1: \begin{abstract} 

2:  We establish large deviation principles (LDPs) for empirical 

3:   measures associated with a sequence of Gibbs distributions

4: %  \footnote{Do we know that these are still Gibbs distributions when

5: %  the potentials are singular?} 

6: on $n$-particle configurations, each of which is defined in terms of an inverse

7: temperature $\beta_n$ and an energy functional that is the sum of a 

8: (possibly singular) interaction and confining potential. 

9:  Under fairly 

10: general  assumptions on the potentials, we establish LDPs both with speeds 

11: $\beta_n/n \rightarrow \infty$, in which case the rate function is

12: expressed in terms of a functional involving the potentials, and with  

13: the speed  $\beta_n =n$,  when the rate function contains an additional

14:  entropic term.    

15: Such LDPs are motivated by questions arising in random matrix theory,

16: sampling and simulated annealing.   

17: Our approach, which uses the weak convergence 

18: methods developed in \cite{Dupuis1997a}, establishes large deviation principles

19: with respect to stronger, Wasserstein-type  topologies, 

20: thus resolving an open question in \cite{DjalilChafaiNathaelGozlan}. 

21: It also provides a 

22: common framework for the analysis of LDPs with all speeds, and includes cases not covered due to technical reasons in previous works such as    

23: \cite{Arous1998,DjalilChafaiNathaelGozlan}.

24:  \end{abstract}