# On the definition of the associated Legendre polynomials

Mathematica computes for n = 1,2,...:

(-1)^n (LegendreP[n, -1, -3]/Sqrt[2])

-I, -3 I, -11 I, -45 I, -197 I, ...


Maple computes for n = 1,2,...:

(-1)^n*LegendreP(n, -1, -3)/sqrt(2)

1, 3, 11, 45, 197, ...


Question: Is there a bug or just different definitions?

Edit: Which one conforms with the (standard) definition in the Handbook of Mathematical Functions respectively with the definition in DLMF?

Yes, they have different definitions:

• Maple

• Mathematica

• Wikipedia

$$P_\ell^m(x)=(-1)^m \left(1-x^2\right)^{m/2} \frac{\mathrm d^m}{\mathrm dx^m} P_\ell(x)$$

This difference produces different complex phases (different branches of the square root).

• @SophiaAntipolis It seems to me that DLMF corresponds to Mathematica: dlmf.nist.gov/14.6 – ybeltukov Sep 6 '14 at 16:33

Something ybeltukov forgot to mention: LegendreP[] takes a third optional parameter corresponding to the "type" of Legendre function needed. There are three types, all agreeing within the unit disk, but having different branch cut structures outside it. By default, the type 1 function is computed: LegendreP[n, m, z] == LegendreP[n, m, 1, z].

In particular, Maple's choice of type corresponds to type 3 in Mathematica's convention. Thus:

Table[(-1)^n LegendreP[n, -1, 3, -3]/Sqrt[2], {n, 5}]
{1, 3, 11, 45, 197}


P.S. Similar remarks hold for LegendreQ[].