Convert a flattened excel to nested json in pandas

I am fairly new to this and have spent the entire day reading numerous posts and figuring out how i can convert this flattened excel table to a nested json. Here is an example of the flattened nested table:

    {'Sample': {0: '1A',
  1: '1A',
  2: '1A',
  3: '1A',
  4: '1A',
  5: '1A',
  6: '1A',
  7: '2A',
  8: '2A',
  9: '2A',
  10: '2A',
  11: '2A',
  12: '2A',
  13: '2A'},
 'Substance category': {0: 'Additive',
  1: 'Additive',
  2: 'Alkali',
  3: 'Alkali',
  4: 'Alkali',
  5: 'Alkali',
  6: 'Alkali',
  7: 'Additive',
  8: 'Additive',
  9: 'Alkali',
  10: 'Alkali',
  11: 'Alkali',
  12: 'Alkali',
  13: 'Alkali'},
 'Substance': {0: 'Irgafos 168',
  1: 'Alkylphenylphosphate',
  2: 'Calcium',
  3: 'Kalium',
  4: 'Lithium',
  5: 'Magnesium',
  6: 'Natrium',
  7: 'Irgafos 168',
  8: 'Alkylphenylphosphate',
  9: 'Calcium',
  10: 'Kalium',
  11: 'Lithium',
  12: 'Magnesium',
  13: 'Natrium'},
 'Value': {0: 0,
  1: 0,
  2: 2,
  3: 2,
  4: 1,
  5: 2,
  6: 3,
  7: 2,
  8: 3,
  9: 2,
  10: 3,
  11: 1,
  12: 2,
  13: 3}}

This table looks like this Sample table

I used the following code to get a nested json, which was taken from this answer.

j = (df.groupby(['Sample','Substance category'])
       .apply(lambda x: x[['Substance','Value']].to_dict('records'))
       .reset_index()
       .rename(columns={0:'Substance'})
       .to_json(orient='records'))

I am getting the following json.

[
  {
    "Sample": "1A",
    "Substance": [
      {
        "Substance": "Irgafos 168",
        "Value": 0
      },
      {
        "Substance": "Alkylphenylphosphate",
        "Value": 0
      }
    ],
    "Substance category": "Additive"
  },
  {
    "Sample": "1A",
    "Substance": [
      {
        "Substance": "Calcium",
        "Value": 2
      },
      {
        "Substance": "Kalium",
        "Value": 2
      },
      {
        "Substance": "Lithium",
        "Value": 1
      },
      {
        "Substance": "Magnesium",
        "Value": 2
      },
      {
        "Substance": "Natrium",
        "Value": 3
      }
    ],
    "Substance category": "Alkali"
  },
  {
    "Sample": "2A",
    "Substance": [
      {
        "Substance": "Irgafos 168",
        "Value": 2
      },
      {
        "Substance": "Alkylphenylphosphate",
        "Value": 3
      }
    ],
    "Substance category": "Additive"
  },
  {
    "Sample": "2A",
    "Substance": [
      {
        "Substance": "Calcium",
        "Value": 2
      },
      {
        "Substance": "Kalium",
        "Value": 3
      },
      {
        "Substance": "Lithium",
        "Value": 1
      },
      {
        "Substance": "Magnesium",
        "Value": 2
      },
      {
        "Substance": "Natrium",
        "Value": 3
      }
    ],
    "Substance category": "Alkali"
  }
]

However what I actually want is to define an addition level for the 'Substance category'. Despite all my efforts, I just could not figure that out and none of the answers could help me.

Thank you very much in advance.

Well creating a multilevel df was not a problem. But when I exported that to a json, it did not maintain the nested structure of the indexes. Anyway, I finally found an answer here. It was just a matter to searching on google with the right keywords link