Most of the numbers we first learn to treat as exact—fractions like 3/7, roots like √2, and even the golden ratio—share a comforting property: you can describe them as solutions to polynomial equations.
x^2 - 2 = 0.x^2 - x - 1 = 0.Numbers that satisfy some nonzero polynomial with integer (equivalently, rational) coefficients are called algebraic numbers.
A transcendental number is a number that does not satisfy any such polynomial. In other words, no matter what integer-coefficient polynomial you try, a transcendental number will never be one of its roots.
That definition is simple to state, but it leads quickly into some of the deepest parts of modern number theory—because ruling out every possible polynomial equation is an incredibly strong claim.
It helps to think of algebraic numbers as “numbers algebra can pin down”:
x^5 - 3x + 1 = 0,” it’s algebraic.Both kinds of numbers can be described by formulas, infinite series, or integrals. The difference is not whether you can write the number down—it’s whether a polynomial equation can capture it.
Famous examples:
Here’s one of the classic surprises:
So, in a precise sense, almost every real number is transcendental.
And yet, proving transcendence for a particular “everyday” constant can be maddeningly difficult. For example, despite overwhelming intuition about how “complicated” these should be, we still do not know whether numbers like e + π or e·π are even irrational, let alone transcendental.
This gap between how common transcendental numbers are and how hard it is to prove one is transcendental is a big part of the subject’s charm.
In the 1840s, Joseph Liouville found a way to produce transcendental numbers by making them “too well approximated” by rationals. Very roughly:
A famous example is Liouville’s constant:
L = Σ_{k=1..∞} 10^{-(k!)} = 0.110001000000000000000001...
The 1s appear at positions 1!, 2!, 3!, …, creating approximations that are “too good to be algebraic.”
This was the first time mathematicians could point to a specific, explicit transcendental number and prove it.
Two landmark theorems soon followed:
Lindemann’s result famously resolves an ancient geometric puzzle: you cannot “square the circle” using only straightedge and compass. The reason is that such a construction would force π to be algebraic—contradicting Lindemann’s theorem.
Even more powerful is the general direction these proofs point toward: exponentials are often where transcendence becomes provable.
A useful takeaway is:
If α ≠ 0 is algebraic, then e^α is transcendental.
This doesn’t answer everything—π itself isn’t of the form e^α—but it sets the tone: the exponential function tends to push algebraic inputs out into transcendental territory.
The Gelfond–Schneider theorem (1934) settled Hilbert’s 7th problem. In a friendly form:
If a is algebraic with a ∉ {0, 1}, and b is algebraic but not rational, then every value of a^b is transcendental.
This instantly proves the transcendence of several “nice-looking” constants, such as:
At first glance, i^i looks like it should be an exotic complex number. But complex exponentiation reveals something charming: a standard (principal) value is
i^i = e^(-π/2)
a perfectly ordinary positive real number (about 0.2079…).
And it’s transcendental.
One way to see why transcendence is plausible is through Gelfond–Schneider: both the base i and the exponent i are algebraic, and the exponent is not rational, so i^i falls under the theorem.
A subtle point (and a great conversation starter) is that complex exponentiation is multi-valued: different choices of the complex logarithm give different values of i^i. Gelfond–Schneider is built to handle that: it says any value of a^b in this setting is transcendental.
| Expression | Status | What we know (today) |
|---|---|---|
| e + π | Unknown | Not known to be irrational; not known to be transcendental. |
| e·π | Unknown | Not known to be irrational; not known to be transcendental. |
| e^π | Transcendental | Proven using Gelfond–Schneider (via the identity e^π = (-1)^(-i)). |
| π^e | Unknown | Not known to be irrational; not known to be transcendental. |
| i^i | Transcendental | One value equals e^(-π/2); transcendence follows from Gelfond–Schneider. |
A neat “in-between” fact you can prove with high-school algebra (plus one key field-theory idea) is:
At least one of (e + π) and (e·π) must be transcendental.
Why? If both (e + π) and (e·π) were algebraic, then e and π would satisfy the quadratic
x^2 - (e + π)x + eπ = 0
forcing e and π to be algebraic—contradiction. The argument doesn’t tell us which of the two is transcendental, but it shows that “both algebraic” is impossible.
If the 19th and early 20th centuries were about proving that specific famous constants are transcendental, much of today’s research asks a broader question:
When do families of naturally occurring constants satisfy algebraic relations, and when are they forced to be independent?
A few lively directions:
Many constants that appear across geometry, number theory, and physics can be expressed as periods—roughly, values of integrals of algebraic functions over algebraically described regions. Kontsevich and Zagier proposed a far-reaching viewpoint: any algebraic relation between periods should come from a small set of “geometric” operations on the defining integrals. This is closely related to Grothendieck’s vision of motives—a framework meant to explain why different cohomology theories “see the same shapes.”
This is modern transcendence theory with a geometric accent: instead of treating constants as isolated objects, it tries to classify which kinds of numbers can appear from which kinds of spaces.
The numbers
ζ(s) = Σ_{n=1..∞} 1 / n^s
are famous; at odd integers s = 3, 5, 7, … they are especially mysterious. A lot of modern work builds “motivic” versions of these constants to predict the web of relations among them. Even when transcendence is out of reach, mathematicians can often prove weaker but meaningful facts—like irrationality or linear independence in large families.
A dramatic modern development is functional transcendence, where one studies algebraic relations not just among numbers, but among values of analytic functions (exponential, elliptic, modular, and more). Tools from o-minimal geometry and results in the spirit of Ax–Schanuel control how algebraic structure can intersect the transcendental behavior of these functions. These ideas connect transcendence to “unlikely intersections” problems in Diophantine geometry—situations where intersections are expected to be tiny unless there is a hidden reason.
Even on classic, concrete questions, there has been real movement. A celebrated theorem of Ball–Rivoal shows that infinitely many odd zeta values ζ(3), ζ(5), ζ(7), … are irrational, and subsequent work has improved quantitative versions of this statement. This isn’t yet transcendence, but it’s a meaningful step in the same direction: proving these constants can’t satisfy “too simple” algebraic descriptions.
Transcendental number theory is, in a sense, the study of the boundary of algebra. It explains why some classic dreams are impossible (like squaring the circle), why exponentials are so powerful in producing new numbers, and why many of the constants that arise naturally in analysis and geometry seem to live in a realm that algebra can only partially tame.
And perhaps most importantly, it reminds us that “simple to write down” and “simple to understand” are very different things.