nLab irrational number

Irrational numbers

Irrational numbers


An irrational number is of course a number that is not rational. As such, the concept is perhaps uninteresting. However, the term ‘irrational number’ is often used for an irrational real number; in this case, it is interesting to consider such numbers for two reasons:

Of course, there are also various theorems about general classes of numbers that distinguish rational from irrational numbers.


In the real numbers

An affine function on the real numbers consists of a function ff and real number coefficients aa and bb such that for all real numbers xx, f(x)=ax+bf(x) = a x + b. The constant function at zero λx.0\lambda x.0 is an affine function where a=0a = 0 and b=0b = 0. A real number is irrational if for all affine functions ff with integer coefficients, if f(x)=0f(x) = 0, then f=λx.0f = \lambda x.0. This is equivalent to saying that a real number is irrational if for all rational numbers aa, xax \neq a.

Alternatively, a real number xx is irrational if given any rational number aa (thought of as a real number), the absolute value |xa|{|x - a|} is positive.

These two definitions are equivalent in classical mathematics. However, these two definitions no longer coincide in constructive mathematics; the former definition of irrational number is called weakly irrational while the latter definition is called strongly irrational or strictly irrational. Strongly irrational numbers are most commonly used in constructive mathematics, since it uses the apartness relation or strict order relation of the real numbers, which, unlike equality, is what is detected of the real numbers in constructive mathematics.

The set of irrational real numbers (a subset of the set of real numbers) is variously denoted 𝕀\mathbb{I}, 𝕁\mathbb{J}, or 𝔹\mathbb{B} (in various fonts). The 𝕀\mathbb{I} and 𝕁\mathbb{J} stand for ‘irrational’, while the 𝔹\mathbb{B} stands for ‘Baire’ (see the next paragraph). Here we will use 𝕁\mathbb{J}.

We may give 𝕁\mathbb{J} a topology as a subspace of the real line \mathbb{R}. With this topology, 𝕁\mathbb{J} is sometimes called Baire space; however, one uses a different uniform structure. (This should be distinguished from the sense of Baire space as a space to which the Baire category theorem applies; however, 𝕁\mathbb{J} is an example of such a space.)

In Archimedean integral domains

Let RR be an Archimedean integral domain with the integers R\mathbb{Z} \subseteq R being a integral subdomain of RR.

An element rRr \in R is irrational if for all aa \in \mathbb{Z} and bb \in \mathbb{Z}, |a|>0\vert a \vert \gt 0 and |arb|>0\vert a \cdot r - b \vert \gt 0.

The set of irrational numbers in RR is defined as

𝕁 R{rR|a,b.(|a|>0)(|arb|>0)}\mathbb{J}_R \coloneqq \{r \in R \vert \forall a, b \in \mathbb{Z}. (\vert a \vert \gt 0) \wedge (\vert a \cdot r - b \vert \gt 0) \}

In integral domains with a p-adic norm

Let RR be an integral domain with a p-adic norm |()| p\vert(-)\vert_p for a prime number pp, with the integers R\mathbb{Z} \subseteq R being a integral subdomain of RR.

An element rRr \in R is irrational if for all aa \in \mathbb{Z} and bb \in \mathbb{Z}, |a| p>0\vert a \vert_p \gt 0 and |arb| p>0\vert a \cdot r - b \vert_p \gt 0.

The set of irrational numbers in RR is defined as

𝕁 R{rR|a,b.(|a| p>0)(|arb| p>0)}\mathbb{J}_R \coloneqq \{r \in R \vert \forall a, b \in \mathbb{Z}. (\vert a \vert_p \gt 0) \wedge (\vert a \cdot r - b \vert_p \gt 0) \}


The followers of Pythagoras believed that ‘All is number’, meaning what we now call (positive) natural numbers. In geometry, this meant that any two lengths (or other geometric magnitudes) xx and yy are commensurable? in the sense that there exists a unit length uu such that x=mux = m u and y=nuy = n u for some natural numbers mm and nn. Identifying the ratios of geometric magnitudes with (positive) real numbers, this becomes the claim that every real number is rational. The discovery that this is false is also attributed to the Pythagoreans (but the legends of punishment for this secret date from several hundred years later). Greek mathematicians developed further the theory of irrational numbers, up to the general theory of magnitudes (which we may now regard as a theory of positive real numbers) attributed to Eudoxus in Book X of Euclid's Elements.

Mathematicians coming from the cultures of the Islamic Golden Age (particularly Abu Kamil?) were the first to treat irrational numbers algebraically as numbers (rather than geometrically as ratios of magnitudes); they applied the algebra of Al-Khwarizmi? to square roots, cube roots, etc. (Ultimately, Omar Khayyam developed a general method to find the real roots of any cubic polynomial.) However, they seem to have implicitly believed that all real numbers were expressible using such roots (radical number?s), which we now know is false even for some algebraic numbers, such as the real root of x 5+2x+1x^5 + 2 x + 1. In any case, they only used such numbers.

Later, European mathematicians of the early modern era (particularly Cardano?, Tartaglia?, and Ferrari?) had begun work with imaginary numbers, which are necessarily irrational. Following this, Lambert? and Legendre succeeded in proving the irrationality of pi, e, and their powers, which ultimately led to the conjecture that they were transcendental (whereas radical numbers, and even the root of x 5+2x+1x^5 + 2x + 1, are by definition algebraic); this conjecture was later established by Hermite and Lindemann?. Around this time, Euler and Lagrange popularized continued fractions (see below) to study both rational and irrational numbers.

During the arithmetization of analysis? in the 19th century, people sometimes wrote of the problem of ‘defining irrational numbers’. The actual issue here was defining real numbers in general; one could define rational numbers algebraically, leaving only the irrational numbers as the problem. However, this may be a red herring; one could just as easily define algebraic numbers algebraically and say that the problem is defining transcendental numbers; indeed, it was only with the discovery that such numbers as π\pi and e\mathrm{e} are irrational that work on this problem came to life. On the other hand, it's not clear that anybody could completely work out the order properties of algebraic numbers without already coming upon Dedekind's solution. In any case, specific irrational algebraic numbers such as 2\sqrt{2} posed no difficulty to the finitist methods used by such algebraists as Leopold Kronecker.

To this day, there are various specific real numbers (such as π+e\pi + \mathrm{e}, the Euler–Mascheroni constant γ\gamma, etc.) whose rationality or irrationality is unknown. In constructive mathematics, this makes it unproved that these numbers are rational or irrational (although the double negation of this statement can be proved for any real number). The question of whether 2 2{\sqrt{2}}^{\sqrt{2}} is rational or irrational is part of a famous illustration of the nature of constructive vs nonconstructive proof. (Namely, there is a cheap and easy nonconstructive proof that there exist irrational numbers aa and bb such that a ba^b is rational: let bb be 2\sqrt{2} and let aa be either 2\sqrt{2} or 2 2{\sqrt{2}}^{\sqrt{2}}, depending on whether the latter is rational or irrational. A constructive proof that decides which of these is the case is much harder: 2 2{\sqrt{2}}^{\sqrt{2}} turns out to be irrational, by a constructive version of the Gelfond–Schneider theorem?.1)


The Baire space 𝕁\mathbb{J} is homeomorphic to the product space \mathbb{N}^{\mathbb{N}} of 0\aleph_0 copies of the discrete space of natural numbers. The homeomorphism is given by continued fractions (see below).

Every inhabited Polish space is a quotient space of 𝕁\mathbb{J}, and 𝕁\mathbb{J} is itself a Polish space.

As a subset of the real line, 𝕁\mathbb{J} is a full set (meaning that its complement, the set of rational numbers, is null).

Cantor space may be identified with a subspace of 𝕁\mathbb{J}, consisting of those irrational numbers whose continued fraction expansion consists only of 11 and 22 (but this does not agree with the usual inclusions into \mathbb{R}).

The fan theorem states precisely that 𝕁\mathbb{J} (when thought of as a topological space) is sober or that 𝕁\mathbb{J} (when thought of as a locale) is topological/spatial/has enough points. This is true in classical mathematics and in intuitionistic mathematics but fails in other forms of constructive mathematics.

Continued fractions

(Main article: continued fraction.)

Let [a 0;a 1,a 2,a 3,][a_0;a_1,a_2,a_3,\ldots] be an infinite sequence of integers, all positive except (possibly) a 0a_0. We interpret this as the number

a 0+1a 1+1a 2+1a 3+. a_0 + \frac{1}{a_1 + \frac{1}{a_2 + \frac{1}{a_3 + \cdots}}} .

By truncating this expression after a ia_i, we produce a rational number; altogether, this is an infinite sequence of rational numbers.


This is a Cauchy sequence whose limit is irrational. Furthermore, every irrational number has a unique representation in this way. Yet more, the bijection thus shown between 𝕁\mathbb{J} and the infinitary cartesian product × +× +× +×\mathbb{Z} \times \mathbb{N}^+ \times \mathbb{N}^+ \times \mathbb{N}^+ \times \cdots is a homeomorphism when the two sets are given their usual topologies.

The usual proofs of these theorems are entirely constructive. Accordingly, in the foundations of mathematics, one may define Baire space either as the space of irrational numbers or as the infinite product \mathbb{N}^{\mathbb{N}}. However, to treat Baire space as a uniform space or as a metric space, one uses the structure from \mathbb{N}^{\mathbb{N}}.



  1. Of course, one could also take a=2a = \sqrt{2} and b=2log3log2)b = 2\frac{\log 3}{\log 2}), which are both irrational by easy constructive proofs, if one is after definite irrational numbers aa and bb such that a ba^b is rational (here the result is a b=3a^b = 3). The irrationality of log3log2\frac{\log 3}{\log 2} follows, as easily as that of 2\sqrt{2}, from the fundamental theorem of arithmetic: there do not exist nonzero integers p,qp, q such that 2 p=3 q2^p = 3^q.

Last revised on February 22, 2024 at 23:35:18. See the history of this page for a list of all contributions to it.