5
$\begingroup$

Let $P_{n}(x)$ the $n-th$ Legendre polynomial. It is well-knonw that $$\int_{-1}^1 P_n(x) P_m(x) P_h(x) \, dx=2\left(\begin{array}{ccc} n & m & h\\ 0 & 0 & 0 \end{array}\right)^{2}\tag{1}$$ where $\left(\begin{array}{ccc} n & m & h\\ 0 & 0 & 0 \end{array}\right)$ is the Wigner 3-j symbol. I know that there exist some possible choices for the generalization of the Wigner 3-j symbol to complex numbers but my question is: is it possible to extend othe Wigner 3-j symbol to complex numbers such that type (1) equation holds? In other words, is it possible to find a generalization of the 3-j symbols such that $$\int_{-1}^1 P_a(x)P_b(x)P_c(x) \, dx = f(a,b,c) \left(\begin{array}{ccc} a & b & c\\ 0 & 0 & 0 \end{array}\right)^{2}$$ for a suitable function $f$ depending on $a,b,c\in\mathbb{C}$ and where $P_{a}(x)$ is the $C^{2}(-1,1)$ solution of the Legendre differential equation with parameter $a\in\mathbb{C}$, that is,

$$P_{a}(x)=\,_{2}F_{1}\left(a+1,-a;1;\frac{1-x}{2}\right)?$$

$\endgroup$
5
  • 1
    $\begingroup$ The 3j with all zero magnetic numbers has a closed form evaluation. This is Dixon's Theorem for 3F2 hypergeometric series, see math.stackexchange.com/questions/262973/… so analytically continuing the RHS is trivial. Whenever you see $a!$ just replace that by $\Gamma(a+1)$, in terms of the Euler Gamma function. The question you should look at instead is what could be a definition of $P_a$ for $a$ complex? Perhaps, use the contour integral extraction formula from the generating function one over distance. $\endgroup$ Commented Jul 12, 2023 at 16:30
  • $\begingroup$ I am not aware of the existence of "some non equivalent definitions" for 3j symbols. Would care to explain? $\endgroup$ Commented Jul 13, 2023 at 13:19
  • 1
    $\begingroup$ @AbdelmalekAbdesselam I'm referring to the work of Raynal, On the definition and properties of generalized 3-j symbols, J. of Mathematical Physics, 19, 467–476 (1978). There doesn't seem to be anything newer about it, but I could be wrong. Maybe I have to specify that I intend "non-equivalent definitions" for the analytic continuation of the 3j symbols. $\endgroup$
    – User
    Commented Jul 13, 2023 at 13:28
  • 1
    $\begingroup$ thanks for the ref to Raynal. Perhaps one can go around this because what you want here is not the analytic continuation of the general 3j with 6 free (or 5 with m conservation) parameters, but the continuation of the all m=0 case with 3 free parameters. There may be also a Carlson Theorem type argument which singles out a unique analytic continuation. $\endgroup$ Commented Jul 13, 2023 at 14:47
  • $\begingroup$ Thank you for the anonymous downvote without any comment, it surely help the community to improve itself. $\endgroup$
    – User
    Commented Jul 17, 2023 at 15:12

1 Answer 1

5
+150
$\begingroup$

Summary: this is a topic of active research [1,2,3,4]; the answer to the general case is not known; answers exist for $a\in\mathbb{C}$, $b,c,\in\mathbb{N}$, and for $a,b,c\in\mathbb{C}$ with $b=c$.


Case 1: Consider first the case that one of the three parameters $a,b,c$ is complex, while the other two are still integers. For the Legendre function I use the representation $$P_\mu(z)=\, _2F_1\left(-\mu,\mu+1;1;\frac{1-z}{2}\right),\;\;|1-z|<2,$$ and for the 3j symbol squared $$\left(\begin{array}{ccc} a & b & c\\ 0 & 0 & 0 \end{array}\right)^{2}=\frac{\binom{a+b-c}{\frac{1}{2} (a+b-c)} \binom{a-b+c}{\frac{1}{2} (a-b+c)} \binom{-a+b+c}{\frac{1}{2} (-a+b+c)}}{(a+b+c+1) \binom{a+b+c}{\frac{1}{2} (a+b+c)}}.$$ Then $$\int_{-1}^1 P_{a}(x)P_{b}(x)P_c(x)\,dx=2\cos ^2\left(\tfrac{1}{2} \pi (a+b+c)\right)\left(\begin{array}{ccc} a & b & c\\ 0 & 0 & 0 \end{array}\right)^{2},$$ $$\text{for}\;\;a\in\mathbb{C},b\in\mathbb{N},c\in\mathbb{N}.$$ (I checked this with Mathematica, which can evaluate the integral in closed form.)

So in this case of one complex variable the function $f(a,b,c)$ in the OP equals $2\cos ^2\left(\tfrac{1}{2} \pi (a+b+c)\right)$. Note that if all three variables are integers, the $\cos^2$ factor can be set to unity, since the 3j symbol equals zero for $a+b+c$ odd.

Case 2: If all three parameters are complex, a closed form expression in terms of generalized 3j symbols (or equivalently, generalized Clebsch-Gordan coefficients) exists for the integral $$T_{\mu,\nu}=\int_{-1}^1 P_{\mu}(x)P_{\nu}(x)P_\nu(-x)\,dx,\;\;\mu,\nu\in\mathbb{C}.$$ Note the sign change of one argument of the Legendre function. For integer $\nu=n$ one would simply have $P_n(-x)=(-1)^nP_n(x)$, but for noninteger $\nu$ the sign change is essential.

The result, derived by Yajun Zhou [1] (see also [2,3,4]) is given below, it is lengthy. A simple case is $\mu=\nu$, when we have $$\int_{-1}^1 P_{\nu}(x)P_{\nu}(x)P_\nu(-x)\,dx=\frac{1}{3} \bigl(2+3 \cos \pi \nu+\cos 2 \pi \nu\bigr)\left(\begin{array}{ccc} \nu & \nu & \nu\\ 0 & 0 & 0 \end{array}\right)^{2},\;\;\nu\in\mathbb{C}.$$ For integer $\nu$ the integral vanishes for $\nu$ odd, and for $\nu$ even the usual formula is recovered.

Here is the general result: \begin{align} T_{\mu,\nu}={}&\frac{2}{\pi^2}\frac{\sin(\mu\pi)\sin(\nu\pi)}{\mu(\mu+1)}\left[ \, \frac{1}{\nu}\,{_4F_3}\left(\left.\begin{array}{c} 1,\frac{1-\mu}{2},\frac{\mu+2 }{2},-\nu \\[4pt] \frac{2-\mu }{2},\frac{\mu+3 }{2},1-\nu \\ \end{array}\right| 1\right) -\frac{1}{\nu+1} \,{ _4F_3}\left(\left.\begin{array}{c} 1,\frac{1-\mu}{2},\frac{\mu+2 }{2},\nu+1\ \\[4pt] \frac{2-\mu }{2},\frac{\mu+3 }{2},\nu+2 \\ \end{array}\right| 1\right) \right]. \end{align} The combination of two hypergeometric functions ensures the symmetry $T_{\mu,\nu}=T_{-\mu-1,\nu}=T_{\mu,-\nu-1}$, required by $P_{\mu}(z)=P_{-\mu-1}(z)$.
  1. Yajun Zhou, Legendre Functions, Spherical Rotations, and Multiple Elliptic Integrals
  2. Marco Cantarini, A note on Clebsch-Gordan integral, Fourier-Legendre expansions and closed form for hypergeometric series
  3. John Campbell, New Clebsch-Gordan-type integrals involving threefold products of complete elliptic integrals
  4. Yajun Zhou, On Some Integrals Over the Product of Three Legendre Functions
$\endgroup$
3
  • $\begingroup$ Sorry, just a question: what is your definition of 3-j symbols when the arguments are complex numbers? $\endgroup$
    – User
    Commented Jul 18, 2023 at 6:53
  • $\begingroup$ see the second equation of my post; the binomial coefficients are written in terms of Gamma functions, which are then evaluated at complex arguments. $\endgroup$ Commented Jul 18, 2023 at 8:04
  • $\begingroup$ @Carlo Beenakker, In the following I am writing in the context of spherical harmonics as Legender polynomials are related to them. Are there references, where complex number generalisation of product of three monopole spherical harmonics discussed? Monopole spherical harmonics differ from the usual spherical harmonics. Usually, integration of product of three monopole harmonics can also be written in terms of the 3j-symbols, I am just wondering are there some complex number generalisation of those. $\endgroup$
    – Rakesh
    Commented Apr 5 at 11:47

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .