1: \begin{abstract}

2: 

3: We study the convergence properties of smoothed particle hydrodynamics

4: (SPH) using numerical tests and simple analytic considerations.  Our

5: analysis shows that formal numerical convergence is possible in SPH

6: only in the joint limit $N \rightarrow \infty$, $h \rightarrow 0$, and

7: $N_{nb} \rightarrow \infty$, where $N$ is the total number of

8: particles, $h$ is the smoothing length, and $N_{nb}$ is the number of

9: neighbor particles within the smoothing volume used to compute

10: smoothed estimates.  Previous work has generally assumed that the

11: conditions $N \rightarrow \infty$ and $h \rightarrow 0$ are sufficient

12: to achieve convergence, while holding $N_{nb}$ fixed.  We demonstrate

13: that if $N_{nb}$ is held fixed as the resolution is increased, there

14: will be a residual source of error that does not vanish as $N

15: \rightarrow \infty$ and $h \rightarrow 0$.  Formal numerical

16: convergence in SPH is possible only if $N_{nb}$ is increased

17: systematically as the resolution is improved.  Using analytic

18: arguments, we derive an optimal compromise scaling for $N_{nb}$ by

19: requiring that this source of error balance that present in the

20: smoothing procedure.  For typical choices of the smoothing kernel, we

21: find $N_{nb} \propto N^{0.5}$.  This means that if SPH is to be used

22: as a numerically convergent method, the required computational

23: cost does not scale with particle

24: number as $O(N)$, but rather as $O(N^{1+\delta})$, where $\delta

25: \approx 0.5$, with a weak dependence on the form of the smoothing

26: kernel.

27: 

28: \end{abstract}

29: