I presume then that the "datetime" string you provided is a string column?
If so, try the following code. This will not perform any transformation to timestamps:

from pyspark.sql.functions import lit, split, concat

data = [{"date_string": "2002-01-01T00:00:00.000"}]

data_df = spark.createDataFrame(data) \
    .withColumn(
        "date_string_reformatted", concat(
            lit(split("date_string", "-")[1]), 
            lit("-"),
            lit(split(split("date_string", "-")[2], "T")[0]),
            lit("-"),
            lit(split("date_string", "-")[0]), 
            lit(" "),
            lit(split(split(split("date_string", "-")[2], "T")[1], "\.")[0]),
        )
    )

data_df.display()	

The result: