex space(C,Tp).
(iii) Let idL0 denote the identity mapping
idL0 :(L0,k.k)→(L0,Tq) L07→idL0(L0) .
=L0
on the spaceL0 , once endowed with the norm topology and once withTq. Consider furthermore the familyXL0 of infinitely often differentiable mappings:
XL0
=. ΞL0: L0∈L0 , ΞL0 :P+↑ →L0 (Λ,x)7→ΞL0(Λ,x) .
=α(Λ,x)(L0).
Then the linear operator idL0 isXL0-differentiable in the sense of DefinitionA.16.