This add-on to pymatgen provides tools for analyzing diffusion in materials.
Home Page: https://guide.materialsvirtuallab.org/pymatgen-diffusion/
License: BSD 3-Clause "New" or "Revised" License
![dependabot-preview[bot] avatar](https://avatars.githubusercontent.com/in/2141?v=4 "dependabot-preview[bot]")
![dependabot[bot] avatar](https://avatars.githubusercontent.com/in/29110?v=4 "dependabot[bot]")
![pre-commit-ci[bot] avatar](https://avatars.githubusercontent.com/in/68672?v=4 "pre-commit-ci[bot]")
Dependabot couldn't authenticate with https://pypi.python.org/simple/.
You can provide authentication details in your Dependabot dashboard by clicking into the account menu (in the top right) and selecting 'Config variables'.
v2023.8.15
The function "get_msd_plot()" is incompatible with the function "pretty_plot" in pymatgen
analyzer.get_msd_plot()TypeError Traceback (most recent call last)
Cell In[10], line 1
----> 1 analyzer.get_msd_plot(plt=None)
File ~/soft/anaconda3/envs/pymatgen/lib/python3.11/site-packages/pymatgen/analysis/diffusion/analyzer.py:561, in DiffusionAnalyzer.get_msd_plot(self, plt, mode)
548 """
549 Get the plot of the smoothed msd vs time graph. Useful for
550 checking convergence. This can be written to an image file.
(...)
557 time will be plotted.
558 """
559 from pymatgen.util.plotting import pretty_plot
--> 561 plt = pretty_plot(12, 8, plt=plt)
562 if np.max(self.dt) > 100000:
563 plot_dt = self.dt / 1000
TypeError: pretty_plot() got an unexpected keyword argument 'plt'In the current implementation of ProbabilityDensityAnalysis class, it does not output the low-energy sites along the diffusion pathways. This piece of information is useful in some diffusion analyses (e.g. effects of local environment on the ionic migration).
v2023.1.1
I'm experiencing an issue accessing the Materials Virtual Lab guide for pymatgen-diffusion at the following URL: https://guide.materialsvirtuallab.org/pymatgen-diffusion/. The website appears to be down or inaccessible.
1. Navigate to [https://guide.materialsvirtuallab.org/pymatgen-diffusion/](https://guide.materialsvirtuallab.org/pymatgen-diffusion/).
2. Observe the inability to access the site or receive an error message.404 File not found
The site configured at this address does not contain the requested file.
If this is your site, make sure that the filename case matches the URL as well as any file permissions.v2024.6.25.1
@shyuep similar problems persist in this one too. 1d plotting works fine tho!
for i, analyzer in enumerate(diff_analyzer):
try:
print(f"Starting Van Hove Analysis for analyzer {i} at {temperatures[i]}K")
van_hove = VanHoveAnalysis(
analyzer,
avg_nsteps=avg_nsteps,
ngrid=ngrid,
rmax=rmax,
step_skip=step_skip,
sigma=sigma,
cell_range=cell_range,
species=diffusing_species,
reference_species=reference_species,
indices=indices
)
# Save 3d plot
van_hove_3d_plot = van_hove.get_3d_plot(mode='distinct')
van_hove_3d_plot_filename = os.path.join(output_dir, f'van_hove_3d_plot_{i}.png')
plt.savefig(van_hove_3d_plot_filename)
plt.close() # Close the figure to free memory
print(f"3D plot saved for analyzer {i} at {temperatures[i]}K to {van_hove_3d_plot_filename}")Starting Van Hove Analysis for analyzer 0 at 800K
Unexpected error for analyzer 0 at 800K: VanHoveAnalysis.get_3d_plot() got an unexpected keyword argument 'ax'
Starting Van Hove Analysis for analyzer 1 at 1000K
Unexpected error for analyzer 1 at 1000K: VanHoveAnalysis.get_3d_plot() got an unexpected keyword argument 'ax'
Starting Van Hove Analysis for analyzer 2 at 1200K
Unexpected error for analyzer 2 at 1200K: VanHoveAnalysis.get_3d_plot() got an unexpected keyword argument 'ax'
Starting Van Hove Analysis for analyzer 3 at 1400K
Unexpected error for analyzer 3 at 1400K: VanHoveAnalysis.get_3d_plot() got an unexpected keyword argument 'ax'
Starting Van Hove Analysis for analyzer 4 at 1600K
Unexpected error for analyzer 4 at 1600K: VanHoveAnalysis.get_3d_plot() got an unexpected keyword argument 'ax'Dear Developers,
Thanks for your work, i'm so interested in the the probability_analysis.ipynb. I want to ask a simple question that how could we get the .npy file on our work for generating the .vasp file .By vasp or some other code?could you tell me some details about it?
Thanks for your help in advance,
Regards,
Joey
v2022.7.21
diffusivities = [d.diffusivity for d in analyzers.values()]
plt = get_arrhenius_plot(temperatures, diffusivities)Graph is empty@jmmshn
Please see attachments for the structure files and jupyter notebook for the minimum breaking code.
uc_sc_trans_error.zip
At present the RadialDistributionFunction class is initialised using species and reference_species arguments, to select the atoms used in calculating the RDF. The species labels are used to generate index lists inside the __init__ method: e.g.
indices = [j for j, site in enumerate(structures[0])
if site.specie.symbol in species]This would be more flexible if atoms could be selected directly by index. e.g.
RadialDistributionFunction(spec_indices=[0, 1, 2, 3], ref_spec_indices=[4, 5, 6, 7])as it would allow you to calculate RDFs for a subset of a particular species (e.g. Li only at specific sites in the lattice).
To allow this and selecting by species labels, I can think of two possible approaches:
Add new optional arguments to the current __init__ method, along with logic to allow defined indices to supersede the default species arguments.
This has the advantage of not changing the API for RDFs generated using species labels, at the cost of making the API and code more complex.
Change the __init__ method to only accept selection by indices. Selection by species label is moved to a new class method e.g. RadialDistributionFunction.from_species_labels('Li') which contains the code to identify the correct indices, and then calls the class __init__() method. I think this keeps the code logic cleaner, but breaks backwards compatibility with the current API.
Hi, in the given example file examples/radical_distribution_function.ipynb, setting n_jobs !=1 caused ValueError: not enough values to unpack (expected 3, got 2)
This error make the fast rdf functions less appealing and practically slow down the calculation.
Hope this be fixed some day.
No response
v2023.8.15
I encountered a TypeError when trying to use the get_arrhenius_plot function from the pymatgen.analysis.diffusion.analyzer module.
https://github.com/materialsvirtuallab/Data-driven-First-Principles-Methods-for-the-Study-and-Design-of-Alkali-Superionic-Conductors/blob/master/notebooks/3%20-%20Diffusivity%20and%20Ionic%20Conductivity.ipynbFile "/home/lws/Desktop/SSE/analysis.py", line 22, in <module>
get_arrhenius_plot(temperatures, diffusivities).savefig("arrhenius_plot.png")
File "/home/lws/anaconda3/envs/deepmd/lib/python3.10/site-packages/pymatgen/analysis/diffusion/analyzer.py", line 1099, in get_arrhenius_plot
ax = plt.axes()
TypeError: 'Axes' object is not callablev2024.6.24
I reinstalled pymatgen and pymatgen-diffusion-analysis today but still cannot use plotting functions.
analyzer.get_msd_plot(msd_plot_filename)Traceback (most recent call last):
File "/gpfs/fs0/scratch/o/ovoznyy/orhan/conductivity_pymatgen/conductivity.py", line 123, in <module>
analyzer.get_msd_plot(msd_plot_filename)
File "/gpfs/fs1/home/o/ovoznyy/orhan/general/lib/python3.12/site-packages/pymatgen/analysis/diffusion/analyzer.py", line 552, in get_msd_plot
plt.tight_layout()
^^^^^^^^^^^^^^^^
AttributeError: 'Axes' object has no attribute 'tight_layout'. Did you mean: '_in_layout'?I'm trying to add the PR that was previously intended for pymatgen.
#materialsproject/pymatgen#1435
My understanding of the current implementation of DistinctPathFinder is that it only looks for distinct paths and does not keep track of all paths.
My usage scenario requires the full list of hops (I'm planning on analyzing full migrations pathways using either charge density of IDPP). This will require some redesign of the class.
If there are no objections, I would like to rewrite the DistinctPathFinder with the StructureGraph to keep track of all the symmetry equivalent hops. I think I can keep the output as a list of MigrationPaths so it won't affect the rest of the codebase very much.
However, doing this means that we will have to compare any two hops to check if they are equivalent. The only way I can see for doing this is by adding two dummy atoms at the endpoints for each hop and performing structure matching.
So I will have to replace how the set of symmetrically inequivalent hops are generated.
This is going to take some effort and I just wanted to make sure the devs agree with my approach before I start coding this up. And make sure doing this will not disrupt other ongoing efforts.
Thanks in advance.
2024.3.1
I am attempting to plot the van hove correlation for bulk Na metal:
`analyzer = collections.OrderedDict()
for temp, xdatcar_path in temp_path.items():
if temp == 300:
ionicstep_start = 10764
elif temp == 600:
ionicstep_start = 4943
st = Xdatcar(xdatcar_path, ionicstep_start=ionicstep_start).structures
print(len(st))
analyzer[temp] = DiffusionAnalyzer.from_structures(st, specie="Na", temperature=temp, time_step=1, step_skip=1, smoothed=False)`
It give me below error:
10000
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Cell In[17], line 16
14 st = Xdatcar(xdatcar_path, ionicstep_start=ionicstep_start).structures
15 print(len(st))
---> 16 analyzer[temp] = DiffusionAnalyzer.from_structures(st, specie="Na", temperature=temp, time_step=1, step_skip=1, smoothed=False)
File ~/.repos/venv/py312/lib/python3.11/site-packages/pymatgen/analysis/diffusion/analyzer.py:703, in DiffusionAnalyzer.from_structures(cls, structures, specie, temperature, time_step, step_skip, initial_disp, initial_structure, **kwargs)
700 if initial_disp is not None:
701 disp += initial_disp[:, None, :]
--> 703 return cls(
704 structure,
705 disp,
706 specie,
707 temperature,
708 time_step,
709 step_skip=step_skip,
710 lattices=l,
711 **kwargs,
712 )
File ~/.repos/venv/py312/lib/python3.11/site-packages/pymatgen/analysis/diffusion/analyzer.py:411, in DiffusionAnalyzer.__init__(self, structure, displacements, specie, temperature, time_step, step_skip, smoothed, min_obs, avg_nsteps, lattices, c_ranges, c_range_include_edge, structures)
409 self.corrected_displacements = dc
410 self.max_ion_displacements = np.max(np.sum(dc**2, axis=-1) ** 0.5, axis=1)
--> 411 self.max_framework_displacement = np.max(
412 self.max_ion_displacements[framework_indices]
413 )
414 self.msd = msd
415 self.mscd = mscd
File ~/.repos/venv/py312/lib/python3.11/site-packages/numpy/core/fromnumeric.py:2810, in max(a, axis, out, keepdims, initial, where)
2692 @array_function_dispatch(_max_dispatcher)
2693 @set_module('numpy')
2694 def max(a, axis=None, out=None, keepdims=np._NoValue, initial=np._NoValue,
2695 where=np._NoValue):
2696 """
2697 Return the maximum of an array or maximum along an axis.
2698
(...)
2808 5
2809 """
-> 2810 return _wrapreduction(a, np.maximum, 'max', axis, None, out,
2811 keepdims=keepdims, initial=initial, where=where)
File ~/.repos/venv/py312/lib/python3.11/site-packages/numpy/core/fromnumeric.py:88, in _wrapreduction(obj, ufunc, method, axis, dtype, out, **kwargs)
85 else:
86 return reduction(axis=axis, out=out, **passkwargs)
---> 88 return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
ValueError: zero-size array to reduction operation maximum which has no identity
```
I am unable to understand that the XDATCAR file is not empty then why it is unable to read? Kindly @shyuep
Thanks in advance!
### Code snippet
```python
st = Xdatcar(xdatcar_path, ionicstep_start=ionicstep_start).structures
print(len(st))
analyzer[temp] = DiffusionAnalyzer.from_structures(st, specie="Na", temperature=temp, time_step=1, step_skip=1, smoothed=False)`
```
### Log output
```shell
10000
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Cell In[17], line 16
14 st = Xdatcar(xdatcar_path, ionicstep_start=ionicstep_start).structures
15 print(len(st))
---> 16 analyzer[temp] = DiffusionAnalyzer.from_structures(st, specie="Na", temperature=temp, time_step=1, step_skip=1, smoothed=False)
File ~/.repos/venv/py312/lib/python3.11/site-packages/pymatgen/analysis/diffusion/analyzer.py:703, in DiffusionAnalyzer.from_structures(cls, structures, specie, temperature, time_step, step_skip, initial_disp, initial_structure, **kwargs)
700 if initial_disp is not None:
701 disp += initial_disp[:, None, :]
--> 703 return cls(
704 structure,
705 disp,
706 specie,
707 temperature,
708 time_step,
709 step_skip=step_skip,
710 lattices=l,
711 **kwargs,
712 )
File ~/.repos/venv/py312/lib/python3.11/site-packages/pymatgen/analysis/diffusion/analyzer.py:411, in DiffusionAnalyzer.__init__(self, structure, displacements, specie, temperature, time_step, step_skip, smoothed, min_obs, avg_nsteps, lattices, c_ranges, c_range_include_edge, structures)
409 self.corrected_displacements = dc
410 self.max_ion_displacements = np.max(np.sum(dc**2, axis=-1) ** 0.5, axis=1)
--> 411 self.max_framework_displacement = np.max(
412 self.max_ion_displacements[framework_indices]
413 )
414 self.msd = msd
415 self.mscd = mscd
File ~/.repos/venv/py312/lib/python3.11/site-packages/numpy/core/fromnumeric.py:2810, in max(a, axis, out, keepdims, initial, where)
2692 @array_function_dispatch(_max_dispatcher)
2693 @set_module('numpy')
2694 def max(a, axis=None, out=None, keepdims=np._NoValue, initial=np._NoValue,
2695 where=np._NoValue):
2696 """
2697 Return the maximum of an array or maximum along an axis.
2698
(...)
2808 5
2809 """
-> 2810 return _wrapreduction(a, np.maximum, 'max', axis, None, out,
2811 keepdims=keepdims, initial=initial, where=where)
File ~/.repos/venv/py312/lib/python3.11/site-packages/numpy/core/fromnumeric.py:88, in _wrapreduction(obj, ufunc, method, axis, dtype, out, **kwargs)
85 else:
86 return reduction(axis=axis, out=out, **passkwargs)
---> 88 return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
ValueError: zero-size array to reduction operation maximum which has no identity
I am unable to understand when the XDATCAR file is not empty then why it is unable to read? Kindly @shyuep
Thanks in advance!
```
### Code of Conduct
- [X] I agree to follow this project's Code of Conduct
For some complicated alkali conductors, the default 5 Angstrom max_path_length will lead to quite many paths and may be more than needed. If we reduce the path length cutoff and can still form the percolating pathways, it would help to either save efforts in unnecessary NEB calculations or quick estimate if a material is a 3D conductor before doing any calculations.
Right now, the van Hove plots only support one type of visualization - the spectrum like plot.
I would like options to have this plotted as a 3D surface, or a series of stacked plots (e.g., how people usually plot it in the literature).
Dear Developers,
Thank you for your interesting work on pymatgen-diffusion.
In my testing, I use the following script to generate an initial path of NEB:
the_solver = IDPPSolver.from_endpoints(endpoints=[init_struct, final_struct], nimages=nimages,
sort_tol=sort_tol)
new_path = the_solver.run(maxiter=5000, tol=1e-5, gtol=1e-3, species=species)
This functionality is very convenient for me, however, it is found that the order of atoms in the generated path can be different from those in the given end points (initial_struct, final_struct). For example, the order in O, H will be changed into H, O using the IDPP solver. Since we need to fix some substrate layers in some NEB calculations and these "fix" flags have already been set for the initial/final_struct, it is desired that the generated path can use the same atomic order as given in the initial/final_struct.
Can I easily achieve this by some minor revisions of scripts? Thank you.
Outputting PDA data in chgcar format can give atoms in the wrong order. This function uses the self.structure.symbol_set to extract the element symbols, but these are stored in alphabetical order, which is not necessarily the order of the atom coordinate data (e.g., as read in from an XDATCAR file) or PDA data, resulting in element symbols and PDA data possibly becoming out of sync.
One way to fix this is to use the same method as in pymatgen.core.trajectory to extract the element symbols in correct order, using itertools functionality.
v2024.6.25
@shyuep could you please check pypi installer. I downloaded these two files from repo manually and try to run some calcualtions now..
from pymatgen.analysis.diffusion.analyzer import DiffusionAnalyzer, get_extrapolated_conductivity, \
get_extrapolated_diffusivity, get_conversion_factor, fit_arrhenius, get_arrhenius_plotTraceback (most recent call last):
File "/gpfs/fs0/scratch/o/ovoznyy/orhan/conductivity_pymatgen/conductivity.py", line 3, in <module>
from pymatgen.analysis.diffusion.analyzer import DiffusionAnalyzer, get_extrapolated_conductivity, \
ModuleNotFoundError: No module named 'pymatgen.analysis.diffusion.analyzer'Please update pymatgen-diffusion in PyPI with the version in this github.
Occupancy analysis based on AIMD simulations should be added.
When you run migrationhop.get_sc_structures(vac_mode=True) the vacancy-based structure is not returned. Instead we just receive an empty host lattice with no diffusing atoms.
As an example, the following code can reproduce the error:
# Load our structure. Note that YbOF is mp-7028
from pymatgen.core.structure import Structure
structure = Structure.from_file("ybof.cif")
# grab the first (and only) pathway found by DistinctPathFinder as an example
from pymatgen.analysis.diffusion.neb.pathfinder import DistinctPathFinder
pathfinder = DistinctPathFinder(structure, migrating_specie="F")
pathway = pathfinder.get_paths()[0]
# now generate the corresponding supercells for this pathway
structure_start, structure_end, base_supercell = pathway.get_sc_structures(vac_mode=True)
# Let's look at the output and confirm the error.
# Here we expect to see Yb48O48F47 but instead see Yb48O48, which indicates
# we are only getting the host lattice back -- and are missing the diffusing specie
structure_start.composition
structure_end.compositionI have made a fix for this bug and will make a pull request for this shortly.
Would it be possible to increase the tolerance when deleting atoms for start/end structures?
I have a database setup where I'm initializing MigrationHop objects from past data -- so this tolerance is too strict compared to what I'm storing metadata at. When I switch the tolerance from 1e-8 to 1e-5 everything works though. I don't think anyone would need a tolerance lower than a 1e-3 (unless someone is using a physically unreasonable structure). Would it be okay if I opened a pull request for this?
Hello,
I get a strange result from ProbabilityDensityAnalysis. Li and transition metal layers are not arranged like POSCAR/CONTCAR. Here is the code I use:
file = ["tem-1000/out_vasprun.xml"]
analyzer = DiffusionAnalyzer.from_files(file, specie="Li", smoothed=False)
structure = analyzer.structure
trajectories = [s.frac_coords for s in analyzer.get_drift_corrected_structures()]
pda = ProbabilityDensityAnalysis(structure, trajectories, species="Li")
pda.to_chgcar("CHGCAR.vasp")
I attach vasprun file below. Hope that someone help me to check.
Dear Developers,
Thanks for your contribution for the pymatgen, I am interested in the probability_analysis.ipynb,I thought it is fantacy. But i don't know how to generate the .npy or the .json file of our own work. By vasp or some other code ?.Could you give me some advice?
Thanks for your help in advance,
Regards,
Joey
pymatgen-analysis-diffusion/pymatgen/analysis/diffusion/analyzer.py
Lines 229 to 243 in 5923fb5
DiffusionAnalyzer will fail if there is no framework ions (namely if all ions are mobile)
Basically if framework_indices is an empty list. The first dimension of framework_disp would be 0. Then drift would be a matrix filled with NAN, which leads to failure in the downstream.
This should a minor bug that is easy to fix.
No response
latest
I want to plot arrhenius, msd and van hove plots but all of them return similar error as you can see below. I just downloaded the last commit by @shyuep but nothing changed. Could you please fix the plotting issues in genereal?
Thank you in advance!
analyzer.get_msd_plot(msd_plot_filename)Traceback (most recent call last):
File "/gpfs/fs0/scratch/o/ovoznyy/orhan/conductivity_pymatgen/conductivity.py", line 123, in <module>
analyzer.get_msd_plot(msd_plot_filename)
File "/gpfs/fs1/home/o/ovoznyy/orhan/general/lib/python3.12/site-packages/pymatgen/analysis/diffusion/analyzer.py", line 547, in get_msd_plot
plt.xlabel(f"Timestep ({unit})")
^^^^^^^^^^
AttributeError: 'Axes' object has no attribute 'xlabel'. Did you mean: '_label'?"get_structure" can not properly read machine learning steps in a vasprun.xml file. Any plans to fix the issue or is there a work around (e.g. create a json file containing all the steps)?
I am trying an example and plot with get_arrhenius_plot function. No error, but figure does not show information.
temperatures = [600, 800, 1000, 1200]
analyzers = collections.OrderedDict()
for temp in temperatures:
with open("aimd_data/%d.json" % temp) as f:
d = json.load(f)
analyzers[temp] = DiffusionAnalyzer.from_dict(d)
diffusivities = [d.diffusivity for d in analyzers.values()]
#print(temperatures, diffusivities)
plt = get_arrhenius_plot(temperatures, diffusivities)
This would make it a lot easier to use than to specially pass in structure and trajectories.
No response
v2023.8.15
Due to the breaking change of pymatgen (Breaking: Have plot methods return plt.Axes object, not matplotlib module), the following error comes out in the current version.
pymatgen\analysis\diffusion\analyzer.py", line 560, in get_msd_plot
plt = pretty_plot(12, 8, plt=plt)
^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: pretty_plot() got an unexpected keyword argument 'plt'
No response
No response
Hi, I think the MSD calculation is incorrect. I obtained the smearing, and spiky MSD for the smooth option = False. Because we need to find the fitting slope from MSD(t) line, it is hard to find the best fitting time lag from the spiky MSD. �Maybe a function log(MSD) vs log(t) should be introduced to find the ballistic or diffusive regime, which can not see from a spiky MSD. The user should have an option to choose a suitable time for estimating slop, and diffusivity because the tail of MSD is often removed.
Note: It is surprising that I got the different MSD lines from the same MD output files using the same MSD equation. (compared with MSD obtained from VASPKIT, or Yfei Mo code)
Dear Developers,
I have used RadialDistributionFunction class in vanhove module of this package and found one ambiguity. Input data are 5 structures with same composition.
While passing single structure, each of them have the first RDF peak at around 2 Angstrom but when I pass all the five structures, the resulting RDF (single plot) shows a peak at 2 Angstrom along with another peak below 2 Angstrom, which to me seems ambiguous. Will you please have a look at it and let me know.
Feel free to ask me if you need the input files for testing.
Thanks for your help in advance,
Regards,
Vishank
Hi @shyuep, can we have a new PyPI release of pymatgen-diffusion?
Thanks!
-Jimmy
Hi, I would like to report a possible bug in:
When indx is equal to len(interval) - 1, then IndexError will be raised when we try to get indx + 1 of interval in line 115.
A possible fix would be to change line 111 to: if indx >= len(interval) - 1
Thanks for looking into this!
Hi all,
I just wanted to point out a logical bug in the MigrationPath._eq_ function. When we look at extremely long pathways, there are cases where the isite, msite, and esite are all equivalent, but the paths are not. Yes, the longer path is likely unreasonable, but it can mask the pathway of interest. Here's an example using mp-22526 (LiCoO2):
from pymatgen import MPRester
mpr = MPRester(MY_API_KEY)
structure = mpr.get_structure_by_material_id("mp-22526") # LiCoO2
from pymatgen_diffusion.neb.pathfinder import DistinctPathFinder
dpf = DistinctPathFinder(
structure = structure,
migrating_specie = 'Li',
max_path_length = 5,
symprec = 0.1,
perc_mode = None,
)
paths = dpf.get_paths() # I print out some results of this below
Here, paths will identify the following:
[Path of 4.9240 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [1.000, -2.000, 1.000] (ind: 0, Wyckoff: 1a),
Path of 4.9927 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [0.000, -1.000, 0.000] (ind: 0, Wyckoff: 1a)]
Likewise, if we run the same code with max_path_length = 10 (this a beyond-reasonable cutoff, but I'm just using it to help illustrate the bug) we will obtain the following paths:
[Path of 9.4649 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [3.000, -2.000, -2.000] (ind: 0, Wyckoff: 1a),
Path of 9.5722 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [0.000, -1.000, -1.000] (ind: 0, Wyckoff: 1a),
Path of 9.8826 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [3.000, -1.000, -3.000] (ind: 0, Wyckoff: 1a)]
This is clearly in error as the original pathways identified at 5Ang are not even seen anymore. I've already gone ahead and implemented the fix and tested it -- you simply add the following lines to pymatgen_diffusion.neb.pathfinder at line 330:
# Check that lengths are within 0.0001 Angstroms
if (self.length - other.length) > 1e-4:
return False
I choose to hardcode a tolerance of 1e-4 rather than a straight comparison (self.length != other.length) because the lengths often give differences of <1e-6 Angstroms in cases that should be identical. I think it would be fix to loosen this tolerance as well -- even if the path comparison passes on this check, it will continue to symmetrical comparison of sites as it currently does. While this adds an extra check to _eq_, it actually improves the speed of the function! This is because the distance comparison is cheaper than symmetry comparison, which is done as the last step. After the fix, we get the following outputs for the same code used above:
# After Fix; cutoff = 5Ang
# [Path of 2.8429 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [0.000, -1.000, 1.000] (ind: 0, Wyckoff: 1a),
# Path of 4.9240 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [1.000, -2.000, 1.000] (ind: 0, Wyckoff: 1a),
# Path of 4.9927 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [0.000, -1.000, 0.000] (ind: 0, Wyckoff: 1a)]
# After Fix; cutoff = 10Ang
# [Path of 2.8429 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [0.000, -1.000, 1.000] (ind: 0, Wyckoff: 1a),
# Path of 4.9240 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [1.000, -2.000, 1.000] (ind: 0, Wyckoff: 1a),
# Path of 4.9927 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [0.000, -1.000, 0.000] (ind: 0, Wyckoff: 1a),
# Path of 5.7454 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [1.000, -1.000, -1.000] (ind: 0, Wyckoff: 1a),
# Path of 6.4103 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [1.000, 0.000, -2.000] (ind: 0, Wyckoff: 1a),
# Path of 7.5216 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [2.000, -3.000, 1.000] (ind: 0, Wyckoff: 1a),
# Path of 7.5667 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [2.000, -1.000, -2.000] (ind: 0, Wyckoff: 1a),
# Path of 8.5685 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [2.000, 0.000, -3.000] (ind: 0, Wyckoff: 1a),
# Path of 9.4649 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [3.000, -2.000, -2.000] (ind: 0, Wyckoff: 1a),
# Path of 9.5722 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [0.000, -1.000, -1.000] (ind: 0, Wyckoff: 1a),
# Path of 9.8826 A from Li [0.000, 0.000, 0.000] (ind: 0, Wyckoff: 1a) to Li [3.000, -1.000, -3.000] (ind: 0, Wyckoff: 1a)]
Let me know if you think there's a mistake in my breakdown. Otherwise, if you could add these two lines and release a new version, it would be greatly appreciated! Thanks in advance.
-Jack
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.