1: \begin{abstract}

2:   Self-gravity computation by multipole expansion is a common approach

3:   in problems such as core-collapse and Type Ia supernovae, where

4:   single large condensations of mass must be treated.  The standard

5:   formulation of multipole self-gravity in arbitrary coordinate systems suffers from two significant

6:   sources of error, which we correct in the formulation presented in

7:   this article.  The first source of error is due to the numerical

8:   approximation that effectively places grid cell mass at the central

9:   point of the cell, then computes the gravitational potential at that

10:   point, resulting in a convergence failure of the multipole

11:   expansion. We describe a new scheme that avoids this problem by

12:   computing gravitational potential at cell faces. The second source

13:   of error is due to sub-optimal choice of location for the expansion

14:   center, which results in angular power at high multipole $l$ values

15:   in the gravitational field, requiring a high --- and expensive ---

16:   value of multipole cutoff \lmax.  By introducing a global measure of

17:   angular power in the gravitational field, we show that the optimal

18:   coordinate for the expansion is the square-density-weighted mean

19:   location.  We subject our new multipole self-gravity algorithm, implemented in the FLASH simulation framework, to

20:   two rigorous test problems: MacLaurin spheroids for which exact

21:   analytic solutions are known, and core-collapse supernovae.  We show

22:   that key observables of the core-collapse simulations, particularly

23:   shock expansion, proto-neutron star motion, and momentum

24:   conservation, are extremely sensitive to the accuracy of the

25:   multipole gravity, and the accuracy of their computation is greatly

26:   improved by our reformulated solver.

27: 

28: \keywords{supernovae: general -- hydrodynamics -- gravitation -- stars:

29:   interiors -- methods: numerical}

30: 

31: \end{abstract}

32: