Where can I find the manual or references for the three-body embedding descriptor in v2.0.0?

[kmyeo1007]

I'm interested in the Three-body embedding descriptor.
I already find this manual https://deepmd.readthedocs.io/en/latest/train-input.html.
But I can't find the theoretical detail of all the new version descripters.
What's the difference in se_e2_a(in v2) with se_a(in v1)?
Also, the se_e2_r and se_r too.
Thank you.

[njzjz]

What's the difference in se_e2_a(in v2) with se_a(in v1)?
Also, the se_e2_r and se_r too.

There's no difference.

[kmyeo1007]

Thank you very much :)
And where can I find any theoretical detail of the three-body embedding descriptor(se_e3) that I can refer to?

[njzjz]

Hi, it is now available in the following preprint: https://arxiv.org/abs/2111.04281