Originally posted by PinkFloyd
What does the Q mean in your formula? I'd like to try it.
There's not really anything to try; I've just given a definition, but I will elaborate.
Q is the field of rational numbers, and it's standard to construct Q[x], which consists of all polynomials with coefficients in Q. Polynomials can be added componentwise and multiplied the way you were taught in school, and with this addition and multiplication, Q[x] becomes a commutative ring with a 1.
Now declare that two polynomials in Q[x] are equivalent if and only if their difference is a multiple (in the Q[x] multiplication) of x^2 + 1. This partitions Q[x] into equivalence classes (these are called the cosets of the principal ideal generated by x^2 + 1, if you feel like Wikipediating).
Now we call the set of these equivalence classes Q[x]/(x^2 + 1). We can define multiplication and addition in here, too, by letting this new set "inherit" the operations from Q[x]. Explicitly, the sum and product of the cosets containing p(x) and q(x) are the cosets containing p(x) + q(x) and p(x)q(x), respectively. It's not hard to show that these operations are well-defined, and that Q[x]/(x^2 + 1) becomes a commutative ring, with the coset containing 1 as its 1 and the zero coset (i.e. the set of polynomials divisible by x^2 + 1) as its zero.
Now do a little algebra to show that there are no polynomials in Q[x] whose product is x^2 + 1. From this it follows that Q[x]/(x^2 +1) is actually a field -- we have multiplicative inverses in here!
There are now a couple of equivalent ways to define i with this machinery. First, the field we just built is a 2-dimensional vector space over Q, and Q embeds in it, so it has a Q-basis consisting of 1 and something not in Q. We can pick the other basis element to have minimal polynomial x^2 + 1, by that construction, so we call it i, i.e. the field we constructed just now is Q extended by a root of x^2 + 1 which we just "imagined" into existence. We haven't brought up order yet, so right now there is nothing distinguishing i and -i, since they are both roots -- either one works.
Now make the real vector space with 1 and i ans a basis, form the subspace generated by i, knock out 0, and you have the imaginary numbers.
Alternatively (and this is what I said in my first post), hunt around in the field we just made for elements which are the roots of polynomials with INTEGER coefficients. These form a ring, and only four of them have multiplicative inverses, namely 1, -1 and two other elements. These form a cyclic group of order four. Any element that generates this group can be called i, and the other generator -i.
The really interesting thing here is that the more essential definition says: "take the algebraic closure of Q. It will be a degree-2 extension of R. It has to contain a root of x^2 + 1 because it is an algebraic closure, so take the subspace generated by this root, knock out 0, and you have the imaginary numbers." The reason this isn't appropriately simple is the second sentence, which is essentially the fundamental theorem of algebra. Now there is a Galois-theoretic proof of this, which uses the construction above and some machinery from group theory. However, most proofs I've seen come from complex analysis, or even wketchy geometry, and all of these proofs rely on a definition of the complex numbers that is in some sense independent of the one given above. It's almost, but not quite, enough to make one able to claim that an analyst and an algebraist
mean something different when they write "i".