# Well-order as a construction principle for physical theories

## Abstract

Physics has up to now missed to express in mathematical terms the fundamental idea of events of a path in time and space uniquely succeeding one another. An appropriate mathematical concept that reflects this idea is a well-ordered set. In such a set every subset has a least element. Thus every element of a well-ordered set has as its definite successor the least element of the subset of all elements larger than itself. This is apparently contradictory to the densely ordered real number lines which conventionally constitute the coordinate axes in any representation of time and space and in which between any two numbers exists always another number. In this article it is shown how decomposing this disaccord in favour of well-ordered sets causes spacetime to be discontinuous.

## 1 Introduction

Whenever the motion of a particle is described in classical, special relativistic or general relativistic space and time it is always implicitly assumed that the particle moves from an initial event to a final event through a sequence of consecutive intermediate events. In other words, it is presumed that every element of the path has an unique successor. But even though this is a fundamental concept, it is never disclosed. It is neither named in Newton’s laws of motion or his law of gravitation nor mentioned in the basic principles of Special or General Relativity. Consequently, it is never expressed in mathematical terms.

Also without further ado these theories make constitutively use of the real number line by employing it as (local) coordinate axes or coordinate axes and parameter line for their respective manifolds which represent space and time.

Unfortunately these very two unnamed axioms are contradictory to each other. This becomes apparent most clearly in terms of ordered sets. An ordered set (*A*, *R*) is a set *A* together with a binary order relation *R* that is itself a subset of the set of all ordered pairs \(A\times A\) of *A*; see [1, p 10, p 17], [2, p 64, p 80].

A densely ordered set is an ordered set in which between any two elements of the set exists another element of the set; see [2, p 205]. Well known examples for this are the rational numbers or the real numbers together with the natural order of numbers; see [2, p 213]. By employing the real number line as coordinate axes this is transferred into the resulting \({\mathbb {R}}^n\). The open balls as the base of the Euclidean topology are a reflection of this. For any radius \(r>0\) exists a non-empty open ball around any \(x\in {\mathbb {R}}^n\). Since Riemannian manifolds are locally homeomorphic to a suitable \({\mathbb {R}}^n\) the dense order of \({\mathbb {R}}\) is not only a characteristic of Newtonian and Special relativistic spacetime but also of the description of space and time in General Relativity.^{1}

Another type of ordered sets are well-ordered set in which every non-empty subset has a least element; see [2, p 224]. Therefore every element of a well-ordered set has as its definite successor the least element of the subset of all elements larger than itself. This evidently provides appropriate means to mathematically represent the idea of events in spacetime uniquely succeeding one another. The defining example for well-ordered sets are ordinal numbers. Among these the natural numbers together with the natural order of numbers are the best known instance. But they are by far not the only members of this class of numbers. In fact, they just open the gates into this transfinite realm.

Densely ordered sets and well-ordered sets are obviously irreconcilable. A familiar example for this incongruity is the countability of \({\mathbb {Q}}\): bijections form \({\mathbb {Q}}\) to \({\mathbb {N}}\) are known to exist but none of them preserve the order of \({\mathbb {Q}}\).

Since each of the two unnamed axioms of spacetime respectively incorporates one of these types of ordered sets they cannot be assumed in combination.

As it is shown in [4] or [5, chapter 1.6] the concept of ordered sets is already a part of the definition of Minkowski space. However, with regard to the question if Minkowski space should be a densely ordered set or a well-ordered set nothing is specified. In Sect. 2 it is explicated that settling this unanswered question in favour of well-ordered sets causes spacetime to be discontinuous. More specifically, well-ordered Minkowski spacetime proves to consist of representations of ordinal numbers. For the sake of completeness the alternative of a densely ordered spacetime is stated as well with the tacit understanding that this choice is equivalent to renouncing the conception of a particle moving along a sequence of consecutive events. All this is achieved without recourse to the notion of quantum gravity. Thus, already a reformulation of electrodynamics over a well-ordered spacetime which produces the existing (large amount of) experimental results can verify that time and space are discontinuous. But more importantly this will provide a link between the fundamental principle of events uniquely succeeding one another and experimental data.

As a first step towards a rephrasing of electrodynamics (or the theory of any other force) an approach to well-ordered spacetime is presented in Sect. 3 which is different and more general than that of Sect. 2: for the reason of being a field which naturally incorporates ordinal numbers (and other infinite elements) besides real numbers Conway’s class of surreal numbers \(\text {No}\) prove to be an appropriate mathematical ground for investigating well-ordered spacetimes. Especially a concise way of navigating discontinuous spacetimes by virtue of pairs of infinite elements and their additionally included multiplicative inverse elements is described. Furthermore they allow for a consistent way of handling spacetimes based on representations of ordinal numbers that contain limit ordinals. The latter leads to the definition of virtual events. This is followed by opposing the exponential map of Riemannian geometry to the idea of visiting all elements of a geodesic in their natural order by consecutively transiting to a distinct next neighbour in Sect. 4. Section 5 exhibits in terms of cardinality how much unnecessary information is introduced into path integration when a densely ordered spacetime is used instead of a well-ordered spacetime. Finally, Sect. 6 takes a look on the quantisation of the area operator of Loop Quantum Gravity from the angle of well-ordered sets and arrives at the result that LQG (unwittingly) uses the lines of argumentation presented in the preceding sections. The reconstruction of \({\mathbb {R}}^4_1\) from a \({\mathbb {R}}^4\) by means of ordered sets in Appendix A is not fully part of the main line of reasoning of this article. But it is nevertheless worth noting because it describes how the need to keep the disjoint ordered and unordered subsets of a partial causal order apart during linear transformations leads to *c*.

## 2 Causality and order types

As mentioned in the introduction, the mathematical representation of the idea of “a distinct next neighbour” is the concept of well-order. In a well-ordered set \((A,<^{\text {ord(A)}})\) every subset has a least element. Thereby, every element \(\alpha \) of a well-ordered set *A* has as its unique successor \(\alpha +1\) the least element of the subset of all elements of \((A,<^{\text {ord(A)}})\) larger than itself.

In contrast to this, between any two elements *a*, *c* of a densely ordered set \((A,<^{\text {den}})\) always exists a third element *b* of \((A,<^{\text {den}})\). Thus, a subset of \((A,<^{\text {den}})\) consisting of elements larger than a specific member *a* of \((A,<^{\text {den}})\) has no least element. The strict mathematical formulation of the conception of infinitesimal vicinity is based on this. Consequently, there are no definite next neighbours in an infinitesimal vicinity, or more succinctly, the expression “infinitesimal next neighbour” contains a contradiction.

*A*,

*R*), the issue of choosing a well-ordered or a densely ordered set is rephrased to the question which set of numbers is well-ordered or densely ordered by the natural order <.

With regard to the natural order neither \(({\mathbb {R}},<)\) nor \(({\mathbb {Q}},<)\) is well-ordered, because both contain subsets which have no least element. Moreover, every uncountable subset of \(({\mathbb {R}},<)\) is not well-ordered. But all of these sets are densely ordered by <, and for this reason allow infinitesimal vicinities for all events.

In summary, if events are to have distinct successors and if this order is to be expressed by the natural order <, they have to be elements of a countable ordinal number \(\alpha <\omega _1\) to which \(({\mathbb {R}},<)\) merely serves as an embedding space. Notably, admissible events necessarily form a discontinuous subset of \(({\mathbb {R}},<)\). Quantisation is the special case of discontinuity where events are evenly timed and spaced.

*M*. But this requirement is already a member of said set because, as has just been shown, the causality condition (2) actually states the formation of an ordered type, or respectively, an ordered set \(\left( A,R\subset A\times A\right) \). Firstly, the constraint of events being put in sequence along the \(x^0\)-axis by the standard order < determines the binary relation

*R*:

*M*by \(M(\alpha )\) in (1); cf the rhs of (8).

For more details of the prevalence of time as the carrier of causal order over the spatial dimensions as it is expressed by \(\Rightarrow \) in (6) and (7) see Appendix A.

A question that immediately arises when utilising well-ordered spacetime is how to handle an how to interpret spacetimes that are based on ordinal numbers containing limit ordinals larger than \(\omega =|{\mathbb {N}}|\). To this end \(M(\alpha )\) is not really apt. A mathematical framework that is more suitable for these purposes are the surreal numbers which are discussed in Sect. 3.

## 3 Surreal spacetime

An appropriate mathematical ground for investigating well-ordered spacetimes is provided by Conway’s surreal numbers \(\text {No}\) [7]. This class of numbers comprises real numbers as well as infinite and subfinite elements like \(\omega \) and \(\frac{1}{\omega }\). In [7] elements like \(\frac{1}{\omega }\) are called infinitesimal elements. To avoid confusion with anything that is based on the concept of infinitesimal vicinity (see second paragraph of Sect. 2), \(\frac{1}{\omega }\in \text {No}\) and alike elements are called subfinites in this article. For the conceptual reasons for this distinction see Appendix B.

^{2}Thus the exponents of the orders of magnitude establish a causal order of events, which is the task already committed to the \(x^0\)-components of events. These two orders are accordable by an order-reversing bijection between the sets \(\{y_\beta \}\) and \(\{r^0_\beta \}\):

*s*of \(\text {No}^4\) in its normal form is harmonised with the natural order of real numbers, \(\text {No}^4\) relegates to its subset of elements of countable normal length \(\alpha =\text {nl}(s)<\omega _1\) with causally ordered members of different \(\omega ^{y_\beta }{\mathbb {R}}^4\) as candidates for events in spacetime:

It should be clear by now that \(\text {No}^4\) is utilised in terms of paths. Every element of (23) constitutes a discontinuous path to be paced out by multiplication by eligibly selected \(\omega ^{r^0_\delta }\). The information about the transition from \(r_0\) to \(r_\delta \) is contained in the pair \((r_0+\frac{r_\delta }{\omega ^{r^0_\delta }},\omega ^{r^0_\delta })\); for a detailed discussion of this with regard to topology see Appendix B. Obviously forces act whenever the \(r_\beta \) do not form a straight line and/or the distance between consecutive coefficients \(r_{\beta +1}-r_\beta \) and \(r_{\beta +2}-r_{\beta +1}\) varies. Furthermore every \(r_\delta \) is a point of intersection for many paths each of which offers a different succeeding event. In this vein \({\mathbb {M}}(\alpha )\) (23) acts as a stage of potential next events.

But even that there are \(\omega _1\)-many paths intersecting at \(r_\delta \), i.e. that there is a full \({\mathbb {R}}^4_1\)-cone of successors \(r_\pi \) available at \(r_\delta \), this does not cure the contradiction in the statement “infinitesimal next neighbour”, because each of this potential *next* events \(r_\pi \) can only be reached by means of a discontinuous path. And what is even more important, all potential events with time component less than that of the successor \(r_\eta \) which became reality \(r^0_\delta<r^0_\pi <r^0_\eta \) cannot be revisited, as it would be necessary to establish the density needed for an infinitesimal vicinity, or else causality is breached. See Sect. 4 for more on this. On the other hand the interpretation of every \(r_\pi \) having a probability to be picked as the next real event is readily at hand. A \(s\in {\mathbb {M}}(\alpha )\) is then established step-by-step instead of being regarded as an already complete entity which deterministically fixes a path.

Reversely, a stage-like \(M(\alpha )\) is transformed into \({\mathbb {M}}(\alpha )\) by firstly generating the set of all path (2) and (3) allow in (8) and repeating this step for all combinations \(\times _{i=0}^3(R^i(\alpha ),<)\) of all order embeddings of \(\alpha \). After this, aligning all these paths along orders of magnitude by using the negative of the \(r^0\)-components of their points as exponents for \(\omega \) leads to \({\mathbb {M}}(\alpha )\).

*s*.

Finally, the independent \({\mathbb {R}}^4\) at different orders of magnitude \(\omega ^\beta \) need not have parallel bases. Whenever the bases \(\{e_{\beta i}\}\) between the respective \({\mathbb {R}}^4\) at each \(\omega ^\beta \) are nonparallel, \({\mathscr {Q}}(c;r_\beta -x_0)\) in (21) has to be replaced by \({\mathscr {Q}}(c;Tr_\beta -x_0)\) where *T* is a transformation that aligns \(\{e_{\beta i}\}\) and \(\{e_{0i}\}\).

## 4 Comparing subsets of \(\text {No}^n\) to Riemannian geometry

^{3}In all formal glory:

*s*resulting from combining \(\varXi \) with all possible \(\times _{i=0}^nr^i\) among these bijections. Multiplying an element

*s*of (27) by every element of the sequence \(\{\omega ^\chi \}\) yields the sequence \(\{s\omega ^\chi \}\) in which the respective coefficients of the order of magnitude \(\omega ^0=\omega ^{-\chi +\chi }\) form an uncountable sequence \(\{r(\chi )\}\) of members of \({\mathbb {R}}^n\) which is of course the sequence of coefficients of \(\omega ^{-\chi }\) in (27). But no matter which

*s*is the basis for this sequence, none of the involved images \(r^i(\chi )\) reflects the order of \(\{r(\chi )\}\). Inversely, regrouping \(\{r(\chi )\}\) by using one of the \(r^i\) in its natural order calls up the \(\chi \) erratically; see (26). Since the orders of magnitude in the definition of the normal form of a \(s\in \text {No}^n\) decline along a well-order, surreal numbers sustain the idea of a definite next element. In the orders of magnitude’s set of real coefficients in a natural order the transitions between orders of magnitudes are always leaps even when the well-order and the natural order are harmonised as it is done in the following of (18). (26) and the ensuing choice is also present in the exponential map of Riemannian geometry, though not as obvious as in (27). By definition the exponential map takes tangent vectors \(v\in T_o(M)\) at a point

*o*in a complete manifold

*M*to the point \(\gamma _v(1)\) of the geodesic with \(\gamma _v(0)=o\) whose tangent vector at

*o*is

*v*[3, p 70]:

*o*determines by virtue of the exponential map a transition from \(o=\gamma _v(0)\in M\) to \(p=\gamma _v(1)\in M\) in such a way that the portion of

*M*belonging to \(\gamma _v:(0,1)\subset {\mathbb {R}}\rightarrow M\) is omitted. So \(\exp _o\) mediates a leap from

*o*to

*p*(with

*v*being the designator of

*p*at

*o*) instead of facilitating a smooth passage between them. The attempt to restore the idea of a smooth progression from

*o*to

*p*is made by virtue of the density of \({\mathbb {R}}\) that transpires \(\exp _o\); see [3, p 71]:

*o*and

*p*is accessible by means of \(\exp _o\). But the idea of smoothly progressing from

*o*to

*p*, i.e. visiting all elements of \(\gamma _v\) in their natural order by consecutively transiting to a distinct next neighbour is not included in (29). If it were a part of (29), the real interval \([0,1]\subset {\mathbb {R}}\) in its entirety would have to be the representation of an (necessarily innumerable) ordinal number which it is not. Since Riemannian geometry is fundamentally build on the concept of infinitesimal vicinities it is not very surprising that the notion of elements of tangent spaces link elements of a Riemannian manifold as definite next neighbours that (28) hints at does not resonate with the rest of the theory. The ignorance of members of a interval of \(\gamma _v\) this requires is for example in contradiction with the topological concept of neighbourhoods which are constitutively woven into the idea of a manifold; cf [3, p 71, proposition 30] and footnote 1.

For discontinuously charting a geodesic (or any other curve) (27) is the instrument of choice. In \({\mathbb {S}}^n((\text {ord}({\mathbb {R}}))\supseteq \omega _1)\) surely exists a *s* which produces \(\gamma _v\) when successively multiplied by all \(\omega ^\chi \); see also Appendix C. For \(n=4\) all subsets of this *s* corresponding to a countable ordinal number and whose coefficients of the involved orders of magnitudes additionally satisfy (23) are eligible for physics. As only the coefficients of orders of magnitude associated with a limit ordinal may become reality (24), the coefficients of all other orders of magnitude, i.e. the virtual events, do not need to adhere to \(\gamma _v\). Thus all members of (27) with physical subsets whose real events are congruent, are affiliated to \(\gamma _v\) from the angle of physics.

## 5 Well-ordered spacetime and path integration

*dt*. Accordingly for the associated sum the transition into an integral is barred, even though \(\varDelta t\) is assumed to be already small enough to permit \(H(Q(t_k),P(t_{k-1})) \rightarrow H(q_k,p_{k-1})\). Now this first simple deduction only states a restriction. But starting out from the definition of the Riemann integral, there is another way to obtain the same result which casts some light on the full ordinal quality of

*N*:

*f*by simple functions \(f_n\) supports this claim:

*f*by a single indicator function, the limit must aim at least at \(\omega _1\). Otherwise, there will be not enough terms in the sum (33b) to exhaust the range of

*f*completely. In unison with this, \(2^n\) and \(2^{2n}\) should be read as \(2^{{\mathbb {N}}}\) and \(2^{2{\mathbb {N}}}\) to indicate the necessary cardinality of the process (33). Replacing them by an ordinal number \(\gamma \ge \omega _1\) converts the rhs of (33b) into the transfinite expression that it is at its heart.

^{4}See [7] for an axiomatic definition of terms like \(\frac{1}{\gamma }\).

As it was shown in Sect. 2, a well-ordered spacetime consists of representations of a countable ordinal number \(\alpha <\omega _1\). Hence, if (30) is evaluated over such a spacetime, *N* has to be strictly less than \(\omega _1\) and the associated sum in the exponential function cannot turn into an integral. For \(\omega<N<\omega _1\) see Sect. 3. Section 4 gives the details on the unavoidable violation of causality by bijections \(({\mathbb {R}},<^{\text {ord}({{\mathbb {R}}})}\supseteq \omega _1)\leftrightarrow ({\mathbb {R}},<)\).

As a matter of course, the sum over *a* in (30) cannot be made into an integral for the same reasons that apply to the sum over *k*. Furthermore, (30) should be adjusted to the use of a well-ordered spacetime by replacing \(\int [\prod _{k=1}^N\prod _adq_{k,a}]\) and \(\frac{1}{2\pi }\int [\prod _{k=1}^{N+1}\prod _adp_{k-1,a}]\) with \(\sum _{q\in {}^{\alpha }{\mathbb {R}}}\) and \(\sum _{p\in {}^{\alpha }{\mathbb {R}}}\) where *q* and *p* are order embeddings. In terms of surreal spacetime the sum over order embeddings *q* is replaced by \(\sum _{s\in {\mathbb {M}}{\alpha }}\) whereas the information contained in the *p* is now embodied in the according set of pairs \((r_0+\frac{r_\delta }{\omega ^{r^0_\delta }},\omega ^{r^0_\delta })\); cf (23) and the subsequent discussion.

## 6 Well-order and loop quantum gravity

^{5}spacetime will be endowed with more (sub)structure than it obtains form the natural numbers alone; see also Sect. 3 for more on this matter, especially the discussion below equation (24).

## Footnotes

- 1.
In Riemannian geometry \({\mathbb {R}}^n\) is more than a mere vehicle to locally handle a manifold. As stated right at the top of [3, p 1] a smooth manifold is resembling Euclidean space enough to permit the establishment of all essential features of calculus. This statement is realised in the following by defining local coordinate systems to be homeomorphisms, i.e. smooth manifolds and \({\mathbb {R}}^n\) are topologically equivalent. By this the standard topology of \({\mathbb {R}}^n\), whose base are open balls, and hence the dense order of \({\mathbb {R}}\), on which calculus hinges, is fundamentally introduced into smooth manifolds.

- 2.
\(\omega ^a\ll \omega ^b\) says that \(\omega ^b\) is of a higher order of magnitude than \(\omega ^a\); see [8, p 54].

- 3.
Inversely there are at best countable many pairs \(x<z\) of elements of \(({\mathbb {R}},<)\) which are mapped to \(f^{-1}(x)=\chi \in ({\mathbb {R}},<^{\text {ord}({\mathbb {R}})})\) and its direct successor \(f^{-1}(z)=\chi +1\).

- 4.
It is strange that in both Riemann integration and Lebesgue integration results already emerge even if only an initial subset of \(\omega _1\) as plain as the natural numbers \(\omega \subset \omega _1\) is considered (The usual interpretation of \(\lim _{n\rightarrow \infty }\) is to aim for \(\omega \); there is no hint that the use of limit ordinals larger than \(\omega \) and their internal structure is intended.). The condition of continuity does not immediately impose itself to be powerful enough to enforce this. Maybe there have been up to now only taken functions into account which by their definition are structurally plain beyond the requirements of continuity to allow this. Indeed, the author is not aware of any attempt to integrate the Weierstrass function. The same inner liveliness within the restrictions of continuity that prevents its differentiability should also prohibit (32) to yield results when being regarded for this function on \(\omega \) only.

- 5.
Since \(\lim _{N\rightarrow \infty }\) is used in [10] in the usual way, i.e. just unspecificly aiming at larger values and thus to \(\omega \), and moreover it is claimed that (41) exhibits a continuous result already after this most initial subset of \(\omega _1\), it may be assumed that (41) is an expression which is structurally as plain as the natural numbers; see footnote 4.

- 6.
There is one subspace of simultaneity at the origin stating to which events it is unrelated. This subspace is readily transferred to other elements by translation. Talking subsequently about a single unordered subspace refers to this blueprint at the origin.

## References

- 1.T. Jech,
*Set Theory*(Springer, Berlin, 2002)zbMATHGoogle Scholar - 2.K. Kuratowski, A. Mostowski,
*Set Theory*(North Holland Publishing Company, Amsterdam, 1976)zbMATHGoogle Scholar - 3.B. O’Neill,
*Semi-Riemannian Geometry*(Academic Press, San Diego, 1983)zbMATHGoogle Scholar - 4.E.C. Zeeman, Causality Implies the Lorentz Group. J. Math. Phys.
**5**, 490 (1964)ADSMathSciNetCrossRefGoogle Scholar - 5.G.L. Naber,
*The Geometry of Minkowski Spacetime*(Springer, New York, 2012)CrossRefGoogle Scholar - 6.Oliver Deiser,
*Einführung in die Mengenlehre*(Springer, Berlin, 2010)CrossRefGoogle Scholar - 7.J.H. Conway,
*On Numbers and Games*(A K Peters Ltd, Wellesley, 2001)zbMATHGoogle Scholar - 8.H. Gonshor,
*An Introduction to the Theory of Surreal Numbers*(Cambridge University Press, Cambridge, 1986)CrossRefGoogle Scholar - 9.Steven Weinberg,
*The Quantum Theory of Fields*, vol. 1 (Cambridge University Press, Cambridge, 2010)zbMATHGoogle Scholar - 10.Carlo Rovelli,
*Quantum Gravity*(Cambridge University Press, Cambridge, 2010)zbMATHGoogle Scholar

## Copyright information

**Open Access**This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Funded by SCOAP^{3}