1: \begin{abstract}

2: We develop an efficient sampling and free energy calculation technique

3: within the adaptive biasing potential (ABP) framework.  By mollifying

4: the density of states we obtain an approximate free energy and an

5: adaptive bias potential that is computed directly from the population

6: along the coordinates of the free energy. Because of the mollifier,

7: the bias potential is ``nonlocal'' and its gradient admits a simple 

8: analytic expression. 

9: A single observation of the reaction

10: coordinate can thus be used to update the approximate free energy at

11: every point within a neighborhood of the observation.  This greatly

12: reduces the equilibration time of the adaptive bias potential.  This

13: approximation introduces two parameters: strength of mollification and

14: the zero of energy of the bias potential.  While we observe that the

15: approximate free energy is a very good estimate of the actual free

16: energy for a large range of mollification strength, we demonstrate that the

17: errors associated with the mollification may be removed via

18: deconvolution.  The zero of energy of the bias potential, which is

19: easy to choose, influences the speed of convergence but not the

20: limiting accuracy.  This method is simple to apply to free energy or

21: mean force computation in multiple dimensions and does not involve

22: second derivatives of the reaction coordinates, matrix manipulations

23: nor on-the-fly adaptation of parameters.  For the alanine dipeptide

24: test case, the new method is found to gain as much as a factor of ten in

25: efficiency as compared to two common adaptive biasing force

26: formulations and it is shown to be as efficient as well-tempered metadynamics 

27: with the post-process deconvolution giving a clear advantage to the mollified density of states method. 

28: \end{abstract}

29: