1: \begin{abstract}

2: Boundary integral equations and Nystr\"om discretization provide a powerful

3: tool for the solution of Laplace and Helmholtz boundary value problems.

4: %

5: However, often a weakly-singular kernel arises, in which case

6: specialized quadratures that modify the matrix entries near the diagonal

7: are needed to reach a high accuracy.

8: %

9: We describe the construction of four different quadratures

10: which handle logarithmically-singular kernels.

11: %

12: Only smooth boundaries are considered, but some of the techniques

13: extend straightforwardly to the case of corners.

14: %

15: Three are modifications of the global periodic trapezoid rule,

16: due to Kapur--Rokhlin, to Alpert, and to Kress.

17: %

18: The fourth is a modification to a quadrature based on

19: Gauss-Legendre panels due to Kolm--Rokhlin;

20: this formulation allows adaptivity. %and the handling of corners simply.

21: %

22: We compare in numerical experiments the convergence of the four schemes in

23: various settings, including low- and high-frequency planar Helmholtz problems,

24: and

25: 3D axisymmetric Laplace problems.

26: %

27: We also find striking differences in performance in an iterative

28: setting.

29: %

30: We summarize the relative advantages of the schemes.

31: %compare their performance in an iterative setting, and

32: %discuss their relative advantages.

33: \end{abstract}

34: