3
$\begingroup$

I am trying to get some deeper intuition for the representations of $\rm SO(3)$ and how they combine with each other, and I ran into an odd object that I'm hoping that folks here might help me identify.

As a general setting, I am considering the combinations of the solid harmonics $$S_{\ell m}(\mathbf r) = r^\ell Y_{\ell m}(\theta,\phi)$$ (which are, conveniently, homogeneous polynomials of degree $\ell$, and are much easier to work with than the $Y_{\ell m}$'s), and I am trying to understands what actually happens, at the coordinate level, when you use Clebsch-Gordan coefficents to combine a set of these into a single representation. That is, I define $$ F_{\ell_1\ell_2\ell_3}(\mathbf r_1, \mathbf r_2, \mathbf r_3) = \sum_{m_1,m_2,m_3} ⟨\ell_1m_1\ell_2m_2|\ell_3m_3⟩ S_{\ell_1m_1}(\mathbf r_1) S_{\ell_2m_2}(\mathbf r_2) S_{\ell_3m_3}(\mathbf r_3)^* , $$ which naturally gives a scalar-valued function, and I then look to understand the resulting structure.


As a first example, consider the case when $(\ell_1,\ell_2,\ell_3) = (1,1,1)$. It is an easy calculation (meaning: Mathematica can do it easily) to show that $$ F_{111}(\mathbf r_1, \mathbf r_2, \mathbf r_3) = c_{111} \mathbf r_1 \cdot (\mathbf r_2 \times \mathbf r_3), $$ where $c_{111}$ is a boring numerical constant. Similarly, if one takes the slightly more complex example of $(\ell_1,\ell_2,\ell_3) = (1,2,2)$, there is again a simple and pleasing structure: $$ F_{122}(\mathbf r_1, \mathbf r_2, \mathbf r_3) = c_{122} \mathbf r_1 \cdot (\mathbf r_2 \times \mathbf r_3)(\mathbf r_2 \cdot\mathbf r_3). $$


My question here is what happens a couple of levels up, at $(\ell_1,\ell_2,\ell_3) = (2,2,3)$. Here the polynomials are no longer something that I can recognize, but Mathematica has very little trouble reducing the scalar function to the following form: $$ F_{223}(\mathbf r_1, \mathbf r_2, \mathbf r_3) = c_{223} \left(\mathbf r_1 \cdot (\mathbf r_2 \times \mathbf r_3)\right) \ \mathbf r_1 \cdot \begin{pmatrix} x_2 (y_3^2 + z_3^2 - 4 x_3^2) - 5 y_2 x_3 y_3 - 5 z_2 x_3 z_3 \\ y_2 (z_3^2 + x_3^2 - 4 y_3^2) - 5 z_2 y_3 z_3 - 5 x_2 x_3 y_3 \\ z_2 (x_3^2 + y_3^2 - 4 z_3^2) - 5 x_2 x_3 z_3 - 5 y_2 y_3 z_3 \end{pmatrix} . $$ And this brings me to my real question here: what is that thing on the right?

It is clear to me that, as a cubic-polynomial function of $\mathbf r_2$ and $\mathbf r_3$, this thing transforms as a vector and it is therefore some kind of $\ell=1$ representation of $\rm SO(3)$. But what representation is this, exactly? Does it have a name? Is there some systematic way to treat it? (Say, analogously to the formalism in this answer, maybe?) I have tried to get it as clean as I can, but I'm still not super sold on it, and it feels like there should be a better way to express it.

I would be particularly like to know how this object fits within a larger framework, as I would like to keep going to higher $\ell$'s (at least as high as $(\ell_1,\ell_2,\ell_3) = (2,3,4)$) and the more I can understand about the setting, the better.

$\endgroup$
3
  • $\begingroup$ Are any symmetries more apparent when written in terms of Wigner 3j symbols? $\endgroup$ Commented Aug 30, 2023 at 16:53
  • $\begingroup$ @CosmasZachos Not really. When written out in full, whether it be in terms of 3j symbols or Clebsch-Gordan coefficients, the whole thing is a mess. It's only after simplifying that one can factor out the triple product and get polynomials that are small enough to see anything. $\endgroup$ Commented Aug 30, 2023 at 17:08
  • $\begingroup$ (... but, that said, you are right, 3j symbols are a better approach than Clebsch-Gordan for what I want to do.) $\endgroup$ Commented Aug 30, 2023 at 17:08

1 Answer 1

1
$\begingroup$

OK, folks, I managed to get this on my own =).

Short answer: the vector in question can be expressed as $$ \begin{pmatrix} x_2 (y_3^2 + z_3^2 - 4 x_3^2) - 5 y_2 x_3 y_3 - 5 z_2 x_3 z_3 \\ y_2 (z_3^2 + x_3^2 - 4 y_3^2) - 5 z_2 y_3 z_3 - 5 x_2 x_3 y_3 \\ z_2 (x_3^2 + y_3^2 - 4 z_3^2) - 5 x_2 x_3 z_3 - 5 y_2 y_3 z_3 \end{pmatrix} = \mathbf r_2 r_3^2-5\mathbf r_3 (\mathbf r_2\cdot\mathbf r_3) $$ so that the target function reads $$ F_{223}(\mathbf r_1, \mathbf r_2, \mathbf r_3) = c_{223} \left(\mathbf r_1 \cdot (\mathbf r_2 \times \mathbf r_3)\right) \left[ (\mathbf r_1\cdot\mathbf r_2) r_3^2-5(\mathbf r_1\cdot\mathbf r_3)(\mathbf r_2\cdot\mathbf r_3) \right] . $$

Slightly longer answer.

How does one get to this?

For one, the vector representation is a bit of a red herring, and it is better to work directly with the scalar form (i.e. the big vector dotted with $\mathbf r_1$), as this allows for more freedom in collecting terms and recognizing patterns.

And secondly, the terms of the form $y_3^2 + z_3^2 - 4 x_3^2$ are the key things to look for, and these should be simplified as $$ y_3^2 + z_3^2 - 4 x_3^2 = r_3^2 - 5 x_3^2. $$ Once this separation is done, it is much easier to break those two terms into separate groups and simplify them individually.

$\endgroup$
1
  • $\begingroup$ Not that it's meaningful to me, but you could write the 2nd factor, symmetric in 1,2, as $\vec r_3 M \vec r_3$, with M being the dyadic $M= \vec r_1\cdot \vec r_2 {\mathbb I} -5 \overleftarrow {r}_1 \vec r_2$ . $\endgroup$ Commented Aug 30, 2023 at 20:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.