The other crucial figure of the time in France was Joseph, Baron Fourier. His major contribution, presented in The Analytical Theory of Heat (1822), was to the theory of heat diffusion in solid bodies. He proposed that any function could be written as an infinite sum of the trigonometric functions cosine and sine; for example,

Expressions of this kind had been written down earlier, but Fourier’s treatment was new in the degree of attention given to their convergence. He investigated the question “Given the function f(x), for what range of values of x does the expression above sum to a finite number?” It turned out that the answer depends on the coefficients a_n, and Fourier gave rules for obtaining them of the form

Had Fourier’s work been entirely correct, it would have brought all functions into the calculus, making possible the solution of many kinds of differential equations and greatly extending the theory of mathematical physics. But his arguments were unduly naive: after Cauchy it was not clear that the function f(x) sin (nx) was necessarily integrable. When Fourier’s ideas were finally published, they were eagerly taken up, but the more cautious mathematicians, notably the influential German Peter Gustav Lejeune Dirichlet, wanted to rederive Fourier’s conclusions in a more rigorous way. Fourier’s methodology was widely accepted, but questions about its validity in detail were to occupy mathematicians for the rest of the century.

Elliptic functions

The theory of functions of a complex variable was also being decisively reformulated. At the start of the 19th century, complex numbers were discussed from a quasi-philosophical standpoint by several French writers, notably Jean-Robert Argand. A consensus emerged that complex numbers should be thought of as pairs of real numbers, with suitable rules for their addition and multiplication so that the pair (0, 1) was a square root of −1 (i). The underlying meaning of such a number pair was given by its geometric interpretation either as a point in a plane or as a directed segment joining the coordinate origin to the point in question. (This representation is sometimes called the Argand diagram.) In 1827, while revising an earlier manuscript for publication, Cauchy showed how the problem of integrating functions of two variables can be illuminated by a theory of functions of a single complex variable, which he was then developing. But the decisive influence on the growth of the subject came from the theory of elliptic functions.

The study of elliptic functions originated in the 18th century, when many authors studied integrals of the form

where p(t) and q(t) are polynomials in t and q(t) is of degree 3 or 4 i. Such integrals arise naturally, for example, as an expression for the length of an arc of an ellipse (whence the name). These integrals cannot be evaluated explicitly; they do not define a function that can be obtained from the rational and trigonometric functions, a difficulty that added to their interest. Elliptic integrals were intensively studied for many years by the French mathematician Adrien-Marie Legendre, who was able to calculate tables of values for such expressions as functions of their upper endpoint, x. But the topic was completely transformed in the late 1820s by the independent but closely overlapping discoveries of two young mathematicians, the Norwegian Niels Henrik Abel and the German Carl Jacobi. These men showed that if one allowed the variable x to be complex and the problem was inverted, so that the object of study became

considered as defining a function x of a variable u, then a remarkable new theory became apparent. The new function, for example, possessed a property that generalized the basic property of periodicity of the trigonometric functions sine and cosine: sin (x) = sin (x + 2π). Any function of the kind just described has two distinct periods, ω₁ and ω₂:

These new functions, the elliptic functions, aroused a considerable degree of interest. The analogy with trigonometric functions ran very deep (indeed, the trigonometric functions turned out to be special cases of elliptic functions), but their greatest influence was on the burgeoning general study of functions of a complex variable. The theory of elliptic functions became the paradigm of what could be discovered by allowing variables to be complex instead of real. But their natural generalization to functions defined by more complicated integrands, although it yielded partial results, resisted analysis until the second half of the 19th century.

The theory of numbers

While the theory of elliptic functions typifies the 19th century’s enthusiasm for pure mathematics, some contemporary mathematicians said that the simultaneous developments in number theory carried that enthusiasm to excess. Nonetheless, during the 19th century the algebraic theory of numbers grew from being a minority interest to its present central importance in pure mathematics. The earlier investigations of Pierre de Fermat had eventually drawn the attention of Leonhard Euler and Lagrange. Euler proved some of Fermat’s unproven claims and discovered many new and surprising facts; Lagrange not only supplied proofs of many remarks that Euler had merely conjectured but also worked them into something like a coherent theory. For example, it was known to Fermat that the numbers that can be written as the sum of two squares are the number 2, squares themselves, primes of the form 4n + 1, and products of these numbers. Thus, 29, which is 4 × 7 + 1, is 5² + 2², but 35, which is not of this form, cannot be written as the sum of two squares. Euler had proved this result and had gone on to consider similar cases, such as primes of the form x² + 2y² or x² + 3y². But it was left to Lagrange to provide a general theory covering all expressions of the form ax² + bxy+ cy², quadratic forms, as they are called.

Lagrange’s theory of quadratic forms had made considerable use of the idea that a given quadratic form could often be simplified to another with the same properties but with smaller coefficients. To do this in practice, it was often necessary to consider whether a given integer left a remainder that was a square when it was divided by another given integer. (For example, 48 leaves a remainder of 4 upon division by 11, and 4 is a square.) Legendre discovered a remarkable connection between the question “Does the integer p leave a square remainder on division by q?” and the seemingly unrelated question “Does the integer q leave a square remainder upon division by p?” He saw, in fact, that when p and q are primes, both questions have the same answer unless both primes are of the form 4n − 1. Because this observation connects two questions in which the integers p and q play mutually opposite roles, it became known as the law of quadratic reciprocity. Legendre also gave an effective way of extending his law to cases when p and q are not prime.

All this work set the scene for the emergence of Carl Friedrich Gauss, whose Disquisitiones Arithmeticae (1801) not only consummated what had gone before but also directed number theorists in new and deeper directions. He rightly showed that Legendre’s proof of the law of quadratic reciprocity was fundamentally flawed and gave the first rigorous proof. His work suggested that there were profound connections between the original question and other branches of number theory, a fact that he perceived to be of signal importance for the subject. He extended Lagrange’s theory of quadratic forms by showing how two quadratic forms can be “multiplied” to obtain a third. Later mathematicians were to rework this into an important example of the theory of finite commutative groups. And in the long final section of his book, Gauss gave the theory that lay behind his first discovery as a mathematician: that a regular 17-sided figure can be constructed by circle and straightedge alone.

The discovery that the regular “17-gon” is so constructible was the first such discovery since the Greeks, who had known only of the equilateral triangle, the square, the regular pentagon, the regular 15-sided figure, and the figures that can be obtained from these by successively bisecting all the sides. But what was of much greater significance than the discovery was the theory that underpinned it, the theory of what are now called algebraic numbers. It may be thought of as an analysis of how complicated a number may be while yet being amenable to an exact treatment.

The simplest numbers to understand and use are the integers and the rational numbers. The irrational numbers seem to pose problems. Famous among these is Square root of√2. It cannot be written as a finite or repeating decimal (because it is not rational), but it can be manipulated algebraically very easily. It is necessary only to replace every occurrence of (Square root of√2)² by 2. In this way expressions of the form m + nSquare root of√2, where m and n are integers, can be handled arithmetically. These expressions have many properties akin to those of whole numbers, and mathematicians have even defined prime numbers of this form; therefore, they are called algebraic integers. In this case they are obtained by grafting onto the rational numbers a solution of the polynomial equation x² − 2 = 0. In general an algebraic integer is any solution, real or complex, of a polynomial equation with integer coefficients in which the coefficient of the highest power of the unknown is 1.

Gauss’s theory of algebraic integers led to the question of determining when a polynomial of degree n with integer coefficients can be solved given the solvability of polynomial equations of lower degree but with coefficients that are algebraic integers. For example, Gauss regarded the coordinates of the 17 vertices of a regular 17-sided figure as complex numbers satisfying the equation x¹⁷ − 1 = 0 and thus as algebraic integers. One such integer is 1. He showed that the rest are obtained by solving a succession of four quadratic equations. Because solving a quadratic equation is equivalent to performing a construction with a ruler and a compass, as Descartes had shown long before, Gauss had shown how to construct the regular 17-gon.

Inspired by Gauss’s works on the theory of numbers, a growing school of mathematicians were drawn to the subject. Like Gauss, the German mathematician Ernst Eduard Kummer sought to generalize the law of quadratic reciprocity to deal with questions about third, fourth, and higher powers of numbers. He found that his work led him in an unexpected direction, toward a partial resolution of Fermat’s last theorem. In 1637 Fermat wrote in the margin of his copy of Diophantus’s Arithmetica the claim to have a proof that there are no solutions in positive integers to the equation xⁿ + yⁿ = zⁿ if n > 2. However, no proof was ever discovered among his notebooks.

Kummer’s approach was to develop the theory of algebraic integers. If it could be shown that the equation had no solution in suitable algebraic integers, then a fortiori there could be no solution in ordinary integers. He was eventually able to establish the truth of Fermat’s last theorem for a large class of prime exponents n (those satisfying some technical conditions needed to make the proof work). This was the first significant breakthrough in the study of the theorem. Together with the earlier work of the French mathematician Sophie Germain, it enabled mathematicians to establish Fermat’s last theorem for every value of n from 3 to 4,000,000. However, Kummer’s way around the difficulties he encountered further propelled the theory of algebraic integers into the realm of abstraction. It amounted to the suggestion that there should be yet other types of integers, but many found these ideas obscure.

In Germany Richard Dedekind patiently created a new approach, in which each new number (called an ideal) was defined by means of a suitable set of algebraic integers in such a way that it was the common divisor of the set of algebraic integers used to define it. Dedekind’s work was slow to gain approval, yet it illustrates several of the most profound features of modern mathematics. It was clear to Dedekind that the ideal algebraic integers were the work of the human mind. Their existence can be neither based on nor deduced from the existence of physical objects, analogies with natural processes, or some process of abstraction from more familiar things. A second feature of Dedekind’s work was its reliance on the idea of sets of objects, such as sets of numbers, even sets of sets. Dedekind’s work showed how basic the naive conception of a set could be. The third crucial feature of his work was its emphasis on the structural aspects of algebra. The presentation of number theory as a theory about objects that can be manipulated (in this case, added and multiplied) according to certain rules akin to those governing ordinary numbers was to be a paradigm of the more formal theories of the 20th century.