Do Gödel's incompleteness theorems have any significance or application to axiomatic theories of classical mechanics like Newton's for example?
-
I don't know, but the only "undecidable statements" I've ever heard of were self-referential. That is, a statement expressed in the language of some formal system, that describes a statement expressed in the same formal system, that happens to be itself. The whole point of classical mechanics is to describe the physical world. Would it even make sense to talk about a statement of classical mechanics that describes a statement of classical mechanics instead of describing some physical system? – Solomon Slow Feb 07 '20 at 15:39
-
3@SolomonSlow - The trick is to encode the classical mechanics statement plus classical mechanics in a formal system that is then implemented using a classical mechanics system. A very contrived situation, surely, but totally allowed. – Anders Sandberg Feb 07 '20 at 15:45
-
1Related: https://physics.stackexchange.com/q/14939/2451 and links therein. – Qmechanic Feb 07 '20 at 16:44
-
1Backreaction, Monday, February 03, 2020 Guest Post: “Undecidability, Uncomputability and the Unity of Physics. Part 1.” by Tim Palmer – Keith McClary Feb 07 '20 at 22:44
-
Let $p$ denote a physical theorem, and $q$ a mathematical statement that remains undecidable even given all physics. Then $p\land q$ would qualify, but you probably want a more interesting example. So what are the rules here? – J.G. Feb 08 '20 at 22:36
-
1https://physics.stackexchange.com/questions/403574/what-situations-in-classical-physics-are-non-deterministic is related. – Daniel Darabos Feb 08 '20 at 23:47
-
That's a great question but too bad it had a poor answer. – Kugutsu-o Feb 09 '20 at 20:28
-
Related https://physics.stackexchange.com/q/87239/226902 https://physics.stackexchange.com/q/94560/226902 – Quillo Jun 30 '23 at 07:27
4 Answers
Yes. One simple (trivial) way of seeing it is to consider a mechanical computer, and ask whether one can predict its end-state given the initial state without running it. Since this is exactly the Turing halting problem, any method that would allow you to make that prediction in Newtonian mechanics would either be a counterexample to the theorem (contradiction), or require some uncomputable element.
This should not be surprising. Chaotic dynamical systems lack analytic solutions and are often undecidable. This appears to be true (one needs to define deciding carefully here) for the N-body problem (see also this dissertation).
Does this matter? Probably less to physicists and more the philosophers and computer scientists. The issue of how computation is embodied in the physical world is a big topic, with interesting questions about whether hypercomputation is possible for real (it is allowed in Newtonian mechanics - one can in principle solve the halting problem using a Newtonian configuration) and whether physical laws are prognostic.

- 32,067
-
4I am not sure if the "undecidability" in Gödel, and in Turing means the same. – peterh Feb 08 '20 at 01:43
-
The Turing halting problem just says that you can't make a Turing machine that can tell you, for all Turing machines, whether they halt. It doesn't say that, for a particular Turing machine, we can't say whether it halts. – Acccumulation Feb 08 '20 at 01:44
-
Yes, thus there is no Turing-machine what could compute the result of all mechanical machines [or say it, if there is no result, because the mechanical machine never halts]. But does it mean that the NM is undecidable? In the mathematical logic, a statement is undecidable, iff it can't be proven and can't be refuted in the axiom system in which it is formulated. If it is equivalent that no TM can solve it in the general case, that is for me non-trivial. – peterh Feb 08 '20 at 01:49
-
1@peterh-ReinstateMonica I believe the Curry-Howard correspondence implies they are the same. Given a mathematical proposition, build a Turing machine that outputs 1 if it is a theorem and 0 if it is a non-theorem. If it is undecidable, then it cannot output anything, and thus, it must not halt. Not rigorous, but I think that's the gist of it.
https://en.wikipedia.org/wiki/Curry%E2%80%93Howard_correspondence
– Lawnmower Man Feb 08 '20 at 03:52 -
7I cannot agree with this answer since the Halting problem is intimately tied to the unbounded nature of Turing machines, whereas any physical computer would have to be bounded and therefore be computationally equivalent to a finite automaton, for which halting can be determined. – kviiri Feb 08 '20 at 09:39
-
@kviiri I think the idealistic "physical computer", in the frame of this post, must be finite, but it can be any big. ("finite but limit-less"? I am not sure this is the correct terminology) – peterh Feb 08 '20 at 13:19
-
1
-
@peterh-ReinstateMonica Any finite computer, no matter how big, is "immune" to the halting problem, so to say, because it will have a finite amount of states and therefore will either halt or visit one of these states more than once eventually, never halting. – kviiri Feb 08 '20 at 14:00
-
It's a nice try but I agree with @kviiri. Any physical computer must be made of $ N$ particles with $N<10^{200}$ say. The Halting problem for any such computer is decidable. – lcv Feb 08 '20 at 14:02
-
@JiK I'm not well-versed enough with classical mechanics to say whether its usual formulations permit objects of infinite size, which'd be required for a truly general Turing machine. However, in case Anders wants to argue based on that, at least they should point out they're explicitly talking about a very special mechanical computer with highly unusual properties. – kviiri Feb 08 '20 at 14:04
-
6Note that the question was about Newtonian mechanics, not about standard physics. Infinite numbers of particles and arbitrarily high momenta are definitely allowed in the formalism. – Anders Sandberg Feb 08 '20 at 15:41
-
-
@lcv Any proof that works out the decidability for that specific physical compute may well take more than $10^{200}$ particles to write it down ... – Hagen von Eitzen Feb 08 '20 at 18:15
-
Wait so basically undecidable questions always are only about Infinities? – Kugutsu-o Feb 08 '20 at 18:28
-
1@Ezio - An undecidable question cannot be answered using finite computational means. Note that it does not have to be about an infinite system: one can set up a N-body system for some finite N and ask whether it will always stay bounded or particles will eventually shoot away, and that question can be undecidable. – Anders Sandberg Feb 08 '20 at 23:43
-
So yes it's enough to show Turing undecidability since you are working in a system with an enumerable set of axioms so if you could prove for every Turing machine which doesn't half that it doesn't half you would have enumerated all the non-halting machines and (as you can easily enumerate halting machines) u would have a computable solution to halting problem (just see which enumeration your machine shows up on). – Peter Gerdes Oct 16 '21 at 13:16
-
However, I think a full proof of this would require picking a particular axiomitazation and language to work in and verifying you really can write down a finite formula in that language which corresponds to the Turing machine halting. For instance, is it even possible to exactly model a Turing machine in classical mechanics? You need to check it's not merely that you can find a config which will compute first 10^20th steps right but all of them. I'd worry that u could only represent better and better T machine approximations. Probably not but would need to be checked. – Peter Gerdes Oct 16 '21 at 13:20
The theorem does have consequences for any formal system that includes arithmetic. There will always be theorems that will be undecidable. Those theorems though, always involve infinite quantities, one example is: Do this system of Diophantine equations have a finite or infinite number of solutions?". The problem is in general undecidable, although for particular examples you might find it decidable. Other more interesting examples can be found here, and also in @Anders Sanberg's answer

- 56,248
I would like to share two examples that one could consider a "physics version" of Gödel's incompleteness theorem. This will not really answer the question about whether axiom systems in physics can be formally incomplete in the Gödel sense. However, it will show how an incompleteness of a physical theory is usually connected to some physical insight that is not captured by the axiom system, and that a mathematical incompleteness would therefore simply indicate that our theory does not describe all the physics. (This argument assumes determinism, but since the question is about classical mechanics I think that is fair).
The first example is Newton's cradle, or a simplified version where three perfectly incompressible identical spheres are lined up perfectly in space and collide along a line. Pictorially (o = sphere, o> = is moving sphere, -=spatial separation):
--o>---o---o---
From an axiomatic perspective, we have Newton's laws (mainly momentum conservation relevant here) and the perfectly incompressible sphere assumption (which gives energy conservation).
For two balls (--o>----o--), the problem is solved by momentum and energy conservation and one finds that the first ball stops while the second ball takes over the momentum of the first one. There is no problem here.
For three balls (--o>----o----o---), the problem is simply a succession of two ball collision.
The problem becomes interesting when two of the balls touch at the start (--o>---oo----). When setting up momentum and energy conservation for this system, one finds that the equations have multiple physical solutions, with the two successive two-ball collision solution only being one of them. In some sense, the problem is undecidable within our "axiom system".
What is happening here? The answer is very simple: The perfectly incompressible sphere assumption does not make much sense when two balls touch. In reality, even if the balls are in a limiting sense perfectly incompressible, there will be show waves travelling through the material and cause a unique trajectory to be realized physically. Our axioms do not capture this situation.
Another set of famous examples are various paradoxes in special relativity (e.g. the ladder paradox), which in the end require the concept of a perfect rigid body to be abandoned.
The Newton's cradle example is undecidable in a physics sense. In a mathematical sense, this has, however, nothing to do with Gödel's incompleteness theorem. The special relativity paradoxes are an example of axioms that physically do not go together.

- 11,535
-
2Nice example (+1). Although it seems that all apparently Goedel-like, undecidable, situations in physics, become decidable when one properly insert back physical limitations (like, no perfect elastic body, no perfect rigid,..) – lcv Feb 08 '20 at 13:57
-
4I do not think that "having multiple solutions" is in any sense the same as the problem being undecidable. In the case of multiple solutions we have solved the problem, it's just that its solution is not unique. In the case of undecidability we have proven that it is impossible to arrive at a general solution at all.. – ACuriousMind Feb 08 '20 at 16:48
-
@ACuriousMind Depends on what you mean by "the problem". If you want to know what happens physically, the considered axiom system does not answer the question. I am aware that this is by no means the same as undecidability in Gödel's sense and I specified that in my answer. My point is that investigating Gödel theorems for physics theories is only useful in a limited sense, since an incompleteness is inherently unphysical. Any suggestions to make that clearer in my answer? – Wolpertinger Feb 08 '20 at 19:54
If you are being strict about the kinds of things Gödel was talking about, no. Godel was operating on systems which could prove all true statement in arithmetic. In physics these are generally statements that are assumed rather than proven. Physics builds on top of math, building upon what mathematics asserts to be true.
To start off, you would need to find something quantized in classical mechanics. Most classical mechanics topics deal with real numbers, and thus most meaningful statements are proven using Second Order Logic(SOL). Gödel demonstrated that SOL systems cannot admit a proof of their own consistency, but beyond that he has little to say about such systems.
You woudl also need to find a quantized self-referential system, and then one which admits something akin to a proof of its own consistency.
Indeed, true self reference itself is quite rare in physics, and a bit of a novelty. But you may be able to find something to sink your teeth into in the world of trying to model how the brain thinks. But the more we learn about the brain, the more we learn how little of it relies on something as brittle as logic.

- 48,357
-
That's just not really true. Incompleteness theorem applies for any formal system that contains arithmetic. Ultimately whatever the deductive system you use in physics you should be able to find undecidable statements. – Kugutsu-o Feb 09 '20 at 09:11
-
I think any formal system that contains arithmetic is self referential by default – Kugutsu-o Feb 09 '20 at 09:23
-
@Ezio I believe that true arithmetic is a formal system which is complete and consistent. It, of course, its not effectively generated. This does bring up a funny quirk of the question. If the systems physicists look at do not strive to meet the very specific requirements on the systems Godel was making proofs about, does that mean they are affected or unaffected by his theorems? – Cort Ammon Feb 09 '20 at 15:06
-
I don't think it's a matter of belief. Im not sure how affected physics axiomatic systems are this is the question. – Kugutsu-o Feb 09 '20 at 19:31
-
@Ezio Well, at the heart of physics is the fundamental philosophical question of science. Are we discovering fundamental laws of physics, or are we identifying patterns we see everywhere we look? The latter is completely immune to Godelian thinking in every way, because it isn't seeking to prove things. – Cort Ammon Feb 11 '20 at 03:31
-
Look I for one don't subscribe to artificial distinctions between math physics and philosophy. To understand natural philosophy is to understand the nature of reality and that is the same thing as understanding mathematics. In a of these fields we are looking for the common( fundamental) patterns =laws that are there in some form regardless of scales or particular areas. – Kugutsu-o Feb 11 '20 at 09:47
-
@Ezio Then, using those words, physics is not a field which permits us to identify natural laws, merely things which appear to our limited faculties as natural laws and can be tested as such. Most of the time, this nuance is of limited importance, but when we are getting into theoretical limits like those Godel proved. We have not proven the existence of all integers in the real world, merely found systems which suggest numbers so large that we can believe that they reach onwards forever. – Cort Ammon Feb 11 '20 at 14:41
-
Natural laws might come "from beyond just physics", from philosophy or some elegant set of math rules that happens to describe all interactions in the concrete real world. I don't even know how would one precisely distinguish between the platonic realm of math and this real world and which one has more" real" existence. What godel essentialy proved is that provability is limited which is kind of obvious since one always has to assume something to prove anything. So there will always be something outside the logical or physical ability to prove or measure both in real and platonic worlds. – Kugutsu-o Feb 11 '20 at 15:14
-
One needs to distinguish two things. The systems Godel (well Godel, Rosser etc) proved to contain formally undecidable statements (those which extend certain weak systems of arithmetic) and those axiom systems whose decision problem (can u computable decide if phi is provable for any phi) is decidable. There are many systems that don't extend arithmetic that still aren't decidable. All this stuff about QM and nature of physics isn't really relevant as the Q is only well defined relative to a specific set of axioms (undecidability can depend on how u represent system) – Peter Gerdes Oct 16 '21 at 13:26