Add Input-Output HEOM feature#2590
Draft
nwlambert wants to merge 4 commits into
Draft
Conversation
commit 09e222c Author: Neill Lambert <nwlambert@gmail.com> Date: Mon Dec 23 15:21:04 2024 +0900 Add additional missing docstring commit d01a9f6 Author: Neill Lambert <nwlambert@gmail.com> Date: Mon Dec 23 15:17:29 2024 +0900 one more CR to fix commit a03e6f0 Author: Neill Lambert <nwlambert@gmail.com> Date: Mon Dec 23 15:14:48 2024 +0900 undo some whitespace change Some experiences with qobjevo left unwanted whitespace, just undoing it. commit e40fa6e Author: Neill Lambert <nwlambert@gmail.com> Date: Mon Dec 23 15:12:19 2024 +0900 Update environment.py docstrings, add tests Note: in adding tests, i fixed a couple of bugs in existing tests. commit 912a6ed Merge: 8946824 c5761bb Author: Neill Lambert <nwlambert@gmail.com> Date: Mon Dec 23 10:29:05 2024 +0900 Merge branch 'qutip:master' into heom_mauro_inputoutpuit commit 8946824 Author: Neill Lambert <nwlambert@gmail.com> Date: Sun Dec 22 20:46:55 2024 +0900 Fix some docstrings and comments in bofin_solvers commit 50dc67a Author: Neill Lambert <nwlambert@gmail.com> Date: Sun Dec 22 14:31:56 2024 +0900 change definition to be consistent with paper commit 422f2ea Author: Neill Lambert <nwlambert@gmail.com> Date: Fri Dec 20 16:57:48 2024 +0900 fix some linting issues commit f9b55f0 Author: Neill Lambert <nwlambert@gmail.com> Date: Fri Dec 20 16:20:16 2024 +0900 Fix docstrings and linter things commit 141ccd9 Author: Neill Lambert <nwlambert@gmail.com> Date: Thu Dec 19 22:41:06 2024 +0900 small bug fix commit 609a23b Author: Neill Lambert <nwlambert@gmail.com> Date: Mon Dec 16 16:23:33 2024 +0900 fix bug in output functions commit 49c886d Author: Neill Lambert <nwlambert@gmail.com> Date: Mon Dec 16 16:18:54 2024 +0900 add output in function form commit 9401c30 Author: Neill Lambert <nwlambert@gmail.com> Date: Wed Dec 11 16:36:09 2024 +0900 refactor to custom inputoutputbath Undid changes to environment.py, instead collated this all in a new inputoutput class in bofin_baths. i think this is conceptually much simpler and easier to extend in the future commit 3d9b5f5 Author: Neill Lambert <nwlambert@gmail.com> Date: Wed Dec 11 15:41:03 2024 +0900 stable working version all bugs are fixed, but having this in environment.py doesn't make much sense. rework in progress commit 690afd7 Author: Neill Lambert <nwlambert@gmail.com> Date: Mon Dec 9 16:52:20 2024 +0900 Merge conflict changes Gradually merging with environments.py commit 87abdae Merge: 9175faa c185699 Author: Neill Lambert <nwlambert@gmail.com> Date: Mon Dec 9 14:43:07 2024 +0900 Merge remote-tracking branch 'upstream/master' into heom_mauro_inputoutpuit commit 9175faa Merge: 6ca7c1b 464112a Author: Neill Lambert <nwlambert@gmail.com> Date: Fri Nov 1 11:40:53 2024 +0900 Merge branch 'qutip:master' into heom_mauro_inputoutpuit commit 6ca7c1b Merge: 3734237 ad35160 Author: Neill Lambert <nwlambert@gmail.com> Date: Mon Sep 9 15:03:27 2024 +0900 Merge branch 'qutip:master' into heom_mauro_inputoutpuit commit 3734237 Author: Neill Lambert <nwlambert@gmail.com> Date: Thu Sep 5 19:46:44 2024 +0900 spacing fix commit 6a250ab Author: Neill Lambert <nwlambert@gmail.com> Date: Thu Sep 5 17:35:11 2024 +0900 optimize time-dep ops remove cython experiments The time-dep ops are now collated per ck_func, and then turned into qobjevos. this is much faster than old method. commit 1635862 Author: Neill Lambert <nwlambert@gmail.com> Date: Wed Sep 4 18:01:45 2024 +0900 added output fixed output exponents definitions and solver equations commit 3f19f25 Author: Neill Lambert <nwlambert@gmail.com> Date: Thu Aug 29 17:24:36 2024 +0900 refactor to just include TD on input terms Reverted back to using data objects of Ops, and made a new list to store only time-dependent ops for input ADOS. Solves slow RHS construction problem (partially) commit 233627c Author: Neill Lambert <nwlambert@gmail.com> Date: Mon Aug 19 07:53:06 2024 +0900 Experiments with speedup Trying to speed up RHS construction. all failures so far. commit 8efc188 Author: Neill Lambert <nwlambert@gmail.com> Date: Tue Aug 13 09:58:25 2024 +0900 small refactor rename and reduce number of variables needed todefine input/output, etc commit 70dd299 Author: Neill Lambert <nwlambert@gmail.com> Date: Sat Aug 10 13:14:27 2024 +0900 small big fixes fixed some small bugs! now works for input commit 2f7c2a7 Author: Neill Lambert <nwlambert@gmail.com> Date: Fri Aug 9 17:05:42 2024 +0900 Refactor and time-dependent terms Regressed to the old kron based construction so we can use qobjev in blocks and support time-dep Rin terms. commit f0cc7f2 Author: Neill Lambert <nwlambert@gmail.com> Date: Fri Aug 9 08:35:11 2024 +0900 add input/output exponents Add 4 new exponents for real imaginary parts of input output
2 tasks
Added args parameter to solver, and have it passed to the gather function. Also, made sure args for TD hamiltonian were supported in heomsolve.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Checklist
Description
This is a first attempt at adding support for what is called ''input output'' baths in the HEOM solver, first introduced here https://arxiv.org/abs/2408.12221
Usually HEOM comes with 3 caveats: (i) the bath operator the system couples to be must be linear and the free bath evolution should preserve Gaussianity, (ii) the initial state of the bath must be Gaussian and (iii) we can only obtain limited bath information (usually some kind of statistics about the bath coupling operator, but not, e.g., bath occupation at different frequencies/spatial positions).
Input-output HEOM removes (ii) and (iii) via the use of additional ADOs incorporating the statistics about linear bath operators applied to the bath at arbitrary times. If this is done at t=0 we can interpret them as changing the initial state of the bath, hence ''input'', and resolving (ii). If they are done at the final observation time, we can interpret them as informing us about bath properties, hence "output", and resolving (iii) (to some degree). In principle they can also be done at intermediate times, but for simplicity here we stick with this input-output language.
In the HEOM construction we augment the bath with
InputOutputBathobjects, which are a lot like normal HEOM baths except: they truncate exactly at tier 2, and the couple to other ADOs with time-dependent coupling operators. Essentially, generally speaking we don't need to do the normal exponential decomposition for the statistics these baths encode (unless we want to, so for output we also support such decompositions, this will be demonstrated in the example notebooks/docs).To include these in the QuTiP HEOM solver, I did two slightly suspicious things, hence the draft nature of this PR:
InputOutputBath, which inherits from the existing parent bath classes introduced recently, and uses the exponents type to store the information about the correlation information needed for both input and output. The interface to this new bath needs some polishing and checks, and the use of the exponents type is a little bit bad, as now the 'ck' information for some of the terms must be time-dependent functions. When then sent tobofin_solversI check for these new terms, by their new type names, and treat them a bit differently in_GatherHEOMRHS(see below).@ytdHuang suggested instead introducing a new
DynamicalFieldstype. This would be conceptually nicer, but require a bit more sweeping changes to bofin_solvers, and I kinda like that we can still use a lot of the existing exponents functions for this. Anyways, open to suggestions!QobjEvoand using kronecker products made the RHS very slow on large problem sizes. Generalizing_from_csr_blocks()to support qobjevo somehow failed miserably (much beyond my paygrade).After some chats with @pmenczel the optimal solution was a half-way house where we treat TD-terms separately, group them by common TD-function pre-factors, use the existing
_from_csr_blocks()to build those groups and thenQojbEvo()them. Seems to work well so far. But a more robust solution would be nice in the long term if we want to support totally general RHS time dependence for other reasons.On a final point, using the results requires a bit of finesse, manipulating ADOs with
.extract()and so on (see examples). I would like to automate this a bit, and also add some predefined input/output functions for common existing baths. I feel like this will be a separate PR though, as there is a lot of work to be done in that direction.Related issues or PRs
I will add examples to qutip-tutorials in a separate PR. qutip/qutip-tutorials#118
Additional edit comment:
Forgot to mention, no support for fermions yet.