Loading field map
Transcendental number theory is the branch of number theory that studies numbers which are not algebraic. An algebraic number is any complex number that satisfies a nonzero polynomial equation with integer coefficients; examples include all rational numbers, square roots, cube roots, and many other constructed quantities. A transcendental number is a complex number that is not algebraic—it satisfies no such polynomial equation. The existence of transcendental numbers was established in the nineteenth century, and the field has since grown into a sophisticated discipline that uses tools from analysis, algebra, and geometry to determine whether specific numbers of interest are transcendental, to measure how "far" they are from being algebraic, and to understand the structural properties of the sets they form.
The most direct question in transcendental number theory is: Given a specific number, is it transcendental? This is often surprisingly hard. For example, it is known that π and e are transcendental, but the transcendence of many simple combinations, such as π + e or π·e, remains unknown. The field's central problems cluster around a few enduring themes.
The first theme is proving transcendence for individual constants. The classical constants—π, e, and their relatives—were the original targets. Beyond these, mathematicians ask about values of special functions at algebraic points, such as the gamma function, the zeta function, or elliptic functions.
The second theme is linear independence and algebraic independence. Two numbers are algebraically independent if no nonzero polynomial with integer coefficients in two variables vanishes at the pair. For instance, it is known that π and e are both transcendental, but whether they are algebraically independent is open. More generally, given a collection of numbers, one asks how many independent algebraic relations they satisfy. This is a much deeper question than individual transcendence, and progress has been slow.
The third theme is measure and distribution. Given a transcendental number α, one can ask how well it can be approximated by algebraic numbers of bounded degree and height. This leads to the theory of transcendence measures, which quantify the "distance" from α to the nearest algebraic number of a given complexity. Relatedly, the field studies the distribution of transcendental numbers in the complex plane, though this is less developed than the analogous theory for algebraic numbers.
A fourth theme, more recent, concerns transcendence in arithmetic geometry. Here one asks about the transcendence of periods—integrals of algebraic differential forms over algebraic cycles—and about the values of modular forms and other automorphic functions at algebraic points. This connects transcendental number theory to algebraic geometry, Hodge theory, and the theory of motives.
The prehistory of the field lies in the eighteenth-century distinction between rational and irrational numbers. Euler conjectured that logarithms of rational numbers are irrational, and Lambert proved that π is irrational in the 1760s. But the concept of an algebraic number as a unified class, and the question of whether a number could be non-algebraic, only emerged in the nineteenth century.
The first explicit transcendental numbers were constructed by Joseph Liouville in 1844. His method was based on approximation: he showed that algebraic numbers cannot be approximated too well by rationals, and he exhibited numbers (now called Liouville numbers) that are approximable beyond this bound. This established that transcendental numbers exist, but Liouville's examples were artificial—they were defined by rapidly converging series with no obvious connection to natural constants.
The next major step came in 1873, when Charles Hermite proved that e is transcendental. His proof was a tour de force of analysis, using integrals and estimates that later became standard tools. In 1882, Ferdinand von Lindemann adapted Hermite's method to prove that π is transcendental, thereby resolving the ancient Greek problem of squaring the circle: if π were algebraic, a straightedge-and-compass construction of a square with the same area as a given circle would be possible, but Lindemann's result ruled this out. The Hermite–Lindemann method also yielded the transcendence of e^α for any nonzero algebraic α, and hence of logarithms of algebraic numbers (when nonzero).
A second major method arrived in the 1920s with the work of Carl Ludwig Siegel. Siegel developed a general framework, often called the Siegel method or the method of E-functions, for proving transcendence of values of certain entire functions. His work was motivated by the desire to prove transcendence for values of Bessel functions and other special functions, and it introduced powerful analytic techniques that were later refined by others. Siegel's approach was not a complete solution, but it established a research programme that continues today.
The mid-twentieth century saw the development of the Baker method, named after Alan Baker, who in the 1960s proved deep results on linear forms in logarithms of algebraic numbers. Baker's theorem gives explicit lower bounds for the absolute value of expressions like β₁ log α₁ + ⋯ + βₙ log αₙ, where the αᵢ and βᵢ are algebraic numbers. This has had enormous consequences not only in transcendental number theory but also in Diophantine approximation, the theory of Diophantine equations, and even in computational number theory. Baker received the Fields Medal in 1970 for this work.
A parallel development, initiated by Aleksandr Gelfond and Theodor Schneider independently in the 1930s, resolved a conjecture of Hilbert: if α and β are algebraic, α ≠ 0, 1, and β is irrational, then α^β is transcendental. This is the Gelfond–Schneider theorem, and it settled the seventh of Hilbert's famous 23 problems posed in 1900. The theorem also implies the transcendence of 2^√2 and of e^π (since e^π = (−1)^{−i}).
The modern era of the field is often associated with the work of Yuri Manin, Michel Waldschmidt, and others who connected transcendence to algebraic geometry and to the theory of periods. A period is a complex number that can be expressed as an integral of an algebraic function over a domain defined by algebraic inequalities. Many classical constants—π, log 2, values of the gamma function at rational arguments—are periods. The conjecture of Kontsevich and Zagier, formulated in the late 1990s, proposes that any linear relation among periods with rational coefficients can be derived from a small set of elementary manipulations (additivity, change of variables, and Stokes's theorem). This conjecture remains open and is one of the central problems connecting transcendental number theory to algebraic geometry and mathematical physics.
Transcendental number theory is not organized into sharply separated schools, but rather into several methodological traditions that overlap and borrow from each other. The most important distinction is between analytic methods, which use complex analysis and estimates of integrals, and algebraic methods, which use algebraic geometry, commutative algebra, and the theory of algebraic groups. In practice, most proofs combine both.
The oldest method, originating with Hermite and Lindemann, is based on constructing auxiliary functions with many zeros and then deriving a contradiction from the assumption that a given number is algebraic. The core idea is to build an entire function that vanishes to high order at certain points, then evaluate it at the number in question and show that the value is both too small (by analytic estimates) and too large (by algebraic properties) to be zero. This method is direct and powerful for individual constants, but it is difficult to generalize to families of numbers or to algebraic independence.
Siegel's approach is a systematic refinement of the Hermite–Lindemann idea. An E-function is an entire function whose Taylor coefficients are algebraic numbers with controlled denominators and growth. Siegel showed that if an E-function satisfies a linear differential equation with polynomial coefficients, then its values at algebraic points are either transcendental or satisfy certain algebraic relations. This method is more flexible than the classical approach and applies to a wide class of special functions. However, it requires detailed knowledge of the differential equations satisfied by the functions in question, and it does not always yield the sharpest possible results.
The Gelfond–Schneider theorem was proved by a different analytic technique, often called the method of interpolation. The idea is to construct an auxiliary function that interpolates certain values, then use estimates from complex analysis to show that the function must vanish identically, leading to a contradiction. This method was later extended by Gelfond to prove results on algebraic independence of values of the exponential function, though the full algebraic independence of π and e remains open.
Baker's work is a culmination and extension of the Gelfond–Schneider method. The key innovation was to handle linear forms in several logarithms, rather than just a single expression. Baker's theorem provides explicit, effective lower bounds for such forms, meaning that the bounds are computable in principle. This effectiveness is crucial for applications: it allows number theorists to bound the sizes of solutions to Diophantine equations, to solve certain exponential Diophantine equations, and to prove results in computational number theory. The method is analytic, but it uses deep facts about algebraic number fields and their embeddings into the complex numbers.
Since the 1970s, a different tradition has grown out of the observation that many transcendental numbers arise as periods of algebraic varieties. This approach, associated with the work of Manin, Deligne, and others, treats transcendence as a question about the structure of the fundamental group and the de Rham cohomology of algebraic varieties. The central object is the period torsor, which encodes the relations among periods. The Kontsevich–Zagier conjecture is the most famous open problem in this direction, but even partial results have led to new insights. This approach is more algebraic and structural than the classical analytic methods, and it connects transcendental number theory to the Langlands programme and to the theory of motives. However, it has so far produced fewer concrete transcendence results for specific constants than the analytic methods.
A separate but important tradition studies transcendence in the p-adic setting, where the absolute value is the p-adic one rather than the usual complex absolute value. The p-adic analogue of the Gelfond–Schneider theorem was proved by Gelfond and by others, and the p-adic analogue of Baker's theorem is also known. These results are used in number theory, particularly in the study of Diophantine equations and in the theory of p-adic L-functions. A related but distinct area is transcendence in function fields of positive characteristic, where the role of the integers is played by polynomials over a finite field. Here, the theory is often more complete than in the number field case, because the analytic tools are simpler, but the results are not directly transferable.
These methods are not rivals in the sense of competing schools; rather, they are complementary tools that address different aspects of the same problems. The Hermite–Lindemann and Siegel methods are best for proving transcendence of specific values of special functions. The Gelfond–Schneider and Baker methods are best for proving transcendence of exponentials and logarithms of algebraic numbers, and for obtaining quantitative bounds. The algebraic-geometric approach is best for understanding the global structure of periods and for formulating conjectures that unify many individual results.
There is also a hierarchy of difficulty. Proving that a single number is transcendental is easier than proving that two numbers are algebraically independent, which in turn is easier than proving that a collection of numbers is algebraically independent. The known methods reflect this hierarchy: the analytic methods can often handle individual transcendence, but algebraic independence results are rare and difficult. For example, it is known that π and e^π are transcendental, but the algebraic independence of π and e is open, and even the algebraic independence of π and e^π is not known in full generality.
A recurring theme is the role of effective versus ineffective results. Baker's theorem is effective: it gives explicit constants. Many other transcendence results are ineffective: they prove that a number is transcendental but do not provide a way to compute a bound on the degree of any polynomial that might vanish at it. Effectiveness matters for applications, and the search for effective versions of known theorems is an active area.
The current state of transcendental number theory is characterized by a few well-established results, a large body of conjectures, and a set of tools that are powerful but limited.
Among the established results, the following are central:
These results are all special cases of a more general conjecture, often attributed to Schanuel, which would unify them. Schanuel's conjecture states that if α₁, …, αₙ are complex numbers that are linearly independent over the rationals, then the field generated by α₁, …, αₙ, $e^{α₁}$, …, $e^{αₙ}$ has transcendence degree at least n over the rationals. This conjecture implies the Hermite–Lindemann theorem (take $n = 1$), the Gelfond–Schneider theorem (take $n = 2$ with α₁ = 1, α₂ = β log α), and many other results. It also implies that π and e are algebraically independent (take α₁ = 1, α₂ = iπ, and use Euler's identity). Schanuel's conjecture is widely believed but completely unproved; it is perhaps the central open problem in the field.
Beyond Schanuel's conjecture, the field is shaped by several other major open problems:
The field also has a substantial computational and algorithmic component. Effective bounds from Baker's theorem are used in algorithms for solving Diophantine equations, and there is ongoing work on improving these bounds. The LLL algorithm and other lattice-reduction techniques are used to make Baker-type bounds practical.
A notable feature of the present landscape is the increasing interaction with other areas. Transcendental number theory now connects to:
These connections are not one-way: ideas from these fields have led to new conjectures and new techniques in transcendental number theory, even if they have not yet resolved its central problems.
It is important to be clear about what transcendental number theory can and cannot do. The field has no general method for deciding whether an arbitrary number is transcendental. In fact, the set of transcendental numbers is uncountable, while the set of numbers for which transcendence is known is countable, so in a precise sense almost all transcendental numbers are beyond the reach of current methods. The known results apply to numbers that arise from algebraic operations and the exponential function, or from periods of algebraic varieties, but even within this class many basic questions are open.
The distinction between transcendence and algebraic independence is crucial. Proving that a number is transcendental is often a matter of showing that it satisfies no polynomial equation of any degree. Proving algebraic independence of several numbers is a matter of showing that they satisfy no multivariate polynomial equation. The latter is much harder, and the known methods for algebraic independence are far less developed. For example, it is not known whether π and e are algebraically independent, nor whether π and ζ(3) are algebraically independent.
Another limitation is the lack of a general theory of transcendence measures. For a given transcendental number α, one would like to know how well it can be approximated by algebraic numbers of bounded degree and height. Such measures are known for some numbers (such as e and π) but not for others, and the optimal measures are almost never known.
Finally, the field is marked by a striking gap between what is conjectured and what is proved. Schanuel's conjecture, if true, would resolve many of the field's problems, but it seems to require fundamentally new ideas. The Kontsevich–Zagier conjecture is similarly ambitious. The field's progress has come from the development of increasingly sophisticated analytic and algebraic tools, but each new tool has revealed new difficulties rather than a general solution.
Transcendental number theory thus remains a field with a clear central object—the distinction between algebraic and transcendental numbers—a rich set of methods, a body of deep and beautiful results, and a frontier of open problems that are both central to mathematics and likely to require new ideas to resolve.