Skip to content

Chapter 3 — Polynomials and Rational Functions

Close of Chapter 3 — The Impossible That Worked: the Legitimation of Complex Numbers

A quantity does not need to be understood to be used; sometimes it must be used first, for centuries, in order to begin to understand it afterwards.


Complex numbers were not born, as one might suppose, from the question x2=−1x^2=-1. That equation any algebraist of the sixteenth century could simply declare without solution and move on without much discomfort, exactly as §1.2 of this book did, implicitly. They were born, instead, from a problem where declaring “it has no solution” was not an option: Cardano’s formula for solving cubic equations, published in 1545. For certain cubics with three real roots, perfectly computable and verifiable by direct substitution, Cardano’s formula demanded passing, in an intermediate step, through the square root of a negative number. The final result, once the computation was completed, was real, correct, verifiable; the road to it, however, crossed a territory that nobody knew how to justify. This phenomenon, today called casus irreducibilis, could not be resolved by ignoring the impossible quantity: one had to operate with it, even though nobody knew what it was.

It was Rafael Bombelli who, in 1572, dared to take that territory seriously. In his Algebra, he proposed explicit rules for adding and multiplying these quantities, which he called più di meno and meno di meno, “plus of minus” and “minus of minus”, and he showed, by direct computation, that when they were applied systematically to a concrete case of Cardano’s formula, the impossible parts canceled each other, leaving exactly the real root already known by other means. Bombelli offered no philosophical justification of what those quantities were; he offered, instead, something more persuasive for an algebraist: rules of calculation that, applied with discipline, always produced correct answers. It was, in essence, the same kind of operational confidence with which Bombelli and his contemporaries already treated negative numbers, also “impossible” for a good part of the Greek geometric tradition, and yet indispensable for keeping the account of a debt.

The name that ended up prevailing did not help dissolve the suspicion. It was Descartes who, in 1637, called these quantities imaginary, as opposed to real, and he did so, according to the context of his Géométrie, with a connotation close to disdain: an imaginary root was, for him, a root that algebra produced without any geometric construction being able to exhibit it as a length. The name remained fixed in the mathematical language for centuries, even after the suspicion that motivated it ceased to have any basis, a situation not very different from the one that still pursues today the word “irrational” of §1.1, loaded with a value judgment that the proof of irrationality never intended to issue.

The first genuine legitimation did not come from algebra, but from geometry, and it took more than a century and a half to arrive. John Wallis, around 1673, attempted a first graphical representation without achieving a satisfactory formulation; it was the Norwegian-Danish surveyor Caspar Wessel, in 1799, and independently the Parisian bookkeeper Jean-Robert Argand, in 1806, who proposed the idea that finally took hold: to represent a+bia+bi as the point with coordinates (a,b)(a,b) in a plane, with the horizontal axis for the real part and the vertical for the “imaginary” part. Under this view, multiplying by ii ceased to be a mysterious operation and became a rotation of ninety degrees about the origin, a transformation as geometrically concrete as any reflection or translation. Carl Friedrich Gauss, whose authority ended up consolidating the idea before the mathematical community around 1831, was also the one who proposed the name that would gradually replace Descartes’s: complex numbers, without the pejorative load of “imaginary”, precisely to point out that there was nothing less real in them than in any other geometric object of the plane.

But the geometric representation, however persuasive, still left one question unanswered: why should it be accepted that a point of the plane is, genuinely, a number, with the right to be added and multiplied as such? It was William Rowan Hamilton who, in 1837, took the step that this very chapter repeated in §3.3: instead of appealing to a geometric intuition or to an “imaginary unit” of uncertain nature, Hamilton defined complex numbers, without adornment, as ordered pairs of real numbers, with a sum and a product specified by explicit rules, (a,b)+(c,d)=(a+c,b+d)(a,b)+(c,d)=(a+c,b+d) and (a,b)(c,d)=(ac−bd,ad+bc)(a,b)(c,d)=(ac-bd,ad+bc), exactly those of §3.3 §1. Under this construction, ii is not a mysterious quantity that someone decided to “invent” to solve x2+1=0x^2+1=0; it is, quite simply, the pair (0,1)(0,1), an object as concrete as any other element of R×R\mathbb{R}\times\mathbb{R}, whose square turns out to be, by direct and verifiable computation, the pair (−1,0)(-1,0). The question “what is ii, really?”, which had tormented generations of mathematicians since Bombelli, dissolves in Hamilton’s construction, not because it is answered, but because it ceases to be necessary: ii is exactly what its axioms say it is, neither more nor less, in the same way that −1-1 is exactly what the field axioms of §1.1 say it is, without any need to ask whether it “exists” in some deeper sense.

That construction, however, did not come without a cost, and that cost reveals something structural about what is gained and what is lost when a number system is extended. §1.7 of this book constructed R\mathbb{R} as an ordered field: for every pair of distinct reals, one is greater than the other, in a way compatible with sum and product. C\mathbb{C} admits no order with that compatibility: if it admitted one, i2=−1i^2=-1 would force a contradiction, since in any ordered field the square of any nonzero element is positive. Gaining a root for x2+1x^2+1 cost, exactly, the possibility of saying whether one complex number is “greater” or “less” than another. And the pattern did not stop at C\mathbb{C}: Hamilton himself, seeking to extend the construction once more to represent rotations in three dimensions, discovered in 1843 the quaternions, a number system of dimension four that gains still more structure at the expense of sacrificing, this time, the commutativity of the product. Each extension seems to obey the same logic of exchange: no number system has everything, and each is defined as much by what it adds as by what it decides to leave behind.


What remains, to close the chapter, is the same question that the close of the previous chapter already hinted at, now in its sharpest version. If C\mathbb{C} is constructed, without mystical residue, as R×R\mathbb{R}\times\mathbb{R} with two well-defined operations, with no act of faith, no “imaginary unit” that has to be accepted by decree, in what sense, then, can it be said that Bombelli discovered something in 1572, if the object he was manipulating would not be properly constructed until almost three centuries later? Is the history of C\mathbb{C} the history of an object that was waiting, already complete, to be found, or the history of a tool that worked for generations before anyone knew, or needed to know, exactly what it was they were using?