Skip to content

Latest commit

 

History

History
100 lines (65 loc) · 4.09 KB

toDoList.md

File metadata and controls

100 lines (65 loc) · 4.09 KB

To-Do List

Goals

  • Ensure the accuracy of the dataset

  • Format the data to be human and machine readable

  • Generate informative visualizations

  • Checkpoint: Publish Data In Data In Brief

  • Long-term: To use dataset for machine learning, relating composition to properties

Low-hanging fruits

  • Make the DOIs all consistent links

  • Remove rows containing missing DOIs

    • 840 total (573 mrl, 19 personal, 250 blanks)
    • 8548 datapoints remaining

  • Download all papers and put them in the repo

    • Use legit sources, not scihub. So need to login with credentials.

Missing Data Still

  • Synthesis data need to consolidate / separate mixing parameters, synthesis method and preparative route

  • Source 1 (data collected by Layla and Leng Ze)

    • Bulk or Thin Film
    • Calcination and Mixing Parameters - Maybe do it after consolidating the synthesis data

Columns to consider deleting

  • Space group, ICSD stuff

  • Thermal Diffusivity, Weighted Mobility, Measurement Atmosphere, Carrier Concentration, Carrier Mobility

Functions for Column Conversion

  • Function to convert Mass and Mixed Formula to Stoichiometric compositions

    • Eg. Some formula are now in weight ratios ((W18O40)0.3(ZnO)0.7) and some are in a mixture of stoichiometric ratios and weight ratios (0.9Cu2Se-0.1(Bi0.88Pb0.06Ca0.06CuSeO)-0.01 wt% Graphene)
    • WORST CASE: Delete these entries (count how many of them are there first)
    • 302 total (all mixed formula, i.e. some parts are stoichiometrically represented, typically the compound itself, while other parts are mass represented)
    • Done initial function
    • Need to check if the conversion is correct

  • Host-Dopant Based on a Function

    • Function to take in a threshold % (like 1%) to separate dopant from host
    • Potential issues: It cannot parse dopants that are compounds
      • E.g. (AgCl)0.001PbTe(0.999) is supposed to be (PbTe)0.999 as host and (AgCl)0.001 as dopant. But we might get Ag0.001, Cl0.001.
      • E.g. (AgCl)0.001PbTe(0.999)Cu0.001 is supposed to be (PbTe)0.999 as host and (AgCl)0.001 and Cu0.001 as dopant. But we might get Ag0.001, Cl0.001 and Cu0.001

  • Might want to remove base and alloy formulas

    • If we don't remove it, need to verify base + alloy = pretty formula

  • Generation of electronic and lattice thermal conductivity from total thermal conductivity

    • Need to have at least two of those properties
      • If 2 properties, take a simple difference or sum will get you the other
      • If 3 properties, need to verify if that electronic + lattice = total

  • Generation of power factor from Seeback coefficient and electrical conductivity

    • If we have 2 properties, can use the formula PF = S * electrical conductivty ** 2
    • If we have 3 properties, need to verify that the products are correct

  • Need to rebase the temperature of some of Sparks' Thermoelectric Data as per comments - some may be 1000K but taken from another temperature.

Data Verification

  • Resolve all comments in Excel

  • Verification of TE numerical values

    • Outlier analysis → Then manually check through said outliers

  • Verification of textual data (synthesis, comments)

    • Use LLM
    • For pretty formula, ask if the formula is present in the paper?

  • Check for duplicates

    • Logic: There is the DOI and source column. Given two different sources ,are there same DOIs?

Data Visualization

  • Thermoelectric values (e.g. zT) over the years

    • Line plot
    • Either take average (with std dev?) or maximum

  • doped vs undoped

    • single bar plot
    • might want to isolate DOIs containing both doped and undoped data for fairer comparison?

  • properties by family - scatterplot with different colours + size of scatterplot giving a third property

    • Check the material family first - some are ambiguous
    • Alternatively, just show common material family (maybe at least 2 different DOIs of the same family after removing those unsure)

  • Best performing thermoelectric material by temperatures