Manifold-Topology from K-Causal Order
Rafael D. Sorkin1,2 , Yasaman K. Yazdi3 , and Nosiphiwo Zwane4
arXiv:1811.09765v2 [gr-qc] 16 May 2019
1
Perimeter Institute for Theoretical Physics, 31 Caroline St. N., Waterloo ON, N2L
2Y5, Canada
2
Department of Physics, Syracuse University, Syracuse, NY 13244-1130, U.S.A.
3
Department of Physics, University of Alberta, Edmonton AB, T6G 2E1, Canada
4
University of Swaziland, Private Bag 4, Kwaluseni, M201, Swaziland
E-mail: rsorkin@pitp.ca, kouchekz@ualberta.ca, ntzwane@uniswa.sz
Abstract. To a significant extent, the metrical and topological properties of
spacetime can be described purely order-theoretically. The K + relation has proven to
be useful for this purpose, and one could wonder whether it could serve as the primary
causal order from which everything else would follow. In that direction, we prove, by
defining a suitable order-theoretic boundary of K + (p), that in a K-causal spacetime,
the manifold-topology can be recovered from K + . We also state a conjecture on how
the chronological relation I + could be defined directly in terms of K + .
Keywords: Causal structure, Causal order, K-causality, Manifold topology
1. Introduction
There is much information in the causal structure of spacetime, including information
about the topology, differentiable structure, and metric (indeed the full metric up to a
conformal factor) [1, 2, 3]. Accordingly, a field of study that could be called global causal
analysis has grown up to utilize the causal structure directly, beginning with the well
known singularity theorems [4, 5]. In [6] and [7] a positive energy theorem was proven
using arguments similar to those that feature in the proofs of the singularity theorems. In
[8] the authors show that the causality relation J + turns a globally hyperbolic spacetime
into a bicontinuous poset, which in turn allows one to define an intrinsic topology called
the interval topology that turns out to be the manifold topology. Moreover, several
approaches to understanding relativity theory, some of them as old as relativity itself,
make primary use of causal structure ([9, 10]). Applications of causal structure are farreaching and extend beyond the description of spacetime at the classical level. Causal
set theory is an approach to quantum gravity for which a type of causal structure is
fundamental. The causal set itself is a discrete set of elements structured as a partial
order, and its defining order-relation corresponds macroscopically to the causal order of
spacetime. (See [11, 12, 13] for more details on causal set theory.)
2
A natural question is how far can one get with describing manifold properties using
nothing more than order-theoretic concepts. This is of interest generally, but also in
the context of theories such as causal set theory. Workers in other fields, like computer
science, have been interested in this question as well [8]. But if one wishes to conceive of
order-theoretic relationships as the foundation of spacetime structure, it seems simplest
to have in mind a single order-relation, rather than the two or more that one sees
in most presentations.‡ The relation K + , defined in [7], was originally conceived for
this purpose. A closed and transitive generalization of I + and J + , it has properties
that make it suitable to serve as the primary causal order of a spacetime. Because
it is transitive (and acyclic in a K-causal spacetime) it lends itself naturally to ordertheoretic reasoning. And because it is topologically closed it avoids the problems with
supremums that would arise if one took a relation like I + as basic.
The relation K + was also designed to be a tool of use when one attempts to
generalize causal analysis to C 0 metrics that may fail to be everywhere smooth and
invertible. In [7] Sorkin and Woolgar used K + to extend certain results like the
compactness of the space of causal curves from C 2 Lorentzian metrics to C 0 Lorentzian
metrics, as needed for the positive-energy proof in [6]. In [14] Dowker, Garcia, and
Surya showed that K + is robust against the addition and subtraction of isolated points
or metric degeneracies. This allows such degeneracies to be present in metrics that
contribute to the gravitational path-integral, thus enabling one to include in the space
of histories topology-changing spacetimes (Lorentzian cobordisms) which contain such
degeneracies. This is of course a very interesting physical consideration. For some recent
articles on K + see [15] and [16].
Another line of thought, which comes from causal sets, also points to the desirability
of a sole order-relation (and to K + as a reasonable choice thereof). In a fundamentally
discrete context the fine topological differences that distinguish I, J, and K from each
other lose their meaning. On the other hand, the discrete-continuum correspondence is
normally introduced (at the kinematic level) in terms of so-called sprinklings which place
points at random in a Lorentzian manifold M via a Poisson process with a constant
density of ρ = 1 in natural units. By definition, the probability of sprinkling N points
N
into a region with spacetime volume V is P (N) = (ρVN )! e−ρV . This produces a causal set
whose elements are the sprinkled points and whose partial order relation is most simply
taken to be that of the manifold’s causal order restricted to the sprinkled points. Since
with a Poisson process, the probability for two sprinkled points to be lightlike related
vanishes, it makes no difference which continuum order one uses to induce an order
on the sprinkled points. However, it is sometimes convenient to consider, instead of a
random sprinkling, something like a “diamond lattice” in two-dimensional Minkowski
space, in which case either J + or K + would be the most useful choice. In all such cases
‡ The two most common are I + and J + . The “chronological future” of a point-event p, I + (p), is the
set of events accessible from p by future-directed timelike curves starting from p. The “causal future”
J + (p) is the set of points accessible from p by future-directed timelike or null curves starting from p.
The pasts sets, I − (p) and J − (p), are defined analogously, with future replaced by past.
3
one loses nothing by thinking of K + as the basic continuum-order. One sees again how
it is more natural to work with only one causal relation, i.e. one does not distinguish
between lightlike and timelike related pairs of elements, only between causally related
and unrelated pairs.
If K + is to be taken as primitive, then it must be possible in particular to recover
the manifold-topology from it, something which was not addressed in [7]. We could
perhaps obtain the manifold-topology indirectly by first defining I + in terms of K + ,
but we will not do this herein (although we will provide a conjecture suggesting how
it might be done). Instead we proceed directly from K + to the topology by defining
an order-theoretic boundary of K + (p) and demonstrating that it coincides with the
topological boundary. Removing it, we obtain a family of open sets A(p, q) from which
the topology can be reconstructed.
We begin in Section 2 by reviewing the definition and properties of K ± . We then
introduce the derived sets A± (p), which we use throughout this paper. In Section 3
we show that A± (p) is open and locally equivalent to I ± (p). From this it follows that
the order-interval A(p, q) is locally equal to I(p, q), which completes the proof that the
K-causal order yields the manifold topology. In Section 4 we state a conjecture for how
to obtain I + (p) more directly from K + (p). The appendix collects the lemmas from [7]
that we use in this paper.
Our results do not depend on the spacetime-dimension of the manifold M that we
work in.
2. K-Causality
2.1. Some Definitions
Definition 2.1. K + := ≺ (respectively K − ) is the smallest relation that contains I +
(respectively I − ) and is both transitive§ and topologically closed [7].
Regarded as a subset of M × M, the relation K + can be obtained by intersecting all
the closed and transitive sets Ri that include I + [7]:
\
K+ =
Ri ,
where
∀i, I + ⊂ Ri .
(1)
i
+
+
By K (p, M) or K (p) we denote all the points q such that p ≺ q, where p, q ∈ M.
Let O be an open subset of M. For q ∈ K + (p, O) we write p ≺O q. Figure 1 shows an
example of K + (p) and how it differs from I + (p) and J + (p). In the figure, q, r, s ∈ K + (p)
while q ∈ J + (p), r, s ∈
/ J + (p) (and also q ∈ I + (p), r, s ∈
/ I + (p)).
An open set is K-causal if and only if ≺ induces a (reflexive) partial ordering on
it, in other words iff ≺ restricted to the open set is asymmetric.k
§ p≺q≺q≺r⇒p≺r
k Minguzzi has proven [17] that K-causality coincides with stable causality, and that in consequence,
K + coincides with the the Seifert relation JS+ provided that K-causality is in force.
4
s
o
r
o
q
p
Figure 1. An example of K + (p). The wiggly lines are removed from the manifold.
q, r, s ∈ K + (p) while q ∈ J + (p), r, s ∈
/ J +.
Definition 2.2. The order theoretic or causal boundary of K + (p), denoted ∂K + (p), is
the set indicated by
∂K + (p) = {x ∈ K + (p)|x = sup{yi } for some yi ∈
/ K + (p)}.
(2)
More precisely, a point x of K + (p) belongs to ∂K + (p) iff it is the supremum with respect
to ≺ of an increasing sequence¶ of points yi all belonging to the complement of K + (p).
(Intuitively this says that one can approach x arbitrarily closely from outside of K + (p),
which is a precise order-theoretic counterpart of the topological definition of boundary.)
We next define the open sets A+ (p) which we will use in the next section to recover
the manifold topology.
Definition 2.3. A+ (p) is the set K + (p) without its causal boundary defined in (2):
A+ (p) = K + (p)\∂K + (p).
(3)
By definition, A+ (p) is open in the order-theoretic sense. We will see in the next
section that the order-theoretic boundary coincides with the topological one, therefore
making A+ (p) open in the topological sense as well.
Definition 2.4. The set A(p, q) is
A(p, q) = A+ (p) ∩ A− (q).
(4)
The set A(p, q) is a kind of “open order-interval”. Figure 2 shows an example.
3. Manifold Topology from the sets A(p, q)
Let us demonstrate that the sets A(p, q) are locally the same as the order-intervals
I(p, q). To that end, we first prove that the future- and past- sets A± (p) are topologically
open and locally equivalent to I ± (p).
¶ Instead of “increasing sequence”, one could say “directed set”
5
q
o
o
p
Figure 2. The region in gray is an example of a K-causal open interval A(p, q).
To show that A+ (p) is open, it suffices to show that the causally defined boundary
∂K + (p) that we removed from K + (p) to yield A+ (p) is also a topological boundary.
We will assume henceforth that M is K-causal and without boundary.
Lemma 3.1. x = sup yi ⇒ x = limi→∞ {yi } in the topological sense. Hence if x is (with
respect to K + ) the supremum of an increasing sequence (or directed set) of points yi
then x is also the limit of {yi } with respect to the manifold topology.
Proof. Let the increasing sequence y1 ≺ y2 ≺ y3 ... have x as its supremum, and let U be
an arbitrarily small open neighborhood of x. From Lemma A.4 we can assume without
loss of generality that U is K-convex.+ It suffices to prove that the yi eventually enter
U (after which they will necessarily remain in U because of the latter’s convexity, and
because yi ≺ x).
Now there are two possibilities. Either the yi enter U or they don’t. In the former
case we are done, so suppose the latter case, and let B = Fr(U) be the topological
(not causal) boundary of U. The symbol ‘Fr’ stands for “Frontier”, and we follow the
convention of [18] in referring to the topological boundary this way. Since U is arbitrarily
small and B is closed, we can assume without loss of generality that it is compact. By
Lemma A.3 there then exist points zi in B such that (∀i) yi ≺ zi ≺ x.
Recalling that B is compact, and passing if necessary to a subsequence, we can
assume without loss of generality that the zi converge to some point z∞ of B. Then
since the relation ≺ is closed, and since all of the zi ≺ x, we see immediately that
z∞ ≺ x. On the other hand, if (for fixed i) k > i, then we have yi ≺ yk ≺ zk , from which
follows yi ≺ zk . Therefore we see in the same way from yi ≺ zk that yi ≺ z∞ ∀i. But by
the definition of supremum, x is the least point of M for which this holds, hence x ≺ z∞ ,
which together with z∞ ≺ x implies that they are equal. This is a contradiction, since
+
A set or subset is said to be K-convex if and only if it contains the closed interval K(p, q) =
K + (p) ∩ K − (q) between every pair of its elements.
6
x ∈ U where U is an open set, whereas z∞ ∈ B and B ∩ U = ∅. Therefore {yi } must
enter U and in doing so, they converge (topologically) to x, as desired.
Lemma 3.2. Every point p of M has a neighborhood in which J + and K + agree.
Proof. Given our standing assumption that M is K-causal, Lemma A.4 tells us that
p has an arbitrarily small K-convex neighborhood U, and then Lemmas A.1-A.2 (with
O = M) tell us that K + restricted to U coincides with K + relativized to U (i.e.
computed as if U were the whole spacetime). But we know that in a small enough U,
the relation J + is closed∗ , transitive, and includes I + . It follows (from the definition of
K + ) that J + (U) includes K + (U), and therefore coincides with it (since∗ J + (U) is the
closure of I + (U)). Notice here, that we can take J + to be J + (U).
Theorem 3.3. The topological boundary of K + (p) equals its causal boundary ∂K + (p),
∀p ∈ M.
Proof. We use the criterion that x is in Fr(K + (p)), the topological boundary of K + (p),
iff every neighborhood of x contains points both inside and outside of K + (p).
First let’s show that ∂K + (p) ⊆ Fr(K + (p)). Let x be any point of ∂K + (p). Since
x itself is in K + (p), it suffices to show that every neighborhood of x contains points in
the complement of K + (p). This follows directly from Lemma 3.1 and the definition of
∂K + (p).
Conversely let us show that Fr(K + (p)) ⊆ ∂K + (p). Let x be any point of Fr(K + (p)).
First of all, we easily check that x ∈ K + (p), because the topological boundary of
any set lies within its closure, and because K + (p) is closed.
Now let U be a small K-convex open neighborhood of x (which exists by Lemma
A.4). By Lemmas A.1-A.2, we can reason as if U were all of M. And we also know
from Lemma 3.2 that if U is sufficiently small then within it, K + and J + coincide.
It’s also clear that I − (x) must be disjoint from K + (p). Otherwise choose any
z ∈ K + (p) ∩ I − (x) and notice that (since I + ⊆ K + ) I + (z) ⊆ K + (p) would be an open
neighborhood of x in K + (p), hence x would be in K + (p)’s topological interior and not
in Fr(K + (p)).
Obviously I − (x) will contain a timelike increasing sequence of points yi converging
topologically to x, and this sequence will be disjoint from K + (p) since I − (x) is. It
remains to be proven that x = sup{yi }.
That x bounds the yi from above is obvious. And because the relation ‘≺’ is
topologically closed, we have for any other upper bound z that z ≻ yi → x ⇒ z ≻ x.
Therefore x is a least upper bound, as required.
Lemma 3.4. A+ (p) is open. In fact it is the interior of K + (p).
∗ See Proposition 4.5.1 in [4] for a proof of this.
7
Proof. As defined above in (3), A+ (p) is what remains of K + (p) after we remove
its order-theoretic boundary ∂K + (p). But in the theorem just proven we have seen
that ∂K + (p) is also the topological boundary of K + (p), and removing the topological
boundary of any set whatsoever produces its interior, which by definition is open.
It follows immediately that the sets A(p, q) are open for all p and q. We claim
furthermore that A+ (p) coincides locally with I + (p), whence A(p, q) coincides locally
with I(p, q). This follows from Lemmas 3.2 and 3.4, which inform us that locally A+ (p)
is the interior of K + (p) and K + (p) = J + (p). To establish our claim, then, simply notice
that locally the interior of J + (p) is I + (p).♯
We now have all the pieces we need to derive the manifold topology from the order
‘≺’, which we do by proving that the open sets A(p, q) furnish a basis for the topology
of M in the sense that every open subset of M is a union of sets of the form A(p, q).
For this, it suffices in turn that:
Criterion: for any x ∈ M and any open set U containing x, we can find p and q such
that x ∈ A(p, q) ⊆ U.
Clearly this criterion is local in the sense that it’s enough for it to hold for U being
arbitrarily small.
Theorem 3.5. The sets A(p, q) furnish a basis for the manifold-topology.
Proof. Since K-causality is in force, strong causality also holds [17], whereby the
manifold-topology is the same as the Alexandrov topology, for which by definition the
sets I(p, q) are a basis. (See theorem 4.24 in [19].) But because locally I(p, q) = A(p, q),
as we have just seen, the sets A(p, q) are also a basis.
To summarize: because the I(p, q) constitute a basis they satisfy the criterion stated
above, and because this criterion is purely local, the sets A(p, q), which locally coincide
with the I(p, q), also satisfy it.
Remark. As seen in figure 2, there will in general be pairs of points, p, q, for which
the set A(p, q) does not agree with I(p, q). Such sets A(p, q) are still included in our
basis, but this does no harm, since by definition, any basis for a topology remains a
basis when more sets are added to it, provided that the additional sets are also open,
which of course the A(p, q) are.
4. From K + (p) to I + (p) and J + (p)
Once we have access to the manifold topology, it is a relatively easy matter to define
continuous curve, and from there to characterize I ± and J ± . Thus, for example, a curve
could be the image of a continuous function from [0, 1] ⊆ R into M, and we might
♯ This follows from the discussion on pages 33-34 and 103-105 of [4], which establishes that the
exponential map at p induces a diffeomorphism between a neighborhood of p in M and a neighborhood
of 0 in Minkowski-space which preserves the sets I + (p) and J + (p).
8
define a causal (respectively timelike) curve as one which is linearly ordered by K +
(respectively A+ ).
Nevertheless, it might be nice to characterize I + and J + more directly in terms of
K + . We conclude with a conjecture of that nature (concerning I + ).
Conjecture 4.1. K + (p)\S = I + (p), where S = {r ∈ K + (p)| every ‘full chain’ from p
to r meets ∂K + (p)}
Here, by a “full chain from p to r” we mean a subset C of M containing p and r that
is: linearly ordered by ≺ (it is a chain); order-theoretically closed in the sense that it
contains all its suprema and infima; and dense in the sense that it contains between any
two of its points a third point [(∀x, y ∈ C)(∃z ∈ C)(z 6= x, y ∧ x ≺ z ≺ y)].
Remark. We didn’t need to remove ∂K + (p) explicitly from K + (p), because it is
automatically included within S.
Acknowledgements: This research was supported in part by Perimeter Institute for
Theoretical Physics. Research at Perimeter Institute is supported by the Government
of Canada through the Department of Innovation, Science and Economic Development
Canada and by the Province of Ontario through the Ministry of Research, Innovation
and Science. YY acknowledges support from the Avadh Bhatia Fellowship at the
University of Alberta. This research was supported in part by NSERC through grant
RGPIN-418709-2012.
A. Lemmas from Sorkin-Woolgar (SW) [7]
In this appendix we collect the lemmas from [7] that we have used in this paper.
Lemma A.1. (Lemma 12 of SW): Let U and O be open subsets of M and U ⊆ O. For
p, q ∈ U, p ≺U q implies p ≺O q.
Lemma A.2. (Lemma 13 of SW): Let U and O be open subsets of M and U ⊆ O. For
p, q ∈ U, p ≺O q implies p ≺U q, if U is causally convex relative to ≺O .
Lemma A.3. (Lemma 14 of SW): Let S be a subset of M with compact boundary
Fr(S), and let x ≺ y with x ∈ S, y ∈
/ S (or vice versa). Then ∃w ∈ Fr(S) such that
x ≺ w ≺ y.
Lemma A.4. (Lemma 16 of SW): If M is K-causal then every element of M possesses
arbitrarily small K-convex open neighborhoods (M is locally K-convex).
References
[1] Malament D B 1977 Journal of Mathematical Physics 18 1399–1404
[2] Hawking S W, King A R and Mccarthy P J 1976 J. Math. Phys. 17 174–181
[3] Geroch R, Kronheimer E H and Penrose R 1972 Proceedings of the Royal Society of London
A: Mathematical, Physical and Engineering Sciences 327 545–567 ISSN 0080-4630 URL
http://rspa.royalsocietypublishing.org/content/327/1571/545
9
[4] Hawking S W and Ellis G F R 1973 The Large Scale Structure of Space-Time Cambridge
Monographs on Mathematical Physics (Cambridge University Press)
[5] Penrose R 1965 Phys. Rev. Lett. 14(3) 57–59 URL https://link.aps.org/doi/10.1103/PhysRevLett.14.57
[6] Penrose R, Sorkin R D and Woolgar E 1993 arXiv:gr-qc/9301015
[7] Sorkin R D and Woolgar E 1996 Class. Quant. Grav. 13 1971–1994 arXiv:gr-qc/9508018
[8] Martin K and Panangaden P 2006 Commun. Math. Phys. 267 563–586 arXiv:gr-qc/0407094
[9] Robb A A 1914 A Theory of Time and Space (Cambridge: University Press)
[10] Robb A A 1936 Geometry Of Time And Space (At The University Press)
[11] Bombelli L, Lee J, Meyer D and Sorkin R 1987 Phys. Rev. Lett. 59 521–524
[12] Sorkin R D 2003 Causal sets: Discrete gravity Lectures on quantum gravity. Proceedings, School
of Quantum Gravity, Valdivia, Chile, January 4-14, 2002 pp 305–327 arXiv:gr-qc/0309009
[13] Henson J 2006 arXiv:gr-qc/0601121
[14] Dowker H F, Garcia R S and Surya S 2000 Class. Quant. Grav. 17 4377–4396 arXiv:gr-qc/9912090
[15] Saraykar R V and Janardhan S 2014 arXiv:1411.1836
[16] Miller T 2017 Universe 3 27 arXiv:1702.00702
[17] Minguzzi E 2009 Commun. Math. Phys. 290 239–248 arXiv:0809.1214
[18] Dieudonné J 2009 A History of Algebraic and Differential Topology, 1900 - 1960
Modern Birkhäuser Classics (Birkhäuser Boston) ISBN
9780817649074 URL
https://www.springer.com/la/book/9780817649067
[19] Penrose R 1972 Techniques of Differential Topology in Relativity (Society for Industrial and Applied
Mathematics)