Chapter 5 — Trigonometric Functions of Real Numbers
5.1 - Construction of π
Before being able to “wrap” the real line onto a circle, it is necessary to know, precisely, what length that circle has. This book already constructed without taking anything for granted; will be no exception.
§1. The problem: what length does a circle have?
Section titled “§1. The problem: what length does a circle have?”The following sections of this chapter will define the trigonometric functions by means of a “wrapping” function that associates, to each real number , a point of the unit circle reached by traveling a distance along that circle, starting at . For that definition to make sense —to know, for example, when the trip completes a turn and begins to repeat— a prior datum is needed: the total length of the unit circle.
That datum is, by definition, . But “length of a curve” is not, yet, a concept that this book has defined with rigor —the general theory of arc length belongs to integral calculus—, and appealing to it here would be exactly the kind of concession that the book avoided with the derivative in Chapter 4. The way out, for the particular case of the circle, is the one found by Archimedes in the third century B.C.: to approximate the circle by regular inscribed and circumscribed polygons, whose lengths —finite sums of straight segments— are completely defined by the Pythagorean theorem, without any notion of curve length. It is, moreover, a construction that cannot rely on sine or cosine —that would be circular, since sine and cosine are not yet defined, and will be defined in §5.3 using —; everything that follows uses, exclusively, coordinates, distances and the Pythagorean theorem.
§2. Inscribed polygons: an increasing sequence of perimeters
Section titled “§2. Inscribed polygons: an increasing sequence of perimeters”To avoid any implicit appeal to an angle or to a fraction of a turn —notions that depend on —, the starting polygon is constructed directly by coordinates.
The initial square. The four points , , , satisfy , hence they are on the unit circle; by symmetry and the distance formula of §1.8 §2, they form a square. Its side is
and its perimeter, .
The doubling lemma. Let a regular polygon of sides be inscribed in the unit circle, with side . Let be two consecutive vertices, the midpoint of the segment , and the point of the circle on the ray from the center passing through , on the side opposite to with respect to —the “midpoint of the arc” —. Since is the perpendicular bisector of —the triangle is isosceles, with —, the point is, by construction, equidistant from and from ; it will be the new vertex when doubling the number of sides.
Lemma. With —Pythagoras in the right triangle , with hypotenuse and leg —, the new side satisfies
Proof. In the right triangle —right-angled at , since is on the line —, with legs and , the Pythagorean theorem gives
Since —by the very definition of —, it follows that .
Proposition (strict growth). If is the perimeter of the regular inscribed polygon of sides and that of the polygon of sides obtained by doubling each side according to the Lemma, then .
Proof. For each side of the original polygon, the point of the Lemma does not belong to the segment —it is on the circle, strictly farther from the chord than any point of , except in the degenerate case —, so that , , form a genuine triangle. The strict triangle inequality gives , that is, the contribution of this side to the new perimeter exceeds its contribution to the original perimeter. Summing over the sides, .
With increasing, a sequence is needed that bounds it from above —the role that played for in §4.2—. The circumscribed polygons play exactly that role. It is the content of §3.
§3. Circumscribed polygons: the dual decreasing sequence
Section titled “§3. Circumscribed polygons: the dual decreasing sequence”The circumscribed square. By the definition of a tangent line to a circle —§1.8—, the tangent at is perpendicular to the radius at that point, that is, the vertical line ; similarly, the tangents at , , are , , . These four lines form a square of side —from to —, with perimeter .
Lemma (equal tangent segments). If is a point exterior to the unit circle, and are the points of tangency of the two tangent lines from , then .
Proof. The triangles and are right-angled at and respectively —tangent perpendicular to the radius—, with and common hypotenuse . By Pythagoras, , and since both are positive, .
Proposition (strict decrease). If is the perimeter of the circumscribed polygon of sides —tangent at the same points as the vertices of the inscribed polygon of sides— and the one obtained by doubling, tangent also at the new points of §2, then .
Proof. Consider a vertex of the circumscribed polygon of sides, where the tangents at two consecutive points meet. The new point of tangency —midpoint of the arc , according to §2— determines a new tangent line, which cuts the segments and at points and respectively —by convexity, the new tangent separates from the arc—. The new perimeter replaces, locally, the path —of length — by the path —of length , since , , are collinear—. By the strict triangle inequality in the triangle —non-degenerate, since does not belong to the new tangent—, . Hence the new path is strictly shorter than the previous one, for each vertex of the original polygon. Summing, .
Computation formula. With and as in §2, the same triangle —now compared, by similarity of shared angle at , with the triangle formed by the radius to a point of tangency and the corresponding tangent— gives the ratio , where is the side of the circumscribed polygon of sides; whence , and
With increasing and decreasing, exactly what §4.2 §3 and §4 already did for is missing: the crossed bound, and the definition by means of supremum and infimum. It is the content of §4.
§4. The crossed bound and the definition of π
Section titled “§4. The crossed bound and the definition of π”Proposition. For all of the form , .
Proof. Same index: since the circumscribed polygon strictly contains the inscribed one —each side of the inscribed one is a chord, and the corresponding circumscribed side is a tangent, farther from the center—, ; formally, with —since for —, hence . Different indices: with , increasing gives , the equal-index case gives , and decreasing gives . Chaining, .
By the supremum axiom, there exists ; by the Lemma of existence of the infimum —section 4.1, §7—, there exists , and by the previous Proposition. The equality is obtained by bounding : since ,
and since —because , according to the doubling Proposition of §2 applied repeatedly— and is bounded —by —, this difference can be forced below any , with the same type of Archimedean argument that closed the Proposition of §4.2 §4. We conclude .
Definition. .
With defined without ambiguity, a couple of iterations of the method already suffice to bound it between two concrete numbers. It is the content of §5.
§5. Numerical bounds
Section titled “§5. Numerical bounds”With , the doubling Lemma of §2 gives .
Bound of . and , whence .
Bound of . From the above, . Since and , it follows that .
Bound of . , so that . In particular, .
Bound of and of . With , , whence ; since , it follows that . Hence .
Conclusion. .
§6. Close
Section titled “§6. Close”The complete method of this section —inscribed and circumscribed perimeters, doubling of sides, the crossed bound— appears, practically unchanged, in the treatise On the Measurement of the Circle by Archimedes of Syracuse, written around the year 250 B.C. It is, of the three historical precedents that this chapter kept citing —Napier in 1614, Newton in 1687, Euler in 1748—, the oldest by almost eighteen centuries, and the only one that did not need to wait for any later development of analysis in order to be, in its own terms, fully rigorous: Archimedes did not have the completeness axiom under that name, but his argument that the two sequences “squeeze together” leaving no margin is, in essence, exactly the one formalized in §4.
It is worth, before continuing with the unit circle proper, situating the importance of this method beyond having produced a number. Its structure —improving an approximation by doubling the computational effort at each step, with an error bound that is reduced in a controlled and demonstrable way— is the same principle that sustains, two thousand years later, a good part of computational numerical analysis: every iterative algorithm that refines a solution step by step, from the bisection method to the numerical methods that solve differential equations without a closed formula, shares with Archimedes the same underlying logic, although none of them needs to double polygons any longer. And the number itself that this method produced turns out, once the trigonometric functions are defined in the following sections, to be inseparable from every periodic phenomenon that science describes —from a sound wave to a planetary orbit—; but that ubiquity cannot yet be shown here, because it depends on tools —sine, cosine, and their periodic behavior— that this chapter is only beginning to construct.
With constructed and bounded, the chapter can, at last, define the unit circle and the wrapping function that translates real numbers into points of that circle. That is the content of §5.2.
Copyright © 2026 Carlos Dagorret — Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)