A periodic table in pandas

A table you can ask questions

A printed periodic table is good to look at, but you cannot ask it "which elements have a radius above 170 pm?" A DataFrame can answer at once. It is a table in Python, with named columns and numbered rows, from the library pandas.

In this lesson each row is one element. The columns are:

• symbol, z (the atomic number), group, period and block;
• radius, the atomic radius in pm;
• ie, the first ionisation enthalpy in kJ/mol;
• en, the Pauling electronegativity.

The noble gases have no radius or electronegativity in the table. pandas shows such a gap as NaN, "not a number".

Building the table

The data is typed into the program as plain text, one element to a line, with spaces between the values:

read_csv normally reads a file. io.StringIO(data) makes the text behave like a file, so no download is needed. sep=" " says the values are split by spaces, and names gives the column names. A value written nan becomes a missing value.

Two quick looks tell you the table came out right: df.shape gives the number of rows and columns, and df.head() prints the first five rows.

Picking columns and rows

• df["ie"] is one column, a Series.
• df[["symbol", "ie"]] is a smaller table with two columns. Note the two pairs of square brackets: the inner pair is a list of names.
• df.set_index("symbol") makes the symbols the row labels. Then .loc["Na", "ie"] looks up one value by its labels.

Sorting and filtering

df.sort_values("radius") puts the smallest atoms first; add ascending=False for the largest first. The original table is not changed unless you store the result.

A filter keeps the rows where a condition is true:

df["radius"] > 170 gives a column of True and False, one for each row, and df[...] keeps the True rows. To join two conditions, write & for "and" or | for "or", with brackets around each condition:

Hydrogen sits in group 1, but it is not an alkali metal, so the second condition leaves it out.

A computed column

A new column can be worked out from the others, for every row at once. One kJ/mol is about $\dfrac{1}{96.49}$ of an electron volt per atom, so:

There is no loop: pandas divides the whole column in one step.

Practise with new numbers

Try it in Python

The program builds the table of the first 36 elements. It prints its size, the three highest ionisation enthalpies, the alkali metals and period 3, and then adds the column in electron volts.

Common mistakes

• The wrong case in a column name. df["IE"] stops with KeyError: 'IE', because the column is called ie.
• and between two conditions. df[(df["group"] == 1) and (df["ie"] > 600)] stops with ValueError: The truth value of a Series is ambiguous. Use &.
• Leaving out the brackets. Without them, & is done before ==, and the same ValueError appears.
• Looking up a label that is not there. .loc["Kr"] on a table indexed by symbol works only if krypton is a row; otherwise it is KeyError: 'Kr'.
• Expecting sort_values to change the table. Store the result: df = df.sort_values("ie").

Remember these

• pd.read_csv(io.StringIO(text), sep=" ", names=cols) turns typed text into a DataFrame; nan marks a missing value.
• df["col"] is a column, df[["a", "b"]] is a smaller table, and df.set_index("symbol").loc[row, col] is one value.
• Filter with df[condition], and join conditions with & or |, each in brackets.
• A computed column is one line: df["new"] = df["old"] / number.

Check yourself

Try each one on your own first. If you are stuck, open a hint. Then check with Show the answer.

1. What do df.shape and df.head() tell you about a new table?

2. Why does the filter for the alkali metals have a second condition?

3. Write the filter that keeps the p-block elements of period 2.

4. A program prints the table sorted by radius, but df is still in order of Z afterwards. Why?

5. Why is there no loop in df["ie_eV"] = df["ie"] / 96.49?