Math 2.0

Numerical analysis Definition

Fast multipole method

Builds on Multipole expansionFundamental solution of the Laplacian

Definition

Let x1,…,xNx_1, \dots, x_N be points in ℝd\R^d with d=2d = 2 or 33, and let q1,…,qNq_1, \dots, q_N be real numbers. Let GG be a kernel such as the fundamental solution of the Laplacian. The task is to compute the potentials ϕi=∑j≠iqjG(xi,xj)\phi_i = \sum_{j \ne i} q_j G(x_i, x_j) for i=1,…,Ni = 1, \dots, N. The fast multipole method with expansion order pp computes approximations of all ϕi\phi_i as follows.

  1. Enclose the points in a box. Subdivide it recursively into 2d2^d children, giving a tree, until each leaf box holds a bounded number of points.
  2. For each leaf, form the multipole expansion of its points truncated at order pp. Going up the tree, translate the expansions of the children to the centre of the parent and add them.
  3. For each box, convert the multipole expansions of the well-separated boxes in its interaction list into one local expansion about its centre. Add to it the local expansion inherited from its parent.
  4. At each leaf, evaluate the local expansion at its points, and add the interactions with points in the neighbouring leaves directly.

Remarks

Example: in the plane, identify ℝ2\R^2 with ℂ\C and take G(z,w)=log⁡∣z−w∣G(z, w) = \log \abs{z - w}. For points zjz_j in a disc of centre cc and radius rr, and for ∣z−c∣>r\abs{z - c} > r, ∑jqjlog⁡(z−zj)=Qlog⁡(z−c)−∑k≥11k∑jqj(zj−c)k(z−c)k,Q=∑jqj.\sum_j q_j \log(z - z_j) = Q \log(z - c) - \sum_{k \ge 1} \frac{1}{k} \sum_j q_j \frac{(z_j - c)^k}{(z - c)^k}, \qquad Q = \sum_j q_j. This follows from log⁡(z−zj)=log⁡(z−c)+log⁡(1−zj−cz−c)\log(z - z_j) = \log(z - c) + \log\big(1 - \frac{z_j - c}{z - c}\big) and the series of log⁡(1−u)\log(1 - u). The potential is the real part, and keeping pp terms gives an error of order (r/∣z−c∣)p(r / \abs{z - c})^p.

Non-example: the Barnes–Hut tree code also groups far points by boxes, but it has no step 3: each point is evaluated against far boxes one by one. Its cost is O(Nlog⁡N)O(N \log N), not O(N)O(N).

For a fixed order pp and points that are not too unevenly spread, the cost of the method is O(N)O(N), and the error decreases geometrically as pp grows.

Reading path

To understand this page from the ground up, read these 19 pages in order. The foundations come first.

Show the path as a graph
Each page uses the pages above it. The colour of the bar on a box shows how strong the evidence is, and the box also gives it as text. The list has the same pages.
  • no evidence level
  1. Euclidean space object
  2. Function concept
  3. Fundamental solution concept
  4. Multipole expansion concept
  5. Total derivative concept
  6. Function of several variables concept
  7. Vector-valued function concept
  8. Laplace transform definition
  9. Differentiable function of several variables concept
  10. Partial derivative concept
  11. Vector field concept
  12. Heaviside step function object
  13. Gradient concept
  14. Divergence concept
  15. Dirac delta function object
  16. Laplacian concept
  17. Harmonic function concept
  18. Fundamental solution of the Laplacian definition
  19. Fast multipole method definition

Further reading

  1. Web https://en.wikipedia.org/wiki/Fast_multipole_method
About this page
TagB4Z227Q9PN
URIm2://B4Z227Q9PN
Kinddefinition
WritingW0 stub
MSC65Y20, 78M16, 70F10
DefinesFast multipole method
UsesMultipole expansion, Fundamental solution of the Laplacian
Changed3 November 2026

Markdown · JSON · Edit on GitHub

What this page needs for W1

  • an inventory row: the row of the inventory that the node accounts for (0 of 1)
  • a source with a locator (0 of 1)

The writing levels say how complete the text of a page is. They do not measure its truth.

What links here

No page links to this page yet.