Repository navigation
Canonical index ordering - treat base tensors special? #16
Krzmbrzl
started this conversation in
Theory & Implementation
Replies: 1 comment 2 replies
|
As the "canonical" spin case we should better define |
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
This is in the context of #15 (comment)
The abba spin case originates as follows:
The original term is
which after spin integration has one spin case baba, which reads
transforming
H_T1into the canonical spin case abba via that i<->j antisymmetry, followed by an index renaming (j->i and i->j) leads to(which coincides with the abba spin case that has already been produced during spin integration of the original term. Therefore, only one of these is kept.)
introducing skeleton tensors here leads to
As you can see, index canonicalization (in the program from my master's thesis) is performed from left to right, starting with the result tensor. During this, all tensors are treated equally (no one has intrinsic preference over the other). That seemed like a reasonable approach.
@ak-ustutt if I understood you correctly, you would propose to prioritize base tensors like H, T1, T2 and so on when determining the canonical index ordering in a given term, correct? So instead of going left to right, we would process tensors by their priority and only after that from left to right.
And in addition to this, we would also require prioritizing index names over spin case when determining the canonical order. However, iirc I specifically prioritized the spin case over index names, because otherwise the spin case was rarely changed during canonicalization and then later in the code (spin summation) we would require everything to be in specific spin cases, in order for skeleton tensors to be introduced. Thus, it seemed simpler to make sure we arrive at the desired spin cases as soon (and as consistently) as possible, without having to reinvent the canonicalization routine for spin summation 🤔
All reactions