BUG: Unable to run phonons with force fields and FileStore

[computron]

Describe the bug
I am trying to run a phonon calculation. The calculation appears to run OK in that

• Phonon band structure PDF is correctly printed
• Several YAML files seem to be correctly printed

Issues:

• The PDF of Phonon DOS seems to be not openable / corrupted
• I also get the following error (my blob_output.yaml file is also corrupted after this):

2026-03-25 08:14:58,274 INFO generate_frequencies_eigenvectors failed with exception:
TypeError: 'NoneType' object is not callable

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/Users/ajain/Documents/code_venvs/atomate2/lib/python3.11/site-packages/jobflow/managers/local.py", line 117, in _run_job
    response = job.run(store=store)
               ^^^^^^^^^^^^^^^^^^^^
  File "/Users/ajain/Documents/code_venvs/atomate2/lib/python3.11/site-packages/jobflow/core/job.py", line 695, in run
    store.update(data, key=["uuid", "index"], save=save)
  File "/Users/ajain/Documents/code_venvs/atomate2/lib/python3.11/site-packages/jobflow/core/store.py", line 360, in update
    store.update(blobs, key="blob_uuid")
  File "/Users/ajain/Documents/code_venvs/atomate2/lib/python3.11/site-packages/maggma/stores/mongolike.py", line 747, in update
    self.update_json_file()
  File "/Users/ajain/Documents/code_venvs/atomate2/lib/python3.11/site-packages/maggma/stores/mongolike.py", line 770, in update_json_file
    bytesdata = orjson.dumps(
                ^^^^^^^^^^^^^
TypeError: Type is not JSON serializable: ScalarFloat

To Reproduce
Steps to reproduce the behavior:

1. Set up jobflow.yaml as in this example:

jobflow_filestore.yaml

2. Run the following script:

ff_phonon.py

The full trace of execution is:

(atomate2) computron:test_scripts ajain$ python ff_phonon.py 
2026-03-25 08:14:01,012 INFO Started executing jobs locally
2026-03-25 08:14:01,017 INFO Starting job - Force field relax (254ed4ae-d944-4d92-99ae-23229508fff1)
2026-03-25 08:14:05,620 INFO Finished job - Force field relax (254ed4ae-d944-4d92-99ae-23229508fff1)
2026-03-25 08:14:05,620 INFO Starting job - get_supercell_size (db50f5ed-b577-48b1-be9d-d34c46ffc8b2)
2026-03-25 08:14:05,687 INFO Finished job - get_supercell_size (db50f5ed-b577-48b1-be9d-d34c46ffc8b2)
2026-03-25 08:14:05,687 INFO Starting job - Force field static (acda7fc8-4f6b-45df-aea7-1baad0378aa7)
2026-03-25 08:14:06,041 INFO Finished job - Force field static (acda7fc8-4f6b-45df-aea7-1baad0378aa7)
2026-03-25 08:14:06,041 INFO Starting job - generate_phonon_displacements (404c4c95-49e6-4029-85c4-d0b27813795b)
/Users/ajain/Documents/code_venvs/atomate2/lib/python3.11/site-packages/jobflow/core/job.py:629: UserWarning: Initial magnetic moments will not be considered for the determination of the symmetry of the structure and thus will be removed now.
  response = function(*self.function_args, **self.function_kwargs)
2026-03-25 08:14:09,002 INFO Finished job - generate_phonon_displacements (404c4c95-49e6-4029-85c4-d0b27813795b)
2026-03-25 08:14:09,003 INFO Starting job - run_phonon_displacements (220f2348-bac7-47c4-8034-79dc11627a59)
2026-03-25 08:14:09,156 INFO Finished job - run_phonon_displacements (220f2348-bac7-47c4-8034-79dc11627a59)
2026-03-25 08:14:09,182 INFO Starting job - Force field static 1/1 (f68ed623-cb82-4395-9ea2-edc265557e83)
2026-03-25 08:14:13,127 INFO Finished job - Force field static 1/1 (f68ed623-cb82-4395-9ea2-edc265557e83)
2026-03-25 08:14:13,128 INFO Starting job - store_inputs (220f2348-bac7-47c4-8034-79dc11627a59, 2)
2026-03-25 08:14:13,152 INFO Finished job - store_inputs (220f2348-bac7-47c4-8034-79dc11627a59, 2)
2026-03-25 08:14:13,153 INFO Starting job - generate_frequencies_eigenvectors (9e576f73-c979-4b92-9041-db8db78cfc85)
The PostScript backend does not support transparency; partially transparent artists will be rendered opaque.
2026-03-25 08:14:58,274 INFO generate_frequencies_eigenvectors failed with exception:
TypeError: 'NoneType' object is not callable

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/Users/ajain/Documents/code_venvs/atomate2/lib/python3.11/site-packages/jobflow/managers/local.py", line 117, in _run_job
    response = job.run(store=store)
               ^^^^^^^^^^^^^^^^^^^^
  File "/Users/ajain/Documents/code_venvs/atomate2/lib/python3.11/site-packages/jobflow/core/job.py", line 695, in run
    store.update(data, key=["uuid", "index"], save=save)
  File "/Users/ajain/Documents/code_venvs/atomate2/lib/python3.11/site-packages/jobflow/core/store.py", line 360, in update
    store.update(blobs, key="blob_uuid")
  File "/Users/ajain/Documents/code_venvs/atomate2/lib/python3.11/site-packages/maggma/stores/mongolike.py", line 747, in update
    self.update_json_file()
  File "/Users/ajain/Documents/code_venvs/atomate2/lib/python3.11/site-packages/maggma/stores/mongolike.py", line 770, in update_json_file
    bytesdata = orjson.dumps(
                ^^^^^^^^^^^^^
TypeError: Type is not JSON serializable: ScalarFloat

2026-03-25 08:14:58,410 INFO Finished executing jobs locally

[JaGeo]

@computron thanks for all the bug reports. I am currently quite busy with other tasks and I am not sure I can debug in the next time.

[computron]

I will ask someone on my side to investigate

[JaGeo]

Thank you!

[hrushikesh-s]

Issues:

• The PDF of Phonon DOS seems to be not openable / corrupted

This happens because the following function call to save the DOS plot doesn't properly mention the file type. Even though the file extension is .pdf, it still considered it as a default EPS image format -->
Old code (see line # 451-454 here)

new_plotter_dos.save_plot(
    filename=kwargs.get("filename_dos", "phonon_dos.pdf"),
    units=kwargs.get("units", "THz"),
)

Update this function call with:

new_plotter_dos.save_plot(
    filename=kwargs.get("filename_dos", "phonon_dos.pdf"),
    img_format="pdf",
    units=kwargs.get("units", "THz"),
)

Let me also create a PR to fix this issue in the main branch

[hrushikesh-s]

Issues:

I also get the following error (my blob_output.yaml file is also corrupted after this):

Okay, so this is a maggma issue. I have started a PR to fix this -- see the PR here . The test for the new feature is also passing, and the PR can be merged upon approval from the maggma maintainers.

Once this PR is merged, you should do one additional change to your jobflow.yaml file -->

Existing file:

JOB_STORE:
  docs_store:
    type: JSONStore
    paths: /path/to/output.json
    read_only: False
  additional_stores:
    data:
      type: JSONStore
      paths: /path/to/blob_output.json
      read_only: False

New file:

JOB_STORE:
  docs_store:
    type: JSONStore
    paths: /path/to/output.json
    read_only: False
    sanitize_on_write: true
  additional_stores:
    data:
      type: JSONStore
      paths: /path/to/blob_output.json
      read_only: False
      sanitize_on_write: true

You should then get two json files, each for the docs_store as well as the data store, saved at your provided path.

[computron]

Is there any reason to not sanitize_on_write by default?

[hrushikesh-s]

I think we can make this a default as well.
In fact, it is helpful to make it the default because if users want to serialize complex objects like Phonon band structure or, in general, other objects that are stored in blobs using jobflow --> then it's better to sanitize these objects before trying to serialize them into JSON.
And it doesn't harm to sanitize non-complex objects as well.

Let me update the maggma PR to make it the default.

Aside -- do you know who I should tag to get the maggma PR merged into main?

[computron]

I am not sure about maggma default - there may be speed reasons to not jsanitize by default there. Perhaps let's just leave things how they are in your PR with the extra line for now unless maggma prefers to sanitize by default.

Aside -- do you know who I should tag to get the maggma PR merged into main?
Probably Aaron, Patrick, or Tyler. I believe Aaron is OOO

[hrushikesh-s]

Okay, sounds good. Let me tag @tschaume and Tyler in the maggma PR.

[computron]

@hrushikesh-s I just realized the 2nd half of this issue is probably the same as this:

#1395

[hrushikesh-s]

@computron Yes, this looks like the same underlying issue as #1395.

From digging into it a bit more (and also based on Tyler’s comment on the Maggma PR), the expectation seems to be that the user is responsible for sanitizing data before it reaches Maggma, rather than Maggma handling arbitrary Python objects during serialization.

Given that, one solution on our side could be to handle this in jobflow -- we can sanitize the blob in JobStore.update() (see here) before they’re handed off to Maggma.

Would you be open to going that route? If yes, then I can make a PR for implementing this change in Jobflow

If not, then Aaron’s custom serializer approach also already works.

[computron]

Can you talk to @esoteric-ephemera about solutions when he returns? Not urgent.

Mostly:
#1400

as referenced in #1395 may already solve the problem?

@esoteric-ephemera can likely give you an opinion quickly

[esoteric-ephemera]

The yaml issues are deeper in monty: ruamel-yaml uses non JSON-serializable data types which make it hard to go from reading in yaml using monty.serialization.loadfn to dumping to JSON with monty.serialization.dumpfn. The two options would be, both on the monty side:

• Move from ruamel-yaml to PyYAML. This might drop support for more complex yaml
• Update monty.json.MontyDecoder to support JSON serialization of ruamel-yaml types

[esoteric-ephemera]

@computron can you to try seeing if atomate2==0.1.0 fixes this for you? If not, we'll patch and put out a 0.1.1