Werner Heisenberg on Thermodynamics

“[I]n the nineteenth century, even the theory of heat could be reduced to mechanics by the assumption that heat really consists of a complicated statistical motion of the smallest parts of matter. By combining the concepts of the mathematical theory of probability with the concepts of Newtonian mechanics Clausius, Gibbs and Boltzmann were able to show that the fundamental laws in the theory of heat could be interpreted as statistical laws following from Newton's mechanics when applied to very complicated mechanical systems.”

“The second closed system of concepts was formed in the... nineteenth century... with the theory of heat. Though the theory... could finally be connected with mechanics through... statistical mechanics, it... [was not] a part of mechanics. ...[T]he phenomenological theory of heat uses... concepts that have no counterpart in other branches of physics, like: heat, specific heat, entropy, free energy, etc. If... one goes... to a statistical interpretation... considering heat as energy, distributed statistically among... many degrees of freedom due to... atomic structure of matter, then heat is no more connected with mechanics than with electrodynamics or other parts of physics. The central concept... is... probability, closely connected with the concept of entropy... [T]he statistical theory of heat requires the concept of energy. But any coherent set of axioms and concepts in physics will necessarily contain the concepts of energy, momentum and angular momentum and the law that these ...must under certain conditions be conserved. This follows if the coherent set is intended to describe... features of nature... correct at all times and everywhere... [i.e.,] features that do not depend on space and time... [i.e.,] that are invariant under arbitrary translations in space and time, rotations in space and the Galileo or Lorentz transformation. Therefore, the theory of heat can be combined with any of the other closed systems of concepts.”