Opened 3 years ago
Last modified 3 years ago
#25163 needs_work enhancement
Implement symmetric functions in super space
Reported by:  zabrocki  Owned by:  

Priority:  major  Milestone:  sage8.4 
Component:  combinatorics  Keywords:  CHAs, sf, super partitions, IMA coding sprint 
Cc:  alauve, darij, zabrocki, tscrim  Merged in:  
Authors:  Mike Zabrocki  Reviewers:  
Report Upstream:  N/A  Work issues:  
Branch:  public/symsuperspace/25163 (Commits, GitHub, GitLab)  Commit:  0c5eb095eb7523405b4a200c72ff29ac34055989 
Dependencies:  #25132, #25133  Stopgaps: 
Description
Symmetric functions in superspace is the Hopf algebra of super partitions (ticket #25132). It has the symmetric functions as a subalgebra.
 [DLM06] P. Desrosiers, L. Lapointe, P. Mathieu, Classical symmetric functions in superspace, J. Algebr Comb. (2006) 24:209238, :arXiv:
0509408
 [JL16] M. Jones, L. Lapointe, Pieri rules for Schur functions in superspace, :arXiv:
1608.08577
Change History (38)
comment:1 Changed 3 years ago by
 Branch set to public/symsuperspace/25163
 Commit set to 38280f78641f626acc1689e9b15d9dd33d802e6d
 Dependencies changed from #25132 to #25132, #25133
 Milestone changed from sage8.2 to sage8.3
 Type changed from PLEASE CHANGE to enhancement
comment:2 Changed 3 years ago by
 Commit changed from 38280f78641f626acc1689e9b15d9dd33d802e6d to e7d352b10cd33670fedbec98175e4e13fa737ec4
comment:3 Changed 3 years ago by
 Commit changed from e7d352b10cd33670fedbec98175e4e13fa737ec4 to c9cacf9737d2e1c2f1241315bab2d2415de50472
Branch pushed to git repo; I updated commit sha1. New commits:
bc276dd  Merge branch 'public/combinat/implement_wqsym25133' of git://trac.sagemath.org/sage into public/combinat/implement_wqsym25133

c9cacf9  Merge branch 'public/symsuperspace/25163' of git://trac.sagemath.org/sage into public/symsuperspace/25163

comment:4 Changed 3 years ago by
Just did a trivial rebase over the trivial conflict in #25133.
comment:5 Changed 3 years ago by
 Commit changed from c9cacf9737d2e1c2f1241315bab2d2415de50472 to e3d942f5b1556fc9ef412f81353ed3826ed4524d
comment:6 Changed 3 years ago by
 Commit changed from e3d942f5b1556fc9ef412f81353ed3826ed4524d to fef75637cb6f245bc0895f817eaa40b15cc3c787
comment:7 Changed 3 years ago by
 Commit changed from fef75637cb6f245bc0895f817eaa40b15cc3c787 to e581b38a7fa209a99dcfdad2e956e0df91bac29b
comment:8 Changed 3 years ago by
 Status changed from new to needs_review
I think that this is ready to go.
comment:9 Changed 3 years ago by
Some bikeshedding on the modulelevel doc:
Let `P_{\Theta_n, X_n}` be the polynomial ring in two sets of variables `\Theta_n = \theta_1, \theta_2, \ldots, \theta_n` and `X_n = x_1, x_2, \ldots, x_n`
The "n" here clashes with the "n" in the later definitions of p_{n;} et al; maybe use a capital N instead? Or work with infinitely many variables right away, if this is possible?
where the first set of variables anticommute
If this includes squaring to 0, then please say so.
The symmetric group acts diagonally on this polynomial ring and the symmetric functions in superspace are isomorphic to the invariants in this polynomial ring.
I'd put a comma before the "and" so it's not misread as "on this polynomial ring and the symmetric functions".
p_{(;i)} p_{(;j)} = p_{(;j)} p_{(;i)} \qquad p_{(i;)} p_{(;j)} = p_{(;j)} p_{(i;)} \qquad p_{(i;)} p_{(j;)} =  p_{(j;)} p_{(i;)}
And p_{(i;)}^2 = 0
, I assume.
p_{(; n)} = p_n \quad e_{(; n)} = e_n \quad h_{(; n)} = h_n
Please use , \quad
, not just \quad
.
from `Theta_n`
Don't forget the backslash. (And again, that's an "N", not an "n".)
bosnic degree
The adjective is "Bosnian".
h_{(n; )} = \sum_{\Lambda \in SPar(n1)} m_\Lambda \qquad
Again, use a comma before the \qquad.
`\left<\left< p_\Lambda, p_\Gamma \right>\right> = \delta_{\Lambda\Gamma} z_{\Lambda^s}` where `z_{\Lambda^s}` is the usual constant `z_\lambda` which is the size of
What is \Lambda^s
?
in pairs. For all `\Lambda, \Gamma in SPar(nm)` ,
in
> \in
As with the space of symmetric functions, there is an involution `\omega` that for all super partitions `\Lambda`, `\omega(h_\Lambda) = e_\Lambda`.
add "satisfies".
Hopf algebra and the coproduct is defined bby declaring that the power sum
"by".
I don't quite get: How is e_\Lambda
defined for a strict partition \Lambda? Just by multiplying e_{n;}
over all fermionic entries n
and e_{;n}
over all bosonic entries n
? In what order?
comment:10 Changed 3 years ago by
 Commit changed from e581b38a7fa209a99dcfdad2e956e0df91bac29b to 4c26ad5edf5f5df4235eac8af4011d141122f9fd
Branch pushed to git repo; I updated commit sha1. New commits:
4c26ad5  Darij's suggestions for documentation

comment:11 Changed 3 years ago by
 Commit changed from 4c26ad5edf5f5df4235eac8af4011d141122f9fd to e9055047f4c462294f0bd89d67ef6551845009c0
Branch pushed to git repo; I updated commit sha1. New commits:
e905504  n > N in two places

comment:12 Changed 3 years ago by
 Commit changed from e9055047f4c462294f0bd89d67ef6551845009c0 to 2c67a04b4f5b541768474dc2b43924817b3c5ba4
Branch pushed to git repo; I updated commit sha1. New commits:
2c67a04  mostly missing periods

comment:13 Changed 3 years ago by
 Commit changed from 2c67a04b4f5b541768474dc2b43924817b3c5ba4 to 82f92cc6e8468d21d995114f717cda6d4d73eb6d
Branch pushed to git repo; I updated commit sha1. New commits:
82f92cc  minor corrections to doc

comment:14 Changed 3 years ago by
 Commit changed from 82f92cc6e8468d21d995114f717cda6d4d73eb6d to 58fb4a41a268c6a2e7122f36733c3b026045991e
Branch pushed to git repo; I updated commit sha1. New commits:
58fb4a4  correction of a few typos

comment:15 Changed 3 years ago by
 Commit changed from 58fb4a41a268c6a2e7122f36733c3b026045991e to 9966e882b0992dbe7011dee65390b3d3a2810d87
Branch pushed to git repo; I updated commit sha1. New commits:
9966e88  manual merge with develop

comment:16 Changed 3 years ago by
+ class Schur_sb(CombinatorialFreeModule, BindableClass): + r""" + The Schur basis arising from `q=t=0` specialization of Macdonald.
+ class Schur_s(CombinatorialFreeModule, BindableClass): + r""" + The Schurstar basis arising from `q=t=0` specialization of Macdonald.
+ class Schur_b(CombinatorialFreeModule, BindableClass): + r""" + The Schurbar basis as the `q=t=\infty` specialization of Macdonald.
+ class Schur(CombinatorialFreeModule, BindableClass): + r""" + The Schurbar basis as the `q=t=\infty` specialization of Macdonald.
Can it be that these descriptions are messed up? Why is the "Schur" doc claiming to be Schurbar? Shouldn't Schur be from q=t=0, not from q=t=\infty?
Also, is there no simpler way to define these bases than by Macdonald specialization? I'd expect some Berezinian analogue of the JacobiTrudi formulas?
comment:17 Changed 3 years ago by
 Commit changed from 9966e882b0992dbe7011dee65390b3d3a2810d87 to 832dca6a320c1880c846477ca36971c03654f719
Branch pushed to git repo; I updated commit sha1. New commits:
832dca6  change the description of the Schur bases

comment:18 Changed 3 years ago by
You are right the descriptions were confused there. The Schurstar and Schurstarbar are related to these by duality and I looked in the paper and noticed that they are not listed as a specialization so I cut that description.
Also, is there no simpler way to define these bases than by Macdonald specialization? I'd expect some Berezinian analogue of the JacobiTrudi formulas?
That would be welcome. Here they are implemented by Pieri rules which are given in [JL2016].
comment:19 Changed 3 years ago by
 Commit changed from 832dca6a320c1880c846477ca36971c03654f719 to b7104306ee3f7a24b8f04f123590242d145d9590
Branch pushed to git repo; I updated commit sha1. New commits:
b710430  added reference [BFM2015] and examples from that reference

comment:20 Changed 3 years ago by
 Commit changed from b7104306ee3f7a24b8f04f123590242d145d9590 to d50bdb386250c606d4e32c0135f85154f4019ac7
Branch pushed to git repo; I updated commit sha1. New commits:
d50bdb3  missing def for monomial basis, mark test long

comment:21 Changed 3 years ago by
 Commit changed from d50bdb386250c606d4e32c0135f85154f4019ac7 to f8e9a967c1086cbe84d858d89ceb00ea24054dfa
Branch pushed to git repo; I updated commit sha1. New commits:
f8e9a96  doc test for antihomomorphism; delete not particularly useful method

comment:22 Changed 3 years ago by
More random comments:
+ The product of monomial basis elements is calculated for a + fixed fermionic partition.
What does that mean?
What is a BindableClass? (no, I don't get the doc at https://doc.sagemath.org/html/en/reference/misc/sage/misc/bindable_class.html ) and why is it being used here?
+where `\Lambda` is a super partition with femionic sector `m`.
feRmionic
Is this a Hopf algebra or a superHopf algebra? I.e., is there a twist involved in the bialgebra axiom? My suspicion is that it is, because otherwise the squarezero relations p_{(i;)}^2 = 0
would contradict the primitivity of the p_{(i;)}
wrt the coproduct. But do you ever tell this to Sage? If you don't, I doubt that it's correctly computing coproducts of nongenerators! Also, of course, this should be said in the doc. (Is the Hopf structure anywhere in the literature?)
+ Test if super partitions an element all have the same bidegree.
Probably should be "if the superpartitions in the support of self
have the same bidegree".
Maybe decide between "symmetric generators" and "bosonic generators"  currently you seem to be using both languages (or do they mean different things?).
comment:23 Changed 3 years ago by
 Commit changed from f8e9a967c1086cbe84d858d89ceb00ea24054dfa to 0b7c97f6c675e9f2bc876e6479ef42b858a12cfe
Branch pushed to git repo; I updated commit sha1. New commits:
0b7c97f  corrections to documentation

comment:24 Changed 3 years ago by
I used BindableClass
because that seems to be what all bases of chas need to inherit (see WQSymBasis_abstract
in wqsym.py
, FSymBasis_abstract
in fsym.py
and bases in ncsf.py
and qsym.py
). I tried deleting the BindableClass
and the command h = self.Complete()
raised a TypeError: __init__() takes exactly 2 arguments (1 given)
. And, no, I do not understand that documentation either.
About the Hopf algebra structure:
My suspicion is that it is, because otherwise the squarezero relations
p_{(i;)}^2 = 0
would contradict the primitivity of thep_{(i;)}
wrt the coproduct.
As far as I can tell this is a Hopf algebra and not a superHopf algebra. There is no posted paper that I can point to, but I do have a nonposted preprint. However I will continue to check this carefully because I am reviewing the paper. Can you explain your suspicion more carefully? Perhaps there is an error in the paper.
comment:25 Changed 3 years ago by
In characteristic 0, any nilpotent primitive element of a (nonsuper) Hopf algebra must be 0. More strongly: If a
is a primitive element of a Hopf algebra in characteristic 0, then the powers of a
are linearly independent.
comment:26 Changed 3 years ago by
Well then. Something is not right with the universe. No test I've performed has identified a problem with the Hopf structure. I'll need to check it against the proof. Can you provide me a reference?
comment:27 Changed 3 years ago by
How about Appendix A here: https://arxiv.org/pdf/1105.5572.pdf
comment:28 Changed 3 years ago by
You are probably not testing enough. The error should materialize when you take the coproduct of a product of p_{n;}s. Basically, the coproduct of p_{n;}p_{m;} should switch signs if you swap n with m; but if I compute it naively in a Hopf algebra, I get
p_{n;} p_{m;} \otimes 1 + p_{n;} \otimes p_{m;} + p_{m;} \otimes p_{n;} + 1 \otimes p_{n;} p_{m;},
which does not swap signs.
Further typos: "Summetric", "fermonionic".
comment:29 Changed 3 years ago by
 Commit changed from 0b7c97f6c675e9f2bc876e6479ef42b858a12cfe to 0c5eb095eb7523405b4a200c72ff29ac34055989
Branch pushed to git repo; I updated commit sha1. New commits:
0c5eb09  typos in documentation

comment:30 followup: ↓ 32 Changed 3 years ago by
What do you mean by swap signs? The sign does arise in the product. Does this seem like it is not correct?
sage: p[4,2].coproduct() p[; ] # p[4, 2; ] + p[2; ] # p[4; ] + p[4; ] # p[2; ] + p[4, 2; ] # p[; ] sage: (p[2]*p[4]).coproduct() p[; ] # p[4, 2; ]  p[2; ] # p[4; ]  p[4; ] # p[2; ]  p[4, 2; ] # p[; ]
Aaron, looking at Proposition 36, "Let K denote the associated graded Hopf algebra with respect to the coradical filtration of H." In order to compare to that proof, do I need to know what the coradical filtration of symmetric functions in super space is?
comment:31 Changed 3 years ago by
sage: (p[2]*p[4]).coproduct() p[; ] # p[4, 2; ]  p[2; ] # p[4; ]  p[4; ] # p[2; ]  p[4, 2; ] # p[; ]
I'm fairly sure you don't want this to be the result. The general version of this formula would be
sage: (p[i]*p[j]).coproduct() p[; ] # p[j, i; ]  p[i; ] # p[j; ]  p[j; ] # p[i; ]  p[j, i; ] # p[; ]
but this doesn't survive swapping i with j.
The fact that I'm cited is incomplete, by the way (sorry): It should require a
to be nonzero, and the base ring to be a field.
comment:32 in reply to: ↑ 30 Changed 3 years ago by
Replying to zabrocki:
Aaron, looking at Proposition 36, "Let K denote the associated graded Hopf algebra with respect to the coradical filtration of H." In order to compare to that proof, do I need to know what the coradical filtration of symmetric functions in super space is?
Probably not. The braiding (by 1) enters during use of twist maps. So axioms for coalgebra will be unchanged, I believe. This means it's the same as the coradical filtration for ordinary symmetric functions? (I have not looked at the coproduct defined...) This means the kth level would have basis p[lam;mu]
where the sum of the lengths of lam and mu is lessorequal to k. See Sec. 2.3 http://pi.math.cornell.edu/~maguiar/a.pdf for more on qHopf algebras, though they do not address coradical filtration.
I expect the associated graded qHopf algebra will be embeddable in a qshuffle Hopf algebra, which is also a domain.
comment:33 Changed 3 years ago by
 Status changed from needs_review to needs_work
Thank you. You spotted a very subtle and potentially fatal mistake on my part. I missed the line that says (f_1 \otimes g_1) (f_2 \otimes g_2) = (1)^{ab} f_1 f_2 \otimes g_1 g_2
where a
and b
are the fermionic degrees of g_1
and f_2
. If I am not mistaken then, this is a superHopf algebra and I need to use the super algebra category: see #18044 and the discussion in #18675 .
Is this still a CHA?
comment:34 Changed 3 years ago by
If I am not mistaken, the coproduct for a supercoalgebra is not implemented. Is this worth doing first on a separate ticket?
comment:35 Changed 3 years ago by
Likely relevant: #25603.
comment:36 Changed 3 years ago by
Maybe also move the coproduct out of this ticket for the time being  this ticket is already too large for comfortable reviewing...
comment:37 Changed 3 years ago by
I'm debating taking the coproduct out an moving it to another ticket. I want to see what happens with #25603 because I would like to know what needs to be done in order to implement it.
Identifying this mistake helped me locate an error in the paper. I pulled in #25603 and tried to verify that this was a Hopfsuperalgebra (or superHopfalgebra, whatever). The formula for the antipode in the paper used the fact that S(f*g) == S(g)*S(f)
to derive the antipode on the powersum basis and this was leading to doctest failures on _test_antipode
. I don't know why, but _test_antipode
passed on other bases even though the antipode is computed by coercion.
comment:38 Changed 3 years ago by
 Milestone changed from sage8.3 to sage8.4
update milestone 8.3 > 8.4
A first implementation. I merged #25133 so that the CHAS directory is there.
Last 10 new commits:
Corrected doc tests, moved theorem_10 as a method, removed super_compositions
doc changes and rename one pieri rule
add superpartition to the doc
doc corrections, added to list of catalog_partitions and enumerated_sets, pyflakes
Refactored class to use ClonableArray. Other reviewer changes.
corrected reference
Merge branch 'public/superpartitions' into public/symsuperspace/25163
first implementation
remove spaces
import statements, documentation lists, minor correction