1: \begin{abstract}

2:   The numerical convergence of smoothed particle hydrodynamics (SPH) can be

3:   severely restricted by \corr{random force errors induced by} particle

4:   disorder, especially in shear flows\corr{, which are} ubiquitous in

5:   astrophysics. The increase in the number $\NH$ of neighbours when switching

6:   to more extended smoothing kernels \emph{at fixed resolution} (using an

7:   appropriate definition for the SPH resolution scale) is insufficient to

8:   combat these errors. Consequently, trading resolution for better convergence

9:   is necessary, but for traditional smoothing kernels this option is limited

10:   by the pairing (or clumping) instability. Therefore, we investigate the

11:   suitability of the Wendland functions as smoothing kernels and compare them

12:   with the traditional B-splines. Linear stability analysis in three

13:   dimensions and test simulations demonstrate that the Wendland kernels avoid

14:   the pairing instability for \emph{all} $\NH$, despite having vanishing

15:   derivative at the origin (disproving traditional ideas about the origin of

16:   this instability; instead, we uncover a relation with the kernel Fourier

17:   transform and give an explanation in terms of the SPH density

18:   estimator). The Wendland kernels are computationally more convenient than

19:   the higher-order B-splines, allowing large $\NH$ and hence better numerical

20:   convergence (note that computational costs rise sub-linear with $\NH$). Our

21:   analysis also shows that at low $\NH$ the quartic spline kernel with

22:   $\NH\approx60$ obtains much better convergence then the standard cubic

23:   spline.

24: \end{abstract}

25: