Volume/Nsites not correct for at least mp-1783


mp-1783 (MnS) is listed as Volume = 84.62 A^3 and Nsites = 2. This leads to a large and unrealistic atomic density of ~42 A^3/atom. However, checking the POSCAR, there are actually 8 sites and the volume is 166.375 A^3 leading to a true atomic density of ~20 A^3/atom, which is reasonable.

It’s difficult to say how pervasive this issue is, but it’s worth looking into as quantities like atomic density are often used for screening, data analysis, etc. and these quantities seem easy enough to pull from the corresponding POSCARs.



Hi @CJBartel,

Welcome to the forum!

In this case I’m not seeing 2 sites, I’m seeing 4: there are 2 Mn sites and 2 S sites, so the atomic density is correct given the 84.62 Å^3 volume.



Hi Matt,

I’m referring to “Nsites” as displayed on the website and “nsites” as queried from the database using the api. Both of these return 2.

1 Like

As an update to this, 3 of the 4 MnS entries have the incorrect atomic density (volume/nsites), suggesting this is not a singular issue.

If you manually check these POSCARs, mp-2562 is the only one with the correct atomic density.

mp-2065 has atomic density = 17 A^3/atom from the POSCAR and 33 from the query, mp-1783 - 21 A^3/atom from POSCAR and 42 from query, mp-556853 - 15 from the POSCAR and 30 from the query.

See below dictionary which results from an MP query for formula = ‘Mn1S1’. It appears the nsites that is stored for querying purposes sometimes differs by a factor of 2 compared with the actual number of sites that corresponds with the reported volume.

{‘mp-2562’: {‘volume’: 85.19846634546657,
‘nsites’: 4,
‘material_id’: ‘mp-2562’,
‘pretty_formula’: ‘MnS’},
‘mp-2065’: {‘volume’: 67.25531630763525,
‘nsites’: 2,
‘material_id’: ‘mp-2065’,
‘pretty_formula’: ‘MnS’},
‘mp-556853’: {‘volume’: 59.52666598810431,
‘nsites’: 2,
‘material_id’: ‘mp-556853’,
‘pretty_formula’: ‘MnS’},
‘mp-1783’: {‘volume’: 84.61861307639987,
‘nsites’: 2,
‘material_id’: ‘mp-1783’,
‘pretty_formula’: ‘MnS’}}


Thanks for updating on this Chris, agreed that definitely looks like an error. Will follow up.