1: \begin{abstract}  

2: If an absolute reference frame with respect to time, position, or 

3: orientation

4: is missing one can only implement

5: quantum operations which are covariant with respect 

6: to the corresponding unitary symmetry group $G$. 

7: Extending observations of Vaccaro et al., I argue that

8: the free energy of a quantum system with $G$-invariant Hamiltonian

9: then splits up into the Holevo information

10: of the orbit of the state under 

11: the action of $G$  and the free energy of its orbit average.

12: These two kinds of free energy cannot be converted into each other. 

13: The first component is subadditive and the second superadditive; 

14: in the limit of infinitely many copies only the usual

15: free energy matters. 

16: 

17: Refined splittings of free energy 

18: into more than two independent (non-increasing) terms can be defined by averaging

19: over probability measures on $G$ that differ from the Haar measure. 

20: 

21: Even in the presence of a reference frame,

22: these results provide lower bounds  on the amount of 

23: free energy that is lost after applying 

24: a covariant channel. 

25: If  the channel 

26: properly decreases 

27: one  of these quantities, it decreases the free energy 

28: necessarily at least by the same amount, since

29: it is unable to convert

30: the different forms of free energies into each other. 

31: \end{abstract}

32: