Mathematical logic and multiverses

The concepts of mathematical logic, introduced to explain Godel’s theorem, can also be exploited to shed further light on the question of multiverses in mathematical physics.

Recall that any physical theory whose domain extends to the entire universe, (i.e. any cosmological theory), has a multiverse associated with it: namely, the class of all models of that theory. Both complete and incomplete theories are capable of generating such multiverses. The class of models of a complete theory will be mutually non-isomorphic, but they will nevertheless be elementarily equivalent. Two models of a theory are defined to be elementarily equivalent if they share the same truth-values for all the sentences of the language. Whilst isomorphic models must be elementarily equivalent, there is no need for elementarily equivalent models to be isomorphic. Recalling that a complete theory T is one in which any sentence s, or its negation Not(s), belongs to the theory T, it follows that every model of a complete theory must be elementarily equivalent.

Alternatively, if a theory is such that there are sentences which are true in some models but not in others, then that theory must be incomplete. In this case, the models of the theory will be mutually non-isomorphic and elementarily inequivalent.

Hence, mathematical logic suggests that the application of mathematical physics to the universe as a whole can generate two different types of multiverse: classes of non-isomorphic but elementarily equivalent models; and classes of non-isomorphic and elementarily inequivalent models.

The question then arises: are there any conditions under which a theory has only one model, up to isomorphism? In other words, are there conditions under which a theory doesn’t generate a multiverse, and the problem of contingency (‘Why this universe and not some other?’) is eliminated?

A corollary of the Upward Lowenheim-Skolem theorem provides an answer to this. The latter entails that any theory which has a model of any infinite cardinality, will have models of all infinite cardinalities. Models of different cardinality obviously cannot be isomorphic, hence any theory, complete or incomplete, which has at least one model of infinite cardinality, will have a multiverse associated with. (In the case of a complete theory, the models of different cardinality will be elementarily equivalent, even if they are non-isomorphic). Needless to say, general relativity has models which employ the cardinality of the continuum, hence general relativity will possess models of every cardinality.

For a theory of mathematical physics to have only one possible model, it must have only a finite model. A Theory of Everything must have a unique finite model if the problem of contingency, and the potential existence of a multiverse is to be eliminated.

Published in: on July 25, 2009 at 10:57 am  Comments (1) Theories of everything and Godel’s theorem

Does Godel’s incompleteness theorem entail that the physicist’s dream of a Theory of Everything (ToE) is impossible? It’s a question which, curiously, has received scant attention in the philosophy of physics literature.

To understand the question, first we’ll need to introduce some concepts from mathematical logic: A theory T is a set of sentences, in some language, which is closed under logical implication. In other words, any sentence which can be derived from a subset of the sentences in a theory, is itself a sentence in the theory. A model M for a theory T is an interpretation of the variables, predicates, relations and operations of the langauge in which that theory is expressed, which renders each sentence in the theory as true. Theories generally have many different models: for example, each different vector space is a model for the theory of vector spaces, and each different group is a model for the theory of groups. Conversely, given any model, there is a theory Th(M) which consists of the sentences which are true in the model M.

Now, a theory T is defined to be complete if for any sentence s, either s or Not(s) belongs to T. A theory T is defined to be decidable if there is an effective procedure of deciding whether any given sentence s belongs to T, (where an ‘effective procedure’ is generally defined to be a finitely-specifiable sequence of algorithmic steps). A theory is axiomatizable if there is a decidable set of sentences in the theory, whose closure under logical implication equals the entire theory.

It transpires that the theory of arithmetic (technically, Peano arithmetic) is both incomplete and undecidable. Moreover, whilst Peano arithmetic is axiomatizable, there is a particular model of Peano arithmetic, whose theory is typically referred to as Number theory, which Godel demonstrated to be undecidable and non-axiomatizable. Godel obtained sentences s, which are true in the model, but which cannot be proven from the theory of the model. These sentences are of the self-referential form, s = ‘I am not provable from A’, where A is a subset of sentences in the theory.

Any theory which includes Peano arithmetic will be incomplete, hence if a final Theory of Everything includes Peano arithmetic, then the final theory will also be incomplete. The use of Peano arithmetic is fairly pervasive in mathematical physics, hence, at first sight, this appears to be highly damaging to the prospects for a final Theory of Everything in physics.

In some mitigation, for the application of mathematics to the physical world one’s conscience may be fairly untroubled by the difficulties of self-referential statements. However, undecidable statements which are free from self-reference have been found in various branches of mathematics. For example, it has been proven that there is no general means of proving whether or not a pair of ‘triangulated’ 4-dimensional manifolds are homeomorphic (topologically identical).

Crucially, however, whilst the theory of a model, Th(M), may be undecidable, it is guaranteed to be complete, and it is the models of a theory which purport to represent physical reality. A final Theory of Everything might have no need of Peano arithmetic, and might well be complete and decidable. However, even if a final Theory of Everything is incomplete and undecidable, the physical universe will be a model M of that theory, and every sentence in the language of the theory will either belong or not belong to Th(M). 