We begin by constructing all possible higher-order and higher-rank tensor operators from powers of the derivative , where are the spin-projection components of the vector (rank-1) operator . It is obvious that all possible th-order powers of the derivative can be written as sums of terms . Therefore, any th-order power is simply obtained by multiplying some th-order power by a sum of operators. Then, powers of a given rank can be obtained iteratively by vector coupling.
In the second order, the two nabla operators can be coupled to angular momenta 0 and 2. The coupling to angular momentum 0, , corresponds to the Laplacian operator. Furthermore, the coupling to angular momentum 2, , gives the second-order, rank-2 derivative operator. The rank-1 coupling, , vanishes because the derivatives commute. Similarly, in each one higher order, a rank- symmetric operator can be coupled with only to and . Hence, all the th-order powers have the form of multiplied by the th-order rank- (stretched) coupled operators for . Then, up to NLO, one obtains 16 different operators listed in Table 1. Any arbitrary tensor formed by coupled operators can always be rewritten as a sum of operators through the repeated use of the symbols.
Exactly in the same way, we define 16 different operators , which are spherical tensors built of the relative momentum operators coupled up to NLO, i.e., for and . In the remainder of this section, we only discuss operators , while all the results mutatis mutandis also pertain to operators .
The stretched coupled operators for ,
(11) |
Equivalently, derivative operators can be written in the Cartesian representation, in which their components are numbered by Cartesian indices, , . The order of these indices does not matter (totally symmetric tensors) and all traces vanish,
(12) |
We note here in passing that we could have equally well used the
Cartesian derivative operators with traces not subtracted out, i.e.,
In principle, below one could replace the spherical representations of derivative operators shown in Table 1 by their Cartesian counterparts (13)-(17) or (18)-(22), and work entirely in the Cartesian representation. However, in our opinion, the use of the spherical representation is superior and more economical. Moreover, whenever calculation of the Cartesian derivatives is more suitable, we may express spherical components of the derivative operators through the Cartesian derivatives, as shown in Table 2. An example of using the Cartesian representation (18)-(22) is given in Sec. 4.