Constructing a number field with complex embedding

I am using the function number_field_elements_from_algebraics in sage.rings.qqbar to write some algebraic numbers as elements of a number field. I am passing embedded=True to also construct an embedding into QQbar. This works fine if the algebraic numbers are real but fails if they are complex:

sage: number_field_elements_from_algebraics([1 + sqrt(7)], embedded=True)
(Number Field in a with defining polynomial y^2 - 7 with a = 2.645751311064591?,
[a + 1],
Ring morphism:
From: Number Field in a with defining polynomial y^2 - 7 with a = 2.645751311064591?
To:   Algebraic Real Field
Defn: a |--> 2.645751311064591?)

sage: number_field_elements_from_algebraics([1 + sqrt(-7)], embedded=True)
---------------------------------------------------------------------------
NotImplementedError                       Traceback (most recent call last)
----> 1 number_field_elements_from_algebraics([Integer(1) + sqrt(-Integer(7))], embedded=True)

/usr/lib/python2.7/site-packages/sage/rings/qqbar.pyc in number_field_elements_from_algebraics(
numbers, minimal, same_field, embedded, prec)
2242         real_case = False
2243         if embedded:
-> 2244             raise NotImplementedError
2245     # Make the numbers algebraic
2246     numbers = [mk_algebraic(_) for _ in numbers]

NotImplementedError:


What is the reason for why only embeddings of real numbers are supported?

I modified the code in number_field_elements_from_algebraics to support complex embeddings by essentially removing the check for whether the numbers are real and replacing RealIntervalField by ComplexIntervalField. This seems to work and the doctests still pass with these modifications. However I am worried that there was a good reason (either mathematically or related to some sage internals) for restricting this function to the real case.

The relevant code in sage.rings.qqbar was introduced in these commits:

• 26dc3e4e6a26f6f613a69d57929ea492c278dad0
• a9045bf8a3aab2d0aa00be17e91227bc1b50262a
edit retag close merge delete

Sort by » oldest newest most voted

I would say that 1 + sqrt(7) could be uniformly chosen to the be the unique positive root of (x-1)^2-7, while there is no canonical choice for (x-1)^2+7. However, the problem should be in transforming the symbolic expression 1 + sqrt(-7) into an algebraic number, but once an element of QQbar is defined, i do not see any problem (an element of QQbar is embedded in the complex field). So for me, there should be an error, but it should appear when calling QQbar(1 + sqrt(-7)).

more

Thanks!

I am a bit confused though: Why can we not choose the solution with positive imaginary part for (x-1)^2+7? In fact, isn't this what sage is doing when it returns +I (as opposed to -I) for QQbar(sqrt(-1))?

( 2019-07-12 17:57:11 +0200 )edit

Having defined:

sage: a = sqrt(-7)


This works:

sage: number_field_elements_from_algebraics([a])
(Number Field in a with defining polynomial y^2 - y + 2,
[-2*a + 1],
Ring morphism:
From: Number Field in a with defining polynomial y^2 - y + 2
To:   Algebraic Field
Defn: a |--> 0.50000000000000000? - 1.322875655532296?*I)


This fails:

sage: number_field_elements_from_algebraics([a], embedded=True)
Traceback (most recent call last)
...
NotImplementedError:


The reason for why only embeddings of real numbers are supported is that nobody has done it yet. If you can do it, your contribution will be appreciated.

Regarding your worry that there was a good reason: if so, the correct thing to do would have been to add as a comment about the reason, and to raise a ValueError rather than a NotImplementedError.

more