setting index value to np.uint64 in a panda dataframe changes its type to int64

Question

I am trying to set the dtype on the index of a panda dataframe

import pandas as pd
import numpy as np
a = pd.DataFrame(columns=['value'])
a.loc[np.uint64(3)] = [3]

# Here I would expect to get UINT64 but it's INT64
a.index.dtype
> dtype('int64')

a.index = a.index.astype(np.uint64)

# Works as expected
a.index.dtype
> dtype('uint64')

a.loc[np.uint64(4)] = [4]

# Now I am getting totally confused. Why do I get float64 ?
a.index.dtype
> dtype('float64')

• Why is Panda not using the DType of the index value passed to the lob function
• Why is the dtype changing of what it seems to be randomly ?

What am I missing ? What is the proper way to set the index type once and for all ?