Hey Nitin!

The rankings are from the relative sorted positions (lowest to highest) of each element’s pymatgen mendeleev number.

For example, if you only have 3 elements in a dimension, with mendeleev numbers [50, 1, 22] your ranked mendeleev numbers would be:

[3, 1, 2]

In two of our dimensions, we had 52 elements,. so the mendeleev ranks were from 1-52.

For polyatomic anions, we took the average (weighted by stoichiometric proportion of each element) of the mendeleev numbers for the constituent elements. Then we ranked each anion’s average mendeleev number (lowest to highest) in the same way as above.

In our experiment the third dimension was 7 different anions. So they were ranked between 1-7. So for example, the ABX ranked mendeleev vector [1, 52, 3] represents

[the lowest mendeleev number element, the highest mendeleev number element, the 3rd highest weighted average mendeleev number anion]

Thanks,

Alex

## ···

On Tuesday, September 10, 2019 at 8:12:55 AM UTC-7, Nitin Kumar wrote:

Hi Alex,

I read through the Rocketsled paper (very nice details on the rocketsled framework) and was wondering how is the Mendeleev number calculated for cation and anion. In the Supplementary section [last paragraph on the page containing Section 3.1]:

"The search dimensions input to Rocketsled for A and B are integers from 1-52, representing the rank of the specie derived from its Mendeleev number. The search dimensions for anions were defined as integers 1-7, representing their rank based on their average Mendeleev number. "

- How is the cation’s Mendeleev number derived and then given the rank between 1-52 (Why this range?).

- For anion, how was the average Mendeleev number determined and then given range between 1-7?

Thanks.

best,

Nitin