search

mp-interns / eratosthenes

Intelligent sieving of material candidates for the computationally budget-conscious
BSD 2-Clause "Simplified" License
0 stars 2 forks source link

List of our (~8k?) materials for which reasonable elastic constants can be computed #5

Closed dwinston closed 9 years ago

dwinston commented 9 years ago

@chenweis can you help out here? Getting the materials for which data is available is trivial, but can you provide some code/guidance on how to get the full "potential" set of MP materials?

chenweis commented 9 years ago

The criteria I used for the current material list for elastic constant calculations are:

On Tue, Jun 23, 2015 at 12:31 PM, Donny Winston notifications@github.com wrote:

@chenweis https://github.com/chenweis can you help out here. Getting the materials for which data is available is trivial, but can you provide some code/guidance on how to get the full "potential" set of MP materials?

— Reply to this email directly or view it on GitHub https://github.com/mp-interns/eratosthenes/issues/5.

dwinston commented 9 years ago

@chenweis, does the following fit the bill? I constructed it from your description using https://github.com/materialsproject/mapidoc.

elements = ["C", "Mg", "Al", "Zn", "Ni", "Fe", "W", "Co", "Nb",
            "Hf", "Sc", "V", "Cu", "Zr", "Ti", "Pd", "Pt"]
# `m` is a pymatgen.MPRester instance
m.query(criteria={"nelements": {"$lte": 3},
                  "nsites": {"$lte": 12},
                  "e_above_hull": {"$lte": 0.1},
                  "elements": {"$in": elements}},
        properties=["material_id"])

It returns ~6800 materials. Does the root "nsites" key for a material correspond to a conventional cell?

chenweis commented 9 years ago

The nsites key is for primitive cell, but I think you can use that number instead.

Also, if you want to have more materials in the pool, "nelements<=4" can be used.

On Tue, Jun 23, 2015 at 1:17 PM, Donny Winston notifications@github.com wrote:

@chemweis, does the following fit the bill? I constructed it from your description using https://github.com/materialsproject/mapidoc.

elements = ["C", "Mg", "Al", "Zn", "Ni", "Fe", "W", "Co", "Nb", "Hf", "Sc", "V", "Cu", "Zr", "Ti", "Pd", "Pt"]

m is a pymatgen.MPRester instance

m.query(criteria={"nelements": {"$lte": 3}, "nsites": {"$lte": 12}, "e_above_hull": {"$lte": 0.1}, "elements": {"$in": elements}}, properties=["material_id"])

It returns ~6800 materials. Does the root "nsites" key for a material correspond to a conventional cell?

— Reply to this email directly or view it on GitHub https://github.com/mp-interns/eratosthenes/issues/5#issuecomment-114630272 .

tschaume commented 9 years ago

How severe would the consequences be on the computational time when opening up the nsites <= 12 requirement? In case our algorithm works, we might be able to include materials with more sites since we're saving time by only calculating the promising ones.

chenweis commented 9 years ago

Large structures are very expensive for elastic constant calculations. DFT approximately scales with the O(n^3), n=number of electrons. If you can come up with a short list of large structures, it's still possible to calculate them individually.

On Tue, Jun 23, 2015 at 1:30 PM, Patrick Huck notifications@github.com wrote:

How severe would the consequences be on the computational time when opening up the nsites <= 12 requirement? In case our algorithm works, we might be able to include materials with more sites since we're saving time by only calculating the promising ones.

— Reply to this email directly or view it on GitHub https://github.com/mp-interns/eratosthenes/issues/5#issuecomment-114633339 .

dwinston commented 9 years ago

I think we can call this closed. My MPRester.query example earlier in this thread should be sufficient.