The Metaphysics of Time Reversal: Hutchison on Classical Mechanics (Q4940877)

The author starts his paper with Hutchison's question why nonconservative forces like friction are justifiably ignored when judging classical mechanical systems to be time-reversal invariant (TRI). According to the author, Hutchison in this sense believes to identify a genuine problem in mechanics being as much as a theory of phenomenal forces like friction or of more fundamental geometric forces, and hints that there is a real risk that the claim that classical mechanics is TRI is true by fiat alone. What kind of ``classical mechanics'' is considered here? Callender claims (with Feynman) that non-conservative forces in reality do not exist. Or, in view of Helmholtz' famous contribution, that our surrounding world is conservative in the sense of energy conservation: nothings gets lost anywhere. However, to make the mechanics ``practicable'', the first step of investigation is usually ``modelling'', i.e. an abstraction of the outer world onto an intellectual image to which the ``natural laws'' are applied and which is capable to answer the question(s) one actually has. A simple example may give more insight into the modelling process: one has sometimes to deal with differential equations with time-dependent coefficients. Where does this time dependence come from? It must have a source anywhere, or, mechanically spoken, there exist forces which cause this time behaviour. Going back to the root, one will immediately find that, in a wider description (more degrees of freedom (d.o.f.)), one part of the mechanical system acts on the remainder without feedback from the latter, such that the corresponding equation of motion can be solved separately, and the solution enters the actually considered part via time dependent coefficients. The basic assumption of no-feedback will, however, in general contradict Newton's basic law of interaction. Therefore, what is indeed considered here goes along with assuming (some of the) interactions to be negligable (like e.g. the gravitational influence of the Earth on the Sun). This consideration is totally identical to looking at a (reduced) mechanical system with ``phenomenal'' forces like friction and proves Callender's equivalence with a conservative (but consequently widened) system. Now, mechanical modelling (d. o. f. reduction) is the only way to make mechanical problems calculable (although it does not simplify mechanics at all). But is it adequate for philosophy? Probably not, because it generally goes along with some approximative assumptions. What are the ``natural laws''? Obviously the so-called Newton's second law: on page 334, the classical mechanics is represented by \(F=dp_i/dt = - \nabla_i V(z_i(t))\). Although not mentioned in the text, \(z_i\) ranges from 1 to 3 denoting the position of \(i\)-th particle in a three-dimensional space frame which is inertially fixed (because no Coriolis and centrifugal terms occur); here \(V\) is potential energy, and \(\nabla\) is nabla-operator (should read \(\nabla_i V(\sum_j (z_1,z_2,z_3)_j), j=1\cdots n\)). This takes into account possible interactions between all particles). No momentum of momentum arises here, no constraints enter the consideration. Therefore, along with the claim that non-conservative forces do not exist, classical mechanics here seems at a glance to be reduced to conservative particle dynamics abandoning the benefits of mechanical modelling, thus leaving any chance of solving equations of motion for any ``terrestrial'' problem (e.g. such (seemingly) simple devices like a double pendulum), but reducing considerations on a pure intellectual (philosophical?) reliable base at the atomic (spin-free) level to (trivially) assure TRI. However, the problem of ``natural laws'' is not yet answered. How do these look like? Nobody knows. In mechanics, their ``application'' is more the use of a theory, where the theory has to be understood in the sense of collected experience which is in agreement with the observable world. But does it hold for the non-observable atomistic world? This is a very old question which has (mostly) already been denied by physicists. Thus, once more, what kind of classical mechanics is considered here? There is obviously only one answer: classical mechanics is, in the present context, reduced to trivial conservative ``mass-point mechanics''. Once more: What are the ``laws of nature''? Obviously, see above, Newton's second law on the one hand. But do they also include what Hutchison calls the interconnections (here: the potential forces) -- or not? In other words: are ``dynamical laws'' and ``force laws'' treated as one unit or separately? Assuming the latter, the definition of TRI is (or seems) too weak because it does not yet describe the entire mechanical system but allows for any force capable to bring back the particle into its initial state. To give an example which is mentioned in the context: a particle attached to the ground by a spring can be forced go back to its initial state by removing the spring and applying any other but capable force. Then, the particle's state variables between final position and initial position may differ on its way back from those on its way forward. If the problem under consideration is (as I assume): what was the (unknown) initial state of the considered mechanical system with known actual state? Then this TRI definition would not be adequate. Although the potential forces need by no way to be linear, one should have in mind that linear systems must also fulfill the TRI condition. Once more the spring-mass system: for a linear spring the question of time reversal and initial condition is easily answered by the general solution \( x (t) = \Phi (t) x _o\), where \( x ^T = ( q ^T,p ^T)\) is state vector (position and momentum as used in the paper), \( x _o\) is (unknown) initial condition, and \(\Phi (t)\) is fundamental matrix showing the characteristic \( \Phi (t)^{-1} = \Phi (-t)\). Given the final state \(x _F\), one has \( x = \Phi (t) \Phi^{-1}(t_F) \Phi (t_F) x _o = \Phi (t) \Phi^{-1} (t_F) x _F = \Phi (-\tau) x _F\), where \(\tau = t_F-t\), leading to the initial condition \( x _o\) for \(t=0\). This calculation holds also for (linear) damping. Is it, therefore, necessary to generally exclude nonconservative forces from TRI considerations? Callender's arguments for time reversal are not easy to understand: ``why should the momentum \(p\) change sign for \(t\rightarrow -t\), and what does canceling of effects mean? A particle with a linear spring (spring coefficient \(c\)) may, for instance, have the momentum \(p=p_o\cos(\sqrt c/m t)\) with initial state \(q_o=0, p_o=m\dot q _o\) and will, for \(t\rightarrow -t\), not ``automatically change sign because they are functions of time. Likewise for all other functions of time.'' What is obviously meant here is: time reversal is achieved, if, for a given state, the positions are kept while the velocities (velocity vectors) change sign, leaving the laws of interaction (here potential forces) unchanged. Although many arguments are hard to understand, which may come from a lack in definitions (what is really considered?) and/or a mixup of arguments (nonconservative forces may be excluded on the atomic level but one is then definitely leaving classical mechanics in the common sense), the discussion is interesting. It has, however, not at all come to an end with this paper. Callender and Hutchison may continue their fight, for the amusement of the readers.