1

Microlocal Analysis of a Deformed Quantum Field Theory

Rishabh Ballal
Institute for Theoretical Physics, Leipzig University
E-mail: [email protected]
arXiv:2412.20511v1 [math-ph] 29 Dec 2024

Albert Much
Institute for Theoretical Physics, Leipzig University
E-mail: [email protected]

Rainer Verch
Institute for Theoretical Physics, Leipzig University
E-mail: [email protected]

Abstract: A deformation technique, known as the warped convolution, takes quantum

fields in Minkowski spacetime to quantum fields in noncommutative Minkowski space-
time. Since a quantum field is an operator-valued regular distribution and the warped
convolution is (weakly) an oscillatory integral of Rieffel-type, we prove that the sym-
bol classes introduced by Hörmander admit extensions which are suited to the warped
convolutions of scalar quantum field operators. We further show that, if a particular
vector state on the undeformed algebra of field operators fulfills the microlocal spec-
trum condition, then every vector state on the deformed algebra generated by these
warped convolutions fulfills the microlocal spectrum condition.

1 Introduction
Due to Heisenberg’s uncertainty principle, the precision of localisation of a system in
spacetime is inversely proportional to that of its energy-momentum. If a measurement
is then so precise as to initiate gravitational collapse, the information of the system
will be obscured. This operational obstacle can be overcome by [DFR95, § 1] uncer-
tainty relations of Planckian order on spacetime itself, and these relations are realised
by a noncommutative algebra.
Let X0 , . . . , Xd−1 denote the unbounded self-adjoint position operators in a Hilbert
space V . Imposing the commutation relations [Xµ , Xν ] = i Qµν idV , where Q is a real
and skew-symmetric matrix, results in a noncommutative Minkowski spacetime of d-
2

dimensions. Using techniques of Weyl quantisation, the scalar quantum field Φ in the
Fock space F should now be represented in the tensor product V ⊗ F [DFR95, § 6],
which we denote formally as Φ⊗ . However, there exists a deformation ΦQ ([ABJJ07],
[GL07]) of Φ in F alone such that the commutation relations of Φ⊗ and ΦQ coincide.
In particular, Grosse and Lechner [GL07] constructed a unitary map from V ⊗ F to
F , and with it, proved the unitary equivalence of the mentioned quantum fields.
Without recourse to the Fock space, this deformation is implemented in the frame-
work of warped convolutions ([BS08], [BLS11]). Let the scalar quantum field Φ be a
regular distribution that is valued as an unbounded operator in a Hilbert space H .
If U is a strongly continuous unitary representation on H of the identity component
R R
of the Poincaré group d ⋊ O(1, d − 1), with d as the translation subgroup, then the
warped convolution WQ [Φ(x)] of the distributional kernel Φ(x) acts as
Z
−d
WQ [Φ(x)] Ψ = (2π) e−iη(θ,ξ) U (Qθ) Φ(x) U (−Qθ) U (ξ) Ψ d(θ, ξ) (1.1)
Rd ×Rd
R
on vectors Ψ that lie in the domain of Φ(f ) = Φ(x) f (x) dx and are smooth under
R
the action d ∋ a 7→ U (a) Ψ ∈ H of translations. The Minkowski metric η endows,
in a weak sense, the warped convolution with the structure of an oscillatory integral
of Rieffel-type [Rie93, Chap. 1]. In lieu of Rieffel’s analysis, his oscillatory integral
Z
Iη (s) = (2π)−k e−iη(θ,ξ) s(θ, ξ) d(θ, ξ) (1.2)
Rk ×Rk

R
will be understood as a continuous linear functional on a subspace Sρm ( kθ × kξ ) of R
R R R
Hörmander’s symbol classes Sρm ( sx ×( kθ × kξ )) [Hö71, Chap. I]. This is possible since
I
the oscillatory integral η itself (the part excluding the symbol s) has no dependence
on x. A notable consequence [LW11, Thm. 3.2] is the consideration of ρ in the range
(−1, 1] as opposed to Hörmander’s range (0, 1]. However, the dependence of symbols
on x needs to be re-introduced for the sake of Φ. As per Hörmander’s definition, all
R R R R R
symbols belong in C ∞ ( sx × ( kθ × kξ )), and so any Sρm ( kθ × kξ )-valued distribution
R R R R
on sx which integrates a symbol in Sρm ( sx ×( kθ × kξ )) against functions in Cc∞ ( sx ) R
has an empty wavefront set. Contrariwise, the correlations of field operators within a
‘physical’ state, which arise in the symbols of the products of warped convolutions, are
notorious for their singular character and are merely integrable locally. This poses the
first main question to be addressed:

Is there an extension of Hörmander’s symbol classes for Rieffel’s oscillatory integral,

R R
such that the symbols are locally integrable on sx and smooth on kθ × kξ ? R
 3

m,ρ
An affirmation is given in Section 3 by introducing the extended symbol classes Xloc .
1 R
s
These spaces have the structure of Lloc ( x )-modules (Lemma 3.8), and the extended
R R R
symbols induce Sρm ( kθ × kξ )-valued regular distributions on sx (Proposition 3.9), or
symbolic distributions in short. Since an oscillatory integral composes with a symbolic
R
distribution to define a distribution on sx , it is a simple consequence of the continuity
of oscillatory integrals that the wavefront set of this composition lies within the wave-
front set of the symbolic distribution (Theorem 3.12).
After the construction of the scalar quantum field Φ from the Borchers-Uhlmann
algebra in Section 4, the above constructions are applied in Section 5 to show that the
warped convolution of a field operator is an oscillatory integral; if Ψ1 and Ψ2 belong
to a suitable domain of smoothness, then the inner product

hΨ1 , U (Qθ) Φ(x) U (−Qθ) U (ξ) Ψ2 i (1.3)

and its smeared version

hΨ1 , U (Qθ) Φ(f ) U (−Qθ) U (ξ) Ψ2 i (1.4)

define an extended symbol and a symbolic distribution, respectively (Theorem 5.2). In

the light of Poincaré covariance for warped convolutions, a deformed algebra can only
be generated from the warped convolutions of field operators associated with matrices
in the orbit ΣQ := {ΛQΛT : Λ ∈ L↑+ }, under the action of the identity component L↑+
of the Poincaré subgroup O(1, d − 1). Consequently, it becomes necessary to alter the
definition of n-point distributions for admissible vector states (Definition 5.7).
The interest in wavefront sets stems from the aforementioned correlations of fields,
which depend on the state being considered. If ω is a quasi-free state on the ∗-algebra
of the Klein-Gordon field Φ in a globally hyperbolic spacetime, then it is characterised
as being a physical state if the singular part of the function ω2 (x, y) = ω(Φ(x) Φ(y)) is
identical to that of the Minkowski vacuum. Before Radzikowski, this characterisation
expressed ω2 as a sum of the Hadamard parametrix [KW91, § 3.3] and a smooth term,
and referred to ω as a Hadamard state. With microlocal analysis, Radzikowski proved
[Rad96] that a quasi-free state ω is Hadamard if and only if the (smooth) wavefront
set of the two-point distribution, induced from ω2 and denoted identically, satisfies

WF(ω2 ) = (x1 , ζ1 ; x2 , −ζ2 ) ∈ N+ × N− : (x1 , ζ1 ) ∼1 (x2 , ζ2 ) , (1.5)

where N± ⊂ T∗ R4 represent the bundles of future(+)/past(−)-directed null covectors,

4

and the equivalence relation (x1 , ζ1 ) ∼1 (x2 , ζ2 ) marks the existence of a null geodesic
γ from x1 to x2 such that ζ1 is cotangent to γ at x1 and ζ2 is cotangent to γ at x2 .
This also applies to vector-valued quantum fields ([SV00], [SV01]). In the absence of
the Klein-Gordon equation, Brunetti et al. [BFK96] presented a weaker characterisa-
tion, designated the microlocal spectrum condition (µSC; Definition 6.1), in which

WF(ω2 ) ⊆ (x1 , ζ1 ; x2 , −ζ2 ) ∈ J+ × J− : (x1 , ζ1 ) ∼2 (x2 , ζ2 ) , (1.6)

R
where J± ⊂ T∗ 4 denote the bundles of future(+)/past(−)-directed causal covectors,
while the equivalence relation (x1 , ζ1 ) ∼2 (x2 , ζ2 ) denotes the existence of a piecewise
smooth curve γ from x1 to x2 such that the parallel transport of ζ1 along γ coincides
with ζ2 at x2 . Subsequently, an analytic counterpart of the µSC was developed using
analytic wavefront sets in analytic spacetimes ([SVW02], [SW24]). In both, (classes of)
globally hyperbolic spacetimes and analytic spacetimes, Gérard and Wrochna utilised
techniques of pseudo-differential calculus to tackle questions concerning the existence
and construction of corresponding Hadamard states ([GW14], [GW19]). Here, we will
not consider the analytic situation; grounded in Minkowski spacetime, we implement
simple methods of microlocal analysis in addressing the second main question:

When does a vector state on the ∗-algebra generated by the warped convolutions
of scalar quantum field operators fulfill the microlocal spectrum condition?

Note that every admissible vector state is defined with a vector from the underlying
dense domain of the algebra. As it turns out, the criterion for this fulfillment is quite
natural and accumulates in the main physical result of this work (Theorem 6.4):

Theorem.
R
If the state ω on the algebra P( 4 ) of field operators fulfills the µSC, then so does every
R
admissible vector state on the algebra P( 4 , ΣQ ) of warped field operators.

R
Here, ω is viewed as a vector state on the algebra P( 4 ) of field operators, but
is more importantly recognised as the state on the Borchers-Uhlmann algebra which
constructs the Gelfand-Naimark-Segal Hilbert space in which the field operators exist.
R
On the other hand, P( 4 , ΣQ ) is the ∗-algebra of field operators deformed under the
warped convolutions associated with the orbit ΣQ = {ΛQΛT : Λ ∈ L↑+ }. Clearly, ΣQ
presupposes the choice of a reference matrix Q, which can lead to different properties
of the resulting theory; for example, it can impose locality in wedges ([GL07], [GL08],
[BS08], [BLS11]) or enable consistent Wick rotations [GLLV13]. The choice of Q in the
 5

latter is to ensure that time remains commutative, and has previously been considered
by one of the present authors in analysing quantum Dirac fields in Moyal-Minkowski
spacetime [Ver11]. This, in itself, is an example of the joint development with Paschke
[PV04] of (locally covariant) quantum field theory in a Lorentzian analogue of Connes’
noncommutative geometry, so as to provide a rigorous understanding of the geometry
of quantum spacetime. In contrast, the deformation in commutative Minkowksi space-
time does not reflect any mathematical obstacle to the notion of a spacetime point.
An imposition of this sort, on physical grounds, would then necessitate a re-definition
of the wavefront set, or possibly of the asymptotic correlation spectrum [Ver99]. Here,
we maintain the standard definitions and treat the deformation in its own right.

2 Distributions and Wavefront Sets

So as to keep this article self-contained, we recall some basic definitions and results of
distribution theory and wavefront sets.
First, consider the following spaces of functions. Let (Kj )j∈N be an exhaustion of
R s by compact sets, that is, a sequence of compact sets such that K is contained in
j
R S
R
the interior of Kj+1 , and s = j Kj . The space C ∞ ( s ) of smooth complex-valued
R R
functions on s is a Fréchet space E ( s ) equipped with the family of semi-norms

tα,Kj (ψ) := sup |D α ψ(x)|, ψ ∈ E( Rs), (2.1)

x∈Kj

lending the topology of uniform convergence in all derivatives on compact sets. The
R R
space Cc∞ ( s ) of compactly supported smooth functions on s is a subspace D( s ) R
R
of E ( s ), endowed with a locally convex topology as the strict inductive limit [SW99,
§ II.6] of the Fréchet spaces

Rs ) := {ψ ∈ E (Rs) : supp ψ ⊂ Kj },
DKj ( (2.2)

each of which inherits its topology from E ( Rs) [Gru09, § 2.1].

2.1 Definition. ([Gru09, § 3.3, 3.1], [Str09, Defs. 1, 2])
R R
The topological dual of E ( s ) is the space E ′ ( s ) of compactly supported distributions,
R R
and the topological dual of D( s ) is the space D ′ ( s ) of distributions. ⋄

R R R R R
The inclusion D( s ) ⊂ E ( s ) implies E ′ ( s ) ⊂ D ′ ( s ). Both spaces, E ′ ( s ) and
R
D ′ ( s ), are locally convex under the weak-∗ topology [Con07, Def. V.1.1]. Often, the
6

characterisation in the next proposition is used as the definition of a distribution.

2.2 Proposition. [Hö03, Thm. 2.1.4]

R
A linear functional u on D( s ) is a distribution if and only if, for every j ∈ N, there
N
exist constants kj ∈ 0 and Cj > 0 such that
X
|u(ψ)| ≤ Cj tα,Kj (ψ), ∀ ψ ∈ DKj ( Rs ). (2.3)
|α|≤kj

R R R R
The evaluation of u ∈ D ′ ( s ) at ψ ∈ D( s ) (or u ∈ E ′ ( s ) at ψ ∈ E ( s )) will be
denoted by the bilinear form ≺ u, ψ ≻ := u(ψ). A class of examples which offer great
insight into the properties of distributions are the regular distributions; every locally
R R
integrable function f defines via the injective map Λ : L1loc ( s ) → D ′ ( s ) a regular
distribution
Z
≺ Λ(f ), ψ ≻ := f (x) ψ(x) dx, ∀ ψ ∈ D( s ), R (2.4)

R
since |≺ Λ(f ), ψ ≻| ≤ kf kL1 (Kj ) t0,Kj (ψ) when ψ ∈ DKj ( s ). Denote the derivative as
D α := (−i)|α| ∂ α . The problem of interchanging integrals and derivatives for regular
distributions generalises to arbitrary distributions as follows.

2.3 Lemma. [FJ98, Thm. 4.1.1]

R R R R
Let u ∈ D ′ ( s1 ) and ψ ∈ E ( s1 × s2 ). If every y ′ ∈ s2 is contained in some open set
R
Uy′ such that x 7→ ψ(x, y) is supported in a compact set Ky′ ⊂ s1 for all y ∈ Uy′ , then
R
≺ u, ψ(−, y) ≻ ∈ D( s2 ) and Dyα ≺ u, ψ(−, y) ≻ = ≺ u, Dyα ψ(−, y) ≻.

This result is instrumental in the construction of tensor products of distributions

[Hö03, Thm. 5.1.1], which, in turn, is useful for the following Schwartz kernel theorem.

2.4 Theorem. [Hö03, Thm. 5.1.2]

R R
Every distribution u ∈ D ′ ( s1 × s2 ) induces a continuous linear map K : D( Rs ) →
2

R
D ′ ( s1 ) under the relation

≺ Kψ, ϕ ≻ = ≺ u, ϕ ⊗ ψ ≻, (2.5)

and conversely, every such continuous linear map K admits a unique distribution u for
which relation (2.5) holds.
 7

R
Let u ∈ E ′ ( s ). Consistent with regular distributions, the Fourier transform Fu of
R
u is defined as ≺ Fu, ψ ≻ := ≺ u, Fψ ≻, for all ψ ∈ E ( s ). When restricted to D( s ), R
the uniqueness of tensor products [Hö03, Thm. 5.1.1] indicates that ≺ u, Fψ ≻ can be
rewritten as [Hö03, Thm. 7.1.14]
Z Z
≺ u, ψ(ζ) eiη(−,ζ) dζ ≻ = ≺ u ⊗ Λ(ψ), eiη(−,−) ≻ = ≺ u, eiη(−,ζ) ≻ ψ(ζ) dζ, (2.6)

R
for all ψ ∈ D( s ). In this case, the Fourier transform of u is simply understood as
a regular distribution induced by the function Fu(ζ) := ≺ u, eiη(−,ζ) ≻. As is the case
for compactly supported smooth functions, u is smooth if and only if Fu is decreas-
ing rapidly ([Hö03, Thm. 7.3.1], [Str09, Thm. 6]). This global characterisation admits
a refinement to ‘microlocally’ capture the singular behaviour of distributions.

2.5 Definition. [FJ98, Def. 11.1.3]

R R
Let u ∈ D ′ ( s ). A point (x, ζ ′ ) ∈ T∗ s \ Z, where Z denotes the zero section, is a
regular direction for u if there exist
R
(i) ϕ ∈ D( s ) with ϕ(x) 6= 0, and
(ii) an open conic1 neighbourhood Γ of ζ ′ in which

sup (1 + kζk)N |F(ϕu)(ζ)| < ∞, ∀N ∈ N. (2.7)

ζ∈Γ

The complement of the set of regular directions determines the closed cone

WF(u) := (x, ζ ′ ) ∈ T∗ Rs \ Z : (x, ζ ′ ) is not a regular direction for u , (2.8)

known as the (smooth) wavefront set of u. ⋄

The wavefront set is a notable tool of ‘microlocalisation’; first, ϕ ∈ D( s ) localises R

R
a distribution u ∈ D ′ ( s ) in the base space and then the Fourier transform F(ϕu)(ζ)
identifies the non-regular directions in the fibre. Note that the second condition of a
regular direction is adjusted if u is valued in a Fréchet space with semi-norms pj ; one
replaces |F(ϕu)(ζ)| in (2.7) with pj (F(ϕu)(ζ)), for all j ∈ . N
Also, the projection π2 (WF(u)) := {ζ : (x, ζ) ∈ WF(u) for some x ∈ s } of the R
wavefront set onto its covectors will prove beneficial in Section 6. If u is a compactly
supported distribution, this projection is equivalent to the set of directions in which
the Fourier transform Fu is not rapidly decreasing [Hö03, Prop. 8.1.3].
1
A cone Γ ⊆ T∗ R s
\ Z is implicitly defined: if (x, ζ) ∈ Γ, then (x, tζ) ∈ Γ for all t > 0.
8

3 Rieffel’s Oscillatory Integrals

R R R
Let η be a non-degenerate bilinear form on k and let χ ∈ Cc∞ ( kθ × kξ ) be such that
χ(0, 0) = 1. Rieffel considered the oscillatory integral [Rie93, Chap. 1]
Z
Iη (s) := (2π)−k ǫց0
lim e−iη(θ,ξ) χ(ǫθ, ǫξ) s(θ, ξ) d(θ, ξ), (3.1)
Rk ×Rk

R
with technical assumptions on s suited to an isometric action of k . In its stead, s will
R R
belong to a ‘symbol class’ Sρm ( kθ × kξ ) in a style similar to Hörmander’s [Hö71, § 1.1],
but with minor adjustments introduced by Lechner and Waldmann [LW11, § 2].

3.1 Definition.
Let m ∈ Rand −1 < ρ ≤ 1. The vector space of all s ∈ C ∞ ( Rkθ × Rkξ ) that fulfill
|Dθβ Dξα s(θ, ξ)| ≤ Cα,β (1 + k(θ, ξ)k)m−ρ(|α|+|β|) (3.2)

is a symbol class Sρm (Rkθ × Rkξ ), and s is a symbol of order m and type ρ. ⋄

Contrary to Hörmander’s definition [Hö71, Def. 1.1.1], the symbols are assumed to
be polynomially bounded in both, θ and ξ, with the same type ρ. This is in anticipa-
R R R R R
tion of kθ × kξ being fibres in the bundle sx × ( kθ × kξ ), whose introduction will be
R
delayed so as to accommodate lower regularity in the base space sx . Another point of
contrast is the larger domain of ρ; its validity will be shown in Theorem 3.5.
R R
Henceforth, Sρm ( kθ × kξ ) will be abbreviated simply as Sρm . Every symbol class Sρm
is a Fréchet space with the countable family of semi-norms

pα,β (s) := sup |Dθβ Dξα s(θ, ξ)| (1 + k(θ, ξ)k)−m+ρ(|α|+|β|) (3.3)
(θ,ξ)∈Rk ×Rk

that determine the topology [GS94, Chap. 1]. As evident from the definition, symbol
classes become larger if m increases or ρ decreases: that is, if m ≤ m′ and ρ ≥ ρ′ ,
′ S
then Sρm ⊆ Sρm′ . For constant ρ, the vector space m Sρm can therefore be equipped
with the strict inductive limit topology of Fréchet spaces.

3.2 Examples. ([Gru09, § 7.1], [Jos99, § 3])

R R
Functions in C ∞ ( kθ × kξ ) that are positively homogeneous of degree t are in S1t since

 θ ξ 
|s(θ, ξ)| = k(θ, ξ)kt s , ≤ C (1 + k(θ, ξ)k)t (3.4)
k(θ, ξ)k k(θ, ξ)k
 9

for k(θ, ξ)k > 0. Functions in Cc∞ ( Rkθ × Rkξ ) are obviously in Tm S1m . ⋄

3.3 Lemma. [Hö71, Prop. 1.1.6]

′
If s ∈ Sρm and t ∈ Sρm are symbols of the same type ρ, then
′
(i) the pointwise product st belongs to Sρm+m , and
m−ρ(|α|+|β|)
(ii) the derivatives Dθβ Dξα s belong to Sρ .

S m
Both properties are easily proven. This product bestows upon m Sρ the structure
of a topological ∗-algebra, with Sρ0 being a unital ∗-subalgebra.

3.4 Lemma. [Hö71, § 1.2]

R
If m < −2k, the integral Rk ×Rk (1 + k(θ, ξ)k)m d(θ, ξ) is finite.

Proof. Let k−kE denote the Euclidean norm and let α > 0. In spherical coordinates,
Z Z ∞
m
(1 + αk(θ, ξ)kE ) d(θ, ξ) = Ω2k−1 (1) (1 + αr)m r 2k−1 dr, (3.5)
Rk ×Rk 0

where Ω2k−1 (1) is the surface measure of the unit (2k −1)-sphere. If t := −2k −m > 0,
Z ∞ Z ∞
−2k+1 2k−1 −2k+1
Ω2k−1 (1) α m
(1 + αr) (αr) dr ≤ Ω2k−1 (1) α (1 + αr)−t−1 dr
0 0
= Ω2k−1 (1) α−2k t−1 . (3.6)

By equivalence of all norms on a finite-dimensional vector space [Con07, Thm. III.3.1],

R R
for every norm k−k on kθ × kξ , there exists C > 0 such that C k(θ, ξ)kE ≤ k(θ, ξ)k.
Consequently, (1 + k(θ, ξ)k)m ≤ (1 + Ck(θ, ξ)kE )m since m < 0. 

I
It follows that the oscillatory integral η is absolutely convergent when the symbol
s has order m < −2k since
Z Z
k
I
(2π) | η (s)| ≤ lim |χ(ǫθ, ǫξ)| |s(θ, ξ)| d(θ, ξ) ≤ p0,0 (s) (1 + k(θ, ξ)k)m d(θ, ξ). (3.7)
ǫց0

I
From the preceding lemmata, η can be extended to a continuous linear functional on
S m
m Sρ by iterating a differential operator which lowers the order of the symbol. As
the next theorem will illustrate, the standard construction ([Hö71, Prop. 1.2.2], [Jos99,
§ 3]) derives a partially different result [LW11, Thm. 3.2] owing to the bilinear form η
being positively homogeneous of degree two.
10

3.5 Theorem.
R R R
Let η be a non-degenerate bilinear form on k . If χ ∈ Cc∞ ( kθ × kξ ) fulfills χ(0, 0) = 1,
then the oscillatory integral
Z
Iη (s) = (2π)
−k
lim
ǫց0
e−iη(θ,ξ) χ(ǫθ, ǫξ) s(θ, ξ) d(θ, ξ) (3.8)
Rk ×Rk
S m
is a continuous linear functional on m Sρ when −1 < ρ ≤ 1. That is, if s is a symbol
of order m and type ρ, then
X
I
| η (s)| ≤ Ch pα,β (s) (3.9)
|α|+|β|≤h

for h > (ρ + 1)−1 (m + 2k).

Proof. The idea is simple but clouded with technicalities. A differential operator will
be constructed such that it leaves the exponential term invariant. It can therefore be
introduced within the integral, after which partial integration lowers the order of the
symbol by Lemma 3.3. Formally, this is depicted as
Z Z Z
−iη(θ,ξ) −iη(θ,ξ)



e s(θ, ξ) d(θ, ξ) = Me s(θ, ξ) d(θ, ξ) = e−iη(θ,ξ) M ∗ s(θ, ξ) d(θ, ξ),
(3.10)

where M is a first order differential operator and M ∗ its formal adjoint under partial
integration. Iterating this step eventually makes the symbol integrable by Lemma 3.4.
However, the technicalities arise since M is not defined at the origin in kθ × kξ . One R R
workaround is to separate the integral over a ball centred at the origin and its com-
plement; the outlined method is only applied to the latter, while the contribution of
the former can be made arbitrarily small by reducing the size of the ball.
Let B(δ) denote the open ball of radius δ centred at the origin in kθ × kξ , and R R
I I I
B(δ) its closure. Split the integral as η (s) = η (1B(δ) s) + η (1B(δ)c s). As χ and s are
I
smooth, (1B(δ) s) converges absolutely:
Z
I I
| η (1B(δ) s)| = | η (1B(δ) s)| ≤ (2π)−k
p0,0 (s) (1 + k(θ, ξ)k)m d(θ, ξ)
B(δ)

≤ (2π)−k p0,0 (s) vol(B(δ)) sup (1 + k(θ, ξ)k)m . (3.11)

(θ,ξ)∈B(δ)

In B(δ)c , the function φ := kη(−, ξ)k2 + kη(θ, −)k2 is always non-zero because η is
R R
non-degenerate. If Ξ = φ−1 (η(−, ξ), η(θ, −)) ∈ kθ × kξ , then the differential operator
 11

M := i Ξ · grad(θ,ξ) leaves e−iη(θ,ξ) invariant:

M eiη(θ,ξ) = i φ−1 η(−, ξ) · gradθ (e−iη(θ,ξ) ) + i φ−1 η(θ, −) · gradξ (e−iη(θ,ξ) )

= φ−1 (kη(−, ξ)k2 + kη(θ, −)k2 ) e−iη(θ,ξ)
= e−iη(θ,ξ) . (3.12)

I
On replacing e−iη(θ,ξ) with M e−iη(θ,ξ) in η (1B(δ)c s), the Leibniz product rule yields a
divergence and a formal adjoint operator M ∗ as

(M e−iη(θ,ξ) ) χs = i Ξ · grad(θ,ξ) (e−iη(θ,ξ) ) χs

= i div(θ,ξ) (e−iη(θ,ξ) χs Ξ) − i e−iη(θ,ξ) Ξ · grad(θ,ξ) (χs) + div(θ,ξ) (Ξ) χs
=: i div(θ,ξ) (e−iη(θ,ξ) χs Ξ) − i e−iη(θ,ξ) M ∗ (χs). (3.13)

The integral of the first term over B(δ)c reduces, by the Gauß theorem and compact
support of χ, to an integral over the boundary ∂B(δ). Proper application of the Gauß
theorem pertains to the region enclosed by B(δ)c and any concentric ball B(δ′ ) which
contains supp χ. The integral over ∂B(δ′ ) vanishes and thus the limit δ′ → ∞ exists.
If n is the inward unit normal vector field and dσ the surface measure on ∂B(δ), then
Z
div(θ,ξ) (e−iη(θ,ξ) χ(ǫθ, ǫξ) s(θ, ξ) Ξ(θ, ξ)) d(θ, ξ)
B(δ)c
Z


= e−iη(θ,ξ) χ(ǫθ, ǫξ) s(θ, ξ) Ξ(θ, ξ) · n(θ, ξ) dσ(θ, ξ) (3.14)
∂B(δ)

and from the compactness of ∂B(δ), this integral is absolutely convergent:

Z
(2π)−k lim |χ(ǫθ, ǫξ)| |s(θ, ξ)| |Ξ(θ, ξ) · n(θ, ξ)| d(θ, ξ)
ǫց0 ∂B(δ)

≤ (2π)−k p0,0 (s) Ω2k−1 (δ) sup |Ξ(θ, ξ) · n(θ, ξ)| (1 + k(θ, ξ)k)m . (3.15)
(θ,ξ)∈∂B(δ)

It remains to analyse the integral of the second term in (3.13): e−iη(θ,ξ) M ∗ (χs). First,
T
the function χ(ǫθ, ǫξ) belongs to m S1m for every non-zero ǫ, but is considered as an
element of S10 since it is the smallest symbol class in which limǫց0 χ(ǫθ, ǫξ) converges.
Then, the pointwise product χ(ǫθ, ǫξ) s(θ, ξ) remains in Sρm . Second, from Example 3.2
and Lemma 3.3, the symbol classes of the coefficients of M ∗ = Ξ·grad(θ,ξ) +div(θ,ξ) (Ξ)
m−(ρ+1)
are computed: Ξj ∈ S1−1 and div(θ,ξ) (Ξ) ∈ S1−2 . Thus, M ∗ maps χs ∈ Sρm to Sρ .
12

After an h-fold iteration of the above procedure,

X
(M ∗ )h χ(ǫθ, ǫξ) s(θ, ξ) = zα,β (ǫ, θ, ξ) Dθβ Dξα s(θ, ξ) (3.16)
|α|+|β|≤h

m−h(ρ+1) m−ρ(|α|+|β|)
belongs to Sρ . By Lemma 3.3, Dθβ Dξα s is in Sρ and so zα,β must be
ρ(|α|+|β|)−h(ρ+1)
in Sρ With the respective symbol estimates and h > (ρ + 1)−1 (m + 2k),
.
Lemma 3.4 is applicable and ensures absolute convergence:
Z


(2π)−k lim |(M ∗ )h χ(ǫθ, ǫξ) s(θ, ξ) | d(θ, ξ)
ǫց0 B(δ)c
Z X
−k
≤ (2π) lim |zα,β (ǫ, θ, ξ)| |Dθβ Dξα s(θ, ξ)| d(θ, ξ)
ǫց0 B(δ)c
|α|+|β|≤h
X Z
−k
≤ (2π) lim Cα,β pα,β (s) (1 + k(θ, ξ)k)m−h(ρ+1) d(θ, ξ)
ǫց0 B(δ)c
|α|+|β|≤h
X
≤ Ch′ pα,β (s). (3.17)
|α|+|β|≤h

I I I
As a result, | η (s)| ≤ | η (1B(δ) s)| + | η (1B(δ)c s)| is bound from above by the sum
of (3.11), (3.15), and (3.17). If p0,0 (s) 6= 0, choose δ < 1 so that k(θ, ξ)k < 1 implies
(1 + k(θ, ξ)k)m < max(1, 2m ) in B(δ). Given ε > 0, reduce δ further until vol(B(δ))
and Ω2k−1 (δ) are both less than (2π)k ε/(2p0,0 (s) max(1, 2m )). It follows that
X
I
| η (s)| < ε + Ch pα,β (s) (3.18)
|α|+|β|≤h

since the sum of (3.11) and (3.15) is bound from above by ε. 

Of course, the inequality before (3.17) allows exchanging the limit and the integral.
The unused condition χ(0, 0) = 1 then ensures that the choice of χ does not affect η . I
However, this use of dominated convergence eliminates χ from the oscillatory integral
and explicitly requires M ∗ to make the symbol absolutely integrable. When the limit
I
is maintained outside the integral, η can be seen as the limit of convergent integrals.
As a consequence of the higher degree of homogeneity of η, this oscillatory integral
R
accommodates symbols of order m ∈ and type −1 < ρ ≤ 1. Rieffel considers [Rie93,
R R
Chap. 1] a space of functions kθ × kξ → A, where A is a Fréchet algebra, and then
restricts to the dense subspace that is smooth under an isometric action of k . These R
maps precisely corresponds to the symbol class S00 when A = . However, Rieffel uses C
 13

the Laplacian, a second order differential operator, in deriving the criterion h > k for
S00 , thereby proving that h > (ρ + 1)−1 (m + 2k) is not optimal.

3.6 Proposition. ([LW11, Prop. 3.9], [Rie93, Cor. 1.12])

If s ∈ Sρm is independent of θ or ξ, that is, either s(θ, ξ) = s(0, ξ) for all θ ∈ Rk or
R I
s(θ, ξ) = s(θ, 0) for all ξ ∈ k , then η (s) = s(0, 0).

Proof. Without loss of generality, assume that s(θ, ξ) = s(0, ξ) for all θ ∈ Rk . Let χ
be of the separable form χ1 ⊗ χ2 such that
Z
η I
(s) = (2π)−k
lim
ǫց0
e−iη(θ,ξ) χ1 (ǫθ) χ2 (ǫξ) s(0, ξ) d(θ, ξ)
Z
−k −k
= (2π) lim ǫ Fχ1 (ǫ−1 ξ) χ2 (ǫξ) s(0, ξ) dξ
ǫց0
Z
−k
= (2π) lim Fχ1 (ξ) χ2 (ǫ2 ξ) s(0, ǫξ) dξ. (3.19)
ǫց0

Then, dominated convergence and the inverse Fourier transform gives

Z
η I
(s) = χ 2 (0) s(0, 0) (2π)−k
Fχ1 (ξ) dξ = χ1 (0) χ2 (0) s(0, 0), (3.20)

and the result follows from the defining condition χ(0, 0) = χ1 (0) χ2 (0) = 1. 

R R R
Consider symbols on sx × ( kθ × kξ ). According to Hörmander’s definition [Hö71,
R R R
Def. 1.1.1], every symbol would belong in C ∞ ( sx × ( kθ × kξ )). Since the differential
operator M constructed on e−iη(θ,ξ) in the proof of Theorem 3.5 will not introduce any
derivatives in x, the assumption of smoothness in x is unnecessary. Moreover, regular
R
distributions on sx induced by these symbols will have empty wavefront sets. Hence,
it seems wise to reduce smoothness in x to local integrability. For this, let (Kj )j∈N be
R
an exhaustion of sx by compact sets.

3.7 Definition.
R R R R
Let m ∈ and −1 < ρ ≤ 1. The vector space L1loc ( sx ; Sρm ( kθ × kξ )) is designated an
m,ρ
extended symbol class Xloc , with a Fréchet topology determined by the semi-norms
Z
qα,β,Kj (u) := pα,β (u(x)) dx, (3.21)
Kj

m,ρ
where u ∈ Xloc is an extended symbol of order m and type ρ. ⋄
14

′ m,ρ m′ ,ρ′ S m,ρ

If m ≤ m′ and ρ ≥ ρ′ , then Sρm ⊆ Sρm′ implies Xloc ⊆ Xloc . Thus, m Xloc
can also be equipped with the strict inductive limit topology of Fréchet spaces when ρ
m,ρ
is fixed. That an extended symbol class Xloc is indeed an extension of a symbol class
Sρ follows trivially from the observation that the elements of Sρm are just functions in
m
m,ρ
Xloc which are independent of x. The algebraic structure of extended symbol classes
m,ρ m′ ,ρ
carries over from symbol classes: for u ∈ Xloc and v ∈ Xloc , the pointwise product
m+m′ ,ρ S m,ρ
uv(x) := u(x) v(x) defines an element in Xloc by Lemma 3.3, providing m Xloc
0,ρ
with the structure of a topological ∗-algebra and Xloc that of a unital ∗-subalgebra.

3.8 Lemma.
m,ρ
Each extended symbol class Xloc is an L1loc ( Rsx )-module.
′
0,ρ
Proof. The canonical embedding L1loc ( sx ) ֒→ Xloc R
as functions constant in (θ, ξ)
S
m,ρ m ,ρ m+m′ ,ρ m,ρ
and the pointwise product Xloc × Xloc → Xloc in m Xloc together define
1 s R m,ρ
a ring homomorphism Lloc ( x ) → End(Xloc ). 

3.9 Proposition.
m,ρ
Let u ∈ Xloc e
. The map Λ(u) : D( Rsx ) → Sρm (Rkθ × Rkξ ), given by
Z
e
≺ Λ(u), ψ ≻ := u(x) ψ(x) dx, Rsx ),
∀ ψ ∈ D( (3.22)

is an Sρm -valued regular distribution on Rsx.

Proof. Let ψ ∈ DKj ( Rsx). By Lemma 3.8, uψ belongs to Xlocm,ρ and so
Z Z
pα,β (u(x) ψ(x)) dx = pα,β (u(x)) |ψ(x)| dx ≤ qα,β,Kj (u) t0,Kj (ψ), (3.23)
Kj Kj

implying [Tho75, Thm. 3] that the pointwise product uψ is Pettis integrable; for each
R
Σ in the Borel σ-algebra of sx , there exists an element uψ
R
ρ
f =
f ∈ S m such that ϕ(uψ)
m
Σ ϕ u(x) ψ(x) dx for all continuous linear functionals ϕ on Sρ . Then, the relation
Z


e
pα,β (≺ Λ(u), ψ ≻) ≤ pα,β u(x) ψ(x) dx (3.24)

only holds if pα,β (u(x) ψ(x)) is measurable, which is true by the definition of extended
symbols. Following (3.24) with (3.23) hence proves, under Proposition 2.2, that Λ(u) e
is an Sρm -valued regular distribution. 
 15

3.10 Definition.
e
Every Sρm -valued regular distribution Λ(u) R
on sx induced by an extended symbol u ∈
m,ρ
Xloc is a symbolic distribution of order m and type ρ. ⋄

Extended symbols and symbolic distributions bestow upon the oscillatory integral,
in two ways, the structure of a distribution. First, the function η ◦ u : sx → I is R C
s I
locally integrable and determines a regular distribution Λ( η ◦ u) on x . Second, the R
I e
composition η ◦ Λ(u) R
: D( sx ) → C R
is a distribution on sx by Proposition 2.2 since

|≺ Iη ◦ Λ(u),
e ψ ≻| = |Iη (≺ Λ(u),
e ψ ≻)|
X
≤ Ch e
pα,β (≺ Λ(u), ψ ≻)
|α|+|β|≤h
X
≤ Ch t0,Kj (ψ) qα,β,Kj (u), Rsx),
∀ψ ∈ DKj ( (3.25)
|α|+|β|≤h

where the first inequality follows from continuity and the second from the inequalities
(3.24) and (3.23). As the next proposition will show, both distributions are equal and
the oscillatory integral, in a heuristic sense, behaves as an intertwiner of regular and
symbolic distributions.

3.11 Proposition.
m,ρ
If u ∈ Xloc I e
, then η ◦ Λ(u) = Λ( Iη ◦ u). That is, the following diagram commutes.
Rkθ × Rkξ )
Sρm (
e
Λ(u) Iη (3.26)

D( Rsx ) Λ(Iη ◦ u)
C

Proof. The functional δ(θ,ξ) (s) := (1 + k(θ, ξ)k)−m s(θ, ξ) on Sρm evaluates s at (θ, ξ)
analogous to the δ-distribution and is clearly continuous since |δ(θ,ξ) (s)| ≤ p0,0 (s). By
the Pettis integrability shown in the proof of Proposition 3.9,

e
≺ Λ(u), e
ψ ≻(θ, ξ) = (1 + k(θ, ξ)k)m δ(θ,ξ) (≺ Λ(u), ψ ≻)
Z Z
= (1 + k(θ, ξ)k)m δ(θ,ξ) (u(x)) ψ(x) dx = u(x)(θ, ξ) ψ(x) dx,
(3.27)
16

for all ψ ∈ D( Rsx ). Applying this relation along with Fubini’s theorem gives
Z
≺ Iη e
◦ Λ(u), ψ ≻ = (2π)−k lim
ǫց0
e
e−iη(θ,ξ) χ(ǫθ, ǫξ) ≺ Λ(u), ψ ≻(θ, ξ) d(θ, ξ)
Z Z 
= (2π)−k lim e−iη(θ,ξ) χ(ǫθ, ǫξ) u(x)(θ, ξ) ψ(x) dx d(θ, ξ)
ǫց0
Z  Z 
= (2π)−k lim ψ(x) e−iη(θ,ξ) χ(ǫθ, ǫξ) u(x)(θ, ξ) d(θ, ξ) dx.
ǫց0
(3.28)

In deriving (3.17), the limit ǫ ց 0 played no part and so the estimate

Z X
−k
(2π) |ψ(x) e−iη(θ,ξ) χ(ǫθ, ǫξ) u(x)(θ, ξ) d(θ, ξ)| ≤ |ψ(x)| Ch pα,β (u(x))
|α|+|β|≤h
(3.29)

still holds. Since the upper bound is integrable, dominated convergence results in
Z  Z 
I e −k
≺ η ◦ Λ(u), ψ ≻ = (2π) lim
ǫց0
ψ(x) e−iη(θ,ξ) χ(ǫθ, ǫξ) u(x)(θ, ξ) d(θ, ξ) dx
Z  Z 
= −k
(2π) lim e−iη(θ,ξ) χ(ǫθ, ǫξ) u(x)(θ, ξ) d(θ, ξ) ψ(x) dx
ǫց0

= ≺ Λ( Iη ◦ u), ψ ≻ (3.30)

R
for all ψ ∈ D( sx ). Due to the separating nature of the semi-norms on D ′ ( Rsx), the
desired equivalence is obtained. 

R R R
The extension of symbol classes to sx × ( kθ × kξ ) required local integrability in
x and symbolic estimates in (θ, ξ). In lieu of the introduced extended symbol classes
R R R
L1loc ( sx ; Sρm ( kθ × kξ )), these conditions can also be realised with the vector-valued
R R R
symbols Sρm ( kθ × kξ ) → L1loc ( sx ). Lechner and Waldmann [LW11] analysed symbols
that are more generally valued in sequentially complete locally convex spaces, and fed
to an oscillatory integral which is then valued in the same space. Here, the extended
m,ρ
symbol classes Xloc are preferred owing to their interesting algebraic structure and
I
since the oscillatory integral η does not itself require any modification.
Finally, the distributional structures introduced with oscillatory integrals permit a
little discussion about wavefront sets. To be precise, oscillatory integrals (or any other
continuous linear functional on Sρm , since the following proof only requires continuity)
never enlarge the wavefront set of symbolic distributions.
 17

3.12 Theorem.
m,ρ
If u ∈ Xloc , then WF(Λ( Iη ◦ u)) = WF(Iη ◦ Λ(u))
e e
⊆ WF(Λ(u)).

Proof. The equality follows immediately from Proposition 3.11. For the inclusion, let
e
(x, ζ ′ ) be a regular direction for Λ(u). R
Then there exist ψ ∈ D( sx ) with ψ(x) 6= 0 and
an open conic neighbourhood Γ of ζ ′ in which

e
sup (1 + kζk)N pα,β (F[ψ Λ(u)](ζ)) < ∞, ∀N ∈ N. (3.31)
ζ∈Γ

By the continuity of the oscillatory integral (Theorem 3.5), there exists h such that

|F[ψ( Iη ◦ Λ(u))](ζ)|
e = |Iη (≺ Λ(u),
e eiη(−,ζ) ψ ≻)|
X
≤ Ch e
pα,β (≺ Λ(u), eiη(−,ζ) ψ ≻), (3.32)
|α|+|β|≤h

e
and the left hand side rapidly decreases in Γ since ≺ Λ(u), e
eiη(−,ζ) ψ ≻ = F[ψ Λ(u)](ζ).
′ I e
Thus, (x, ζ ) is a regular direction for η ◦ Λ(u). 

The implications of these results will become apparent in their application to the
warped convolution of quantum field operators, which are (weak) oscillatory integrals
of Rieffel-type. Explicitly, the extended symbol and symbolic distribution are given in
Theorem 5.2, while the wavefront set inclusion offered by the action of the oscillatory
integral on the symbolic distribution (in the case of a product of warped convolutions)
will initiate the proof of Theorem 6.4.

4 The Scalar Quantum Field

R
Consider d to be Minkowski spacetime with metric η = diag(+1, −1, . . . , −1). Let P
R
denote the Poincaré group d ⋊ O(1, d − 1) of symmetries for Minkowski spacetime,
and P↑+ the component connected to its identity element. Then the Borchers-Uhlmann
algebra [Bor62, Uhl62] is defined as
∞ O
M n
B(Rd) := D( Rd ), (4.1)
n=0

where
N0
R C R
D( d ) := . An element f ∈ B( d ) is a sequence (f (n) )n∈N0 with finitely
R
many non-zero components, and the operations on B( d ) can easily be extrapolated
from linear (or anti-linear, for the involution) and component-wise extension of
18

P
(i) the associative product (f × g)(n) = nj=0 f (j) ⊗ g (n−j) ,
(ii) the involution (f1 ⊗ · · · ⊗ fn )∗ = fn ⊗ · · · ⊗ f1 by complex conjugation, and
(iii) for every (a, Λ) ∈ P, the ∗-automorphism

α(a,Λ) (f1 ⊗ · · · ⊗ fn ) = (α(a,Λ) f1 ⊗ · · · ⊗ α(a,Λ) fn ), (4.2)

where (α(a,Λ) fj )(x) := fj (Λ−1 (x − a)).

The product fulfills the involutive property (f × g)∗ = g∗ × f ∗ and admits an identity
R
element 1B := (1, 0, . . .). Hence, B( d ) is a unital ∗-algebra [Sch20, § 2.1]. Barring
R
the action of the Poincaré group, the structure of B( d ) is that of a tensor algebra
R N
R
[Sch20, § 6.1] over D( d ). On providing n D( d ) with the projective topology, so as
to ensure that the multilinear maps ⊗n : (f1 , . . . , fn ) 7→ f1 ⊗ · · · ⊗ fn are continuous,
R
B( d ) obtains the structure of a topological tensor algebra [Sch20, § 6.6] over D( d ) R
under the direct sum topology.
Let ωn be a continuous linear functional on
Nn
R
D( d ). By the universal property
R C
of tensor products, there exists a multilinear map ω n : D( d )×n → such that ω n =
ωn ◦ ⊗n . The projective topology on
Nn
R
D( d ) makes ωn continuous. When n = 2,
R
the bilinear map ω 2 is clearly equivalent to a linear map D( d ) → D ′ ( d ), and theR
Schwartz kernel theorem (Theorem 2.4) extends ω 2 to a distribution on d × d . For R R
n ≥ 2, a similar iterative argument shows that ωn is a distribution on ( d )×n . Thus,R
the functionals ωn are known as n-point distributions and together determine a con-
R
tinuous linear functional ω on B( d ) as
∞
X
ω(f ) := ≺ ωn , f (n) ≻. (4.3)
n=0

R
This is a state on the ∗-algebra B( d ) if two constraints are imposed: first, ω0 (1) = 1
ensures normalisation ω(1B ) = 1, and second, ω(f ∗×f ) ≥ 0 ensures positivity. Assume
further that ω is Poincaré-invariant: α∗(a,Λ) ω = ω for all (a, Λ) ∈ P. For a state on a
unital ∗-algebra, the following Gelfand-Naimark-Segal (GNS) construction applies.

4.1 Theorem. ([Mor19, Thm. 8.9], [Sch20, Thm. 4.38])

R
The pair (B( d ), ω) determines a quadruple (Hω , Dω , πω , Ψω ), where
(i) Hω is a Hilbert space,
(ii) Dω is a dense subspace of Hω ,
R
(iii) πω is a ∗-representation of B( d ) on Hω as (possibly unbounded) linear operators
with a common invariant domain Dω , and
 19

R
(iv) Ψω is an algebraically cyclic unit vector: πω (B( d )) Ψω = Dω .
Furthermore, the two tuples are related by the equation

ω(f ) = hΨω , πω (f ) Ψω i, ∀ f ∈ B( Rd ). (4.4)

The quadruple (Hω , Dω , πω , Ψω ) is unique up to unitary equivalence.

Let L+ (Dω ) denote the ∗-algebra of operators A with domain Dω , each of which
admits an adjoint A∗ whose domain contains Dω , such that the subspace Dω remains
invariant under A and A+ := A∗ ↾Dω [Sch20, Def. 3.1, Lem. 3.2]. Moreover, L+ (Dω ) is
locally convex since every Ψ ∈ Dω determines a semi-norm A 7→ kA Ψk.

4.2 Definition.
The scalar quantum field Φ is the operator-valued distribution2

Φ : D( Rd ) → L+(Dω )


f 7→ Φ(f ) := πω (0, f, 0, . . .) , (4.5)

and the operator Φ(f ), for any f ∈ D( Rd ), is referred to as a field operator. ⋄

From the definition, it is clear that the scalar quantum field is linear: Φ(λf + g) =
C
λΦ(f )+Φ(g) for all λ ∈ , and that the field operators are Hermitian: Φ(f )+ = Φ(f ).
R
Hence, the identity idHω and the field operators Φ(f ), for all f ∈ D( d ), generate the
R
∗-algebra P( d ) ∼ R
= πω (B( d )). A simple computation using (4.3), (4.4), the product
R
on B( d ), as well as the fact that πω is a ∗-representation, yields the relation

≺ ωn , f1 ⊗ · · · ⊗ fn ≻ = hΨω , Φ(f1 ) . . . Φ(fn ) Ψω i. (4.6)

Since the products of field operators evaluated within the GNS vector Ψω correspond
precisely to the n-point distributions, the state ω can be viewed as a vector state on
R
P( d ). Every Ψ ∈ Dω of unit norm defines an admissible vector state ω Ψ on P( d ) R
as ω Ψ (−) := hΨ, (−) Ψi, and the n-point distributions for ω Ψ take the form

≺ ωnΨ , f1 ⊗ · · · ⊗ fn ≻ := hΨ, Φ(f1 ) . . . Φ(fn ) Ψi. (4.7)

Finite convex combinations of vector states also define states on P( d ). R

Due to the Poincaré-invariance of ω, the symmetries are implementable on Hω as
2
For the quantum field, the conventional notation Φ(f ) is preferred to the dual pairing ≺ Φ, f ≻.
20

a strongly continuous unitary representation U of P↑+ which fulfills, for all (a, Λ) ∈ P↑+ ,

U (a, Λ) Ψω = Ψω , U (a, Λ) Dω ∼
= Dω ,
(4.8)
U (a, Λ) πω (f ) U (a, Λ)−1 = πω (α(a,Λ) f ).

This establishes the covariance of the field: U (a, Λ) Φ(f ) U (a, Λ)−1 = Φ(α(a,Λ) f ). For
R
the translation subgroup d of P↑+ , denote the adjoint action of U (a) := U (a, idRd ) by
AdU(a) [−] := U (a) (−) U (−a), noting that U (a)
−1 = U (−a), and let f (x) := f (x − a).
(a)
With these abbreviations, translational covariance can be stated as

U (a) Ψω = Ψω , U (a) Dω ∼
= Dω , AdU
(a) [Φ(f )] = Φ(f(a) ). (4.9)

By generalising Stone’s theorem [Sch12, Thm. 6.2] to d-parameter unitary groups, the
unitary representation of the translation subgroup takes the form U (a) = eiη(a,P ) for
a d-tuple P of commuting self-adjoint operators. Usually, it is assumed that the joint
spectrum of P is supported in the future light cone, but this spectrum condition will
be replaced in Section 6 by a microlocal counterpart. Without the spectrum condition,
the state ω need not be the unique Minkowski vacuum state.
As is the case for free field theories, Φ is considered to be a regular distribution in
the weak sense. That is, there are pointlike-localised “operators” Φ(x) such that3
Z
hΨ1 , Φ(f ) Ψ2 i = hΨ1 , Φ(x) Ψ2 i f (x) dx, ∀ Ψ1 , Ψ2 ∈ Dω . (4.10)

Translational covariance for Φ(x) can be written as AdU (a) [Φ(x)] = Φ(x + a). However,
Φ(x) is itself not a well-defined operator [Wig96] and should instead be interpreted as
R
a sesquilinear form on Dω . Integrating it against functions in D( d ) localises the field
in a manner which allows the expression of local commutativity: [Φ(f ), Φ(g)] = 0 when
the supports of f and g are spacelike separated. For our purpose, local commutativity
of the field is not required.
Altogether, the assumptions on the field fall short of the Wightman axioms [SW01,
§ 3.1] due to the exclusion of local commutativity and the spectrum condition.

5 Warped Algebras and States

R
Let Q ∈ End( d ) define a skew-symmetric bilinear form with the metric: η(θ, Qξ) =
−η(ξ, Qθ), and consider the dense subspace D∞
ω := {Ψ ∈ Dω : a 7→ U (a) Ψ is smooth}.
3
For the regular form of the quantum field, the map Λ : L1loc ( R ) → D (R ) will not be used.
d ′ d
 21

5.1 Definition. ([BS08], [BLS11])

The warped convolution wQ [A] of an operator A ∈ L+ (Dω ) is the sesquilinear form
Z
wQ [A](Ψ1 , Ψ2 ) := (2π)−d e−iη(θ,ξ) hΨ1 , AdU
(Qθ) [A] U (ξ) Ψ2 i d(θ, ξ), (5.1)
Rd ×Rd

where Ψ1 and Ψ2 belong to the domain D∞

ω of wQ [A]. ⋄

By Proposition 3.6, the warped convolution of idHω reduces to the inner product.
For a field operator, the warped convolution is well-defined as an oscillatory integral
due to the constructions in Section 3; this is depicted in the next theorem. Note that
R R
the function χ ∈ Cc∞ ( dθ × dξ ) will be notationally suppressed until required.

5.2 Theorem.
If Ψ1 , Ψ2 ∈ D∞
ω , then

uΨ
Q
1 ,Ψ2
(x)(θ, ξ) := hΨ1 , AdU
(Qθ) [Φ(x)] U (ξ) Ψ2 i (5.2)

is an extended symbol of order 0 and type 0, which induces the symbolic distribution

e Ψ1 ,Ψ2 ), f ≻(θ, ξ) = hΨ1 , AdU [Φ(f )] U (ξ) Ψ2 i,

≺ Λ(uQ (Qθ) ∀ f ∈ D( Rd ). (5.3)

Proof. The derivative of U (a) = eiη(a,P ) is generalised from one-parameter unitary

groups [Sch12, Prop. 6.1] as Daβ U (a) = U (a) η(−, P )β in the strong operator topology.
The Leibniz product rule and the skew-symmetry of Q in the metric η then give

|Dθβ Dξα uΨ
Q
1 ,Ψ2
(x)(θ, ξ)|
= |Dθβ Dξα hU (−Qθ) Ψ1 , Φ(x) U (−Qθ) U (ξ) Ψ2 i|
X
≤ Cγ,β |hU (−Qθ) (QP )β−γ Ψ1 , Φ(x) U (−Qθ) (QP )γ U (ξ) P α Ψ2 i|. (5.4)
γ≤β

Let ǫ > 0. By the characterisation of suprema, there exists (θs , ξs ) such that
X
sup Cγ,β |hU (−Qθ) (QP )β−γ Ψ1 , Φ(x) U (−Qθ) (QP )γ U (ξ) P α Ψ2 i|
(θ,ξ)∈Rd ×Rd γ≤β
X
<ǫ+ Cγ,β |hU (−Qθs ) (QP )β−γ Ψ1 , Φ(x) U (−Qθs ) (QP )γ U (ξs ) P α Ψ2 i|. (5.5)
γ≤β

Since Ψ1 and Ψ2 belong in D∞ e β,γ,Q := U (−Qθs ) (QP )β−γ Ψ1 and

ω , so do the vectors Ψ1
22

e α,γ,Q := U (−Qθs ) (QP )γ U (ξs ) P α Ψ2 . As a result, the semi-norms of X 0,0 satisfy

Ψ 2 loc
Z


qα,β,K (uΨ
Q
1 ,Ψ2
)= pα,β uQΨ1 ,Ψ2
(x) dx
ZK Z X
< ǫ dx + e β,γ,Q , Φ(x) Ψ
Cγ,β |hΨ e α,γ,Q i| dx, (5.6)
1 2
K K γ≤β

which is finite due to the regular form of the scalar quantum field Φ.
Ψ1 ,Ψ2 0,0 e Ψ1 ,Ψ2 ) via
The extended symbol uQ ∈ Xloc induces a symbolic distribution Λ(u Q
Proposition 3.9, which takes the form
Z
Ψ1 ,Ψ2
e
≺ Λ(uQ ), f ≻(θ, ξ) = uΨQ
1 ,Ψ2
(x)(θ, ξ) f (x) dx
ZR
d

= hΨ1 , AdU (Qθ) [Φ(x)] U (ξ) Ψ2 i f (x) dx

Rd

= hΨ1 , AdU
(Qθ) [Φ(f )] U (ξ) Ψ2 i, (5.7)

for all f ∈ D( Rd ). 

Associated with every sesquilinear form wQ [Φ(f )] is an operator WQ [Φ(f )], which
will also be referred as the warped convolution of a field operator. Its domain consists
of vectors Ψ2 ∈ D∞ ω , each of which admit the existence of a vector Ψ3 ∈ Hω such that
wQ [Φ(f )](Ψ1 , Ψ2 ) = hΨ1 , Ψ3 i for all Ψ1 ∈ D∞ ∞
ω . Since Dω is dense in Hω , the vector
Ψ3 is uniquely determined. The operator is then defined by setting WQ [Φ(f )] Ψ2 = Ψ3
[Sch12, Def. 10.4], that is,

wQ [Φ(f )](Ψ1 , Ψ2 ) = hΨ1 , WQ [Φ(f )] Ψ2 i, ∀ Ψ1 ∈ D∞

ω . (5.8)

In fact, the following proposition shows that the covariance of the field Φ extends the
domain of WQ [Φ(f )] to D∞ ω (while the form domain must be strictly larger than the
corresponding operator domain, there is no inconsistency as D∞ ω is not the maximal
form domain) and permits Ψ1 to be any vector in Hω . This is equivalently written as
the following integrability condition.

5.3 Theorem.
If Ψ ∈ D∞
ω , then WQ [Φ(f )] Ψ is weakly integrable.

Proof. Subject to a first order approximation of U (a) = eiη(a,P ) , the covariance of

the field (4.9) reduces to the known commutator [η(−, P )ej , Φ(f )] Ψ = −Φ(D ej f ) Ψ,
 23

N
where {ej } is the standard basis of d0 . An arbitrary product of the generators can be
shifted across the field operators by iterating this expression:
X
η(−, P )β Φ(f ) Ψ = Cδ,β Φ(D δ f ) η(−, P )β−δ Ψ. (5.9)
δ≤β

Now, the Riesz representation theorem [Con07, Thm. I.3.4] identifies continuous linear
functionals on Hω with vectors in Hω under the inner product, which is why it must
be shown first that the oscillatory integral
Z
−d
(2π) lim e−iη(θ,ξ) χ(ǫθ, ǫξ) hΨ′ , AdU
(Qθ) [Φ(f )] U (ξ) Ψi d(θ, ξ) (5.10)
ǫց0 Rd ×Rd

is finite for all Ψ′ ∈ Hω . Applying the Leibniz product rule, the derivative Daβ U (a) =
U (a) η(−, P )β , and the iterated expression (5.9) altogether gives

|Dθβ Dξα hΨ′ , U (Qθ) Φ(f ) U (−Qθ) U (ξ) Ψi|

X
≤ Cγ,β |hΨ′ , U (Qθ) (QP )γ Φ(f ) U (−Qθ) (QP )β−γ U (ξ) P α Ψi|
γ≤β
X
′
≤ Cδ,γ,β |hU (−ξ) Ψ′ , AdU
(Qθ−ξ) [Φ(D
TQ δ
f )] (QP )γ−δ (QP )β−γ P α Ψi|
δ≤γ≤β
X
′′
≤ Cδ,β |hU (−ξ) Ψ′ , Φ(D TQ δ f(Qθ−ξ)) (QP )β−δ P α Ψi|, (5.11)
δ≤β

N
where TQ ∈ End( d0 ) satisfies (TQ )µν = 1 if Qµν 6= 0 and (TQ )µν = 0 if Qµν = 0, in
N R
the standard bases of d0 and d . Using the Cauchy-Schwarz inequality yields

|hU (−ξ) Ψ′ , Φ(D TQ δ f(Qθ−ξ) ) (QP )β−δ P α Ψi|

≤ kU (−ξ) Ψ′ k kΦ(D TQ δ f(Qθ−ξ)) (QP )β−δ P α Ψk
≤ kΨ′ k Cj′′′ tTQ δ,Kj (f(Qθ−ξ) ) k(QP )β−δ P α Ψk, (5.12)

by the definition of Φ as an operator-valued distribution. Since Ψ belongs to D∞ ω , so

does (QP ) β−δ α
P Ψ. Moreover, for each (θ, ξ), the compact set Kj must contain the
support of f(Qθ−ξ), and can be chosen from a compact exhaustion of d such that it R
also includes the support of f . Then a translation of f within Kj does not affect the
suprema of its derivatives, allowing the replacement of tTQ δ,Kj (f(Qθ−ξ) ) in the above
estimate with tTQ δ,Kj (f ). On taking the supremum over (θ, ξ), the semi-norms of S00
are thus seen to be finite, and by Theorem 3.5, so is the oscillatory integral.
24

For every ǫ > 0, the compact support of χ ensures that the integral
Z
−d
(2π) e−iη(θ,ξ) χ(ǫθ, ǫξ) hΨ′ , AdU
(Qθ) [Φ(f )] U (ξ) Ψi d(θ, ξ) (5.13)
Rd ×Rd

converges absolutely for all Ψ′ ∈ Hω , and so there exists Ψ′′ǫ ∈ Hω′′ ∼

= Hω such that
′ ′′
this integral equals hΨ , Ψǫ i [DU77, Lem. II.3.1]. In the limit ǫ ց 0, the convergence
of the oscillatory integral ensures the weak convergence of Ψ′′ǫ , and its weak limit is
precisely the weak integral of WQ [Φ(f )] Ψ. 

Since Ψ′ is arbitrary in Hω , the weak integral WQ [Φ(f )] Ψ is uniquely determined.

The warped convolution of a field operator, abbreviated as a warped field operator, is
an operator which acts on Ψ ∈ D∞ ω as
Z
−d
WQ [Φ(f )] Ψ = (2π) e−iη(θ,ξ) AdU
(Qθ) [Φ(f )] U (ξ) Ψ d(θ, ξ), (5.14)
Rd ×Rd

in the sense of weak oscillatory integrals. By Proposition 3.11, a warped field operator
can be determined from the warped convolution of its pointlike localisation Φ(x):

hΨ′ , WQ [Φ(f )] Ψi = ≺Iη ◦ Λ(uQ

′
e Ψ ,Ψ ), f ≻
Z
= ≺ Λ(Iη ◦ uQ ), f ≻ = hΨ′ , WQ [Φ(x)] Ψi f (x) dx,
Ψ ,Ψ ′
(5.15)

for all Ψ′ ∈ Hω and Ψ ∈ D∞ ω . As the first step towards the construction of a ∗-algebra
of warped field operators, it must be shown that the product of warped field operators
is well-defined on the domain D∞ ω . This is indirectly verified as follows.

5.4 Proposition.
Every warped field operator WQ [Φ(f )], with f ∈ D( Rd ), leaves D∞ω invariant.
Proof. Let Ψ ∈ D∞ ∞
ω . The vector WQ [Φ(f )] Ψ will belong to Dω if it belongs to the
domain of the derivative Daβ U (a) = U (a) η(−, P )β . Due to the covariance of the field
(4.9), as well as the iterated expression (5.9), it holds for all Ψ′ ∈ Hω that

hΨ′ , U (a) η(−, P )β WQ [Φ(f )] Ψi

Z
−d
= (2π) e−iη(θ,ξ) hΨ′ , U (a) η(−, P )β AdU(Qθ) [Φ(f )] U (ξ) U (a) Ψi d(θ, ξ)
Z X
= (2π)−d Cδ,β e−iη(θ,ξ) hΨ′ , AdU δ
(Qθ) [Φ(D f(a) )] U (ξ) U (a) η(−, P )
β−δ
Ψi d(θ, ξ)
δ≤β
 25

X
= Cδ,β hΨ′ , WQ [Φ(D δ f(a) )] U (a) η(−, P )β−δ Ψi. (5.16)
δ≤β

This is a finite sum of warped field operators acting on vectors U (a) η(−, P )β−δ Ψ in
D∞ ′
ω , which is well-defined. Since this relation holds for all Ψ ∈ Hω , it represents an
equality of weak integrals. Hence, the map a 7→ U (a) WQ [Φ(f )] Ψ is smooth. 

Also relevant to the construction of the ∗-algebra of warped field operators are the
following standard properties of warped convolutions.

5.5 Proposition. [BLS11, Lem. 2.2, Prop. 2.9, Prop. 2.11]

Let A ∈ L+ (Dω ) and Ψ ∈ D∞ ω . The warped convolution
(i) can be restricted, up to a constant factor, to an integral over ker Q⊥ × ker Q⊥ ,
(ii) can equivalently be written as
Z
−d
WQ [A] Ψ = (2π) e−iη(θ,ξ) U (ξ) AdU
(Qθ) [A] Ψ d(θ, ξ), (5.17)

(iii) preserves adjoints: WQ [A+ ] ⊆ WQ [A]∗ ,

(iv) is covariant in the sense that,
 
U (a, Λ) WQ [A] U (a, Λ)−1 = WΛQΛT U (a, Λ) A U (a, Λ)−1 , (5.18)

where ΛT denotes the transpose of Λ under η. Here, ΛQΛT is denoted as a matrix

R
product in the standard basis of d rather than a composition of endomorphisms.

If the warped convolution of an operator acts on a translation-invariant vector such

as Ψω , then it is a redundant deformation of that operator under Proposition 3.6:

WQ [A] Ψω = A Ψω , ∀ A ∈ L+ (Dω ). (5.19)

Another case of redundancy is when Q = 0, as is evident from either Proposition 3.6

or Proposition 5.5(i). Therefore, Q is also known as a deformation matrix, and to be
consistent, Q will henceforth refer to its matrix representation in the standard basis.
However, the covariance of warped field operators, as depicted in Proposition 5.5(iv),
exposes the fallacy of considering just one deformation matrix Q for the warped field
operators. Rather, a reference deformation matrix Q must be chosen, and the warped
convolutions associated with its orbit ΣQ := {ΛQΛT : Λ ∈ L↑+ } should be incorporated
into the constructed ∗-algebra. Note that the proper orthochronous Lorentz group L↑+
26

is the identity component of the Poincaré subgroup O(1, d − 1).

5.6 Definition.
Let Q be a deformation matrix. In dimensions d ≥ 3, a warped algebra P( Rd , ΣQ ) is
the polynomial algebra generated by idHω and the warped field operators

WQ′ [Φ(f )] : f ∈ D( Rd ), Q′ ∈ ΣQ , (5.20)

where ΣQ = {ΛQΛT : Λ ∈ L↑+ } is the orbit of Q under the action of L↑+ . ⋄

After restricting the adjoints of the warped field operators to the domain D∞ ω , the
d R
warped algebra P( , ΣQ ) can be identified as a topological ∗-subalgebra of L (D∞+
ω )
due to Proposition 5.5(iii) and the Hermiticity of the field. The natural next step is to
consider states on P( d , ΣQ ).R
5.7 Definition.
Every Ψ ∈ D∞ Ψ
ω of unit norm defines an admissible vector state ω on P( Rd , ΣQ) as
ω Ψ (−) := hΨ, (−) Ψi, and a map ωnΨ : (ΣQ )×n → D ′ ( nd ) satisfying R
≺ ωnΨ (Q1 , . . . , Qn ), f1 ⊗ · · · ⊗ fn ≻ = hΨ, WQ1 [Φ(f1 )] . . . WQn [Φ(fn )] Ψi, (5.21)

for all f1 , . . . , fn ∈ D( Rd ), is a warped n-point distribution for ωΨ. ⋄

Technically, referring to ωnΨ as a distribution is a misnomer, because it is only the

image of ωnΨ which is a distribution. Explicitly, (the image of) ωnΨ takes the form

ωnΨ (Q1 , . . . , Qn ) = IL n
η
e Ψ (Q1 , . . . , Qn )),
◦ Λ(un (5.22)

where the symbolic distribution subjected to the oscillatory integral is given by

e Ψ (Q1 , . . . , Qn )), f1 ⊗ · · · ⊗ fn ≻(θ1 , . . . , θn , ξ1 , . . . , ξn )

≺ Λ(un

:= hΨ, AdU U
(Q1 θ1 ) [Φ(f1 )] U (ξ1 ) . . . Ad(Qn θn ) [Φ(fn )] U (ξn ) Ψi
Xn
= hΨ, Φ(f1(Q1 θ1 ) ) . . . Φ(fn(Qn θn +Pn−1 ξk ) ) U ( ξi ) Ψi, (5.23)
k=1
i=1

due to the covariance of the field (4.9). Admissible vector states that are pulled-back
along the diagonal map ΣQ → (ΣQ )×n can be reformulated as states on an algebra
deformed by the Rieffel product [BLS11, Lem. 2.4].
 27

6 Microlocal Spectrum Condition

Henceforth, we set d = 4 as the Minkowski spacetime dimension. Since an admissible
vector state is entirely determined by its n-point distributions, wavefront sets offer a
characterisation of the singular correlations of field operators within that state. After
Radzikowski’s [Rad96] initial contribution, Brunetti et al. [BFK96] provided a broader
description using elements of graph theory.
Let Gn denote the set of undirected graphs having n vertices {vi } and finitely many
edges {er }. An immersion (x, γ, k) of a graph G ∈ Gn into 4 assigns R
R
(i) a point x(vi ) in 4 to every vertex vi in G,
R
(ii) a piecewise smooth curve γr in 4 which connects the points x(vi ) and x(vj ), to
every edge er in G which connects the vertices vi and vj , and
(iii) a constant future-directed covector field kr on each γr .
R
The immersion instantiates a point (x1 , ζ1 ; . . . ; xn , ζn ) ∈ (T∗ 4 )×n \ Z ×n if
X X
(i) xi = x(vi ), (ii) ζi = kr (xi ) − kr (xi ), (6.1)
er between er between
vi and vj>i vi and vj<i

for i = 1, . . . , n.

6.1 Definition. ([BFK96], [San10])

Let ω be a state. If the wavefront sets of its n-point distributions ωn satisfy

WF(ωn ) ⊆ Γn := (x1 , ζ1 ; . . . ; xn , ζn ) ∈ (T∗ R4 )×n \ Z ×n :
∃ G ∈ Gn and an immersion (x, γ, k) of G into R4 which
instantiates the point (x1 , ζ1 ; . . . ; xn , ζn ) , (6.2)

then the state ω fulfills the microlocal spectrum condition (µSC). ⋄


It is not difficult to observe from the definition that Γ2 = (x, ζ; y, −ζ) ∈ J+× J− ,
R
where J± ⊂ T∗ 4 denote the bundles of future(+)/past(−)-directed causal covectors.
Sometimes, the piecewise smooth curves γr of graph immersions are restricted to be
light-like (or causal), resulting in the µSC with light-like (or causal) immersions. Due
to the non-locality of the warped convolution, admissible vector states on P( 4 , ΣQ ) R
are not expected to fulfill these stronger versions of the µSC. However, there exists a
sufficient condition under which these states fulfill the µSC; this answers the question
posed in the introduction.
28

6.2 Lemma.
N
Let n ∈ . The wavefront set of the symbolic distribution (5.23) satisfies
[
e Ψ
π2 (WF(Λ(un (Q1 , . . . , Qn )))) ⊆

(ζ1 , . . . , ζn ) ∈ R4n \ {0} :
s,t∈{0,...,N }

(0, . . . , 0, ζ1 , . . . , ζn , 0, . . . , 0) ∈ π2 (WF(ωs+n+t )) , (6.3)

for some N ∈ N0 and for arbitrary Q1, . . . , Qn ∈ ΣQ.

Proof. Let Q1 , . . . , Qn ∈ ΣQ . If θ = (θ1 , . . . , θn ) ∈ R4n , ξ = (ξ1 , . . . , ξn ) ∈ R4n , and
ζ = (ζ1 , . . . , ζn ) ∈ R4n , then the microlocalisation of Λ(u
e Ψ (Q1 , . . . , Qn )) is
n

 

e Ψ (Q1 , . . . , Qn )) (ζ1 , . . . , ζn )
pα,β F (f1 ⊗ · · · ⊗ fn ) Λ(un


e Ψ
= pα,β ≺ Λ(un (Q1 , . . . , Qn )), e
iη(−,ζ1 )
f1 ⊗ · · · ⊗ eiη(−,ζn ) fn ≻
= sup Dθβ Dξα hΨ, Φ((eiη(−,ζ1 ) f1 )(Q1 θ1 ) ) . . . Φ((eiη(−,ζn ) fn )(Qn θn +Pn−1 ξi ) ) U (ξ+ ) Ψi ,
i=1
(θ,ξ)
(6.4)
P
where ξ+ := ni=1 ξi . By the linearity of the field, the translations of the exponential
functions can be extracted from the inner product; for example, the j-th element of
this n-fold product of field operators becomes
Pj−1
Φ((eiη(−,ζj ) fj )(Qj θj +Pj−1 ξi ) ) = eiη(θj ,Qj ζj ) e−i i=1 η(ξi ,ζj )
Φ(eiη(−,ζj ) fj(Qj θj +Pj−1 ξi ) ).
i=1 i=1

(6.5)

Noting that α = (α1 , . . . , αn ) ∈ 4n N

0 and β = (β1 , . . . , βn ) ∈
4n , the derivatives in
0 N
the semi-norms pα,β act via the Leibniz product rule
(i) on the exponential prefactors to introduce monomials in (ζ1 , . . . , ζn ):

β −δj
Dθjj eiη(θj ,Qj ζj ) = η(−, Qj ζj )βj −δj eiη(θj ,Qj ζj ) , (6.6)
Pn Pn
κ
Dξjj e−iη(ξj , i=j+1 ζi ) = (−1)|κj | η(−, ζj+1 + · · · + ζn )κj e−iη(ξj , i=j+1 ζi ) , (6.7)

where the multi-indices δ = (δ1 , . . . , δn ) ∈ 4n N

0 and κ = (κ1 , . . . , κn−1 , 0) ∈ N4n
0
satisfy the constraints δ ≤ β and κ ≤ α respectively,
(ii) on the translations of the functions f1 , . . . , fn by application of Lemma 2.3:

δ γ γ
Φ(eiη(−,ζj ) Dθjj Dξ11,j−1 . . . Dξj−1
j−1,1
fj(Qj θj +Pj−1 ξi ) ), (6.8)
i=1
 29

where γ = {(γi,1 , . . . , γi,n−i ) ∈ 0

Pn−i
N4(n−i)
: i = 1, . . . , n − 1} satisfies the constraint
κi + l=1 γi,l ≤ αi , and
(iii) on the unitary operator U (ξ+ ) = U (ξ1 ) . . . U (ξn ) in the strong operator topology:
Pn−j Pn−j
α −κj − γj,i
Dξjj i=1
U (ξj ) Ψ = U (ξj ) η(−, P )αj −κj − i=1 γj,i
Ψ, (6.9)

since Ψ ∈ D∞ ω .
After differentiation, the exponential prefactors and the effect of the metric η are both
redundant when considering the absolute value in the semi-norms pα,β . Altogether,
 

e Ψ
pα,β F (f1 ⊗ · · · ⊗ fn ) Λ(un (Q1 , . . . , Qn )) (ζ1 , . . . , ζn )
X
≤ sup Cκ,γ,δ,α,β |(Q1 ζ1 )β1 −δ1 . . . (Qn ζn )βn −δn (ζ2 + · · · + ζn )κ1 . . . (ζn )κn−1 |
(θ,ξ) κ,γ,δ
γ γ
|hΨ, Φ(eiη(−,ζ1 ) Dθδ11 f1(Q1 θ1 ) ) . . . Φ(eiη(−,ζn ) Dθδnn Dξ11,n−1 . . . Dξn−1
n−1,1
fn(Qn θn +Pn−1 ξi ) )
i=1
Pn−1
α1 −κ1 − γ1,i αn
U (ξ+ ) P i=1 ...P Ψi|. (6.10)

The monomials in ζ are of finite degree and will not affect the rapid decrease. Thus,
the focus is on the inner product in each summand.
R
Since Ψω is cyclic for P( 4 ), there exist constants C ′ ∈ and N ′ ∈ , as well as C N
R
functions h(s,1) , . . . , h(s,s) ∈ D( 4 ) for s = 1, . . . , N ′ such that

N ′
X
′
Ψ = C Ψω + Φ(h(s,s) ) . . . Φ(h(s,1) ) Ψω . (6.11)
s=1

Similarly, each σ := (κ, γ, α) admits the existence of constants Cσ′′ ∈ and Nσ′′ ∈ C N,
(t,1) (t,t)
R
as well as functions gσ , . . . , gσ ∈ D( 4 ) for t = 1, . . . , Nσ′′ such that
Pn−1
U (ξ+ ) P α1 −κ1 − i=1 γ1,i
. . . P αn Ψ
Nσ ′′
X
= Cσ′′ U (ξ+ ) Ψω + U (ξ+ ) Φ(gσ(t,1) ) . . . Φ(gσ(t,t) ) Ψω
t=1
Nσ′′
X
= Cσ′′ Ψω + Φ((gσ(t,1) )(ξ+ ) ) . . . Φ((gσ(t,t) )(ξ+ ) ) Ψω , (6.12)
t=1

due to the covariance of the field and the invariance of Ψω under translations (4.9).
Recalling
N0
R
D( 4 ) := C
from the Borchers-Uhlmann algebra allows the expression
30

of each inner product in (6.10) as the sum

Nσ
′ ′′
N X
X
≺ ωs+n+t , h(s)∗ ⊗ eiη(−,ζ1 ) Dθδ11 f1(Q1 θ1′ ) ⊗ · · ·
s=0 t=0
γ γ
⊗ eiη(−,ζn ) Dθδnn Dξ11,n−1 . . . Dξn−1
n−1,1
fn(Qn θn +Pn−1 ξi ) ⊗ (gσ(t) )(ξ+ ) ≻, (6.13)
i=1

(t) (t,1)
where h(s)∗ := h(s,1) ⊗ · · · ⊗ h(s,s) with h(0) := C ′ , and (gσ )(ξ+ ) := (gσ )(ξ+ ) ⊗ · · · ⊗
(t,t) (0)
(gσ )(ξ+ ) with (gσ )(ξ+ ) := Cσ′′ . In every summand of (6.13), the Fourier transform is
evaluated only with respect to the slots of the functions f1 , . . . , fn . Hence, it will be
useful to restructure the (s + n + t)-point distributions.
The final part is, in essence, similar to Hörmander’s inclusion [Hö03, Thm. 8.2.12]
for the wavefront set of a distribution in the context of the Schwartz kernel theorem
N
(Theorem 2.4). Consider s, t ∈ 0 and (0, . . . , 0, ζ1′ , . . . , ζn′ , 0, . . . , 0) ∈
/ π2 (WF(ωs+n+t )).
By definition, there exists an open conic neighbourhood Γs,t of this point in which
 (s) 
F (e
h ⊗ fe1 ⊗ · · · ⊗ fen ⊗ ge(t) ) ωs+n+t (0, . . . , 0, ζ1 , . . . , ζn , 0, . . . , 0)
= ≺ ωs+n+t, e
h(s) ⊗ eiη(−,ζ1 ) fe1 ⊗ · · · ⊗ eiη(−,ζn ) fen ⊗ ge(s) ≻ (6.14)

decreases rapidly, for all e

N
R N
R
h(s) ∈ s D( 4 ), fe1 , . . . , fen ∈ D( 4 ), and ge(t) ∈ t D( 4 ). R
This can be extended to identify when (6.13) is rapidly decreasing, by accounting for
the different values of s and t. Set N as the maximum value of the set {N ′ } ∪ {Nσ′′ :
σ = (κ, γ, α)}, which must be finite under the implemented cyclicity, and consider a
point (ζ1′ , . . . , ζn′ ) from the set
\ 
(ζ1 , . . . , ζn ) ∈ R4n \ {0} : (0, . . . , 0, ζ1 , . . . , ζn, 0, . . . , 0) ∈/ π2 (WF(ωs+n+t)) .
s,t∈{0,...,N }
(6.15)

Since a finite intersection of open sets is open, there exists4 an open conic neighbour-
T
hood Γ := s,t∈{0,...,N } Γs,t of the point (ζ1′ , . . . , ζn′ ) in which (6.13) decreases rapidly.
The dependence of (6.13) on (θ, ξ) is relevant only to the localisation of ωs+n+t, and
has been bypassed by considering, in (6.15), only the covectors which never belong to
(the projection of) the wavefront set, such that the Fourier transform of ωs+n+t will
rapidly decrease regardless of where it is localised. Consequently, the right hand side
of (6.10) is rapidly decreasing in Γ, and obviously so is the left hand side, irrespective
of α and β. This implies that (ζ1′ , . . . , ζn′ ) ∈
/ π2 (WF(Λ(u e Ψ (Q1 , . . . , Qn )))). 
n
4
Under the canonical embeddings of Γs,t ⊂ R 4(s+n+t)
into R 4(N+n+N)
.
 31

6.3 Lemma. [BFK96, Thm. 4.5]

N N
Let n ∈ and let s, t ∈ 0 . If ω fulfills the µSC and

(c1 , 0; . . . ; cs , 0; a1 , ζ1 ; . . . ; an , ζn ; b1 , 0; . . . ; bt , 0) ∈ WF(ωs+n+t), (6.16)

then there exists a graph H ∈ Gn and an immersion of H into R4 which instantiates

R
the point (a1 , ζ1 ; . . . ; an , ζn ) ∈ (T∗ 4 )×n \ Z ×n .

Proof. Assume that s and t are both non-zero; the other cases can be proven by an
obvious reduction of the following procedure.
Since ω fulfills the µSC, there exists a graph G ∈ Gs+n+t and an immersion (x, γ, k)
R
of G into 4 which instantiates the point

(c1 , 0; . . . ; cs , 0; a1 , ζ1 ; . . . ; an , ζn ; b1 , 0; . . . ; bt , 0) ∈ (T∗ R4)×(s+n+t) \ Z ×(s+n+t) . (6.17)

In particular, the following covector relations are satisfied:

X X
0= kr (ci ) − kr (ci ), i = 1, . . . , s, (6.18)
er between er between
vi and vj>i vi and vj<i
X X
0= kr (bi ) − kr (bi ), i = 1, . . . , t. (6.19)
er between er between
vs+n+i and vs+n+i and
vj>s+n+i vj<s+n+i

At c1 , the covector 0 is a sum of only future-directed covectors kr (c1 ), each of which

must be zero. Due to their constancy, the covector fields kr vanish along all piecewise
smooth curves γr which start or end at c1 . As a result, the covector 0 at c2 is again a
sum of only future-directed covectors kr (c2 ), each of which must be zero. This process
can be iterated up to cs , nullifying the covector fields kr along curves γr which either
start or end at ci , for all i ∈ {1, . . . , s}. Since these covector fields, curves, and points
are redundant in the immersion, they can be dropped. Similarly, the covector 0 at bt
is a sum of only past-directed covectors kr (bt ), each of which must be zero. Traced in
the reverse direction, an identical argument eliminates the covector fields kr along all
curves γr which either start or end at bi , for all i ∈ {1, . . . , t}, allowing these covector
fields, curves, and points to be removed from the immersion. Hence, the graph G can
be restricted to the subgraph H ∈ Gn , induced by the vertex subset that maps to the
point (a1 , . . . , an ) under the immersion (x, γ, k). In other words, this is an immersion
of H which instantiates the point (a1 , ζ1 ; . . . ; an , ζn ). 
32

6.4 Theorem.
R
If the state ω on the algebra P( 4 ) of field operators fulfills the µSC, then so does every
R
admissible vector state on the algebra P( 4 , ΣQ ) of warped field operators.

Proof. Let ω Ψ be an admissible vector state, and let Q1 , . . . , Qn ∈ ΣQ be arbitrary.

From the inclusion offered by Theorem 3.12 regarding the wavefront sets of oscillatory
integrals acting on symbolic distributions, (5.22) gives

WF(ωnΨ (Q1 , . . . , Qn )) = WF( IL n

η
e Ψ (Q1 , . . . , Qn )))
◦ Λ(un
e Ψ
⊆ WF(Λ(u n (Q1 , . . . , Qn )))

⊆ (x1 , ζ1 ; . . . ; xn , ζn ) ∈ (T∗ R4 )×n \ Z ×n :
e Ψ
(ζ1 , . . . , ζn ) ∈ π2 (WF(Λ(un (Q1 , . . . , Qn )))) , (6.20)

where the last inclusion is a simple consequence of the definition of the wavefront set.
Lemma 6.2 then provides, for some N ∈ 0 , the inclusion N

R e Ψ (Q1 , . . . , Qn ))))
(x1 , ζ1 ; . . . ; xn , ζn ) ∈ (T∗ 4 )×n \ Z ×n : (ζ1 , . . . , ζn ) ∈ π2 (WF(Λ(un
[
⊆

(x1 , ζ1 ; . . . ; xn , ζn ) ∈ (T R
∗ 4 ×n
) \Z :×n

s,t∈{0,...,N }

(0, . . . , 0, ζ1 , . . . , ζn , 0, . . . , 0) ∈ π2 (WF(ωs+n+t )) . (6.21)

R
Consider (x1 , ζ1 ; . . . ; xn , ζn ) ∈ (T∗ 4 )×n \ Z ×n such that (0, . . . , 0, ζ1 , . . . , ζn , 0, . . . , 0)
belongs to π2 (WF(ωs+n+t )) for some s, t ∈ {0, . . . , N }. This implies the existence of a
R
point (c1 , . . . , cs , a1 , . . . , an , b1 , . . . , bt ) ∈ 4(s+n+t) such that

(c1 , 0; . . . ; cs , 0; a1 , ζ1 ; . . . ; an , ζn ; b1 , 0; . . . ; bt , 0) ∈ WF(ωs+n+t), (6.22)

and, by Lemma 6.3, there exist a graph H ∈ Gn and an immersion of H into 4 which R
instantiates the point (a1 , ζ1 ; . . . ; an , ζn ).
Let H ∈ Gn be a reduced version of H; it has the same vertex set, but admits only
one edge between every pair of vertices connected by one or more edges in H. Define
R
an immersion (x′ , γ ′ , k′ ) of H = ({vi }, {er }) into 4 in terms of H = ({vi }, {er }) as
(i) x′ (vi ) = xi ,
(ii) γr′ is the straight line connecting x′ (vi ) and x′ (vj ) if er connects vi and vj ,
P
(iii) kr′ is the parallel transport along γr′ of kr′ (x′ (vi )) = er between vi and vj kr (x(vi )),
where vi and vj are the vertices corresponding to the endpoints of γr′ .
As this immersion is constructed precisely to instantiate the point (x1 , ζ1 ; . . . ; xn , ζn ),
 33

it can be concluded that


(x1 , ζ1 ; . . . ; xn , ζn ) ∈ (T∗ R4 )×n \ Z ×n :
(0, . . . , 0, ζ1 , . . . , ζn , 0, . . . , 0) ∈ π2 (WF(ωs+n+t )) ⊆ Γn , (6.23)

for every s, t ∈ {0, . . . , N }. Altogether, the inclusions (6.20), (6.21), and (6.23) prove
the fulfillment of the µSC for the admissible vector state ω Ψ . 

Since the wavefront set of a sum of distributions is contained within the union of
their individual wavefront sets, a finite convex combination of admissible vector states
also fulfills the µSC under the criterion of Theorem 6.4.

7 Conclusion and Outlook

To recapitulate, an admissible vector state on the warped algebra P( 4 , ΣQ ) fulfills R
R
the µSC if the state ω on the original algebra P( 4 ) fulfills the µSC. Recall that ω
R
refers to the state on the Borchers-Uhlmann algebra B( 4 ) that constructs the GNS
R
Hilbert space in which it is then viewed as a vector state on P( 4 ). On restoring the
spectrum condition, as is reasonable in Minkowski spacetime, ω naturally fulfills the
µSC [BFK96, Thm. 4.6]. Hence, so does every admissible vector state on P( 4 , ΣQ ). R
R
Similar to the warped convolution, the algebra P( 4 ) can be deformed with the
Rieffel product ([BLS11], [Rie93]). However, admissible vector states are generally not
positive, and must be composed with some operation such as the warped convolution
[BLS11, Lem. 2.4] or a (modified) Gaußian convolution [KNW09] to enforce positivity.
In both cases, identical results on the fulfillment of the microlocal spectrum condition
can be derived, based on almost identical proofs.
The analysis of singular field correlations can be further refined by using Sobolev
wavefront sets [JS02, App. B] instead of the smooth wavefront set. This corresponds
to the generalisation from Hadamard states to adiabatic states [JS02]. It perhaps has
R R
the potential to distinguish between P( 4 ) and P( 4 , ΣQ ), since the monomials in
(6.10) suggest that the microlocal Sobolev regularity of the symbolic distribution of a
warped n-point distribution, and hence of the warped n-point distribution itself (with
a suitable adaptation of Theorem 3.12), will be lower than that of the corresponding
R
n-point distribution. In other words, an adiabatic state on P( 4 ) and P( 4 , ΣQ ) is R
expected to be of lower order on the latter. One hopes that this distinction relates to
the degree of nonlocality in noncommutative Minkowski spacetime.
34

References
[ABJJ07] E. Akofor, A. P. Balachandran, S. G. Jo, and A. Joseph, Quantum Fields on the
Groenewold-Moyal Plane: C, P , T and CP T , J. High Energy Phys. 08 (2007), no. 45,
doi:10.1088/1126-6708/2007/08/045.

[BFK96] R. Brunetti, K. Fredenhagen, and M. Köhler, The Microlocal Spectrum Condition

and Wick Polynomials of Free Fields on Curved Spacetimes, Commun. Math. Phys.
180 (1996), 633–652, doi:10.1007/BF02099626.

[BLS11] D. Buchholz, G. Lechner, and S. J. Summers, Warped Convolutions, Rieffel Defor-

mations and the Construction of Quantum Field Theories, Commun. Math. Phys 304
(2011), 95–123, doi:10.1007/s00220-010-1137-1.

[Bor62] H.-J. Borchers, On Structure of the Algebra of Field Operators, Nuovo Cim 24 (1962),
214–236, doi:10.1007/BF02745645.

[BS08] D. Buchholz and S. J. Summers, Warped Convolutions: A Novel Tool in the Con-
struction of Quantum Field Theories, Quantum Field Theory and Beyond (2008),
107–121, doi:10.1142/9789812833556 0007.

[Con07] J. B. Conway, A Course in Functional Analysis, Graduate Texts in Mathematics 96,

Springer New York, 2007, doi:10.1007/978-1-4757-4383-8.

[DFR95] S. Doplicher, K. Fredenhagen, and J. E. Roberts, The Quantum Structure of Space-

time at the Planck Scale and Quantum Fields, Commun. Math. Phys. 172 (1995),
187–220, doi:10.1007/BF02104515.

[DU77] J. Diestel and J. J. Uhl, Vector Measures, Mathematical Surveys and Monographs
15, American Mathematical Society, 1977, doi:10.1090/surv/015.

[FJ98] F. G. Friedlander and M. Joshi, Introduction to the Theory of Distributions, Cam-

bridge University Press, 1998, doi:10.3138/9781442615151.

[GL07] H. Grosse and G. Lechner, Wedge-Local Quantum Fields and Noncommutative

Minkowski Space, J. High Energy Phys. 11 (2007), no. 012,
doi:10.1088/1126-6708/2007/11/012.

[GL08] , Noncommutative Deformations of Wightman Quantum Field Theories, J.

High Energy Phys. 09 (2008), no. 131, doi:10.1088/1126-6708/2008/09/131.

[GLLV13] H. Grosse, G. Lechner, T. Ludwig, and R. Verch, Wick Rotation for Quantum Field
Theories on Degenerate Moyal Space(-time), J. Math. Phys. 54 (2013), no. 022307,
doi:10.1063/1.4790886.

[Gru09] G. Grubb, Distributions and Operators, Graduate Texts in Mathematics 252, Springer
New York, 2009, doi:10.1007/978-0-387-84895-2.
 35

[GS94] A. Grigis and J. Sjöstrand, Microlocal Analysis for Differential Operators, London
Mathematical Society Lecture Note Series 196, Cambridge University Press, 1994,
doi:10.1017/CBO9780511721441.

[GW14] C. Gerard and M. Wrochna, Construction of Hadamard States by Pseudo-Differential

Calculus, Commun. Math. Phys. 325 (2014), 713–755,
doi:10.1007/s00220-013-1824-9.

[GW19] , Analytic Hadamard States, Calderón Projectors and Wick Rotation near
Analytic Cauchy Surfaces, Commun. Math. Phys. 366 (2019), 29–65,
doi:10.1007/s00220-019-03349-z.

[Hö71] L. Hörmander, Fourier Integral Operators I, Acta Math. 127 (1971), 79–183,
doi:10.1007/BF02392052.

[Hö03] , The Analysis of Linear Partial Differential Operators I: Distribution Theory

and Fourier Analysis, 2 ed., Classics in Mathematics, Springer Berlin, Heidelberg,
2003, doi:10.1007/978-3-642-61497-2.

[Jos99] M. Joshi, Introduction to Pseudo-Differential Operators, 1999, arXiv:math/9906155.

[JS02] W. Junker and E. Schrohe, Adiabatic Vacuum States on General Spacetime Manifolds:
Definition, Construction, and Physical Properties, Ann. Henri Poincaré 3 (2002),
1113–1181, doi:10.1007/s000230200001.

[KNW09] D. Kaschek, N. Neumaier, and S. Waldmann, Complete Positivity of Rieffel’s Defor-

R
mation Quantization by Actions of d , J. Noncommut. Geom. 3 (2009), 361–375,
doi:10.4171/jncg/40.

[KW91] B. Kay and R. Wald, Theorems on the Uniqueness and Thermal Properties of Station-
ary, Nonsingular, Quasifree States on Spacetimes with a Bifurcate Killing Horizon,
Physics Reports 207 (1991), no. 2, 49–136, doi:10.1016/0370-1573(91)90015-E.

[LW11] G. Lechner and S. Waldmann, Strict Deformation Quantization of Locally Convex

Algebras and Modules, 2011, arXiv:1109.5950v1 [math.QA].

[Mor19] V. Moretti, Fundamental Mathematical Structures of Quantum Theory: Spectral The-

ory, Foundational Issues, Symmetries, Algebraic Formulation, Springer Cham, 2019,
doi:10.1007/978-3-030-18346-2.

[PV04] M. Paschke and R. Verch, Local Covariant Quantum Field Theory over Spectral Ge-
ometries, Class. Quantum Grav. 21 (2004), 5299–5316,
doi:10.1088/0264-9381/21/23/001.

[Rad96] M. Radzikowski, Micro-Local Approach to the Hadamard Condition in Quantum Field

Theory on Curved Space-Time, Commun. Math. Phys. 179 (1996), 529–553,
doi:10.1007/BF02100096.
36

[Rie93] R
M. A. Rieffel, Deformation Quantization for Actions of d , Memoirs of the American
Mathematical Society 106, American Mathematical Society, 1993,
doi:10.1090/memo/0506.

[San10] K. Sanders, Equivalence of the (Generalised) Hadamard and Microlocal Spectrum

Condition for (Generalised) Free Fields in Curved Spacetime, Commun. Math. Phys.
295 (2010), 485–501, doi:10.1007/s00220-009-0900-7.

[Sch12] K. Schmüdgen, Unbounded Self-Adjoint Operators on Hilbert Space, Graduate Texts

in Mathematics 265, Springer, 2012, doi:10.1007/978-94-007-4753-1.

[Sch20] , An Invitation to Unbounded Representations of ∗-Algebras on Hilbert Space,

Graduate Texts in Mathematics 285, Springer Cham, 2020,
doi:10.1007/978-3-030-46366-3.

[Str09] A. Strohmaier, Microlocal Analysis, Quantum Field Theory on Curved Spacetimes:

Concepts and Mathematical Foundations (C. Bär and K. Fredenhagen, eds.), Lecture
Notes in Physics 786, Springer Berlin, Heidelberg, 2009,
doi:10.1007/978-3-642-02780-2 4.

[SV00] H. Sahlmann and R. Verch, Passivity and Microlocal Spectrum Condition, Commun.
Math. Phys. 214 (2000), 705–731, doi:10.1007/s002200000297.

[SV01] , Microlocal Spectrum Condition and Hadamard Form for Vector-Valued

Quantum Fields in Curved Spacetime, Reviews in Mathematical Physics 13 (2001),
no. 10, 1203–1246, doi:10.1142/S0129055X01001010.

[SVW02] A. Strohmaier, R. Verch, and M. Wollenberg, Microlocal Analysis of Quantum Fields

on Curved Space-times: Analytic Wave Front Sets and Reeh-Schlieder Theorems, J.
Math. Phys. 43 (2002), 5514–5530, doi:10.1063/1.1506381.

[SW99] H. H. Schaefer and M. P. Wolff, Topological Vector Spaces, Graduate Texts in Math-
ematics 3, Springer New York, 1999, doi:10.1007/978-1-4612-1468-7.

[SW01] R. J. Streater and A. S. Wightman, PCT, Spin and Statistics, and All That, Princeton
Landmarks in Mathematics and Physics, Princeton University Press, 2001,
doi:10.1515/9781400884230.

[SW24] A. Strohmaier and E. Witten, Analytic States in Quantum Field The-

ory on Curved Spacetimes, Ann. Henri Poincaré 25 (2024), 4543–4590,
doi:10.1007/s00023-024-01419-0.

[Tho75] E. Thomas, Integration of Functions with Values in Locally Convex Suslin Spaces,
Transactions of the American Mathematical Society 212 (1975), 61–81,
doi:10.1090/s0002-9947-1975-0385067-1.

[Uhl62] A. Uhlmann, Über die Definition der Quantenfelder nach Wightman und Haag, Wiss.
Zeitschr. Karl-Marx Univ. Leipzig 11 (1962), 213–217.
 37

[Ver99] R. Verch, Wavefront Sets in Algebraic Quantum Field Theory, Commun. Math. Phys.
205 (1999), 337–367, doi:10.1007/s002200050680.

[Ver11] , Quantum Dirac Field on Moyal-Minkowski Spacetime – Illustrating Quantum

Field Theory over Lorentzian Spectral Geometry, Acta. Phys. Polon. Supp. 4 (2011),
no. 3, 507–527, doi:10.5506/APhysPolBSupp.4.507.

[Wig96] A. S. Wightman, How it was Learned that Quantized Fields are Operator-Valued
Distributions, Fortschr. Phys. 44 (1996), 143–178, doi:10.1002/prop.2190440204.