# Programming Multipolar Expansions in Spherical Tensors

We have that a density matrix can be written in a basis of spherical tensors:

$\rho =\sum _{K=0}^S\sum_{q=-K}^K \rho _{\text{Kq}}^{(S)} T_{\text{Kq}}^{(S)}$

for example for a 3x3 matrix, $S=1$, dimension of the matrix is $2S+1$, I will drop superscript (S) for simplicity

$\rho =\rho_{00}T_{00}+\rho_{1-1}T_{1-1}+\rho_{10}T_{10}+\rho_{11}T_{11}+\rho_{2-2}T_{2-2}+\rho_{2-1}T_{2-1}+\rho_{20}T_{20}+\rho_{21}T_{21}+\rho_{21}T_{21}$

Now, I have a command that gives me the matrix representation of $T_{Kq}^{(S)}$

I want to know now how to make a program so that if you give as an imput the parameter S, they write you the resultant matrix dependent of the coefficients $\rho_{Kq}$. Also as $\rho_{Kq}=(-1)^q\rho_{Kq}^*$, this should also be implemented so that the final matrix depends on fewer parameters.

Edit: Answering to Hector below. I have not programmed the command which gives $T_{Kq}^{(S)}$, to say the truth I have not made code in Mathematica never, and hardly ever elsewhere. Well command is sphtens[K,q,S]

I've been thinking about the algorithm to get the desired result. To me the step are the following:

• Input the value of S
• Create a square matrix of zeros of dimension 2S+1.

Call it $\rho$ for example.

• Create two loops one for the K and another one for the q

for $K=0$ till $2S$

for $q=-K$ till $K$

if $q\geq 0$

$\rho=\rho + \rho_{Kq}$ \times sphtens[K,q,S]

if not $\rho=\rho + (-1)^q\rho_{Kq}^* \times$ sphtens[K,q,S]

end for

end for

Basically what I don't know is to write this in Wolfram language

-
After "I have a command that gives me the matrix representation of …" is a good place to put the code you have been playing with. –  Hector Nov 21 '13 at 16:44
I edited to give an answer to your suggerence, I guess someone that has some level of programmimg with mathematica can built the program/function with some ease from my algorithm. –  user10712 Nov 21 '13 at 18:40
don't know is to write this in mathematica language, excuse me but it is the Wolfram Language or WL for short. Please see wolfram.com/wolfram-language that is the official name of the language from now on. –  Nasser Nov 21 '13 at 18:46
Edited, Wolfram name everything after his own name, a bit weird. –  user10712 Nov 22 '13 at 14:34

To transform back from spherical tensors to Cartesian tensors, the unitary transform is

\mathbf{U}_l = \frac{1}{\sqrt{2}}\left\{ \begin{align} Y^{-m}_l + (-1)^m Y^m_l &,\ m >0\\ \sqrt{2} Y^0_l &,\ m=0\\ i(Y^{-m}_l - (-1)^m Y^m_l) &,\ m<0 \end{align}\right.

where $m$ runs from $-l$ to $l$. This can be built in many ways, usually involving Table, but here is an implementation I used.

Clear[ExchangeMatrix, AntiDiagonalMatrix, RealHarmonics];
ExchangeMatrix[a_Integer?Positive]:= Permutations[ConstantArray[0, a-1]~Join~{1}]
AntiDiagonalMatrix[vec_?VectorQ, k_Integer:0]:=
DiagonalMatrix[vec,k].ExchangeMatrix[Length@vec+Abs[k]]

RealHarmonics[l_Integer?NonNegative]:=
Block[{harmmat,ord},
harmmat =(DiagonalMatrix[
Table[(-1)^m Which[m<0,-I, m==0, 1/Sqrt[2],m>0, 1], {m, l,-l,-1}]]+
+ AntiDiagonalMatrix[
Table[Which[m<0,I, m==0, 1/Sqrt[2],m>0, 1], {m, l,-l,-1}]
])/Sqrt[2];

(* This moves the z^n terms to the end, thus maintaining x, y, z order *)
harmmat[[Drop[Range[1,2l+1], {l+1}]~Append~(l+1)]]
]


Note, ExchangeMatrix and AntiDiagonalMatrix are part of a larger package of matrix utilities I have put together.

To use RealHarmonics, just enter the angular quantum number, e.g.

RealHarmonics[1]
(* {{-(1/Sqrt[2]), 0, 1/Sqrt[2]}, {I/Sqrt[2], 0, I/Sqrt[2]}, {0, 1, 0}} *)


Then, to transform to spherical tensors, you take the conjugate transpose,

ConjugateTranspose@RealHarmonics[1]
(*{{-(1/Sqrt[2]), -(I/Sqrt[2]), 0}, {0, 0, 1}, {1/Sqrt[2], -(I/Sqrt[2]), 0}}*)

-