There are several variants of the Haag-Kastler axioms for algebraic quantum field theory. Usually one associates an algebra $\mathcal{A}(O)$ to each open region $O$ of spacetime. An often-suggested axiom is that the algebra associated to a region is the same as the algebra associated to its causal completion. To be precise: Let the causal complement $O'$ of a spacetime region $O$ be the set of points that are spacelike separated from all points in $O$. Then $O''=(O')'$ is called the causal completion, and the axiom in question axiom states $$\mathcal{A}(O'')=\mathcal{A}(O).$$ (For instance, see Haag, "Local Quantum Physics," Equation III.1.10.) Due to inconsistent terminology, I'll just call this the causal shadow axiom.

Can one provide a more elaborate explanation/intuition for the causal shadow axiom or show that it holds for free scalar field theory?

In some presentations -- including, it seems, the original Haag-Kastler paper -- the authors make an alternative statement: $$\mathcal{A}(D(O))=\mathcal{A}(O),$$ where $D(O)$ is the "domain of dependence" or "causal shadow" of $O$, i.e. the set of points $p$ such that every inextendible causal curve through $p$ intersects $O$. This weaker axiom makes more sense to me. (I think the causal completion axiom implies the causal shadow axiom.)

To highlight the difference between the causal shadow and causal completion, consider $O$ as the union of two small balls centered at the points $(1,0)$ and $(-1,0)$ in $(t,x)$ coordinates. Then $O''$ will approximately be the "causal diamond" with vertices at $(\pm 1,0), (0,\pm 1)$, but the causal shadow is much smaller.

In response to comments: Another axiom is the "Haag duality" axiom: $$\mathcal{A}(O')=\mathcal{A}(O)',$$ where $\mathcal{A}'$ is the commutant. Because $\mathcal{A}''=\mathcal{A}$ for von Neumann algebras, you can iterate the duality axiom to see it implies the causal completion axiom. In some places (e.g. Haag and Araki textbooks), I only found the duality axiom stated for the case that $O$ is a causal diamond, but e.g. in "Introduction to Algebraic Quantum Field Theory" by Horuzhy (p21 and p48), Haag duality is discussed for arbitrary regions, thus implying the causal completion axiom for arbitrary regions. Rather than getting lost in a web of claims, I'd rather try to justify/refute the causal completion axiom, perhaps considering the case of a free theory.

  • $\begingroup$ Indeed, it seems that the book uses two different notions of causal completion. In Eq. (III.1.10) (where a completely misleading comment addressing to sec.4 appears) and Eq. (III.3.44) the "causal completion" seems to be nothing but the domain of dependence. Later in Sec. 4 the causal completion should be to understood as $O''$ which does not coincide with the domain of dependence in general. $\endgroup$ Oct 19, 2015 at 10:35
  • $\begingroup$ Thanks. Haag may be somehow referring to the domain of dependence as you suggest, but many other authors list this axiom and quite explicitly mean the causal completion as I defined it -- so the question remains. $\endgroup$ Oct 19, 2015 at 10:40
  • $\begingroup$ Could you give me another problematic reference where the problem appears? $\endgroup$ Oct 19, 2015 at 10:41
  • $\begingroup$ What about Araki's textbook? $\endgroup$ Oct 19, 2015 at 10:47
  • 1
    $\begingroup$ @user39158 Assume A(Q')=A(Q)'. Apply to Q=O'. Then A(O'')=A(O')'. But also A(O')=A(O)', so taking commutant of both sides gives A(O')'=A(O)''=A(O). Putting together gives A(O'')=A(O). $\endgroup$ Oct 19, 2015 at 17:18

1 Answer 1


I got a personnal proof for $ D(O)= O'' $ in Minkowski spacetime and I'm sure it holds in more general spacetimes...

Notations and conventions: $S\subset \mathcal M$ subset of the Minkowski spacetime, $\eta$ has signature (+,-,-,-) $$\textbf{causal cone}\qquad J(p) := \left\lbrace q\in\mathcal{M},\ (q-p)^2 \geq 0,\ \forall\ p\in S\right\rbrace,\quad J(S)= \bigcup_{p\in S} J(p)$$ $$\textbf{causal complement}\qquad S' := \left\lbrace q\in\mathcal{M},\ (q-p)^2 < 0,\ \forall\ p\in S\right\rbrace = \mathcal{M} \backslash J(S) $$

One wants to show (cf. comment, this is not the correct def. of $D(S)$) $$\textbf{domain of dependence} \qquad S'' \stackrel{!}{=} \left\lbrace q\in\mathcal{M},\ J(q) \subseteq J(S) \right\rbrace =: D(S) $$

This can be written as an equivalence $\ J(q) \subseteq J(S) \enspace \Longleftrightarrow\enspace q\in S''$ and note that each side can be understood as an implication $ A\ \Rightarrow\ B$ or with logical connectives $\neg A\ \text{or}\ B$. (i.e. $p\in J(q)\ \Rightarrow\ p\in J(S)$ and $ p\in S'\ \Rightarrow \ (q-p)^2 <0$)

$\underline{(\Rightarrow):}$ assuming that for all $ p\in \mathcal{M}$, either $ p\in \{q\}'\ \text{or}\ p\in J(S)$, one checks that $\ p\in S'\enspace\Rightarrow \enspace (q-p)^2 <0$.

$\underline{(\Leftarrow):}$ the contraposition of the last implication reads $\ (q-p)^2 \geq 0 \enspace\Rightarrow \enspace \exists\ s\in S,\ (p-s)\geq 0 $.

Remark: the causal completion of two points $p,q \in \mathcal{M},\ q\in I^+(p)$ (chronological future) is indeed the double cone, but not the open one $I^+(p)\cap I^-(q)$ but actually the "causal" one $J^+(p)\cap J^-(q)$

  • $\begingroup$ Consider $S= \{(0,1), (0,-1)\}$ referring to Minkowskian coordinates $x,t$ (whith the given order). What are $D(S)$ and $S''$? $\endgroup$ Oct 19, 2015 at 10:53
  • $\begingroup$ yeah, it is really a counter example... $\endgroup$
    – Noix07
    Oct 19, 2015 at 10:57
  • $\begingroup$ With the standard notion of $D(S)$ is seems to me that $D(S)=S$. With yours $D(S)$ is the diamond generated by $S$. Am I wrong? $\endgroup$ Oct 19, 2015 at 10:57
  • $\begingroup$ Thanks -- I appreciate your domain of independence definition, regardless $\endgroup$ Oct 19, 2015 at 11:00
  • $\begingroup$ Got to eat, but I think I know where the problem is: there is a difference between $J(q)\subseteq J(1,0) \cup J(-1,0)$ and $\exists\ p\in S,\ J(q)\subseteq J(p)$ $\endgroup$
    – Noix07
    Oct 19, 2015 at 11:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.